Abstract / Summary
Introduction Despite the identification of >700 genes associated with inherited kidney disease, many affected individuals presumably remain genetically undiagnosed through testing of known disease genes. Novel disease gene discovery requires interpretation of large-scale sequencing data, which remains challenging. Methods We developed the Nephro Candidate Score (N-CS), a machine learning (ML) framework for variant prioritization combining a Nephro Gene Score (N-GS), Nephro Variant Score (N-VS), and heuristic Inheritance Score. The ML-based N-GS and N-VS predict gene-disease relevance and variant pathogenicity from diverse genomic features. Clinical utility was assessed in 1452 individuals with suspected inherited kidney disease using exome-wide candidate discovery and gene-level burden analysis, with external validation using recently proposed inherited kidney disease genes. Results The N-CS showed strong predictive performance, with the XGBoost-based N-GS achieving an AUC of 0.94 and the Logistic Regression-based N-VS an AUC of 0.99 on independent test sets. A gene set enrichment analysis demonstrated significant enrichment of kidney-related biological processes among top-ranked genes (p < 0.001). In our patient cohort, N-CS-guided variant filtering recovered established disease genes and highlighted ERMP1 as a compelling candidate. Incorporating the N-GS in a polycystic kidney disease burden analysis reduced multiple-testing penalties, highlighting a moderate-effect candidate. Externally, the N-CS assigned high scores to 10 of 11 newly proposed kidney disease genes. Conclusion The N-CS is a digital framework to accelerate disease gene discovery and diagnosis. It is freely available as the NC-Scorer web tool (https://nc-scorer.kidney-genetics.org/) and a command-line interface enabling standardized, scalable candidate variant evaluation.