El Agente F\'armaco: A Scalable AI Scientist for Drug Discovery

| Source: arXiv AI

Tags: El Agente Fármaco, Drug Discovery, AI Scientist, LLM Agents, Molecular Docking, Codex

El Agente Fármaco is an LLM-based AI scientist that turns plain-language drug discovery goals into scalable screening workflows, scoring highest on a 22-task benchmark against plain tool calls and Codex.

Details

Computational drug discovery has many tools for proposing and evaluating molecules, but chaining them and choosing next steps still takes expert effort. El Agente Fármaco turns natural-language discovery goals into screening workflows. Its harness selects tools, reviews results, and adjusts steps to the target and user objectives. A shared memory called Fármaco Atlas stores molecules, structures, and results with provenance, and the agent pulls evidence through targeted database queries. One request can launch calculations across a whole molecule collection without a separate LLM decision per compound, which limits context growth. In a retrospective comparison with an expert-led campaign it proposed chemically distinct candidates with competitive docking scores and higher calculated drug-likeness. On a 22-task benchmark it had the highest mean model-judged score versus plain tool calls and Codex, with lower median token use.