PerturbAI and OpenAI Launch AI Reasoning Agents for Biomedical Discovery
A year-long collaboration involving scientists at PerturbAI, OpenAI, and their other institutions yielded an AI reasoning framework that turns datasets into testable hypotheses, with thousands of agents analyzing whole-brain genomic data in minutes instead of months. The post PerturbAI and OpenAI Launch AI Reasoning Agents for Biomedical Discovery appeared first on GEN - Genetic Engineering and…
PerturbAI and OpenAI have unveiled AI reasoning agents designed to assist scientists in interpreting vast experimental datasets, connecting discoveries to existing research, and proposing testable hypotheses. The breakthrough is detailed in a preprint titled "Distributed AI agents enable comprehensive discovery of biological insights at scale."
The framework, developed over a year-long collaboration with scientists from Scripps Research Institute, University of California, San Diego, and Howard Hughes Medical Institute, involved deploying thousands of agents to examine cellular responses, search scientific literature, and link experimental outcomes with established biological knowledge.
This resulted in a reasoning atlas that connected genes, cell responses, literature findings, and potential biological mechanisms. Remarkably, the system achieved these feats in just 32 minutes, analyzing data that would otherwise take much longer. The researchers tested the agents' interpretations against those of human experts in a blind comparison, finding that the AI agents were preferred 75% of the time.
The collaboration aims to extend this technology to drug-target discovery, leveraging PerturbAI's proprietary causal biological data to uncover new drug targets more efficiently.
Written by urgent.news from GEN Biotechnology's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
