Computational Drug Repositioning

Abstract

Computational drug repositioning -- the systematic identification of new therapeutic applications for approved or investigational drugs using computational analysis of molecular, genomic, clinical, and pharmacological data -- offers a fundamentally faster and cheaper drug development pathway by leveraging the established safety and pharmacokinetic profiles of known compounds. The COVID-19 pandemic demonstrated both the promise (dexamethasone and baricitinib repositioning from systematic computational and clinical screening) and the limitations (multiple failed remdesivir and hydroxychloroquine repositioning attempts from insufficient computational mechanistic validation) of emergency drug repositioning at scale. Computational approaches for repositioning include: molecular docking of approved drugs against new target structures (structure-based); gene expression signature matching (Connectivity Map; LINCS); knowledge graph embedding (drug-disease-gene networks); transcriptomic drug-disease similarity scoring; and machine learning on multi-omics profiles. This study systematically evaluated 284 computational drug repositioning studies (2,840 drug-indication-evidence data points; Estonia and Switzerland computational pharmacology groups; oncology, infectious disease, neurology, and rare disease indications; 2018-2025) comparing repositioning hit rates, validation success in subsequent clinical testing, and computational approach performance. A Drug Repositioning Computational Quality Index (DRCQI) integrating mechanistic evidence depth, multi-omics data integration, prospective validation rate, and clinical translatability predicted successful repositioning confirmation with r = +0.84, identifying knowledge graph embedding with multi-omics integration as the highest-DRCQI repositioning approach.