Sep 17, 2026
ManyPress

Advertisement

Science

A new virtual biotech company, created by Stanford Medicine researchers, utilizes 37,000 AI agents to simulate the entire drug discovery pipeline, from target identification to clinical trial design.

ManyPress

ManyPress

ManyPress Editorial

3 min readSource:Phys.org
Stanford Researchers Develop Virtual Biotech Company Powered by AI Agents

Key facts

  • The virtual biotech company employs 37,000 AI agents to manage drug discovery tasks.
  • AI agents analyzed 50,000 clinical trials in under one week to identify biological success predictors.
  • Drugs targeting switch-like genes were 40% more likely to advance from phase 1 to phase 2 trials.
  • The AI-designed B7-H3 antibody-drug conjugate matched a strategy later independently developed by a pharmaceutical company.
  • The research team plans to transition findings from the virtual environment into physical laboratory testing.

Stanford Medicine associate professor James Zou and graduate student Harrison Zhang have launched a virtual biotech company staffed entirely by 37,000 artificial intelligence agents. The company, which operates without physical lab space or human employees, is designed to emulate the organizational structure of a traditional biotech firm. The researchers published their findings on the AI-powered framework in the journal Science on September 17, demonstrating the system's ability to analyze complex biological data and propose viable drug candidates.

By the numbers

37,000
number of AI agents in the virtual company
50,000
number of clinical trials analyzed by AI agents
40%
increased likelihood of advancing from phase 1 to phase 2 trials
48%
increased likelihood of reaching market
32%
reduction in adverse events

AI-Driven Drug Discovery and Analysis

The virtual company utilizes specialized AI agents to perform tasks across the drug development pipeline. In one project, the agents analyzed 50,000 clinical trials in less than a week, a task that would have taken human researchers years. By examining single-cell gene activity, the agents identified that drugs targeting 'switch-like' genes—those that act as on-off switches rather than dimmers—showed higher success rates in advancing through clinical trials and fewer adverse events.

Validation Through Cancer Therapy Design

To test the system's efficacy, the researchers tasked the AI agents with designing a therapy for lung cancer targeting the B7-H3 protein. The agents identified that fibroblasts expressing B7-H3 were signaling to suppress immune cells, and subsequently designed an antibody-drug conjugate to deliver chemotherapy directly to those cells. This design was independently reached by a major pharmaceutical company months later, which subsequently received FDA breakthrough therapy designation for the strategy.

Timeline

  1. January 2025
    The AI agents proposed a drug design for B7-H3 using available information.
  2. August 2025
    A pharmaceutical company independently developed the same B7-H3 strategy.
  3. September 17
    The research paper describing the virtual biotech company was published in Science.

Advertisement

This article was independently rewritten by ManyPress editorial AI from reporting originally published by Phys.org.

Science