S 6 Structural Variant Calling

S6.1 Genome-wide Structural Variant Detection We used whole-genome shotgun paired-end sequence data generated with both Illumina and Applied Biosystems SOLiD platforms from the genomes of six canid samples (including a additional Basenji only sequenced to low coverage on the Illumina platform, but excluding the Chinese wolf), to estimate the fraction of the genome with segmental duplications. Our goal was to determine potentially duplicated regions to filter out for the final SNP call set. We identified the segmental duplication (SD) content in these genomes using the Whole-genome Shotgun Sequence Detection (WSSD) approach [1]. This strategy is based on determining regions with a significant excess of depth of coverage. Briefly, WGS reads are allowed to map to multiple locations to a reference genome, and therefore we expect that paralogous copies map into all locations. Highly identical duplicated genomic regions would be detected with an excess of depth of coverage. In our case, we used the dog assembly (canFam2) downloaded from the UCSC Genome Browser. Repeats detected by RepeatMasker and simple tandem repeats with period smaller than 12 detected by the Tandem Repeat Finder were pre-masked. We aligned the Illumina reads allowing 94% of sequence identity using mrFAST v2.0.0.5 [2] and SOLID reads with drFAST v0.0.0.3 [3]. We calculated the absolute copy numbers of non-overlapping windows of 1 kb of unmasked sequence using mrCaNaVaR version 0.31 (http://mrcanavar.sourceforge.net/). We identified SDs as regions with at least 5 consecutive windows with a copy number higher than 2.5. We detected between 1,379 and 1,413 SD segments larger than 10 kb in the five genomes we analyzed. These regions comprise 52.77 to 55.01 Mb in total that correspond to 2.09% to 2.17% of the reference assembly (Table S6.1.1).