AI agents revolutionize biological data analysis, uncovering new disease insights

OpenAI, PerturbAI, and Scripps Research have deployed thousands of AI agents to analyze biological data from millions of mouse brain cells, identifying potential leads for Alzheimer’s and Parkinson’s research in just 32 minutes. The agents, powered by GPT-Rosalind, examined a whole-brain CRISPR atlas, connecting gene disruptions with scientific literature to uncover unexpected biological connections. This approach, termed distributed biological reasoning, enabled the agents to analyze data far more efficiently than traditional methods, which could take months or even years. Human experts preferred the distributed-agent interpretations over single-agent reports in 75% of blinded comparisons. The research revealed insights such as differences in how Tsc1 and Tsc2 affect brain cells, and a potential link between cholesterol regulation and neurodegeneration. One key finding involved DCTN1, a gene linked to Perry syndrome, where disrupting Dctn1 increased cholesterol-related gene expression. This led to new hypotheses about cholesterol distribution in neurons and its role in neurodegeneration. Scientists emphasized that AI can help turn complex data into testable ideas, accelerating the discovery of new treatments for brain diseases.