Evaluation of Bayesian Classification Framework on the Variant Classification of Hereditary Cancer Predisposition Genes

AbstractPurposeTo assess the differences in the variants classifications using the ACMG/AMP 2015 guidelines and the Bayesian point-based classification system (here referred to as point system) in 115 hereditary cancer predisposition genes and explore the utility of the point system in variant reanalysis.MethodsGermline variant classifications from 721 pediatric patients were evaluated using the two scoring systems and compared with our reported classification.Results2376 unique variants were identified. The point system exhibited a propensity to decrease the rate of variants of unknown significance (VUS) to 15% compared to 36% by the ACMG/AMP 2015 (Cochran-Armitage with Z-score of −16.686; p-value < 0.001). This reduction in VUS rate is attributed to 1) single benign supporting evidence (12%); 2) single benign strong evidence (4%), each of which independently could downgrade a VUS to likely benign in the point system; and 3) resolving conflicting evidence or evidence not recognized by the ACMG/AMP 2015 guidelines (5%). Examination of the point system scores of the reported VUS (28%) facilitated tiering and prioritizing the variants for reanalysis.ConclusionThe point system facilitates the reduction of the VUS rate and provides a systematic way for periodic reanalysis of VUS in hereditary cancer predisposition genes.