OpenAI partners with Boston Children’s Hospital: Breakthrough in AI-assisted diagnosis of rare genetic diseases

OpenAI collaborates with Boston Children's Hospital to use o3 Deep Research models to assist in the diagnosis of rare genetic diseases. The research results were published in NEJM AI (New England Journal of Medicine AI sub-journal) and attracted widespread attention in the medical community.

The research team used the o3 Deep Research model to conduct genetic reanalysis of 376 cases of long-undiagnosed rare diseases. The model successfully identified 18 previously undetected diagnoses, a diagnostic rate of approximately 4.8%. This ratio has important clinical implications for the rare disease field—many patients have experienced years of diagnostic woes.

o3 Deep Research specializes in cross-analysis of large-scale medical literature and genetic data, and can discover disease-causing mutations that are easily missed by traditional analysis methods from massive genomic data. OpenAI said that this case demonstrates the unique value of AI in "long-tail medical problems" - although rare diseases have fewer patients per disease, the total number is huge, and AI-assisted diagnosis can significantly reduce the cost and time of diagnosis.

The publication of NEJM AI provides top academic endorsement for this case. AI diagnosis of rare diseases is one of the scenarios with the highest ROI in medical AI - although the patient base is small, there is an urgent need for diagnosis and a strong willingness to pay. OpenAI's continued deep penetration in medical vertical fields (from HealthBench to rare disease diagnosis) is building a complete product narrative: from general model capabilities to professional domain verification.