AI-Designed Nanobodies Accelerate Battle Against SARS-CoV-2 Variants

Recent advancements in artificial intelligence have led to groundbreaking developments in biomedical research, exemplified by the creation of AI-powered virtual laboratories. Researchers at Chan Zuckerberg Biohub and Stanford University have developed a platform where autonomous AI agents, each with specialized scientific roles, collaborate seamlessly to address complex problems such as designing effective nanobodies against emerging SARS-CoV-2 variants. This platform allows a human scientist to initiate and oversee research, while the AI team independently formulates hypotheses, conducts virtual experiments, and refines its strategies in real-time.

In a study published in Nature on July 29, 2025, the researchers describe how this AI-driven Virtual Lab platform operates. It features a Principal Investigator AI agent that leads a team of specialized AI agents, emulating the collaborative environment of human scientific labs. Powered by large language models, these agents possess reasoning and decision-making capabilities, enabling them to tackle open-ended research questions.

The team applied this innovative system to design nanobodies—tiny proteins similar to antibodies—that can bind to unique spike proteins of new SARS-CoV-2 variants. Within days, the virtual lab produced dozens of candidate nanobodies, two of which demonstrated promising binding capabilities when tested in the lab of researcher Kyle Swanson.

This achievement marks a significant leap in AI applications in biomedical sciences, showing that autonomous AI agents can work together to solve real-world scientific challenges quickly and effectively. The collaborators highlight that such AI systems not only augment human expertise but also foster diverse perspectives, leading to more innovative solutions.

The Virtual Lab platform also includes a Scientific Critic agent, which questions findings and improves the robustness of the results. The majority of discussions and decisions are made by the AI agents, with humans intervening at key points. This collaboration resulted in the identification of 92 novel nanobodies, with two promising candidates against SARS-CoV-2 variants.

The success of this project demonstrates the potential for AI to revolutionize scientific research, moving beyond assistive tools to become primary drivers of discovery. Zou and Pak emphasize that their system is adaptable to various scientific disciplines, promising a new era of accelerated research powered by AI teamwork.