Fig. 1

Overview of the Study Design. First, SNPscan assay and HL gene panel sequencing of samples from 22,125 cases and 7,258 controls from the CDGC project were performed. After routine bioinformatic analysis and multidisciplinary panel discussion, variants were classified into five categories: pathogenic (P), likely pathogenic (LP), variants with uncertain significance (VUS), likely benign (LB), and benign (B), based on the ACMG/AMP-HL guidelines, with minor modifications. VUS were further classified into benign-leaning VUS (BL-VUS) and pathogenic-leaning VUS (PL-VUS), according to the ACGS Best Practice Guidelines. Then, samples and variants were filtered to obtain a homogeneous dataset, to minimize biases in defining the PS4 thresholds due to heterogeneity in phenotypes, inheritance patterns, and genetic background. Next, a truth set was generated by removing PL-VUS and variants with conflicting interpretation between CDGC and external databases (DVD, HGMD, and ClinVar). Further, the truth set was divided into three subsets to evaluate and optimize PS4 evidence. For truth subset 1 and 3, positive likelihood ratio (LR⁺) was calculated for each tested cutoff by counting the number of disease-causing and non-pathogenic variant above or below the tested cutoff. In truth subset 2, local positive likelihood ratio (Lr⁺) was estimated by comparing the density ratio of odds ratio (OR) distributions between disease-causing and non-pathogenic variants within a given interval. The lower boundary of LR⁺ /lr⁺ was utilized to determine evidence strength by matching with the thresholds defined for each subset. Finally, the adjusted PS4 criteria were applied in variant reclassification and patient reanalysis. PTA, pure tone audiometry; MDT, multidisciplinary team; GQ, genotype quality; DP, depth; AF_case, AF in cases; TP, true positive (number of P/LP variants above a tested cutoff); FP, false positive (number of BL-VUS/B/LB variants above a tested cutoff); TN, true negative (number of BL-VUS/B/LB variants below a tested cutoff); FN, false negative (number of P/LP variants below a tested cutoff); TPR, true positive rate (sensitivity); TNR, true negative rate (specificity); PPV, positive predictive value; NPV, negative predictive value