China’s DeepRare AI Sets Record in Disease Diagnosis

A joint research team from Shanghai Jiao Tong University has unveiled DeepRare, an AI-powered system that sets a global record for diagnostic accuracy in identifying rare diseases ^[1][2]. The system utilizes an "agentic" architecture to analyze clinical phenotypes and genetic data through a traceable reasoning process ^[2][3]. Testing showed that DeepRare achieved 57.18 percent accuracy using only clinical information, a significant improvement over previous international models ^[1]. When genetic sequencing was included, accuracy rose to over 70.6 percent, outperforming expert physicians in head-to-head evaluations ^[2][3]. Unlike traditional "black-box" AI, DeepRare provides a clear evidence chain for every conclusion, allowing clinicians to verify its logic ^[1][3]. The system is already in use by over 1,000 professionals across 600 medical institutions worldwide ^[1].