NG00067 – ADSP Umbrella – 2018.07.30

Genetic Intelligence (GI) will analyze de-identified whole genome sequences (WGS) from Alzheimer’s disease (AD) and healthy patients using it’s AI-based platform to discover novel causal genes and variants for AD.GI has previously obtained a NSF Phase I SBIR grant (#1819331) to develop and validate it’s computational platform using Amyotrophic Lateral Sclerosis (ALS) WGS (obtained from our partner the New York Genome Center) as a proof-of-principle. GI successfully validated the platform by rediscovering known ALS genes SOD1 and C9orf72, as well as discovering new ALS genes including STMN2, which was independently discovered by two labs this year using experimental approaches. We are currently validating two additional ALS targets experimentally. We plan to focus on applying our platform to uncover novel genes and variants in AD using a similar study design as ALS.Objective 1: Obtain de-identified WGS data (e.g., ADSP and ADNI) from NIAGADS and preprocess them for input into GI’s computational pipeline. Objective 2: Run the AD WGS through our in-house genetic background dissector tool, Cato, that uses several machine learning systems to stratify genomes before analysis to avoid spurious results arising from differences in case and control backgrounds. Objective 3: Rediscover known AD genes. Chromosomes containing known AD genes (e.g., APOE, PSEN1 and PSEN2) will be input into our causal gene discovery platform, Bergspitze, to confirm if it can rediscover the known AD genes. The output from Bergspitze will be input into Franklin, GI’s interpretation module that provides a coherent etiology model for the disease with awareness of alternative etiologies advanced in the literature. Objective 4: Discovery of new AD genes. Run the full genomes from all cluster groups and ancestry cohorts in Bergspitze and Franklin to identify and prioritize new AD genes and variants. There are no plans to collaborate with other institutions.

Alzheimer’s disease (AD) is the leading cause of dementia affecting over 44 million people worldwide, and for which currently there are no effective prevention or cures. Enormous efforts and funding have been put into the discovery of the root cause of AD so as to find an effective treatment, but even with advancing genetic sequencing and analysis technology, no smoking gun has been found yet. Part of the issue has been scientific focus on analyzing the human exome, the ~1% of the human genome that codes for proteins, due to ease of analysis and lower noise in the data. However, this methodology precludes ~99% of the rest of the human genome, which harbors critical regulatory features that affect many of the processes in the body. Genetic Intelligence (GI) aims to solve this problem using its novel whole genome analysis platform that blends advanced genetic principles with state-of-the-art machine learning to identify causal disease targets. This information can then be used to create new drug candidates that are not only effective and precise, but also affordable.

Neuropsychiatric Symptoms (NPS) (e.g. aggression, psychosis, anxiety, apathy, depression, agitation, sleep disturbances, repetitive behaviors) occur in 85% of AD patients, and are associated with greatly increased suffering of patients and families. Despite this, our understanding of the etiology of NPS in AD is inadequate, with treatments for NPS often being ineffective and associated with serious adverse effects. This knowledge gap is particularly egregious in underserved racial and ethnic groups. The aim of the current project (U01AG079850) is to collate, harmonize, and analyze the AD-associated NPS data collected on ADSP/ADSP-FUS samples. We plan to (1) expand the racially and ethnically diverse datasets of the ADSP-FUS and related efforts to include harmonized NPS data, creating the largest and most diverse genomic resource on NPS in AD to date allowing researchers to assess a wide range of additional critical hypotheses through these resources; (2) utilize these harmonized data to identify and describe genetic determinants, pathways, and polygenic effects underlying specific NPS in AD; (3) explore the shared genetic architecture across AD-associated NPS and with primary psychiatric disorders; and (4) disentangle the role of ancestry in NPS genetic risk. We anticipate that this work will lead to a better understanding of the genetic basis of NPS in AD which is vital to infer the mechanistic pathways underlying these highly disabling symptoms and develop more effective pharmacological targets. To collate NPS data on all ADSP-FUS cohorts we closely collaborate with the ADSP-Phenotype Harmonization Consortium. Creation of refined harmonized NPS phenotypes will be conducted by Dr. Ted Huey’s group at Brown University. Genomic data analyses will be conducted by Dr. Reitz’ group at Columbia University and Dr. Beecham’s group at the Wake Forest School of Medicine.

Although neuropsychiatric symptoms (e.g. aggression, psychosis, anxiety, apathy, depression, and sleep disturbances) occur in ~85% of Alzheimer disease patients and are associated with accelerated decline, increased cost, out-of-home placement, and greatly increased suffering of patients and families, our understanding of their etiology is still inadequate, with treatments often being ineffective and even associated with serious adverse effects (including increased mortality). This knowledge gap is particularly egregious in underserved racial and ethnic groups such as Hispanics and African-Americans. We propose to expand the racially and ethnically diverse ADSP-FUS and related resources to include harmonized neuropsychiatric symptom data allowing researchers to assess a variety of additional critical hypotheses, and to utilize these harmonized data to identify ancestry-specific genetic determinants, molecular pathways, and polygenic effects underlying neuropsychiatric symptoms in Alzheimer disease.

In complex diseases like Alzheimer’s disease (AD), the disease risk is conveyed by genetic variants each one with a small effect on disease susceptibility, which may add up to a relevant impact in combination. Such complex synergistic effects can be detected by means of machine learning (ML) models such as decision tree ensembles, which are excellent candidates for detecting genetic variant interactions even beyond only pairwise interactions in candidate gene approaches. Explainable artificial intelligence can then explain the models and deliver information about effects and importance of the single variants. In a first application of this technique, we have formulated an aquaporin-4 (AQP4) single nucleotide polymorphisms (SNPs) based risk score for brain amyloid burden using data from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database (adni.loni.usc.edu). We plan now to replicate and extend our method with a larger dataset and include further candidate genes, e.g. circadian, sleep related or blood-brain-barrier related genes. The data from the NIAGADS data repository will expand the previously used database significantly. Of special interest are the data from the Alzheimer’s Disease Sequencing Project (ADSP). Besides a general increase in data, the racial/ethnic diversity of ADSP is also of high interest. Genetic variants might occur at different frequencies in different populations, so that rare combinations of variants in one population might be more frequent in another population. Access to these data will not only allow to replicate our previous findings, but also to extend our methodology to other candidate genes. Information on SNP status, together with further information like age, gender, APOE4 status, diagnosis and endophenotypes of AD, as available, will be used to train and explain a ML model. Information gathered from the explanations will be used to assign importance as well as protecting or deleterious effects to the SNPs, eventually allowing to formulate a risk score for (endo)phenotypes of AD or diagnosis.

In diseases like Alzheimer’s, genetic variations each with a small effect may together affect the disease risk. These variations can add up to have a significant impact. We can use machine learning, like decision tree ensembles, and explainable AI to find these combined effects. We've already used this method to create a risk score for brain amyloid burden in Alzheimer's. Now, we want to expand this work using a larger dataset and looking at more genes, e.g. genes related to sleep. We want to use a new dataset that includes more data and diverse genetic information. Different populations may have different genetic variations, so having more diverse data will help us to understand these differences. We'll use this data, along with other information like age and gender, to train a machine learning model. By understanding how the model makes its decisions, we can figure out which variants of a gene are most important and how their combination affect Alzheimer's risk. This will help us to create gene-specific risk scores for Alzheimer's and related conditions.

• Objectives: In this project, we seek to holistically investigate the genetic and molecular drivers of sex dimorphism in Alzheimer’s disease across ancestries. • Study design: This study integrates large-scale population genetics with multi-omics and endophenotype analyses. We are integrating all data available from ADGC and ADSP, together with other data from AMP-AD and biobanks such as UKB, FinnGen, and MVP to conduct large-scale multi-ancestry GWAS, rare-variant gene aggregation analyses, QTL studies, PWAS, TWAS, etc. We also particularly focus on X chromosome association studies. The study design also interrogates interactions with ancestry, hormone exposures, and with APOE*4, as well as comparisons to non-stratified GWAS/XWAS of Alzheimer’s disease. Further, we will also employ genetic correlation analyses, mendelian randomization, colocalization, and pleiotropy analyses, to interrogate overlap with other complex traits to better understand the mechanisms underlying sex dimorphism in Alzheimer’s disease. • Analysis plan, including the phenotypic characteristics that will be evaluated in association with genetic variants: Our phenotypes will include Alzheimer’s disease risk, conversion risk, various endophenotypes (including amyloid/tau biomarkers, brain imaging metrics, etc.) as well as molecular traits. As noted above, we will conduct large-scale multi-ancestry GWAS, XWAS, rare-variant gene aggregation analyses, QTL studies, PWAS, TWAS, etc. Specific aims include interrogating these question and analyses on (1) the autosomes, (2) the X chromosome, and (3) leveraging sex stratified QTL studies to drive discovery of risk genes.

Alzheimer’s disease (AD) manifests itself differently across men and women, but the genetic and molecular factors that drive this remain elusive. AD is the most common cause of dementia and till today remains largely untreatable. It is thus crucial to study the genetics of AD in a sex-specific manner, as this will help the field gain important insights into disease pathophysiology, identify novel sex-specific risk factors relevant to personalized genetic medicine, and uncover potential new AD drug targets that may benefit both sexes. This project uses large-scale genomics and multi-omics to elucidate novel sex agnostic and sex-specific AD risk genes. We will interrogate sex dimorphism for AD risk on the autosomes and the sex chromosomes. We similarly interrogate sex dimorphism in the genetic regulation of gene expression and protein levels, which we will integrate with genetic risk for Alzheimer’s disease to further discovery risk genes. Throughout, we will also interrogate how sex-specific risk for AD interactions with hormone exposures, ancestry, and the APOE*4 risk allele.

Genome-Wide Association Studies (GWAS) are powerful techniques for linking genetic variation to complex phenotypes but have limited utility in small samples. Because the overwhelming majority of genotyped cohorts consist of individuals of European ancestry, conducting well-powered GWAS in diverse populations will be a challenge for years to come. This project overcomes this barrier by developing a meta-analysis framework for GWAS conducted in populations of different ancestries. In particular, multi-ancestry meta-analysis (MAMA) implements a generalized method of moments estimator based on differences in local linkage disequilibrium (LD) structure across the relevant populations. In doing so, MAMA allows for genetic signal in one population to be shared across other populations, substantially boosting statistical power in small samples and allowing for novel genetic associations to be detected. The primary goal of our proposed project is to use MAMA to jointly analyze GWAS summary statistics for Alzheimer’s disease corresponding to several different ancestries. Preliminary applications of MAMA to other phenotypes has yielded many additional genomewide significant loci for each ancestry. MAMA summary statistics can be interpreted similar to the original GWAS summary statistics, but with greater statistical power. We hope to incorporate NIAGADS data from the study Kunkle et al. (2020) into our analysis pipeline. We will jointly analyze these data with summary statistics from Kunkle et al. (2018) and from Zhou et al. (2018), studies based on European-ancestry and East-Asianancestry samples, respectively. We anticipate that doing so will yield many novel genetic discoveries about the genetics of Alzheimer’s disease in each of these populations.

Understanding the link between genetic variation and the incidence of diseases like Alzheimer’s remains an urgent task in the genetics community. One outstanding challenge involves incorporating people with non-European ancestries in such studies. This is an important effort for two reasons. First, generalizing findings from homogenous populations is difficult and hinders a broader understanding of the relevant disease. Second, such an inability to generalize findings will likely perpetuate existing health inequalities. It will likely take many years before enough people with non-European ancestries have been genotyped to match the precision currently available in European-ancestry samples. Rather than waiting for such resources to materialize, our project develops a statistical framework that allows existing cross-ancestry data to be meta-analyzed to improve our ability to detect genetic associations in understudied populations. The NIAGADS dataset, combined with our methodology, will allow for a broader understanding of Alzheimer’s to be developed.

Alzheimer’s disease (AD) is a common, progressive, neurodegenerative disorder with a strong genetic component with heritability estimates ranging from 58–79% for late-onset AD and over 90% for early onset AD. Genetic association studies are important to highlight key biological mechanisms contributing to the etiology of AD and provide key insights into potential pathways that can ultimately be targeted for future therapeutic development. The objective of this study is to perform a retrospective analysis of genetic data collected from large-scale population-based and case-control cohorts including the UK Biobank, the Alzheimer’s Disease Sequencing Project (ADSP), and FinnGen and integrate them with publicly available multi-omics datasets including, but not limited to, Genotype-Tissue Expression (GTEx), Microglia Genomic Atlas (MiGA), and neuroimaging data to identify novel and existing evidence for genetic determinants of AD. No attempt will be made to try and identify subjects. Aim 1: Identify novel and replicate existing gene associations for AD. We will perform case-control and family-based genetic analyses with AD diagnosis as the outcome of interest. Covariates include age, sex, and principal components. ADSP, UKB, and FinnGen will be analyzed separately and combined with meta-analysis. Biobank cases will be defined using ICD-9/ICD-10 codes, and proxy cases and controls will be carefully defined using questionnaire data on parental history of AD. Both true and proxy cases will be considered to maximize the number of AD cases. Aim 2: Prioritize novel gene associations identified in Aim 1. We will perform genetic fine-mapping and leverage tissue and cell-type specific datasets (e.g. GTEx and MiGA) to prioritize targets for further functional and analytical interrogation. Statistical methods used for target prioritization include colocalization, statistical fine-mapping, and Mendelian randomization. Furthermore, multi-omics-based network approaches will be used to identify disease-related molecular modules and tissue-specific regulatory circuits.

Alzheimer’s disease (AD) is a common, progressive, neurodegenerative disorder with a strong genetic component with heritability estimates ranging from 58–79% for late-onset AD and over 90% for early onset AD. To date, there is only one treatment option intended to mediate the disease progression of AD, while all others treat symptoms associated with AD. Genetic association studies are important to highlight key biological mechanisms contributing to the etiology of AD and provide key insights into potential pathways that can ultimately be targeted for future therapeutic development. The objective of this study is to perform a retrospective analysis of genetic data collected from large-scale population-based and case-control cohorts including the UK Biobank, the Alzheimer’s Disease Sequencing Project (ADSP), and FinnGen and integrate them with publicly available multi-omics datasets including, but not limited to, Genotype-Tissue Expression (GTEx), Microglia Genomic Atlas (MiGA), and neuroimaging data to identify novel and existing evidence for genetic determinants of AD.

WE have published a first paper on this RUS, but we will continue with the basic RUS goals regarding adaptive immune receptors and HLA allele typing.The below RUS, part of which is below, is still applicable but has been truncated to stay consistent with character limitations. As far as updates are concerned, we have the following paper in press at Medical Hypotheses: CMV as a factor in the development of Alzheimer's disease? And, we have the following manuscript submitted: Association of the HLA-DQB1*02 allele with lower tauopathy in Alzheimer’s disease. Thus, we will first identify immune receptor (IR} recombinations found within the AD, blood exomes; and then match features of these IRs, such as the chemical aspects of the IR antigen binding sites [1, 4, 5] with clinical characteristics, e.g., age of onset, diagnostic status, and cognitive measures. We thus expect to identify certain IR recombinations associated with distinct prognoses. It should also be noted that there have been previous studies linking HLA-DR alleles to late-onset AD [10]. This is of particular interest, because of our previous work linking T-cell receptor, V or J usage, HLA allele combinations to distinct cancer survival rates [2, 3, 7, 8]. Thus, we will also programmatically obtain the HLA alleles from the exome files and determine whether any T-cell receptor, V or J usage, HLA allele combinations are associated with particular features of AD development.References available by email: gblanck@usf.eduAs best as applicant understands, all attachments are still in effect. If any difficulties with attachments, please email applicant.

The purpose of this project is to learn whether there are any features of the immune system that are unique to Alzheimer's patients. If so, such unique features might help understand disease progression better and might provide targets for therapies.

Objective: Utilize ADSP data sets to identify genes & specific genetic variants that confer risk for or protection from Alzheimer disease. Aim 1: Using combined WGS/WES across the ADSP Discovery, Disc-Ext, and FUS Phases, including single nucleotide variants, small insertion/deletions, and structural variants. We will: Aim 1a. Perform whole genome single variant and rare variant case/control association analyses of AD using ADSP and other available data; Aim 1b. Target protective variant identification via association analysis using selected controls within the ADSP data and performing meta-analysis across association results based on selected controls from non-ADSP data sets. Aim 1c. Perform endophenotype analyses including cognitive function measures, hippocampal volume and circulation beta-amyloid ADSP data in subjects for which these measures are available. Meta analysis will be conducted across ADSP and non-ADSP analysis results. Aim 2: To leverage ethnically-diverse and admixed populations to identify AD variants we will: Aim 2a. Estimate and account for global and local ancestry in all analyses; Aim 2b. Perform admixture mapping in samples of admixed ancestry; and Aim 2c. Perform ethnicity-specific and trans-ethnic meta-analyses. Aim 3: To identify putative therapeutic targets through functional characterization of genes and networks via bioinformatics, integrative ‘omics analyses. We will: Aim 3a. Annotate variants with their functional consequences using bioinformatic tools and publicly available “omics” data. Aim 3b. Prioritize results, group variants with shared function, and identify key genes functionally related to AD via weighted association analyses and network approaches. Analyses will be performed in coordination with the following PIs. Coordination will involve sharing expertise, analysis plans or analysis results. Anita DeStefano, Boston University, Philip De Jager, Columbia University; Eric Boerwinkle & Myriam Fornage, University of Texas Health Science Center, Houston; Sudha Seshadri, University of Texas, San Antonio; Ellen Wijsman, University of Washington; Richard Gibbs, Baylor College of Medicine.

This proposal seeks to analyze existing genetic sequencing data generated as part of the Alzheimer’s Disease Sequencing Project (ADSP) including the ADSP Follow-up Study (FUS) with the goal of identifying genes and specific changes within those genes that either confer risk for Alzheimer’s Disease or provide protection from Alzheimer’s Disease. Analytic challenges include analysis of whole genome sequencing data, appropriately accounting for population structure across European ancestry, Hispanic, and African American participants, and interpreting results in the context of other genomic data available.

Objectives of the proposed research: Alzheimer’s disease and its related dementias (ADRD) are a growing public health crisis with no known cure. ADRD diagnosis remains challenging due to its inherent heterogeneity, variability of early symptoms, overlap and possibility of combined etiologies. The overarching objective of this project to develop novel interpretable deep learning methods to integrate multimodal neuroimaging, genomic and clinical datasets collected from demographically diverse cohorts for early detection of ADRD. Study design: Our central hypothesis is that an integrative deep learning model that operate on both MRI data, and NGS data will be an ideal solution for isolating ADRD biomarkers that can lead to early diagnosis. We aim to integrate multi-modal datasets including genomic, clinical, demographic, neuroimaging from diverse populations in studies such as ADNI, ADSP, and HEBLA and develop a multimodal fusion-based approach for optimized and accurate classification of various presentations of ADRD. Further, we will design and develop an interpretability-framework that would help explain the decisions made by deep-learning models leading to knowledge discovery for neuroscientists, and transparent analysis for clinicians. Analysis plan: We plan to utilize genetic variants as features in our deep learning model. To this end, we will perform GWAS to find SNPs that associate with disease diagnosis and endophenotypes such as neuroimaging, biospecimen measures, and cognitive performance. We will utilize genetic data to perform a network propagation study to discovery AD-associated genes.

Alzheimer’s disease and related dementias (ADRD) are a growing health crisis, with current global management costs over $350 billion and cases expected to grow to 14 million globally by 2050. Emerging evidence of advantages of early detection of the disorder, and the current limitation of the clinical practices suggest that more quantitative methods are needed to identify, and early diagnose ADRD. In this project, we aim to develop interpretable deep learning models to integrate genetic, imaging, and clinical datasets for diagnosis and early prediction of ADRD.

Objectives To identify genetic variants that are overrepresented in sporadic dementias when compared with controls. To identify genetic variants that are found uniquely in apparently sporadic dementia cases. To determine if rare variants identified in our cohorts of neurodegenerative disease are present in the ADSP controls. Study Design We have performed exome-sequencing in over 3,000 samples from a variety of neurodegenerative dementias. These data were generated using Illumina sequencing and called using GATK’s Best Practices v3. In these samples, we have identified genetic variants that have either much lower frequency in controls and in publicly available databases of genetic data (gnomAD), or that are absent from these cohorts. The proposed study design is largely a case-control study in ADSP data to replicate our findings as well as a simple lookup for rarer variants in cases and controls, where sample size isn’t enough to perform meaningful associations. We will perform single variant and gene-based associations using standard methods (fisher test/logistic regression and SKAT-O) using gender, age and principal components as covariates. These tests are either implemented in PLINK or can be performed in R. To allow us to dissect the associations between genetic variants and phenotype we will require access to gender, age at onset, age at death (where available), Braak staging and CERAD scores for all cases in ADSP. Funding Funding for the study is currently from Van Andel Research Institute’s internal funds.

The main objective of the study is to identify genetic variants that cause or predispose to neurodegeneration. To accomplish this, we will analyze data previously generated for a variety of these conditions and use data from ADSP to replicate findings and improve our statistical power to detect these associations with disease. The identification of genetic variants, even if rare, that have a strong impact on dementia phenotypes will be of significant importance in advancing our understanding of disease biology. These variants will also be candidate targets for future diagnostic or therapeutic approaches for these diseases.

The ADSP-PHC was established to harmonize the rich endophenotype data across cohort studies to enable modern genomic analyses of ADRD with the ultimate goal to generate harmonized data that will become a “legacy” dataset perpetually curated and shared NIAGADS. The ADSP-PHC will facilitate phenotypic data harmonization for ADSP participants with genetic and genomic data. This effort represents a multi-disciplinary approach leveraging interdisciplinary expertise in large-scale genetic and genomic studies, clinical neuroscience, neuroimaging, psychometrics, and bioinformatics. This study will utilize a team structure consisting of two coordinating centers to oversee activities of all harmonization teams and to oversee coordination with other ADSP workgroups and initiatives, a Storage and Informatics Core that will oversee the coordination between LONI and NIAGADS for all data storage, compliance, and dissemination, a CHARGE Coordination Core to oversee the alignment of data and protocols with CHARGE workgroups, an Integration & Analytics Core that will enable data integration across phenotypes to facilitate downstream machine learning applications, and eight Domain-Specific Harmonization Teams tasked with harmonization in their area of expertise. The endophenotypes that will be harmonized by this project include Cognition, Fluid Biomarkers, Amyloid PET, Structural MRI, White Matter Hyperintensities, White Matter Integrity, Vascular Risk Factors, and Neuropathology. Site PIs on this project include: Jesse Mez (Boston University), Adam Brickman (Columbia University), Andy Saykin (Indiana University), Elizabeth Mormino (Stanford University), Pauline Maillard (UC Davis), Duygu Tosun-Turgut (UC San Francisco), Christos Davatzikos (University of Pennsylvania), Arthur Toga (USC); Mohamad Habes (University of Texas Health Science Center at San Antonio), Michael Cuccaro (University of Miami), Paul Crane (University of Washington), Bennett Landman (Vanderbilt University), Timothy Hohman (Vanderbilt University Medical Center), and Carlos Cruchaga (Washington University in St. Louis).

The growing availability of endophenotypic data in cohort studies of Alzheimer’s disease and related dementias (ADRD) provides an exciting opportunity to further characterize the genetic architecture of this devastating disease. However, there is a pressing need to develop and apply advanced harmonization approaches to align ADRD endophenotypes across cohorts. The goal of this coordinated national initiative – the AD Sequencing Project Phenotype Harmonization Consortium, or “ADSP-PHC” – is to provide a centralized database of robust endophenotypes for large-scale genomic analyses that will accelerate the identification of novel targets for therapeutic intervention in ADRD. will accelerate the identification of novel targets for therapeutic intervention in ADRD.

Non-invasive and low-cost tools for early detection of individuals at high risk for developing Alzheimer’s Disease (AD) are of crucial importance. While genetic studies (GWAS) exist, most focus on European populations, limiting their application to diverse groups. Given the Eurocentric biases in GWASs, PRSs are better at predicting AD risk for European ancestry as opposed to others. This research addresses this gap by developing a multimodal hazard score (MHS) that incorporates ethnically and genetically diverse populations. Our team established a successful polygenic hazard score (PHS) that predicts AD onset in European cohorts. Higher PHS predicted greater cognitive decline in CN, entorhinal cortex volume loss and predicted conversion from cognitively normal (CN), to mild cognitive impairment (MCI) to AD. The PHS was also associated with neuropathological AD phenotypes and systematically varied AD biomarkers such as amyloid and total tau PET. This suggests its potential as a robust genetic risk indicator to be included in the MHS. Leveraging data from the largest AD cohorts to date, including ADSP, we will conduct Cox proportional models to develop the MHS combining age, PHS, brain atrophy, biofluid-based data, and clinical outcomes to predict neurocognitive decline trajectory. This score will be able to predict AD progression and identify individuals at high risk of transitioning from cognitive normalcy to mild cognitive impairment (MCI) and eventually AD in a generalizable population. Additionally, we will also compute power calculations to estimate required clinical trial sample sizes after hypothetical enrichment using the MHS. Finally, we aim to replicate these findings in other diverse cohorts to ensure broader applicability. We believe this MHS surpasses single-modal and European data-driven models in predicting AD for diverse populations. This could pave the way for earlier intervention, more efficient research, and ultimately, improved outcomes for individuals at risk of AD.

Alzheimer’s Disease (AD) disproportionately affects diverse populations, highlighting the need for non-invasive and low-cost tools for early detection. This research aims to develop a multimodal hazard score (MHS) in ethnically and genetically diverse populations. The MHS combines various risk factors, including an individual's genetic profile, neuroimaging, and specific blood markers (biomarkers). This comprehensive approach offers the potential to identify individuals at high risk of transitioning from cognitively normal (CN) to mild cognitive impairment (MCI), ultimately leading to AD. To develop the MHS, we will leverage data from the largest AD cohorts to date, including ADSP. We will then evaluate MHS on its ability to identify individuals with greater risk for converting from CN to MCI and AD on the generalizable population and compute power calculations to estimate required clinical trial sample sizes. This technique could pave the way for earlier intervention, more cost-efficient research, participant burden and ultimately, improved outcomes for individuals at risk of AD.

The goal of this collaborative project is to identify and prioritize genetic variations that contribute to late-onset Alzheimer’s disease (AD), with a particular focus on developing mouse models to understand the biology and of and develop new treatments for AD. We use a variety of approaches to analyze high-throughput genomic and genetic data, including statistical methods for identifying causal variants and genetic interaction networks, as well as epigenetic analyses of cellular regulation. By investigating and quantifying the effects of these mutations individually and interactively, we hope to begin to understand the biology of AD and unravel the complexity of AD genetic risk. Our analysis will use the human genetic data in the Alzheimer’s disease sequencing project (ADSP), from which we will identify high-priority candidate genes and variants for further study in mouse models. Thus the goals of this work are twofold: to understand how brain irregularities progress into full AD, and provide the research community with a valuable mouse model for further studies and therapeutic testing. The described research will pilot novel analytical tools for the study of the biology of AD and provide insights into the genetics of neurodegeneration. We will use independent evidence to identify candidates with putative functional roles. Transcriptome data from Alzheimer’s studies will be used to identify candidate genes and ENCODE and other regulatory data will be used to identify putative regulatory regions. Variants will be ranked based on computationally predicted mutation severity and differential expression in AD. Candidates in regulatory regions will be prioritized based on expression differences in nearby genes. Finally, functional data sources such as Gene Ontology, Allen Brain Atlas, and mouse phenotypes will be used to determine the potential role in neurodegeneration. The result will be a list of candidate variants prioritized for study in mouse models. Since these data will be integrated with a focus on the role of individual genes generated from a population-wide analysis, we do not foresee creating any additional risk to individual participants.

While many genetic loci have been identified as contributing to the risk of late-onset Alzheimer’s disease, the biological underpinnings and interdependence of these mutations are generally poorly understood. By investigating and quantifying the effects of these mutations individually and interactively, we hope to begin to unravel the complexity of genetic risk. Data from the Alzheimer’s Disease Sequencing Project will be essential in this approach, allowing us to identify the best candidate genes and perform computational analysis to design advanced mouse models of AD. These models will be an experimental basis for understanding the biology of Alzheimer’s and performing early-stage testing of candidate therapeutics.

The objectives of this proposal are to identify rare genetic risk factors that are unique to or shared by Alzheimer’s disease (AD) and Progressive Supranuclear Palsy (PSP).We hypothesize we will identify a set of rare genetic risk factors associated with AD and another set of rare genetic risk factors associated with PSP, some of which may be shared between AD and PSP. To determine shared or unique genetic risk, we will compare AD to controls and PSP to controls. Both analyses will use the same controls and have similar number of cases. AD subjects will be included from Alzheimer’s Disease Sequencing Project, Alzheimer’s Disease Neuroimaging Initiative, and Accelerating Medicines Partnership – Alzheimer’s disease. From these studies, we will include adult controls. We will use roughly 1900 whole genomes from PSP subjects. Given the availability of AD sequencing, we will replicate the association of rare genetic risk with an independent holdout AD dataset, which will include AD and controls from the ADSP Follow Up Phase. We will also validate our finding in multi-ethnic cohorts from the ADSP.Traditional rare variant analyses have limited power due to the large number of variants and small variant effect size. Although one solution is to group variants into genes, genes do not act in isolation, but rather interact with one another in networks. Grouping variants in a network can improve power. Additionally, since most genetic risk lies in large noncoding regions of the genome, focusing analyses on noncoding regulatory regions should further increase power. We will incorporate network connectivity in rare variant statistical tests and prioritize functional noncoding variants will identify rare genetic risk factors in AD and PSP by overcoming deficiencies in traditional methods. Analyzing rare variants in protein coding, promoter and distal noncoding regions, we will compare the proposed network and non-coding prioritization methods to traditional gene-based, unprioritized non-coding methods.

Neurodegenerative diseases including Alzheimer’s disease and Progressive Supranuclear Palsy are characterized by abnormal tau protein accumulation and do not currently have disease modifying treatments. Analyzing whole genome sequencing with novel genomic and genome informatic methods may identify rare genetic risk factors that lead to these diseases. The shared or unique rare genetic risk factors of Alzheimer’s disease and Progressive Supranuclear Palsy may become future therapeutic targets.

We plan to analyze whole exome and whole genome sequence data generated from subjects with progressive supranuclear palsy (PSP), corticobasal degeneration (CBD), Alzheimer's disease (AD) and elderly normal controls. The goal is to detect mutations and variants that cause, contribute to risk, or protect against PSP and/or CBD. We want to compare PSP and CBD genotypes to those from AD and normal controls sequenced by the Alzheimer's Disease Sequence Project. We would like both whole genome and whole exome data from the Alzheimer's Disease Sequence Project for AD and normal controls. We would also like whole genome and whole exome data for PSP and CBD generated by the PSP and Tau consortiums. We will use these data to determine which mutations and variants are associated with PSP or CBD versus benign variants. All PSP and CBD subjects being sequenced are deceased. The requested data sets will have variants recalled as a batch and combined to evaluate allele frequencies of called variants. The AD and control variant frequencies will then be compared to allele frequencies from PSP and CBD subjects as described above. We will also compare structural variants (insertion-deletions, copy number variants, and chromosomal rearrangements) identified in PSP and CBD subjects to those found in AD and in cognitively normal controls in order to determine structural variants involved in PSP and CBD pathogenesis. All of the investigators that are listed will be using a joint called VCF generated from the requested data sets. PSP is a neurodegenerative disease closely related to Alzheimer's disease (AD). PSP, CBD and AD have neurofibrillary tangles as part of the signature neuropathology defining these disorders. PSP and CBD are considered Alzheimer’s Disease Related Disorders (ADRD).

We are attempting to identify all the inherited elements that contribute to progressive supranuclear palsy (PSP) and corticobasal degeneration (CBD) risk. To do this we will analyze DNA sequence data from subjects with AD, PSP, CBD, and subjects who are cognitively normal. The sequence data from these groups will be compared to identify differences that contribute to the risk of developing PSP and CBD, or that protect against these diseases. These DNA differences can be at a single site in the genetic code, or can span multiple sites, changing the copy number of DNA sequences. Both types of genetic variants will be examined.

Alzheimer's disease(AD) is the most common cause of dementia, accounting for 60% to 80% of cases that affect over six million people in the United States. The disease gradually progresses from mild cognitive impairment(MCI) to dementia, which takes more than a decade. Identifying individuals who have a high risk of AD earlier is essential for AD prevention and intervention. As the heritability of AD is high(up to 79%), genetic data should be powerful to identify individuals at high risk. Indeed, polygenic risk score (PRS), designed to estimate individual genetic liability by integrating large GWAS summary statistics and individual genotype data, has been shown to be promising for AD risk prediction(AUCs up to 84%). However, the prediction accuracy using a single PRS is still not sufficient for MCI and AD classification in clinical practice. We hypothesize that convolution neural network(CNN) models can improve the classification of AD and MCI by multiple integrating PRSs from multiple traits, multi-omics data (genotyping data, scRNA-seq), clinical data, and imaging data. The objective is to develop advanced AI algorithms and build data-driven models for disease risk assessment, earlier identifying individuals with high risk for MCI and AD. Our long-term goal is to develop and validate a prediction model that can be translated into clinical practice. Our CNN model has recently shown an improved performance for AD with PRSs from multiple traits(AUC 92.4%). We want to extend our approach to predicting AD and MCI in different ethnic groups and validate the results with independent datasets. To this end, we would like to apply for multi-omics data in NG00067.v9 from https://dss.niagads.org/datasets/ng00067/. With an extensive experience in genetic studies on complex disorders and disease modeling, we are confident that we will achieve the specified goals and promote the integration of genetic data with AI algorithms, facilitating data-driven, personalized care of AD. We expect to finish this study within 2 years with publication and grant application. We have IRB approval and will follow the rules for data sharing and acknowledgment.

Alzheimer’s disease (AD), the most common form of dementia, that usually develops from mild cognitive impairment to dementia. There is currently no treatment to slow the progression of this disorder. But earlier identification of the individuals with higher risk maybe critical to prevent the disease. We propose a new approach to create models for classification of AD and MCI with artificial intelligence and genetic data. This study will have a significant value in personalized medicine for AD risk assessment, classification, and earlier intervention.We don’t have the planned collaboration with researchers outside Cleveland Clinic in the current analytic plans.

We propose to develop capable and intelligent computer-based toolboxes that enable searching, sharing, visualizing, querying, and analyzing genetics, genomics, multi-omics, and clinical data for AD. The central unifying hypothesis of this U01 project (U01AG073323) is that a genome-wide, multimodal artificial intelligence (AI) framework to identify novel risk genes and networks from human WGS/WES and multi-omics findings will offer drug targets for targeted therapeutic development in AD. Aim 1 will identify rare coding variant-based risk genes using a sequence and structure-based deep learning model. Aim 2 will identify rare non-coding variant-based risk genes using a multiple kernel learning approach. Aim 3 will test whether GWAS common variants linked to AD pathobiology and endophenotypes are enriched in gene regulatory networks in a cell-type specific manner using a Bayesian framework. These analyses will leverage variants from ethnically diverse WGS/WES and clinical data (i.e., imaging, biomarkers, and cognitive measures) from Alzheimer's Disease Sequencing Project (ADSP), and publicly available chromatin interactomic data from NIH RoadMap, FANTOM5, and NIH 4D Nucleome. We will validate our findings using WGS/WES data and protein expression data from our existing cohorts: The Cleveland Clinic Lou Ruvo Center for Brain Health Aging and Neurodegenerative Disease Biobank (CBH-Biobank) and the Cleveland Alzheimer's Disease Research Center (CADRC). We will compile information for clinical data harmonization, including functional imaging, AD biomarkers, and cognitive measures for all integrative analyses. There are no any PHI information will collected or used in the data analysis. We don’t have the planned collaboration with researchers outside Cleveland Clinic in the current analytic plans.

It is estimated that more than 16 million people with AD live in the United States by 2050 and the predisposition to AD involves a complex, polygenic, and pleiotropic genetic architecture. This project will develop intelligent computer-based network medicine and systems biology tools, capable of identifying and validating human genome sequencing findings for novel risk gene discoveries and targeted therapeutic development in AD. The innovative network-based, artificial intelligence toolboxes and novel risk genes and biologically relevant targeted therapeutic approaches developed in this proposal will prove to be novel and effective ways to improve outcomes in long-term brain care for the rapidly growing AD population, an essential goal of AD precision medicine.

We seek access to ADSP data for two purposes. First, my lab is focused on understanding regulatory mechanisms of rare non-coding variation in neurodegeneration-associated loci. As a part of these efforts, we have generated data internally including 10X multiomics (matched single nucleus RNA-seq and ATAC-seq) and genome-wide restriction fragment resolution HiC data, which, along with publicly available data, we have used to nominate regulatory elements for neurodegeneration-associated genes. One question is if genetic variation in cases is enriched in these regulatory regions compared to controls, which we will use ADSP data to assess.As a second independent effort, we are a part of collaborations with the Yokoyama lab at UCSF and the Kosik lab at UCSB as well as The Multi-Partner Consortium to Expand Dementia Research in Latin America (ReDLat) to analyze genomes for early onset Alzheimer’s and frontotemporal dementia cohorts compared to unaffected controls. A critical part of these efforts is replication of any findings in independent cohorts. Access to Alzheimer's Disease Sequencing Project (ADSP) data is ideal for this purpose. We will analyze ADSP data for association signals identified in our independent cohorts using either single variant or burden analysis approaches. Phenotypic characteristics that will be evaluated in association with genetic variants will be either case/control status or age of symptom onset as available. Although we conduct these projects as collaborations, this application is for analysis of ADSP data at HudsonAlpha.We seek access to all consent groups. Our research does not include the study of population origins or ancestry, and thus qualifies for HMB-designated studies. Our research is also applicable to each of the disease-specific (DS) categories: for example, we are interested in effects in Alzheimer's and related dementias as well as other phenotypes related to aging, brain and memory. We also note that this analysis involves methods development research (MDS) (new approaches to understand effects of non-coding variation). Finally, this is a genetics study only (GSO).

Our lab aims to understand function of the genome to gain confidence in the precise way in which genetic changes lead to risk for disease. We are working to identify stretches of DNA near genes associated with Alzheimer’s and related dementias and/or aging, brain and memory that may be involved in turning these genes on and off. A key test is if genetic changes in these regions are enriched in people with disease, which we will use these ADSP data to assess.In a second project, we work with members of The Multi-Partner Consortium to Expand Dementia Research in Latin America (ReDLat) to to analyze the DNA from patients with early onset Alzheimer’s and frontotemporal dementia. A critical part of this type of work is checking to see if findings from one set of patients are reproducible in different sets of patients. Access to ADSP data would allow for us to answer this question. The types of data that will be evaluated in association with genetics will be either if the individuals assessed have disease or not, or if their genetics affects when they develop disease.

Alzheimer’s disease (AD) is a highly heterogeneous disease with diverse clinical manifestations and rate of progression. AD subtypes identification will improve our understanding of the underlying disease mechanisms, enable us to predict disease trajectory and develop new disease-modifying treatments. Objective. To identify distinct AD subtypes from WGS data of AD individuals Analysis plan. We will use 3000 WGS data derived from the ADSP Discovery Case-Control Based Extension Study. We will use the available SNVs and INDELS and infer structural variants (SVs) with our in-house multi-caller pipelines. Rare variants will be retained for further analysis. We will then split the dataset in training and tests set, and use the identified set of genetic variants (i.e. SNVs, INDELS and SVs) as input to a deep neural network (an autoencoder architecture) to learn an unsupervised latent representation of the data. AD subtypes will be identified within this reduced space and characterized using, demographics and clinical data. We will then contrast each subtype with the control groups to identify subtype relevant variants (i.e. putative subtype biomarkers), which will be used as input features to a gradient boosted tree model, to generate a subtype predictive model and subtype specific features. Planned collaboration. Each member of the team will devote effort in specific areas of investigation, nevertheless, all the team members will discuss, through regular meeting, individual progress and potential challenges. In particular, Dr Coppola (Research Scientist, Department of Pathology, Yale University, USA), together with Dr Dean Palejev (Associate Professor, GATE Institute, Sofia University, Bulgaria) will be involved in the deep learning model generation and validation, and subtype identification; Dr Fredrik Johansson (Assistant Professor, Department of Computer Science & Engineering, Chalmers University of Technology. Sweden), will work on the supervised machine learning model; Dr Alexander Schliep, Associate Professor, Department of Computer Science & Engineering, University of Gothenburg, Sweden), will work on the SVs inference.

Alzheimer’s disease (AD) is a highly heterogeneous disease with diverse clinical manifestations and rate of progression. The heterogeneity of AD has complicated both clinical trial design and outcomes, and thus the need for better models of AD, and/or better strategies for selection of participants into specific clinical trials is evident. The identification of more homogeneous disease subgroups (i.e. AD subtypes) will improve our understanding of the underlying disease mechanisms, enable us to predict disease trajectory and develop new disease-modifying treatments. We will use a comprehensive set of genetic variants in combination with deep learning algorithms to identify AD subtypes. Subtypes will be characterized using clinical and demographic data. Finally, variants specific to each cluster will be identified and used to train a predictive machine-learning model to classify new individuals.

Objective: To leverage large-scale electronic health care record data and biomarkers of neurodegeneration to improve gene discovery in Alzheimer’s disease. Study design: The goal of our work is to integrate common and rare genetic variation across multiple cohorts with the genetic data collected in the ADSP to improve gene discovery in Alzheimer’s disease (AD) and related neurodegenerative conditions. We plan to perform both genome-wide and exome-wide association analyses (GWAS/ExWAS) with phenotypes including any dementia, AD, mild cognitive impairment (MCI), age of onset for cognitive decline, and quantitative readouts of neurodegeneration including cognitive tests and brain MRI measures where available. We will integrate our findings with expression quantitative trait loci and single cell gene expression to understand pathways and mechanisms modulating risk for these phenotypes.Analysis plan: First, we will integrate data from ADSP with electronic health and genetic data at the Regeneron Genetics Center (RGC) to harmonize both genetic variation and phenotype data. Next, we will use standard approaches to perform GWAS/ExWAS, perform meta-analysis across cohorts, and post-GWAS analyses to annotate our findings across traits. All data will remain anonymized and securely stored, we will not share any of the individual level data outside of Regeneron or beyond the researchers on our application. We have a secure computational environment to store these data and IT staff dedicated to ensuring we comply with the necessary requirements delineated by the NIAGADS.

Therapeutic development in Alzheimer’s disease has greatly benefitted from understanding the genetics of this disorder. However, the role of rare genetic variation and the impact of genetics on biomarkers in Alzheimer’s remain mostly unknown. Our goal is to develop new therapies, and the data from the ADSP will help us prioritize potential molecules and pathways to pursue in Alzheimer’s disease and related conditions.

Objective: To nominate putatively functional rare noncoding variants in ADStudy Design: We have developed a novel pipeline for noncoding variant prioritization that combines principles from statistical genetics, gene regulation, and machine learning. We have previously used this type of pipeline to prioritize noncoding variants within known genetic risk loci for AD (PMID: 33106633). In brief, this pipeline: 1. Identifies all known LD-expanded variants that have been significantly associated with AD 2. Filters these variants for those overlapping gene regulatory elements in specific cell types of the brain 3. Uses machine learning to predict which variants will have strong effects on transcription factor binding 4. Uses functional genomics technologies including massively parallel reporter assays and CRISPR-based genome editing to pinpoint which of the nominated variants have validated functional effectsSo far, this pipeline has only been applied to common variants identified by GWAS but we aim to apply this same methodology to nominate functional noncoding rare variants. With this in mind, we will:1. Download all ADSP WGS datasets to identify all variants discovered in AD cases and controls 2. Annotate each variant with its frequency both within the ADSP cohorts and within the general population using resources such as gnomAD and TOPMED. 3. Input rare variants implicated in AD (i.e. either only observed in AD or observed more frequently in AD than in the general population) in the above described pipeline to functionally validate a subset of rare variants as putative noncoding drivers of disease. 4. Link any putative functional variants to their cell type-specific target genesThe result of this work would be a list of variants with putative functional roles in AD and their putative target genes. In this study, we do not plan to associate any phenotypic characteristics other than AD vs Non-AD. No collaboration is anticipated.

Alzheimer's disease (AD) is driven by both genetic and non-genetic factors. Previous studies have estimated that ~60% of the susceptibility to AD can be attributed to genetic factors. This heritability can be roughly evenly divided between (i) common genetic variation and (ii) rare or structural variation. Of the rare genetic variation driving AD, we understand vanishingly little. This is because >90% of rare genetic variants lie within the noncoding regions of the genome. These noncoding regions harbor gene regulatory elements but do not code for proteins. As we lack a fundamental understanding of how genetic variation impacts the noncoding genome, it has remained challenging to predict which of these noncoding rare variants might have functional effects in AD. We have developed a pipeline to prioritize these rare noncoding variants using a combination of epigenetics and machine learning approaches. We will use this pipeline to nominate putatively functional rare noncoding variants and then use functional genomics assays to validate these predicted effects.

The AI4AD (Artificial Intelligence for Alzheimer’s Disease) Initiative aims to create and develop advanced AI methods, and apply them to genomic, imaging and cognitive data, in order to 1) identify AD genotypes and endophenotypes that dissect AD’s heterogeneity; 2) relate said genotypes and endophenotypes with clinical progression in pre-dementia patients; 3) identify novel treatment targets for AD by analyzing whole genome and associated phenotypic data. The goals of this multisite initiative (Paul Thompson, USC; Christos Davatzikos, Li Shen, Penn; Andy Saykin, IU; Heng Huang, Pitt, Paul Crane, UW; Adam Brickman, Columbia; Tim Hohman, Vanderbilt; Guyngah Jun, BU; Duygu Tosun, UCSF; Alexander Zaranek, Curii) leverage the promise of machine learning (ML) to contribute to precision diagnostics, prognostication, and targeted and novel treatments. We will develop ML and deep learning methods to apply to large scale biobanks of whole genome sequences (WGS), neuroimaging, cognitive, and clinical data, aiming to discover new genomic features that influence biological processes of AD. We will apply methods of genome representation and tiling to WGS repositories to create inputs for AI methods. We will develop novel, interpretable, biological knowledge guided deep learning methods to discover genomic motifs associated with AD, AD risk, and biological processes of AD as defined by NIA-AA criteria. To quantify subtypes and disentangle biological processes of AD, we will apply computational methods to multimodal MRI and amyloid- and tau-sensitive PET to stratify and subtype patient groups; novel imaging genomics methods will detect genomic markers and pathways that modulate the developing pathology as detected in the images, and that predict future clinical decline or resilience. We hypothesize that advanced deep learning methods combined with whole genome data will outperform traditional methods and GWAS for predicting AD onset and progression, and will assist with disease subtyping and discovering treatable targets in the genome. A team will rank and repurpose existing and identify novel drugs and targets in the genome based on the discovered genetic motifs affecting AD.

The AI4AD (Artificial Intelligence for Alzheimer’s Disease) Initiative aims to create and develop advanced AI methods, and apply them to extensive and rich genomic, imaging and cognitive data, in order to 1) identify genotypes and endophenotypes of AD that dissect the heterogeneity of the disease; 2) relate these genotypes and endophenotypes with clinical progression, in pre-dementia patients; 3) identify novel treatment targets for AD, by analyzing whole genome and associated phenotypic data at a previously impossible scale. Collectively, the goals of this highly collaborative multi-site initiative leverage the promise of machine learning to contribute to precision diagnostics, prognostication, and targeted and novel treatments.

The ADSP-PHC was established to harmonize the rich endophenotype data across cohort studies to enable modern genomic analyses of ADRD with the ultimate goal to generate harmonized data that will become a “legacy” dataset perpetually curated and shared NIAGADS. The ADSP-PHC will facilitate phenotypic data harmonization for ADSP participants with genetic and genomic data. This effort represents a multi-disciplinary approach leveraging interdisciplinary expertise in large-scale genetic and genomic studies, clinical neuroscience, neuroimaging, psychometrics, and bioinformatics. This study will utilize a team structure consisting of two coordinating centers to oversee activities of all harmonization teams and to oversee coordination with other ADSP workgroups and initiatives, a Storage and Informatics Core that will oversee the coordination between LONI and NIAGADS for all data storage, compliance, and dissemination, a CHARGE Coordination Core to oversee the alignment of data and protocols with CHARGE workgroups, an Integration & Analytics Core that will enable data integration across phenotypes to facilitate downstream machine learning applications, and eight Domain-Specific Harmonization Teams tasked with harmonization in their area of expertise. The endophenotypes that will be harmonized by this project include Cognition, Fluid Biomarkers, Amyloid PET, Structural MRI, White Matter Hyperintensities, White Matter Integrity, Vascular Risk Factors, and Neuropathology. Site PIs on this project include: Jesse Mez (Boston University), Adam Brickman (Columbia University), Andy Saykin (Indiana University), Elizabeth Mormino (Stanford University), Pauline Maillard (UC Davis), Duygu Tosun-Turgut (UC San Francisco), Christos Davatzikos (University of Pennsylvania), Arthur Toga (USC); Mohamad Habes (University of Texas Health Science Center at San Antonio), Michael Cuccaro (University of Miami), Paul Crane (University of Washington), Bennett Landman (Vanderbilt University), Timothy Hohman (Vanderbilt University Medical Center), and Carlos Cruchaga (Washington University in St. Louis).

The growing availability of endophenotypic data in cohort studies of Alzheimer’s disease and related dementias (ADRD) provides an exciting opportunity to further characterize the genetic architecture of this devastating disease. However, there is a pressing need to develop and apply advanced harmonization approaches to align ADRD endophenotypes across cohorts. The goal of this coordinated national initiative – the AD Sequencing Project Phenotype Harmonization Consortium, or “ADSP-PHC” – is to provide a centralized database of robust endophenotypes for large-scale genomic analyses that will accelerate the identification of novel targets for therapeutic intervention in ADRD. will accelerate the identification of novel targets for therapeutic intervention in ADRD.

The goal of this study is to identify new genes and mutations that cause or increase risk for Alzheimer disease (AD), as well as protective factors. Individuals and families were selected from the Knight-ADRC (Washington University) and the NIA-LOAD study. Only families with at least three first-degree affected individuals were included. Families with pathogenic variants in the known AD or FTD genes, or in which APOE4 segregated with disease were excluded. At least two cases and one control were selected per family. Cases had an age at onset (AAO) after 65 yo and controls had a larger age at last assessment than the latest AAO within the family. Whole exome (WES) and whole genome sequencing (WGS) was generated for 1,235 individuals (285 families) that together with data from our collaborators and the ADSP family-based cohort (3,449 individuals and 757 families) will provide enough statistical power to identify new genes for AD. Dr. Tanzi (Harvard Medical School) will provide WGS from 400 families from the NIMH Alzheimer disease genetics initiative study. We will perform single variant and gene-based analyses to identify genes and variants that increase risk for disease in AD families. Single variant analysis will consist of a combination of association and segregation analyses. We will run family-based gene-based methods to identify genes that show and overall enrichment of variants in AD cases. We will also look for protective and modifier variants. To do this we will identify families loaded with AD cases, that also include individuals with a high burden of known risk variants but that do not develop the disease (escapees). We will use the sequence data and the family structure to identify variants that segregate with the escapee phenotype. The most promising variants and genes will be replicated in independent datasets (ADSP case-control, ADNI, Knight-ADRC, NIA-LOAD ). We will perform single variant and gene-based analyses to replicate the initial findings, and survival analysis to replicate the protective variants. We will select the most promising variants/genes for functional studies

Family-based approaches led to the identification of disease-causing Alzheimer’s Disease (AD) variants in the genes encoding APP, PSEN1 and PSEN2. The identification of these genes led to the A?-cascade hypothesis and to the development of drugs that target this pathway. Recently, we have identified rare coding variants in TREM2, ABCA7, PLD3 and SORL1 with large effect sizes for risk for AD, confirming that rare coding variants play a role in the etiology of AD. In this proposal, we will identify rare risk and protective alleles using sequence data from families densely affected by AD. We hypothesize that these families are enriched for genetic risk factors. We already have sequence data from 695 families (2,462 individuals), that combined with the ADSP and the NIMH dataset will lead to a dataset of more than 1,042 families (4,684 individuals). Our preliminary results support the flexibility of this approach and strongly suggest that protective and risk variants with large effect size will be found, which will lead to a better understanding of the biology of the disease.

The ADSP-PHC was established to harmonize the rich endophenotype data across cohort studies to enable modern genomic analyses of ADRD with the ultimate goal to generate harmonized data that will become a “legacy” dataset perpetually curated and shared NIAGADS. The ADSP-PHC will facilitate phenotypic data harmonization for ADSP participants with genetic and genomic data. This effort represents a multi-disciplinary approach leveraging interdisciplinary expertise in large-scale genetic and genomic studies, clinical neuroscience, neuroimaging, psychometrics, and bioinformatics. This study will utilize a team structure consisting of two coordinating centers to oversee activities of all harmonization teams and to oversee coordination with other ADSP workgroups and initiatives, a Storage and Informatics Core that will oversee the coordination between LONI and NIAGADS for all data storage, compliance, and dissemination, a CHARGE Coordination Core to oversee the alignment of data and protocols with CHARGE workgroups, an Integration & Analytics Core that will enable data integration across phenotypes to facilitate downstream machine learning applications, and eight Domain-Specific Harmonization Teams tasked with harmonization in their area of expertise. The endophenotypes that will be harmonized by this project include Cognition, Fluid Biomarkers, Amyloid PET, Structural MRI, White Matter Hyperintensities, White Matter Integrity, Vascular Risk Factors, and Neuropathology. Site PIs on this project include: Jesse Mez (Boston University), Adam Brickman (Columbia University), Andy Saykin (Indiana University), Elizabeth Mormino (Stanford University), Pauline Maillard (UC Davis), Duygu Tosun-Turgut (UC San Francisco), Christos Davatzikos (University of Pennsylvania), Arthur Toga (USC); Mohamad Habes (University of Texas Health Science Center at San Antonio), Michael Cuccaro (University of Miami), Paul Crane (University of Washington), Bennett Landman (Vanderbilt University), Timothy Hohman (Vanderbilt University Medical Center), and Carlos Cruchaga (Washington University in St. Louis).

The growing availability of endophenotypic data in cohort studies of Alzheimer’s disease and related dementias (ADRD) provides an exciting opportunity to further characterize the genetic architecture of this devastating disease. However, there is a pressing need to develop and apply advanced harmonization approaches to align ADRD endophenotypes across cohorts. The goal of this coordinated national initiative – the AD Sequencing Project Phenotype Harmonization Consortium, or “ADSP-PHC” – is to provide a centralized database of robust endophenotypes for large-scale genomic analyses that will accelerate the identification of novel targets for therapeutic intervention in ADRD. will accelerate the identification of novel targets for therapeutic intervention in ADRD.

The ADSP-PHC was established to harmonize the rich endophenotype data across cohort studies to enable modern genomic analyses of ADRD with the ultimate goal to generate harmonized data that will become a “legacy” dataset perpetually curated and shared NIAGADS. The ADSP-PHC will facilitate phenotypic data harmonization for ADSP participants with genetic and genomic data. This effort represents a multi-disciplinary approach leveraging interdisciplinary expertise in large-scale genetic and genomic studies, clinical neuroscience, neuroimaging, psychometrics, and bioinformatics. This study will utilize a team structure consisting of two coordinating centers to oversee activities of all harmonization teams and to oversee coordination with other ADSP workgroups and initiatives, a Storage and Informatics Core that will oversee the coordination between LONI and NIAGADS for all data storage, compliance, and dissemination, a CHARGE Coordination Core to oversee the alignment of data and protocols with CHARGE workgroups, an Integration & Analytics Core that will enable data integration across phenotypes to facilitate downstream machine learning applications, and eight Domain-Specific Harmonization Teams tasked with harmonization in their area of expertise. The endophenotypes that will be harmonized by this project include Cognition, Fluid Biomarkers, Amyloid PET, Structural MRI, White Matter Hyperintensities, White Matter Integrity, Vascular Risk Factors, and Neuropathology.Site PIs on this project include: Jesse Mez (Boston University), Adam Brickman (Columbia University), Andy Saykin (Indiana University), Elizabeth Mormino (Stanford University), Pauline Maillard (UC Davis), Duygu Tosun-Turgut (UC San Francisco), Christos Davatzikos (University of Pennsylvania), Arthur Toga (USC); Mohamad Habes (University of Texas Health Science Center at San Antonio), Michael Cuccaro (University of Miami), Paul Crane (University of Washington), Bennett Landman (Vanderbilt University), Timothy Hohman (Vanderbilt University Medical Center), and Carlos Cruchaga (Washington University in St. Louis).

The growing availability of endophenotypic data in cohort studies of Alzheimer’s disease and related dementias (ADRD) provides an exciting opportunity to further characterize the genetic architecture of this devastating disease. However, there is a pressing need to develop and apply advanced harmonization approaches to align ADRD endophenotypes across cohorts. The goal of this coordinated national initiative – the AD Sequencing Project Phenotype Harmonization Consortium, or “ADSP-PHC” – is to provide a centralized database of robust endophenotypes for large-scale genomic analyses that will accelerate the identification of novel targets for therapeutic intervention in ADRD. will accelerate the identification of novel targets for therapeutic intervention in ADRD.

The objectives are to identify and characterise genes and genetic variants which increase or decrease the risk of developing Alzheimer's disease.Exome sequenced and whole genome sequenced cases and controls will be analysed. The predicted function of DNA variants will be obtained using software such as VEP, PolyPhen, SIFT. Weighted burden analysis will be performed wherein variants are given higher weights if they are predicted to have a major effect on protein function and/or if they are rare. For each gene, in each subject the weights for the DNA variants possessed by that subject will be summed to produce a score. The scores between cases and controls will be compared using logistic regression and incorporating relevant covariates such as sex, age, principal components. If scores are on average higher in cases this indicates that damage to the gene increases risk of Alzheimer's disease. If scores are higher in controls this indicates that damage to the gene reduces risk. Sets of genes will also be analysed in a similar way. The method has been applied to a smaller ADSP dataset: https://www.biorxiv.org/content/10.1101/596007v1

We will analyse whether variants in DNA can interfere with the functioning of particular genes and either increase or decrease the risk of developing Alzheimer's disease. We will examine all the variants in a gene observed in large samples of people with and without Alzheimer's disease to see if variants are more commonly seen in one or other group. We will weight the variants to that more attention is paid to those which are rare and those which are predicted to have a major effect on the functioning of the gene. If we see more variants in the people with Alzheimer's disease then this suggests that damaging that gene could increase risk of illness. If the people without disease have more variants in a gene then that could suggest that damaging that gene would actually protect against Alzheimer's disease. Understanding these effects will ultimately assist in the development of methods to treat or prevent the disease.

The objectives are to identify and characterise genes and genetic variants which increase or decrease the risk of developing Alzheimer's disease. Exome sequenced and whole genome sequenced cases and controls will be analysed. The predicted function of DNA variants will be obtained using software such as VEP, PolyPhen, SIFT. Weighted burden analysis will be performed wherein variants are given higher weights if they are predicted to have a major effect on protein function and/or if they are rare. For each gene, in each subject the weights for the DNA variants possessed by that subject will be summed to produce a score. The scores between cases and controls will be compared using logistic regression and incorporating relevant covariates such as sex, age, principal components. If scores are on average higher in cases this indicates that damage to the gene increases risk of Alzheimer's disease. If scores are higher in controls this indicates that damage to the gene reduces risk. Sets of genes will also be analysed in a similar way. The method has been applied to a smaller ADSP dataset:Curtis D, Bakaya K, Sharma L, Bandyopadhyay S. Weighted burden analysis of exome-sequenced late onset Alzheimer's cases and controls provides further evidence for a role for PSEN1 and suggests involvement of the PI3K/Akt/GSK-3β and WNT signalling pathways. Ann Hum Genet 2020 https://doi.org/10.1111/ahg.12375

We will analyse whether variants in DNA can interfere with the functioning of particular genes and either increase or decrease the risk of developing Alzheimer's disease. We will examine all the variants in a gene observed in large samples of people with and without Alzheimer's disease to see if variants are more commonly seen in one or other group. We will weight the variants to that more attention is paid to those which are rare and those which are predicted to have a major effect on the functioning of the gene. If we see more variants in the people with Alzheimer's disease then this suggests that damaging that gene could increase risk of illness. If the people without disease have more variants in a gene then that could suggest that damaging that gene would actually protect against Alzheimer's disease. Understanding these effects will ultimately assist in the development of methods to treat or prevent the disease.

Objectives of the proposed research: Alzheimer's disease (AD) is a progressive neurological disorder that affects millions globally. Despite significant strides in developing treatments aimed at specific neuropathological markers, a substantial subset of individuals retains normal cognitive function even in the presence of AD-related neuropathology. This phenomenon suggests the existence of "cognitive resilience" factors that help preserve cognitive function independently of the disease's typical pathological processes. The primary objective of our proposed research is to conduct a comprehensive secondary analysis of the extensive genotype and phenotype data curated within the National Institute on Aging Genetics of NIAGADS. We aim to identify genetic and molecular signatures of individuals with high resilience to AD. Study design: To achieve our objectives, we will develop a novel computational framework that integrates polygenic risk scores (PRS), incorporating additional clinical variables into clinical risk scores. This will allow for a more precise measurement of AD risk at the individual level within the ADSP and provide a comprehensive assessment of AD risk and resilience factors. Analysis plan: Following our previous GRPa-PRS framework [1], we will identify participants within the top quantile of AD risk who, despite their elevated risk, do not develop AD. These individuals will be classified as highly resilient. Given that the effects of resilience are generally subtle, we will aggregate genetically regulated effects at the pathway level, enabling us to characterize resilience at the individual level. Following this, we will conduct a comparative analysis of the identified resilient vs AD individuals to further understand the underlying genetic and molecular basis of their resilience. To further validate our findings, we will link these results with additional molecular evidence from the ROS/MAP cohort. 1. GRPa-PRS: A risk straticifcation method to identify genetically regulated pathways in polygenic diseases. medRxiv. 2. Polygenic resilience scores capture protective genetic effects for Alzheimer’s disease. Translational psychiatry, 12(1), 296.

Alzheimer's disease (AD) is a progressive neurological disorder that poses substantial societal and economic challenges, affecting over 55 million people worldwide. Significant research efforts and therapeutic developments have focused on addressing AD's neuropathological features, such as beta-amyloid plaques and neurofibrillary tangles. Despite these efforts, clinical-pathological studies have revealed that a notable proportion of individuals maintain normal cognitive function even in the presence of these neuropathological markers. Through our GRPa-PRS framework, we aim to stratify individuals with high resilience and identify factors associated with cognitive resilience—the ability to preserve cognitive function despite a high genetic risk for developing AD. The success of this project will help to uncover these resilience factors could pave the way for innovative therapeutic strategies that enhance resistance to AD.

The goal of this project is to identify and prioritize genetic variations that contribute to late-onset Alzheimer’s disease (AD). The planned multi-modal meta-analysis study will integrate genetic information from multiple AD datasets (ADSP, EADB GWAS, ADGC, FinGen, UK Biobank, CHARGE consortia) with other publicly available multi-omics data including gene expression (GTEx), epigenetics (GEO) and neuro-imaging. The objective, here, is to extend and refine polygenic risk scores (PRS) to work across ethnicities and prioritize genes and pathways that are causal in nature to development of late-onset AD in different ancestry cohorts. Retrospective, case-control analysis setup against the backdrop of other neurodegenerative diseases will also help identify lead variants that are specific to AD enabling the elucidation of mechanistic pathways that lead to the disease etiology. A well-known limitation of PRS scores derived from single ethnicity datasets is that its predictive power declines in cross-ethnicity cohorts. To address this, we want to investigate the possibility of creating orthogonal effect scores (OES) from multimodal analysis for gene prioritization and causality determination. The OES will be used to further refine PRS in different ethnicities and possibly identify common emerging pathways. Linkage disequilibrium (LD), quantitative trait loci (QTL) analysis, tissue colocalization, pathway networks and siRNA screening data along with functional implications of mutations, evolutionarily conservation of genomic regions and overlap with non-coding regulatory sites (eg. DNAse I hypersensitivity sites) will be used in the generation of OES. This orthogonal scoring system will be tested for its ability to refine PRS for achieving higher AD risk prediction in known AD cases. The contrived OES will enable creation of in-silico admixed populations and association of AD risk in these populations. The ability to predict AD risk correctly in admixed populations will be immensely useful in associating PRS signatures to AD risk in real patients.

The genetic landscape of late onset AD is complex and its pathophysiology has been elusive in spite of advances in high-throughput genomic techniques. The main limitations in characterizing the disease are, presence of multiple pathways resulting in the disease pathophysiology, lack of large data cohorts that encompass multiple ethnicities leading to low predictive power for AD risk across studies, and difficulty in identifying the causal genomic changes from the many AD-associated loci.This study aims at addressing the latter two limitations by integrating AD data containing mutations, expression, modifications and changes in AD-associated gene loci to guide their prioritization in prediction of AD risk. This prioritization method has the potential to lead us to causative genomic changes that can be traced to key AD related biological pathways and provide the elusive pathophysiological signature for AD across ethnicities.

The ADSP-PHC was established to harmonize the rich endophenotype data across cohort studies to enable modern genomic analyses of ADRD with the ultimate goal to generate harmonized data that will become a “legacy” dataset perpetually curated and shared NIAGADS. The ADSP-PHC will facilitate phenotypic data harmonization for ADSP participants with genetic and genomic data. This effort represents a multi-disciplinary approach leveraging interdisciplinary expertise in large-scale genetic and genomic studies, clinical neuroscience, neuroimaging, psychometrics, and bioinformatics. This study will utilize a team structure consisting of two coordinating centers to oversee activities of all harmonization teams and to oversee coordination with other ADSP workgroups and initiatives, a Storage and Informatics Core that will oversee the coordination between LONI and NIAGADS for all data storage, compliance, and dissemination, a CHARGE Coordination Core to oversee the alignment of data and protocols with CHARGE workgroups, an Integration & Analytics Core that will enable data integration across phenotypes to facilitate downstream machine learning applications, and eight Domain-Specific Harmonization Teams tasked with harmonization in their area of expertise. The endophenotypes that will be harmonized by this project include Cognition, Fluid Biomarkers, Amyloid PET, Structural MRI, White Matter Hyperintensities, White Matter Integrity, Vascular Risk Factors, and Neuropathology. Site PIs on this project include: Jesse Mez (Boston University), Adam Brickman (Columbia University), Andy Saykin (Indiana University), Elizabeth Mormino (Stanford University), Pauline Maillard (UC Davis), Duygu Tosun-Turgut (UC San Francisco), Christos Davatzikos (University of Pennsylvania), Arthur Toga (USC); Mohamad Habes (University of Texas Health Science Center at San Antonio), Michael Cuccaro (University of Miami), Paul Crane (University of Washington), Bennett Landman (Vanderbilt University), Timothy Hohman (Vanderbilt University Medical Center), and Carlos Cruchaga (Washington University in St. Louis).

The growing availability of endophenotypic data in cohort studies of Alzheimer’s disease and related dementias (ADRD) provides an exciting opportunity to further characterize the genetic architecture of this devastating disease. However, there is a pressing need to develop and apply advanced harmonization approaches to align ADRD endophenotypes across cohorts. The goal of this coordinated national initiative – the AD Sequencing Project Phenotype Harmonization Consortium, or “ADSP-PHC” – is to provide a centralized database of robust endophenotypes for large-scale genomic analyses that will accelerate the identification of novel targets for therapeutic intervention in ADRD. will accelerate the identification of novel targets for therapeutic intervention in ADRD.

Objective: Utilize ADSP data sets to identify genes & specific genetic variants that confer risk for or protection from Alzheimer disease. Aim 1: Using combined WGS/WES across the ADSP Discovery, Disc-Ext, and FUS Phases, including single nucleotide variants, small insertion/deletions, and structural variants. We will: Aim 1a. Perform whole genome single variant and rare variant case/control association analyses of AD using ADSP and other available data; Aim 1b. Target protective variant identification via association analysis using selected controls within the ADSP data and performing meta analysis across association results based on selected controls from non-ADSP data sets. Aim 1c. Perform endophenotype analyses including cognitive function measures, hippocampal volume and circulation beta-amyloid ADSP data in subjects for which these measures are available. Meta analysis will be conducted across ADSP and non-ADSP analysis results. Aim 2: To leverage ethnically-diverse and admixed populations to identify AD variants we will: Aim 2a. Estimate and account for global and local ancestry in all analyses; Aim 2b. Perform admixture mapping in samples of admixed ancestry; and Aim 2c. Perform ethnicity-specific and trans-ethnic meta-analyses. Aim 3: To identify putative therapeutic targets through functional characterization of genes and networks via bioinformatics, integrative ‘omics analyses. We will: Aim 3a. Annotate variants with their functional consequences using bioinformatic tools and publicly available “omics” data. Aim 3b. Prioritize results, group variants with shared function, and identify key genes functionally related to AD via weighted association analyses and network approaches. Analyses will be performed in coordination with the following PIs. Philip De Jager, Columbia University; Eric Boerwinkle & Myriam Fornage, U of Texas Health Science Center, Houston; Sudha Seshadri, U of Texas, San Antonio; Ellen Wijsman, U of Washington. Richard Gibbs, Baylor College of Medicine.

This proposal seeks to analyze existing genetic sequencing data generated as part of the Alzheimer’s Disease Sequencing Project (ADSP) including the ADSP Follow-up Study (FUS) with the goal of identifying genes and specific changes within those genes that either confer risk for Alzheimer’s Disease or provide protection from Alzheimer’s Disease. Analytic challenges include analysis of whole genome sequencing data, appropriately accounting for population structure across European ancestry, Hispanic, and African American participants, and interpreting results in the context of other genomic data available.

The objective of this study is to develop machine learning models using genetic and phenotype data from the NIAGADS database https://dss.niagads.org/. We will develop both unsupervised and supervised learning models to characterize the heterogeneity and risk of Alzheimer’s disease (AD). This is an MPI study in collaboration with Dr. Honghuang Lin at University of Massachusetts Chan Medical School.For the first aim, we will build an expandable multimodal unsupervised machine learning framework to investigate AD heterogeneity. We will perform AD subtyping by harnessing the rich multimodality information across a wide spectrum of data (e.g., genetics, images and blood biomarkers). A Bayesian kernel network will be built to estimate the relative weight of each individual data modality, which would also allow the addition of new data modalities as they become available. The analyses will be performed both within and between ethnic populations.For the second aim, we will build an expandable multimodal supervised machine learning framework to quantify AD risk from longitudinal follow-up of clinically normal elders. We will build a separate deep learning network for each data modality in consideration of its unique feature sets. A multiplicative strategy will then be taken to aggregate information from different modalities with weighted contributions. Feature selection will also be performed to identify the most informative features predictive of AD risk.For the third aim, we will build AD-related gene regulatory networks in post-mortem human brain samples. We will examine the association of multi-omics data with AD, which will be used to assign gene priority based on the combinatorial evidence from each type of omics data. A gene ontology-guided greedy search strategy will then be implemented to build gene regulatory networks, and identify key drivers that might be potential therapeutic targets for AD. The analyses will be stratified by ethnic populations and AD phenotypic clusters.

Alzheimer's disease (AD) is the most common form of dementia characterized by progressive loss of cognitive function. There are very limited treatment options for AD. For the current application, we seek to develop multimodal machine learning models by leveraging the rich collection of AD-related omics data and phenotypical data recently generated from the Alzheimer's Disease Sequencing Project (ADSP). Three aims will be pursued in the current application. For Aim 1, we will build an expandable multimodal unsupervised machine learning framework to investigate AD heterogeneity. For Aim 2, we will build an expandable multimodal supervised machine learning framework to quantify AD risk from longitudinal follow up of cognitively normal elders. For Aim 3, we will build AD-related gene interaction networks in post-mortem human brain samples. The present application represents an innovative approach to identify individuals at high risk of AD. The outlined strategy will provide new insights into the risk stratification and prevention strategies for AD.

A majority of the human genome has been well characterized through the initial Human Genome Project and numerous large-scale sequencing studies such as the 1000 Genomes Project, Alzheimer's Disease Neuroimaging Initiative (ADNI), Alzheimer’s Disease Sequencing Project, and others. There are, however, many genome regions that are challenging to characterize using standard approaches that are important to human health and disease. We intend to (1) develop and test new methods to characterize mutations in these regions, and (2) test associations between these mutations and disease phenotypes. Data from the ADSP may be combined with other datasets, such as the Alzheimer's Disease Neuroimaging Initiative. All appropriate precautions will be taken to verify proper population stratification and eliminate any sample redundancy. Combining these data will not increase risk to participants, as all individual-level data will remain confidential. We may also use portions of the ADSP data as controls for other diseases such as amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD), though only in situations that do not violate genetic or data-use principles. Specifically, data that where participants consented for use only within Alzheimer's disease studies will not be used for any purpose outside Alzheimer's disease research.

Many regions of the human genome present challenges that prohibit scientists from discovering potential disease-causing mutations. We are developing methods to characterize mutations in these regions to identify new genes involved in disease.

A majority of the human genome has been well characterized through the initial Human Genome Project and numerous large-scale sequencing studies such as the 1000 Genomes Project, Alzheimer's Disease Neuroimaging Initiative (ADNI), Alzheimer’s Disease Sequencing Project, and others. There are, however, many genome regions that are challenging to characterize using standard approaches that are important to human health and disease. We intend to (1) develop and test new methods to characterize mutations in these regions, (2) test associations between these mutations and disease phenotypes, and (3) re-align the raw ADSP data to the CHM13 reference genome.Data from the ADSP may be combined with other datasets, such as the Alzheimer's Disease Neuroimaging Initiative. All appropriate precautions will be taken to verify proper population stratification and eliminate any sample redundancy. Combining these data will not increase risk to participants, as all individual-level data will remain confidential. We may also use portions of the ADSP data as controls for other diseases such as amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD), though only in situations that do not violate genetic or data-use principles. Specifically, data that where participants consented for use only within Alzheimer's disease studies will not be used for any purpose outside Alzheimer's disease research.

Many regions of the human genome present challenges that prohibit scientists from discovering potential disease-causing mutations. We are developing methods to characterize mutations in these regions to identify new genes involved in disease.

Common variant polygenic scores (PGSs) have been a central approach to predicting genetic risk of AD, however, we know that rare (minor allele frequency [MAF] less than 0.5%) and low frequency (MAF 0.5% to 5%) variants can account for the missing heritability in AD. The objectives of this project are to develop a PGS that incorporates variants across the full allele frequency spectrum and to evaluate how well it predicts AD compared to a common-variant PGS. To accomplish these objectives, we will leverage sequencing data from the ADSP (our study has contributed over 1,000 samples to the Follow Up phase). To generate the common variant PGS, we will calculate a weighted sum of 39 variants previously associated with AD and use their effect sizes from a large meta-analysis as the weights. To generate the PGS comprised of all frequencies, we will add rare and low frequency variants associated with AD, weighted by their effect sizes, to the common variant polygenic score. Prediction of AD case-control status for both the common variant and full frequency PGS will be characterized with an empirical receiver operating characteristic (ROC) curve.In this renewal, we are also requesting the NG00113 - Metabolic and Lipidomics signatures in Alzheimer disease brains dataset. We have metabolic and proteomic data (that we are working to deposit into a repository). We would like to use the data in the NG00113 dataset to replicate the findings in our study.

Currently, no combined measure of genetic risk for Alzheimer’s disease (AD) includes genetic variants that are less common in the population in addition to the more common ones. However, we know that the less common variants can account for the missing heritability in AD. The goals of this project are to develop a genetic risk score that incorporates variants across the full allele frequency spectrum and to evaluate how well it predicts AD compared to a common-variant only genetic risk score. To accomplish these goals, we will leverage genetic data from the ADSP. We will characterize the prediction of AD case-control status for both the common-variant only and full frequency genetic risk score with an empirical receiver operating characteristic (ROC) curve, which summarizes the sensitivity in relationship to the specificity of a genetic risk score at multiple thresholds that separate AD cases and controls.We are now requesting access to metabolomic data to replicate findings from our study of the metabolomics of the pre-clinical and clinical phases of AD.

This U19 aims to bridge these knowledge gaps for discovery and validation of Centrally-linked Longitudinal pEripheral biomARkers of AD (CLEAR-AD) in multi-ethnic populations. CLEAR-AD U19 is based on the premise that AD is a complex disorder in which many biological pathways are disrupted due to multi-omic perturbations, which can be detected in brain and reflected in blood. The specific aims of CLEAR-AD are: 1) To discover CLPMS of the complex and heterogeneous AD pathophysiology and its co-pathologies. 2) To identify longitudinal CLPMS that detect and predict dynamic neuroimaging, fluid biomarker, and clinical changes across AD spectrum. 3) To characterize differences and similarities in CLPMS profiles across NHW, African American (AA) and Latino American (LA) participants to uncover biomarker patterns in multi-ethnic groups. 4) To make these vast resources available to the scientific community to amplify and accelerate its impact. In this U19, we will leverage NIH-funded ADNI, MCSA and ADRC cohorts of >3,700 multi-ethnic participants to generate >20,000 multi-omics measures (Omics Core) that will be processed and integrated with >48,000 harmonized AD cognitive, neuroimaging and fluid endophenotypes (Analytic Core). Using these data, we will identify brain region and cell-type specific CLPMS, which reflect biological subtypes of AD and disease stage (Project 1). We will discover longitudinal changes in CLPMS that predict cognitive and A/T/N/V progression (Project 2). We will define longitudinal cognitive and A/T/N/V changes and CLPMS in URP that are either conserved with NHW or population-specific (Project 3). This U19 will a) Identify the next generation of AD biomarkers with mechanistic insights; b) Establish a precision medicine approach for rigorous multi-omics biomarker discovery and validation in AD; c) Discover molecules that can serve as biomarkers and therapeutic targets; d) Enhance biomarker research in trial-ready multi-ethnic populations; and e) Generate and share a vast and harmonized resource of endophenotype and multi-omics data in NIH-funded cohorts.

There is a clear and immediate need for the discovery of peripheral molecular signatures linked to central disease processes, core and co-pathologies in Alzheimer’s Disease (AD), that will serve as precision medicine blood-based biomarkers for diagnostic, prognostic, theragnostic and therapeutic purposes. AD is a complex disorder in which many biological pathways are disrupted due to multi-omic perturbations, which can be detected in brain and reflected in blood, i.e. centrally-linked peripheral molecular signatures (CLPMS). This U19 will leverage deeply phenotyped, longitudinal NIH-funded multi-ethnic cohorts and cross-disciplinary expertise for multi-omics data generation and its integration with harmonized AD endophenotypes, will share these data and utilize them in integrated U19 projects to discover CLPMS that will serve as the next generation of AD biomarkers.

We aim to develop robust, calibrated, and reliable predictive models by utilizing machine learning techniques on a rich dataset of imaging, omics, and phenotypic data. Our approach is twofold:1. We will employ supervised machine learning methods to construct predictive models for defined outcomes such as clinical disease stages, meaningful cognitive decline, and imaging and neuropathologic markers (e.g., amyloid or tau positivity). We acknowledge that a major challenge in implementing predictive models is their integration into real-world settings. Therefore, once validated, we will convert these models into web applications or simplify them into risk scores to improve their usability and practical application. 2. Given the significant heterogeneity in Alzheimer's disease and existing gaps in our mechanistic understanding, we will utilize semi-supervised machine learning to subtype the disease. This method will incorporate genetics, biofluid biomarkers, imaging, and neuropsychological tests. Our methodology offers a significant advantage over the unsupervised machine and deep learning techniques commonly used in the field. By incorporating "partial labeling" and incorporating knowledge of disease states, our semi-supervised approach enables more clinically relevant disease subtyping and results.For this project, we are collaborating with experts from other institutions, including Dr. Christos Davatzikos from the University of Pennsylvania, who is an expert in neuroimaging and machine learning, and Dr. Richard Lipton from Albert Einstein College of Medicine, known for his expertise in epidemiology and the design of trials and cohorts.

Despite the extensive research on Alzheimer's and other dementias, the success rate of randomized clinical trials has been disappointingly low. This is primarily due to the significant clinical variations among patients, which limits our ability to predict their outcomes accurately. To address this challenge, we aim to develop a predictive framework specifically focusing on Alzheimer's and related dementias. We will use several cohorts and will employ machine learning and advanced statistical modeling on various types of data, including demographics, neuropsychological measures, biofluid and imaging biomarkers from existing international cohorts and concluded clinical trials. The framework will enable us to predict disease progression which will facilitate early intervention and personalized care for individuals at risk of developing dementia. Our project seeks to harness the power of machine learning and advanced analytics to improve our understanding of Alzheimer's Disease, enhance prediction accuracy, and ultimately contribute to more effective treatments and interventions.

Novel treatments for Alzheimer’s Disease (AD) are urgently needed. Observational data indicate that hypertension and hypercholesterolemia are associated with increased risk of both AD and cognitive status (CS) in non-demented persons. Because hypertension and hypercholesterolemia can be treated effectively, confirmation of causal links between them and AD/CS would provide an appealing therapeutic opportunity. Because mutations are randomly distributed during meiosis, mutation-disease associations are immune to confounding by postnatal exposures. In this setting, mutations strongly associated with an exposure of interest constitute ideal instrumental variables to evaluate the causal effect of that exposure on an outcome of interest. This is an appealing strategy for hypertension/hypercholesterolemia (exposures of interest) and AD/CS (outcomes of interest) because genetic variation explains a substantial proportion of the variance of these two vascular risk factors. We will combine novel methods in statistical genetics and well-established instrumental variable techniques to test the overarching hypothesis that genetically-determined hypertension and hypercholesterolemia influence risk of both late-onset AD and CS in non-demented persons. Our proposal leverages our team’s expertise and successful track record of impactful contributions in the fields of Aging; the robust research infrastructure available through Yale’s OAIC; and access, through the NIAGAD Data Storage Site and UK Biobank, to clinical and genomic data from 550,990 persons to pursue the following aims: determine whether genetically-determined hypertension and hypercholesterolemia are associated, individually or jointly, with increased risk of late-onset AD; and determine whether genetically-determined hypertension and hypercholesterolemia are associated with CS in community-dwelling individuals not yet diagnosed with dementia. This administrative supplement to Yale’s OAIC will deploy an innovative strategy for causal inference based on genetic information to clarify whether observed associations between hypertension/ hypercholesterolemia and AD/CS reflect true causal relationships.

Novel treatments for Alzheimer’s Disease are urgently needed. Observational data indicate that hypertension and hypercholesterolemia are associated with increased risk of both late-onset Alzheimer’s Disease and cognitive decline in non-demented persons; however, it is not clear whether these relationships are causative or associative. We will combine novel methods in statistical genetics and well-established instrumental variable techniques to test the overarching hypothesis that genetically-determined hypertension and hypercholesterolemia influence risk of both late-onset Alzheimer’s Disease and cognitive decline in nondemented persons.

This is another update (11/12/2024), adding additional subjects.This is an update to a currently-active project through dbGaP in order to garner GRCh38 mapped reads and additional WES subjects (a total of 19,922) for downstream analysis of regions identified through the scan statistic, endophenotype and causal approaches. The original edited RUS is copied below: We aim to better isolate causal variants within putative ADRD disease genes via two primary approaches. First, we will use an empirical Bayes scan statistic to detect regions of disease variant enrichment. In addition, we will employ novel causal inference methodology to estimate variant-specific causal risk for ADRD. These complementary approaches will allow for discovery of novel ADRD genes as well as enumeration/localization of important variants within putative AD risk genes. We will also employ more conventional approaches (e.g., SKAT, endophenotype development) as appropriate. We have read and approved the Data Use Agreement as signed and submitted on dbGaP and plan to upload results of our findings in a timely manner.

The main goals of the Alzheimer’s Disease Sequence Project (ADSP) include the identification of novel genomic variants contributing to risk of Late-Onset Alzheimer’s Disease or to protection against Alzheimer’s Disease (AD), as well as providing information as to why at-risk individuals may not develop AD or related dementias, especially in multi-ethnic populations. The aim of our data analysis aligns with these goals to identify novel genomic variants associated with AD. We will aim to do so via a scan-based statistic at each variant, where the statistic is specially designed for the analysis of genomic data. We will also explore alternative methodologies for these discoveries including the calculation of a causal estimate of variants within putative AD genes.

As part of the Collaborative for Alzheimer's Disease genetics REsearch (CADRE: NIA grant U01-AG058654), we plan to analyze whole exome and whole genome sequence data generated from subjects with Alzheimer's disease (AD) and elderly normal controls. These data will be generated by the National Human Genome Institute Large-Scale Sequence Program. The goal of the planned analyses is to identify genes that have alleles that protect against or increase susceptibility to AD. We will evaluate variants detected in the sequence data for association with AD to identify protective and susceptibility alleles using the whole exome case-control data. We will also evaluate sequence data from multiplex AD families to identify variants associated with AD risk and protection, and evaluate variant co-segregation with AD. The family data will be whole genome data. The family-based data will be used to inform the cases control analysis and visa versa. We also will focus on structural variants (insertion-deletions, copy number variants, and chromosomal rearrangements). Evaluation of structural variants will involve both whole genome and whole exome data. Structural variants will be analyzed with single nucelotide variants detected and analyzed in the case-control and family-based data.

We are attempting to identify all the inherited elements that contribute to Alzheimer's disease risk. To do this we will analyze DNA sequence data from subjects with Alzheimer's disease and elderly subjects who are cognitively normal. The sequence data from these 2 groups will be compared to identify differences that contribute to the risk of developing Alzheimer's disease of that protect against Alzheimer's disease. These DNA differences can be at a single site in the genetic code, or can span multiple sites, changing the copy number of DNA sequences. Both types of genetic variants will be examined.

Objective: To gain insights into retrotransposon activation in specific cell types, our first objective is to analyze differential transposable element expression in bulk sequenced microglia from Alzheimer’s disease patient brain tissue versus controls (NG00105). Study design: Reads will be aligned to the GRCh38 human reference genome with STAR using parameters optimized for aligning transposon derived multi-aligning reads. Read counts for transposon and gene loci will be obtained using TEtranscripts. Differential expression of genes and transposons will then be calculated using Deseq2. Analysis plan: Unsupervised machine learning techniques will be applied to cluster transcription counts by variance to make associations between specific retrotransposons and microglial/immune response associated genes.Objective: We have identified multiple candidate non-reference mobile element insertion variants using nanopore long read sequencing of DNA extracted from frontal cortex of patients at Braak 0, III, and V/VI. Our second objective is to utilize the ADSP umbrella whole genome sequencing dataset (NG00067) to determine if our findings are conserved in a larger cohort of patients with Alzheimer’s disease. Study Design: CRAM alignment files aligned to the GRCh38 reference genome from the ADSP discovery (snd10000) and PSP-UCLA (snd10017) WGS data sets will be analyzed with xTea (Chu et al. 2021) to identify the presence of mobile element insertions previously identified via nanopore. Only genomic regions containing insertions of interest will be analyzed. Analysis Plan: Non-reference mobile insertions identified via nanopore will be compared in control, Alzheimer’s disease, and PSP NIAGADS datasets. Insertions meeting the designated criteria will be considered for a replication analysis using cohorts from the ADSP umbrella dataset. We will determine whether these variants can predict the longitudinal clinical rate of disease progression and correlate with other features such as tau PET positivity, CSF tau, and cognitive testing. We will also consider sex, age, and high-risk genotypes.

Objective 1: Almost half of the human genome is composed of transposable elements, or “jumping genes.” Retrotransposons are activated in human Alzheimer’s disease and related “tauopathies,” as well as in Drosophila and mouse models of tauopathy. In the current study, we will analyze retrotransposon activation specifically in microglia, the immune cells of the brain, in the context of tauopathy. In addition, we will determine if retrotransposons activation correlates with expression of neighboring immune response genes. Objective 2: We have previously identified tau-induced retrotranpsoson activation as driver of neurodegeneration. In a preliminary analysis of Alzheimer’s disease patient samples and controls, we have used long-read whole genome DNA sequencing technology to discover non-reference retrotransposon insertions that are unique to Alzheimer’s disease patients. In the current study, we expand these analyses to determine if our findings are conserved in a larger patient cohort, and how these novel insertions relate to disease progression.

Objective: The HLA region was previously identified in an Alzheimer’s GWAS study. The role of the adaptive immune system in Alzheimer’s is not well understood, despite emerging evidence suggesting infectious agents may be contributing to the disease.Study design: Using the data requested from NIAGADS, we will attempt to remap the HLA region to produce a more better defined haplotypes for each sample. We will also look at other variants of interest associated with Alzheimer’s and/or the innate and adaptive immune system, to pursue hypotheses around how these two arms of the immune system may interact. We will also use these data to perform association testing, identifying variants associated with AD risk or infection and evaluating their sensitivity to covariates such as APOE genotype, sex, and ancestry.Analysis plan: Data from NIAGADS with be downloaded to an AWS instance. Regions of interest, such as the HLA region on chromosome 6, will be extracted using samtools. Additional genotype data will be imputed using the Michigan Imputation Server and reference data selected to best match the ancestry(s) represented in the data. Association testing will adjust for population structure and genetic relatedness. Variants of interest will be annotated using resources such as the Variant Effect Predictor and the Genotype-Tissue Expression project to facilitate the interpretation of association results. Pathway analyses may be used to better understand potential relationships between implicated genes and genes previously implicated in AD and related disorders. We will perform genome scans in large data sets representing diverse ancestries. We will use imputed genotype data within association signals to fine-map the location of variants associated with Alzheimer’s disease. Association testing across independent data sets will be used to replicate these signals. We will use variant annotation to describe the potential relationships between implicated variants and gene function, regulation, and pathways. This work will attempt to identify genes involved in the innate and adaptive immune responses in connection with AD.

We will be investigating the possible connections between genes in the immune system and Alzheimer's disease. We will be looking at both the innate and adaptive arms of the immune system. We will also include approaches that consider the potential role of pathogens in contributing to Alzheimer's etiology.

Mitochondrial dysfunction has been hypothesized to be the primary event in Alzheimer's disease (AD) pathology. Extracellular amyloid-ß (Aß) plaques and intracellular neurofibrillary tangles are histopathological hallmarks of the disease. The so called mitochondrial cascade hypothesis proposes that mitochondrial dysfunction is the primary event in AD pathology. Numerous experiments demonstrated that accumulation of Aß in mitochondria begins before the occurrence of extracellular deposition, and this leads to mitochondrial dysfunction with increased oxidative stress, impaired mitodynamics and decrease in ATP production that leads to synaptic dysfunction, apoptosis, and neurodegeneration. Few mitochondrial enzymes are closely linked to the degradation of accumulated intra-mitochondrial Aß, but also to degradation of the mitochondrial targeting sequences that are cleaved off from the imported precursor proteins by the mitochondrial matrix peptidase. If these peptides fail to be cleared from the mitochondrial matrix, they may act as detergent-like, toxic agents, forming pores in the membranes.This project stems from preliminary, original observations from mouse models, yeast KO models, and a screening of a gene encoding a mitochondrial protease in small cohorts of patients with neurodegenerative disorders. In summary, this MitoAD project will be focused on trying to answer to the following question: Are there variants in mitochondrial protease genes associated with increased risk for AD?We have already selected 7 genes, and we will compare the frequency of: 1)Loss of function variants and 2)Missense variants with frequency <1% in the ADRD database vs. GnomAD database. In case of significant differences in this first analysis, we will then possibly subdivide the ADRD cohort in two groups, i.e. AD and FTD, in order to see if the observed mitochondrial impairment is directly linked to AD/ Aß accumulation or is a more general defect, not specifically associated with neurodegeneration.

Mitochondrial dysfunction has been hypothesized to be the primary event in Alzheimer's disease (AD) pathology. Extracellular amyloid-beta (Aß) deposition is the key histopathological hallmark of AD, but Aß accumulation occurs also in mitochondria causing impairment in different mitochondrial pathways. Diverse mitochondrial enzymes have a role in degradation of unfolded proteins including accumulated intramitochondrial Aß. The scope of this project is to better define the role of these mitochondrial proteases on Aß processing and to evaluate how/if impairment in this pathway is linked to AD development.We want to exploit the ADRD database to evaluate if variants in selected genes encoding mitochondrial proteases are indeed associated with increased risk for AD (and, in case, if this risk is specific for AD or is common to other dementia neurodegenerative conditions).

Objective: Utilize ADSP data sets to identify genes & specific genetic variants that confer risk for or protection from Alzheimer disease. Aim 1: Using combined WGS/WES across the ADSP Discovery, Disc-Ext, and FUS Phases, including single nucleotide variants, small insertion/deletions, and structural variants. We will: Aim 1a. Perform whole genome single variant and rare variant case/control association analyses of AD using ADSP and other available data; Aim 1b. Target protective variant identification via association analysis using selected controls within the ADSP data and performing meta analysis across association results based on selected controls from non-ADSP data sets. Aim 1c. Perform endophenotype analyses including cognitive function measures, hippocampal volume and circulation beta-amyloid ADSP data in subjects for which these measures are available. Meta analysis will be conducted across ADSP and non-ADSP analysis results. Aim 2: To leverage ethnically-diverse and admixed populations to identify AD variants we will: Aim 2a. Estimate and account for global and local ancestry in all analyses; Aim 2b. Perform admixture mapping in samples of admixed ancestry; and Aim 2c. Perform ethnicity-specific and trans-ethnic meta-analyses. Aim 3: To identify putative therapeutic targets through functional characterization of genes and networks via bioinformatics, integrative ‘omics analyses. We will: Aim 3a. Annotate variants with their functional consequences using bioinformatic tools and publicly available “omics” data. Aim 3b. Prioritize results, group variants with shared function, and identify key genes functionally related to AD via weighted association analyses and network approaches. Analyses will be performed in coordination with the following PIs. Coordination will involve sharing expertise, analysis plans or analysis results. No individual level data will be shared across institutions. Anita DeStephano, Boston University, Philip De Jager, Columbia University; Eric Boerwinkle & Myriam Fornage, U of Texas Health Science Center, Houston; Sudha Seshadri, U of Texas, San Antonio; Ellen Wijsman, U of Washington.

This proposal seeks to analyze existing genetic sequencing data generated as part of the Alzheimer’s Disease Sequencing Project (ADSP) including the ADSP Follow-up Study (FUS) with the goal of identifying genes and specific changes within those genes that either confer risk for Alzheimer’s Disease or provide protection from Alzheimer’s Disease. Analytic challenges include analysis of whole genome sequencing data, appropriately accounting for population structure across European ancestry, Hispanic, and African American participants, and interpreting results in the context of other genomic data available.

Alzheimer's disease (AD) is the most common form of dementia but has no effective prevention or treatment. Developing a comprehensive picture of the genetic architecture of AD including a network level functional assessment of risk/resilience genes is essential to develop novel therapeutic targets. The overarching goals of this study are to use genetic and genomic approaches to: 1) identify genes and variants that are involved in the development of AD and related disorders; 2) identify functional networks enriched for AD or related disorder risk and protective loci; 3) determine how cellular function and physiology is impacted by these genetic factors in disease-relevant cell types and animal models. This study will use publicly available whole genome/exome sequence data generated by the Alzheimer’s Disease Sequencing Project (ADSP) and genome-wide association study (GWAS) data from the International Genomics of Alzheimer’s Project (IGAP) and others. We will apply a suite of case-control and family approaches to investigate genetic association with dichotomous and continuous disease traits. This study will not only further our understanding of the genetic architecture of AD but also provide key information regarding the molecular mechanisms, setting the stage for novel therapeutic development.

Alzheimer’s disease (AD) is the only disease among the top ten killers in the U.S. without a disease modifying therapy. Genetic studies provide a powerful means to identify genes and pathways that are causally linked to disease etiology. We propose to use genomic and functional approaches to identify genes that alter the risk of AD and investigate how these genes disrupt cellular pathways leading to disease.

We are studying the effects of rare (minor allele frequency < 5%) genetic variants on the risk of developing late-onset Alzheimer’s Disease (AD). We are interested in variants that have a protective effect in subjects who are at an increased genetic risk, or variants that lead to multiple dementias. Our aim is to identify any genetic variants that are present in the “case” group but not the “AD control” groups for both types of variants. The raw data we receive will be annotated to identify SNP locations and frequencies using existing databases such as 1,000 Genomes. We will filter the data based on genetic models such as compounded heterozygosity, recessive and dominant models to identify different types of variants.

Current genetic understanding of Alzheimer’s Disease (AD) does not fully explain its heritability. The APOE4 allele is a well-established risk factor for the development of Alzheimer’s Disease (AD). However, some individuals who carry APOE4 remain cognitively healthy until advanced ages. Additionally, the cause of mixed dementia pathology development in individuals remains largely unexplained. We aim to identify genetic factors associated with these “protected” and mixed pathology phenotypes.

Despite recent advancements, existing genetic risk prediction models (GRPMs) for late-onset Alzheimer's Disease (LOAD) lack sufficient discrimination ability to support clinical applications. Given the lack of treatments that meaningfully affect disease progression once symptoms have manifested and the socioeconomic consequences at stake, there is a serious unmet need for more accurate GRPMs able to assess a patient's LOAD risk in middle age or earlier, before presymptomatic neurodegeneration begins. Additionally, the lack of diversity in traditional GRPMs further exacerbates inequities in health care for non-Europeans. This project builds off of our previous work in which we produced a GRPM for LOAD using single SNPs and epistatic interactions between SNPs in an ensemble model with polygenic risk scores. In this phase we will develop methods for cross-ancestry SNP analysis to select features to be included in a LOAD GRPM that can be generalized to individuals of any ancestry. The ADSP data will be incorporated with data from the Alzheimer's Disease Neuroimaging Initiative, the Alzheimer's Disease Sequencing Project, the UK BioBank and the Framingham Heart Study. We will develop methods for association testing for homogeneous cross-ancestry SNP effects, followed by fine-mapping using diverse subjects to identify genetic features with functional relevance. We will also create and implement methods for association testing for heterogeneous cross-ancestry SNP effects for identification of epistatic interactions across ancestries. Finally we will train a cross-ancestry AD risk model using a cross-validation framework and replicate our findings in an independent cohort. The model will include age, sex, APOE genotype, and ancestry PCs as covariates and we will test gradient boosting, deep learning, super learning, and ensemble modeling methods.

There are no treatments that significantly slow progression of Alzheimer’s Disease (AD) once symptoms manifest, making early intervention crucial to reduce the burden of this disease. A genetic risk prediction model (GRPM) for determining AD risk early in life, would allow early intervention, life planning, and improved patient stratification for clinical trials. Despite advancements, GRPMs for AD lack sufficient discrimination to support such applications. To address this need, Parabon developed a GRPM using machine learning that is able to predict an individual's risk of developing AD at any age that achieves state-of-the-art prediction accuracy. However, like most genetic risk scores, it was built using European subjects and thus has reduced accuracy in non-Europeans. The goal of this project is to enhance our modeling pipeline to identify genetic variants and interactions both across and within ancestral groups, then build a predictive model and test it in an independent replication set.

The ADSP-PHC was established to harmonize the rich endophenotype data across cohort studies to enable modern genomic analyses of ADRD with the ultimate goal to generate harmonized data that will become a “legacy” dataset perpetually curated and shared NIAGADS. The ADSP-PHC will facilitate phenotypic data harmonization for ADSP participants with genetic and genomic data. This effort represents a multi-disciplinary approach leveraging interdisciplinary expertise in large-scale genetic and genomic studies, clinical neuroscience, neuroimaging, psychometrics, and bioinformatics. This study will utilize a team structure consisting of two coordinating centers to oversee activities of all harmonization teams and to oversee coordination with other ADSP workgroups and initiatives, a Storage and Informatics Core that will oversee the coordination between LONI and NIAGADS for all data storage, compliance, and dissemination, a CHARGE Coordination Core to oversee the alignment of data and protocols with CHARGE workgroups, an Integration & Analytics Core that will enable data integration across phenotypes to facilitate downstream machine learning applications, and eight Domain-Specific Harmonization Teams tasked with harmonization in their area of expertise. The endophenotypes that will be harmonized by this project include Cognition, Fluid Biomarkers, Amyloid PET, Structural MRI, White Matter Hyperintensities, White Matter Integrity, Vascular Risk Factors, and Neuropathology. Site PIs on this project include: Jesse Mez (Boston University), Adam Brickman (Columbia University), Andy Saykin (Indiana University), Elizabeth Mormino (Stanford University), Pauline Maillard (UC Davis), Duygu Tosun-Turgut (UC San Francisco), Christos Davatzikos (University of Pennsylvania), Arthur Toga (USC); Mohamad Habes (University of Texas Health Science Center at San Antonio), Michael Cuccaro (University of Miami), Paul Crane (University of Washington), Bennett Landman (Vanderbilt University), Timothy Hohman (Vanderbilt University Medical Center), and Carlos Cruchaga (Washington University in St. Louis).

The growing availability of endophenotypic data in cohort studies of Alzheimer’s disease and related dementias (ADRD) provides an exciting opportunity to further characterize the genetic architecture of this devastating disease. However, there is a pressing need to develop and apply advanced harmonization approaches to align ADRD endophenotypes across cohorts. The goal of this coordinated national initiative – the AD Sequencing Project Phenotype Harmonization Consortium, or “ADSP-PHC” – is to provide a centralized database of robust endophenotypes for large-scale genomic analyses that will accelerate the identification of novel targets for therapeutic intervention in ADRD. will accelerate the identification of novel targets for therapeutic intervention in ADRD.

We plan to analyze SNP array, whole exome and whole genome sequence data generated from subjects with Alzheimer disease and related dementias (ADRD) and elderly normal controls. The goal of the planned analyses are to identify genes and other functional elements that have variations that protect against or increase susceptibility to ADRD. We will evaluate variants detected in the sequence data or imputed in the array data for association with ADRD to identify protective and susceptibility alleles using the SNP array, whole exome, and whole genome data. We will also evaluate similar sequence data from multiplex ADRD families to identify variants associated with ADRD risk and protection, and evaluate variant co-segregation with ADRD. We also will focus on structural variants (e.g. insertion-deletions, copy number variants, and chromosomal rearrangements, etc.) detected using both whole genome and whole exome data. All data will be analyzed separately and in an integrated fashion and will incorporate additional genetic and functional data. Further, we will examine the variability in genetic effects by genetic ancestry.

We are attempting to identify all the inherited elements that contribute to the risk of Alzheimer's disease and related disorders (ADRD). To do this we will analyze DNA sequence data from subjects with ADRD and elderly subjects who are cognitively normal. The sequence data from these two groups will be compared to identify differences that contribute to the risk of developing Alzheimer's disease of that protect against ADRD. These DNA differences can be at a single site in the genetic code, or can span multiple sites, changing the copy number of DNA sequences. Both types of genetic variants will be examined.

While single gene rare variants have been shown to play a significant role in Early-Onset Alzheimer’s Disease (EOAD), their role in Late-Onset (LOAD) has not been emphasised. The gene discovery methodology we have developed at Population Bio allows for unbiased exploration of highly informative genomic variants in any cohort of interest. Our approach is based on ultra-high resolution copy number variant (CNV) analysis. We have invested heavily in such analysis on normal populations. These are used as comparators for cohorts of interest, such as LOAD. In our LOAD work, this analysis generated a list of CNVs which were either absent in the normal populations we studied or else present at significantly higher frequency in the LOAD cohort. Such CNVs are routinely annotated to determine if they overlie known genes and/or regulatory regions. As an example, we have discovered a deletion in 3% of our LOAD cases, which is present in <= 1% of normals. This deletion disrupts a transcription factor binding site in the intron of a gene, which, via GeneHancer, is known to control exon 1 of the gene. The gene in question is novel to LOAD, and is an important metabolic gene, with known biology. It is vital that we validate this finding by analysis of independent LOAD datasets. In addition, we wish to validate other genes discovered in the same manner We have very deep experience of analyzing WGS/WES datasets. Our focus will be to pull out of the available WGS/WES datasets all the variants for the candidate genes of interest. Such variants, including SNVs, indels and CNVs (called using a variety of tools we have experience with) will be analyzed by reference to databases of normal individuals: i.CNVs, by reference to our own internal database but also gnomad (https://gnomad.broadinstitute.org) CNV data and DGV (http://dgv.tcag.ca) ii.SNVs/indels, by reference to gnomad These analyses will allow us to determine whether there exists a mutational burden for our candidate genes of interest in independent LOAD cohorts, and will serve as validation/refutation. The main phenotype of interest will be definitive diagnoses of LOAD, based on neuropathological and clinical cognitive analyses

Most of the common conditions that affect large numbers of the general population have a genetic basis. While progress has been rapid in the field of cancer, the same cannot be said for common, non-cancer, conditions, such as Late-Onset Alzheimer's Disease (LOAD). It is pretty clear now that not all cases of LOAD represent the same disease, in terms of what is the cause. Our approach has been to consider common diseases as collections of rare subgroups, each of which has a specific cause and which, in due course, will have a specific treatment. We have pioneered and implemented a method to rapidly uncover potentially causal genes in common disorders and will use the data generated from this study to strengthen our discoveries, by validating a set of novel candidate genes we have identified in LOAD Our project will allow us to: 1.Define subsets of disease 2.Work with pharmaceutical companies to develop drugs that will specifically target each subset of disease. In some cases, disease progression may be halted by the therapies developed. In some cases, reversal and/or cure may be possible

“Asymptomatic” Alzheimer’s disease (AD) is a phenomenon in which 30% of individuals over age 65 meet criteria for autopsy-confirmed pathological AD (beta-amyloid plaques and tau aggregation) but do not clinically manifest cognitive impairment.1-3 The resilience that underlies asymptomatic AD is marked by both protection from neurodegeneration (brain resilience)4 and preserved cognition (cognitive resilience).Our central hypothesis is that genetic effects allow a subset of individuals to endure extensive AD neuropathology without marked brain atrophy or cognitive impairment. We are uniquely positioned to identify resilience genes by leveraging the Resilience from Alzheimer’s Disease (RAD) database, a local resource in which we have harmonized a validated quantitative phenotype of resilience across 8 large AD cohort studies.Our strong interdisciplinary team represents international leaders in genetics, neuroscience, neuropsychology, neuropathology, and psychometrics who will leverage the infrastructure and rich resources of the AD Genetics Consortium, IGAP, ADSP, and our recently established and harmonized continuous metric of resilience to fulfill the following aims:Aim 1. Identify and replicate common genetic variants that predict cognitive resilience (preserved cognition) and brain resilience (protection from brain atrophy) in the presence of AD pathology. We hypothesize that common genetic variation will explain variance in resilience above and beyond known predictors like education. Replication analyses will leverage age of onset data from IGAP to demonstrate that resilience loci predict a later age of AD onset.Aim 2. Identify and replicate rare and low-frequency genetic variants that predict cognitive and brain resilience. Rare and low-frequency variants with large effects have been identified in AD case/control studies, providing new insight into the genetic architecture of AD.Aim 3: Identify sex-specific genetic drivers of cognitive and brain resilience to AD pathology. Our preliminary results highlight sex differences in the downstream consequences of AD neuropathology, including sex-specific genetic markers of resilience.

As the population ages, late-onset Alzheimer’s disease (AD) is becoming an increasingly important public health issue. Clinical trials targeted a reducing AD progression have demonstrated that patients continue to decline despite therapeutic intervention. Thus, there is a pressing need for new treatments aimed at novel therapeutic targets. A shift in focus from risk to resilience has tremendous potential to have a major public health impact by highlighting mechanisms that naturally counteract the damaging effects of AD neuropathology. The goal of the present project is to characterize genetic factors that protect the brain from the downstream consequences of AD neuropathology. We will identify both rare and common genetic variants using a robust metric of resilience developed and validated by our research team. The identification of such genetic effects will provide novel targets for therapeutic intervention in AD.

The ADSP-PHC was established to harmonize the rich endophenotype data across cohort studies to enable modern genomic analyses of ADRD with the ultimate goal to generate harmonized data that will become a “legacy” dataset perpetually curated and shared NIAGADS. The ADSP-PHC will facilitate phenotypic data harmonization for ADSP participants with genetic and genomic data. This effort represents a multi-disciplinary approach leveraging interdisciplinary expertise in large-scale genetic and genomic studies, clinical neuroscience, neuroimaging, psychometrics, and bioinformatics. This study will utilize a team structure consisting of two coordinating centers to oversee activities of all harmonization teams and to oversee coordination with other ADSP workgroups and initiatives, a Storage and Informatics Core that will oversee the coordination between LONI and NIAGADS for all data storage, compliance, and dissemination, a CHARGE Coordination Core to oversee the alignment of data and protocols with CHARGE workgroups, an Integration & Analytics Core that will enable data integration across phenotypes to facilitate downstream machine learning applications, and eight Domain-Specific Harmonization Teams tasked with harmonization in their area of expertise. The endophenotypes that will be harmonized by this project include Cognition, Fluid Biomarkers, Amyloid PET, Structural MRI, White Matter Hyperintensities, White Matter Integrity, Vascular Risk Factors, and Neuropathology.Site PIs on this project include: Jesse Mez (Boston University), Adam Brickman (Columbia University), Andy Saykin (Indiana University), Elizabeth Mormino (Stanford University), Pauline Maillard (UC Davis), Duygu Tosun-Turgut (UC San Francisco), Christos Davatzikos (University of Pennsylvania), Arthur Toga (USC); Mohamad Habes (University of Texas Health Science Center at San Antonio), Michael Cuccaro (University of Miami), Paul Crane (University of Washington), Bennett Landman (Vanderbilt University), Timothy Hohman (Vanderbilt University Medical Center), and Carlos Cruchaga (Washington University in St. Louis).

The growing availability of endophenotypic data in cohort studies of Alzheimer’s disease and related dementias (ADRD) provides an exciting opportunity to further characterize the genetic architecture of this devastating disease. However, there is a pressing need to develop and apply advanced harmonization approaches to align ADRD endophenotypes across cohorts. The goal of this coordinated national initiative – the AD Sequencing Project Phenotype Harmonization Consortium, or “ADSP-PHC” – is to provide a centralized database of robust endophenotypes for large-scale genomic analyses that will accelerate the identification of novel targets for therapeutic intervention in ADRD.

The purpose of this study is to find new Alzheimer related variants and genes, by combining exome and whole genome data from healthy controls and Alzheimer patients from different studies. Data will be analyzed using association, burden and variant component statistics.

Some individuals develop dementia, while others do not. A large part is likely determined by ones genes, Alzheimer’s disease has a heritability of up to 80%. What are the key genetic factors that determine if one will get Alzheimer disease? In this study, we will thoroughly explore genomic data of a large group of healthy persons and dementia patients to answer this question.

Our goal is to use machine learning (ML) to understand and predict genetic cause and risk to Alzheimer's disease (AD).Our study is purely data informatic in nature and follows common ML-practices, including data cleaning and quality control of both samples and genotypes, pre-processing (such as correcting for covariates like age, sex and general population structure), training and parameter optimization, and evaluation using hold-out sets. Our primary goal is to increase the predictive power for disease status from only the genetic information. We use several different ML-algorithms to build predictors, such as compressed sensing (also known as LASSO), neural networks, and horseshoe Bayesian regression. The underlying genetic architectures are studied through dissecting and analyzing the trained predictors — which loci are important? how important are they (in particular the polygenetic importance beyond APOE)? what are the genetic correlations for these loci within and across different ancestries or other population groups? etc. — informing both fundamental disease research and future predictor algorithm designs.A priori, we will at least use the phenotypic characteristics sex, age/age of onset, race, ethnicity, AD status, and family history. We will also investigate whether a more informative case variable can be constructed as a function of the mentioned variables in conjunction with AD status comments. Other phenotypic characteristics may also be used in the continuous improvement of our predictor algorithms.All analysis will be performed on high-performance computing clusters at Michigan State University (MSU), where they will be stored under strict security, accessible only to PI and three other MSU staff (who also sign the Data Transfer Agreement), in accord with regulations.We will publish all scientific results in peer-reviewed journals and make developed general algorithms public. Published predictors will be made available, both to the public and returned to the NIAGADS.

Alzheimer’s disease (AD) is affecting more than 6 million Americans with an enormous impact on their lives and surrounding families. We aim to improve existing and develop new methods to predict the AD risk of an individual from his/her genetic information. Such a predictor can inform both life and treatment decisions, and its inner workings shed light on the genetic causes of AD. As such, it furthers the basic disease understanding and both the development and employment of preventive medications.Our methods use modern machine learning techniques (e.g. LASSO and neural networks) which are trained on carefully processed trait and genetic data. The amount of data is a crucial factor for success and the more than 30,000 participating individuals in the NIAGADS database constitute a state-of-the-art resource for this type of research.All data is handled under strict security policies while all scientific results will be made publicly available through publications and downloadable files.

The purpose of our study is to identify effective therapeutic targets for Alzheimer’s disease (AD) through stratifying this heterogeneous disorder into subtypes. We will (i) define specific patient segments based on known genetic risks such as ApoE genotype and AD-related phenotypes, and (ii) identify novel genetic factors and understand the biological and pathological mechanism for the specific segment. The specific patient segments will be defined and characterized by the known genetic risks such as ApoE genotype, and AD-related phenotypes, including symptoms, clinical progression, brain imaging and Braak stages. The whole genome sequencing data and the whole exome sequencing data will be analyzed to identify novel genetic variants or genes associated in case-control cohort for each specific segment. These findings will be confirmed with family-based association analyses. We plan to analyze the whole genome sequencing data and the whole exome sequencing data with several phenotype-variant analysis approaches. First, we will stratify and characterize the patients by the known genetic risks and AD-related phenotypes, including symptoms, clinical progression, brain imaging and Braak stages. Then, we will analyze common variants, rare variants, including loss of function mutation, to identify novel genetic variants associated to each specific segment. These findings will be confirmed with family-based association analyses. System biological approaches will be used to determine perturbation of specific genes or pathways related to the phenotypes and to understand the molecular mechanism in each segment by integrating public data such as gene expression data that could be useful to identify the relevant brain region and cell types. All data will remain anonymized and securely stored, and only those listed on our application and their staff will have access to these data. We will not share any of the individual level data outside of Sumitomo Dainippon Pharma nor beyond the researchers on our application. We will adhere to all agreements through the DSS NIAGADS. We have a secure computational environment where we will use these data.

The purpose of our study is to identify effective therapeutic targets for Alzheimer’s disease. Alzheimer’s disease is heterogeneous disease, which result from different combinations of genetic factors as well as environmental factors. Stratifying this heterogeneous disorder into subtypes based on genetic factors and objective phenotypes is important step to discover effective therapeutic targets. The whole genome sequencing data and the whole exome sequencing data will be analyzed to identify novel genetic variants associated with disease and/or subtypes.

Approximately 15-20% of people age 70 or older carry a cancer-associated mutation in a substantial proportion of their blood cells, even though the vast majority do not have cancer. This condition has been termed Clonal Hematopoiesis of Indeterminate Potential (CHIP). Past research has demonstrated that CHIP is associated with an increased risk of both all-cause mortality and several age-associated diseases, such as atherosclerotic cardiovascular disease. Using data from TOPMed, we found a surprising correlation: after controlling for competing risks, such as death, age, sex, and APOE genotype, CHIP is associated with protection from AD and AD-related pathologies, and that the degree of protection is proportional to the size of the mutant clone. We would like to improve our understanding of this phenomenon by leveraging the richly annotated whole genome (WGS) and whole exome sequencing (WES) datasets provided by the NIAGADs under the Alzheimer’s Disease Sequencing Project (ADSP). Understanding of this phenomenon could lead to possible treatments and greater ability to predict who will suffer from AD and its related pathologies. The NIAGADs has already performed the screening, recruitment, consent, and specimen collection. We are requesting access to datasets that include de-identified phenotypic and genetic information from both healthy controls and patients with AD and AD-related pathologies, specifically Dementia and Frontotemporal Dementia. We plan to control for and examine the following phenotypic characteristics: presence or absence of diagnosed neurological disease, age of disease onset (or age at the first visit for control groups), age at specifimen collection, sex, ethnic origin, and measures of cognitive and clinical decline. These phenotypes will be evaluated against genetic information (WES, WGS, single-nucleotide polymorphisms, and polygenic risk scores). Specifically, we plan to use mutation callers in order to identify somatic mutations, such as mutect, varscan, and mocha (from the WES data, etc.).

Approximately 15-20% of people age 70 or older carry a cancer-associated mutation in a substantial proportion of their blood cells, even though the vast majority do not have cancer. This condition has been termed Clonal Hematopoiesis of Indeterminate Potential (CHIP). Past research has demonstrated that CHIP is associated with an increased risk of both all-cause mortality and several age-associated diseases, such as atherosclerotic cardiovascular disease. Using data from TOPMED and Alzheimer’s Disease (AD) Sequencing Project, we found a surprising correlation: CHIP is associated with protection from AD and AD-related pathologies, and that the degree of protection is proportional to the size of the mutant clone. We would like to improve our understanding of this phenomenon by leveraging the richly annotated whole genome and whole exome sequencing datasets provided by the NIAGADs to replicate this association. Understanding of this phenomenon could lead to possible treatments and greater understanding of individual AD risk.

Objectives: Affecting more than 3 million people per year, Alzheimer's Disease is the only leading cause of death without a treatment or cure. Genetics plays a major role in developing Alzheimer's Disease. For example, APOEe4 is the strongest genetic risk factor for sporadic Alzheimer's disease. The APOEe4 allele increases the disease risk by 3 times in heterozygotes and by 15 times in homozygotes. Notably, - 40% of African Americans have at least one APOEe4 allele. Yet, African Americans with APOE e4 do not have an elevated risk of developing Alzheimer's. The underlying protective mechanism in Africans is unknown. The TMEM173 gene encodes a protein called STING that is critical in host defense, anti-tumor immunity, and tissue inflammation. Besides the WT allele, the human TMEM173 gene contains two additional common alleles: R71H­G230A-R293Q (HAQ) and G230A-R293Q (AQ). In East Asians, WT/HAQ (34.3%), not WT/WT (22.0%), is the most common TMEM173 genotype. Intriguingly, the AQ allele is exclusively carried by Africans (~40%). The objective is to explore/establish an association between the common African AQ allele and protection against Alzheimer’s Disease. Study design: Our hypothesis is that the common African-specific TMEM173 AQ allele is associated with a decreased risk of Alzheimer’s disease. Using the Data Portal, https://dss.niagads.org/datasets/, we will select the relevant dataset. For example, one of the datasets we are interested in is (https://dss.niagads.org/sample-sets/snd 10003/) ADGC African American samples (Accession Number:snd 10003) that has 1648 controls and 1290 cases. We will download the sequence file, analyze the TMEM173 alleles in the 1648 controls and 1290 cases data, calculate the odds ratio with p values, and determine if the TMEM173 AQ alleles impact AD incidence in Africans. We will also compare ADSP cohorts with Caucasians, Asians, and Hispanics for the common TMEM173 alleles. Analysis plan: The analyzed cohorts consist of cases and controls. We will compare the allele frequency differences in the case and controls. We will match gender, age, and APOE e4 allele but will not conduct detailed phenotypic characterization.

Genetic factors influence people with Alzheimer’s. For example, just one copy of the APOEe4 allele increases the odds of Alzheimer’s disease by ~3 folds. - 40% of Africans have at least one copy of APOEe4 allele. Yet, Africans with APOEe4 do not have an elevated risk of developing Alzheimer's. The underlying protective mechanism is unknown. Accumulating evidence suggests that the TMEM173 gene promotes neuroinflammation and neurodegeneration, such as Parkinson’s and Alzheimer’s diseases. Intriguingly, while the majority of people have the WT allele of the TMEM173 gene, many have the R71H-G230A-R293Q (HAQ), G230A-R293Q (AQ) alleles. More East Asians are WT/HAQ (34.3% of East Asians) than WT/WT (22.0%). Meanwhile, the AQ allele is Africans-specific. Less than 1% of non-Africans have the AQ allele, while ~40% of Africans carry the AQ allele. One copy of the AQ allele can functionally suppress TMEM173-promoted tissue inflammation in mice. Here, we explore the potential protective role of the AQ allele in Alzheimer’s disease.

Objectives of the proposed research: Alzheimer's disease (AD), the most common type of dementia, is a neurodegenerative disease that generally affects people greater than 45 years old. AD patients show a persistent cognitive decline that leads to total disability at the end stage. Tau protein is one of major proteins linked to AD progression; it accumulates in neurons and forms paired helical filaments. As a result, Tau protein loses its capability to bind with microtubules and leading to neurodegeneration. We have performed Cdc37 chaperone based mass spectrometry to identify novel proteins linked to AD. We found Shroom-3 interaction with Cdc37 completely abolished in AD brain tissues compared to normal human brain tissues. These data provide strong evidence for potential role of Shroom-3 in AD. Currently, there is no genomic data available on Shroom-3 in AD cases. Hence, with this data access we aimed to perform genomic analysis for Shroom-3 and identify any potential mutations (SNPs) in Shroom-3 in AD. After analyzing Shroom-3, we will look at Cdc37 chaperones and other related proteins to fully characterize Shroom-3 and associated proteins.Study design: As a pilot study, we will aim for sample size n=100. Depending on available data and information for analysis, we will group samples as follows: male & female, different ethnicity, and age groups. Depending on data analysis results sample size will be adjusted to higher numbers after completion of pilot study with n=100. We will use bioinformatic software to compare gene sequences from AD patients with normal healthy individual (wild-type gene sequences) to identify any potential mutations/ Single nucleotide polymorphisms (SNPS). Based on results, we will plan cellular and animal model studies for further characterization.Analysis plan, including the phenotypic characteristics that will be evaluated in association with genetic variants: We will carefully look at AD and normal aligned sequences for any changes in a particular nucleotide or set of nucleotides to identify mutations/SNPs in different groups (gender, ethnicity, & age).

Alzheimer's disease (AD), the most common type of dementia, is a neurodegenerative disease that generally affects people greater than 45 years old. AD patients show a persistent cognitive decline that leads to total disability at the end stage. Tau protein is one of major proteins linked to AD progression; it accumulates in neurons and forms paired helical filaments. As a result, Tau protein loses its capability to bind with microtubules and leading to neurodegeneration. We have performed Cdc37 chaperone based mass spectrometry to identify novel proteins linked to AD. We found Shroom-3 interaction with Cdc37 completely abolished in AD brain tissues compared to normal human brain tissues. These data provide strong evidence for potential role of Shroom-3 in AD. Currently, there is no genomic data available on Shroom-3 in AD cases. Hence, with this data access we aimed to perform genomic analysis for Shroom-3 and identify any potential mutations (SNPs) in Shroom-3 in AD. We will also look at Cdc37 chaperones and related proteins to fully characterize Shroom-3 and associated proteins.

Alzheimer’s disease (AD) is a highly heterogeneous disease with diverse clinical manifestations and rate of progression. AD subtypes identification will improve our understanding of the underlying disease mechanisms, enable us to predict disease trajectory and develop new disease-modifying treatments.Objective. To identify distinct AD subtypes from WGS data of AD individualsAnalysis plan. We will use 3000 WGS data derived from the ADSP Discovery Case-Control Based Extension Study. We will use the available SNVs and INDELS and infer structural variants (SVs) with our in-house multi-caller pipelines. Rare variants will be retained for further analysis. We will then split the dataset in training and tests set, and use the identified set of genetic variants (i.e. SNVs, INDELS and SVs) as input to a deep neural network (an autoencoder architecture) to learn an unsupervised latent representation of the data. AD subtypes will be identified within this reduced space and characterized using, demographics and clinical data. We will then contrast each subtype with the control groups to identify subtype relevant variants (i.e. putative subtype biomarkers), which will be used as input features to a gradient boosted tree model, to generate a subtype predictive model and subtype specific features.Planned collaboration. Each member of the team will devote effort in specific areas of investigation, nevertheless, all the team members will discuss, through regular meeting, individual progress and potential challenges. In particular, Dr Coppola (Research Scientist, Department of Pathology, Yale University, USA), together with Dr Dean Palejev (Associate Professor, GATE Institute, Sofia University, Bulgaria) will be involved in the deep learning model generation and validation, and subtype identification; Dr Fredrik Johansson (Assistant Professor, Department of Computer Science & Engineering, Chalmers University of Technology. Sweden), will work on the supervised machine learning model; Dr Alexander Schliep, Associate Professor, Department of Computer Science & Engineering, University of Gothenburg, Sweden), will work on the SVs inference.

Alzheimer’s disease (AD) is a highly heterogeneous disease with diverse clinical manifestations and rate of progression. The heterogeneity of AD has complicated both clinical trial design and outcomes, and thus the need for better models of AD, and/or better strategies for selection of participants into specic clinical trials is evident. The identication of more homogeneous disease subgroups (i.e. AD subtypes) will improve our understanding of the underlying disease mechanisms, enable us to predict disease trajectory and develop new disease-modifying treatments. We will use a comprehensive set of genetic variants in combination with deep learning algorithms to identify AD subtypes. Subtypes will be characterized using clinical and demographic data. Finally, variants specic to each cluster will be identied and used to train a predictive machine-learning model to classify new individuals.

The AI4AD (Artificial Intelligence for Alzheimer’s Disease) Initiative aims to create and develop advanced AI methods, and apply them to genomic, imaging and cognitive data, in order to 1) identify AD genotypes and endophenotypes that dissect AD’s heterogeneity; 2) relate said genotypes and endophenotypes with clinical progression in pre-dementia patients; 3) identify novel treatment targets for AD by analyzing whole genome and associated phenotypic data. The goals of this multisite initiative (Paul Thompson, USC; Christos Davatzikos, Li Shen, Penn; Andy Saykin, IU; Heng Huang, Pitt, Paul Crane, UW; Adam Brickman, Columbia; Tim Hohman, Vanderbilt; Guyngah Jun, BU; Duygu Tosun, UCSF; Alexander Zaranek, Curii) leverage the promise of machine learning (ML) to contribute to precision diagnostics, prognostication, and targeted and novel treatments. We will develop ML and deep learning methods to apply to large scale biobanks of whole genome sequences (WGS), neuroimaging, cognitive, and clinical data, aiming to discover new genomic features that influence biological processes of AD. We will apply methods of genome representation and tiling to WGS repositories to create inputs for AI methods. We will develop novel, interpretable, biological knowledge guided deep learning methods to discover genomic motifs associated with AD, AD risk, and biological processes of AD as defined by NIA-AA criteria. To quantify subtypes and disentangle biological processes of AD, we will apply computational methods to multimodal MRI and amyloid- and tau-sensitive PET to stratify and subtype patient groups; novel imaging genomics methods will detect genomic markers and pathways that modulate the developing pathology as detected in the images, and that predict future clinical decline or resilience. We hypothesize that advanced deep learning methods combined with whole genome data will outperform traditional methods and GWAS for predicting AD onset and progression, and will assist with disease subtyping and discovering treatable targets in the genome. A team will rank and repurpose existing, and identify novel drugs and targets in the genome based on the discovered genetic motifs affecting AD.

The AI4AD (Artificial Intelligence for Alzheimer’s Disease) Initiative aims to create and develop advanced AI methods, and apply them to extensive and rich genomic, imaging and cognitive data, in order to 1) identify genotypes and endophenotypes of AD that dissect the heterogeneity of the disease; 2) relate these genotypes and endophenotypes with clinical progression, in pre-dementia patients; 3) identify novel treatment targets for AD, by analyzing whole genome and associated phenotypic data at a previously impossible scale. Collectively, the goals of this highly collaborative multi-site initiative leverage the promise of machine learning to contribute to precision diagnostics, prognostication, and targeted and novel treatments.

Objectives: We are requesting access to the NIAGADS datasets to augment our ongoing studies on the genetics of Alzheimer’s disease (AD) and AD-related endophenotypes being carried out by Kamboh and his group since 1995. We are doing GWAS using array genotypes, whole-exome sequencing and whole-genome sequencing on datasets derived from University of Pittsburgh ADRC and ancillary population-based longitudinal studies on dementia and biomarkers. Different available phenotypes include AD and non-AD dementia, age-at-set, disease progression and survival, neuroimaging, cognitive decline, plasma biomarkers for the core ATN and non-ATN pathologies. We also plan to expand on gene-gene interaction and sex-stratified analyses which require the actual genotype data. The NIAGADS datasets will be used for replication and meta-analysis, and for gene-gene interaction and sex-stratified analyses. Study Design: A case-control design will incorporate a diverse cohort of individuals with AD and age-matched controls. For quantitative traits (neuroimaging and plasma biomarkers, cognitive performance measures, indicators of disease progression), linear regression analyses will be performed to identify genetic loci. To ensure the findings are robust and inclusive, participants from diverse demographic backgrounds will be included, enabling the exploration of potential genetic variations across populations. Analysis Plan: We will conduct GWAS and targeted analyses on candidate genes on different AD and AD-related phenotypes. Primary phenotypic variables include AD disease status, age-at-onset, last age for controls, APOE genotype, cognitive decline trajectories, sex, and race. Analyses will evaluate the influence of specific genetic variants on disease risk, cognitive performance, and biomarker levels, considering both individual and interactive effects of the APOE genotype. Results will be adjusted for potential confounders, such as demographic factors, to ensure valid associations. Detail analytical methods are described in our published papers for case-control (PMID: 32651314;35694926), quantitative traits (PMID: 30361487;37666928), and cognitive decline (PMID: 37089073; 30954325).

Our research group at the University of Pittsburgh (Pitt), has been working on the genetics of Alzheimer’s disease (AD) and AD-related endophenotypes for almost three decades, on data derived largely from the University of Pittsburgh Alzheimer’s Disease Research Center and ancillary dementia studies. We are requesting access to the NIAGADS genotype and phenotype datasets to augment our sample size to increase power to detect novel genetic associations with AD and related endophenotypes.

Dementia with Lewy Bodies (DLB) is an understudied Alzheimer’s-related dementia characterized neuropathologically by the accumulation of Lewy bodies in the brain. Like other dementias, definitive diagnosis can only be made upon autopsy after death, though numerous clinical features such as visual hallucinations, Parkinsonism, and REM sleep behavior disorder are associated with the condition. DLB is thought to account for approximately 5% of dementia diagnoses. A major barrier to understanding the natural history and pathology of this condition is the lack of definitive diagnoses that can be made during a patient’s lifetime.Recently, a polygenic risk score (PRS) was developed that can help identify people at higher risk of developing DLB based on their genetic profile (Chia et al, 2021). Intriguingly, the genetic variants associated with DLB identified in this study are also associated with risk of either Alzheimer’s disease or Parkinson’s disease, consistent with other observations that have identified pathological features in DLB similar to those two.This study will examine the association of the DLB polygenic risk score with various phenotypic measures in the ADSP. The goal of this analysis is to determine if specific features are enriched in people with a high propensity for DLB based on genetics. To the extent that these people have another diagnosed dementia, this would suggest the possibility of a misdiagnosis. Specifically, we will examine individuals whose data is shared through the NIA’s NIAGADS Data Sharing Service.For each individual, we will compute the DLB PRS previously described [3]. We will then ask if the score correlates with a diagnosis of Alzheimer’s disease, presence of amyloid, Braak stage, or Parkinson’s Disease Braak stage. We will also look at subsets of the DLB PRS consisting only of SNPs associated with Alzheimer’s disease or Parkinson’s disease.

Dementia with Lewy bodies (DLB) is an understudied dementia with features similar to both Alzheimer’s disease and Parkinson’s disease. Recent studies have developed a polygenic risk score that can predict who is at higher risk of DLB on the basis of their genetics. Genetic factors in this score have also been associated with Alzheimer’s disease and Parkinson’s disease. Here we will ask how the DLB risk score correlates with Alzheimer’s diagnosis and various features of the brains of people with Alzheimer’s to better understand the relationship between DLB, Alzheimer’s, and Parkinson’s.

Our overarching objective is to apply machine learning techniques to predict and interpret the functional effects of genetic variants including Single Nucleotide Variants (SNVs), indels and Structural Variants (SVs) from AD WGS data at the levels of DNA regulation and RNA processing, and link these effects directly to pathways and network context. We will leverage WGS generated by the ADSP and others together with harmonized endophenotypes and clinical data, multi-omics data from the AMP-AD, functional genomics data from Roadmap Epigenomics, PsychENCODE and GTEx Projects, and microglia and monocytes specific transcriptomic and single-cell RNA-seq data sets. Our central hypothesis is that many AD-associated genetic risk or protective variants influence pre- and post-transcriptional gene regulation, resulting in changes to gene expression and cellular pathways/networks, and ultimately contribute to protein aggregation in AD. The objective of this aim is to leverage deep-learning-based models capable of predicting functional effects of genomic variants on pre- and post-transcriptional gene regulation. We will train existing and novel sequence-based deep learning models of epigenomic state and RNA regulation and processing specific to AD-relevant cell types and states. in silico mutagenesis under these trained models will be used to calculate functional impact “delta scores” for every SNV, indel and structural variants (SV) detected from AD WGS. We will use these delta scores to empower non-coding rare variant tests of association with AD at the regulatory region, gene and pathway levels. We will conduct functional fine-mapping through the integration of (i) the CNN delta scores (ii-iii) expression and splicing quantitative trait loci (eQTL and sQTL), (iv) AD endophenotypes and (v) multi-ethnic AD WGS data. We will use probabilistic ML methods, combined with cell-type-specific and single-cell RNA-seq datasets, to build gene regulatory networks. This NIH funded project is a close collaboration with Dr. Towfique Raj at Mount Sinai Medical School.

Despite decades of research and enormous investment, no disease-modifying treatment is available for Alzheimer’s disease (AD). Combining population-scale data collection, human genetics and machine learning provide a way forward to uncover and characterize new causal cellular processes involved in AD. Effectively integrating diverse genomic data to better understand AD represents a substantial computational challenge, both in terms of data scale and analysis complexity. We will train machine learning models to predict epigenomic signals from the genomic sequences to estimate the functional impact of any genetic variant. These analyses will highlight variants and genes involved in AD. However, genes do not operate in a vacuum so robust machine learning will be used to learn cell-type and disease- specific networks. Such pathways will be prime candidates for future functional and therapeutic studies of AD.

Objective: to identify personalized polygenic profiles, comprised of the APOE e2 allele, other SNPs in the APOE region, and SNPs spread through the entire genome, with stronger protection in aging and Alzheimer’s Disease (AD) framework, and identify the role of AD risk factors in these profiles using ADSP, ADGC, and other studies. Study design: Aim 1 will identify molecular signatures of aging-related traits (AD, cardiovascular diseases, longevity, etc.), defined as differences in linkage disequilibrium (LD) patterns between affected and unaffected subjects using methods of LD analysis. In Aim 2 we will dissect heterogeneity in the molecular signatures using methods of stratification analyses. We will examine the impact of age at onset, sex, race/ethnicity, Braak stage, AD risk factors (diabetes, lipids, hypertension, body mass index, education), and other factors. Aim 3 will identify personalized polygenic profiles of aging-related traits using traditional and advanced bio-demographic methods. In Aim 4 we will perform bioinformatics analysis and characterize transcription pathways using summary statistics and individual-level data from the expression quantitative trait loci studies. In some cases, we may need to pool several datasets to increase power of the analyses in a mega sample. This will be done by pooling individuals’ records for genotypes and selected phenotypes described above from different studies. This pooling will not create any additional risks to participants because neither genetic nor phenotypic information for the same individual will increase. This research is consistent with data use restrictions for ADSP. We will not conduct non-genetic research, will not investigate individual pedigree structures, population origins, ancestry, individual participant genotypes, perceptions of racial/ethnic identity, variables that could be considered as stigmatizing an individual or group, or issues such as non-maternity. The research is designed to protect data confidentiality and follow local and institutional policies and procedures for data handling. The results of this research will be broadly shared with the scientific community.

Increasing population of the elderly individuals worldwide raises serious concerns about burden of geriatric conditions in future, especially Alzheimer’s disease, cardiovascular diseases, and other common aging-related diseases. These diseases can cluster in families suggesting that they can have genetic origin. Understanding their genetic origin could lead to breakthrough in preventing or curing such diseases. Despite continuing efforts, understanding their genetic basis remains very limited. Particular problem is to better understand genetic basis of Alzheimer’s disease, its relationship to other aging-related diseases, and identify genetic variants which could help protect against such diseases. This project focuses on identifying personalized polygenic profiles involving the Alzheimer’s disease protective genetic variant, so-called APOE e2 allele, which could strengthen protective effects against Alzheimer’s disease and investigate which factors can improve this protection. This research will facilitate the development of interventional strategies aiming to promote healthy aging.

Objective: to identify personalized genetic profiles of risks and resilience to Alzheimer’s disease (AD) and vascular diseases in the disease-specific and pleiotropic contexts in prioritized loci leveraging information from the AD-centered pleiotropic meta-analysis planned in this project and previous analyses by our and other research groups, and identify the role of AD risk and other factors in these profiles using ADSP, ADGC, and other studies. Study design: Aim 1 will identify specific and pleiotropic loci for AD and vascular traits from new analyses and the existing publications by: (i) performing pleiotropic genome-wide analysis focused on AD, cardiovascular diseases (CVD), and AD risk factors and (ii) identifying promising loci from this analysis and the results of previous analyses by our and other research groups. Aim 2 will dissect heterogeneity leveraging the analysis of molecular signatures defined as differences in linkage disequilibrium patterns in affected and unaffected subjects. Aim 3 will identify personalized genetic profiles of AD-specific and pleiotropic risks and resilience. Aim 4 will leverage biological, bioinformatics, and omics analyses to make sense of statistical inferences. In some cases, we may need to pool several datasets to increase power of the analyses in a mega sample. This will be done by pooling individuals’ records for genotypes and selected phenotypes described above from different studies. This pooling will not create any additional risks to participants because neither genetic nor phenotypic information for the same individual will increase. This research is consistent with data use restrictions for ADSP. We will not conduct non-genetic research, will not investigate individual pedigree structures, population origins, ancestry, individual participant genotypes, perceptions of racial/ethnic identity, variables that could be considered as stigmatizing an individual or group, or issues such as non-maternity. The research is designed to protect data confidentiality and follow local and institutional policies and procedures for data handling. The results of this research will be broadly shared with the scientific community.

Increasing population of the elderly individuals worldwide raises serious concerns about burden of geriatric conditions in future, especially Alzheimer’s disease, cardiovascular diseases, and other common aging-related diseases. These diseases can cluster in families suggesting that they can have genetic origin. Understanding their genetic origin could lead to breakthrough in preventing or curing such diseases. Despite continuing efforts, understanding their genetic basis remains very limited. Particular problem is to better understand genetic basis of Alzheimer’s disease, its relationship to other aging-related diseases, and identify genetic variants which could help protect against such diseases. This project focuses on identifying personalized genetic profiles of risk and resilience to AD and vascular diseases. This research will facilitate the development of interventional strategies aiming to promote healthy aging.

The purpose of this study is to find new Alzheimer related variants and genes, by combining exome and genome data from healthy controls and Alzheimer patients from different studies. Data will be analyzed using association, burden and variance component statistics.

Some individuals develop dementia, while others do not. A large part is likely determined by ones genes, Alzheimer’s disease has a heritability of up to 80%. What are the key genetic factors that determine if one will get Alzheimer's disease ? In this study, we will thoroughly explore genomic data of a large group of healthy persons and dementia patients to answer this question.

The ADSP-PHC was established to harmonize the rich endophenotype data across cohort studies to enable modern genomic analyses of ADRD with the ultimate goal to generate harmonized data that will become a “legacy” dataset perpetually curated and shared NIAGADS. The ADSP-PHC will facilitate phenotypic data harmonization for ADSP participants with genetic and genomic data. This effort represents a multi-disciplinary approach leveraging interdisciplinary expertise in large-scale genetic and genomic studies, clinical neuroscience, neuroimaging, psychometrics, and bioinformatics. This study will utilize a team structure consisting of two coordinating centers to oversee activities of all harmonization teams and to oversee coordination with other ADSP workgroups and initiatives, a Storage and Informatics Core that will oversee the coordination between LONI and NIAGADS for all data storage, compliance, and dissemination, a CHARGE Coordination Core to oversee the alignment of data and protocols with CHARGE workgroups, an Integration & Analytics Core that will enable data integration across phenotypes to facilitate downstream machine learning applications, and eight Domain-Specific Harmonization Teams tasked with harmonization in their area of expertise. The endophenotypes that will be harmonized by this project include Cognition, Fluid Biomarkers, Amyloid PET, Structural MRI, White Matter Hyperintensities, White Matter Integrity, Vascular Risk Factors, and Neuropathology. Site PIs on this project include: Jesse Mez (Boston University), Adam Brickman (Columbia University), Andy Saykin (Indiana University), Elizabeth Mormino (Stanford University), Pauline Maillard (UC Davis), Duygu Tosun-Turgut (UC San Francisco), Christos Davatzikos (University of Pennsylvania), Arthur Toga (USC); Mohamad Habes (University of Texas Health Science Center at San Antonio), Michael Cuccaro (University of Miami), Paul Crane (University of Washington), Bennett Landman (Vanderbilt University), Timothy Hohman (Vanderbilt University Medical Center), and Carlos Cruchaga (Washington University in St. Louis).

The growing availability of endophenotypic data in cohort studies of Alzheimer’s disease and related dementias (ADRD) provides an exciting opportunity to further characterize the genetic architecture of this devastating disease. However, there is a pressing need to develop and apply advanced harmonization approaches to align ADRD endophenotypes across cohorts. The goal of this coordinated national initiative – the AD Sequencing Project Phenotype Harmonization Consortium, or “ADSP-PHC” – is to provide a centralized database of robust endophenotypes for large-scale genomic analyses that will accelerate the identification of novel targets for therapeutic intervention in ADRD. will accelerate the identification of novel targets for therapeutic intervention in ADRD.

Using the haplotype-algorithm for FBAT, we will develop a general testing framework that will allow for the implementation of region-based association tests, e.g. SKAT, burden, multi-variate and, using a permutation approach, the calculation of exact p-values. All association test statistics will be computed based on the exact genetic variance/ covariance matrix. Permutation/ simulation-based p-values are obtained using our new haplotype-algorithm. We will also develop higher criticism approaches for the region-based association analysis of the AD samples. Recent developments in theoretical statistics have shown that higher criticism approaches are, by far, the most powerful statistical techniques to detect association signals in very spare data, for instance rare variant WGS data. In the framework that we developed for the previous funding cycle, the higher criticism approaches will be implemented based on the exact genetic exact variance/ covariance matrix. Permutation/ simulation-based p-values will be obtained using our haplotype-algorithm.We will also develop meta-analysis approaches to combine locus-specific and region-based association findings for AD across studies.Using reported association findings in the literature, we will develop polygenic risk score models/ integrated risk models that are based on the methodology of marker assisted selection, and evaluate their performance in terms of prediction of AD in simulation studies.We will evaluate all of our approaches by application to the requested data set (NG00067).

We will develop haplotype-based approaches for the region-based analysis of WGS data in family-based designs for Alzheimer's Disease. We will develop meta-analysis approaches to aggregate locus specific and region-based associations with Alzheimer's Disease across studies. Furthermore, we will develop polygenic risk score/ integrated risk model approaches that model age-at-onset and non-affection status as the primary phenotype for AD, thereby achieving better performances than standard approaches.

Background The gene progranulin is critical for maintenance of brain health. Progranulin is thought to regulate components of the autophagic/lysosomal pathway. Normal expression of progranulin is needed to prevent neurodegenerative disease caused by aberration in this pathway. Mutations that disrupt the function of progranulin cause Frontotemporal Dementia, while mutations that reduce progranulin expression are thought to increase risk for the development of Alzheimer’s disease. However, mutations that reduce progranulin expression remain to be fully characterized.Objectives This research project seeks to evaluate the effect of variation in progranulin’s regulatory regions on the risk of developing Alzheimer’s disease. In doing so, we hope to uncover novel Alzheimer’s risk variants and broaden the current understanding of the role of progranulin in Alzheimer’s disease.Study Design The first phase of this study involves the computational prediction of thousands of variants with the potential to alter progranulin expression. Briefly, transcription factor ChIP-seq data and transcription factor binding site sequence information has been utilized to identify regions within the progranulin gene where variation is expected to alter transcription factor binding. Variation in these transcriptionally active ‘regulatory’ sites is expected to modify progranulin expression. We will query these regulatory sites in Alzheimer’s patient and control genomes to identify novel variants that are associated with increased or decreased risk of Alzheimer’s. Variants of interest will be further characterized to determine the effect of the variant on progranulin expression.Analysis Plan We will assess enrichment of the most common variants (MAF > 0.005) by chi-squared test. Since most of the variants are thought to be rare, enrichment for these variants will be assessed using a rare variant association test called the Optimized Sequence Kernel Association Test (SKAT-O). Our primary analysis will assess whether any variants are enriched in case or control populations. Subsequently, we intend to evaluate if any variants are capable of modifying age of onset or symptom severity.

Maintenance of brain health during aging is a complicated process, involving many players. One such player is a protein called progranulin. Progranulin works to regulate components of the waste disposal system inside brain cells. Without enough functional progranulin, the waste builds up in brain cells and causes dementia. In fact, progranulin mutations are a common genetic cause of Frontotemporal Dementia. Furthermore, some progranulin mutations are thought to increase risk for the development of Alzheimer’s disease. Despite progranulin’s important role in maintaining brain health during aging, we still do not know much about the role of progranulin mutations in Alzheimer’s. To address this knowledge gap, we intend to perform the most comprehensive search to date for progranulin mutations that increase risk for Alzheimer’s. Any progranulin mutations that appear to increase risk for Alzheimer’s will be characterized further. Ultimately, this study seeks to expand our current understanding of the role of progranulin in the development of Alzheimer’s disease.

The goal of analyzing ADSP umbrella cohort data is identifying variants, genes and pathways associated to Alzheimer’s disease (AD), and stratifying patients by genetic risks. Following describes procedure.1) Identification and validation of genetic risks The whole genome and whole exome sequencing data will be analyzed to identify genetic variants or genes associated to phenotypes in case-control cohort, such as AD status and Braak stages. Several latest methods will be applied, such as VEP [William McLaren et al, 2016], LOFTEE [Karczewski, 2015] and PEXT scoring [Beryl B.C et al., 2020] for variant annotation and SAIGE-GENE [Wei Zhou et al., 2020] for the association test. The association will be tested for other endophenotypes such as cognitive scores and brain volumes that available in subset of the cohort. Replication and meta-analysis will be conducted on UK biobank and Tohoku medical megabank organization (ToMMo) cohort data. The ToMMo data consists of Japanese cohort so that we can analyze the effect of the variants among multi ethnic groups.2) Patient stratification in ADSP cohort Leveraging the increased sample size, we will stratify the cohort by genetic risks such as ApoE types, or phenotypes such as Braak stages, and compare the effect size of variants or genes among the patient groups. In addition, the genetic risk score (GRS) will be calculated using LDpred2 [Florian Prive, 2020], RapidoPGS [Guillermo Reales, 2020], and PRSice2 [Choi, S.W., 2020], and validated in independent cohorts and compared to available clinical endophenotypes. Last, we will search the effect of the GRS to extensive phenotypes in UK biobank and ToMMo.3) Identification of variants associated with pathologies and disease progression To further characterize patients by genetic risk, we will conduct GWAS and EWAS on pathology measurements, models of co-pathology, comorbidity with other neurological diseases, and disease progression. NG00127 and NG00154 will be used for this purpose.

The aim of our study is identifying variants or genes potentially causal of the Alzheimer’s disease in whole or subset of patients. To be specific, WES and WGS data will be analyzed to investigate common and rare variants associated with disease status and intermediate phenotypes. In addition, the patients will be stratified and sub-grouped by their genetic or phenotypic characteristics. Last, we will incorporate other large biobanks such as UK biobank or ToMMo to investigate the genetic effects to extensive phenotypes potentially linked to symptoms appearing in sub patient groups.

Aim 1: Identify novel genes and replicate existing gene associations for Alzheimer’s disease (AD). Aim 1a: Common variant genome-wide association analysis. With this approach, we will leverage existing consortium GWAS summary statistics where makes sense (or request leave-one/N summary association statistics out if we see a need to use a different version of phenotype definition from the same cohort) and augment them with additional datasets available internally. Aim 1b: Rare variant gene-level genetic burden analysis. Using the ADSP analysis pipeline, we will aim to use the same analysis pipeline (but reserve the option to use an alternative pipeline) to contribute the whole genome sequencing (WGS) data generated from the internal galantamine samples to ADSP-led consortium analysis. We will perform case-control and/or family-based genetic analyses and/or quantitative trait genetic analyses using AD traits such as diagnosis, age of onset, amyloid positivity, tau positivity, CSF biomarker endophenotypes, disease progression, etc. (where the phenotype is available) as the outcome of interest. Covariates include age, sex, and principal components. ADSP, UKB, and FinnGen will be analyzed separately and combined with a meta-analysis. Biobank cases will be defined using ICD-9/ICD-10 codes, and proxy cases and controls will be carefully defined using questionnaire data on the parental history of AD. Both true and proxy cases will be considered to maximize the number of AD cases. Aim 2: Prioritize novel gene associations identified in Aim 1. We will perform genetic fine-mapping and leverage tissue and cell-type specific datasets (e.g. GTEx, AD Knowledge Portal including AMP-AD, internal datasets, MiGA, Harari et al snRNA-Seq) to prioritize targets for further functional and analytical interrogation. Furthermore, multi-omics-based network approaches will be used to identify disease-related molecular modules and tissue-specific regulatory circuits. Aim 3: utilize single-nuclei sequencing data to more fully catalog cell type heterogeneity in the brains of individuals with AD and how this differs from brain from uninjured, cognitively unimpaired individuals.

Alzheimer’s disease (AD) is a common, progressive, neurodegenerative disorder with a strong genetic component with heritability estimates ranging from 58–79% for late-onset AD and over 90% for early-onset AD. To date, there is only one approved treatment option intended to mediate the disease progression of AD, while all others treat symptoms associated with AD. Genetic association studies are important to highlight key biological mechanisms contributing to the etiology of AD and provide insights into potential pathways that can ultimately be targeted for future therapeutic development. The aim of this study is to perform a retrospective analysis of genetic data collected from large-scale population-based and case-control cohorts including the UK Biobank, the Alzheimer’s Disease Sequencing Project (ADSP), FinnGen, and Janssen internal cohorts. We will also integrate them with available multi-modal datasets including but not limited to, Microglia Genomic Atlas, Harari et al snRNA-Seq, and neuroimaging data to identify novel and existing evidence for genetic determinants of AD.

Project Title: Identification of early biomarkers for Alzheimer’s Disease (AD)(1) Objectives of the proposed research This research project aims to develop a data-driven causal graph framework for the identification of biomarkers for the early detection of subjects who may potentially develop Alzheimer’s Disease (AD), while accounting for potential confounders, including genetic, sociodemographic, environmental, clinical, and behavioral factors.(2) Study designWe will use machine learning techniques to identify biomarkers, such as mutations in the blood, which are highly correlated with the onset of AD. Then we use a causal AI technique to identify the most causal of such biomarkers. Our hypothesis is that these most causal biomarkers can be used to identify presymptomatic subjects who may potentially develop AD. We track such presymptomatic subjects who eventually develop AD to test our hypothesis. (3) Analysis planThis is an interdisciplinary project, marrying AI and neuroscience. We develop a framework that utilizes AI and big-data to speed up the search for the early biomarkers, by incorporating domain knowledge on the complex causal pathological pathways and co-morbidities.We shall evaluate sociodemographic, environmental, clinical, and behavioral factors in association with genetic variants.

Identification of early biomarkers for Alzheimer’s Disease (AD) This will proceed in two stages. First we will use artificial intelligence techniques to identify biomarkers, such as mutations in the blood, which are highly correlated with the onset of AD. The top such markers will then be tracked. Our hypothesis is that these top markers will be able to detect the early onset of AD in presymptomatic subjects. We track presymptomatic subjects who eventually develop AD to test our hypothesis.

BACKGROUND, OBJECTIVES: Our group developed a method to estimate the impact of missense mutations, that we call the “Action” of missense mutations. This method is better than current state-of-the-art approaches at matching experimental data on mutational loss of function, not just in our own controls but also in blind competitions assessed objectively by independent judges (CAGI 2011, and 2012-13). When we used Action on head and neck cancer patient data (TCGA) we obtained significant separation of patient survival among those with a high Action and those with a low Action in somatic TP53 mutations. However, mutations in other genes may also correlate with patient outcome, such as the mutations of IDH1 in glioblastomas (Nobusawa et al., Clin Cancer Res, 2009). Therefore, we plan to integrate mutation impact information over the human proteome and identify how severely they affect the pathways associated with each cancer type. In addition, we like to test the same principles in data from complex diseases such as Alzheimer’s Disease. To do so, we developed a network diffusion method that uses current information of protein interactions (in a physical or broader sense) in order to project the dysfunction of a protein to its near neighbors (Lisewski et.al., Physica A, 2010). Putting these together, our hypothesis is that the diffusion of Action to the human protein network can identify novel Alzheimer’s disease-associated genes and provide a better stratification of patient outcome. STUDY DESIGN, ANALYSIS PLAN: To test our hypothesis we need to access “Individual germline variant data” of patients. For each individual, we will score the germline missense mutations by Action and treat it as the potential dysfunction on the protein. Then, we will diffuse this action over the network and measure the effect on each gene and on each pathway. When we compare these data to those from healthy individuals (1000 Genomes Project), i) we can identify genes associated to each disease and ii) the pathways that affect mostly the disease, and iii) measure the severity of the mutational damage to these genes or pathways. USE RESTRICTIONS: We will follow all restrictions described.

My group is interested in developing computational tools that predict i) disease-associated genes, ii) disease-causing mutations, and iii) the impact of an individual’s mutations to the phenotype. We make these predictions by comparing the mutational patterns of the cases with those expected either by random chance or given the purifying section observed in human polymorphisms. Here, we request access to the database of the NIA Genetics of Alzheimer’s Disease Data Storage Site (NIAGADS) Data Sharing Service (DSS), in order to obtain protein mutation data from patients and healthy individuals.

The objective of this study is to develop machine learning models using genetic and phenotype data from the NIAGADS database https://dss.niagads.org/. We will develop both unsupervised and supervised learning models to characterize the heterogeneity and risk of Alzheimer’s disease (AD). This is an MPI study in collaboration with Dr. Anita DeStefano at Boston University School of Public Health.For the first aim, we will build an expandable multimodal unsupervised machine learning framework to investigate AD heterogeneity. We will perform AD subtyping by harnessing the rich multimodality information across a wide spectrum of data (e.g., genetics, images and blood biomarkers). A Bayesian kernel network will be built to estimate the relative weight of each individual data modality, which would also allow the addition of new data modalities as they become available. The analyses will be performed both within and between ethnic populations.For the second aim, we will build an expandable multimodal supervised machine learning framework to quantify AD risk from longitudinal follow-up of clinically normal elders. We will build a separate deep learning network for each data modality in consideration of its unique feature sets. A multiplicative strategy will then be taken to aggregate information from different modalities with weighted contributions. Feature selection will also be performed to identify the most informative features predictive of AD risk.For the third aim, we will build AD-related gene regulatory networks in post-mortem human brain samples. We will examine the association of multi-omics data with AD, which will be used to assign gene priority based on the combinatorial evidence from each type of omics data. A gene ontology-guided greedy search strategy will then be implemented to build gene regulatory networks, and identify key drivers that might be potential therapeutic targets for AD. The analyses will be stratified by ethnic populations and AD phenotypic clusters.

Alzheimer's disease (AD) is the most common form of dementia characterized by progressive loss of cognitive function. There are very limited treatment options for AD. For the current application, we seek to develop multimodal machine learning models by leveraging the rich collection of AD-related omics data and phenotypical data recently generated from the Alzheimer's Disease Sequencing Project (ADSP). Three aims will be pursued in the current application. For Aim 1, we will build an expandable multimodal unsupervised machine learning framework to investigate AD heterogeneity. For Aim 2, we will build an expandable multimodal supervised machine learning framework to quantify AD risk from longitudinal follow up of cognitively normal elders. For Aim 3, we will build AD-related gene interaction networks in post-mortem human brain samples. The present application represents an innovative approach to identify individuals at high risk of AD. The outlined strategy will provide new insights into the risk stratification and prevention strategies for AD.

Congenital heart disease (CHD) affects approximately 1% of infants born each year. While CHD was previously fatal, surgical palliation now allows most patients to survive into adulthood. With more adults living with CHD, there is increasing appreciation for continuing health problems among CHD patients, such as high risk for dementia and Alzheimer’s disease. Recent studies show that APOE modifies neurodevelopmental outcomes in the CHD population (Gaynor JW, J Thoracic Cardiovascular Surgery, 2014) and that CHD patients have higher risk for Alzheimer’s disease (Bagge CN, Circulation, 2018). We hypothesize that CHD and Alzheimer’s have shared genetic causes and modifiers. Further insights into the genetic causes for CHD and dementia may reveal novel genetic relationships between the two diseases and provide possibilities for improvements in long term neurological outcomes for CHD patients. We have performed whole exome sequencing at 80x coverage on a discovery cohort of over 600 CHD patients recruited at the University of Pittsburgh Children’s Hospital and obtained access to a cohort of ~4000 healthy older individuals sequenced by the Medical Genome Reference Bank (MGRB) for use as population-matched controls. Here we will perform case-control association analysis with human next-generation sequencing data to identify SNVs, indels, and CNVs associated with CHD. We request access to sequencing data from the Alzheimer’s Disease Sequencing Project to perform a separate case-control analysis, comparing the Alzheimer’s cohort to the MGRB controls. We will then compare genes and variants that are significantly associated with each disease to identify shared pathways involved in disease pathogenesis. Processing and statistical analysis will be performed on the Pittsburgh Supercomputing Center using GATK, bcftools, PLINK, SKAT, and MAGMA well as custom shell, Python, and R scripts. These studies should help us elucidate the shared genetic etiology of CHD and Alzheimer’s disease. We intend to publish or share any findings from this study with the scientific community by presenting at national scientific meetings.

Congenital heart disease (CHD) affects approximately 1% of infants born each year. While CHD was previously fatal, surgical palliation now allows most patients to survive into adulthood. With more adults living with CHD, there is increasing appreciation for continuing health problems among CHD patients, such as high risk for dementia and Alzheimer’s disease. We hypothesize that CHD and Alzheimer’s have shared genetic causes and modifiers. Further insights into the genetic causes for CHD and dementia may reveal novel genetic relationships between the two diseases and provide possibilities for improvements in long term neurological outcomes for CHD patients. Here we will compare genes and variants that are significantly associated with each disease based on case-control analysis to identify shared pathways involved in disease pathogenesis. In the future, we will study the functional consequences of such mutations using cells and mouse models. These studies should help us to elucidate the shared genetic etiology of CHD and Alzheimer’s disease.

Objective The objective of the proposed study is to establish the connection between Alzheimer’s Disease-related genomic markers and neuroimaging phenotypes and their association with the clinical onset of dementia. We hypothesize that a) genomic factors are associated with diverse Alzheimer’s Disease-related neuropathological and clinical progression patterns; and b) the genotype-phenotype interaction is dynamic along the Alzheimer’s Disease progression trajectory, which in turn regulates the clinical progression of dementia.Study design We plan to develop data-driven computational models using multi-modal imaging-genomics information, to test these hypotheses with the following two Specific Aims: (1) construct clinically relevant computational neuroimaging-genomic fingerprints to characterize distinctive subtypes of Alzheimer’s Disease neuropathological patterns, and (2) Construct clinically explainable subtype-aware AI models with effective genomic-neuroimaging information fusion to achieve accurate prediction of disease progression of Alzheimer’s Disease.Analysis plan I will construct and validate harmonized models by utilizing the available data from the Alzheimer's Disease Sequencing Project Phenotype Harmonization Consortium, which is a multi-institutional effort that harmonized phenotypical data of 22k participants collected from 31 AD-related cohorts to produce a large-scale, racially diverse, standardized set of clearly defined data.1. We will develop semi-supervised machine-learning-based classification frameworks to explore the complex genotype-phenotype associations that determine distinctive neuroimaging-based pathological progression patterns.2. We will also develop machine-learning model predictions of future AD-specific neuropathological biomarkers. More specifically, we aim to predict the progression of cortical Aβ levels for identifying pre-symptomatic subjects, and progression of tau levels for symptomatic subjects.

Alzheimer’s Disease (AD) is a complex neurodegenerative disease with multiple variations of pathologies that affect the brain function, eventually leading to cognitive decline. Individual variations of our gene might be associated with different subtypes of the disease. Thus, it is important to explore the disease characteristics within the various AD subtypes to achieve personalized diagnosis and precision medicine, and eventually developing effective treatments for AD. The objective of this proposal is to study the connection between AD-related genomic markers and neuroimaging phenotypes and their association with the clinical onset of dementia.

The ADSP-PHC was established to harmonize the rich endophenotype data across cohort studies to enable modern genomic analyses of ADRD with the ultimate goal to generate harmonized data that will become a “legacy” dataset perpetually curated and shared NIAGADS. The ADSP-PHC will facilitate phenotypic data harmonization for ADSP participants with genetic and genomic data. This effort represents a multi-disciplinary approach leveraging interdisciplinary expertise in large-scale genetic and genomic studies, clinical neuroscience, neuroimaging, psychometrics, and bioinformatics. This study will utilize a team structure consisting of two coordinating centers to oversee activities of all harmonization teams and to oversee coordination with other ADSP workgroups and initiatives, a Storage and Informatics Core that will oversee the coordination between LONI and NIAGADS for all data storage, compliance, and dissemination, a CHARGE Coordination Core to oversee the alignment of data and protocols with CHARGE workgroups, an Integration & Analytics Core that will enable data integration across phenotypes to facilitate downstream machine learning applications, and eight Domain-Specific Harmonization Teams tasked with harmonization in their area of expertise. The endophenotypes that will be harmonized by this project include Cognition, Fluid Biomarkers, Amyloid PET, Structural MRI, White Matter Hyperintensities, White Matter Integrity, Vascular Risk Factors, and Neuropathology. Site PIs on this project include: Jesse Mez (Boston University), Adam Brickman (Columbia University), Andy Saykin (Indiana University), Elizabeth Mormino (Stanford University), Pauline Maillard (UC Davis), Duygu Tosun-Turgut (UC San Francisco), Christos Davatzikos (University of Pennsylvania), Arthur Toga (USC); Mohamad Habes (University of Texas Health Science Center at San Antonio), Michael Cuccaro (University of Miami), Paul Crane (University of Washington), Bennett Landman (Vanderbilt University), Timothy Hohman (Vanderbilt University Medical Center), and Carlos Cruchaga (Washington University in St. Louis).

The growing availability of endophenotypic data in cohort studies of Alzheimer’s disease and related dementias (ADRD) provides an exciting opportunity to further characterize the genetic architecture of this devastating disease. However, there is a pressing need to develop and apply advanced harmonization approaches to align ADRD endophenotypes across cohorts. The goal of this coordinated national initiative – the AD Sequencing Project Phenotype Harmonization Consortium, or “ADSP-PHC” – is to provide a centralized database of robust endophenotypes for large-scale genomic analyses that will accelerate the identification of novel targets for therapeutic intervention in ADRD.

The objective of our research is to characterize genomic rearrangements associated with various human disease including Alzheimer’s. The overarching hypothesis guiding our research is that repair of DNA double-strand breaks (DSBs) by using ‘risky’ inaccurate pathways can lead to genomic destabilization. Our focus is on two DSB repair pathways: break-induced replication (BIR) and microhomology-mediated BIR (MMBIR). BIR is initiated by a broken DNA end invading into a homologous template followed by extensive DNA synthesis that is highly mutagenic. Interruptions of BIR leads to initiation of MMBIR, a template-switching event that often leads to complex genomic rearrangements and has been linked to neurological conditions and to cancer. The overall goal of our proposed research is to define the molecular mechanisms of MMBIR, and to identify factors that inhibit or promote cells entering into MMBIR.We aim to achieve this using our MMBSearch tool to detect MMBIR events that are often missed by other methods in human WGS analyses. Using MMBSearch we will analyze data from NIAGADS, specifically data on neurological disease associated whole genome sequencing (WGS) and whole exome sequencing (WES) to detect MMBIR events associated with neurodegenerative disorders.The results of this analyses will be used to determine the frequency of MMBIR in various types of human cells and their association with neurodegenerative disorders. In addition, we will identify chromosomal locations where MMBIR events are especially abundant and specific features in humans that predispose them to MMBIR. We will identify genetic variations predisposing cells to MMBIR, which may uncover that specific SNPs, structural variations, certain gene mutations, etc. are associated with MMBIR events. We specifically hypothesize that mutations in DNA repair, DNA replication, chromatin maintenance, and DNA damage checkpoint genes could promote MMBIR. These studies will shed light on the etiology and mechanism of MMBIR to potentially develop biomarkers for early detection and design targeted therapies to treat human disorders.

The goal of our research is to understand the underlying mechanisms of genomic instability that lead to human disease. In particular, we are interested to investigate the molecular mechanism of an essentially uncharacterized DNA repair pathway, microhomology-mediated break-induced replication (MMBIR) that has been implicated in DNA mutations and found in a variety of human cancers and in association with neurological diseases. We have recently described a diagnostic pattern of mutations associated with MMBIR using a yeast model, which has allowed us to develop a novel algorithm to search for MMBIR events in sequenced human genomes. We are planning to apply this new algorithm to identify MMBIR events in analyzing human genome databases. The proposed research will allow us to further understand mechanisms of leading to various human diseases including cancer and neurological human diseases and to refine our software that is aimed to detect MMBIR in human genomes. The proposed research will be focused on analyzing the data from NIAGADS database.

Project Title: The Australian Imaging Biomarkers and Lifestyle (AIBL) Flagship Study of Ageing: Detecting and Preventing Alzheimer’s disease: Towards Lifestyle Interventions - Somatic mutation in Alzheimer's Disease (sub-project)Objectives -- Somatic Mutation in AD is a project to identify non-congenitally acquired genetic risks associated with disease onset of sporadic Alzheimer’s disease (AD). Somatic mutation can be any form of alteration in DNA that occur after conception. As opposed to congenital, it’s generally not hereditary unless the germ cells are involved. These alterations can (but do not always) cause disease. We aim to identify somatic variants that contribute to sporadic AD. We believe that the detection of somatic mutations can overcome the flaws of the large genome-wide multiple testing and increase the signal-to-noise ratio to pinpoint the rare genetic determinants that were largely neglected by current genetic association studies.Study design -- We have collected 20 paired human brain microglial DNAs (treated as “tumour”) and whole blood DNAs (treated as “normal”) to call somatic mutations by a tumour-normal mode using a software, MuTect2 (Broad Institute). The sequence has been obtained from the whole genome. Hundreds of rare genetic variants have been identified to connect with AD.Analysis plan -- We’d like to validate our results using datasets like NG00067, NG00105 and NG00106. However, it’s ideal if we could access the alignment data (i.e., BAM files) as well. Because technically somatic calling is not simply a difference between normal (germline) and reference; but also calls for tumour against normal (germline) alongside alignment. MuTect2 is developed to identify somatic mutations. It works with or without matching normal. Once we get access to the alignment data, we will reprocess all samples using the MuTect2 without matching the normal pipeline. We'll call somatic mutations using those datasets and validate the rare genetic determinants that contribute to sporadic AD.

Somatic Mutation in Alzheimer's disease is a project to identify non-congenitally acquired genetic risks associated disease onset of a sporadic Alzheimer’s disease (AD). We believe that detection of somatic mutations can pinpoint the rare genetic determinants that were largely neglected by current genetic association studies. In our pilot study, we have identified hundreds of rare genetic mutations that are strongly associated with AD. We'd like to validate our results using an independent cohort. We plan to reprocess NIH datasets using our own pipeline. But we would need to access the raw data rather than the processed data. This research will greatly accelerate the research on the molecular genetics of AD.

For this study, we will analyze data from whole genome sequencing (WGS) of from the Alzheimer's Disease Sequencing Project (ADSP) discovery-replication phase families and independent case control data from ADSP extension study. We will also analyze WGS and whole exome sequencing (WES) data from the Alzheimer's Disease Neuroimaging (ADNI) study and the ADSP follow-up study (ADSP-FUS) as they become available. The overall goal of this project is to identify and annotate causal variants related to LOAD using sequencing data generated from families multiply affected by the disease and validate the results in independent case-control datasets. Using families as discovery and unrelated individuals as replication and having the ability to genotype additional family members can provide direct evidence of causality by establishing which variants co-segregate in families and are associated in the general population with disease.

Analyses of whole genome, whole exome and targeted resequencing will continue to provide important new information regarding potential risk conferring genes, biochemical pathways involved in Alzheimer's disease and targets that may be suitable for pharmacological manipulation. While whole exome and targeted sequencing are powerful technologies, analysis of whole genomes will provide more information and allow discovery of rare, high risk variants.

The objective of the proposed research is to elucidate the genetic risk factors for Inflammatory Bowel Disease (IBD) in a Hispanic cohort. In order to fulfill this objective, we are requesting access to whole genome sequence (WGS) data for 500 Hispanic controls which were sequenced by the National Human Genome Research Institute as part of the Alzheimer Disease Sequencing Project Discovery Case-Control Based Extension Study. More specifically, these data will be combined with WGS data from our Hispanic cohort of ~1600 IBD cases and ~900 controls ascertained through the Crohn's and Colitis Center at the University of Miami in Miami, Florida and Cedars Sinai Medical Center in Los Angeles, California. Drs. Jacob McCauley and Maria Abreu will serve as the primary investigators for this application at the University of Miami and Dr. Dermot McGovern will serve as the primary investigator at Cedars Sinai Medical Center. The use of these data by the noted not-for-profit organizations will be limited to biomedical purposes as related to IBD, in accordance with data use limitations.The analyses to be conducted in this combined Hispanic sample of ~1600 IBD cases and ~1400 controls include case-control association for replication of single variant IBD associations previously identified in European populations as well as assessment for homogeneity of effect size across populations. Polygenic risk scores will be utilized to test for cumulative variant associations with IBD. Fine-mapping of single variant associations will be done using several parallel approaches. Firstly, trans-ethnic meta-analysis with summary statistics from published European and African American studies will be conducted. Linkage disequilibrium (LD) structure around associated variants will be assessed using knowledge of replication and information on LD from ancestral source populations. Secondly, the Bayesian FINEMAP algorithm will also be considered. Additional analyses will include admixture mapping for identification of novel signals and gene-based tests for cumulative association of rare variants.

The objective of the proposed research is to elucidate the genetic risk factors for Inflammatory Bowel Disease (IBD) in a Hispanic cohort. This project combines whole genome sequence data from 500 Hispanic controls which were sequenced as part of the Alzheimer Disease Sequencing Project with ~1600 IBD patient and ~900 control samples which were ascertained by the Crohn’s and Colitis Center at the University of Miami and Cedars Sinai Medical Center. To facilitate our objective, several analyses will be conducted in this combined sample of ~1600 IBD patients and ~1400 controls. First, we will identify DNA segments that contain a genetic variant that occurs more often in IBD patients than in healthy controls. These segments may be novel to Hispanics or have been identified previously in populations of European ancestry. These DNA segments can be large and often the exact location of the predisposing variant within each segment is unclear. Therefore, additional analyses will be done to narrow the width of these DNA segments and identify the predisposing variant(s) within each segment.

The ADSP-PHC was established to harmonize the rich endophenotype data across cohort studies to enable modern genomic analyses of ADRD with the ultimate goal to generate harmonized data that will become a “legacy” dataset perpetually curated and shared NIAGADS. The ADSP-PHC will facilitate phenotypic data harmonization for ADSP participants with genetic and genomic data. This effort represents a multi-disciplinary approach leveraging interdisciplinary expertise in large-scale genetic and genomic studies, clinical neuroscience, neuroimaging, psychometrics, and bioinformatics. This study will utilize a team structure consisting of two coordinating centers to oversee activities of all harmonization teams and to oversee coordination with other ADSP workgroups and initiatives, a Storage and Informatics Core that will oversee the coordination between LONI and NIAGADS for all data storage, compliance, and dissemination, a CHARGE Coordination Core to oversee the alignment of data and protocols with CHARGE workgroups, an Integration & Analytics Core that will enable data integration across phenotypes to facilitate downstream machine learning applications, and eight Domain-Specific Harmonization Teams tasked with harmonization in their area of expertise. The endophenotypes that will be harmonized by this project include Cognition, Fluid Biomarkers, Amyloid PET, Structural MRI, White Matter Hyperintensities, White Matter Integrity, Vascular Risk Factors, and Neuropathology. Site PIs on this project include: Jesse Mez (Boston University), Adam Brickman (Columbia University), Andy Saykin (Indiana University), Elizabeth Mormino (Stanford University), Pauline Maillard (UC Davis), Duygu Tosun-Turgut (UC San Francisco), Christos Davatzikos (University of Pennsylvania), Arthur Toga (USC); Mohamad Habes (University of Texas Health Science Center at San Antonio), Michael Cuccaro (University of Miami), Paul Crane (University of Washington), Bennett Landman (Vanderbilt University), Timothy Hohman (Vanderbilt University Medical Center), and Carlos Cruchaga (Washington University in St. Louis).

The growing availability of endophenotypic data in cohort studies of Alzheimer’s disease and related dementias (ADRD) provides an exciting opportunity to further characterize the genetic architecture of this devastating disease. However, there is a pressing need to develop and apply advanced harmonization approaches to align ADRD endophenotypes across cohorts. The goal of this coordinated national initiative – the AD Sequencing Project Phenotype Harmonization Consortium, or “ADSP-PHC” – is to provide a centralized database of robust endophenotypes for large-scale genomic analyses that will accelerate the identification of novel targets for therapeutic intervention in ADRD. will accelerate the identification of novel targets for therapeutic intervention in ADRD.

Objectives: Analyze the approximately 3,500 DNA sequences from GWAS at NIAGADS for associations between gender, AD, and frequency of SNPs in chromosomal DNA using the P-Link software analytical tool. Rationale: We have performed such analyses of the DNA sequences (VCF files) made available by ADNI and have identified a significantly higher incidence of SNPs (p<10-7) in a few chromosomal genes in males vs. females with the diagnosis of AD. We would like to perform similar analyses to the DNA sequences (in VCF files) of the greater than 3,000 sequenced DNAs at NIAGADS. Plan: The SNP association analyses will be performed on the DNA sequences using information about the gender and diagnosis of each individual whose DNA sequence we would analyze. We will use the P-Link software to generate data tables (Tab delimited P-Link association files) and Manhattan plots of genome-wide associations (gnuplot). There are no multiple research sites participating in the planned analysis of the DNA sequences. All work will be performed at the University of Kansas AD Center.

Variations in the sequences of DNA in our chromosomes and their association with the incidence of Alzheimer's disease (AD) have been identified in the last 10 years and have brought about new thinking regarding possible causes of AD. These variations in DNA do not directly cause the disease but increase the likelihood of the onset of AD in some individuals late in their life. For many years, it has been known that among various populations there is differential incidence of AD between males and females. In our initial study of a relatively small number of individuals with or without AD, we identified that there was a significant association between a few of the variants in DNA sequences and the incidence of AD in males as compared with females. The study planned will use the DNA sequences in the NIAGADS repository to conduct a similar analysis for variants in DNA sequences. The NIAGADS sequences represent a substantially larger population than that which we analyzed previously and should allow us to explore the possible association of gender and AD with variants of DNA.

The goal of this project is to develop artificial intelligence (AI) approaches for extracting unforeseen patterns from clinical, genetic, genomic, and imaging data that could lead to ideas for new drug development or drug repurposing. Our proposed AI methods and software will be open-source, user-friendly, and freely available for all to use. Specifically, we will analyze ADSP data sets using three novel informatics methods to tailor our automated machine learning (AutoML) tool to the analysis of Alzheimer’s disease (AD) data. First, we will develop a Multi-Modal Interaction (M2I) feature selection algorithm for identifying genetic interactions that are predictive of AD (AIM 1). Second, we will develop a Knowledge-driven Multi-omics Integration (KMI) algorithm for combining omics features for AI analysis of AD (AIM 2). Third, we will develop a Multidimensional Brain Imaging Omics (MBIO) integration framework for the joint analysis of multisource large-scale data for predicting AD. Finally, we will integrate all three biomedical informatics methods into our open-source AutoML software package and apply it to the ADSP data sets. We expect our methods will reveal new biomarkers for AD that will open the door for better treatments and clinical decision support.

The goal of this project is to develop artificial intelligence (AI) approaches for extracting unforeseen patterns from clinical, genetic, genomic, and imaging data that could lead to ideas for new drug development or drug repurposing. We will develop three biomedical informatics methods with focuses on genetics, genomics and imaging respectively. We will integrate these methods into our open-source AutoML software package, and apply it to the ADSP data sets. We expect our methods will reveal new biomarkers for AD that will open the door for better treatments and clinical decision support.

A central goal of human genetics is to understand the link between genotypic and phenotypic variation, including disease risk and local adaptations. In order to perform these analyses reliably, we need to characterize population structure reliably. The historical signatures of our past such as population mixtures, contractions, and expansions, as well as human diseases and natural selection, have left traces in our genomes. In this proposal, our objective is to develop and apply computational methods to reliably characterize population history, including admixture (recent and ancient events including Neanderthal and Denisovan gene flow), founder events and natural selection in the diverse populations in the ADSP dataset. We will use these inferences to then reliably characterize disease associations and identify signals of natural selection, leveraging local and global ancestry inferences and estimates of relatedness across samples in a linear mixed model or other frameworks. We will look for associations between Alzheimer's disease and other cognitive traits and identify genes and genomic regions/ pathways associated to these traits. We will also study the distribution of archaic ancestry––from Neanderthals, Denisovans or other unknown archaic hominins––across the genome to find regions of exceptional high (or low) archaic ancestry that might provide hints about the function of these regions. Further, we will characterize the differences in mutation patterns across different human populations and archaic and modern human lineages. Together, this project will provide insights about our evolutionary past, mutation patterns and genes associated to local adaptations and Alzheimer’s disease and cognitive traits in a diverse multi-ethnic cohort. Together, this project will provide insights about our evolutionary past, genes associated to local adaptations and Alzheimer’s disease, and evolution of cognitive traits in a diverse multi-ethnic cohort.

Evolutionary history shapes our genes and traits. The historical signatures of our past such as population mixtures, contractions, and expansions, as well as human diseases and natural selection, have left traces in our genomes. In this proposal, our objective is to develop and apply computational methods to reliably characterize population history, including admixture, founder events and natural selection in multi-ethnic individuals. Additionally, we will use these insights to reliably map genes and pathways associated to Alzheimer's disease and other traits in the diverse individuals in ADSP.

The ADSP-PHC was established to harmonize the rich endophenotype data across cohort studies to enable modern genomic analyses of ADRD with the ultimate goal to generate harmonized data that will become a “legacy” dataset perpetually curated and shared NIAGADS. The ADSP-PHC will facilitate phenotypic data harmonization for ADSP participants with genetic and genomic data. This effort represents a multi-disciplinary approach leveraging interdisciplinary expertise in large-scale genetic and genomic studies, clinical neuroscience, neuroimaging, psychometrics, and bioinformatics. This study will utilize a team structure consisting of two coordinating centers to oversee activities of all harmonization teams and to oversee coordination with other ADSP workgroups and initiatives, a Storage and Informatics Core that will oversee the coordination between LONI and NIAGADS for all data storage, compliance, and dissemination, a CHARGE Coordination Core to oversee the alignment of data and protocols with CHARGE workgroups, an Integration & Analytics Core that will enable data integration across phenotypes to facilitate downstream machine learning applications, and eight Domain-Specific Harmonization Teams tasked with harmonization in their area of expertise. The endophenotypes that will be harmonized by this project include Cognition, Fluid Biomarkers, Amyloid PET, Structural MRI, White Matter Hyperintensities, White Matter Integrity, Vascular Risk Factors, and Neuropathology. Site PIs on this project include: Jesse Mez (Boston University), Adam Brickman (Columbia University), Andy Saykin (Indiana University), Elizabeth Mormino (Stanford University), Pauline Maillard (UC Davis), Duygu Tosun-Turgut (UC San Francisco), Christos Davatzikos (University of Pennsylvania), Arthur Toga (USC); Mohamad Habes (University of Texas Health Science Center at San Antonio), Michael Cuccaro (University of Miami), Paul Crane (University of Washington), Bennett Landman (Vanderbilt University), Timothy Hohman (Vanderbilt University Medical Center), and Carlos Cruchaga (Washington University in St. Louis).

The growing availability of endophenotypic data in cohort studies of Alzheimer’s disease and related dementias (ADRD) provides an exciting opportunity to further characterize the genetic architecture of this devastating disease. However, there is a pressing need to develop and apply advanced harmonization approaches to align ADRD endophenotypes across cohorts. The goal of this coordinated national initiative – the AD Sequencing Project Phenotype Harmonization Consortium, or “ADSP-PHC” – is to provide a centralized database of robust endophenotypes for large-scale genomic analyses that will accelerate the identification of novel targets for therapeutic intervention in ADRD. will accelerate the identification of novel targets for therapeutic intervention in ADRD.

Historically, young onset Alzheimer’s disease (YOAD) is defined as onset of clinical symptoms before the age of 65, and 90% of such patients are not associated with mutations in the main 3 Ab related genes (APP, PSEN1 and PSEN2). The objective of this project is to deeply characterize known and new genetic components of YOAD in the largest pathologically confirmed cohort in the world and to evaluate the impact of SNVs, SVs and repeat expansions. Due to their earlier onset age, and strong heritability, we hypothesize that YOAD patients are enriched in rare pathogenic variants within the Abeta and Tau pathways. In addition, patients with YOAD are often misdiagnosed for frontotemporal dementia due to clinical symptom overlap. In that context, we also hypothesize that there is genetic overlap between both diseases. We generated whole-genome sequencing data from over 900 YOAD patients, including more than 400 autopsy confirmed YOAD cases, over 1000 FTD patients and 800 controls. We are requesting access to the ADSP whole-genome sequencing data (raw and VCF) to i) to perform gene-based, single variant and pathway association analyses in ADSP YOAD and late onset data to replicate our findings, ii) increase our YOAD cohort size for assessing the overlap and differences between FTD and YOAD patients. Single nucleotide variants, as well as structural variants, will be assessed. To do so, we will use already generated SNVs VCF but also generate structural variant calling using our Mayo pipeline. We will utilize several commonly used software programs, such as Plink-seq and SKAT package, to perform our association analyses. All analyses will be done at the single variant, gene, structural variant, and pathway levels. Using these approaches, we hope to identify novel mutations/genes/pathways that are related to both AD and FTD and will benefit the larger scientific community working on neurodegenerative disorders.

We aim at identifying new risk factors for young onset Alzheimer’s disease (age at onset before 65). To do so we utilize deep phenotyping and genetic approaches. Upon completion of our work, we will obtain a comprehensive understanding of young onset Alzheimer’s disease genetics. Altogether our project will identify potential new therapeutic targets for young onset Alzheimer’s disease and will pave the way for individualized therapy development not only for young onset Alzheimer’s disease, but also for the more common late onset AD.

We are a part of collaborations with the Yokoyama lab at UCSF and the Kosik lab at UCSB to analyze genomes for early onset Alzheimer’s and frontotemporal dementia cohorts compared to unaffected controls. A critical part of these efforts is replication of any findings in independent cohorts. Access to Alzheimer's Disease Sequencing Project (ADSP) data is ideal for this purpose. We will analyze ADSP data for association signals identified in our independent cohorts using either single variant or burden analysis approaches. Phenotypic characteristics that will be evaluated in association with genetic variants will be either case/control status or age of symptom onset as available. Although we conduct these projects as collaborations, this application is for analysis of ADSP data at HudsonAlpha.

We work together with the Yokoyama lab at UCSF and the Kosik lab at UCSB to analyze the DNA from patients with early onset Alzheimer’s and frontotemporal dementia in comparison to people without these diseases. A critical part of this type of work is checking to see if findings from one set of patients are reproducible in different sets of patients. Access to Alzheimer's Disease Sequencing Project (ADSP) data would allow for us to answer this question. We will analyze ADSP data for association signals identified in our independent sample sets. The types of data that will be evaluated in association with genetics will be either if the individuals assessed have disease or not, or if their genetics affects when they develop disease.

The purpose of this study is to find new Alzheimer related variants and genes, by combining exome and genome data from healthy controls and Alzheimer patients from different studies. Data will be analyzed using association, burden and variant component statistics.

Some individuals develop dementia, while others do not. A large part is likely determined by gene, Alzheimer’s disease has a heritability of up to 80%. What are the key genetic factors that determine if one will get Alzheimer disease? In this study, we will thoroughly explore genomic data of a large group of healthy persons and dementia patients to answer this question.

The overall objective of the proposed research is to use dataset NG00067 to conduct genetic research for better understanding of genetic protective and risk factors underlying the Alzheimer’s disease (AD) continuum and use artificial intelligence techniques to improve individual-level AD prediction with genomic features. Our central hypothesis are 1) loss-of-function single nucleotide polymorphisms (SNP) of AD-risk genes, and their interactive genes at the protein level, might confer protective effect against AD; and 2) deep convolutional neural network (CNN) can deal with large and colinear feature space, thus the CNN classifier could potentially improve the performance of individual-level AD continuum prediction with genetic features over non-deep classifiers. To test these hypothesis, we plan to examine single loss-of-function mutation SNPs of AD-risk genes, and their protein level interactors, as potential protective factors for AD (Aim1). More specifically, AD-associated genetic risks/hazards will be obtained from previous literatures. Their protein-level interactive genes will be obtained from the Reactome database. Loss-of-function SNPs associated with these genes will be further obtained from the gnomAD database. Phenotypic comparisons along the AD continuum will be conducted among 0-copy, 1-copy, and 2-copy mutation carriers of these SNPs. In addition, we plan to develop a CNN-based deep learning technique to derive high-order meaningful genetic features from the whole-genome or whole-exome sequencing data that could improve the individual-level disease prediction (Aim2). For subjects with brain imaging or CSF biomarker data available from the NACC database, we plan to further jointly evaluate the associations among genetic risks/protectors, brain imaging changes, and CSF biomarkers along the AD continuum (Aim 3) to better understand the pathophysiology underlying AD onset and progression. A separate data request has been made to the NACC database.

The purpose of this study is 1) to better understand genetic protective and risk factors underlying Alzheimer's disease, and 2) to improve individual level disease predictions with artificial intelligence techniques. We seek to identify AD protective genetic features through screening the disease phenotypes of loss-of-function mutation carriers in subjects along the AD spectrum. We also would like to take advantages of the deep neural network in capturing high-order meaningful features from large, high-dimensional, and co-linear feature spaces. Therefore, we will utilize/adapt the well-established deep convolutional neural network classifier with genetic features to predict individual-level disease status. These results will help us to better identify subjects at risks of AD, and improve the accuracy and efficacy for personalized diagnosis, treatment, and prevention.

The objective of this project is to find common regulatory and rare functional genetic variants involved in both aging and Alzheimer’s disease (AD), and suggest new genetic targets for AD prevention. We plan to: (i) evaluate collective effects of genetic interactions using newly developed in our group Interaction Polygenic Risk Score (IPRS), allowing to integrate the additive and interaction effects of genes on AD and aging traits, which presents significant methodological advantage; (ii) leverage the whole exome sequencing data (WXS), to find rare functional variants associated with aging and AD; (iii) focus on genetic regulators of translation that influence levels of proteins and provide connection between genes and phenotypes; and (iv) explore biological pathways involved in aging and AD. For this, we will conduct only secondary analyses of existing genetic and phenotypic data collected in the Alzheimer's Disease Sequencing Project (ADSP), as well as in other studies, including Framingham Cohort (a.k.a., Framingham Heart Study (FHS)), Cardiovascular Health Study (CHS), Alzheimer’s Disease Neuroimaging Initiative (ADNI), and UK Biobank. Current request refers to the ADSP. The analyses will be performed using relevant statistical methods and software. The project does not involve any contact with or participation of the real subjects.

The objective of this project is to significantly improve our understanding of the heterogeneity of Alzheimer’ s disease (AD) and common genetic mechanisms in aging and AD, and find new genetic targets for AD prevention, with emphasis on regulatory and rare functional variants involved in both aging and AD. This objective will be addressed by conducting secondary analyses of existing human data collected in existing human studies, containing genetic and phenotypic information on thousands of individuals.

Alzheimer’s disease (AD) is a highly heterogeneous disease with diverse clinical manifestations and rate of progression. AD subtypes identification will improve our understanding of the underlying disease mechanisms, enable us to predict disease trajectory and develop new disease-modifying treatments.Objective. To identify distinct AD subtypes from WGS data of AD individualsAnalysis plan. We will use 3000 WGS data derived from the ADSP Discovery Case-Control Based Extension Study. We will use the available SNVs and INDELS and infer structural variants (SVs) with our in-house multi-caller pipelines. Rare variants will be retained for further analysis. We will then split the dataset in training and tests set, and use the identified set of genetic variants (i.e. SNVs, INDELS and SVs) as input to a deep neural network (an autoencoder architecture) to learn an unsupervised latent representation of the data. AD subtypes will be identified within this reduced space and characterized using, demographics and clinical data. We will then contrast each subtype with the control groups to identify subtype relevant variants (i.e. putative subtype biomarkers), which will be used as input features to a gradient boosted tree model, to generate a subtype predictive model and subtype specific features.Planned collaboration. Each member of the team will devote effort in specific areas of investigation, nevertheless, all the team members will discuss, through regular meeting, individual progress and potential challenges. In particular, Dr Coppola (Research Scientist, Department of Pathology, Yale University, USA), together with Dr Dean Palejev (Associate Professor, GATE Institute, Sofia University, Bulgaria) will be involved in the deep learning model generation and validation, and subtype identification; Dr Fredrik Johansson (Assistant Professor, Department of Computer Science & Engineering, Chalmers University of Technology. Sweden), will work on the supervised machine learning model; Dr Alexander Schliep, Associate Professor, Department of Computer Science & Engineering, University of Gothenburg, Sweden), will work on the SVs inference.

Alzheimer’s disease (AD) is a highly heterogeneous disease with diverse clinical manifestations and rate of progression. The heterogeneity of AD has complicated both clinical trial design and outcomes, and thus the need for better models of AD, and/or better strategies for selection of participants into specic clinical trials is evident. The identication of more homogeneous disease subgroups (i.e. AD subtypes) will improve our understanding of the underlying disease mechanisms, enable us to predict disease trajectory and develop new disease-modifying treatments. We will use a comprehensive set of genetic variants in combination with deep learning algorithms to identify AD subtypes. Subtypes will be characterized using clinical and demographic data. Finally, variants specic to each cluster will be identied and used to train a predictive machine-learning model to classify new individuals.

We have been developing more powerful statistical methods to detect common variant (CV)- or rare variant (RV)-complex trait associations and/or putative causal relationships for GWAS and DNA sequencing data. Here we propose applying our new methods, along with other suitable existing methods, to the existing ADSP sequencing data and other AD GWAS data provided by NIA, hence requesting approval for accessing the ADSP sequencing and other related GWAS/genetic data. We have the following two specific Aims: Aim 1. Association testing using the ADSP data. We'd like to detect CV- and RV-AD associations based on the ADSP data. Aim 2. Association testing under genetic heterogeneity: For complex traits, genetic heterogeneity, especially of RVs, is ubiquitous as well acknowledged in the literature, however there is barely any existing methodology to explicitly account for genetic heterogeneity in association analysis of RVs based on a single sample/cohort. We propose using secondary and other omic data, such as transcriptomic or metabolomic data, to stratify the given sample, then apply a weighted test to the resulting strata, explicitly accounting for genetic heterogeneity that causal RVs may be different (with varying effect sizes) across unknown and hidden subpopulations. Some preliminary analyses have confirmed power gains of the proposed approach over the standard analysis. Aim 3. Meta analysis of RV tests: Although it has been well appreciated that it is necessary to account for varying association effect sizes and directions in meta analysis of RVs for multi-ethnic cohorts, existing tests are not highly adaptive to varying association patterns across the cohorts and across the RVs, leading to power loss. We propose a highly adaptive test based on a family of SPU tests, which cover many existing meta-analysis tests as special cases. Our preliminary results demonstrated possibly substantial power gains.

We propose applying our newly developed statistical analysis methods, along with other suitable existing methods, to the existing ADSP sequencing data and other AD GWAS data to detect common or rare genetic variants associated with Alzheimer’s disease (AD). The novelty and power of our new methods are in two aspects: first, we consider and account for possible genetic heterogeneity with several subcategories of AD; second, we apply powerful meta-analysis methods to combine the association analyses across multiple subcategories of AD. The proposed research is feasible, promising and potentially significant to AD research. In addition, our proposed analyses of the existing large amount of ADSP sequencing data and other AD GWAS data with our developed new methods are novel, powerful and cost-effective.

Late-onset Alzheimer’s Disease (LOAD) affects over 46 million people worldwide and is expected to double by 2030 and triple by 2050. Accumulating evidence supports a strong genetic component underpinning AD etiology. However, genetic studies of AD have focused primarily on assessing the impact of common variants (either in the APOE epsilon 4 allele or traditional polygenic scores) in European populations. Indeed, there is a sparsity of evidence demarcating the role of rare coding variants on LOAD in European and non-European populations alike, which could provide invaluable insight toward the genetic determinants of LOAD since rare variants are collectively numerous and more deleterious relative to common variants. Using the ethnically diverse ADSP WES data, we aim to systematically demarcate the effect of rare variants on LOAD by constructing a rare variant polygenic risk score (rvPRS), which captures the gene-based burden of rare variants across the genome. The Rare Variant EXome CALIBration using External Repositories (RV-EXCALIBER) method will be used to conduct case-control exome-wide association study using gene burden testing to delineate genes that harbour an enrichment of rare, deleterious variants in LOAD cases relative to controls from the genome aggregation database (gnomAD). Gene-based effect sizes will be used to construct an rvPRS in an independent ADSP validation population, which will be used in multivariable logistic regression models to to predict LOAD status after adjusting for age, sex, APOE epsilon-4-allele, and principal components of ancestry. We also aim to assess the transferability for the predictive power of the rvPRS across the European, African, and Hispanic ancestries present within the ADSP.

Late-onset Alzheimer’s Disease (LOAD) affects over 46 million people worldwide and is expected to double by 2030 and triple by 2050. There is evidence supports that genetic factors underlie AD development. To date, genetic studies of primarily focussed on a small subset of all genetic variants that occur commonly in the population. However, it has been shown that variants that are rare can aid in disease prediction and are better at identifying genes that cause disease. Using the ADSP population, we aim to develop a score based on rare variants that can help identify individuals at high risk for AD from various different ethnic backgrounds.

The “Watson Alzheimer’s Genetics Experiment (WAGE) is a collaboration between IBM TJ Watson Research Center, the Center for Genomics of Alzheimer’s Disease (CGAD U54 AG052427) and AD geneticists at the University of Pennsylvania, Indiana University, Columbia University, and Indiana University (Alzheimer’s Disease Neuroimaging Initiative [ADNI]). We plan to analyze whole genome sequence data generated from subjects with Alzheimer's disease (AD) and elderly normal controls obtained from “R3 17K WGS Project Level VCF” and “phenotypes/pedigree for all subjects”. These data were generated by the National Human Genome Institute Large-Scale Sequence Program, the Alzheimer’s Disease Neuroimaging Initiative, and National Institute on Aging funded investigators. The goal of the planned analyses is to identify genes that have alleles that protect against or increase susceptibility to AD. We will examine both single nucleotide variants and structural variants (indels, deletions, insertions, etc). We will use whole genome sequence data for AD cases from the Alzheimer’s Disease Sequence Project. We will use control data from the Alzheimer’s Disease Sequencing Project (ADSP) and ADNI.

We seek to understand what machine learning algorithms can tell us about Alzheimer’s disease, and are applying machine learning algorithms to all the inherited elements that contribute to Alzheimer's disease risk, and characterizing their statistical power to resolve GW significant alleles. To do this we will analyze DNA sequence data from subjects with Alzheimer's disease and elderly subjects who are cognitively normal. The sequence data from these 2 groups will be compared to identify differences that contribute to the risk of developing Alzheimer's disease of that protect against Alzheimer's disease. These DNA differences can be at a single site in the genetic code, or can span multiple sites, changing the copy number of DNA sequences. Both types of genetic variants will be examined.

Recent projects characterizing genomic variation across large numbers of individuals have revealed that somatic mutations driving clonal expansion in hematopoietic cells occur as part of human aging. This phenomenon, is associated with a number of adverse outcomes, including increased mortality, cardiovascular disease risk, and risk of hematological malignancy. The aim of this proposal is to assess what (if any) association clonal hematopoiesis (CH) has with Alzheimer’s disease (AD) or Dementia.We will use the available exome and whole-genome sequencing to look for somatic mutations associated with CH. In general, distinguishing germline mutations from somatic mutations is non-trivial within a single sample. However, somatic and germline variants are expected to differ in their variant allele fraction distributions. Additionally, many somatic mutations associated with CH are thought to cause severe developmental disease when they occur in the germline (e.g., loss of function in DNMT3A is associated with Tatton-Brown-Rahman syndrome). The poor prognosis of affected patients should make germline mutations in these genes rare. After identifying participants with CH, we will use standard statistical methods (e.g. a fisher test) to determine if CH has any association with AD phenotype. We will also look for sex, race, ethnicity, and APOE specific effects.

Recent projects characterizing genomic variation across large numbers of individuals have revealed that somatic mutations driving clonal expansion in hematopoietic cells occur as part of human aging. This phenomenon, is associated with a number of adverse outcomes, including increased mortality, cardiovascular disease risk, and risk of hematological malignancy. The aim of this proposal is to assess what (if any) association clonal hematopoiesis (CH) has with Alzheimer’s disease (AD) or Dementia.

- Objectives of the proposed research: The primary objective of this research is to validate the findings from our analysis of whole genome sequencing data in Koreans by replicating them using the ADSP dataset, particularly focusing on individuals from diverse ethnic backgrounds. Our previous analyses aimed to identify associations between genetic variants and Alzheimer's disease (AD), with a specific focus on early-onset Alzheimer's disease (EOAD), by analyzing whole genome sequencing data and performing both variant-level and gene-level statistical analyses. By leveraging the requested ADSP data, we aim to confirm these associations and improve the accuracy of our findings.- Study design: This research follows a cross-sectional design. We will utilize data from EOAD patients and control groups recruited at Seoul National University Bundang Hospital to perform rare-variant association analyses, focusing on their association with AD diagnosis. Based on the results of our analysis, we will conduct replication studies using the ADSP dataset, which includes participants of various ethnicities and comprehensive phenotypes. This study is part of our larger project, "Diagnosis of Alzheimer’s disease using GWAS, blood transcriptome and blood biomarkers", which aims to integrate genetic findings with transcriptome data and other biomarkers to improve the diagnostic accuracy of AD.- Analysis plan, including the phenotypic characteristics that will be evaluated in association with genetic variants: The analysis using the ADSP dataset will focus on replicating previous findings related to EOAD diagnosis using whole genome sequencing data in Koreans. We will evaluate the association between variants identified in our analysis and various phenotypic characteristics in the ADSP dataset, such as cognitive function and biomarkers, including β-amyloid, phosphorylated tau, and other Alzheimer-related markers measured by different tools.- Collaborations: There are no planned collaborations with researchers at other institutions for this study.

This study aims to better understand the genetic factors associated with early-onset Alzheimer's disease (EOAD), particularly among diverse populations. By analyzing whole genome sequencing data from Koreans, we have identified certain genetic variants linked to EOAD. We will now replicate these findings using data from the Alzheimer's Disease Sequencing Project (ADSP), which includes individuals of various ethnicities. Through this research, we hope to improve the accuracy of Alzheimer's disease diagnosis by identifying genetic risk factors that contribute to the development of the disease, ultimately aiding in early detection and treatment.

Objectives: The aim of our analyses of the Alzheimer’s Disease Sequence Project (ADSP) cohort data is to improve the identification of reliable gene-drug targets to develop effective therapeutic medicines to prevent or slow-down the progression of Alzheimer’s disease in patients. Study design: Our goal is to identify novel and reliable gene-targets to develop effective therapeutic medicines to prevent or slow-down the human burden caused by Alzheimer’s disease. To identify variants that confer risk or provide protection to Alzheimer’s disease it is essential to obtain deep sequencing (whole-genome and whole-exome) in families with high penetrant variants and in case-control populations. We will capitalize on the Discovery Phase and Extension Phase cohorts that includes WGS data in families by performing family-based association analyses. Additionally, we will perform case-control association analyses on the whole-exome sequence data from the Discovery Phase Case/Control, Extension Case/Control, FUS1, and FUS2 cohorts. Analysis Plan: We plan to analyze the WGS and WES data with several phenotype-variant analysis approaches. We will perform common variant, rare variant gene-based (i.e. stop-gain, frameshift, putatively deleterious non-synonymous, and splice-site variants), pathway-based analysis, and sex-stratified analysis. We plan to perform association analysis with a dichotomous outcome (i.e. affected/unaffected) and with neuropathology quantitative measures (where available). We have expertise in several analyses software to perform the proposed analysis; they include PLINK, PLINKSeq, MENDELSCAN, and varianttools. We plan to analyze the ADSP cohort and other Alzheimer's disease cohorts independently (i.e. UKBB) and to combine the summary statistics (Odds ratio and p-values) by meta-analysis. The planned research is consistent with the data use limitations/restrictions for the requested dataset(s), and we promise to follow all regulations within. Our proposed research will support all conditions specified in the Data Use Agreements associated with the study and will not violate relevant privacy or consent policies.

The aim of our analyses of the Alzheimer’s Disease Sequence Project (ADSP) cohort data is to unravel the genetic architecture of AD with an objective to identify reliable gene-drug targets through various family-based and population-based statistical analyses, followed by prioritizing molecular targets and to develop effective therapeutic medicines to prevent or slow-down the progression of Alzheimer’s disease in patients.

Objectives: ADHD persists throughout the lifespan and has been linked with higher risk for MCI and AD dementia based on large epidemiological studies. Recent findings from our group indicated a correlation between higher genetic liability for ADHD and progressive cognitive decline, as well as the development of AD pathophysiology, in cognitively unimpaired older adults with amyloid deposition at baseline. We aim to further explore the association between ADHD and cognitive decline by investigating the role of genetic risk for ADHD in the cognitive profile, as well as amyloid and tau deposition, in individuals diagnosed with AD or MCI. For that, we will calculate the ADHD-PRS, which is a valid biomarker of ADHD pathology, in individuals that are part of the Alzheimer’s Disease Sequencing Project. Study design: This study will analyze cross-sectional and longitudinal data from cohorts included in the Alzheimer’s Disease Sequencing Project. Analysis plan: We will calculate ADHD-PRS based on the latest GWAS. We will use linear and mixed regression models to test the association between ADHD-PRS and cognitive function (executive function, language, memory, and visuospatial) in cognitively unimpaired, MCI, and AD individuals. Analyses will be controlled for age, sex, and ancestry. Linear and mixed effect models will be used to evaluate the association between ADHD-PRS and fluid biomarkers (amyloid and phosphorylated tau), as well as neuropathology markers of amyloid deposition and neurofibrillary degeneration. We will conduct sensitivity analysis to explore the confounding effects of education, vascular risk factors (using clinical data or post-mortem markers of vascular brain injury), and psychiatric comorbidities (by calculating PRS of major depressive disorder, bipolar disorder, autism spectrum disorder, and schizophrenia). We hypothesized that higher ADHD-PRS will be associated with longitudinal decline in memory and executive function, as well as higher markers of tau pathology. Based on prior results from our group, we also hypothesize that these findings will be observed in individuals with amyloid deposition at baseline.

ADHD is a common neurodevelopmental disorder that persists throughout the lifespan. Recent large epidemiological studies have indicated an increased risk for AD and MCI among individuals with ADHD. The underlying mechanisms linking ADHD and cognitive decline remain unclear, but prior data published by our group supports that individuals with ADHD have reduced resilience to amyloid pathology, leading to a decline in cognition at lower pathological levels. The main goal of this project is to further investigate the association between ADHD and cognitive decline, tau, and amyloid deposition in individuals diagnosed with AD or MCI. For that, we plan to utilize GWAS data from the Alzheimer’s Disease Sequencing Project to calculate ADHD polygenic risk scores (ADHD-PRS), which is a validated marker of ADHD pathology. We will evaluate the association between ADHD-PRS and cognitive function, as well as fluid biomarkers of amyloid and tau pathology. As secondary goals, we aim to investigate the confounding role of the genetics risk for other psychiatric disorders in these associations.

The purpose of our study is to identify novel genetic factors associated with Alzheimer’s Disease, corticobasal degeneration (CBD) and progressive supranuclear palsy (PSP). This includes identifying genetic factors associated with the risk of these conditions, as well as genetic risk factors associated with age-at-onset (AAO) for these conditions. We will also evaluate genetic associations with sub-phenotypes individuals have within these broad disease categories, such as their Braak staging results which provide insights into the level of severity of Alzheimer’s. Thus we are requesting access to the set of genomic Whole Exome and Whole Genome Sequences (WES and WGS) have just been released through the National Institute on Aging Genetics of Alzheimer’s Disease Data Storage Site (DSS NIAGADS). The findings from our genetic association testing have the potential for identification of new therapeutic targets for Alzheimer's Disease, CBD, and PSP. The findings from our studies also have the potential for identification of genetic and phenotypic biomarkers that will be beneficial for subsetting patients in new ways. We will use standard genetic epidemiological methods to handle the WGS and WES data. We will also analyze cell type-specific expression differences in AD to identify biomarkers and disease pathways using standard gene expression analysis methods currently in use. We will also use other multi-omic and other genetic data that has now become available to further understand genetic association results we have found in AD.All data will remain anonymized and securely stored, and only those listed on our application and their staff will have access to these data. We will not share any of the individual level data outside of Genentech nor beyond the researchers on our application. We will adhere to all data use agreement stipulations through the DSS NIAGADS. We have a secure computational environment called Rosalind within Genentech where we will use these data. We have IT security staff that constantly monitor all our research computing, assuring safety and privacy of all of our stored data. We will not collaborate with researchers at other institutions.

Genetic variation and gene expression data allows us to understand more of the genetic contribution to risk and protection from diseases such as Alzheimer’s and dementia. This information also allows us to identify important biological contributors to disease for developing effective treatment strategies, and identifying groups of individuals that would benefit most from new treatments. Our exploration of this relationship between genotype, disease traits, gene expression, and outcomes, through these datasets will allow us to pursue important new findings for disease treatment.

We plan to analyze GWAS, whole exome and whole genome sequence data generated from subjects with Alzheimer's disease (AD) and elderly normal controls. The goal of the planned analyses is to identify genes that have alleles that protect against or increase susceptibility to AD. We will evaluate variants detected in the sequence data for association with AD to identify protective and susceptibility alleles using the whole exome and whole genome case-control data. We will also evaluate sequence data from multiplex AD families to identify variants associated with AD risk and protection, and evaluate variant co-segregation with AD. The family data will be whole genome data. The family-based data will be used to inform the cases control analysis and visa versa. We also will focus on structural variants (insertion-deletions, copy number variants, and chromosomal rearrangements). Evaluation of structural variants will involve both whole genome and whole exome data. Structural variants will be analyzed with single nucelotide variants detected and analyzed in the case-control and family-based data

We are attempting to identify all the inherited elements that contribute to Alzheimer's disease risk. To do this we will analyze DNA sequence data from subjects with Alzheimer's disease and elderly subjects who are cognitively normal. The sequence data from these 2 groups will be compared to identify differences that contribute to the risk of developing Alzheimer's disease of that protect against Alzheimer's disease. These DNA differences can be at a single site in the genetic code, or can span multiple sites, changing the copy number of DNA sequences. Both types of genetic variants will be examined.

Objectives: To develop and validate methods for assessing how genomic foundation models generalize to neuropsychiatric conditions, specifically Alzheimer's disease and related dementias (ADRD). We aim to: - Evaluate foundation models' ability to capture disease-relevant genomic patterns - Develop fine-tuning approaches to improve model performance - Create benchmarks for assessing model generalization to psychiatric genetics - Establish reproducible pipelines for model evaluation and optimizationStudy Design: This secondary data analysis study will utilize NIAGADS datasets to: - Extract embedding representations from pre-trained genomic foundation models - Analyze model performance across different cohorts and phenotypes - Develop and test fine-tuning methods - Create standardized evaluation metricsAnalysis Plan: Phase 1: Data Preparation - Harmonize genomic and phenotypic data across cohorts - Format data for compatibility with foundation models - Implement quality control measures - Create validation datasetsPhase 2: Model Evaluation Phenotypic characteristics to be evaluated: - Clinical diagnosis status (AD/ADRD vs. control) - Cognitive performance metrics - Fluid biomarker levels - Neuropathological findings - Cardiovascular risk factors - Neuroimaging metrics (DTI, FLAIR, PET, T1) Analysis methods: - Embedding space analysis - Layer-wise representation analysis - Linear separability testing - Correlation analysis - Target decoding performance - Distribution matching assessmentPhase 3: Fine-tuning Development - Implement distribution matching techniques - Develop data augmentation strategies - Create iterative optimization procedures - Establish performance benchmarksAll data access and analysis will be conducted under approved institutional protocols with appropriate data security measures in place.

Our research aims to improve how artificial intelligence (AI) systems understand genetic information related to Alzheimer's disease and dementia. Using existing, anonymized genetic data from research participants, we will test and enhance AI systems' ability to identify genetic patterns associated with these conditions.Just as AI has been trained to understand language and images, we will develop methods to help AI better process genetic information. The goal is to create tools that help researchers better understand the genetic factors contributing to Alzheimer's disease, potentially supporting future advances in early detection and treatment.

Objective of the proposed research Generate an AD polygenic risk score (PRS) using neuropathologically confirmed AD cases and controls (ND). The PRS will be used to assess the risk of AD in other neurological and psychiatric disease and to identify potential risk factors in large population studies accounting for the genetic risk.Study Design We will select the neuropathologically confirmed AD and ND samples, and genotypes will be used to conduct a genome-wide association study (GWAS), which results will be used to estimate the PRS.Analysis plan VCF files will undergo quality controls with the following inclusion criteria: QUAL >30, DP >10, and GQ > 20. We will exclude variants with allelic balance for heterozygous <0.3 or >0.7. We will include only SNPs with genotype rate >95%, MAF>0.05, Hardy-Weinberg Equilibrium p > 1.0E-06, and sample genotype rate >95%. Sex mismatch will be conducted using the SNPs located in the X chromosome. Duplicated samples will be identified using the pi-hat metrics, removing samples with pi-hat >0.7. Outliers will be identified by the Identity By State statistics. The GWAS will be conducted using BOLT-LMM (PMID:29892013), adjusting for the top 10 PCs and for sex. PRS will be estimated using the methods described by Marees et al. (PMID:29484742) and from the PR-Sice2 software (PMID:31307061). A wide range of neurological and psychiatric disease will be used as target datasets to assess the risk of AD. The PRS will be also used to investigate AD risk factors accounting for the genetic risk. For the method by Marees et al., we will select the significant SNPs (p<5.0E-08), and the number of alleles present in each individual in the target dataset will be counted and then multiplied by their corresponding GWAS p-value and finally summed across all SNPs for each individual's PRS. The method implemented in PR-Sice2 used a wide range of p-values optimizing the cutoff according to the accuracy prediction in the target phenotype. This project will be conducted in collaboration with Drs. Eric Reiman Valentina Ghisays (Banner Alzheimer's Institute, Phoenix, AZ). Only the PI (Dr. Piras) will have access to the data.

The research aims to develop a polygenic risk score (PRS) for Alzheimer's disease (AD) using neuropathologically confirmed AD cases and controls. This PRS will help assess AD risk in other neurological and psychiatric conditions and identify potential risk factors, considering genetic predisposition. The study will use Whole Genome Sequencing (WGS) and microarray data from AD and non-AD (ND) samples. A genome-wide association study (GWAS) will generate the necessary genetic data for PRS calculation. Quality control and filtering will be applied to ensure the accuracy of the genetic data. The PRS will be calculated using genome-wide significant SNPs and analyzed with PR-Sice2 software. The score will be applied to various neurological and psychiatric datasets to evaluate the risk of AD and explore potential risk factors.

Historically, young onset Alzheimer’s disease (YOAD) is defined as onset of clinical symptoms before the age of 65, and 90% of such patients are not associated with mutations in the main 3 Ab related genes (APP, PSEN1 and PSEN2). The objective of this project is to deeply characterize known and new genetic components of YOAD in the largest pathologically confirmed cohort in the world and to evaluate the impact of SNVs, SVs and repeat expansions. Due to their earlier onset age, and strong heritability, we hypothesize that YOAD patients are enriched in rare pathogenic variants within the Ab and Tau pathways. In addition, patients with YOAD are often misdiagnosed for frontotemporal dementia due to clinical symptom overlap. In that context, we also hypothesize that there is genetic overlap between both diseases. We generate whole-genome sequencing data from over 900 YOAD patients, including more than 400 autopsy confirmed YOAD cases, over 1000 FTD patients and 800 controls. We are requesting access to the ADSP whole-genome sequencing data (raw and VCF) to i) to perform gene-based, single variant and pathway association analyses in ADSP YOAD and late onset data to replicate our findings, ii) increase our YOAD cohort size for assessing the overlap and differences between FTD and YOAD patients. Single nucleotide variants, as well as structural variants, will be assessed. To do so, we will use already generated SNVs VCF but also generate structural variant calling using our Mayo pipeline. We will utilize several commonly used software programs, such as Plink-seq and SKAT package, to perform our association analyses. All analyses will be done at the single variant, gene, structural variant, and pathway levels. Using these approaches, we hope to identify novel mutations/genes/pathways that are related to both AD and FTD and will benefit the larger scientific community working on neurodegenerative disorders.

We aim at identifying new risk factors for young onset Alzheimer’s disease (age at onset before 65). To do so we use deep phenotyping and genetic approaches. Upon completion of our work, we will obtain a comprehensive understanding of young onset Alzheimer’s disease genetics. Altogether our project will identify potential new therapeutic targets for young onset Alzheimer’s disease and will pave the way for individualized therapy development not only for young onset Alzheimer’s disease, but also for the more common late onset AD.

We hypothesize that patients with AD and FTD share common genetic factors. We generated whole-genome sequencing data from over 500 FTD patients and 800 controls and identified new genetic variants associated with FTD. Our findings add to the growing body of genetic factors associated with both diseases (TREM2, MAPT, GRN, and C9ORF72). We have now extended our cohort of whole-genome sequenced FTD and controls. We are requesting access to the ADSP whole-genome sequencing data (raw and VCF) to i) use the ADSP controls in order to increase our statistical power to detect risk variants in FTD by increasing the size of our control set, and ii) to perform gene-based, single variant and pathway association analyses in ADSP data with candidate FTD/ALS genes to determine their impact across dementias. Single nucleotide variants, as well as structural variants, will be assessed. To do so, we will call variants combining our FTD data with ADSP data starting from the raw files and using the same pipeline as the one already used for our FTD genome data. ADSP raw files will be processed through the Mayo Clinic Genome-GPS (GGPS) analytic pipeline and the ANNOVAR variant annotation pipeline. In addition, we will also run the ADSP’s data with specific pipelines more accurate for complex genomic regions such as the HLA risk region, e.g. HISAT2. We will also use new genotyping technologies such as long-read sequencing to impute structural variants in the ADSP dataset and compare them to our findings in our FTD cohort. We will utilize several commonly used software programs, such as Plink-seq and SKAT package, to perform our association analyses. Control data will be used to perform a large association study on FTD patients and controls. Then, association studies within the ADSP dataset will be performed. All analyses will be done at the single variant, gene, structural variant, and pathway levels. Using these approaches, we hope to identify novel mutations/genes/pathways that are related to both AD and FTD and will benefit the larger scientific community working on neurodegenerative disorders.

Frontotemporal dementia (FTD) is the second most common form of early-onset dementia after Alzheimer’s disease (AD). While AD patients initially present with memory problems, FTD patients usually present with changes in personality and behavior and sometimes language problems. However, clinical and genetic overlaps between both diseases have been reported. We propose to perform in-depth genetic studies to identify genetic factors that cause or increase the risk for FTD and comprehensively assess the genetic overlap between AD and FTD. The identification of new disease genes will provide novel insight into the disease biology; may improve genetic counseling in patients, and could provide new targets for future therapeutic interventions.

We are requesting data from the NIAGDS portal for utilization in our project concerning genomic and multiomic analysis of Alzheimer's disease in diverse populations. This is relevant to funded NIA grant R01AG075884 as well as pending NIA applications. We will use deidentified data from NIAGDS, as well as from other data sources including the TOPMed consortium and newly generated data from cohorts such as Jackson Heart Study, to identify putative risk factors and biological mechanisms for identified genetic loci for Alzheimer's disease risk. We will utilize statistical and genetic analysis methods (including polygenic risk score construction, colocalization using GWAS summary statistics, and association analysis for gene transcripts, proteins, and metabolites) to identify putative risk factors for Alzheimer's disease.

Black Americans face a disproportionately high risk of developing cognitive impairment and Alzheimer’s disease and related dementias (ADRD), compared to non-Hispanic White adults, but the biological mediators underlying this elevated risk are not well understood. Additionally, most efforts to identify risk biomarkers have not included diverse populations, making results less relevant to all Americans. High throughput multi-omics data from blood samples in diverse participants, including through newly funded grants and existing funded data from NIA, may allow us to identify predictors of ADRD and incident cognitive impairment risk across diverse US populations, including in Black adults underrepresented in ADRD research. Integration of genetic data may allow us to further clarify the genes, proteins, and metabolites through which Alzheimer’s disease genetic risk variants function, as well as improve risk prediction in populations with substantial non-European ancestry.

Our overarching objective is to apply machine learning techniques to predict and interpret the functional effects of genetic variants including Single Nucleotide Variants (SNVs), indels and Structural Variants (SVs) from AD WGS data at the levels of DNA regulation and RNA processing, and link these effects directly to pathways and network context. We will leverage WGS generated by the ADSP and others together with harmonized endophenotypes and clinical data, multi-omics data from the AMP-AD, functional genomics data from Roadmap Epigenomics, PsychENCODE and GTEx Projects, and microglia and monocytes specific transcriptomic and single-cell RNA-seq data sets. Our central hypothesis is that many AD-associated genetic risk or protective variants influence pre- and post-transcriptional gene regulation, resulting in changes to gene expression and cellular pathways/networks, and ultimately contribute to protein aggregation in AD. The objective of this aim is to leverage deep-learning-based models capable of predicting functional effects of genomic variants on pre- and post-transcriptional gene regulation. We will train existing and novel sequence-based deep learning models of epigenomic state and RNA regulation and processing specific to AD-relevant cell types and states. in silico mutagenesis under these trained models will be used to calculate functional impact “delta scores” for every SNV, indel and structural variants (SV) detected from AD WGS. We will use these delta scores to empower non-coding rare variant tests of association with AD at the regulatory region, gene and pathway levels. We will conduct functional fine-mapping through the integration of (i) the CNN delta scores (ii-iii) expression and splicing quantitative trait loci (eQTL and sQTL), (iv) AD endophenotypes and (v) multi-ethnic AD WGS data. We will use probabilistic ML methods, combined with cell-type-specific and single-cell RNA-seq datasets, to build gene regulatory networks. This NIH funded project is a close collaboration with Dr. David Knowles at the New York Genome Center/ Columbia University.

Despite decades of research and enormous investment, no disease-modifying treatment is available for Alzheimer’s disease (AD). Combining population-scale data collection, human genetics and machine learning provide a way forward to uncover and characterize new causal cellular processes involved in AD. Effectively integrating diverse genomic data to better understand AD represents a substantial computational challenge, both in terms of data scale and analysis complexity. We will train machine learning models to predict epigenomic signals from the genomic sequences to estimate the functional impact of any genetic variant. These analyses will highlight variants and genes involved in AD. However, genes do not operate in a vacuum so robust machine learning will be used to learn cell-type and disease-specific networks. Such pathways will be prime candidates for future functional and therapeutic studies of AD.

Endosomal dysfunction is implicated in the pathogenesis of Alzheimer’s disease (AD). Variants in SORL1 encoding a receptor for the retromer complex (the master regulator of endosomal trafficking) has been strongly linked to AD. Disrupting the retromer-SORLA pathway mediated regulation of endosomal trafficking triggers AD’s molecular/cellular pathologies in various ways: i) disrupting the retromer-SORLA complex reduces glutamate receptor recycling to the surface of dendritic spines, triggering the early stages of the neurodegenerative process; ii) with SORLA as the intermediary, retromer traffics APP away from its amyloidogenic cleavage in early endosomes. Retromer core or SORLA depletion causes endosomal traffic jam and thereby enhancement of APP processing to Aß; iii) dysfunction of the retromer trafficking pathway, by depleting either SORLA or VPS26b, accelerates tau secretion; iv) a recent study showed that retromer dysfunction in neurons triggers morphological alterations that phenocopy microglia abnormalities observed in the AD brain, and that the microglial pathology can be partially rescued by neuronal retromer gene therapy. These studies provide a strong rationale to investigate the retromer-SORLA pathway as a functional druggable target to slow the course of AD by restoring endosomal function. A critical step for this is to identify individuals carrying pathogenic variants in SORL1 and other endosomal trafficking genes and characterize their (endo)phenotypic profiles including cognitive clinical course and AD biomarker profiles. The goal of this proposal is to determine which SORL1 variants and variants in other endosomal trafficking genes are truly pathogenic and may be valuable drug development candidates by identifying variant carriers in ADSP WGS data and assessing critical endophenotypes in these individuals including cognitive profiles and biofluidic biomarker profiles.

Intracellular trafficking of critical proteins is an important mechanistic pathway in the pathogenesis of Alzheimer’s disease (AD). A master regulator of intracellular endosomal trafficking is the retromer complex. An increasing number of studies indicate that the retromer may be a valuable functional druggable target to slow the course of AD, and a critical step for this is to identify individuals carrying pathogenic variants in retromer-related genes and characterize their (endo)phenotypic profiles including cognitive clinical course and AD biomarker profiles. The goal of this proposal is to determine which genetic variants in endosomal trafficking genes are truly pathogenic and may be valuable drug development candidates by identifying variant carriers in ADSP WGS data and assessing critical endophenotypes in these individuals including cognitive profiles and biofluidic biomarker profiles.

Neuropsychiatric Symptoms (NPS) (e.g. aggression, psychosis, anxiety, apathy, depression, agitation, sleep disturbances, repetitive behaviors) occur in 85% of AD patients, and are associated with greatly increased suffering of patients and families. Despite this, our understanding of the etiology of NPS in AD is inadequate, with treatments for NPS often being ineffective and associated with serious adverse effects. This knowledge gap is particularly egregious in underserved racial and ethnic groups. The aim of the current project (U01AG079850) is to collate, harmonize, and analyze the AD-associated NPS data collected on ADSP/ADSP-FUS samples. We plan to (1) expand the racially and ethnically diverse datasets of the ADSP-FUS and related efforts to include harmonized NPS data, creating the largest and most diverse genomic resource on NPS in AD to date allowing researchers to assess a wide range of additional critical hypotheses through these resources; (2) utilize these harmonized data to identify and describe genetic determinants, pathways, and polygenic effects underlying specific NPS in AD; (3) explore the shared genetic architecture across AD-associated NPS and with primary psychiatric disorders; and (4) disentangle the role of ancestry in NPS genetic risk. Included in these analyses will be in particular early-onset samples already recruited and whole-genome sequenced under the READR , EFIGA and NIA-FBS initiatives which have a particularly high prevalence of NPS. We anticipate that this work will lead to a better understanding of the genetic basis of NPS in AD which is vital to infer the mechanistic pathways underlying these highly disabling symptoms and develop more effective pharmacological targets. To collate NPS data on all ADSPFUS cohorts we closely collaborate with the ADSP-Phenotype Harmonization Consortium. Creation of refined harmonized NPS phenotypes will be conducted by Dr. Ted Huey’s group at Brown University. Genomic data analyses will be conducted by Dr. Reitz’ group at Columbia University and Dr. Beecham’s group at the University of Miami.

Although neuropsychiatric symptoms (e.g. aggression, psychosis, anxiety, apathy, depression, and sleep disturbances) occur in ~85% of Alzheimer disease patients and are associated with accelerated decline, increased cost, out-of-home placement, and greatly increased suffering of patients and families, our understanding of their etiology is still inadequate, with treatments often being ineffective and even associated with serious adverse effects (including increased mortality). This knowledge gap is particularly egregious in underserved racial and ethnic groups such as Hispanics and African-Americans. We propose to expand the racially and ethnically diverse ADSP-FUS and related resources to include harmonized neuropsychiatric symptom data allowing researchers to assess a variety of additional critical hypotheses, and to utilize these harmonized data to identify ancestry-specific genetic determinants, molecular pathways, and polygenic effects underlying neuropsychiatric symptoms in Alzheimer disease.

Objective: This research centers on the investigation of the genetic aspects of Progressive Supranuclear Palsy (PSP) compared to Alzheimer's Disease (AD) and unaffected controls through the analysis of whole genome sequencing data. The objective of this study is to comprehensively analyze the immunogenetic landscape in PSP and AD individuals compared to a control group in this dataset.

In our study, we aim to decipher the genetic underpinnings of Progressive Supranuclear Palsy (PSP) related to immune function. We will use sequencing data to run genomics analyses that are focused on a specific region of DNA called the HLA locus and additional immune genes, which play crucial roles in our immune system. Our goal is to compare the genetic makeup of the HLA locus in people affected by PSP compared to those with Alzheimer’s disease (AD) and the general population. In simple terms, we're looking for clues in the DNA that could explain how changes in the immune system might influence why some individuals develop these conditions.

ObjectivesIn our lab we are working on a variety of Alzheimer’s disease (AD) genetics research projects: mitochondrial genetics of AD, haplotype-based analyses in the nuclear genome, AD genetics in Pacific Islanders (PI), and replication of identified disease variants identified in our novel dataset (the Cache County Study). Our objectives include: 1) develop a large dataset of AD mitochondrial genomes (~60,000 total samples), 2) identify mitochondrial genetic variants associated with AD, 3) explore the relationship between mitochondrial inheritance of AD and the mitochondrial genome, 4) identify haplotypes that explain observed AD variants, 5) replicate variants we identify in samples from the Cache County Study, and 6) test PI-specific AD genetics variants for association in the races represented in the ADSP.Study Design/Analysis Plan1. Using our published approach, assemble, annotate, and deposit in NIAGAD whole mitochondrial genome sequences from the ADSP and a variety of sources. 2. Estimate haplotypes in the nuclear genome using ShapeIt and our own novel tool in development. 3. Using an evolutionary based method, TreeScanning, assess the effects of mitochondrial and nuclear haplotypes associated with AD status, including both risk and protective variation. We have published several papers using this approach to study the relationship between the mitochondrial genome and AD. 4. Also using TreeScanning, conduct association studies between mitochondrial haplotypes and maternal family history of AD. 5. Using standard statistical methods to replicate associations from the Cache County Study and PIs. 6. Use SNP data to estimate genetic variances.

Alzheimer’s disease (AD) is the most common form of dementia, is fatal, and causes a substantial burden to both affected individuals and loved ones. We seek to push the boundaries of our current understanding by identifying novel targets. Where a majority of approaches focus on single variants in the nuclear genome, we seek to identify mitochondrial-based targets and haplotypes responsible for influencing disease risk in the nuclear genome. We are sequencing whole genomes from the Cache County Study in large risk and protective pedigrees and will use these samples to confirm our findings.In addition to the work described above, we are studying the AD genetics of Pacific Islanders (PIs). PIs are unrepresented in the ADSP. We will evaluate the genetics we discover in PIs in the populations represented in the ADSP and augment the ADSP with PI genomes.

Alzheimer's disease (AD) is the most common form of dementia and is characterized by cognitive impairment and progressive neurodegeneration. Genome-wide association studies of AD have identified more than 70 risk loci; however, a major challenge in the field is that the majority of these risk factors are harbored within non-coding regions where their impact on AD pathogenesis has been difficult to establish. Therefore, the molecular basis of AD development and progression remains elusive and, so far, reliable treatments have not been found. The overarching goal of this proposal is to examine and validate AD-related changes on chromatin accessibility and the 3D genome at the single cell level. Based on recent data from our group and others, we hypothesize that genotype-phenotype associations in AD are causally mediated by cell type-specific alterations in the regulatory mechanisms of gene expression. To test our hypothesis, we propose the following Specific Aims: (1) perform multimodal (i.e., within cell) profiling of the chromatin accessibility and transcriptome at the single cell level to identify cell type-specific AD-related changes on the 3D genome; (2) fine-map AD risk loci to identify causal variants, regulatory regions and genes; (3) functionally validate putative causal variants and regulatory sequences using novel approaches that combine massively parallel reporter assays, CRISPR and single cell assays in neurons and microglia derived from induced pluripotent stem cells; and (4) develop and maintain a community workspace that provides for the rapid dissemination and open evaluation of data, analyses, and outcomes. Overall, our multidisciplinary computational and experimental approach will provide a compendium of functionally and causally validated AD risk loci that has the potential to lead to new insights and avenues for therapeutic development.

Alzheimer’s disease (AD) affects half the US population over the age of 85 and despite decades of research, reliable treatments for AD have not been found. The overarching goal of our proposal is to generate multiscale genomics (gene expression and epigenome regulation) data at the single cell level and perform fine mapping to detect and validate causal variants, transcripts and regulatory sequences in AD. The proposed work will bridge the gap in understanding the link among the effects of risk variants on enhancer activity and transcript expression, thus illuminating AD molecular mechanisms and providing new targets for future therapeutic development.

Objectives of the proposed research The objective of the proposed research is to analyze the effect of Klotho-VS heterozygosity and other genetic covariates identified during the study on the rate of Alzheimer's Disease progression as measured by neuropsychological data and brain MRI volumes in the context of the APOE genotype.Study design The design of this study involves the creation of a repository consisting of clinical and genetic data from the participants of the NACC's Uniform Data Set. Only participants who transitioned from a clinical diagnosis of mild cognitive impairment to Alzheimer's Disease will be considered.Analysis plan, including the phenotypic characteristics that will be evaluated in association with genetic variant This study will conduct a secondary analysis of the neuropsychological data (MMSE, MOCA, CDR-SB, ADAS-11) and brain MRI volumes in association with APOE genotype and other genetic predictors of clinical decline like Klotho-VS heterozygosity. Effect of Klotho-VS heterozygosity in the context of a specific APOE genotype will be determined using a linear mixed model approach comparing disease progression over time between genetic variables, as well as demographic variables such as sex and age.There will be no collaboration with researchers at other institutions.

Gene set which controls the rate of Alzheimer's disease (AD) progression both in terms of accelerating and attenuating its rate is unknown. Our recently published work demonstrated feasibility of statistical modeling of longitudinal clinical data from AD patients to uncover the effect of certain genes on the rate of disease progression. Thus, we showed that AD patients who harbor the e4 allele of apolipoprotein E gene show accelerates clinical course of AD in comparison to e4 non-carriers. In this project we plant to use statistical modeling of longitudinal clinical data to correlate trajectory of disease progression in with their individual genetic makeup. A gene which effect on AD progression were are planning to study next is Klotho-VS heterozygosity.

AD is the most common cause of dementia and presents a substantial and increasing economic and social burden. Our ability to diagnose and classify AD from cognitive normals (CN), or discriminate among individuals with AD, early mild cognitive impairment [EMCI], or late mild cognitive impairment (LMCI), is essential for the prevention, diagnosis, and treatment of AD. Since individuals with MCI have a high chance of converting to AD, effectively discriminating between those who convert to AD (MCI-C) from those who do not convert (MCINC) is important for early diagnosis of AD. The heterogeneity of AD has motivated attempts to classify distinct subgroups of AD to better inform the underlying physiology. There is evidence to suggest that using data across multiple modalities (e.g. genetics, imaging, metabolomics) has potential to classify AD subgroups better than using single modality. We will apply machine and deep learning methods to gain deeper insight into AD and ADRD pathobiology. We will use datasets that include genomics, genetics, metabolomics, and phenotypic data for this purpose. Data will be divided into discovery and validation sets. On the discovery set, state-of-the-art ML and DL methods for integrative analysis that we and others have developed will be coupled with resampling techniques to determine candidate molecular signatures and pathways discriminating the AD groups considered. Molecular scores will be developed from these candidate biomarkers. The clinical utility of the scores beyond well-known clinical risk factors for AD will be ascertained. We will validate our findings using the validation data. We will visually and quantitatively compare the risk scores across several clinical variables and outcomes. We will use (un)supervised clustering methods to identify molecular clusters, and we will investigate molecular clusters differentiating MCI to AD converters from non-converters. We may explore differences across ethnic subgroups. We will also innovatively apply our multimodal molecular subtyping methods to discover, reproduce, and characterize novel molecular subgroups of AD– this will allow for better risk stratification.

We have been developing novel machine learning (ML) and deep learning (DL) methods that leverage genomics, other omics (including proteomics and metabolomics), clinical and epidemiology data to better understand the pathogenesis of complex diseases. By integrating data from different sources, we have identified molecular signatures contributing to the risk of the development of complex diseases beyond established risk factors. We are proposing to innovatively apply these, and other existing, methods, to data pertaining to Alzheimer’s disease (AD) and Alzheimer’s disease related dementias (ADRD). A deeper understanding of the genes, genetic pathways, and other molecular signatures of AD is essential and could facilitate the identification of potential therapeutic targets for the disease.

Multiple pathologic changes are common in old age and are associated with dementia. While Alzheimer’s disease neuropathologic changes (ADNC) remain the most common pathology observed in older adults, there is also wide recognition for the role of cerebrovascular neuropathologic changes, Lewy body pathology, TAR DNA-binding protein 43 (TDP-43), hippocampal sclerosis in age-related cognitive decline. These pathologies often do not exist in isolation and individuals harboring multiple types of pathologies are found to have worse clinical outcomes compared to individuals with any one specific pathology. The genetic mechanisms and risk factors of the presence of multiple pathologies however are not well understood. While studies have shown shared risk alleles across various pathologies, it remains unclear how multiple pathologies are linked and why individuals with any one pathology develop comorbid pathologies while others do not. Identification of genetic variants distinct and shared between each pathology will highlight potential mechanistic links and dissect differences across pathologic changes. The goal of the planned analyses is to identify genetic variants associated with comorbid pathologies and to assess the relationships in genomic and RNA-level information with clinical outcomes across multiple pathologies. We will use integrated clinical, neuropathologic, and sequencing data from the ADSP-PHC and the 90+ study to 1) identify new and previously identified risk alleles associated with neuropathologic changes and 2) relate genomic information with transcriptomic data from publicly available RNA sequencing datasets, and 3) determine the associations between the genomic and transcriptomic information to clinical outcomes (cognitive, motor, and neuropsychiatric symptoms) across different pathologic changes. We will perform separate analyses on the ADSP-PHC and 90+ study datasets. All analyses will be adjusted for age at death, sex, and ethnicity.  

With advancing age, the presence of multiple brain pathologies is common and associated with dementia. While Alzheimer’s disease is the most common pathology observed in older adults, other changes like cerebrovascular pathology, Lewy body pathology, TDP-43 pathology, and hippocampal sclerosis also play a role in cognitive decline. People with multiple types of brain changes tend to have worse outcomes. Understanding why some individuals develop multiple pathologies while others don't is still not clear, but genetics may provide some insight into mechanisms underlying and linking multiple pathologies. By studying genetic variants associated with these brain changes and their impact on clinical outcomes, we aim to uncover new insights. Our research, using data from the ADSP-PHC and the 90+ study, will explore these connections and help shed light on the complex interplay between genetics, brain changes, and functional decline.

Alzheimer’s disease (AD) and related genomic data sets including sequencing, GWAS and phenotypic data will be combined with longitudinal clinical, demographic, cognitive, MRI, PET, CSF and blood endophenotype data, where available, to investigate the genetic architecture of Alzheimer’s disease and related disorders (ADRD) and brain aging. The overall goal to gain a better understanding of fundamental disease mechanisms, genetic susceptibility and protective factors, and the relationship of genetic factors to disease heterogeneity, progression and different trajectories across biomarker profiles. Data from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) will be combined with ADSP and other data sets to increase detection power and for replication across samples. Analyses will include conventional statistical association, multivariate profiling of endophenotypes, biological pathway and network approaches, longitudinal models and combinatorial machine learning. Deliverables will include reports of new prioritized lists of candidate genes and variants for further investigation in new samples, functional experiments and in model systems. The ultimate goal is discovery of novel potential diagnostic markers and therapeutic targets that will help provide the foundation for a precision medicine approach to AD/ADRD.

Alzheimer’s disease (AD) and related genomic data sets will be combined with longitudinal clinical, demographic, cognitive, MRI, PET, CSF and blood biomarker data to investigate the genetic architecture of Alzheimer’s disease and related disorders (ADRD) and brain aging. The overall goal to gain a better understanding of fundamental disease mechanisms, genetic susceptibility and protective factors, and the relationship of genetic factors to disease heterogeneity, progression and different trajectories across biomarker profiles. Data will be combined across studies to increase detection power and for replication. Analyses will include conventional statistical association and advanced analytic approaches including multivariate profiling, biological pathway and network analysis and machine learning. The ultimate goal is discovery of novel potential diagnostic and therapeutic markers that will help provide the foundation for a precision medicine approach to AD/ADRD.

AIML Consortium Research Use Statement The AI/ML consortium is a collaboration among multiple teams of investigators who will each submit a DAR to use data from NIAGADS. There will not be a single IRB because they are located at different institutions and they will pursue a unique set of research questions. However, the overarching goals of the program described below are shared among the consortium members. We propose Artificial Intelligence (AI), Machine Learning (ML), and/or Deep Learning (DL) approaches, collectively referred to here as "cognitive systems," that lead to the identification of gene mutations/variants that cause or contribute to the risk of or protection against the development of Alzheimer’s disease (AD) and Alzheimer's disease related dementias (ADRD) via analysis of a variety of genetic, genomic, and phenotypic data (including biomarkers) that are currently available to the research community via NIAGADS. The proposed research includes the design and development of intelligent and innovative algorithms, software, methods, and computational tools to enhance the analysis of AD genetic and genomic data. For some proposed research genetic data will be analyzed in combination with a variety of other types of complex data such as biomarker, imaging, “omic,” epigenetic, and phenotypic data. Relevant technological approaches include those that will facilitate the organization, representation, retrieval, harmonization, analysis, recognition, and classification of biological, genetic, and phenotypic/clinical data that will lead to successful identification of therapeutic targets for AD. In summary, in the research proposed GWAS, whole genome, whole exome, genomic, endophenotypic, clinical, and epidemiological data from ethnically diverse affected and unaffected individuals available from NIAGADS will be harmonized and analyzed using cognitive systems analytical approaches. The proposed research will also include data available from other resources outside of NIAGADS. Analysis of these data will identify new genes and genetics pathways that influence risk and protection for AD and guide the field toward novel therapeutic approaches to the disease.

The AI4AD (Artificial Intelligence for Alzheimer’s Disease) Initiative aims to create and develop advanced AI methods, and apply them to extensive and rich genomic, imaging and cognitive data, in order to 1) identify genotypes and endophenotypes of AD that dissect the heterogeneity of the disease; 2) relate these genotypes and endophenotypes with clinical progression, in pre-dementia patients; 3) identify novel treatment targets for AD, by analyzing whole genome and associated phenotypic data at a previously impossible scale. Collectively, the goals of this highly collaborative multi-site initiative leverage the promise of machine learning to contribute to precision diagnostics, prognostication, and targeted and novel treatments.

The ADSP-PHC was established to harmonize the rich endophenotype data across cohort studies to enable modern genomic analyses of ADRD with the ultimate goal to generate harmonized data that will become a “legacy” dataset perpetually curated and shared NIAGADS. The ADSP-PHC will facilitate phenotypic data harmonization for ADSP participants with genetic and genomic data. This effort represents a multi-disciplinary approach leveraging interdisciplinary expertise in large-scale genetic and genomic studies, clinical neuroscience, neuroimaging, psychometrics, and bioinformatics. This study will utilize a team structure consisting of two coordinating centers to oversee activities of all harmonization teams and to oversee coordination with other ADSP workgroups and initiatives, a Storage and Informatics Core that will oversee the coordination between LONI and NIAGADS for all data storage, compliance, and dissemination, a CHARGE Coordination Core to oversee the alignment of data and protocols with CHARGE workgroups, an Integration & Analytics Core that will enable data integration across phenotypes to facilitate downstream machine learning applications, and eight Domain-Specific Harmonization Teams tasked with harmonization in their area of expertise. The endophenotypes that will be harmonized by this project include Cognition, Fluid Biomarkers, Amyloid PET, Structural MRI, White Matter Hyperintensities, White Matter Integrity, Vascular Risk Factors, and Neuropathology.

The growing availability of endophenotypic data in cohort studies of Alzheimer’s disease and related dementias (ADRD) provides an exciting opportunity to further characterize the genetic architecture of this devastating disease. However, there is a pressing need to develop and apply advanced harmonization approaches to align ADRD endophenotypes across cohorts. The goal of this coordinated national initiative – the AD Sequencing Project Phenotype Harmonization Consortium, or “ADSP-PHC” – is to provide a centralized database of robust endophenotypes for large-scale genomic analyses that will accelerate the identification of novel targets for therapeutic intervention in ADRD. will accelerate the identification of novel targets for therapeutic intervention in ADRD.

This U19 aims to bridge these knowledge gaps for discovery and validation of Centrally-linked Longitudinal pEripheral biomARkers of AD (CLEAR-AD) in multi-ethnic populations. CLEAR-AD U19 is based on the premise that AD is a complex disorder in which many biological pathways are disrupted due to multi-omic perturbations, which can be detected in brain and reflected in blood. The specific aims of CLEAR-AD are: 1) To discover CLPMS of the complex and heterogeneous AD pathophysiology and its co-pathologies. 2) To identify longitudinal CLPMS that detect and predict dynamic neuroimaging, fluid biomarker, and clinical changes across AD spectrum. 3) To characterize differences and similarities in CLPMS profiles across NHW, African American (AA) and Latino American (LA) participants to uncover biomarker patterns in multi-ethnic groups. 4) To make these vast resources available to the scientific community to amplify and accelerate its impact. In this U19, we will leverage NIH-funded ADNI, MCSA and ADRC cohorts of >3,700 multi-ethnic participants to generate >20,000 multi-omics measures (Omics Core) that will be processed and integrated with >48,000 harmonized AD cognitive, neuroimaging and fluid endophenotypes (Analytic Core). Using these data, we will identify brain region and cell-type specific CLPMS, which reflect biological subtypes of AD and disease stage (Project 1). We will discover longitudinal changes in CLPMS that predict cognitive and A/T/N/V progression (Project 2). We will define longitudinal cognitive and A/T/N/V changes and CLPMS in URP that are either conserved with NHW or population-specific (Project 3). This U19 will a) Identify the next generation of AD biomarkers with mechanistic insights; b) Establish a precision medicine approach for rigorous multi-omics biomarker discovery and validation in AD; c) Discover molecules that can serve as biomarkers and therapeutic targets; d) Enhance biomarker research in trial-ready multi-ethnic populations; and e) Generate and share a vast and harmonized resource of endophenotype and multi-omics data in NIH-funded cohorts.

There is a clear and immediate need for the discovery of peripheral molecular signatures linked to central disease processes, core and co-pathologies in Alzheimer’s Disease (AD), that will serve as precision medicine blood-based biomarkers for diagnostic, prognostic, theragnostic and therapeutic purposes. AD is a complex disorder in which many biological pathways are disrupted due to multi-omic perturbations, which can be detected in brain and reflected in blood, i.e. centrally-linked peripheral molecular signatures (CLPMS). This U19 will leverage deeply phenotyped, longitudinal NIH-funded multi-ethnic cohorts and cross-disciplinary expertise for multi-omics data generation and its integration with harmonized AD endophenotypes, will share these data and utilize them in integrated U19 projects to discover CLPMS that will serve as the next generation of AD biomarkers.

We plan to analyze SNP array, whole exome and whole genome sequence data generated from subjects with Alzheimer disease and related disorders (ADRD) and elderly normal controls. The goals of the planned analyses are to identify genes and other functional elements that have variations that protect against or increase susceptibility to ADRD. We will evaluate variants detected in the sequence data for association with ADRD to identify protective and susceptibility alleles using the SNP array, whole exome, and whole genome data. We will also evaluate similar sequence data from multiplex ADRD families to identify variants associated with ADRD risk and protection and evaluate variant co-segregation with ADRD. We also will focus on structural variants (e.g. insertion-deletions, copy number variants, and chromosomal rearrangements, etc.) detected using both whole genome and whole exome data. All data will be analyzed separately and in an integrated fashion and will incorporate additional genetic and functional data. Further, we will examine the variability in genetic effects by genetic ancestry.

We are attempting to identify all the inherited elements that contribute to Alzheimer's disease risk. To do this we will analyze DNA sequence data from subjects with Alzheimer's disease and elderly subjects who are cognitively normal. The sequence data from these 2 groups will be compared to identify differences that contribute to the risk of developing Alzheimer's disease of that protect against Alzheimer's disease. These DNA differences can be at a single site in the genetic code, or can span multiple sites, changing the copy number of DNA sequences. Both types of genetic variants will be examined.

Objective of the proposed research: Recent studies have found that single nucleotide polymorphism (SNPs) and copy number variations (CNVs) can both play significant roles in missing heritability of Alzheimer's Disease (AD). In this project, we propose to conduct a comprehensive investigation on both variant types and understand their contributions in AD risk.Study design: We will us whole-genome (WGS) and whole-exome (WES) sequence data in the Alzheimer's Disease Sequencing Project (ADSP) and conduct case-control association analyses of SNPs and CNVs.Analysis plan: Using the ADSP sequence data, we will start with CNV detection and characterization of CNV sequence features (e.g., microhomology, non-template insertions, and segmental duplications) to understand potential mechanisms of CNV formation. Next, we will study the association of AD status with SNPs and CNVs (common and rare) using standard association methods and adjusting for population regressions to assure efficient modeling of joint SNP-CNV effects from common and rare variants. We will perform ethnic-specific and ethnic-combined association analyses. We will use principle-component-based methods to adjust for PS, but also explore the efficacy of other PS adjustment methods. Finally, we will conduct biological annotation on identified risk variants.Collaborators: This team includes researchers from University of Pennsylvania (UPenn) and North Carolina State University (NCSU). UPenn includes Gerard Schellenberg (PI: Professor of Pathology and Laboratory Medicine), Li-San Wang (PI: Professor of Pathology and Laboratory Medicine, Wan-Ping Lee (PI: Research Assistant Professor of Pathology and Laboratory Medicine), Adam Naj (Assistant Professor of Biostatistics and Epidemiology) and Yuk Yee Leung (Research Assistant Professor of Pathology and Laboratory Medicine). NCSU includes Jung-Ying Tzeng (PI: Professor of Statistics and Bioinformatics Research Center), Wenbin Lu (Professor of Statistics), Arnab Maity (Associate Professor of Statistics) and Jessie Jeng (Associate Professor of Statistics).

Copy number variants (CNVs) are DNA regions that have gains (duplications) or losses (deletions). CNVs affect a considerable number of base pairs in the human genome. Unlike single-nucleotide polymorphisms (SNPs) that has been broadly studied in diseases, CNVs were not intensively discovered. The large-scale Alzheimer’s Disease Sequencing Project (ADSP) provides a systematic way to capture nearly all genomic variations and to study the genetic basis of Alzheimer’s Disease (AD). In this project, using the data of affected and unaffected samples from ADSP, we propose to conduct a comprehensive investigation on both variant types (SNPs and CNVs) and study their contributions in AD risk and etiology. We will start with CNV genotyping, followed by conducting standard association analysis of AD with SNPs and CNVs. We will also develop and apply new analytical methods for efficient modeling of joint SNP-CNV effects from common and rare variants. Finally, we will conduct functional annotation on identified risk variants to uncover possible biological mechanisms.

We plan to analyze whole exome and whole genome sequence data generated from subjects with progressive supranuclear palsy (PSP), corticobasal degeneration (CBD), Alzheimer's disease (AD) and elderly normal controls. The goal is to detect mutations and variants that cause, contribute to risk, or protect against PSP and/or CBD. We want to compare PSP and CBD genotypes to those from AD and normal controls sequenced by the Alzheimer's Disease Sequence Project. We would like both whole genome and whole exome data from the Alzheimer's Disease Sequence Project for AD and normal controls. We would also like whole genome and whole exome data for PSP and CBD generated by the PSP and Tau consortiums. We will use these data to determine which mutations and variants are associated with PSP or CBD versus benign variants. All PSP and CBD subjects being sequenced are deceased. The requested data sets will have variants recalled as a batch and combined to evaluate allele frequencies of called variants. The AD and control variant frequencies will then be compared to allele frequencies from PSP and CBD subjects as described above. We will also compare structural variants (insertion-deletions, copy number variants, and chromosomal rearrangements) identified in PSP and CBD subjects to those found in AD and in cognitively normal controls in order to determine structural variants involved in PSP and CBD pathogenesis. All of the investigators that are listed will be using a joint called VCF generated from the requested data sets. PSP is a neurodegenerative disease closely related to Alzheimer's disease (AD). PSP, CBD and AD have neurofibrillary tangles as part of the signature neuropathology defining these disorders. PSP and CBD are considered Alzheimer’s Disease Related Disorders (ADRD).

We are attempting to identify all the inherited elements that contribute to progressive supranuclear palsy (PSP) and corticobasal degeneration (CBD) risk. To do this we will analyze DNA sequence data from subjects with AD, PSP, CBD, and subjects who are cognitively normal. The sequence data from these groups will be compared to identify differences that contribute to the risk of developing PSP and CBD, or that protect against these diseases. These DNA differences can be at a single site in the genetic code, or can span multiple sites, changing the copy number of DNA sequences. Both types of genetic variants will be examined.

Alzheimer’s disease (AD) is a highly heterogeneous disease with diverse clinical manifestations and rate of progression. AD subtypes identification will improve our understanding of the underlying disease mechanisms, enable us to predict disease trajectory and develop new disease-modifying treatments. Objective. To identify distinct AD subtypes from WGS data of AD individuals Analysis plan. We will use 3000 WGS data derived from the ADSP Discovery Case-Control Based Extension Study. We will use the available SNVs and INDELS and infer structural variants (SVs) with our in-house multi-caller pipelines. Rare variants will be retained for further analysis. We will then split the dataset in training and tests set, and use the identified set of genetic variants (i.e. SNVs, INDELS and SVs) as input to a deep neural network (an autoencoder architecture) to learn an unsupervised latent representation of the data. AD subtypes will be identified within this reduced space and characterized using, demographics and clinical data. We will then contrast each subtype with the control groups to identify subtype relevant variants (i.e. putative subtype biomarkers), which will be used as input features to a gradient boosted tree model, to generate a subtype predictive model and subtype specific features. Planned collaboration. Each member of the team will devote effort in specific areas of investigation, nevertheless, all the team members will discuss, through regular meeting, individual progress and potential challenges. In particular, Dr Coppola (Research Scientist, Department of Pathology, Yale University, USA), together with Dr Dean Palejev (Associate Professor, GATE Institute, Sofia University, Bulgaria) will be involved in the deep learning model generation and validation, and subtype identification; Dr Fredrik Johansson (Assistant Professor, Department of Computer Science & Engineering, Chalmers University of Technology. Sweden), will work on the supervised machine learning model; Dr Alexander Schliep, Associate Professor, Department of Computer Science & Engineering, University of Gothenburg, Sweden), will work on the SVs inference

Alzheimer’s disease (AD) is a highly heterogeneous disease with diverse clinical manifestations and rate of progression. The heterogeneity of AD has complicated both clinical trial design and outcomes, and thus the need for better models of AD, and/or better strategies for selection of participants into specific clinical trials is evident. The identification of more homogeneous disease subgroups (i.e. AD subtypes) will improve our understanding of the underlying disease mechanisms, enable us to predict disease trajectory and develop new disease-modifying treatments. We will use a comprehensive set of genetic variants in combination with deep learning algorithms to identify AD subtypes. Subtypes will be characterized using clinical and demographic data. Finally, variants specific to each cluster will be identified and used to train a predictive machine-learning model to classify new individuals.

Objective: Utilize ADSP data sets to identify genes & specific genetic variants that confer risk for or protection from Alzheimer disease. Aim 1: Using combined WGS/WES across the ADSP Discovery, Disc-Ext, and FUS Phases, including single nucleotide variants, small insertion/deletions, and structural variants. We will: Aim 1a. Perform whole genome single variant and rare variant case/control association analyses of AD using ADSP and other available data; Aim 1b. Target protective variant identification via association analysis using selected controls within the ADSP data and performing meta analysis across association results based on selected controls from non-ADSP data sets. Aim 1c. Perform endophenotype analyses including cognitive function measures, hippocampal volume and circulation beta-amyloid ADSP data in subjects for which these measures are available. Meta analysis will be conducted across ADSP and non-ADSP analysis results. Aim 2: To leverage ethnically-diverse and admixed populations to identify AD variants we will: Aim 2a. Estimate and account for global and local ancestry in all analyses; Aim 2b. Perform admixture mapping in samples of admixed ancestry; and Aim 2c. Perform ethnicity-specific and trans-ethnic meta-analyses. Aim 3: To identify putative therapeutic targets through functional characterization of genes and networks via bioinformatics, integrative ‘omics analyses. We will: Aim 3a. Annotate variants with their functional consequences using bioinformatic tools and publicly available “omics” data. Aim 3b. Prioritize results, group variants with shared function, and identify key genes functionally related to AD via weighted association analyses and network approaches. Analyses will be performed in coordination with the following PIs. Coordination will involve sharing expertise, analysis plans or analysis results. No individual level data will be shared across institutions. Philip De Jager, Columbia University; Eric Boerwinkle & Myriam Fornage, U of Texas Health Science Center, Houston; Sudha Seshadri, U of Texas, San Antonio; Ellen Wijsman, U of Washington. William Salerno, Baylor College of Medicine

This proposal seeks to analyze existing genetic sequencing data generated as part of the Alzheimer’s Disease Sequencing Project (ADSP) including the ADSP Follow-up Study (FUS) with the goal of identifying genes and specific changes within those genes that either confer risk for Alzheimer’s Disease or provide protection from Alzheimer’s Disease. Analytic challenges include analysis of whole genome sequencing data, appropriately accounting for population structure across European ancestry, Hispanic, and African American participants, and interpreting results in the context of other genomic data available.

The principal goal of this study is to identify novel genetic signatures associations with subtypes of Alzheimer’s disease (AD). Specifically, we aim to (a) identify subtypes of AD using phenotypic variables including age-of-onset, sex, years of education, clinical measurements, and neurocognitive measurements, and (b) identify novel genetic signatures that include rare variants, APOE status and/or polygenic risk scores (PRS), which are associated with these subtypes. Current knowledge from literature and databases will be utilized to perform feature engineering such as calculation of PRS and genomic-region-based-bins for rare variant burden. We will employ unsupervised learning such as community detection and clustering algorithms to identify subtypes of AD, and supervised learning such as decision tree and regularization algorithms to find genetic signatures that are associated with the subtypes. The identified genetic signatures will be evaluated using appropriate performance metrics for the predictive models used. As more phenotypic variables are made available, the models will be updated to refine the subtypes with better characterization. It is anticipated that the novel genetic signatures will yield insight into the etiologies of the heterogeneous Alzheimer’s disease, and therefore provide opportunities to develop personalized treatments.Initially, we will use the NIAGADS data to generate preliminary data for grant submissions. After receiving award, the project would utilize the very valuable dataset fully.

Alzheimer’s disease (AD) is a heterogeneous condition and is high heritable (58-79% heritability for late-onset and >90% heritability for early-onset). We have yet to identify large proportion of genetic variants that either increase or decrease risk for different subtypes of AD such as early-onset or late-onset. In the proposed study, we plan to identify genetic signatures for the different subtypes of AD. Unraveling the heterogeneity of AD and its associated genetic signatures is critical for implementation of precision medicine to combat such a devastating disease.

Variation in tandem repeats (TRs), particularly large expansions of triplet repeats (eg. polyCAG), are known to cause a number of late-onset neurological diseases. Due to their repetitive and degenerate nature, variations in TRs are typically ignored by standard genome analysis pipelines. Furthermore, pathogenic repeat expansions typically span hundreds to thousands of bases, making variations in them difficult to detect in short read data. However, recently a number of specialized algorithms have been developed which enable expansions of short TRs (motif sizes 1-12bp) to be detected in short-read sequencing data. Our lab has also developed approaches that allow the copy number of repeats with larger motifs (motif size ranging from 12bp-200kb) to be estimated based on read depth. We hypothesize that variation in TR regions contributes to risk of AD, and will test two hypotheses:1. That rare pathogenic expansions of TRs (either in the “full” or “pre-mutation” range) occur at increased frequency in AD patients compared to controls. 2. That length variation in TRs of all sizes represents a class of common genetic variation that may alter an individual’s susceptibility to AD.In Aim 1, we will search for expansions of microsatellite repeats using tools such as ExpansionHunter, exSTRa and STRetch, that analyze WGS BAM files for signatures of expansion. We will look for loci with an excess of rare outlier genotypes in cases vs controls. If loci showing rare expansions in cases are identified, if possible, we would request aliquots of DNA from the specific individuals to perform long-read sequencing to validate the presence of potentially pathogenic repeat expansions. In Aim 2, we will use read depth approaches to estimate copy number of large TRs. We will compare estimated copy numbers of these repetitive regions in cases vs. controls to identify TR loci that show significant associations of copy number with AD compared to controls. Analysis will incorporate technical and biological covariates, such as principal components of WGS read depth data, ethnicity from SNV data, gender, etc, and will utilize a multiple testing correction for genome-wide analysis.

Some types of neurodegenerative diseases are known to be caused by a specific type of genetic mutation where a short piece of DNA becomes repeated hundreds of time. Termed “repeat expansions”, these types of mutation can be difficult to find using standard methods scientists use to sequence DNA. We believe that some cases of Alzheimer’s disease may also be caused by repeat expansions. We will apply new analysis approaches that are specifically designed to find these repeat expansions in genome sequencing data, with the aim of finding novel types of genetic mutation that contribute to some cases of Alzheimer’s disease.

The AI4AD (Artificial Intelligence for Alzheimer’s Disease) Initiative aims to create and develop advanced AI methods, and apply them to genomic, imaging and cognitive data, in order to 1) identify AD genotypes and endophenotypes that dissect AD’s heterogeneity; 2) relate said genotypes and endophenotypes with clinical progression in pre-dementia patients; 3) identify novel treatment targets for AD by analyzing whole genome and associated phenotypic data. The goals of this multisite initiative (Paul Thompson, USC; Christos Davatzikos, Li Shen, Penn; Andy Saykin, IU; Heng Huang, Pitt, Paul Crane, UW; Adam Brickman, Columbia; Tim Hohman, Vanderbilt; Guyngah Jun, BU; Duygu Tosun, UCSF; Alexander Zaranek, Curii) leverage the promise of machine learning (ML) to contribute to precision diagnostics, prognostication, and targeted and novel treatments. We will develop ML and deep learning methods to apply to large scale biobanks of whole genome sequences (WGS), neuroimaging, cognitive, and clinical data, aiming to discover new genomic features that influence biological processes of AD. We will apply methods of genome representation and tiling to WGS repositories to create inputs for AI methods. We will develop novel, interpretable, biological knowledge guided deep learning methods to discover genomic motifs associated with AD, AD risk, and biological processes of AD as defined by NIA-AA criteria. To quantify subtypes and disentangle biological processes of AD, we will apply computational methods to multimodal MRI and amyloid- and tau-sensitive PET to stratify and subtype patient groups; novel imaging genomics methods will detect genomic markers and pathways that modulate the developing pathology as detected in the images, and that predict future clinical decline or resilience. We hypothesize that advanced deep learning methods combined with whole genome data will outperform traditional methods and GWAS for predicting AD onset and progression, and will assist with disease subtyping and discovering treatable targets in the genome. A team will rank and repurpose existing, and identify novel drugs and targets in the genome based on the discovered genetic motifs affecting AD.

The AI4AD (Artificial Intelligence for Alzheimer’s Disease) Initiative aims to create and develop advanced AI methods, and apply them to extensive and rich genomic, imaging and cognitive data, in order to 1) identify genotypes and endophenotypes of AD that dissect the heterogeneity of the disease; 2) relate these genotypes and endophenotypes with clinical progression, in pre-dementia patients; 3) identify novel treatment targets for AD, by analyzing whole genome and associated phenotypic data at a previously impossible scale. Collectively, the goals of this highly collaborative multi-site initiative leverage the promise of machine learning to contribute to precision diagnostics, prognostication, and targeted and novel treatments.

The goal of this project is to develop artificial intelligence (AI) approaches for extracting unforeseen patterns from clinical, genetic, genomic, and imaging data that could lead to ideas for new drug development or drug repurposing. Our proposed AI methods and software will be open-source, user-friendly, and freely available for all to use. Specifically, we will analyze ADSP data sets using three novel informatics methods to tailor our automated machine learning (AutoML) tool to the analysis of Alzheimer’s disease (AD) data. First, we will develop a Multi-Modal Interaction (M2I) feature selection algorithm for identifying genetic interactions that are predictive of AD (AIM 1). Second, we will develop a Knowledge-driven Multi-omics Integration (KMI) algorithm for combining omics features for AI analysis of AD (AIM 2). Third, we will develop a Multidimensional Brain Imaging Omics (MBIO) integration framework for the joint analysis of multisource large-scale data for predicting AD. Finally, we will integrate all three biomedical informatics methods into our open-source AutoML software package and apply it to the ADSP data sets. We expect our methods will reveal new biomarkers for AD that will open the door for better treatments and clinical decision support.

The goal of this project is to develop artificial intelligence (AI) approaches for extracting unforeseen patterns from clinical, genetic, genomic, and imaging data that could lead to ideas for new drug development or drug repurposing. We will develop three biomedical informatics methods with focuses on genetics, genomics and imaging respectively. We will integrate these methods into our open-source AutoML software package, and apply it to the ADSP data sets. We expect our methods will reveal new biomarkers for AD that will open the door for better treatments and clinical decision support.

We aim to evaluate the potential modification of AD risk and manifestation of AD phenotypes through varying patterns of admixture, local ancestry, and global ancestry proportions within and between Latíno populations. We aim to develop an ancestry informed risk model and report how consideration of local and global ancestry may influence estimated effects of known AD risk factors to influence AD risk and phenotypes. We will leverage this information to build improved PRS and compare local ancestry aware methods to other common methods for improving multi-ancestry portability of PRS scores. Specifically our study aims to classify: 1. The risk of AD conferred by the local ancestry of genetic risk factors in admixed populations 2. Methods and computational pipeline for establishing an ancestry informed risk score for AD in admixed populations from the US and Latin America 3. How patterns of admixture across a genome influence AD risk and modify associations with common co-morbidities or AD risk in response to social determinants of health or heath and other non-genetic AD risk factors. 4. How local ancestry informed PRS compare to other common methods for PRS derivation in admixed populations (PRS-CSx for example, as well as PolyPred-S, SBayesRC, etc, and other to-be published methods) including PRS methods which leverage ADRD-relevant functional annotation and QTL resources, such as for transcriptome based risk scoring Since we will be using multiple data sets and harmonization of those data might not be the most statistically appropriate route to answer the specific questions our project seeks to address, we will in some cases perform a meta-analysis. To complete a meta- analysis specific to each objective of this proposal, we will employ a meta-analysis framework that has utility in both multiethnic and ethnic-specific analysis. Additionally, where appropriate, we will use linear mixed effects models and linear regression models to measure associations/correlations of the variables we will examine in our investigation (e.g. associations/correlations of longitudinal or cross-sectional global cognition with gene expression).

AD disparities between populations are well documented and are influenced by both social determinants of health and the genetic architecture specific ethnic groups. Our research aims to build a comprehensive model that elucidates how AD risk is modified in response of varying patterns of admixture within and between Latino populations. This study will report how ancestry differentiated and specific genetic loci (in terms of both variant frequency and effect size) modify AD risk, if differential patterns of admixture influence AD risk in the context of global ancestry proportions, and whether the expressivity of AD phenotypes is modified by admixture patterns across the genome in Latíno populations. Furthermore, this study will determine whether non-genetic risk factors along in combination with environmental risk factors that may be specific to sub-populations of Latínos mediate risk for AD in response to differential patterns of admixture.

The goal of this project is to utilize standard genetics tools and ensemble/deep learning methods to predict/classify the etiological aspects of Alzheimer's disease and other neurodegenerative diseases based on genetic data and genomic data (including individual level data e.g. genotype and sequencing data, transcriptomic, and epigenomics data, and also by the use of summary statistics). Our primary phenotypes of interest include case:control status, age at onset, survival time (in terms of disease duration from diagnosis to loss to follow-up) and related biomarker data, although there may be other phenotypes of interest that are derived later based on available data.

We are attempting to identify and predict risk of Alzheimer's disease and other neurodegenerative diseases based on genetic and genomic data using standard tools and advanced machine-learning methods.

Alzheimer’s disease (AD) is a devastating multifactorial neurodegenerative disorder caused by interactions among multiple genetic and environmental factors. A major challenge of studying AD pathogenesis and developing and testing new drugs is the disease heterogeneity in both clinical phenotype and the underlying pathophysiology. Sex differences both play a significant role in disease risk and are a major source of disease heterogeneity in AD. Although the sex differences in the risk of AD, vulnerability to genetic load and severity of AD pathology burden have been well established, the molecular underpinnings and pathways that are differentially mediated in male and female AD patients are still poorly understood. The goal of this project is to analyze publicly available, large-scale genomic datasets of AD patients and age-matched controls to identify genomic regions that are associated with AD differentially in male and female patients and examine their interactive effects with apoE genotypes.

We would like to leverage GWAS and other sequencing efforts in AD to identify sex specific markers associated with the disease.

Objectives of the proposed research: Our research aims to investigate the role of genetic variants in Alzheimer's Disease (AD). We seek to identify and understand genetic factors that may contribute differently to AD risk in males and females. We also aim to identify the genetic variants for a certain cohort of AD patients receiving drug treatment.Study design: We will utilize data from the Alzheimer's Disease Sequencing Project (ADSP) resource to perform rigorous genetic analyses. This includes genome-wide association studies (GWAS) to identify sex-specific and cohort-specific genetic variants associated with AD. We will also employ bioinformatics and functional genomics approaches to annotate and characterize these variants, exploring their potential impact on gene function and regulation.Analysis plan: We will collect the available genetic and phenotype data to perform association analyses (e.g., logistic regression) to identify genetic variants associated with AD within each sex/cohort. Multiple testing correction using appropriate methods (e.g., Bonferroni correction or false discovery rate control) will be performed. We will annotate identified genetic variants with functional information (e.g., functional impact on genes, pathways, regulatory elements). Enrichment analysis will be conducted to understand the biological processes affected by sex-specific or cohort-specific variants.Planned collaboration at other institutions: Name of collaborator: Adam Naj; Institution: University of Pennsylvania This collaboration involves joint efforts to achieve shared goals, with specific objectives, tasks, and outcomes in mind. Such collaborations may involve the exchange of expertise, resources, or personnel to leverage collective strengths and achieve more impactful results than would be possible independently.

Our study aims to uncover if and how Alzheimer's Disease affects sex-specific and drug-treated sub-cohorts differently. We'll analyze genomics data to find specific genetic factors associated with the disease in each sub-cohort. Then, we'll explore how these factors work and their impact on Alzheimer's development. Our goal is to provide insights that could lead to more personalized approaches for Alzheimer's diagnosis and treatment, ultimately improving our understanding of this disease.

The data will undergo meta-analysis, combining it with another extensive whole-genome sequenced Alzheimer's project encompassing participants from Iceland, Denmark, Norway, Sweden, and the UK. Our analysis will specifically concentrate on the identification of rare sequence variants with significant impact, utilizing burden analysis and investigating loci associated with risk through a recessive mode of inheritance. By pooling these datasets, we hope to gain further insights into the genetic architecture of Alzheimer's disease, shedding light on rare variants and potential recessive risk factors. Our main focus will be on genetics of Alzheimer's disease (G30* + F00*).

The data will undergo meta-analysis, combining it with another extensive whole-genome sequenced Alzheimer's project encompassing participants from Iceland, Denmark, Norway, Sweden, and the UK. Our analysis will specifically concentrate on the identification of rare sequence variants with significant impact, utilizing burden analysis and investigating loci associated with risk through a recessive mode of inheritance. By pooling these datasets, we hope to gain further insights into the genetic architecture of Alzheimer's disease, shedding light on rare variants and potential recessive risk factors.

Alzheimer’s disease (AD) has a strong genetic component, and several studies have identified genetic variants that influence AD. A majority of those variants are common variants that appear frequently in a population, and studies have also found that those variants do not explain all of genetic basis of AD. This finding has led genetic studies to investigate effect of rare variants that may have larger effect than that of common variants. To better understand effect of rare variants on AD, we aim to ADSP whole-genome sequencing (WGS) and whole-exome datasets. We will determine whether rare variants in genes appear more frequently among AD patients than controls. Our lab has developed several statistical approaches for the rare variant association method (both case/control and family), and we will apply these methods to the ADSP dataset. Through this analysis, we will quantify the effect of rare variants on AD. We will also estimate polygenic risk scores (PRS) of individuals with AD and compare them to those without AD. We will check how much phenotypic variance of AD is explained by PRS.

Alzheimer’s disease (AD) has a strong genetic component, and although several studies found several genetic variations associated with AD, they do not explain all of genetic basis of AD. Those genetic variations are mostly common in population, and recent studies have shown that rare genetic variations may also influence AD. In this study, we propose to identify rare genetic variations that are associated with AD by applying the statistical approaches that combine effect of multiple rare variants. Our lab has developed several methods to identify effect of rare variants both among unrelated individuals and among family members. We will apply these methods to the ADSP dataset and find rare variants associated with AD. In addition to rare variants, we will also investigate effect of common variants using a method called polygenic risk score. These analyses in this study will elucidate impact of both common and rare variants in AD.

• Objectives of the proposed research: To identify genetic correlates of psychotic symptoms, defined as the occurrence of delusions or hallucinations, in individuals with Alzheimer Disease (AD+Psychosis, AD+P)• Study design: Individuals who were analyzed in our genome-wide meta-analysis of psychosis in AD (https://www.medrxiv.org/content/10.1101/2020.08.07.20139261v1 and Mol Psychiatry, in revision) who have available whole exome or whole genome sequence data will be included. We will analyze the association of psychosis with the presence of missense, stop/gain/truncating, and canonical splice site mutations with the presence of psychosis at all loci in which our GWAS identified a suggestive (p<10-6) or a significant (p<5x10-8) association with psychosis.• Analysis plan: Analyses as described above will be conducted in individuals of European ancestry using Fisher’s exact test and the SKAT family of tests as appropriate. Psychosis presence/absence will be as defined in https://www.medrxiv.org/content/10.1101/2020.08.07.20139261v1. Covariates (sex, CDR score, age) will be included as appropriate.• If applicable, a brief description of any planned collaboration with researchers at other institutions, including the name of the collaborator(s) and their institutions(s). N/A

Individuals who develop psychotic symptoms such as delusions or hallucinations during Alzheimer disease (AD) have a more rapid deterioration and worse outcomes. We have found that the risk for developing psychosis during AD is influenced by genetic factors. In this proposed research plan, we build on our prior genome-wide association study (GWAS) of psychosis in the context of AD by asking if promising GWAS signal corresponds to association signal from whole-exome sequencing. To do so, we evaluate the correspondence between our GWAS association signals and those from whole-exome sequencing. We will also conduct an exploratory analysis of individuals with AD, with and without psychosis, for sequence variants that predispose to psychotic symptoms in Alzheimer disease.

Late-onset AD (LOAD) is caused by a complex polygenic and environmental background. Whole genome sequencing provides comprehensive coverage of the genome and has several advantages over exome sequencing and genotyping. We plan to use an aggregated collection of whole genome sequenced family-based and case-control datasets to address the following goals. 1) Identify variants (specifically rare) and regions associated with AD (and related or derived phenotypes) or showing an interaction pattern; 2) Functionally finemap associated loci and identify the functional impact of associated variants in non-coding regions; 3) Use identified variants to validate them in a 3D neural-glial culture model. We will utilize several datasets with whole genome sequencing data, including AD datasets from National Institiute of Mental Health (NIMH) AD family sample and Alzheimer’s Disease Sequencing Project. We will use family-based association tests robust to population confounding and other approaches suitable for case-control studies. Novel analysis approaches will be developed and tested. Analysis results and derived data will be made available to the research community.

Alzheimer’s disease (AD) is the most common neurodegenerative disorder with a huge burden on the healthcare system and the 6th leading cause of death in the United States. Sequenced DNA from people will help us to better describe the genetic architecture of AD. We will utilize two types of genomic datasets: genomes from related individuals (family-based) and genomes from unrelated individuals (case/control). Identified functional variants will be validated in a 3D neural-glial culture model and enhance the biological understanding of AD.

Population-based genetic association studies have identified over 30 risk loci for Alzheimer’s disease (AD). However, such approaches do not directly reveal the true causal variants or unfold the functional mechanisms of the risk variants. To bridge these gaps, we need to investigate the functional impacts of genetic variants on molecular traits (e.g., mRNA, proteins, and epigenetic modifications) in disease-relevant tissues (e.g. brains) and cell types. We hypothesize that many of the SNPs influence multiple clinical and molecular features. By integrating the genetic associations with functional quantitative trait loci (QTLs), we aim to investigate the potential cascading causal effect of genetic variations in multiple layers of omics data in AD.In this proposal, we plan to perform a multi-omic quantitative trait locus (xQTL) analyses to RNA-seq, proteomics, and DNA methylation data from the large number of postmortem brain tissues available from the AMP-AD project. This study will be part of NIH/NIA Alzheimer's Disease Sequencing Project (ADSP) Functional Genomics xQTL Consortium, a joint effort to generate a reference map of Alzheimer's-related quantitative loci (QTLs). Specifically, our team will be responsible for the xQTL analysis in the Mount Sinai Brain Bank (MSBB) cohort, which contains whole-genome sequencing (WGS), RNA-seq gene expression, proteomics, and DNA methylation data from over 300 AD and control brains. We will follow the unified xQTL calling pipelines developed by the xQTL Consortium to predict QTLs and conduct the subsequent fine mapping, causal inference, and functional annotation integration. We will validate the identified xQTLs in independent samples from the ROSMAP cohort. Through the AMP-AD portal, we already have access to the MSBB and ROSMAP sample meta data, RNA-seq raw read files, proteomics and DNA methylation data. However, the WGS genotype data in the AMP-AD portal were called based on the human hg19 genome, inconsistent with the hg38 genome-based genotype data for other studies in the xQTL consortium. Thus, we are applying for the access to the MSBB and ROSMAP WGS hg38 genome-based genotype data available in NIAGADS.

Numerous genetic loci have been identified to associate with when Alzheimer's disease (AD) occurs and how it progresses. However, the widely used genome-wide association studies (GWAS)-based approaches do not directly reveal the true causal variants or unfold the functional mechanisms of the risk variants. To bridge these gaps, we plan to study the functional impacts of genetic variants on molecular traits (e.g., mRNA, proteins, and epigenetic modifications) in postmortem brain tissues of large AD cohorts, including Mount Sinai Brain Bank and Religious Orders Study/Memory and Aging Project (ROSMAP). By integrating the genetic association signals with the functional quantitative trait loci, we aim to investigate the potential cascading causal effect of genetic variations in multiple layers of omics data in AD.

The AI4AD (Artificial Intelligence for Alzheimer’s Disease) Initiative aims to create and develop advanced AI methods, and apply them to genomic, imaging and cognitive data, in order to 1) identify AD genotypes and endophenotypes that dissect AD’s heterogeneity; 2) relate said genotypes and endophenotypes with clinical progression in pre-dementia patients; 3) identify novel treatment targets for AD by analyzing whole genome and associated phenotypic data. The goals of this multisite initiative (Paul Thompson, USC; Christos Davatzikos, Li Shen, Penn; Andy Saykin, IU; Heng Huang, Pitt, Paul Crane, UW; Adam Brickman, Columbia; Tim Hohman, Vanderbilt; Guyngah Jun, BU; Duygu Tosun, UCSF; Alexander Zaranek, Curii) leverage the promise of machine learning (ML) to contribute to precision diagnostics, prognostication, and targeted and novel treatments. We will develop ML and deep learning methods to apply to large scale biobanks of whole genome sequences (WGS), neuroimaging, cognitive, and clinical data, aiming to discover new genomic features that influence biological processes of AD. We will apply methods of genome representation and tiling to WGS repositories to create inputs for AI methods. We will develop novel, interpretable, biological knowledge guided deep learning methods to discover genomic motifs associated with AD, AD risk, and biological processes of AD as defined by NIA-AA criteria. To quantify subtypes and disentangle biological processes of AD, we will apply computational methods to multimodal MRI and amyloid- and tau-sensitive PET to stratify and subtype patient groups; novel imaging genomics methods will detect genomic markers and pathways that modulate the developing pathology as detected in the images, and that predict future clinical decline or resilience. We hypothesize that advanced deep learning methods combined with whole genome data will outperform traditional methods and GWAS for predicting AD onset and progression, and will assist with disease subtyping and discovering treatable targets in the genome. A team will rank and repurpose existing, and identify novel drugs and targets in the genome based on the discovered genetic motifs affecting AD.

The AI4AD (Artificial Intelligence for Alzheimer’s Disease) Initiative aims to create and develop advanced AI methods, and apply them to extensive and rich genomic, imaging and cognitive data, in order to 1) identify genotypes and endophenotypes of AD that dissect the heterogeneity of the disease; 2) relate these genotypes and endophenotypes with clinical progression, in pre-dementia patients; 3) identify novel treatment targets for AD, by analyzing whole genome and associated phenotypic data at a previously impossible scale. Collectively, the goals of this highly collaborative multi-site initiative leverage the promise of machine learning to contribute to precision diagnostics, prognostication, and targeted and novel treatments.

The ADSP-PHC was established to harmonize the rich endophenotype data across cohort studies to enable modern genomic analyses of ADRD with the ultimate goal to generate harmonized data that will become a “legacy” dataset perpetually curated and shared NIAGADS. The ADSP-PHC will facilitate phenotypic data harmonization for ADSP participants with genetic and genomic data. This effort represents a multi-disciplinary approach leveraging interdisciplinary expertise in large-scale genetic and genomic studies, clinical neuroscience, neuroimaging, psychometrics, and bioinformatics. This study will utilize a team structure consisting of two coordinating centers to oversee activities of all harmonization teams and to oversee coordination with other ADSP workgroups and initiatives, a Storage and Informatics Core that will oversee the coordination between LONI and NIAGADS for all data storage, compliance, and dissemination, a CHARGE Coordination Core to oversee the alignment of data and protocols with CHARGE workgroups, an Integration & Analytics Core that will enable data integration across phenotypes to facilitate downstream machine learning applications, and eight Domain-Specific Harmonization Teams tasked with harmonization in their area of expertise. The endophenotypes that will be harmonized by this project include Cognition, Fluid Biomarkers, Amyloid PET, Structural MRI, White Matter Hyperintensities, White Matter Integrity, Vascular Risk Factors, and Neuropathology. Site PIs on this project include: Jesse Mez (Boston University), Adam Brickman (Columbia University), Andy Saykin (Indiana University), Elizabeth Mormino (Stanford University), Pauline Maillard (UC Davis), Duygu Tosun-Turgut (UC San Francisco), Christos Davatzikos (University of Pennsylvania), Arthur Toga (USC); Mohamad Habes (University of Texas Health Science Center at San Antonio), Michael Cuccaro (University of Miami), Paul Crane (University of Washington), Bennett Landman (Vanderbilt University), Timothy Hohman (Vanderbilt University Medical Center), and Carlos Cruchaga (Washington University in St. Louis).

The growing availability of endophenotypic data in cohort studies of Alzheimer’s disease and related dementias (ADRD) provides an exciting opportunity to further characterize the genetic architecture of this devastating disease. However, there is a pressing need to develop and apply advanced harmonization approaches to align ADRD endophenotypes across cohorts. The goal of this coordinated national initiative – the AD Sequencing Project Phenotype Harmonization Consortium, or “ADSP-PHC” – is to provide a centralized database of robust endophenotypes for large-scale genomic analyses that will accelerate the identification of novel targets for therapeutic intervention in ADRD. will accelerate the identification of novel targets for therapeutic intervention in ADRD.

Our goal is an improved understanding of the genetic contributors to cognitive resilience in neurodegenerative disease (ND) and normal aging, and the development of machine-learning (ML) models for the prediction of expected cognition and its decline. Our specific aims are: I: Identify novel genetic factors for cognitive resilience through the development of genetically informed, ML models of cognition. We will define a novel measure of residual cognition using ML to identify novel genetic associations for cognitive resilience. To achieve this aim, we will: (1) Characterize and compare the degree of association between polygenic risk scores (PRS) and measured risk and resilience factors with risk of diagnosis, cognition, and cognitive decline across NDs. (2) Develop integrated, multi-PRS predictive models of cognition in normal aging, which will be further refined to account for neuropathology in AD, PD, and ADRSs. Comparison of these models will highlight differences in factors driving cognition across ND conditions. (3) Conduct genome-wide and rare variant association studies of cognitive resilience, as quantified by a novel continuous residual cognition measure defined in (2). II: Identify and characterize genetic signatures of cognitive resilience in healthy aging. Similarly, we will characterize if and how ND, cognitive reserve, and brain structure PRS associated with cognition in healthy aging - in cohorts including ADSP and Wellderly. To identify novel genetic contributors, we will identify individuals with outlying genetic signatures of risk to cognitive decline – and perform association studies and analyses as in I.3. III: Genetically informed dissection of the causal contributions of cognitive / brain reserve in cognitive resilience. We will dissect the causal relationship between cognitive resilience, educational attainment, and neuroimaging features. Using polygenic Mendelian randomization analysis, educational attainment, and brain structure PRS as instrument variables, and measured educational attainment and brain imaging as exposures, we can dissect the relationship between cognitive reserve, brain reserve, and cognitive decline in ND.

Cognitive decline occurs during normal brain aging and during neurodegeneration. Understanding shared genetic factors that protect from cognitive decline in aging and neurodegeneration may help inform strategies to stop its decline. The focus of this proposal is to identify those genetic factors that protect against cognitive decline. In particular, we aim to identify novel genetic factors involved in cognitive resilience through associative, predictive, and causal analysis strategies. The overarching goal is an improved understanding of the genetic contributors to cognitive resilience in neurodegenerative disease and normal aging, and the development of machine-learning (ML) models for the prediction of expected cognition and its decline, which will have downstream utility in clinical trial design, clinical decision-making, and have the potential to reveal determinants of therapeutic response in neurodegenerative diseases.

The AI4AD (Artificial Intelligence for Alzheimer’s Disease) Initiative aims to create and develop advanced AI methods, and apply them to genomic, imaging and cognitive data, in order to 1) identify AD genotypes and endophenotypes that dissect AD’s heterogeneity; 2) relate said genotypes and endophenotypes with clinical progression in pre-dementia patients; 3) identify novel treatment targets for AD by analyzing whole genome and associated phenotypic data. The goals of this multisite initiative (Paul Thompson, USC; Christos Davatzikos, Li Shen, Penn; Andy Saykin, IU; Heng Huang, Pitt, Paul Crane, UW; Adam Brickman, Columbia; Tim Hohman, Vanderbilt; Guyngah Jun, BU; Duygu Tosun, UCSF; Alexander Zaranek, Curii) leverage the promise of machine learning (ML) to contribute to precision diagnostics, prognostication, and targeted and novel treatments. We will develop ML and deep learning methods to apply to large scale biobanks of whole genome sequences (WGS), neuroimaging, cognitive, and clinical data, aiming to discover new genomic features that influence biological processes of AD. We will apply methods of genome representation and ‘tiling’ to WGS repositories to create inputs for AI methods. We will develop novel, interpretable, biological knowledge guided deep learning methods to discover genomic motifs associated with AD, AD risk, and biological processes of AD as defined by NIA-AA criteria. To quantify subtypes and disentangle biological processes of AD, we will apply computational methods to multimodal MRI and amyloid- and tau-sensitive PET to stratify and subtype patient groups; novel imaging genomics methods will detect genomic markers and pathways that modulate the developing pathology as detected in the images, and that predict future clinical decline or resilience. We hypothesize that advanced deep learning methods combined with whole genome data will outperform traditional methods and GWAS for predicting AD onset and progression, and will assist with disease subtyping and discovering treatable targets in the genome. A team will rank and repurpose existing, and identify novel drugs and targets in the genome based on the discovered genetic motifs affecting AD.

The AI4AD (Artificial Intelligence for Alzheimer’s Disease) Initiative aims to create and develop advanced AI methods, and apply them to extensive and rich genomic, imaging and cognitive data, in order to 1) identify genotypes and endophenotypes of AD that dissect the heterogeneity of the disease; 2) relate these genotypes and endophenotypes with clinical progression, in pre-dementia patients; 3) identify novel treatment targets for AD, by analyzing whole genome and associated phenotypic data at a previously impossible scale. Collectively, the goals of this highly collaborative multi-site initiative leverage the promise of machine learning to contribute to precision diagnostics, prognostication, and targeted and novel treatments.

The ADSP-PHC was established to harmonize the rich endophenotype data across cohort studies to enable modern genomic analyses of ADRD with the ultimate goal to generate harmonized data that will become a “legacy” dataset perpetually curated and shared NIAGADS. The ADSP-PHC will facilitate phenotypic data harmonization for ADSP participants with genetic and genomic data. This effort represents a multi-disciplinary approach leveraging interdisciplinary expertise in large-scale genetic and genomic studies, clinical neuroscience, neuroimaging, psychometrics, and bioinformatics. This study will utilize a team structure consisting of two coordinating centers to oversee activities of all harmonization teams and to oversee coordination with other ADSP workgroups andinitiatives, a Storage and Informatics Core that will oversee the coordination between LONI and NIAGADS for all data storage, compliance, and dissemination, a CHARGE Coordination Core to oversee the alignment of data and protocols with CHARGE workgroups, an Integration & Analytics Core that will enable data integration across phenotypes to facilitate downstream machine learning applications, and eight Domain-Specific Harmonization Teams tasked with harmonization in their area of expertise. The endophenotypes that will be harmonized by this project include Cognition, Fluid Biomarkers, Amyloid PET, Structural MRI, White Matter Hyperintensities, White Matter Integrity, Vascular Risk Factors, and Neuropathology. Site PIs on this project include: Jesse Mez (Boston University), Adam Brickman (Columbia University), Andy Saykin (Indiana University), Elizabeth Mormino (Stanford University), Pauline Maillard (UC Davis), Duygu Tosun-Turgut (UC San Francisco), Christos Davatzikos (University of Pennsylvania), Arthur Toga (USC); Mohamad Habes (University of Texas Health Science Center at San Antonio), Michael Cuccaro (University of Miami), Paul Crane (University of Washington), Bennett Landman (Vanderbilt University), Timothy Hohman (Vanderbilt University Medical Center), and Carlos Cruchaga (Washington University in St. Louis).

The growing availability of endophenotypic data in cohort studies of Alzheimer’s disease and related dementias (ADRD) provides an exciting opportunity to further characterize the genetic architecture of this devastating disease. However, there is a pressing need to develop and apply advanced harmonization approaches to align ADRD endophenotypes across cohorts. The goal of this coordinated national initiative –the AD Sequencing Project Phenotype Harmonization Consortium, or “ADSP-PHC” –is to provide a centralized database of robust endophenotypes for large-scale genomic analyses that will accelerate the identification of novel targets for therapeutic intervention in ADRD. willaccelerate the identification of novel targets for therapeutic intervention in ADRD.

Objective of the proposed research: Recent studies have found that single nucleotide polymorphism (SNPs) and copy number variations (CNVs) can both play significant roles in missing heritability of Alzheimer’s Disease (AD). In this project, we propose to conduct a comprehensive investigation on both variant types and understand their contributions in AD risk. Study design: We will use the whole-genome (WGS) and whole-exome (WES) sequence data in the Alzheimer's Disease Sequencing Project (ADSP) and conduct case-control association analyses of SNPs and CNVs. Analysis plan: Using the ADSD sequence data, we will start with CNV detection and characterization of CNV sequence features (e.g., microhomology, non-template insertions, and segmental duplications) to understand potential mechanisms of CNV formation. Next, we will study the association of AD status with SNPs and CNVs (common and rare) using standard association methods and adjusting for population structure (PS) and ages of onset. We will also develop and apply new methods using kernel and regularized regressions to assure efficient modeling of joint SNP-CNV effects from common and rare variants. We will perform ethnic-specific and ethnic-combined association analyses. We will use principle-component-based methods to adjust for PS, but also explore the efficacy of other PS adjustment methods. Finally, we will conduct biological annotation on identified risk variants. Collaborators: The team includes researchers from University of Pennsylvania (UPenn) and North Carolina State University (NCSU). UPenn researchers include Gerard Schellenberg (PI: Professor of Pathology and Laboratory Medicine), Li-San Wang (PI: Professor of Pathology and Laboratory Medicine), Wan-Ping Lee (PI: Research Assistant Professor of Pathology and Laboratory Medicine), Adam Naj (Assistant Professor of Biostatistics and Epidemiology) and Yuk Yee Leung (Research Assistant Professor of Pathology and Laboratory Medicine).

Copy number variants (CNVs) are DNA regions that have gains (duplications) or losses (deletions). CNVs affect a considerable number of base pairs in the human genome. Unlike single-nucleotide polymorphisms (SNPs) that has been broadly studied in diseases, CNVs were not intensively discovered. The large-scale Alzheimer’s Disease Sequencing Project (ADSP) provides a systematic way to capture nearly all genomic variations and to study the genetic basis of Alzheimer’s Disease (AD). In this project, using the data of affected and unaffected samples from ADSP, we propose to conduct a comprehensive investigation on both variant types (SNPs and CNVs) and study their contributions in AD risk and etiology. We will start with CNV genotyping, followed by conducting standard association analysis of AD with SNPs and CNVs. We will also develop and apply new analytical methods for efficient modeling of joint SNP-CNV effects from common and rare variants. Finally, we will conduct functional annotation on identified risk variants to uncover possible biological mechanisms.

Our proposed research has two main objectives: 1) identify genetic variants that can influence appropriate splicing of Alzheimer’s disease (AD) related genes and 2) estimate the length of tandem repeats present throughout the genome to identify novel genetic risk factors for AD. We have evaluated RNA sequencing data from brain samples from patients with AD and age-matched controls. Our analysis has revealed several alternative splice products and intronic tandem repeat sequences that are enriched in patients with AD in genes implicated in disease (PSEN1 and PSEN2 and others). Our first objective is to integrate our RNA sequencing results with whole genome sequence data to identify intronic variants that in combination with alternative splicing products may predispose to Alzheimer’s disease (AD). We have identified microhomology to bacterial genomes at some of these alternative splice sites and will test whether the corresponding bacteria can be detected in DNA samples. We will then build on evaluated RNA sequenced data in combination with whole genome sequence data from the NIAGADS database to find novel genomic areas such as tandem repeat expansions that predispose humans to develop AD. We will study AD cases and controls to identify rare variant burden analysis across defined genomic regions. Our analysis plan is as follows: we will extract genomic regions corresponding from CRAM files. We will quantify the presence of small insertions, deletions and variants in cases and controls. We will determine the burden of nucleotide changes in cases, controls and large sequencing databases (e.g. gnomAD) and perform t-tests and Chi-square tests to quantify whether an enrichment of variants are present in patients with AD.

The genetic information that can be ascertained from large scale sequencing projects can enable novel discoveries for genes that can contribute to disease. The primary objective of many of these projects is to detect nucleotide changes that alter the protein encoded by the host gene. However, the same sequencing information can be used to identify non-coding elements of the genome that can contribute to disease, including variants that can influence splicing – the appropriate assembly of exons that are spliced together to form a gene. As we age, certain elements that preserve tight regulation of exon choice lose effectiveness, particularly when faced with injury or various stressors. We wish to detect variants that influence alternative splicing products in the context of relaxed regulation of exon choice and integrate findings with RNA sequencing databases to identify methods to preserve appropriate splicing.

The first objective is to perform an exome-wide burden analysis of variants in all ADSP cases and controls. We will count all alleles per individual across frequencies (AF< 5%, 1%, 0.1%, singleton, ultra-rare variants), functional annotations (eg. PAVs, nonsynonymous, LOF, synonymous, noncoding) and damaging predictions (eg. CADD scores >12.37=damaging). We will stratify cases according to age of onset (early onset, midlife onset, late onset). Then we will run a logistic regression modeling the number of alleles per individual against disease status, including relevant covariates, such as age, sex, population structure (PCA) and sequencing coverage if applicable. The second objective is to perform gene-set burden analysis of the most enriched variants, using gene-list from for example highly constrained genes according to gnomAD (pLI>0.9), the Molecular Signatures database Hallmark and C2 curated gene-sets, and highly expressed genes from 54 specific GTEx tissues, to identify molecular pathways, biological processes and tissue-specific expression patterns enriched. Here we will use the SKAT-O and/or CMC Fisher software to perform the variant enrichment with the same covariates above. The third objective will be to run a gene-wise burden test and perform a protein-protein interaction network along with enrichment in brain single-cell expression data to prioritize significant candidate genes. Here we will map variants to single genes and use SKAT-O/CMC Fisher. Then we will take all genes with uncorrected P< 0.05 and run a protein-protein interaction network with WebgestaltR, using the network-topology analysis and random walk algorithm, and Gene-Ontology enrichment of the resulting network using BIOGRID. We will use STRING in order to get network interaction significance. As a fourth objective, we aim to expand the previous analysis calling short-tandem repeats, copy number variants and other complex structural variants using software such as gangSTR, expansionhunter, MegaHit (de novo alignment) etc. We would also aim to investigate the burden and single variants association of specific candidate genes for experimental follow-ups.

We will assess the load of rare variants in the ADSP Case Control NGS samples across the whole-exome, whole-genome, biological pathways and genes in a “hypothesis-free” approach, leveraging state-of-the art variant annotation tools and databases. We aim to detect the most enriched type of variants in cases that increase risk, including different ages of onset. This approach will help us to increase our power to reveal novel biological pathways and genes associated with AD, expanding the understanding of rare variation and their implication on disease risk.

We propose a multifaceted approach to elucidate and interpret genetic risk factors for Alzheimer's disease. First, we propose to perform a whole-genome sequencing meta-analysis of the Alzheimer's Disease Sequencing Project with the UK Biobank and All of Us to associate rare coding and non-coding variants with Alzheimer's disease and related dementias. We will explore a variety of case definitions in the UK Biobank and All of Us, including those based on ICD codes from electronic medical records (inpatient, primary care and/or death), self-report of Alzheimer's disease or Alzheimer's disease and related dementias, and/or family history of Alzheimer's disease or Alzheimer's disease and related dementias. We will perform single-variant, coding-variant burden, and non-coding variant burden tests using the REGENIE genome-wide association study toolkit.Second, we propose to develop statistical and machine learning models that can effectively infer (“fine-map”) the causal gene(s), variant(s), and cell type(s) underlying each association we find, as well as associations from existing genome-wide association studies and other Alzheimer's- and aging-related cohorts found in NIAGADS. In particular, we propose to improve causal gene identification by incorporating knowledge of gene function as a complement to functional genomics. For instance, we plan to develop improved methods for inferring biological networks, particularly from single-cell data, and integrate these networks with the results of the non-coding associations from our first aim to fine-map causal genes. To fine-map causal variants and cell types, we plan to integrate the associations from our first aim with single-nucleus chromatin accessibility data from postmortem brain cohorts to simultaneously infer which variant(s) are causal for each discovered locus and which cell type(s) they act through.

We have a comprehensive plan to understand and explain the genetic factors that contribute to Alzheimer's disease. Our approach involves two main steps.First, we'll analyze genetic information from large research databases to identify rare genetic changes associated with Alzheimer's and related memory disorders. We'll look at both specific changes in genes and other parts of the genetic code. We'll use data from different studies and combine them to get a clearer picture.Second, we'll create advanced computer models that can help us figure out which specific genes, genetic changes, and cell types are responsible for these associations. This will help us pinpoint the most important factors contributing to Alzheimer's disease. We'll also analyze data from previous studies to build a more complete understanding of these genetic links.

NIAGADS is the data coordinating center for ADSP. This request will allow us to access genotype and phenotype data for all ADSP samples and perform data processing and quality assurance, before distributing to the scientific community.Currently a data deposition plan is being developed by ADSP: 1. Plans for aggregating phenotype, GWAS, and exome chip genotype data are in place, and NIAGADS will work with data contributors to organize and review files before data are distributed to study investigators.2. NIAGADS will work with ADSP investigators to develop a plan for reviewing incoming sequencing data. This will be done in parallel with basic quality assurance procedures by sequencing center partners before data are promoted to archival status and ready for analysis.NIA is in discussion with other similar whole-genome and whole-exome sequencing projects. We plan to harmonize these additional datasets with the ADSP WGS/WES data so the community can combine these datasets for analysis. All associated phenotypes are minimized and there is minimal risk to the participants.

NIAGADS is the data coordinating center for ADSP. This request will allow us to access genotype and phenotype data for all ADSP samples and perform data processing and quality assurance, before distributing to the scientific community.

Sex chromosomes have long been excluded in genetic and genomic research of human diseases, partially due to the technical challenges of analyzing the data from large-scale molecular profiling methods, such as next-generation sequencing (NGS). Yet their contributions are significantly implicated in the sex dimorphism of Alzheimer’s disease (AD). In this project, the team will apply a sex-chromosome-completement aware alignment approach to re-align large amounts of NGS data from large clinical cohorts of AD, for systematically investigating the genetic contributions of sex chromosomes to AD. Variants will be called based on the realignments, and both X and Y chromosomes will be specifically investigated for the variants’ associations with the harmonized AD phenotypes from ADSP-PHC. The team will also consider the contributions of the homologous X-Y gene pairs to gene expression and perform gene network analysis to evaluate how the expressions affect sex differences in AD. The reprocessed data will be disseminated to the public for broad use to benefit the scientific community. Taken together, the work will lay the foundation for determining the genetic components of sex chromosomes to AD and pave the way for future in-depth study of these genomic features and their implication in sexual dimorphism in AD.

Alzheimer’s disease (AD) impacts millions of US populations, especially women, yet the molecular mechanism behind the sex differences remains poorly understood. This study aims to specifically study the genetic contributions of the sex chromosomes to the observed sex disparities in AD, which will build a benchmark of the molecular mechanism driving sex differences in AD and greatly facilitate untangling the genetic contributions from other factors in its etiology.

Objective: Utilize ADSP data sets to identify genes & specific genetic variants that confer risk for or protection from Alzheimer disease. Aim 1: Using combined WGS/WES across the ADSP Discovery, Disc-Ext, and FUS Phases, including single nucleotide variants, small insertion/deletions, and structural variants. We will: Aim 1a. Perform whole genome single variant and rare variant case/control association analyses of AD using ADSP and other available data; Aim 1b. Target protective variant identification via association analysis using selected controls within the ADSP data and performing meta analysis across association results based on selected controls from non-ADSP data sets. Aim 1c. Perform endophenotype analyses including cognitive function measures, hippocampal volume and circulation beta-amyloid ADSP data in subjects for which these measures are available. Meta analysis will be conducted across ADSP and non-ADSP analysis results. Aim 2: To leverage ethnically-diverse and admixed populations to identify AD variants we will: Aim 2a. Estimate and account for global and local ancestry in all analyses; Aim 2b. Perform admixture mapping in samples of admixed ancestry; and Aim 2c. Perform ethnicity-specific and trans-ethnic meta-analyses. Aim 3: To identify putative therapeutic targets through functional characterization of genes and networks via bioinformatics, integrative ‘omics analyses. We will: Aim 3a. Annotate variants with their functional consequences using bioinformatic tools and publicly available “omics” data. Aim 3b. Prioritize results, group variants with shared function, and identify key genes functionally related to AD via weighted association analyses and network approaches. Analyses will be performed in coordination with the following PIs. Coordination will involve sharing expertise, analysis plans or analysis results. No individual level data will be shared across institutions. Philip De Jager, Columbia University; Eric Boerwinkle & Myriam Fornage, U of Texas Health Science Center, Houston; Sudha Seshadri, U of Texas, San Antonio; Ellen Wijsman, U of Washington. William Salerno, Baylor College of Medicine.

This proposal seeks to analyze existing genetic sequencing data generated as part of the Alzheimer’s Disease Sequencing Project (ADSP) including the ADSP Follow-up Study (FUS) with the goal of identifying genes and specific changes within those genes that either confer risk for Alzheimer’s Disease or provide protection from Alzheimer’s Disease. Analytic challenges include analysis of whole genome sequencing data, appropriately accounting for population structure across European ancestry, Hispanic, and African American participants, and interpreting results in the context of other genomic data available.

We aim to uncover new genetic risk variants for Alzheimer’s disease (AD) by analysis of an integrated analysis of proteomics and genetic sequencing performed at Emory University. Results of these analyses will be used to weight analysis of whole-genome sequencing (WGS), whole-genome genotyping (WGG), and whole-exome sequencing (WES) data from dbGaP and ADSP. We plan to publish our findings, so they are shared with the scientific community.Outcomes that will be tested include: (1) clinical disease status, (2) pathologic characterization (e.g., measures of beta-amylodi, tau, etc.), and (3) cognitive decline. For sequencing data, we will perform joint calling from samples previously mapped by ADSP using PECaller using default settings. Variant annotation will be performed using Bystro and quality control will follow Wingo et al., 2017. For rare variants, we will use burden- and variance-based tests to estimate association between genetic variants and each outcome for every gene in the genome. External weights from proteomic analyses will be optionally used, as well as measures of genomic conservation for each site. For common variants, we plan to test for differences in allele frequencies using maximum likelihood tests. For all analyses, we plan to control for population structure deriving principal components from the underlying sequencing or genotyping data.

Our aim is to identify genetic variants that are associated with Alzheimer's Disease (AD) either using genomic data (from dbGap or from Emory University) or brain protein sequencing data (from Emory University) as a starting point. Each center’s data will be analyzed separately, and we will determine whether the findings are consistent among the centers. Additionally, we will use protein data from brain or cerebrospinal fluid of individuals with or without AD to guide the analysis of the genomic data to identify genetic variants that influence AD risk. Our overarching aim is to use genetic discoveries to identify mechanisms of AD pathogenesis and creation of more meaningful models of the disease.

Understanding the neurobiological mechanisms of Alzheimer's disease is crucial for developing potential treatments. However, most research currently focuses on European populations. It is anticipated that by conducting analyses on diverse populations, such as Asians and Africans, new signals not previously identified in existing research may be uncovered.In this study, we aim to discover new signals through large-scale Alzheimer's disease Whole Genome Sequencing (WGS) analysis across different populations. First, we plan to perform single variant and gene burden association analyses specific to each ethinic group to observe how signals may vary among them. We will identify associated risk factors for AD from independent genomics datasets including the Alzheimer's Disease Sequencing Project (ADSP) and other trans-ethnic genomics datasets. We will employ the REGENIE program, widely used in large-scale genomics datasets. Our phenotypes of interest include Alzheimer's disease status and related variables.Secondly, we will conduct causal variant discovery through fine mapping analysis. We will use various brain tissue expression Quantitative Trait Locus (eQTL) datasets like MetaBrain to identify genes influenced by genetic variants. Additionally, we plan to investigate enhancer or candidate Cis-Regulatory Element (CRE) regions using single-cell level ATAC sequencing data or HiC data to determine whether AD associated variants exist within these regions.Through this research, we aim to uncover differences in GWAS signals across different ethnic groups and identify causal variants by leveraging a variety of data sources.

Alzheimer's disease is known to have a significant genetic influence, and this genetic influence varies across different ethnic groups. In this study, we aim to identify which genetic variations contribute to Alzheimer's disease on an ethnic-specific basis. Identifying ethnic-specific genetic variations could assist in developing tailored treatments for each ethnic group. Additionally, we intend to explore various datasets to understand how certain genetic variations influence particular genes, ultimately contributing to the development of Alzheimer's disease. Through these analyses, we aim to enhance our understanding of the mechanisms underlying Alzheimer's disease.

Our primary goal is to evaluate/utilize/develop the most powerful rare variant analysis methods in the discovery of novel cancer predisposition genes. The advance of sequencing technology and the refinement of reference sequence bring us cost-effective sequencing platforms and better accuracies, however, they have also generated very heterogeneous sequencing data sets, such as using different reference builds, different mapping software (versions).Objective: 1) evaluate/develop methods for building a pipeline in harmonizing different available cohorts to reduce false positives due to sequencing artifacts or batch effects; 2) evaluate/develop rare variant methods in analyzing large scale sequencing data, for example, identify which tests are robust and powerful, how to define regions or SNP sets, how to incorporate variant information.Study design: because large sample size is critical to achieve enough power, we will combine the data sets including TCGA, ADSP, as well as our inhouse sequencing data such as pediatric cancer sequence data or Amyotrophic lateral sclerosis (ALS) sequencing data. We need to deal with potential batch effects of different datasets. Then we will do case-control based association test, e.g., using one trait versus other traits to evaluate the enrichment of rare variants in each particular trait. Other public control data sets, such as 1000 Genomes or gnomAD might also help filtering potential pathogenic variants.

For sequenced rare variants, it is very important to develop powerful methods in identifying trait specific predisposition genes or variants, which can incorporate as much prior information as possible and control potential batch effects due to different processing platforms. We will use the TCGA, ADSP and our inhouse pediatric and ALS sequencing dataset for evaluation and development of methods for identifying predisposition variations of each trait, for example, cancer risk genes, or neurodegenerative risk genes.

High-throughput sequence technologies enable us to easily genotype dozens of single nucleotide polymorphisms (SNPs) within any interesting gene. Such genome-wide SNP data are rapidly growing in disease association studies. The association analysis includes single-SNP and haplotype-based disease associations. Haplotype-based association analysis has several advantages over single-SNP association analysis. We propose to develop novel statistical methods for haplotype-based association using the whole-genome sequencing data from the Alzheimer’s Disease Sequencing Project (ADSP) sponsored by the National Institute of Aging. The phenotype is the case-control status. We will use current programs for haplotyping, then use the haplotypes for genetic association studies. We will develop new methods for modeling the recombination interference.

High-throughput sequence technologies enable us to easily genotype dozens of single nucleotide polymorphisms (SNPs) within any interesting gene. Such genome-wide SNP data are rapidly growing in disease association studies. The association analysis includes single-SNP and haplotype-based disease associations. Haplotype-based association analysis has several advantages over single-SNP association analysis. We propose to develop novel statistical methods for haplotype-based association using the whole-genome sequencing data from the Alzheimer’s Disease Sequencing Project (ADSP) sponsored by the National Institute of Aging.

The proposed project aims to derive novel statistical methods to integrate multi-omics data with genome-wide association studies (GWAS) summary data for studying complex phenotypes. First, we will develop novel statistical methods to integrate transcriptomic, proteomic, and DNA methylation molecular trait data with GWAS data, for the goal of identifying risk genes with interpretable biological functions. We will develop novel statistical methods to learn molecular QTL information by using external individual-level transcriptomics data sets such as GTEx V8 and summary-level molecular QTL datasets, and then integrate this information with the whole genome sequence data from ADSP to prioritize risk genes associated with AD-related phenotypes. We are interested in studying all AD-related phenotypes profiled for the ADSP samples, especially Alzheimer’s disease (AD) and AD-related complex phenotypes. Especially, our lab has access to the ROS/MAP multi-omics data shared by the Rush Alzheimer’s Disease Center (http://www.radc.rush.edu/). All samples in the ROS/MAP study are well-characterized with extensive complex phenotypes profiled, including clinical diagnosis of AD, AD-related complex phenotypes, and psychological phenotypes. We will combine the whole genome sequence data from both ADSP and ROS/MAP samples to increase the total sample size in our study, thus improving the study power.Second, we will investigate integrating multi-omics data and polygenic risk scores, with clinical variables to derive biomarkers for Alzheimer's disease. We will also use both ROS/MAP and ADSP data to improve the study power.We plan to use ADSP data as a validation data set to test our derived methods. We are not limited to studying AD only. We are flexible to study any complex phenotypes that are profiled for both ADSP and ROS/MAP samples.

This proposed project is to develop novel statistical methods to integrate multi-omics data such as summary-level molecular QTL data with GWAS summary data to study complex phenotypes, prioritizing potential causal genes. i) We will develop novel statistical methods to leverage summary-level molecular QTL data of multiple cohorts and ancestries. ii) We will apply our developed methods to study AD-related phenotypes. iii) We will derive risk prediction models using multiple modalities of omics, clinical, and image data, for predicting AD dementia. We propose to apply our proposed methods to the applied genomic analysis data and ROS/MAP multi-omics data to study AD-related phenotypes that are profiled for both ADSP and ROS/MAP samples, including AD, AD-related pathology traits, and related psychological disorders.

This goal of this study is to examine the presence of Variable Number Tandem Repeats (VNTRs) in Alzheimer’s disease (AD) population. Within a genome exists short sequences of repeating DNA. While the repeated sequence (usually >6 bases in length) is usually conserved within a population, the number of times the sequence is repeated in any given individual varies. These genetic variants are known as VNTRs and the number of these repeats can be considered a polymorphism, with individuals or families having a different number of repeats than those seen in the general population. In the past VNTRs have been an overlooked polymorphic component of the genome even though a number of VNTRs have been shown to be associated with neurological disorders and brain functions. We aim to understand not only the presence of specific VNTRs associated with Alzheimer’s disease but also how different phenotypes influences the relationship of these VNTRs to AD, this may inform precise genetic profiles which can be applied to a diverse population. As the associative genetics of AD is known not to be conserved across races it is important to not only assess the dataset as a whole but also the association of the ethno-racial phenotypes to inform these precise genetic profiles. If we can use VNTRs to predict this debilitating disorder it opens up avenues to apply treatments earlier and impact one of the most prevalent social and economic burdens on our society. Whole genome sequence data will be analyzed using VNTRseek against a set of reference tandem repeats generated from the tandem repeats database. This software determines the presence of a particular repeat. Each output will contain the number of repeats for each VNTR for each genome and this will be filtered for repeats that have a variation from the reference to determine the presence of a VNTR. This will be used to determine allelic frequency of specific polymorphic repeats comparing controls to AD. We will also further separate the analysis into gender and the different ancestral phenotypes in the ADSP to assess for specific alleles that may be more associated in one ethno-racial group than another.

Within a genome exists short sequences of repeating DNA. While the repeated sequence (usually >6 bases in length) is usually conserved within a population, the number of times the sequence is repeated in any given individual varies. These variants are known as Variable Number Tandem Repeats (VNTRs) and the number of these repeats can be considered a polymorphism. In the past VNTRs have been an overlooked polymorphic component of the genome even though a number of VNTRs have been shown to be associated with neurological disorders and brain functions. This study will examine the association of VNTRs in Alzheimer’s disease (AD) population using a specialized program, VNTRseek, to explore the presence of these VNTRs in the whole genome sequences from AD cases and controls. We believe analyzing the distribution of VNTRs in a large and diverse AD population may yield new associative genetic alleles which may not only assist in the prediction of AD development but identify new cellular pathways of interest in understanding the pathophysiology of the disorder.

This goal of this study is to examine the presence of Variable Number Tandem Repeats (VNTRs) in Alzheimer’s disease (AD) population. Within a genome exists short sequences of repeating DNA. While the repeated sequence (usually >6 bases in length) is usually conserved within a population, the number of times the sequence is repeated in any given individual varies. These genetic variants are known as VNTRs and the number of these repeats can be considered a polymorphism, with individuals or families having a different number of repeats than those seen in the general population. In the past VNTRs have been an overlooked polymorphic component of the genome even though a number of VNTRs have been shown to be associated with neurological disorders and brain functions. We aim to understand not only the presence of specific VNTRs associated with Alzheimer’s disease but also how different phenotypes influences the relationship of these VNTRs to AD, this may inform precise genetic profiles which can be applied to a diverse population. As the associative genetics of AD is known not to be conserved across races it is important to not only assess the dataset as a whole but also the association of the ethno-racial phenotypes to inform these precise genetic profiles. If we can use VNTRs to predict this debilitating disorder it opens up avenues to apply treatments earlier and impact one of the most prevalent social and economic burdens on our society. Whole genome sequence data will be analyzed using VNTRseek against a set of reference tandem repeats generated from the tandem repeats database. This software determines the presence of a particular repeat. Each output will contain the number of repeats for each VNTR for each genome and this will be filtered for repeats that have a variation from the reference to determine the presence of a VNTR. This will be used to determine allelic frequency of specific polymorphic repeats comparing controls to AD. We will also further separate the analysis into gender and the different ancestral phenotypes in the ADSP to assess for specific alleles that may be more associated in one ethno-racial group than another.

Within a genome exists short sequences of repeating DNA. While the repeated sequence (usually >6 bases in length) is usually conserved within a population, the number of times the sequence is repeated in any given individual varies. These variants are known as Variable Number Tandem Repeats (VNTRs) and the number of these repeats can be considered a polymorphism. In the past VNTRs have been an overlooked polymorphic component of the genome even though a number of VNTRs have been shown to be associated with neurological disorders and brain functions. This study will examine the association of VNTRs in Alzheimer’s disease (AD) population using a specialized program, VNTRseek, to explore the presence of these VNTRs in the whole genome sequences from AD cases and controls. We believe analyzing the distribution of VNTRs in a large and diverse AD population may yield new associative genetic alleles which may not only assist in the prediction of AD development but identify new cellular pathways of interest in understanding the pathophysiology of the disorder.

Identification and characterization of genetic modifiers of risk for Alzheimer’s disease (AD) is paramount to development of a deeper understanding of AD pathogenesis as well as the identification of biomarkers and drug targets. The proposed research seeks to identify novel rare variants that could modulate an individual’s risk for developing sporadic late-onset or early-onset AD, and validate rare variants identified and characterized by our group that contribute to AD risk. In particular, we will focus on characterizing the contributions of changes in repetitive sequences within the coding regions of genes to AD risk. This study will also combine datasets from this project with whole genome or exome sequencing data generated by the University of California, San Francisco from individuals diagnosed with atypical or early-onset (< 65 years of age at diagnosis) AD to identify genetic risk factors unique to these less common forms of AD.

We will use ADSP data in conjunction with existing data from our research center to characterize genetic variation that influence a person’s risk for developing Alzheimer’s disease. After establishing a set of candidate variants, we will functionally characterize their biological effect using cell and biochemical assays. Identification and characterization of Alzheimer’s disease risk modifiers will not only enhance our understanding of disease pathogenesis, but may also facilitate identification of therapeutic targets and biomarkers for preventing Alzheimer’s disease.

Research objective: To examine the polygenic architecture of AD using genetic variants identified in the ADSP data set. We plan to evaluate genetic variation originating from genes, pathways and other functional units for association with AD. Study design: Using whole genome and whole exome sequence, we will perform subject and variant QC using in house pipelines. We will then search for small sets of functionally related variants that show significant, replicable association with AD. Analysis plan: Our analysis plan will include in depth quality control of samples/subjects for the following metrics: sequencing coverage, sample call rate, missing chromosomes, sample contamination, sex check, relatedness, population substructure and a check of APOE genotypes. This will be followed by variant quality control and subsequent single variant association analysis adjusting for variation arising from population substructure, sequencing centers, sex, and APOE genotype. Using approaches suitable for analyzing sets of genes and/or variants (e.g. polygenic score analysis, SKAT-O), we will explore the polygenic architecture of AD of all available datasets by searching for sets of functionally related genes and/or variants which show significant, replicable association with AD.

To identify novel AD genes which can be studied to develop new therapeutic approaches to AD, we will extend conventional analysis of single genetic variants by using methods capable of jointly analyzing all of the genetic variation in individual genes and functionally related sets of genes. These methods should enable us to identify novel, functionally related sets of genes which alter risk of AD.

The project aims at developing a new method to detect rare non-coding variants in enhancers in Alzheimer's disease. Briefly, we limit variants discovery within potential enhancer hotspots from a combination of chromatin marks and eRNAs. We will study the effects of newly discovered non-coding variants using mutations from the whole-genome sequencing data, and examine their associations with phenotypes. The requested datasets will be used to establish and validate our method.

Alzheimer's disease is potentially caused by rare mutations in human genome. Detecting these rare variants proves to be difficult. Our major goal is to develop a new approach to identify these rare mutations, and find out their contribution to the risk of Alzheimer's disease.

The AI4AD (Artificial Intelligence for Alzheimer’s Disease) Initiative aims to create and develop advanced AI methods, and apply them to genomic, imaging and cognitive data, in order to 1) identify AD genotypes and endophenotypes that dissect AD’s heterogeneity; 2) relate said genotypes and endophenotypes with clinical progression in pre-dementia patients; 3) identify novel treatment targets for AD by analyzing whole genome and associated phenotypic data. The goals of this multisite initiative (Paul Thompson, USC; Christos Davatzikos, Li Shen, Penn; Andy Saykin, IU; Heng Huang, Pitt, Paul Crane, UW; Adam Brickman, Columbia; Tim Hohman, Vanderbilt; Guyngah Jun, BU; Duygu Tosun, UCSF; Alexander Zaranek, Curii) leverage the promise of machine learning (ML) to contribute to precision diagnostics, prognostication, and targeted and novel treatments. We will develop ML and deep learning methods to apply to large scale biobanks of whole genome sequences (WGS), neuroimaging, cognitive, and clinical data, aiming to discover new genomic features that influence biological processes of AD. We will apply methods of genome representation and tiling to WGS repositories to create inputs for AI methods. We will develop novel, interpretable, biological knowledge guided deep learning methods to discover genomic motifs associated with AD, AD risk, and biological processes of AD as defined by NIA-AA criteria. To quantify subtypes and disentangle biological processes of AD, we will apply computational methods to multimodal MRI and amyloid- and tau-sensitive PET to stratify and subtype patient groups; novel imaging genomics methods will detect genomic markers and pathways that modulate the developing pathology as detected in the images, and that predict future clinical decline or resilience. We hypothesize that advanced deep learning methods combined with whole genome data will outperform traditional methods and GWAS for predicting AD onset and progression, and will assist with disease subtyping and discovering treatable targets in the genome. A team will rank and repurpose existing, and identify novel drugs and targets in the genome based on the discovered genetic motifs affecting AD.

The AI4AD (Artificial Intelligence for Alzheimer’s Disease) Initiative aims to create and develop advanced AI methods, and apply them to extensive and rich genomic, imaging and cognitive data, in order to 1) identify genotypes and endophenotypes of AD that dissect the heterogeneity of the disease; 2) relate these genotypes and endophenotypes with clinical progression, in pre-dementia patients; 3) identify novel treatment targets for AD, by analyzing whole genome and associated phenotypic data at a previously impossible scale. Collectively, the goals of this highly collaborative multi- site initiative leverage the promise of machine learning to contribute to precision diagnostics, prognostication, and targeted and novel treatments.

Analyzing postmortem phenotype and genomics data from ~1100 human brain samples with machine learning, Vivid Genomics, Inc., has developed prototype genetic biomarker assays that predict the presence of amyloid plaques, Lewy body pathology, cerebral amyloid angiopathy, and rate of cognitive decline. The objective is to increase subject numbers with similar data available through NIAGADS and NACC, along with datasets from several individual academic centers, to further optimize and validate our assays for neurodegenerative/cerebrovascular lesion types including tau, TDP-43, hippocampal sclerosis and microinfarcts, and for predicting rate of cognitive decline. NIAGADS datasets requested are NG00067 (including the newly released data which is new version 9 for dataset NG00067), NG00119, NG00117 and NG00127; data use limitations from these do not exclude our proposed usage. We are targeting >3000 subjects in total to be used for the validation of our models. We will focus on SNP selection and test the effects of different analysis strategies: 1) changing SNP p-value cutoffs 2) using LD-filtered representative SNPs with full genome coverage 3) testing the value of stratifying by APOE genotype 4) determining if it is better to add other covariates including age and sex. A fraction of the genetic data (~30%) will be withheld for validation. Optimization is defined as an area under the curve (AUC) of 80% and positive predictive value (PPV) of 80%, as well as R2 >0.75 for all assays. Values within 10% of this will be considered a successful validation. Through these assays, this project will benefit those suffering from Alzheimer’s disease and other neurodegenerative disorders by increasing clinical trial efficiency through more precise subject selection and/or stratification.

Vivid Genomics is dedicated to developing genetic tests, typically done from DNA obtained from blood, that will predict, for any given older person, the likelihood that they have, or might develop when they become old, the characteristic brain changes of Alzheimer’s disease as well as other brain changes that affect thinking in older people. These changes include amyloid or senile plaques, tangles or tau, amyloid angiopathy, Lewy bodies, TDP-43 pathology, hippocampal sclerosis and brain infarcts (strokes). The objective of this study is to improve upon initial tests developed by Vivid, and to also develop genetic tests to predict the rate at which older people’s thinking ability decreases over time. To do this, Vivid Genomics requests human subject DNA analysis data stored at NIAGADS. Through these new genetic tests, Vivid hopes to benefit those suffering from Alzheimer’s disease and other brain diseases of aging by allowing better selection of subjects for clinical trials of these diseases, which would increase the chances of clinical trials finding useful new treatments.

GWAS, WES and WGS have identified many genes associated with Alzheimer’s Dementia (AD) and its related traits. However, the identified genes thus far collectively explain only a small proportion of disease heritability, suggesting that more genes remained to be identified. Moreover, there is a clear gender and ethnic disparity for AD susceptibility, but little research has been done to identify gender- and ethnic-specific variants associated with AD. Of the many challenges for deciphering AD pathology, lacking of efficient and power statistical methods for genetic association mapping and causal inference represents a major bottleneck. To tackle this challenge, we have developed a set of novel statistical and bioinformatics approaches for genetic association mapping and multi-omics causation inference in large-scale ethnicity-specific epidemiological studies. The goal of this project is to leverage the multi-omics and clinical data archived by the ADSP, ADNI, ADGC as well as other AD-related data repositories to identify novel genes and molecular markers for AD. Specifically, we will (1) validate our novel methods for identifying novel risk and protective genomic variants and multi-omics causal pathways of AD; (2) identify novel ethnicity- and gender-specific genes and molecular causal pathways of AD. We will share our results, statistical methods and computational software with the scientific community.

Although many genes have been associated with Alzheimer’s Dementia (AD), these genes altogether explain only a small fraction of disease etiology, suggesting more genes remained to be identified. Of the many challenges for deciphering AD pathology, lacking of power statistical methods represents a major bottleneck. To tackle this challenge, we have developed a set of novel statistical and bioinformatics approaches for genetic association mapping and multi-omics causation inference in large-scale ethnicity-specific epidemiological studies. The goal of this project is to leverage the rich genetic and other omic data along with clinical data archived by the ADSP, ADNI, ADGC as well as other AD-related data repositories to identify novel genes and molecular markers for AD. Such results will enhance our understanding of AD pathogenesis and may also serve as biomarkers for early diagnosis and therapeutic targets.

Alzheimer’s disease (AD) affects 5.6 million Americans over the age of 65 and exacts tremendous and increasing demands on patients, caregivers, and healthcare resources. Our current understanding of the biology and pathophysiology of AD is still limited, hindering advances in the development of therapeutic and preventive strategies. Existing genetic studies of AD have some success but these explain only a fraction of the overall disease risk, suggesting opportunities for additional discoveries. The proposed project will leverage existing neuroimaging and genetic data resources from the UK Biobank, the Alzheimer’s Disease Sequencing Project (ADSP), the Alzheimer’s Disease Neuroimaging Initiative (ADNI), and the Cohorts for Heart and Aging Research in Genomic Epidemiology (CHARGE) consortium, and will be conducted by a multidisciplinary team of investigators. We will derive AD endophenotypes from neuroimaging data in the UK Biobank using deep learning (DL). We will identify novel genetic loci associated with DL-derived imaging endophenotypes and optimize the co-heritability of these endophenotypes with AD-related phenotypes using UK Biobank genetic data. We will leverage resources and collaborations with AD Consortia and the power of DL-derived neuroimaging endophenotypes to identify novel genes for Alzheimer’s Disease and AD-related traits. Also, we will develop DL-based neuroimaging harmonization and imputation methods and distribute implementation software to the research community. We expect to discover new genes relevant to AD which may leads to understanding of molecular basis of AD and potential new treatment.

Alzheimer’s disease (AD) exacts a tremendous burden on patients, caregivers, and healthcare resources. Our current understanding of the biology of AD is still limited, hindering advances in the development of treatment and prevention. Existing genetic studies of AD have some success but more studies are needed. The proposed project will leverage existing neuroimaging and genetic data resources from the UK Biobank, the Alzheimer’s Disease Sequencing Project (ADSP) and other consortia and will be conducted by a multidisciplinary team of investigators. We will derive new AD relevant intermediate phenotypes from neuroimaging data using deep learning (DL), an AI approach. We will identify novel genetic loci associated with these phenotypes. Also, we will develop imaging harmonization and imputation methods and distribute implementation software to the research community. We expect to discover new genes relevant to AD which may leads to understanding of molecular basis of AD and potential new treatment.

Explore somatic transposable elements and their Alzheimer's disease-related patterns using genomic and phenotypic data from large cohorts:In order to explore the impact of the transposable element in Alzheimer's disease, we propose to conduct a systematic survey in the available large cohorts. The ADSP dataset in NIAGAlzheimer's diseaseS (Accession No. NG00067) includes 16,906 whole-genome sequences and 20,504 whole-exome sequences for case-control and family-based studies of Alzheimer's disease from diverse populations, which is a perfect resource to leverage in this project. Under the support of the Michigan Alzheimer's Disease Center, we will request to access NIAGADS. To detect somatic transposable elements in the ADSP dataset, we will employ established computational pipelines to resolve the transposable elements in the sequencing data, MELT and xTEA for WGS and SCRAMble for WES, respectively. Parameters in these tools, for instance, the calling threshold of supporting reads, will be adjusted accordingly to cooperate with the detection of somatic transposable elements in cells at low frequency. To exclude potential germline transposable elements, we will leverage a master set of polymorphic transposable elements from diverse populations, which are based on our previous projects at the Human Genome Structural Variation Consortium, and the case-control information provided by ADSP. We aim to summarize a spectrum of somatic transposable elements that would be Alzheimer's disease-relevant along with various clinical and phenotypic information. To build Alzheimer's disease-related genetic patterns we will implement Mutect2 (GATK) and Strelka2 to discover SNVs from WGS and WES data and link them with transposable elements in the same haplotype. After obtaining this set of patterns, we will collect phenotypic information from the ADSP dataset to conduct family-based associated analysis and gene-burden analysis. RegulomeDB will be used to annotate the effects of non-coding functional impact and regulatory changes for these Alzheimer's disease-related patterns.

It seeks to explore the connection between the somatic transposable elements in the human genome and Alzheimer’s disease and related dementias. It will leverage large-scale datasets to extensively explore the genome-wide transposable elements and then stratify Alzheimer’s disease-relevant ones by using the rich clinical information from the cohorts. Further analysis pipelines will be built based on the results of the proposed project to investigate the functional impact of these transposable elements on Alzheimer’s disease and would improve the understanding of genetic causes of Alzheimer’s disease and related dementias.

We will use the data to build an AD-related omics database both by using state-of-the-art and by developing advanced deep learning-based methods for harmonization and imputation of multi-omics data, facilitating system biology studies to gain deep understanding of AD.We will use the data to identify the genetic biomarkers with causal effect on behavioral deficits in Alzheimer’s study and use these biomarkers to help predict Alzheimer's disease (AD). SNP data will be screened first using the GWAS and significant SNPs will be selected as genetic biomarkers according to their p-values. Then causal effects will be estimated to evaluate the contribution of genetic biomarkers. For AD prediction, enhanced statistical, machine learning, and deep learning approaches will be explored and compared, which may include but not limited to: the PCA decomposition, ridge regression/elastic net algorithm, boosting algorithms such as XGBoost/lightGBM, deep learning models such as the deep factorization machine.In far future, we want to develop a structured, literature and expert-knowledge based knowledge graph for better prediction, diagnosis and treatment of AD.

Leveraging our newly developed causal inference method, we aim to identify genetic causal pathways for the Alzheimer's disease (AD) from genomic data collected from multiple populations. The identified features will be used to predict cognitive and behavior scores among patients with AD or mild cognitive impairment (MCI). Finally, we want to develop a knowledge graph for better prediction, diagnosis and treatment of AD.

We will run multiple structural variation (SV) callers, including but not limited to Absinthe, Canvas, Manta, MELT, and PanGenie, on the entire Alzheimer’s Disease Sequencing Program (ADSP) WGS data set. We will combine these data into a single coherent SV call set which we will genotype on the entire ADSP WGS data set. We will use these data, together with SNV and indel calls, to generate a phased integrated SNV/indel/SV genotype panel for ADSP. We will then use this to impute the presence of SVs in data from the Alzheimer’s Disease Genetics Consortium (ADGC) genotyping on samples not included in ADSP and also in external phenotyped population sequencing projects including but not necessarily limited to UK Biobank. The resulting directly genotyped and imputed SVs will be used in association studies with AD status and related phenotypes.We will perform this work in collaboration with Badri Vardarajan at Columbia University.We will collaborate with Gerard Schellenberg, Li-San Wang, Wan-Ping Lee and Yuk Yee (Fanny) Leung in GCAD at University of Pennsylvania to share individual SV callsets we have generated distinct from theirs and build a joint SV callset.NYGC will contract the services of Wayne Clarke from Outlier Informatics for this project. Dr. Clarke will be responsible for the cloud implementation of NYGC's SV calling pipelines and for optimizing and running data analysis on AWS and GCP as well as locally on NYGC hardware. In this capacity Dr. Clarke will use NYGC accounts and compute infrastructure to access NIAGADS data in the cloud or downloaded to NYGC's on prem compute infrastructure.

Structural variation refers to alterations in DNA that change the copy number or ordering of large blocks of DNA. Historically, these have been difficult to detect accurately and thus their impact on disease risk has often been ignored. We will apply the latest methods to discover and characterize structural variants in Alzheimer’s Disease (AD) cases and controls, and then use genetic association techniques to determine whether these events are likely to be correlated with increased or decreased risk of developing AD or any of the specific features of AD. Identifying additional genetic risk factors may improve diagnosis of AD and may increase our understanding of the biological mechanisms leading to AD and possible therapeutics.