Human-supervised Agentic AI for Hypothesis Generation and Experimental Assistance in Drug Repurposing

Computational drug repurposing has focused on rapid candidate suggestion, yet real-world campaigns span a broader lifecycle: generating hypotheses, designing experiments, analyzing assay data, and refining the hypotheses. We present RepurAgent, a hierarchical multi-agent system in which a supervisor and planning agent coordinate four specialized sub-agents (research, prediction, data, and report) under human-in-the-loop design, with a three-layer memory architecture and an ensemble retriever grounded in standard operating procedures. We evaluate the system across three scenarios spanning distinct stages within the repurposing lifecycle: in Acute Myeloid Leukemia, a blinded expert evaluation indicated that RepurAgent produced substantially more novel and mechanistically credible candidates compared to a vanilla LLM baseline; in a retrospective COVID-19 antiviral screen, RepurAgent acted as an adaptive experimental collaborator, prioritizing compounds with AUC-ROC up to 0.99 without predefined thresholds and flagging confounders missed in manual review; and for Multiple Sulfatase Deficiency, it prioritized 81 high-confidence candidates from 5159 compounds, one of which was independently corroborated by domain experts. These results demonstrate that agentic AI can provide support across the drug repurposing lifecycle, from hypothesis generation to experimental analysis.