Evaluation ofin silico toolsfor variant classification in missense variants of solid cancer with actionable genetic targets

Advancement in next-generation sequencing technologies has led to a rise in discovery of variants of uncertain significance, which are not clearly categorized as pathogenic or benign. In silico tools, which have been developed to help classify these variants, exhibit variations in outcome. This study aims to evaluate the performance of 6 widely-used in silico tools in predicting the pathogenicity of drug-actionable gene variants in 9 solid cancers. We selected drug-actionable genes according to NCCN guidelines on breast, ovarian, colorectal, melanoma of skin, thyroid, bladder, pancreatic, prostate, and biliary cancer. From these genes, we gathered information on 1161 total missense variants (pathogenic = 606, benign = 555). Each variant’s pathogenicity was determined based on assertions from three databases: ClinVar, OncoKB, and My Cancer Genome. We selected variants with one or more concordant databases and excluded variants with conflicting classifications. The performance of the in silico tools (Align-GVGD, CADD, FATHMM, MutationTaster2021, Polyphen-2 (HumDiv), and Polyphen-2 (HumVar)) was evaluated by calculating and comparing their overall accuracy, sensitivity, specificity, and Matthews correlation coefficient (MCC). Overall, all of the in silico tools demonstrated high sensitivity (0.738-0.927) and moderately-high accuracy (0.555-0.829). Excluding MutationTaster2021, all tools demonstrated low specificity (0.242-0.559) and MCC (0.107-0.413). MutationTaster2021 exhibited the highest performance overall and across solid cancer types. Conversely, Align-GVGD exhibited low performance overall and across cancer types. Tools demonstrating high sensitivity (CADD: 0.983, MutationTaster2021: 0.927, REVEL: 0.851) could be used to rule out the pathogenic variants. MutationTaster2021, with a comparatively high specificity (0.721), could be considered as an additional test to rule in the pathogenic variants. However, given the varying performance and limitations of the tools according to solid cancer type, clinicians should remain cautious in their usage.