**2. Materials and Methods**

A total of 15 specimens of *Capra caucasica*, including East-Caucasian tur from Dagestan (E\_TUR, *n* = 5), West-Caucasian tur from Karachay-Cherkessia (W\_TUR, *n* = 5), and Mid-Caucasian tur from Kabardino-Balkaria (M\_TUR, *n* = 5) were selected for this study. The sampling sites are presented in Figure 2. All the samples were taken from trophy hunters of the Mountain Hunters Club (www.kgo-club.ru), who were licensed to hunt Caucasian tur. DNA extraction was carried out using Nexttec columns (Nexttec Biotechnology GmbH, Leverkusen, Germany) in accordance with the manufacturer's recommendations. SNP genotyping was performed using the Illumina Goat SNP50 BeadChip containing 53,347 SNPs.

SNP quality filtering was performed in PLINK 1.9 [25]. SNPs with unknown positions, located on sex chromosomes as well as those that were genotyped in less than 90% of individuals (− −geno 0.1), with a minor allele frequency (MAF) < 1% (− −maf 0.01) and with deviations from Hardy–Weinberg equilibrium (− −hwe 1e−6) were removed.

For the population genetic analyses (PCA, Admixture, f3 statistics, etc.,) SNPs in linkage disequilibrium (− −indep-pairwise 50 5 0.5) were pruned. The observed (*H*O) and unbiased expected (*H*E) heterozygosity [26], inbreeding coefficient (*F*IS), and rarified allelic richness (*A*R) were calculated in the R package "diveRsity" [27]. Pairwise *F*ST genetic distances [28] were calculated in the R package "StAMPP" [29]. Principle component analysis was performed with PLINK 1.9 (− −pca 4) and visualized in the R package "ggplot2" [30]. An individual tree based on the pairwise identity-by-state (IBS) distance matrix (− −distance 1-ibs) was constructed using the Neighbor-Net algorithm implemented in SplitsTree 4.14.6 [31]. To estimate and visualize the distribution of heterozygosity at the individual level, multilocus heterozygosity (MLH) was calculated in the R package "inbreedR" [32]. Venn diagram was constructed in the R package "VennDiagram" [33]. Cluster analysis was performed in Admixture 1.3 software [34], and visualized in the R package "pophelper" [35]. f3 statistics [36] was computed using ADMIXTOOLS [37], as implemented in the R package "admixr" [38]. The map with sampling sites was created using R packages "maps" [39] and "ggplot2". As long as the Illumina Goat SNP50 BeadChip was developed for domestic goats (*Capra hircus*), the positions on chromosomes in this paper correspond to the reference genome of a domestic goat.
