Figure 5.

AccuSNV demonstrates high accuracy and precision on additional real-world bacterial sequencing data sets. We evaluated SNV concordance between AccuSNV and five other tools across four published bacterial data sets with curated SNVs (Supplemental Fig. S13). UpSet plots (center) display the SNVs overlap count of tool-identified SNVs and reported SNVs. The red-outlined dash box highlights a group of four SNVs identified by all tools except AccuSNV and the original study (corresponding IGV screenshot is shown in the top right panel) (Robinson et al. 2011). The IGV screenshot includes two sample tracks, one carrying the reference allele and one carrying the alternative allele, with the variant position highlighted by a red box. Red reads indicate an inferred insert size larger than expected, which may suggest the presence of a deletion or a possible alignment artifact. The bottom panels display IGV screenshots of five unique SNVs identified by other methods but not supported by AccuSNV or the original study. The colored borders around each IGV screenshot correspond to the dashed boxes in the UpSet plot, indicating the group from which each SNV originated. Below each IGV screenshot is a description of the source of errors, highlighting typical error signatures such as misalignments near large structural variants, indels, and paralogous alignments (as reported by Koboldt 2020). We note that these error modes are not generally tool specific but that these tool-specific SNVs happen to have these error modes. Also see Supplemental Figures S14–S17.

1889f05