Stanford AI Virtual Lab Accelerates Lung Cancer Drug Discovery

Researchers at Stanford University have developed a virtual lab consisting of specialized AI agents that function as co-scientists to accelerate drug discovery. Led by Dr. James Zou, the system employs a fleet of bots, including a professor AI and a professional devil's advocate, to independently generate hypotheses and conduct computer-based experiments.

In a proof-of-concept study published in the journal Science, the agents analyzed the B7-H3 protein to identify an effective treatment approach for lung cancer. This finding was independently corroborated by clinical trial results from a drug developed by Merck. The analysis was completed in less than one day with computing costs totaling approximately $46.

The system has also been applied to the redesign of a Covid-19 vaccine, where it produced two functional novel proteins in a fraction of the usual time and cost. While the agents currently cannot perform physical laboratory experiments with cells, the team demonstrated that the combination of human researchers and AI models can significantly speed up the pace of biomedical progress.