Fig. 3

Accuracy of somatic variant detection based on capture validation of 18 cases with PCGP data. a Design of capture validation for measuring sensitivity and PPV of our analytical pipeline using somatic exonic SNV/indel detection as an example. For exonic SNV/indel, variants that passed the cross-validation filter and had adequate coverage from custom capture sequencing were considered “predicted positives”. Predicted positives that were validated by capture sequencing are considered true positives whereas all variants—including those that failed to be detected or those filtered by cross-validation filtering—that were validated by capture sequencing are considered “actual positives”. b Summary of PPV (true positives/predicted positives) for each variant type. The predicted positive variants for exonic indels and SNVs are the variants that pass the cross-validation filter, whereas other SNVs, which refers to high-confidence non-exonic SNVs, are based on WGS only; as such, they are reported separately. Our test does not report non-exonic indels, so they are omitted. c Summary of sensitivity (true positives/actual positives) for each variant type. For exonic SNV/indel, most of the variants are detected by our variant detection pipeline on both WGS and WES (WGS+WES) platforms; of those that are detected by our pipeline on one platform, results for WGS and WES were comparable, with slightly more detected by WGS. The “Missed” variants are the false negatives; they were removed by cross-validation filtering or only detected by PCGP