Allele Counts and Admixture Analysis of Dog Breeds

Allele counts for the dog breeds and wild canids reported in Boyko et al. Boyko:2010fk were downloaded from http://genome-mirror.bscb.cornell.edu/ on July 30, 2011. These data consist of counts of reference and alternate alleles at 61,468 sites in 85 dog breeds and wild canids. We removed the Jackal and Scottish Deerhound for having relatively high amounts of missing data, and the village dogs because it is unclear if they represent a coherent population. We also removed all SNPs on the X chromosome. This left us with 60,615 SNPs in 82 populations. We ran TreeMix with a window size ( ) of 500. This corresponds to a window size of approximately 20 Mb. For all TreeMix analyses, we set the coyote as the outgroup.
The ascertainment scheme used for SNP discovery in dogs was complicated [68] (link). The largest set of SNPs were ascertained by virtue of being different between the boxer and poodle assemblies. This should lead to an overestimation of the distance between the boxer and the poodle in our analysis. Indeed, in Figure 5B, a considerable negative residual between the boxer and poodle is visible. Another set of SNPs were ascertained by being heterozygous within a boxer individual, and a third set were ascertained by comparison between a boxer and wild canids. These latter SNPs should lead to an overestimation of the distance between the boxer and the wolf in our analysis (as we see for the poodle); in fact, we infer migration between the boxer and the wolf. This ascertainment issue may have led us to underestimate the amount of gene flow in the comparison.