Functional Interpretation of Alzheimer’s Disease Genetic Risk and Systematic Gene Prioritization

Abstract

AbstractBackgroundGenome‐wide association studies (GWAS) of Alzheimer’s disease (AD) risk identified over 100 genome‐wide significant susceptibility loci in the past 15 years, but remaining challenges include resolving the mechanism of action of how these associations functionally contribute to AD risk and prioritizing the candidate risk genes in the complex risk loci. To address these, we have recently developed a systematic gene prioritization pipeline that nominated highly likely candidate risk genes within the majority of 42 new loci identified by our recent AD GWAS meta‐analysis. In this subsequent work, we investigated the complete landscape of AD risk‐associated molecular phenotypes and further extended our analyses to known AD risk loci and subthreshold significant loci.MethodWe investigated downstream effects of AD‐associated variants on molecular phenotypes (e.g. gene expression, splicing, methylation, protein expression, histone modifications) through molecular quantitative trait loci (QTL) catalogues measured in disease‐relevant tissues and cells. We queried the lead variants in these QTL catalogues and performed genetic colocalization analyses (coloc) & transcriptome/methylome/proteome‐wide association studies (TWAS/EWAS/PWAS) using the EADB stage I meta‐analysis GWAS on 85,934 cases and 401,577 controls. Evidence for candidate genes in new, known, and subthreshold loci were assessed by the gene prioritization pipeline that was updated with new analyses. Moreover, we investigated AD‐associated cryptic splicing events using Nanopore sequencing.ResultOur results implicated >140 protein‐coding genes with AD risk in 42 new loci, >275 in 34 known loci, and over >65 in 31 subthreshold loci. Within MINDY2 known locus, evidence from multiple QTL‐GWAS integration domains confirmed ADAM10 as a prioritized gene. Moreover, we obtained first QTL‐based evidences for candidate genes within previously‐unresolved new risk loci such as RASGEF1C. In a subthreshold locus, fine‐mapped expression TWAS associations were observed for SNX31 (a member of sorting nexins together with SNX1 in novel loci) consistently across multiple brain expression reference panels, implicating genetic upregulation of SNX31 with increased AD risk. Furthermore, we could confirm AD‐associated novel splicing events within prioritized genes such as DOC2A. ConclusionOur comprehensive study effectively and systematically prioritized AD risk genes and aided functional interpretation of AD GWAS by contributing to better understanding of genetically‐associated molecular mechanisms underlying AD risk.