OpenAI and PerturbAI Partner to Accelerate Drug Discovery with 2,000 AI Agents
PerturbAI and OpenAI, in collaboration with Scripps Research, developed a multi-agent AI reasoning framework that uses nearly 2,000 parallel AI agents to analyze complex biological data, search literature, and generate hypotheses. The system completed the data analysis and research synthesis of a whole-brain CRISPR atlas in 32 minutes. This framework demonstrates how massively parallel multi-agent AI systems can dramatically speed up biological data interpretation, potentially reducing workloads that usually take months or years down to minutes. It marks a significant step forward in applying LLM-driven reasoning agents to real-world biomedical research and drug target discovery. The analysis utilized PerturbAI's open-source whole-brain CRISPR atlas covering approximately 8 million cells. In a blind evaluation, human biology experts preferred the multi-agent system's explanations 75% of the time over single-agent reports, though these generated insights still require subsequent wet-lab experimental validation.
## BACKGROUND
Multi-agent systems utilize multiple autonomous AI components working together to handle complex, multi-step tasks by dividing labor between information retrieval, reasoning, and synthesis. Meanwhile, CRISPR technology enables precise gene editing and functional mapping in living organisms, generating massive datasets that traditionally require extensive manual analysis.