Stanford Virtual Lab Revolution: AI Scientists Design SARS-CoV-2 Nanobodies Autonomously

Researchers at Stanford University and Chan Zuckerberg Biohub San Francisco have achieved a groundbreaking milestone in artificial intelligence-driven scientific research with the creation of a "Virtual Lab" comprising AI agents capable of conducting complex, autonomous scientific investigations. Published in Nature on July 29, 2025, this revolutionary system successfully designed effective nanobodies against SARS-CoV-2 variants, demonstrating AI's potential to accelerate biomedical breakthroughs.

The Virtual Lab represents a paradigm shift from AI as a research tool to AI as an independent research entity. Unlike traditional AI systems that assist humans with specific tasks, this platform creates a team of AI agents that can formulate hypotheses, design experiments, and generate novel solutions with minimal human oversight.

Led by co-senior authors John Pak of CZ Biohub SF and Stanford's James Zou, the Virtual Lab platform enables a human user to create a "Principal Investigator" AI agent that assembles and directs a team of additional AI agents emulating specialized research roles found in science labs.

The system was put to the test with a pressing real-world challenge: designing antibodies or nanobodies to bind to the spike protein of new variants of the SARS-CoV-2 virus. After just a few days of autonomous collaboration, the Virtual Lab team had designed and implemented an innovative computational pipeline and presented the researchers with blueprints for dozens of binders.

Remarkably, two of these AI-designed nanobodies showed particular promise against new SARS-CoV-2 strains when subsequently tested in Pak's physical laboratory, validating the virtual team's work with real-world results.

The Virtual Lab platform includes several types of AI agents:

Each agent is powered by large language models (LLMs), giving them scientific reasoning and decision-making capabilities. The human researchers participated in AI scientists' meetings and offered guidance at key moments, but their words made up only about 1% of all conversations.

What impressed the researchers most was not just the AI team's efficiency, but their creativity. "The AI agents came up with a pipeline that was quite creative. But at the same time, it wasn't outrageous or nonsensical. It was very reasonable—and they were very fast," said Pak.

The AI agents decided independently that nanobodies would be a more promising strategy than traditional antibodies, providing sound scientific reasoning: nanobodies are typically much smaller than antibodies, making the machine learning scientist's job easier when computationally modeling and designing proteins.

The Virtual Lab team generated 92 new nanobodies, with experiments in Pak's lab finding that two bound effectively to the spike protein of recent SARS-CoV-2 variants. This represents a significant enough finding that Pak expects to publish dedicated studies on these nanobodies.

The AI-designed nanobodies demonstrated:

This breakthrough has profound implications for the future of scientific discovery:

Speed: AI scientists work continuously without fatigue, conducting hundreds of research discussions while human researchers sleep

Scalability: Virtual labs can be replicated and specialized for different research domains

Cost-effectiveness: Reduces the resources needed to explore multiple research directions simultaneously

Accessibility: Makes high-level interdisciplinary expertise available to researchers worldwide

Unlike "black box" AI systems, the Virtual Lab maintains complete transparency. All discussions result in transcripts that human team members can access and review, allowing researchers to understand exactly why certain decisions were made and probe further if needed.

"The agents and the humans are all speaking the same language, so there's nothing 'black box' about it, and the collaboration can progress very smoothly," Pak explained.

While the current platform is designed for biomedical research questions, modifications could allow it to be used across a much wider array of scientific disciplines. Zou envisions expanding the system to tackle challenges in materials science, environmental research, and drug discovery.

"We're demonstrating a new paradigm where AI is not just a tool we use for a specific step in our research, but it can actually be a primary driver of the whole process to generate discoveries," said Zou.

Rather than replacing human scientists, the Virtual Lab enhances human capabilities by handling routine analysis and generating more testable hypotheses. "This project opened the door for our Protein Science team to test a lot more well-conceived ideas very quickly," Pak noted.

The Virtual Lab represents a fundamental shift toward AI-human collaboration in scientific research, where artificial intelligence doesn't just assist discovery but actively drives it forward, promising to accelerate the pace of scientific breakthrough across multiple fields.