Using AI to Reanalyze Unsolved Rare Disease Cases and Find New Diagnoses

Even after genomic sequencing and specialist review, roughly half of people with rare diseases never get a clear genetic diagnosis.

Their medical data contains clues buried across fragmented records, thousands of variants, and rapidly evolving literature.

Researchers from Boston Children’s Hospital’s Manton Center, Harvard, and OpenAI used OpenAI o3 Deep Research to revisit 376 previously unsolved cases, showing that evidence-linked hypotheses from a reasoning model can surface diagnoses that had evaded years of expert analysis.