Evaluation of different in silico tools for the assessment of deleterious variants in acute myeloid leukemia

Abstract Background The utilization of genome/exome sequencing for managing cancer patients is rising. However, deciphering the genomic variations and determining their pathogenicity can be intricate. The widely accepted practice of using in silico pathogenicity predictions as evidence when interpreting genetic variants is an integral part of standard variant classification guidelines. Several algorithms have been developed and evaluated to predict deleterious variants. The objective of this study was to assess the performance of 34 pathogenicity prediction tools (such as BayesDel, CADD, ClinPred, DANN, DEOGEN2, Eigen-PC, FATHMM, GERP++, M-CAP, MetaLR, MutationAssessor, MutationTaster, MutPred, Polyphen2, PROVEAN, REVEL, and SIFT) on the latest version of ClinVar dataset and implement it on the exome sequence data of acute myeloid leukemia (AML) patients to assess the performance to these tools in clinical samples. Results While predicting the pathogenicity of genetic variants, there were 6 in silico tools having specificity > 0.9 and 14 tools having sensitivity > 0.9 on the ClinVar dataset. Further, three tools BayesDel, MetaRNN, and ClinPred demonstrated highest accuracy achieving sensitivity 0.9337–0.9627 and specificity 0.9245–0.9513. By applying these 3 tools on the present study AML exome dataset, 1421, 1235, and 2033 potential deleterious variants in 410 AML-associated genes were observed, respectively. Conclusion This comparison highlighted the in silico tools to predict the potential pathogenicity of the variants which otherwise might have been classified as variants of uncertain significance (VUS). The finding can help in the genetic risk assessment and targeted therapeutic approaches in AML.