A Chinese artificial intelligence system named DeepRare has demonstrated superior performance in diagnosing rare diseases, according to results published this week in the journal Nature. The system aims to reduce the prolonged diagnostic delays faced by patients whose symptoms are often scattered and resemble multiple other conditions.
DeepRare was developed by researchers from Shanghai Jiao Tong University and Xinhua Hospital. Instead of simply matching symptoms against a database, the system uses what developers describe as an agent based workflow. It forms initial hypotheses, tests them against clinical evidence, and then recalibrates probabilities before presenting the most likely diagnosis. This process mirrors traditional clinical reasoning based on suspicion, testing, and revision.
Using only clinical information, DeepRare achieved a top diagnosis accuracy rate of 57.18 percent, significantly outperforming competing tools. When genomic sequencing data was added, accuracy increased to 70.6 percent in complex cases. By comparison, the internationally recognized mutation analysis tool Exomiser achieved 53.2 percent accuracy under similar conditions.
Beyond accuracy, the study highlighted the system’s ability to generate clear reasoning chains. According to the report in Nature, expert agreement with the system’s diagnostic logic reached 95.4 percent. In a field often criticized for opaque algorithmic decision making, this level of explainability represents a structural shift.
Traceable reasoning enables the system to function as a clinical decision support tool rather than a replacement for physicians. It also reduces professional and legal friction when integrating AI into hospital environments.
The demand for such systems reflects persistent diagnostic gaps. A survey conducted by China Alliance for Rare Diseases involving more than 20,000 patients found that 42 percent had previously received a misdiagnosis. The average wait time for a confirmed diagnosis was 4.26 years.
In Europe, published data indicate an average delay of approximately 4.7 years, with more than half of patients waiting over six months after first seeking medical assistance before receiving a diagnosis.
Multiple specialist referrals increase financial costs and psychological strain. Delayed diagnosis frequently postpones therapeutic or preventive intervention. Reducing diagnostic timelines from years to months would produce both clinical and economic impact.
DeepRare has been available on an online diagnostic platform since July 2025. More than 600 medical institutions worldwide have registered to use the system. It is currently undergoing internal evaluation at Xinhua Hospital prior to formal adoption for rare disease diagnostic quality monitoring.
This expansion coincides with broader integration of AI into electronic health records and clinical decision support systems. However, regulatory structures remain incomplete. Reports in Nature Digital Medicine note that governance frameworks for medical AI are still developing, despite more than 1,240 AI enabled medical devices having received regulatory approval.
Adoption disparities persist between urban medical centers and rural regions, indicating that technological advancement alone does not resolve structural healthcare inequality.