The SPARC classifier identified 260 hERG variants as high risk of pathogenicity with SPS ≥ 3.25, confirming structural predictions through functional validation.
Does the SPARC algorithm reliably predict the pathogenicity of hERG variants compared to existing tools?
The SPARC algorithm provides a scalable, cost-effective pre-screening tool to reliably stratify hERG variant pathogenicity and guide clinical interpretation for arrhythmogenic syndromes.
Absolute Event Rate: 0% vs 0%
Abstract Inherited mutations in the KCNH2 gene, which encodes the cardiac hERG potassium channel, are major contributors to arrhythmogenic syndromes such as Long QT and Short QT syndromes. However, clinical interpretation of the growing number of missense variants - many of which are classified as variants of uncertain significance (VUS) - remains a pressing challenge. Here, we present a semi-automated in silico pipeline for predicting hERG variant pathogenicity, acting as a binary classifier and integrating five structural metrics - residue volume, hydrophobicity, charge, steric clashes, and proximity to pathogenic hotspots - into a composite structural pathogenicity score (SPS) scaled from 1 to 5. Applied to 1727 hERG variants from ClinVar and from a French nationwide cohort, this binary classifier, termed SPARC, identified 260 variants as high risk of pathogenicity with SPS ≥ 3.25, of which a representative subset from the French cohort was functionally validated using high-throughput automated patch clamp. Functional phenotyping confirmed the structural predictions, including for several VUS, demonstrating that comprehensive structural scoring can reliably stratify variant pathogenicity. This approach, benchmarked with Alpha Missense and Revel, offers a superior scalable, cost-effective pre-screening tool to guide clinical variant interpretation and prioritization for experimental validation.
Chatelain et al. (Tue,) reported a other. The SPARC classifier identified 260 hERG variants as high risk of pathogenicity with SPS ≥ 3.25, confirming structural predictions through functional validation.