Distributed AI agents turn a brain-cell atlas into testable research leads

OpenAI, PerturbAI and Scripps Research report a 32-minute analysis of existing mouse-cell data; laboratory validation remains essential.

OpenAI, PerturbAI and Scripps Research have used thousands of AI agents to investigate existing experimental data from millions of mouse brain cells. In an October 21 account, OpenAI says the GPT-Rosalind-powered agents completed the analysis in 32 minutes and connected cellular responses with scientific literature.

The team divided work across gene perturbations and combined the reports into a reasoning atlas. One proposed lead concerns DCTN1 and cholesterol regulation in neurons. The account frames such connections as questions for follow-up experiments, rather than proven mechanisms or treatments.

A way to choose the next experiment

The contribution is the organisation of possible explanations. Large experiments can produce more observations than researchers have time to investigate individually. A system that links an observation with relevant literature may help a scientist decide which hypothesis deserves scarce laboratory time.

The 32-minute figure describes the reported computational analysis. It should not be read as the time required to conduct the original experiments, validate a biological mechanism or develop a medicine.

Comparing one agent with many

The team also compared a single agent with the distributed approach while keeping the model and resources fixed. At the largest workload, the distributed reports cited more than 700 references, versus 11 for the single agent. OpenAI says blinded experts preferred the distributed interpretations in 75% of comparisons.

Those measures concern coverage and expert judgement, not successful medicines. The DCTN1 lead asks whether disruption changes cholesterol distribution inside neurons and whether correcting that distribution could help. That is a specific experimental question, which is where computational synthesis becomes useful to a laboratory: it narrows what researchers might test next.

Sources

- [Using thousands of AI agents to uncover new leads in biological data](https://openai.com/index/scaling-ai-research-for-complex-biological-data/) — OpenAI