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

Checked 2 Nov 2026. Open license (CC-BY, CC-BY-SA, CC0, public domain): full text indexed and used in synthesis.

Abstract Computational drug repurposing has largely been focused on rapid hypothesis generation, yet real-world applications span a far broader lifecycle, from drug candidate suggestion to designing experiments, analyzing assay data, and iteratively refining candidates. Here, we demonstrate that agentic AI can operate throughout this lifecycle. To this end, we developed RepurAgent, a hierarchical multi-agent AI system comprising a supervisor agent and a planning agent that coordinate four specialized sub-agents (research, prediction, data, and report), through a human-in-the-loop design, with episodic memory and retrieval-augmented generation. The system is grounded in data, tools, and standard operating procedures specific for drug repurposing, developed within the REMEDi4ALL consortium. We validated the agentic system across three scenarios spanning the various 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 5000 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. RepurAgent is open source and deployed at https://repuragent.serve.scilifelab.se/ .

• Within drug discovery, developers stake out opposed autonomy positions for similar lifecycles: Ghareeb et al. 2026 describe Robin as the first system to fully automate hypothesis generation and data analysis, whereas Huynh et al. 2026 build RepurAgent explicitly around a human-in-the-loop design and Zhao et al. 2026 position agents as pragmatic scientist-in-the-loop co-pilots. Trace →
• Three independent 2026 efforts converge on 'scientist-in-the-loop' as the pragmatic deployment form for research agents: Jeon 2026 prescribes strict boundaries in plant pathology under which a human sets goals, reviews AI output before it is treated as evidence, and approves consequential actions; Zhao et al. 2026 advocate 'pragmatic scientist-in-the-loop co-pilots' for drug discovery; and Huynh et al. 2026 built the drug-repurposing system RepurAgent around a human-in-the-loop design. Trace →
• The self-assessment pattern in evaluating autonomous-research systems has at least a partial counterexample: Huynh et al. 2026 validated RepurAgent through a blinded expert evaluation against a vanilla LLM baseline, which judged its drug-repurposing candidates substantially more novel and mechanistically credible. Trace →
• The statistical-rigor deficit in drug-discovery AI is not only on the machine side: in Huynh et al. 2026's retrospective COVID-19 antiviral screen, the human-supervised RepurAgent prioritized compounds with AUC-ROC up to 0.99 while flagging confounders missed in manual review—agentic analysis supplying scrutiny the original human analysis lacked. Trace →
• Agentic scrutiny of already-published human analyses is recurring across domains: Huynh et al. 2026's RepurAgent flagged confounders missed in manual review of a COVID-19 antiviral screen, and Baker et al. 2026's large-scale agentic reproduction exposed unstated modeling choices, reference mismatches and subsampled tutorial inputs in published studies. Trace →

• AI for science and agentic research