Figure 3: Vendor-specific false-negative rates in trio.
From: Measuring coverage and accuracy of whole-exome sequencing in clinical context
(a) For each trio, we calculate number of vendor-specific false negatives in offspring by focusing on the completely concordant variant calls across all three vendors (3CON) autosomal loci where one parent is heterozygous (i.e., 0/1–0/1–0/1 for V1, V2, and V3, respectively) and the other is reference concordant homozygous. We restrict the analysis to genomic regions covered with ≥10 × by all three vendors. (b) A Venn diagram shows each subgroup of concordant and discordant variant calls for proband. The black solid line represents the total number of variants discovered by two or more vendors, which we use as a denominator to calculate vendor-specific false-negative rates (FNRs) in each trio. (c) FNRs and the numbers of 3CON, concordant variant calls between two vendors (2CONs), and vendor-specific heterozygous variants found in offspring. For instance, in trio A, the FNR for V1 (0.07%) is calculated as the number of loci where V1 did not detect the heterozygous variant (eight variants, corresponding to “f”) divided by the total number of likely true set of heterozygous variants (i.e., 2CONs and 3CON) in the proband (23 + 68 + 8 + 10,907 = 11,006, corresponding to “d + e + f + g”).
