Figure 2: Analytical positive predictive value and sensitivity of variant calls from each vendor.
From: Measuring coverage and accuracy of whole-exome sequencing in clinical context
(a) Biallelic single-nucleotide variants (SNVs) concordant and discordant between V1 and V2 were sequentially checked for genotype quality (GQ) score, per-site depth of coverage (DP), and reported minor allele frequency (MAF) in the Exome Aggregation Consortium (ExAC) server. The number of concordant and discordant SNVs is shown, as is the average number and standard deviation of variants meeting each criterion across nine individuals. In parentheses, the same statistics are expressed as a percentage of the total number of variants seen in either vendor. (b) The scheme for calculating analytical positive predictive value (aPPV) and sensitivity. In each individual, a likely true-positive set of variants is compiled by aggregating all unique variants seen in both V1 and V2 (blue rectangle) that had GQ ≥20, DP ≥10, and (for vendor-specific variants only) MAF >0 in ExAC. All variants discovered by a vendor are used to calculate aPPV (the green shaded squares divided by the red rectangle) and sensitivity (the proportion of the green shaded squares divided by the blue rectangle). (c) The mean and standard deviation across nine individuals for aPPV and sensitivity for each variant type and vendor. FN, false negative; FP, false positive; TP, true positive; WES, whole-exome sequencing.
