A proof of concept study towards multi-omics-based computational drug repositioning in Alzheimer's disease.

Abstract

Alzheimer's disease (AD) is a devastating neurodegenerative disorder for which there currently are no disease-modifying treatments available. To accelerate the path to effective intervention strategies, drug repositioning - the application of available compounds in a novel disease context - has gained increasing attention as a promising alternative to de novo drug development. Rich multi-omics data, generated by large international and interdisciplinary AD consortia, is now enabling the implementation of novel methods that have the potential to drive the computational identification and prioritization of promising repositioning candidates. We recently developed the AD atlas, a web-based multi-omics resource that integrates multiple layers of heterogenous data from different studies and cohorts, including omics QTLs, transcriptomic, proteomic and metabolomic correlation networks, as well as genetic and multi-omics associations with AD and associated biomarkers/endophenotypes. Using this atlas, we generated and analyzed molecular context networks surrounding AD-associated genes as well as those targeted by drug repositioning candidates proposed in the literature. Subsequent enrichment analysis on AD subnetworks was used to identify drugs with overlapping molecular signatures on the gene expression level, while target networks of repositioning candidates were investigated for their potential involvement in AD pathogenesis. We found ample evidence for the potential of integrative multi-omics approaches for drug repositioning in AD. For instance, enrichment analysis of the context network surrounding the AD-associated genes APOE and CLU identified multiple repositioning candidates, where the top hits were drugs that were either previously proposed as promising or already subjected to clinical trials, such as fluoxetine, rosiglitazone, and valproate. Investigating candidate drugs, the exploration of the context network targeted by statins revealed functional links to TYROBP/TREM2 signaling, suggesting a potential protective effect of this drug class through modulation of neuroinflammatory pathways. Our results highlight multiple opportunities to advance drug repositioning efforts in AD by integrative analysis of comprehensive multi-omics data. Automation of our analyses using network-based machine learning approaches and extension of the AD atlas with multi-omics data from drug screens to resolve directionalities will allow us to globally identify molecular pathways disturbed in AD that are targetable by drug repositioning candidates.