AI-Assisted Gene Expression Signature Reversal for Candidate Prioritization in Drug Repurposing

Abstract

: Drug repurposing increasingly relies on transcriptomic evidence to identify new therapeutic uses for existing drugs or drug-like compounds. This review examines how AI-assisted gene expression signature reversal contributes to candidate prioritization in drug repurposing. The approach compares disease-associated gene expression signatures with drug-induced transcriptional profiles, using resources such as the Connectivity Map and LINCS as the data foundation. Signature reversal is valuable because it offers a biologically interpretable ranking logic: compounds that oppose disease-associated expression changes may be prioritized for further study. The reviewed literature suggests that this logic is useful, but uneven. Some studies provide evidence that reversal scores can relate to drug efficacy, whereas validation-focused work also shows that reversal signals may be confounded by broad anti-proliferative effects. AI and machine learning extend this framework by supporting high-dimensional expression analysis, transcriptional response prediction, expression ranking, and context-aware prioritization. Recent deep learning approaches highlight the importance of dose and cellular context when predicting drug-induced responses. Overall, AI-assisted signature reversal is most defensible as a candidate triage strategy rather than as evidence of efficacy. Its value depends on combining reversal scores with pathway interpretation, biological context, and targeted follow-up testing.