PerturbAI and OpenAI Jointly Develop a Parallel Research Framework with Nearly 2,000 AI Agents
Biotechnology company PerturbAI announced a collaboration with OpenAI and Scripps Research to develop an AI reasoning framework in which nearly 2,000 AI agents work in parallel to analyze complex biological experimental data, search the literature, and propose testable hypotheses. When analyzing whole-brain CRISPR maps, the system completed the integration of data and research findings in 32 minutes; the company estimates that humans working serially would need months to a year. In blind-review comparisons against reports from a single AI agent, experts preferred the system's explanations 75% of the time. The collaboration has lasted about a year, and next the two parties will jointly conduct new drug target discovery using causal biological data from living systems; the relevant leads still require follow-up experimental validation.
Why read this
AI score 56
The original text provides the specific workflow of multi-agent collaboration and expert blind-review preference data, offering a look at the current state of AI-assisted scientific research.
Topics
Body · AI translation
OpenAI Teams Up with PerturbAI: Nearly 2,000 AI Agents Conduct Research Together to Accelerate New Drug Development
Phoenix Tech Download the App for Exclusive Early Access
October 21, 2026 21:31:34 From Beijing
Updated October 21, 2026 21:47:24
PerturbAI and OpenAI, together with Scripps Research, have developed an AI reasoning framework in which nearly 2,000 AI agents work in parallel to help scientists analyze complex biological experimental data, search the literature, and propose testable hypotheses. When analyzing a whole-brain CRISPR atlas, the system completed the integration of data and research findings in 32 minutes; the company estimates that humans working serially would need months to a year. In an expert blind evaluation, 75% preferred the reports generated by the system. The two parties plan to jointly carry out new drug target discovery using in vivo causal data.
Phoenix Tech News (Author/Yu Lei) On October 21 Beijing time, biotechnology company PerturbAI announced the latest progress of its collaboration with OpenAI. The two parties, together with Scripps Research, developed an AI reasoning framework that uses nearly 2,000 AI agents working in parallel to help scientists analyze complex biological experimental data, search the literature, and propose hypotheses that can be tested in follow-up experiments.
According to PerturbAI, when analyzing its whole-brain genomic atlas, the system completed data analysis and the integration of research findings in 32 minutes. The company estimates that if humans were to carry out the analysis and literature review serially, the related work could take months to a year. This time comparison comes from the company's estimate and is not a contemporaneous controlled test against a human team.
In the specific workflow, AI agents separately analyze cellular responses, search scientific literature, link experimental results with existing biological knowledge, and then aggregate them into a "reasoning graph" that connects genes, cellular responses, existing research, and possible biological mechanisms, providing scientists with leads for further investigation.
The research team also invited biology experts to evaluate the reports in terms of novelty, degree of evidence support, priority for follow-up research, and confidence. In a blind evaluation comparison with reports generated by a single AI agent, experts preferred the explanations given by the multi-agent system 75% of the time. This result reflects experts' preference for the reports and cannot be directly understood as the accuracy rate of the research conclusions.
The data used in this analysis came from the whole-brain CRISPR atlas published by PerturbAI in March this year, covering about 8 million cells. The company says the atlas was generated within months and has now been downloaded more than 40,000 times. Since then, PerturbAI has also built proprietary atlases covering metabolic disease and aging research.
The collaboration between the two parties has lasted for about a year. Next, they will use causal biological data generated by PerturbAI in living systems to jointly carry out new drug target discovery. The related results have been published through a joint preprint and blog; the results disclosed so far mainly involve data interpretation and research hypothesis generation, and the proposed leads still require follow-up experimental validation.