Using thousands of AI agents to uncover new leads in biological data
Distributed agents cited over 700 references versus 11 for a single agent and were preferred by experts 75 percent of the time in blind comparison.
OpenAI announced on October 21 a collaboration with PerturbAI and Scripps Research using thousands of GPT-Rosalind agents to analyze a whole-brain CRISPR atlas covering millions of mouse brain cells. The distributed biological reasoning approach completed in 32 minutes what typically takes a month to a year, cited over 700 references versus 11 for a single agent, and was preferred by human experts 75 percent of the time in blind comparison. Key findings included a cerebellum-specific divergence between Tsc1 and Tsc2 perturbation, and a proposed link between DCTN1 and cholesterol regulation relevant to Perry syndrome. The reasoning graph is open sourced on GitHub with the paper on bioRxiv.