Researchers from Boston Children’s Hospital, Harvard University, and OpenAI recently completed a study on using artificial intelligence to assist in diagnosing rare genetic diseases in children. Many families face years of uncertainty because their children have genomic data that specialists have been unable to interpret. Even with extensive testing, roughly half of these cases remain unsolved due to the difficulty of connecting fragmented medical records with rapidly evolving scientific literature.
The study involved 376 previously unsolved cases. Researchers used the OpenAI o3 Deep Research reasoning model to analyze clinical and genomic information. Instead of providing a final diagnosis, the model functioned as a research assistant, scanning for patterns and linking candidate explanations for human review. Physicians then performed the necessary clinical confirmation, resulting in a new diagnosis for 4.8% of the cases. This process does not replace the doctor. It provides specialists with focused leads they can investigate further.
The project highlights why periodic reanalysis is necessary. A patient’s genome does not change, but our scientific understanding of genes and variants does. As researchers link new genes to diseases and labs reclassify old variants, old medical records can suddenly contain the answers that were previously impossible to see. In some instances, the model helped identify connections that had been missed by earlier, traditional pipelines.
This work is a step toward making expert-led reanalysis more efficient. It shows that AI can act as an explanation-first reasoning layer, allowing humans to interrogate data more effectively. The study underscores that any diagnosis must come from qualified clinical experts using established, certified laboratory processes. While these initial results are modest, they represent a meaningful change for families who have searched for years for an answer. The next phase of this work, led by the Manton Center, aims to create a more scalable platform to help clinical teams keep pace with ongoing scientific discovery.