OpenAI, PerturbAI and Scripps run 1,946 GPT-Rosalind agents on a mouse brain atlas in 32 minutes

OpenAI, the drug-discovery company PerturbAI and Scripps Research said on October 21 that they had run 1,946 GPT-Rosalind agents in parallel to analyze existing data from millions of mouse brain cells. The analysis took 32 minutes. GPT-Rosalind is OpenAI's model for life sciences research, introduced earlier this year. The work is described in a bioRxiv preprint. The team applied an approach it calls distributed biological reasoning to a whole-brain CRISPR atlas, a dataset from experiments that disrupt genes and measure how cells change in response. Such atlases often contain more observations than any one team can study in depth, which is one reason researchers share them. The authors estimate that comparable human analysis and literature review would take months to a year. That figure is illustrative and based on assumed times of 15 minutes per gene disruption and 10 minutes per paper read. The agents' output consists of hypotheses that still need laboratory testing, not confirmed findings. The study shows one way large numbers of AI agents could help researchers make more use of public biological datasets.