From the supplementary info.......
Unequal sample size of dog and wolf.
To reduce the risk of introducing biases due to unequal sample size in dog and wolf we have used a method to estimate FST 40
that takes differences in sample size into account.
Nevertheless, to test if the unequal dog and wolf representation may have contributed to variance of statistics across the genome, we redid the FST and HP analyses based on a random sub sample of the dog data such that the average coverage in dog equalled that in wolf (6.2x) (Fig. S10 and S11).
First, analysing the sub sampled data we detect 17 regions with Z(HP)<-5, out of which 12 overlap the 14 regions from the original analysis. The 2 regions from the original analysis that did not overlap a region in the sub sampling analysis still show a clear reduction in heterozygosity (Z(HP)<-4) in the sub sampled data.
Secondly, we find 20 regions with Z(FST)>5 in the sub sampled data, out of which 18 overlap the 35 regions detected in the original analysis. The 17 regions from the original analysis that do not overlap a region in the sub sampling analysis still show a clear increase in FST in the sub sampling analysis (8 regions have Z(FST)>4 and the remaining 9 regions have Z(FST)>3).
The majority of the original CDRs were thus identified as CDRs in the sub sampled data, and the remainder still stand out, although not significantly so. These results suggest that the unequal sample sizes of dogs and wolves likely have little effect on the overall results of our selection analyses.