A groundbreaking artificial intelligence system is transforming rare disease diagnosis, with DeepRare demonstrating 70.6% accuracy in complex cases when combining clinical symptoms and genomic data. Published in Nature this week, the findings reveal significant improvements over existing tools like Exomiser (53.2%), offering hope for faster diagnoses in underserved regions.
The system addresses a critical gap: A China Alliance for Rare Diseases survey of 20,000 patients found 42% faced misdiagnoses, with diagnoses taking over four years on average. DeepRare's unique 'agentic' approach mimics human reasoning – forming hypotheses, testing evidence, and refining conclusions – rather than simple symptom matching.
Since its July 2025 deployment on a global diagnostic platform, over 600 medical institutions have adopted the tool. Researchers plan to validate results using 20,000 real-world cases through a new international alliance this year, potentially accelerating care for 300 million people affected by rare diseases worldwide.
