CHOP Researchers Unveil RareDAI Tool to Streamline Genetic Testing

The tool, detailed in the journal NPJ Digital Medicine, utilizes large language models fine-tuned to process seven critical clinical questions. These cover variables ranging from developmental history and congenital abnormalities to cost and accessibility constraints. By codifying these factors, RareDAI reduces the human variability that often plagues diagnostic workflows when guidelines lack highly specific protocols.

Senior study author Kai Wang, a professor of pathology and laboratory medicine at CHOP, noted that the system provides a transparent audit trail for every recommendation. Unlike standard AI models, this framework documents its reasoning, allowing medical professionals to review the logic before proceeding with a test. In performance evaluations, RareDAI outperformed traditional fine-tuned models by up to 20% in accuracy and precision. The project, supported by multiple National Institutes of Health grants, involved collaborative efforts with experts from Columbia University and Boston Children’s Hospital to ensure the model aligns with consistent, evidence-based clinical standards.