Using AI to help physicians diagnose rare genetic diseases affecting children

Researchers from Boston Children's Hospital's Manton Center for Orphan Disease Research, Harvard University, and OpenAI utilized the OpenAI o3 Deep Research reasoning model to help diagnose rare genetic diseases in children. Genomic sequencing often leaves many rare disease patients undiagnosed due to the complexity of sifting through massive variant data, fragmented records, and rapidly evolving scientific literature. In a study published on June 18, 2026, in NEJM AI, the team analyzed de-identified clinical and genomic information from 376 previously unsolved cases that had evaded years of expert analysis. The model functioned as an explanation-first reasoning layer, connecting clinical features, inheritance patterns, variant evidence, and scientific literature into justifications for human review. The AI model did not make clinical decisions or diagnose patients directly. Instead, it produced evidence-linked hypotheses that specialists reviewed using standard clinical frameworks. Qualified experts verified candidates, laboratories performed confirmatory testing, and clinical teams returned results to families. Following this expert review and clinical confirmation, physicians established diagnoses in 18 cases, representing an additional diagnostic yield of 4.8%. Among these, seven were rediscoveries of diagnoses established elsewhere but missing from reviewed records, and one case involved inferring a structural event leading to a DiGeorge syndrome diagnosis. The model also identified multi-gene explanations and a potential novel mechanistic explanation for vitiligo. The findings highlight how AI-assisted research workflows can help experts manage the growing maintenance problem of rare-disease reanalysis as scientific knowledge continues to evolve.