OpenAI's o1 model is beginning to show real-world promise in one of medicine's thorniest problems: identifying rare genetic disorders that often take patients years to diagnose.
Geneticist Catherine Brownstein has been testing the system on cases that typically demand extensive specialist review and detective work. The results suggest the AI can dramatically compress what is usually a grueling diagnostic timeline.
Rare genetic diseases present a particular challenge for medicine because they are uncommon enough that many clinicians have limited exposure to them. Patients often shuttle between providers, undergoing repeated tests and consultations before landing on answers. The diagnostic odyssey can stretch across a decade or longer.
What makes o1 potentially valuable here is its reasoning capability. Unlike earlier language models that can generate plausible-sounding but sometimes inaccurate information, o1 is designed to work through complex logical problems step by step. When applied to genetic data and clinical symptoms, it can consider multiple pathways and cross-reference genetic databases in ways that accelerate human analysis.
Brownstein's work suggests the model can help physicians narrow possibilities and identify patterns that might otherwise require hours of manual research. The speed gain alone matters for patients desperate for answers. But equally important is the potential to reduce missed diagnoses and misidentification in a space where time and expertise are genuinely scarce.
The model is not replacing geneticists. Instead it functions as a research accelerant, handling the computational heavy lifting so experts can focus on clinical judgment and patient context. That division of labor could reshape how rare disease diagnosis works in practice.
Author Emily Chen: "This is the kind of unglamorous, high-impact use case that actually justifies the AI hype, not another chatbot feature."
Comments