Fig. 2: Performance comparison.
From: De novo and somatic structural variant discovery with SVision-pro
a, Comparison of the Mendelian consistency in six family datasets (above) and the twin discordancy in the ChineseQuartet (below). SVision-pro is compared with Sniffles2 (multisample mode) and SVision, cuteSV and debreak (followed by SURVIVOR and Jasmine merging). Each box contains six and three values for HiFi and ONT, respectively (Supplementary Table 5). The boxplot defines the median (Q2, 50th percentile), first quartile (Q1, 25th percentile) and third quartile (Q3, 75th percentile). The bounds of the boxplot, representing interquartile range (IQR), are between Q1 and Q3. The minimum and maximum values are defined as Q1 − 1.5× IQR and Q3 + 1.5× IQR, respectively. The whiskers are values between minima and Q1 and between Q3 and maxima. Values falling outside the Q1–Q3 range are plotted as outliers of the data. b, Comparison of the CSV Mendelian genotype consistency on the simulated trio data. SVision-pro was compared with state-of-the-art CSV caller SVision (followed by SURVIVOR and Jasmine merge). c, In the six families, SVision-pro correctly genotyped a complex locus comprising both an SSV and a CSV. Three distinct alleles are found by SVision-pro, including homologous SSV, homologous CSV and mixed heterozygous SSV and CSV. d, Comparison of the number of de novo calls in the six family datasets. e, Overlapping of 90 de novo calls produced by Sniffles2 with all calls produced by SVision-pro. f, Recall values on the previously published somatic SV callset of HCC1395 tumor-normal paired cell lines. g, The number of somatic SVs and the false-positive rates produced by Vapor validation decrease as the supporting read number increases. h, SVision-pro identified a nonsomatic complex locus that had been reported as a somatic SSV. SVision-pro revealed that the paired normal genome exhibited a heterozygous SSV and CSV, whereas the tumor genome exhibited homozygous CSV.
