From Risk Genes to Functional Connectivity: Multi-Dimensional Integrative Frameworks to Unravel Genetic Mechanisms in Alzheimer’s Disease

Abstract

Alzheimer’s disease (AD) demands integrative approaches to resolve its genetic, molecular, and neurobiological complexity. This dissertation develops frameworks integrating multi-tissue omics and neuroimaging data to dissect AD risk mechanisms across molecular, regulatory, and clinical dimensions. Initially, we introduce Multi-tissue Splicing Gene (MTSG), a tensor decomposition and sparse canonical correlation analysis (sCCA) framework that identifies 174 splicing-mediated AD risk genes by modeling multi-tissue splicing architectures, outperforming single-tissue approaches and revealing novel pathways. Subsequently, we extend this tensor-sCCA framework to identify transcriptional gene networks, i.e., transcription factor (TF) regulons, implicating GCM2 and ZNF732 regulatory networks in AD risk and uncovering cis-/trans-eQTLs linking GWAS variants to dysregulated TF activity. Finally, we bridge genetics to neuroimaging through advanced mediation analysis, employing multi-group structural equation modeling (SEM) to reveal stage-dependent effects of the AD-risk variant rs6733839: its risk allele paradoxically reduces tau pathology in cognitively normal individuals but directly disrupts functional connectivity in subjects with mild cognitive impairment, suggesting its temporally divergent roles. Collectively, this work demonstrates how integration of splicing, regulon, and neuroimaging data can address AD’s genetic heterogeneity, offering methodological advancements in tensor decomposition, causal mediation, and genetics-neuroimaging integration.