Benchmarking of Whole Exome Sequencing and Ad Hoc Designed Panels for Genetic Testing of Hereditary Cancer

Abstract

Next generation sequencing panels have been developed for hereditary cancer, although there is some debate about their cost-effectiveness compared to exome sequencing. The performance of two panels is compared to exome sequencing. Twenty-four patients were selected: ten with identified mutations (control set) and fourteen suspicious of hereditary cancer but with no mutation (discovery set). TruSight Cancer (94 genes) and a custom panel (122 genes) were assessed alongside exome sequencing. Eighty-three genes were targeted by the two panels and exome sequencing. More than 99% of bases had a read depth of over 30x in the panels, whereas exome sequencing covered 94%. Variant calling with standard settings identified the 10 mutations in the control set, with the exception of MSH6 c.255dupC using TruSight Cancer. In the discovery set, 240 unique non-silent coding and canonic splice-site variants were identified in the panel genes, 7 of them putatively pathogenic (in ATM, BARD1, CHEK2, ERCC3, FANCL, FANCM, MSH2). The three approaches identified a similar number of variants in the shared genes. Exomes were more expensive than panels but provided additional data. In terms of cost and depth, panels are a suitable option for genetic diagnostics, although exomes also identify variants in non-targeted genes.

Figure 1. Pedigrees of the families in which a putative pathogenic mutation was identified in the panel genes.

Filled quarters of symbols indicate patients affected by cancer (each color refers to a specific type). Current age, age at death and age at diagnosis (in brackets), when available, are also detailed. Putative pathogenic mutations are shown at the top of each pedigree; proband is marked by an arrow, carrier status was studied in available relatives, and those carrying/not carrying the variant are marked with +/− respectively. CRC, colorectal cancer; GBM; glioblastoma; Liposarc., Liposarcoma; LOH, loss of heterozygosity; MDB; medulloblastoma; NOS, not otherwise specified cancer; STS, soft tissue sarcoma.

Figure 2. Theoretical and observed coverage of the 132-gene Diagnostic Region of Interest (DxROI): base percentage of the DxROI of the 132 genes targeted by any of the panels and the exome, covered by the three different sequencing strategies.

(a) Theoretical coverage. Percentage coverage of the DxROI for each gene is obtained by comparing the designed target regions, as provided by each manufacturer (TSCP and WES) or aimed for in the I2HCP design. (b) Observed coverage. Percentage of DxROI bases of each gene effectively covered at a read depth ≥30x (C30) by each strategy; the median of the 24 samples is shown.

Figure 3. Comparison of main coverage metrics.

Average of all samples and 95% confidence interval are shown in each bar plot for the three approaches. (a) Performance metrics: passing filter (PF) reads; percentage of on-target reads, defined as any read overlapping at least one base the target region defined by the corresponding approach, versus total PF reads; percentage of off-target reads, defined as those within regions more than ±500-bp outside the designed own target regions; and uniformity of the coverage of the 83-gene DxROI (Diagnostic Region of Interest), calculated as the fraction of CCDS coding exons plus 20 bp boundaries reaching a mean read depth within ±70% of the mean read depth over all coding exons plus 20 bp boundaries. (b) Mean read depth, percentage of bases with read depth at least 30x and 10x versus: own target regions, the whole 83-gene DxROI, or considering the coding bases and their 20-bp boundaries separately.