Accurate Drug Repositioning through Non-tissue-Specific Core Signatures from Cancer Transcriptomes

Abstract

Summary: Experimental large-scale screens for drug repositioning are limited by restriction to in vitro conditions and lack of applicability to real human conditions. Here, we developed an in silico screen in human in vivo conditions using a reference of single gene mutations’ non-tissue-specific “core transcriptome signatures” (CSs) of 8,476 genes generated from the TCGA database. We developed the core-signature drug-to-gene (csD2G) software to scan 3,546 drug treatment profiles against the reference signatures. csD2G significantly outperformed conventional cell line-based gene perturbation signatures and existing drug-repositioning methods in both coverage and specificity. We highlight this with 3 demonstrated applications: (1) repositioned category of psychiatric drugs to inhibit the TGF-β pathway; (2) antihypertensive calcium channel blockers predicted to activate AMPK and inhibit AKT pathways, and validated by clinical electronic medical records; and (3) 7 drugs predicted and validated to selectively target the AKT-FOXO and AMPK pathways and thus regulate worm lifespan. : Xu et al. generated core transcriptome signatures representing single-gene mutations for 8,476 human genes from the cancer transcriptomes in the TCGA database and developed a drug-repositioning method to predict high-specificity candidate target genes for 1,938 drugs. Applied to targeting aging pathways, 7 lifespan-extending drugs are found and functionally validated. Keywords: drug repositioning, drug target, cancer genomics, data integrative analysis