OpenAI's GPT-Rosalind proposes cholesterol transport as a neurodegeneration lead

OpenAI says its GPT-Rosalind model suggested that restoring cholesterol transport could protect neurons from degeneration.

The suggestion draws on mouse-brain data linking the DCTN1 gene to Perry syndrome: disrupting DCTN1 raised expression of cholesterol-making genes, which the model read as a sign that cholesterol may not reach the neurons that need it. The analysis was run with PerturbAI and Scripps Research, and OpenAI says the collaboration is expanding to identify new drug targets.

A follow-up post by Rohan Paul describes the setup: 1,946 GPT-Rosalind-5.5 agents each analyzed one knockout's effects and searched the literature, without seeing the atlas paper, which appeared after the model's December 2025 knowledge cutoff. Paul says the agents matched or refined 48 of 55 published findings the expression data could support, and that blinded PhD reviewers preferred their reports 75% of the time. The cholesterol link is a model-generated hypothesis; no report says it has been tested or confirmed.