Agentic RAG Achieves 93.55% Accuracy in Genetic Disease Severity Classification
September 18, 2026
An autonomous agent combining ReAct and RAG classifies over 10,000 Human Phenotype Ontology terms using ACMG-endorsed guidelines. The system retrieves PubMed literature to generate reasoning chains, achieving 93.55% accuracy with high claim support from direct evidence.
HOW THIS AFFECTS YOU
●
researcherThe integration of structured medical guidelines with RAG provides a framework for verifiable clinical reasoning.
●
healthThis demonstrates highly accurate automated screening for genomic disease severity.