OpenAI agents scan mouse brain data for disease leads in 32 minutes
Original title: Using thousands of AI agents to uncover new leads in biological data
Thousands of GPT-Rosalind agents built a reasoning atlas linking gene disruptions, literature, and proposed mechanisms.
OpenAI, PerturbAI, and Scripps Research used thousands of GPT-Rosalind agents to interpret a whole-brain CRISPR atlas from millions of mouse cells in 32 minutes. The agents linked gene-disruption effects to literature and proposed mechanisms, including differences between Tsc1 and Tsc2 and a possible cholesterol-neurodegeneration connection. The team says comparable human review could take months to a year. A reasoning atlas is available for researchers.
Unconfirmed: The disease connections are proposed research leads, not validated mechanisms or treatments.
- A single agent cited 11 references; distributed agents cited more than 700.
- Blinded human experts preferred the distributed-agent interpretations.
- Tsc2 disruption changed dozens of genes in cerebellum neurons; Tsc1 did not.
- The findings are leads for further experiments, not confirmed treatments.
- Vendors
- OpenAI
- Organizations
- PerturbAIScripps Research
- AI models
- GPT-Rosalind
From the original article
5 min read at openai.com
See how OpenAI, PerturbAI, and Scripps Research used distributed AI agents to uncover promising research leads in data from millions of brain cells.
openai.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.