Population genetic structure based on whole genomes, Y-SNPs, and mitochondrial genomes

We conducted PCA for whole-genome SNPs using the program GCTA v1.24.2 (34). A maximum likelihood phylogenetic tree was constructed by RAxML software (35) with the GTRGAMMA model and 100 bootstraps, and the ascertainment bias correction was performed to correct for the impact of invariable sites in the data. Ferret was used as the outgroup (36). Population genetic structure was inferred by ADMIXTURE v1.23 software (37) with default settings. We did not assume any prior information about the genetic structure and predefined the number of genetic clusters (K) from two to seven. We used POPART v1.7 (38) to construct a median-joining network for the Y-SNP haplotypes and mitochondrial genome haplotypes. We constructed the phylogenetic tree based on mitochondrial genomes of 15,238 bp (excluding the D-loop region) using BEAST v1.8.2 (39) with ferret as the outgroup. The best substitution model of GTR + I was selected on the basis of the Bayesian Information Criterion by ModelGenerator v0.85 (40). A strict clock rate was selected on the basis of the assessment of coefficient of variation. A total of 8 × 10⁸ iterations were implemented with 10% burn-ins. The BEAST running results were assessed by Tracer v1.5 and were annotated by TreeAnnotator v1.10. We constructed the phylogenetic tree based on Y-SNPs data using the maximum likelihood method implemented in RAxML (35), with the ascertainment bias correction and ferret as the outgroup.