Novel Dog Genome Content Revealed By Pseudo-De Novo Assembly of Unmapped Sequence Reads

Dog breeds have distinct genetic backgrounds due to historical breeding practices targeted towards desired traits. Intense selective breeding for behavioral and physical characteristics unintentionally unmasked many genetic diseases. The dog reference genome is relied upon to understand the genomic components of heritable disease, yet is based on a single female Boxer, remains poorly annotated, and lacks most breed-specific variants. To examine variation not included in the reference genome, we evaluated whole genome sequence from 26 Border Collies. Raw sequence reads were trimmed on quality score and aligned to the reference genome. 7% of paired reads aligned discordantly and 1% of paired and orphaned reads did not align to the Boxer genome and were de novo assembled using MIRA. Primary assembly produced an average of 454,934 contigs covering 143.5 Mb with an N50 of 301. All primary assembly contigs were assembled into a secondary assembly producing 124,488 contigs covering 125.9 Mb of non-redundant sequence with an N50 of 1158. BLASTn of 1000 randomly selected secondary assembly contigs against NCBI's non-redundant nucleotide database showed that 28% of contigs have no known matches in sequenced species and 9% of contigs preferentially align to the human genome over the dog genome. This study reveals novel genomic content present in Border Collies that does not exist in the Boxer reference genome. Extending the dog genome by including information found in unmapped reads may prove critical in understanding the breed identity and the genetic component of diseases shared by humans and dogs.