This research demonstrates how combining Chain-of-Thought (CoT) reasoning with Retrieval Augmented Generation (RAG) significantly improves rare disease diagnosis from unstructured clinical notes.

• Achieved 72.1% top-10 accuracy for genetic disease diagnosis directly from clinical notes
• Outperformed standard LLM approaches by explicitly prompting for phenotype extraction and reasoning
• Demonstrated that external knowledge retrieval complements LLM reasoning capabilities
• Validated results on real hospital notes, showing potential for clinical implementation

This breakthrough matters because rare disease diagnosis typically takes 5-7 years, with AI-powered approaches potentially reducing diagnostic delays, improving patient outcomes, and reducing healthcare costs.

Full paper: Integrating Chain-of-Thought and Retrieval Augmented Generation Enhances Rare Disease Diagnosis from Clinical Notes