Statistical Framework for Unraveling Age-Dependent Genetic Landscape of Alzheimer's Disease and Related Dementias: Harnessing Large-Scale EHR and DNA-Biobank Integration

Project Summary This proposal seeks to improve our understanding of Alzheimer's Disease (AD) and related dementias (ADRD) by developing a more comprehensive view of its genetic architecture. It aims to leverage large-scale biobanks and Electronic Medical Records (EMR) to systematically study the genetic structure of AD/ADRD over time across a broad spectrum of phenotypes, and in diverse populations. The proposal emphasizes several key areas. First, it underscores the need to integrate age dependency and cross-phenotype dependency into the genetic architecture for AD/ADRD. The motivation comes from the increasing evidence that genetic expression changes throughout an individual's life, and co-morbid conditions may influence the progression of AD/ADRD as individuals age. To facilitate this goal, we will curate a compre- hensive set of age-dependent phenotypes and develop a new network-based approach to uncover genetic links to complex disease networks and their interactions with AD/ADRD throughout the life course. Second, the proposal capitalizes on using large-scale biobanks and EMR, important data resources for AD/ADRD research. The proposal aims to enhance the use of these richly phenotyped and longitudinal data resources to deepen our understanding of the genetic architecture of AD/ADRD, and consequently provide in- sights on potential therapeutic targets and early prevention strategies. That includes a new deep-learning-based imputation algorithm to handle the informative missing data in EMR, and semi-supervised learning algorithms to enhance statistical power. Third, the proposal aims to address methodological gaps in Phenome-Wide Association Studies (PheWAS). This includes proposing new phenotyping algorithms to curate age-related phenotypes, formulating an innovative statistical framework, high-dimensional Dynamic Exponential graphical model (DEG), for a more comprehensive and dynamic view, and handling missing data imputation to integrate lab values into the methods, and developing analytical strategies to enhance the robustness and power of such analyses. By targeting these key areas, this proposal has the potential to significantly advance AD/ADRD research, offer deeper insights into its genetic architecture, and improve early prevention and treatment strategies.