BEIJING -- When symptoms such as "noticing significant developmental delays in mobility, language, and social interaction since age two" are entered into a dialogue box, an AI large model can generate alerts about potential rare genetic disorders, such as Rett syndrome or Angelman syndrome, or complex neurodevelopmental conditions within seconds, alongside medical recommendations including specialized departments for consultation and necessary examinations.

This scenario emerged during a trial of China's first AI large language model dedicated to rare diseases, PUMCH-GENESIS, developed jointly by Peking Union Medical College Hospital (PUMCH) and the Institute of Automation under the Chinese Academy of Sciences.

The hospital said that public testing had recently begun on the model's ability to complete preliminary diagnosis consultations and appointment bookings.

While individual rare diseases are uncommon, their vast diversity creates significant diagnostic hurdles. Misdiagnosis and delayed confirmation remain critical challenges for patients, and AI tools like PUMCH-GENESIS are poised to address these systemic gaps.

Zhang Shuyang, president of PUMCH, explained that developing an AI-assisted diagnostic tool has been a focus for the hospital's expert team for rare diseases over the years.

Based on China's accumulated rare disease knowledge and genetic data from its population, PUMCH-GENESIS is the world's first rare disease model tailored to Chinese demographic characteristics. It enhances diagnostic accuracy and efficiency for clinicians while shortening confirmation timelines.

Traditional AI models face limitations in rare disease applications due to fragmented case data and scarcity of training samples. To overcome this, the research team pioneered a novel technical framework using an approach that integrates minimal initial data with clinical expertise to provide decision support throughout the diagnostic process.

Zhang underscored that PUMCH-GENESIS marks a transformative advancement in China's rare disease diagnostic infrastructure.

As the national leader in rare disease treatment, PUMCH will further drive the integration of AI and clinical practice to bolster primary healthcare capacities and refine the tiered medical system, extending hope for timely diagnosis and treatment to more families, Zhang said.

During the ongoing public testing phase, patients can access the model's preliminary consultation and appointment functions, engaging in multi-round dialogues to obtain initial diagnostic guidance. The next phase will introduce clinician-focused tools, including medical note generation, genetic interpretation, and hereditary counseling support.

Plans are underway to integrate PUMCH-GENESIS into the hospital's online multidisciplinary rare disease clinic, with eventual deployment to all member-hospitals of the national rare disease collaborative network, said the hospital.