Patient-to-patient phenotype matching enables rare disease gene prioritization beyond curated gene-phenotype associations

Abstract

Molecular diagnosis of rare diseases is limited by incomplete knowledge of gene-disease associations, leaving patients with atypical presentations or emerging disorders undiagnosed. We introduce a phenotype-first diagnostic framework that uses previously diagnosed patients as empirical diagnostic evidence, shifting candidate gene prioritization from disease models and predefined gene-phenotype associations to direct patient-to-patient comparison. Implemented as SimPheny, this framework compares an undiagnosed patient with a reference cohort of diagnosed patients and uses the molecular diagnoses of phenotypically similar individuals to prioritize candidate genes, allowing patient-derived clinical evidence to inform diagnosis before association data is incorporated into curated resources. In solved Undiagnosed Diseases Network (UDN) cases, SimPheny outperformed existing prioritization approaches, with the greatest advantage for genes with limited or absent phenotype associations. Clinical review of high-confidence matches in previously unsolved UDN cases confirmed nearly half as diagnostic. These findings establish patient-to-patient matching as a scalable diagnostic strategy for phenotype-driven rare disease diagnostics.