Fig. 2. SVs reported in cancer databases also occur in healthy populations.

a–c, COSMIC contains 46,185 somatic deletions. STIX found evidence for 27.9% of these SVs in SGDP (a) and 27.5% in 1KG (b). In these two plots (and d and e), we summarize the population-level evidence for each recurring SV (blue dot) by the number of (Num.) samples with any concordant evidence (x axis) and the maximum (Max.) amount of per-sample evidence (y axis). c, Only 1% of COSMC SVs appeared in the 1KG SV call set. The agreement between the STIX and the 1KG call sets is plotted using the population frequency estimates from each method for each SV. d–f, PCAWG found 84,083 deletions, 3.4% of which were in SGDP (d) and 2% were in 1KG (e). f, The 1KG call set contained only 0.2% PCAWG SVs. g, A comparison of germline filtering strategies for 183 prostate tumor samples that remove tumor deletions found in matched-normal tissue (SV), the STIX index of 1KG, the 1KG SV calls and the gnomAD SV calls. Histograms show the frequency of sample-level SV counts. Red bars and text give the sample mean. For example, the raw tumor calls had, on average, 3,892.0 SVs and STIX filtering yielded, on average, 35.8 false positives, 23.1 true positives and 10.5 false negatives.