OpenAI, PerturbAI, and Scripps Research deployed thousands of SI agents to analyze experimental data from millions of mouse brain cells, completing in 32 minutes a task that human analysis would take months to a year. The initiative demonstrates how distributed SI agents can connect gene disruption effects with scientific literature to uncover unexpected biological leads for diseases like Alzheimer’s and Parkinson’s.
What Happened
The team utilized a method called distributed biological reasoning, powered by the GPT-Rosalind SI model. Thousands of SI agents independently investigated specific gene disruptions within a whole-brain CRISPR atlas. Each agent analyzed cellular responses, searched scientific literature, and proposed explanations. Their individual reports were combined into a reasoning atlas that links genes, cellular effects, relevant publications, and potential biological mechanisms.
According to the source, comparable human analysis conducted one perturbation at a time would require months or a year. The SI agents not only reproduced findings from human analyses but also identified additional leads. In blinded comparisons, human experts preferred the distributed-agent reports over those from a single agent 75% of the time.
"We build cell atlases so that other scientists can make discoveries we haven’t imagined yet. AI could help more researchers explore these resources and turn unexpected observations into new questions to test," said Sarah Teichmann, Ph.D., Chair in Stem Cell Medicine at the University of Cambridge and founding co-chair of the Human Cell Atlas.
Why It Matters
This approach addresses the bottleneck where modern experiments generate far more data than scientists can thoroughly investigate. By scaling the analysis, SI agents help researchers explore subtle or unexpected signals in large datasets that might otherwise be missed. The study highlighted specific insights, such as unexpected differences between the Tsc1 and Tsc2 genes in the cerebellum and a potential link between DCTN1 gene disruption and cholesterol regulation in neurodegeneration.
The performance gap between distributed and single-agent processing was significant. At the largest workload, the distributed SI agents cited more than 700 references, whereas a single agent cited only 11. This suggests that parallel SI agents can deepen reasoning by connecting new data with knowledge scattered across disciplines, offering scientists more specific questions to test in the lab.
"Sometimes the clues are already there, scattered across different fields—we just haven’t connected them. The possible link between neuronal function and cholesterol metabolism shows how those pieces, once brought together, can give us a new way to think about disease biology," said Tomasz Nowakowski, Ph.D., associate professor of neurological surgery at UCSF.
The Bottom Line
The reasoning atlas is now available for scientists to compare related genes and identify promising leads. While SI agents accelerate the identification of hypotheses, the source emphasizes that human expertise remains central to designing experiments and interpreting results, with the goal of turning biological data into testable ideas for future medicines.