Stanford Medicine researchers built a virtual biotech operation with 37,000 AI agents that analyzed roughly 50,000 clinical trials in under a week and developed a lung cancer therapy strategy.

A pharmaceutical company later independently arrived at the same strategy, and its therapy received FDA Breakthrough Therapy designation.

The results, published Thursday, Sept. 17 in Science, describe what associate professor James Zou and graduate student Harrison Zhang call a "virtual biotech company." The AI system mirrors the organizational chart of a real biotech firm, with a chief science officer agent leading specialized divisions that work in parallel on drug design.

Zou, who leads Stanford's biomedical data science lab, said the project tested whether AI could support work across the drug-development pipeline.

"Could we create a biotech company that takes on everything from looking for drug targets all the way to designing clinical trials? … Could we have a fully agentic company that tackles the extremely complex challenges of drug discovery?" Zou said.

The agents catalogued roughly 50,000 clinical trials in less than a week and built scoring systems from single-cell gene activity data. They found that drugs targeting "switch-like" genes were 40% more likely to advance from phase 1 to phase 2 trials, 48% more likely to reach market and had 32% fewer adverse events compared with drugs that had a broad spectrum of activity. Those patterns held across cancers, brain diseases, heart diseases and kidney and lung conditions.

The agents then zeroed in on B7-H3, a protein long studied by lung cancer researchers. They found it was highly expressed in fibroblasts, connective-tissue cells that often sit near tumors. The fibroblasts appeared to signal nearby immune cells and suppress their activity, effectively cloaking tumors from the body's defenses.

Using that analysis, the agents designed an antibody-drug conjugate targeting B7-H3. An antibody-drug conjugate is a targeted treatment that uses an antibody to deliver a cancer-killing drug to cells carrying a particular protein. The agents proposed the design using information available before January 2025.

In August 2025, a private pharmaceutical company independently arrived at the same antibody-drug conjugate strategy against B7-H3, according to Stanford Medicine. The company's therapy later received FDA Breakthrough Therapy designation. The designation is intended to speed the development and review of treatments for serious conditions when preliminary clinical evidence suggests they may substantially improve on available therapies.

The virtual biotech builds on Zou's earlier "virtual lab," published in Nature in July 2025, which designed 92 novel nanobody binders against SARS-CoV-2 variants, according to The Next Web. Zou's lab has produced or advised more than 10 companies, including Gradio, the open-source machine learning tool acquired by Hugging Face in 2021.

Bloomberg reported in April 2026 that Zou was seeking to raise about $100 million at a roughly $1 billion valuation for a startup called Human Intelligence. Zou did not respond to Bloomberg's requests for comment at the time. The fundraising effort's current status is unclear.

Zou is not pursuing the B7-H3 target further. Another company is already shepherding it into clinics. But the virtual biotech has surfaced other candidate targets, and Zou said the team's next step is bringing those findings into real labs to test how many hold up.