Researchers use 37,000 AI agents to analyze clinical trials and design a drug candidate

Stanford researchers built an experimental system of 37,000 AI agents that analyzed roughly 50,000 clinical trials in less than a week and proposed a lung-cancer therapy, according to work described in Science. Specialized teams handled tasks including molecular analysis, therapeutic design and trial planning. The agents linked trial outcomes with molecular measurements and proposed an antibody-drug conjugate targeting B7-H3, combining a targeting antibody with a chemotherapy payload. A pharmaceutical company independently pursued the same general strategy in August 2025. The analysis was retrospective, and its associations do not establish what causes drugs to succeed. No physical candidate from the virtual system entered a clinical study; its additional proposed targets still require experimental evaluation.