OpenAI, PerturbAI, and Scripps say thousands of GPT-Rosalind agents analyzed data on millions of mouse brain cells in 32 minutes, a job that takes humans months

This capability cuts the time to generate biological hypotheses from months to minutes.

OpenAI, in collaboration with PerturbAI and Scripps Research, has developed a novel method called distributed biological reasoning. This approach utilizes thousands of AI agents powered by GPT-Rosalind to analyze a large dataset of millions of mouse brain cells. The agents collectively processed experimental data, investigated gene disruptions, consulted scientific literature, and proposed biological explanations. The outcome is a publicly accessible reasoning atlas that links genes to cellular effects, relevant publications, and potential mechanisms. This process, reportedly completed in 32 minutes, would typically take humans months or even a year. The research reproduced known findings and uncovered new leads, including unexpected differences in gene interactions and potential links to neurodegenerative diseases. Human experts validated the novelty and scientific value of the AI-generated hypotheses.

The application of thousands of specialized AI agents to biological data analysis signifies a paradigm shift from single-model approaches to distributed, task-specific AI systems. This move allows for parallel processing of massive datasets, dramatically accelerating discovery timelines and potentially uncovering complex interactions that might be missed by a monolithic analysis. The validation by human experts on novelty and scientific value suggests that AI is not just a tool for speeding up known processes but is becoming a genuine partner in scientific ideation and hypothesis generation. This success in biological research could pave the way for similar distributed reasoning frameworks in other complex scientific domains requiring massive data processing and literature synthesis.

This development has significant implications for pharmaceutical research, genetics, and neuroscience, democratizing the ability to derive actionable insights from vast biological datasets. Researchers can now potentially explore hypotheses and drug targets with unprecedented speed, transforming the pace of drug discovery and personalized medicine. The public availability of the reasoning atlas and the underlying framework democratizes access to advanced analytical capabilities. Future research will likely focus on refining these distributed agent systems, expanding their application to other omics data, and integrating them more deeply into experimental design and validation workflows.