We calculated pairwise FST using Weir and Cockerham’s estimator [32 (link)] over 1,000 bootstraps through the hierfstat package for R [33 ]. Due to the relative nature of FST estimates, we conducted a second analysis to compare our pairwise FST estimate to the pairwise values between other free-breeding dog populations. For this second analysis, we used open-source SNP data acquired and published by Pilot et al. ([34 (link), 35 ]; hereafter, Pilot et al. dataset) and used hierfstat to also calculate pairwise FST amongst free-breeding dog populations across Europe and Asia. Pilot et al. [34 (link)] genotyped 324 individuals with the CanineHD Whole-Genome Genotyping BeadChip (Illumina), so, to maintain consistency in our comparisons, we filtered each SNP dataset to retain only those SNPs included on both the Illumina and Affymetrix arrays and which were present in at least 50% of individuals in both datasets. This filtering resulted in 147,592 SNPs (147 K set). We recalculated pairwise FST for Chernobyl City and Nuclear Power Plant populations and then calculated pairwise FST for the free-breeding dog populations included in the Pilot et al. dataset, which included dogs sampled in Poland, Slovenia, Iraq, Saudi Arabia, Armenia, Central Russia, Eastern Russia, Kazakhstan, Tajikistan, China, Mongolia, and Thailand.
We used the hierfstat package to measure observed and expected heterozygosity for each locus, which we compared across populations with a paired t-test with the alpha corrected for multiple comparisons (corrected α = α / 301,023). We used the R package SNPRelate [36 (link)] to calculate individual inbreeding coefficients with Visscher’s estimator as described by Yang et al. [37 (link)] and compared measures between populations through Welch’s two sample t-test. As an additional measure of inbreeding, we analyzed runs of homozygosity (ROH) per individual with PLINK (–homozyg). For this, we utilized the 427 K Set for all 116 unique individuals, subset for each population, because minor allele frequency filtering can overlook homozygous stretches which limits ROH detection [38 (link)]. This also allowed for more comprehensive coverage for the scans. We selected parameters based on Sams and Boyko [39 (link)] and Morrill et al. [40 (link)] to best accommodate SNP coverage for our data set: ROH longer than 500 kb (–homozyg-kb 500), contained at least 50 SNPs (–homozyg-snp 50), and contained no heterozygous calls (–homozyg-window-het 0). Inverse density (Kb/SNP) and gap size thresholds were set high (–homozyg-density 5000, –homozyg-gap 1000) to ignore [39 (link)]. We then calculated each individual’s FROH, using the proportion of genome covered by ROH as a measure of inbreeding [38 (link)]. The FROH scores, averaged across populations, were compared using Welch’s Two Sample t-test.
We used the hierfstat package to measure observed and expected heterozygosity for each locus, which we compared across populations with a paired t-test with the alpha corrected for multiple comparisons (corrected α = α / 301,023). We used the R package SNPRelate [36 (link)] to calculate individual inbreeding coefficients with Visscher’s estimator as described by Yang et al. [37 (link)] and compared measures between populations through Welch’s two sample t-test. As an additional measure of inbreeding, we analyzed runs of homozygosity (ROH) per individual with PLINK (–homozyg). For this, we utilized the 427 K Set for all 116 unique individuals, subset for each population, because minor allele frequency filtering can overlook homozygous stretches which limits ROH detection [38 (link)]. This also allowed for more comprehensive coverage for the scans. We selected parameters based on Sams and Boyko [39 (link)] and Morrill et al. [40 (link)] to best accommodate SNP coverage for our data set: ROH longer than 500 kb (–homozyg-kb 500), contained at least 50 SNPs (–homozyg-snp 50), and contained no heterozygous calls (–homozyg-window-het 0). Inverse density (Kb/SNP) and gap size thresholds were set high (–homozyg-density 5000, –homozyg-gap 1000) to ignore [39 (link)]. We then calculated each individual’s FROH, using the proportion of genome covered by ROH as a measure of inbreeding [38 (link)]. The FROH scores, averaged across populations, were compared using Welch’s Two Sample t-test.
Free full text: Click here
