Somatic mutation detection workflow validity distinctly influences clinical decision

Abstract

AbstractIdentifying somatic mutations from tumor tissues holds substantial clinical consequences for making informed medical decisions. Evaluating the accuracy and robustness of somatic mutation analysis workflows has become essential when employing whole exome sequencing (WES) analysis in clinical settings. In the study, we utilized a set of tumor WES data the Sequencing and Quality Control Phase 2 (SEQC2) project to systematically benchmark the workflow analytical validity, including various combinations of read aligners and mutation callers. The read aligners included BWA; Bowtie2; built-in DRAGEN-Aligner; DRAGMAP; and HISAT2 as well as the callers Mutect2; TNscope; built-in DRAGEN-Caller; and DeepVariant. Among all combinations, DRAGEN showed the best performance with mean F1-score of 0.9659 in SNV detection, while the combination of BWA and Mutect2 showed the second highest mean F1-score of 0.9485. Notably, our results suggested that the mutation callers had a significantly higher impact on the overall sensitivity than the aligners. For drug-related biomarkers, Sentieon TNscope tended to underestimate tumor mutation burden and missed many drug-resistance mutations such as FLT3(c.G1879A:p.A627T) and MAP2K1(c.G199A:p.D67N). Our investigation provides a valuable guide for cancer genomic researchers on tumor mutation identification, accomplished through an in-depth performance comparison among diverse tool combinations.