New research shows that a commercially available large language model helped identify previously missed rare disease diagnoses in pediatric patients.
According to a recent article from NBC News, researchers at Boston Children's Hospital reported that OpenAI's o3 Deep Research model assisted in identifying 18 new diagnoses among 376 children whose rare diseases had remained unexplained despite prior genomic analysis. The findings, published in NEJM AI, suggest AI could help laboratories revisit unresolved genomic cases more efficiently as scientific knowledge continues to evolve.
The research team analyzed patient genomes alongside clinicians' notes, symptom descriptions, and filtered lists of potentially relevant genes. Human experts reviewed every AI-generated finding before confirming diagnoses, reinforcing that the technology is intended to support—not replace—clinical interpretation.
"It's a total game changer," Catherine Brownstein, scientific director of the genetic investigations arm of the Manton Center for Orphan Disease Research at Boston Children's Hospital, told NBC News. She noted that although the AI identified new diagnoses in about 5% of previously unsolved cases, "considering how many times these had already been analyzed, that's a huge number, and each one means an answer for a family."
Potential implications for clinical laboratories
For clinical laboratory professionals, the findings point to a possible new role for AI in genomic workflows. As sequencing volumes continue to increase and more disease-causing genetic variants are discovered, laboratories face growing backlogs of unresolved cases that often require periodic reanalysis.
For example, a May 2025 study in the Journal of Translational Medicine noted that despite the use of sequencing technologies, approximately 60% of rare disease cases remain unsolved.
Rather than replacing laboratory geneticists or molecular pathologists, AI could help prioritize variants, identify newly published gene-disease associations, and accelerate literature reviews that traditionally require significant manual effort. Such capabilities may become increasingly valuable as laboratories expand genomic testing services while facing ongoing workforce shortages.
The researchers emphasized that expert oversight remains essential. Independent experts interviewed by NBC News also cautioned that large language model results require rigorous human review before any diagnosis is reported to patients. In addition, seven of the diagnoses identified during the study were "rediscoveries" of findings that had previously been made elsewhere but were not broadly shared, highlighting continuing challenges with data sharing across the rare disease community.
Although additional validation studies will likely be needed before AI-assisted genome analysis becomes routine, the Boston Children's Hospital experience suggests these tools could eventually help clinical laboratories improve diagnostic yield while reducing the time required to reanalyze complex genomic cases that have remained unsolved for years.
This article was created with the assistance of Generative AI and has undergone editorial review before publishing.
