{"slug": "openais-gpt-rosalind-agents-tear-through-a-7-7-million-cell-brain-atlas-in-32", "title": "OpenAI’s GPT-Rosalind agents tear through a 7.7 million cell brain atlas in 32 minutes", "summary": "OpenAI, PerturbAI and Scripps Research ran nearly 2,000 GPT-Rosalind agents over a whole-brain CRISPR perturbation atlas of roughly 7.7 million mouse brain cells, processing 1,946 genetic perturbations in 32 minutes, OpenAI announced on October 21, 2026, alongside a preprint dated around October 16. The multi-agent system recovered 52 of 55 known findings, a 94.5% recovery rate, and human experts preferred its outputs over single-agent baselines 75% of the time; multi-agent reports cited more than 700 references versus 11 for single-agent models. The agents proposed a link between DCTN1 disruption and altered neuronal cholesterol transport and synthesis, and found that disrupting Tsc2 significantly altered gene expression in cerebellar neurons while Tsc1 did not, results the collaborators frame as a step toward joint drug-target discovery for neurodegenerative disease.", "body_md": "# OpenAI’s GPT-Rosalind agents tear through a 7.7 million cell brain atlas in 32 minutes\n\nOpenAI official logo (public domain, Wikimedia Commons) — CryptoBriefing brand treatment\n\nOpenAI, PerturbAI and Scripps Research used nearly 2,000 AI agents to generate new hypotheses about neurodegenerative disease\n\n[OpenAI](https://cryptobriefing.com/markets/openai/) has teamed up with PerturbAI and Scripps Research to point an army of AI agents at one of neuroscience’s messiest datasets. Nearly 2,000 GPT-Rosalind agents worked through a whole-brain CRISPR perturbation atlas built from around 7.7 million mouse brain cells.\n\nThe whole job took 32 minutes.\n\nThe goal is practical: find new leads for understanding and treating neurodegenerative diseases. The collaborators say the work is a stepping stone toward joint drug-target discovery.\n\n## What the agents actually did\n\nA CRISPR perturbation atlas is a map of what happens to cells when scientists deliberately disrupt specific genes. This atlas covered the whole mouse brain, with data from around 7.7 million cells.\n\nThe GPT-Rosalind agents processed 1,946 genetic perturbations in total. For each one, they connected the observed cellular responses to existing scientific literature, then proposed new hypotheses.\n\nThe output was also published as a public “reasoning atlas.” Rather than just handing over conclusions, the system documents how it got there.\n\nOpenAI announced the work on October 21, 2026, with a preprint dated around October 16.\n\n### AI, tech, and the markets they move—in one daily briefing.\n\nDaily. Free. Join 34,000+ readers across crypto, finance, and policy.\n\n## How well did it work\n\nThe team tested whether the multi-agent system could rediscover findings that scientists had already established. It recovered 52 of 55 known findings, a recovery rate of 94.5%.\n\nHuman experts also compared the multi-agent outputs against single-agent baselines. They preferred the multi-agent work 75% of the time.\n\nThe multi-agent reports cited more than 700 references, compared with just 11 for the single-agent models. Experts reportedly appreciated how comprehensive the reports were.\n\n## The hypotheses worth watching\n\nThe agents proposed a link between disruption of the DCTN1 gene and changes in how neurons transport and synthesize cholesterol. DCTN1 is associated with Perry syndrome, a neurodegenerative condition.\n\nThe agents also looked at disruptions to Tsc1 and Tsc2 in the cerebellum. Disrupting Tsc2 significantly altered gene expression in cerebellar neurons, while disrupting Tsc1 did not produce the same effect. Those results line up with known clinical patterns in tuberous sclerosis.\n\n## Where GPT-Rosalind fits\n\nGPT-Rosalind is OpenAI’s specialized reasoning model for the life sciences. It entered research preview in early 2026.\n\nThe partners bring the other pieces. PerturbAI contributes the perturbation-focused angle, and Scripps Research brings the biomedical research credentials.\n\n**Disclosure:** This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our [Editorial Policy](https://cryptobriefing.com/editorial-policy/).", "url": "https://wpnews.pro/news/openais-gpt-rosalind-agents-tear-through-a-7-7-million-cell-brain-atlas-in-32", "canonical_source": "https://cryptobriefing.com/openai-gpt-rosalind-neurodegenerative-disease-research/", "published_at": "2026-11-20 22:17:24+00:00", "updated_at": "2026-11-20 22:43:51.620574+00:00", "lang": "en", "topics": ["artificial-intelligence", "ai-agents", "ai-research", "large-language-models"], "entities": ["OpenAI", "PerturbAI", "Scripps Research", "GPT-Rosalind", "DCTN1", "Tsc1", "Tsc2", "Perry syndrome"], "also_reported_by": [], "takeaway": "A multi-agent GPT-Rosalind setup recovered 52 of 55 known findings (94.5%) and drew expert preference 75% of the time over single agents, while citing over 700 references versus 11.", "recommendation": {"advice": "Fan out specialized agents over large datasets and compare against single-agent baselines: the multi-agent reports won 75% of expert comparisons and cited 700+ references versus 11.", "playbook": "agent-architecture", "playbook_url": "https://wpnews.pro/playbooks/agent-architecture"}, "alternates": {"html": "https://wpnews.pro/news/openais-gpt-rosalind-agents-tear-through-a-7-7-million-cell-brain-atlas-in-32", "markdown": "https://wpnews.pro/news/openais-gpt-rosalind-agents-tear-through-a-7-7-million-cell-brain-atlas-in-32.md", "text": "https://wpnews.pro/news/openais-gpt-rosalind-agents-tear-through-a-7-7-million-cell-brain-atlas-in-32.txt", "jsonld": "https://wpnews.pro/news/openais-gpt-rosalind-agents-tear-through-a-7-7-million-cell-brain-atlas-in-32.jsonld"}}