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

Computational drug repurposing has largely been focused on rapid hypothesis generation, yet real-world applications span a far broader lifecycle: suggesting drug candidates, designing experiments, analyzing assay data, and iteratively refining hypotheses. Here, we demonstrate that agentic AI can operate throughout this lifecycle. We developed RepurAgent, a hierarchical multi-agent system comprising a supervisor agent and a planning agent that coordinates four specialized sub-agents (research, prediction, data, and report) under human-in-the-loop design, with episodic memory and retrieval-augmented generation. It is grounded in repurposing-specific data, tools, and standard operating procedures developed within the REMEDi4ALL consortium. We validated RepurAgent 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 in agreement with expert selection (AUC-ROC 0.986) ; and for Multiple Sulfatase Deficiency, it prioritized 81 high-confidence candidates from 5159 compounds, which were further corroborated by domain experts. These results demonstrate that agentic AI can support across the drug repurposing lifecycle, from hypothesis generation to experimental analysis.