OpenAI and PerturbAI announce 32-minute AI analysis of mouse-brain data

The data came from a whole-brain CRISPR atlas showing how gene disruptions affect different types of mouse-brain cells. According to OpenAI’s account, each agent investigated a gene disruption, analyzed cellular responses, searched scientific literature and proposed explanations. Their reports were combined into what OpenAI calls a “reasoning atlas,” linking genes, cellular effects, publications and possible mechanisms.

A cholesterol question for researchers to test

One lead concerns DCTN1, a gene whose mutations cause Perry syndrome, a neurodegenerative disease. In the mouse-brain atlas, disrupting Dctn1 increased expression of genes involved in making cholesterol, OpenAI reports. The agents connected that observation with research suggesting problems with cholesterol transport may cause cells to make more cholesterol. They proposed testing whether DCTN1 disruption changes where cholesterol is distributed within neurons, whether that contributes to neurodegeneration and whether restoring normal distribution could protect neurons. Those possibilities are research questions, not demonstrated protective effects or a treatment.

OpenAI reports that, at the largest workload in its comparison, the distributed agents cited more than 700 references versus 11 for a single agent, with the model and resources held fixed. It also says blinded human experts preferred the distributed-agent reports 75% of the time over single-agent reports. Those comparisons concern analysis and reports, not clinical results. PerturbAI says the companies will now apply their framework to pursue drug targets using its data.