MAVEvidence Novel Variant Effect Predictors for Alzheimer

The primary objective of this research is to utilize population data to validate predictors derived from functional assays for Alzheimer’s disease (AD) and Alzheimer’s disease-related dementias (ADRD) risk genes, specifically focusing on APP and SNCA. We aim to validate the functional relevance of these predictors by assessing their correlation with both AD prevalence and clinically meaningful endophenotypes that characterize disease progression.
Study Design:
This study will employ a multi-faceted design consisting of the following components:
Curation of MAVE Data: We will systematically curate existing MAVE studies for APP and SNCA, integrating data from published literature to establish a comprehensive dataset.
Validation of Functional Assays: We will conduct validation studies to quantify the discriminative power of MAVE scores in identifying pathogenic versus benign variants using known variants from ClinVar as a benchmark.
Quantitative Risk Modeling: We will develop a survival analysis framework to model the age- and genotype-dependent penetrance of AD, leveraging MAVE data alongside population incidence data.

Analysis Plan:
The analysis will focus on evaluating the following phenotypic characteristics in association with genetic variants:
Pathogenicity Discrimination: We will assess the ability of MAVE scores to distinguish between pathogenic and benign variants for APP and SNCA using Receiver Operating Characteristic (ROC) curves, aiming for an area under the curve (AUC) greater than 0.7.
Odds of Pathogenicity: The OddsPath for each variant will be calculated using MAVE data in conjunction with ClinVar annotations to quantify the strength of functional evidence.
AD Risk Estimation: We will develop a quantitative model that estimates the risk of developing AD based on genotype, validated against data from the Alzheimer’s Disease Sequencing Project and independent cohorts such as the UK Biobank.

We developed a tool, MAVEvidence, for use by variant scientists to interpret variants in AD/ADRD risk genes. We note that although this use case is not regulated by bodies such as the FDA, it is guided by the published
ACMG/AMP framework, to which MAVEvidence adheres. We aim to collect function scores from MAVEs for APP and SNCA, collect clinical annotations from ClinVar, and evaluate the ability of each MAVE to distinguish known pathogenic and benign variants. To apply functional evidence within the ACMG/AMP framework, function scores must first be converted to evidence strengths. We will calibrate variant-level functional evidence by first calculating the odds of pathogenicity and then applying a Bayesian approach to convert to evidence strength. Finally, we will generate quantitative measures of risk by integrating MAVE data and population AD/ADRD
incidence data into a survival model that predicts age- and genotype-dependent penetrance of Alzheimer’s disease.