AI-driven Virtual Biotech identifies promising lung-cancer therapy
A team of 37,000 AI agents, dubbed Virtual Biotech, analyzed tens of thousands of clinical trials and pinpointed a potential lung-cancer drug targeting protein CD276.
James Zou and his collaborators built a simulated biotech organization called Virtual Biotech, composed of up to 37,000 autonomous AI agents that interact with large language models such as Anthropic's Claude. The AI workforce was tasked with reviewing results from more than 55,000 clinical trials, discovering that targeting proteins expressed in particular cell types nearly doubles the likelihood of a drug reaching the market.
When directed to explore lung-cancer targets, the system highlighted CD276, a protein known to suppress immune responses and be abundant in lung tumors. Using existing data, the AI designed a strategy involving a CD276-recognizing antibody linked to a cytotoxic payload. Independent reviewers assessed the proposal as a viable avenue for further research, though the approach has yet to be validated experimentally or clinically.
Why it matters
It shows how large-scale AI could accelerate drug discovery, potentially shortening the path to new cancer treatments.
In this story
