Fig. 2

Properties and supporting evidence for accurate structural variant identification. a A pathogenic homozygous deletion in the Dystrophin gene. ClinSV’s default annotation tracks are: depth of coverage in an NA12878 control (DOC C), all reads in the sample (DOC), or reads with mapping quality ≥ 20 (DOC MQ20), the average mapping quality of aligned reads from the sample (MQ), coverage standard deviation from 500 controls (DOC SD), segmental duplications [39] (Seg-Dup), annotated ClinSV calls (ClinSV), discordant pairs (DP), split reads (SR), ClinSV variant calls from 500 controls (MGRB), variants from the Database of Genomic Variants (DGV), and RefSeq genes (Genes). b A representative false positive CNV deletion identified by GIAB: the 5′ breakpoint is a mobile element insertion (MEI), the 3′ breakpoint is located in a repeat as suggested by the lower read mapping quality, and there is no intervening drop in coverage or mapping quality. c A representative false positive duplication called by aCGH not supported by WGS. The region shown is deleted in 44% of the samples in the 1000 Genome Project and coverage is frequently altered in controls (see DOC SD), but not in the actual sample (see DOC). d A homozygous deletion identified by ClinSV with reduced DOC, lots of DP support, but no SR support. The CNV is flanked by a pair of repeats. e A model demonstrating how CNV surrounded by repeats can reduce DP and SR support. For repeats greater then read length but smaller than paired-end fragment size only SR support is reduced (top, as in panel d), for repeats greater than the fragment size, both SR and DP support are reduced (bottom). The dashed arrows indicate the expected mapping location of DPs and SRs if the genomic region was not repetitive