AI Agents Analyze Mouse Brain Data in 32 Minutes

OpenAI, in collaboration with biotech startup PerturbAI and The Scripps Research Institute, conducted large-scale biological data analysis from mouse cells. By deploying thousands of AI agents simultaneously, the team successfully processed millions of mouse cell data points in just 32 minutes.

An AI agent is an AI program that autonomously makes decisions and takes actions toward a given objective. Rather than using a single AI model, multiple agents work in parallel, dividing tasks to process vast amounts of data in a short time. What distinguishes this effort from previous experimental attempts is the scale of "thousands" of agents operating in coordination.

In biomedical research, cell-level data analysis traditionally required expert teams working over several months. As the variety and volume of data have increased, the burden of analysis has grown, becoming a significant bottleneck affecting research speed. This effort directly addresses that constraint, and the fact that the same task was completed in 32 minutes stands out even as a simple processing speed comparison.

The Scripps Research Institute is an independent research organization specializing in infectious disease, neuroscience, and immunology, covering everything from basic science to drug development. PerturbAI is known as a startup specializing in life sciences data analysis using AI, and this collaboration brought together three entities—an AI company, a startup, and a research institution—each contributing their respective expertise.

The significance of this effort extends beyond speed alone. If data at this scale can be processed in short timeframes, researchers will be able to test a vastly greater number of hypotheses. Many scientific discoveries depend on the frequency of trial and error, and reducing analysis wait time can accelerate this cycle.

The application of AI agents extends beyond chat and document generation to encompass such specialized data processing. Going forward, verification of analytical accuracy and potential applications to other species and disease domains will be key focal points. It remains to be seen whether the design of AI agents premised on large-scale parallel processing will transform the very approach to life sciences research.