Measuring Coverage and Accuracy of Whole Exome Sequencing in Clinical Context

PURPOSE: To evaluate the coverage and accuracy of whole exome sequencing (WES) across vendors. METHODS: Blood samples from three trios underwent WES at three vendors. Relative performance of the three WES services was measured for breadth and depth of coverage. The false-negative rates (FNR) were estimated using the segregation pattern within each trio. RESULTS: Mean depth of coverage for all genes was 189.0, 124.9 and 38.3 for the three vendor services. Fifty-five of the ACMG 56 genes, but only 56 of 63 pharmacogenes were 100% covered at 10x in at least one of the nine individuals for all vendors; however, there was substantial inter-individual variability. For the two vendors with mean depth of coverage >120x, analytic positive predictive values (aPPV) exceeded 99.1% for SNVs and homozygous indels, and sensitivities were 98.9 – 99.9%; however, heterozygous indels showed lower accuracy and sensitivity. Among the trios, FNRs in the offspring were 0.07 – 0.62% at well-covered variants concordantly called in both parents. CONCLUSION: The current standard of 120x coverage for clinical WES may be insufficient for consistent breadth of coverage across the exome. Ordering clinicians and researchers would benefit from vendors’ reports that estimate sensitivity and aPPV, including depth of coverage across the exome.