AI4AD (Artificial Intelligence for Alzheimer’s Disease): Ultrascale Machine Learning to Empower Discovery in Alzheimer’s Disease Biobanks

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.