How a Research Lab Made Entirely of LLM Agents Developed Molecules That Can Block a Virus

A team of autonomous GPT-4o-based agents, organized as a "Virtual Lab," successfully designed new nanobody molecules to combat variants of the SARS-CoV-2 virus. These AI agents, assigned roles like immunologist and computational biologist, collaborated in structured meetings to devise a research strategy, select tools like AlphaFold, and write the necessary code. The agents decided to mutate existing nanobodies and developed a computational pipeline to rank the most effective mutations. The resulting designs were then synthesized and experimentally validated in a real-world lab, proving the viability of an AI-driven research process from ideation to actionable output.