Horizon: CNV interpretation through rapid automated ACMG-aligned pathogenicity analysis

Abstract

PurposeOur study assesses the Horizon model, a novel CNV classification tool developed in line with American College of Medical Genetics (ACMG) guidelines, to enhance the classification of pathogenicity in CNVs. MethodsHorizon utilizes a ranking-based algorithm, incorporating multiple proprietary databases and variant inheritance models as per ACMG standards. The models effectiveness was verified through Area Under the Curve (AUC) analyses on three datasets comprising 696 pathogenic inherited or de novo variants, as classified by clinical geneticists and several established tools. ResultsHorizon achieved an AUC of 0.97 in the discovery cohort, demonstrating high accuracy in CNV interpretation and proficiency in predicting pathogenicity. We observed an AUC of 0.87 in the de novo variant cohort and an overall AUC of 0.94 across all cohorts, surpassing tools like ClassifyCNV and AnnotSV. It showed particular effectiveness in interpreting duplication CNVs and the highest performance for CNVs sized 3-5 Mb. ConclusionThe Horizon model offers robust and accurate CNV interpretation, outperforming existing tools and aligning closely with clinical evaluations. Its comprehensive approach, integrating a range of genomic features and following ACMG guidelines, makes it a crucial tool in the genomic interpretation landscape, facilitating the rapid and accurate diagnosis of genetic disorders.