Genome-wide evaluation of the pathway activity profiles of APOE ε4 negative Alzheimer’s Disease cohorts using germline genetics

Abstract

APOE is the most clearly established susceptibility gene for late-onset Alzheimer’s Disease where its ε4-typed allele has been suggested to account for 20-30% of AD risk. Nonetheless, epidemiology studies have indicated that APOE alone is neither sufficient nor necessary for the disease, leaving a knowledge gap for APOE ε4 negative patients. Multi-factor analysis approaches, such as the Polygenic Risk Scoring, are accordingly developed to evaluate genetic risks for complex diseases, while the performance in AD models is as yet limited. Therefore, in this study, we have established a network-framed model to incorporate the rare SNP profile across the entire genome and quantify their synergistic impact on pathway activity for profile patterns characterizing the APOE ε4 negative patients. The model was structured as a multi-layer deep learning network and trained with the whole exome sequencing and gene expression profile data of 971 cancer cell lines from the CCLE project. This enabled the projection of any given mutation profile onto a quantitative gene expression profile and further the pathway activity profile that indicated the significance of its deviation from the random simulation. A cohort of 245 APOE ε4 negative patients altogether with 170 agecomparable healthy individuals were analyzed for AD specific patterns. We found there were two diverged types of pathway activity which potentially develop disease via distinct approaches—one via adipogenesis and oxygen stress while the other via interleukin signaling pathways. Our model provides a holistic tool to characterize AD that prompts the heterogeneity of disease among APOE ε4 negative patients.