Particle.news

Download on the App Store

AI-Driven Virtual Lab Validates COVID Nanobodies, Moves to Broader Biomedical Research

Researchers feed experimental data back into the AI-human system to refine molecular designs for future biomedical research.

Scientists Wesley Wu, Nash Bulaong, and John Pak (from left to right), pictured at their lab in San Francisco on July 25, 2025, all contributed to the Virtual Lab paper showing how a group of AI scientists can contribute to original research. (CZ Biohub San Francisco)
A message reading "AI artificial intelligence", a keyboard, and robot hands are seen in this illustration taken January 27, 2025. REUTERS/Dado Ruvic/Illustration
Image
Image

Overview

  • The Virtual Lab leveraged an LLM principal investigator and specialist AI agents to autonomously generate a computational pipeline that produced 92 nanobody candidates within two days.
  • Laboratory testing confirmed two AI-designed nanobodies that bind strongly to emerging JN.1 and KP.3 SARS-CoV-2 variants while retaining affinity for the ancestral spike protein.
  • A dedicated Scientific Critic agent and budget constraints limited human intervention to roughly 1% of the workflow, curbing errors without stifling AI creativity.
  • The system integrates tools such as ESM, AlphaFold-Multimer and Rosetta to facilitate rapid, interdisciplinary molecular design.
  • With validated candidates in hand, researchers are feeding empirical results back into the AI agents and exploring the platform’s application across diverse biomedical questions.