Comprehensive analysis of population structure and genomic diversity in fifteen Anatolian sheep breeds Authors: Mervan Bayraktar, Murat Durmuş, Özge Uyguner, Murat Turan, B Devrim Özcan, Nazan Koluman Published: 2026 ABSTRACT AbstractBackgroundCharacterizing genomic diversity and population structure is essential for the conservation and sustainable utilization of indigenous sheep genetic resources. Anatolia represents an important center of sheep diversity, harboring multiple populations shaped by diverse breeding histories, ecological conditions, and management practices. In this study, we performed a comprehensive genome-wide assessment of fifteen Anatolian sheep populations using SNP-based data to investigate their genomic diversity, population structure, and genetic relationships.ResultsGenetic diversity indices revealed substantial heterogeneity among populations in terms of heterozygosity, allelic richness, minor allele frequency, proportion of polymorphic loci, and inbreeding coefficients, indicating marked differences in within-population genomic variation. Analysis of molecular variance showed that most of the total genetic variation was distributed within populations, although a significant proportion was also attributable to among-population differences. Pairwise FST estimates and principal component analysis demonstrated a structured yet heterogeneous genomic landscape, with some populations exhibiting clear differentiation and others clustering more closely. Phylogenetic analyses further supported this pattern through differences in monophyletic clustering, branch-length dispersion within populations, and patristic distances between populations. Admixture analysis identified three major ancestry components and revealed population-specific differences in ancestry composition and admixture levels. Runs of homozygosity (ROH) analyses showed considerable variation among populations in both the abundance and length distribution of homozygous segments, which was also reflected in class-specific ROH-based genomic inbreeding coefficient (FROH) profiles. Gene flow estimates additionally indicated heterogeneous connectivity across populations.ConclusionsTogether, these findings demonstrate that Anatolian sheep populations retain diverse and structured genomic backgrounds despite varying levels of differentiation, admixture, and genomic inbreeding. This study provides a robust genomic framework for understanding the evolutionary relationships among Anatolian sheep populations and offers valuable insights to inform future conservation, management, and breeding strategies.Keywords: Population structure, Genomic diversity, Anatolian sheep, SNP genotyping, Runs of homozygosity, Gene flow Background Understanding the genetic diversity and population structure of sheep breeds is of both scientific and practical importance. From an evolutionary perspective, patterns of genomic variation and relatedness among breeds can illuminate domestication processes, historical migrations, and breed formation [1, 2]. From an animal breeding and conservation standpoint, assessing genetic diversity is crucial for the sustainable management of genetic resources. High within-breed genetic diversity is associated with greater adaptive potential and a lower risk of inbreeding depression, both of which are crucial for long-term fitness and productivity [3–5]. Local or indigenous breeds, in particular, often harbor unique gene combinations conferring adaptation to their native environments – for example, genes linked to climate resilience (heat or cold tolerance), disease resistance, or the ability to thrive on sparse pastures [6, 7]. These locally adapted breeds can be invaluable for developing resilient production systems, especially under changing environmental conditions and emerging diseases. However, many such breeds are currently at risk due to population declines, uncontrolled cross-breeding, and the replacement by a few high-output commercial breeds. Recognizing these concerns, international organizations like the Food and Agriculture Organization (FAO) have advocated for the conservation of indigenous livestock breeds as a priority [4]. The loss of genetic diversity within domestic species is warned to have serious consequences, potentially diminishing future options for breed improvement and adaptation. In this context, systematically characterizing the genetic diversity and structure of local sheep breeds is a key step toward their effective conservation and utilization [8, 9]. Turkiye is one of the world’s cradles of sheep diversity and remains a center of rich ovine genetic resources. The country currently manages an estimated 42 million sheep, encompassing nearly 30 recognized native breeds or breed types [10]. These breeds represent a broad spectrum of phenotypes – from fat-tailed, hardy meat breeds adapted to the arid steppes, to thin-tailed dairy breeds thriving in more temperate coastal regions – reflecting the varied geography and husbandry practices across Anatolia. For instance, the fat-tailed Akkaraman, renowned for its ability to endure the harsh continental climate of Central Anatolia, is the most populous breed in Turkiye, accounting for roughly 40% of the national sheep flock. Other major breeds, such as the Eastern Anatolian Morkaraman and the Karakaş from the Lake Van region, are likewise well-adapted to high-altitude nomadic grazing and smallholder farming conditions. Despite their distinct regional adaptations, Turkish sheep breeds have not evolved in isolation [11–14]. Genetic studies indicate considerable gene flow and shared ancestry among many of these breeds. A microsatellite-based analysis of Turkish sheep reported that only ~ 9% of the total genetic variation was partitioned between breeds (with the vast majority within breeds), and observed extensive admixture and introgression among local populations [15]. Such findings suggest a complex population structure shaped by both historical connections and more recent breeding practices. However, most earlier genetic surveys of Anatolian sheep have been limited in scope, typically examining a few breeds at a time and relying on mitochondrial DNA or a modest number of microsatellite markers [12, 15–21]. Until recently, no comprehensive genome-wide study had been undertaken to assess the full breadth of genetic variation across Turkiye’s sheep breeds. This represents a critical knowledge gap, given the country’s pivotal role in sheep domestication and its responsibility for a large share of the species’ genetic diversity. The present study aims to evaluate genetic diversity, population structure, phylogenetic relationships, and ROH-based genomic inbreeding across fifteen native Anatolian sheep breeds using 50 K SNP data, with the broader goal of providing genomic evidence to support genetic management, conservation prioritization, and the sustainable use of Anatolian sheep genetic resources. Materials and methods Study populations and sampling The animal material used in this study comprised 359 sheep representing fifteen native Anatolian sheep breeds recognized as distinct breeds or locally maintained breed populations within Turkey. The analyzed populations were Akkaraman (AKN, n = 20; combined production type: meat and milk), Awassi (AWS, n = 32; combined production type: milk and meat), Çukurova Meat Sheep (CMS, n = 19; meat-type), Eşme (ESM, n = 22; combined production type: meat and fertility), Gökçeada (GKD, n = 19; combined production type: milk and meat), Hamdani (HAM, n = 25; combined production type: meat and milk), Hemşin (HMN, n = 20; combined production type: meat and milk), Kangal Akkaraman (KLA, n = 26; combined production type: meat and milk), Koçeri (KOC, n = 20; combined production type: meat, milk, and fertility), Karayaka (KRK, n = 23; combined production type: meat, wool, and milk), Kıvırcık (KVK, n = 24; combined production type: meat, milk, and wool), Morkaraman (MKN, n = 27; combined production type: meat and milk), Norduz (NDZ, n = 28; combined production type: meat and milk), Sakız (SKZ, n = 25; combined production type: milk and fertility), and Zom (ZOM, n = 29; combined production type: meat, milk, and fertility). The sampled breeds were assigned to geographical groups according to their traditional breeding regions and main distribution areas. Detailed geographical information, including the provinces and regional locations of the sampled populations, is provided in Fig. 1. Briefly, the populations represented the Marmara, Aegean, Central Anatolia, Mediterranean, Black Sea, Southeastern Anatolia, and Eastern Anatolia regions. Geographical distribution of the fifteen native Anatolian sheep breeds included in the study (The map shows the sampled breeds according to their traditional breeding regions and sampling provinces within Türkiye. The Black Sea and the Mediterranean Sea are indicated to clarify the northern and southern geographical boundaries of the sampling framework. The sampled populations were grouped as follows: Marmara region: Gökçeada (GKD) and Kıvırcık (KVK); Aegean region: Sakız (SKZ) and Eşme (ESM); Central Anatolia: Akkaraman (AKN) and Kangal Akkaraman (KLA); Mediterranean region: Çukurova Meat Sheep (CMS); Black Sea region: Karayaka (KRK) and Hemşin (HMN); and Eastern–Southeastern Anatolia: Awassi (AWS), Zom (ZOM), Morkaraman (MKN), Hamdani (HAM), Koçeri (KOC), and Norduz (NDZ)) To ensure representativeness, animals were selected from their native or characteristic breeding regions based on geographical origin, breed-typical phenotypic characteristics, and available flock information. The main inclusion criteria were clear breed assignment, geographical representativeness, and suitability of the collected biological material for downstream DNA extraction and genotyping. To maximize within-breed diversity and minimize sampling bias, unrelated or minimally related animals were preferentially sampled. Close relatives were avoided during the sampling stage as far as possible based on breeder information, flock records when available, and field knowledge. Animals were not included when breed identity was uncertain, when sample quality was insufficient, or when close relatedness to already sampled individuals was strongly suspected. Whole-blood samples were collected from the jugular vein by licensed veterinarians using standard animal handling and restraint procedures. Approximately 10 mL of blood was collected from each animal using tubes containing ethylenediamine tetra-acetic acid (EDTA) and stored at − 20 °C until DNA extraction. Animal handling and sampling procedures conformed to national and institutional guidelines for the care and use of animals in research and were approved by the relevant Institutional Animal Care and Use Committee (Approval No. 13-9-2025). DNA extraction and genotyping Genomic DNA was isolated using a commercial GeneAll extraction kit (GeneAll Biotechnology Co., Ltd.) following the manufacturer’s protocol. DNA quantity and purity were assessed using spectrophotometry (NanoDrop; Thermo Fisher Scientific), and integrity was verified by agarose gel electrophoresis. Genotyping was performed using the Axiom Ovine Genotyping Array (51 K) (Thermo Fisher Scientific). Genotype calling and initial array-level quality control were completed according to the manufacturer’s instructions using the Axiom Analysis Suite (Thermo Fisher Scientific). Resulting genotype data were exported in standard PLINK-compatible formats for downstream filtering and population-genomic analyses. Quality control of genotype data Genotype quality control (QC) was performed using the PLINK software version 1.9, following widely accepted protocols for array-based genotype data [22–24]. The raw dataset comprised approximately 51,000 markers generated with the Axiom Ovine Genotyping Array (Thermo Fisher Scientific). SNP- and sample-level filtering steps were applied as follows: SNPs with a call rate < 95% and individuals with a call rate < 95% were removed; markers with a minor allele frequency (MAF) < 0.05 were excluded; SNPs showing significant departures from Hardy–Weinberg equilibrium (HWE; p < 1 × 10⁻⁶) were removed; and sex-discordant or duplicated samples were checked and excluded where appropriate. Pairwise relatedness was assessed using the PI_HAT statistic implemented in the PLINK software, and one member of each highly related pair (PI_HAT > 0.25) was removed to reduce cryptic relatedness. For downstream population-structure analyses, including principal component analysis (PCA) and clustering, linkage disequilibrium (LD) pruning was performed using a sliding-window approach with a 50-SNP window, a step size of 5, and an r² threshold > 0.2 to obtain an approximately independent set of markers. After applying these filters, 46,000 high-quality SNPs remained and were used in all subsequent population-genetic analyses. Diversity statistics Population-level genetic diversity metrics, including observed heterozygosity (Ho), expected heterozygosity (He), Wright’s FIS, and rarefied allelic richness, where Variant Call Format (VCF) files were available, were estimated using a combination of R packages specialized for population genetics. Heterozygosity and F-statistics were computed using the hierfstat R package and cross-checked using the adegenet R package [25, 26]. Allelic richness was estimated by rarefaction using the poppr R package [27]. Summary tables and population-specific diagnostics were generated in the R software environment using the data.table and dplyr R packages for tidy data management [28]. Pairwise genetic differentiation (FST) Pairwise Weir and Cockerham FST estimates were obtained using the StAMPP R package, with bootstrap resampling to obtain confidence intervals [29]. The results were corroborated using the hierfstat R package [26]. The StAMPP R package was used because it efficiently handles large SNP matrices and includes built-in bootstrapping for locus resampling, thereby providing robust estimates of between-population differentiation. Multivariate analyses (PCA) Principal component analysis (PCA) was performed on the LD-pruned SNP set to visualize the principal axes of genomic variation and to compute population centroid positions and centroid distances. The PCA routine was implemented using the SNPRelate R package from Bioconductor [30], which was selected for its high performance with moderately large SNP matrices. The PCA implementation in the PLINK software was also used as an independent verification [23]. Component variances, group centroids, and PC-score distributions were summarized and visualized in the R software environment using the ggplot2 R package [31–33]. Phylogenetic reconstruction Distance matrices based on identity-by-state and allele-sharing measures were computed and used to infer a distance-based phylogeny. Neighbor-joining trees were constructed using the ape R package and were midpoint-rooted [34]. Branch-length heterogeneity was inspected using the phytools R package [35]. Clade compactness and patristic spread were quantified using the cophenetic and patristic functions available in the phangorn and ape R packages [36]. Tree visualization and annotation for publication figures were performed using the ggtree R package [37]. Model-based ancestry inference (sNMF) Admixture and ancestry proportions were inferred using the sNMF program implemented in the LEA R package [38]. The LD-pruned dataset was used as input for sNMF, and K values ranging from 2 to 6 were evaluated with multiple replicates for each K. Cross-entropy was used to select the optimal model. The resulting Q-matrices, representing individual ancestry coefficients, were summarized by population and visualized using the ggplot2 R package [32]. The LEA R package was selected because it provides computationally efficient ancestry inference and robust cross-entropy-based model selection. Analysis of molecular variance (AMOVA) Analysis of molecular variance (AMOVA) was conducted to partition molecular variance among and within predefined populations using the pegas R package and was validated using the poppr R package [27, 39]. Statistical significance was assessed by permutation tests with 999 permutations, and ΦST values were computed from AMOVA variance components to quantify the proportion of variation attributable to among-population differences. Runs of homozygosity (ROH) and FROH analysis Runs of homozygosity (ROH) were identified from genome-wide SNP data using a rule-based sliding-window approach implemented in the detectRUNS R package [40]. Before ROH detection, the filtered genotype data were converted into PED and MAP formats using the PLINK software [23, 24]. ROH segments detected across the autosomal genome were subsequently summarized at both the individual and population levels to characterize the genomic distribution of homozygous tracts and their length profiles. Genomic inbreeding based on ROH was estimated as FROH, calculated as the proportion of the autosomal genome covered by ROH, following the principle described by McQuillan, Leutenegger [41]. Graphical summaries and descriptive visualizations were generated in the R software environment using the ggplot2 R package [33]. Gene flow analysis Gene flow among the studied sheep populations was inferred from genome-wide SNP data using a population differentiation-based framework. After marker filtering and linkage disequilibrium pruning, pairwise genetic differentiation was estimated according to the method of Weir and Cockerham [42]. Approximate gene flow values (Nₘ) were subsequently derived from the pairwise differentiation matrix based on Wright [43] island model. These estimates were used to construct pairwise gene-flow matrices describing the relative extent of historical genetic exchange among populations. The resulting patterns were further summarized using matrix- and network-based visualizations to facilitate the interpretation of inter-population connectivity. All analyses were conducted in the R software environment. Data handling and SNP-based computations were performed using the SNPRelate and gdsfmt R packages [30], whereas network visualization was performed using the igraph and ggraph R packages [44, 45], and graphical outputs were generated using the ggplot2 R package [46]. Visualization, record-keeping, and reproducibility All figures, including PCA scatterplots, admixture barplots, heatmaps, ROH distributions, and phylogenies, were generated in the R software environment using the ggplot2 [32], pheatmap [47], and ggtree R packages [37]. Data wrangling and reporting were performed using the data.table and dplyr R packages [28, 48]. Software versions and R session information were recorded to ensure reproducibility [31]. A provenance file detailing package versions, parameter settings, and input/output file names was prepared for archival deposition and to enable full reproducibility. Results Genetic diversity parameters across populations Genetic diversity parameters estimated across the fifteen Anatolian sheep populations revealed clear variation among groups in heterozygosity, inbreeding coefficients, minor allele frequency, allelic richness, and the proportion of polymorphic loci (Table 1). Observed and expected heterozygosity showed broadly similar patterns across most populations, although some groups displayed comparatively lower diversity than others. GKD, MKN, and ZOM were characterized by relatively higher diversity across multiple indices, including heterozygosity, minor allele frequency, allelic richness, and proportion of polymorphic loci. By contrast, AWS and CMS exhibited comparatively lower diversity across several of these parameters. Table 1Genetic diversity parameters estimated across fifteen Anatolian sheep populationsGrup n HoHeFIS_LFIS_GMAFARPpolyAKN200.2730.259-0.052-0.0550.1971.7070.3AWS320.1350.1760.2000.220.1331.4646.8CMS190.150.151-0.030-0.0280.1141.4343.7ESM220.2720.244-0.102-0.1170.1821.7070.5GKD190.3560.348-0.023-0.0230.2572.00100HAM250.2620.2830.0660.0730.2111.7878.7HMN200.2810.260-0.07-0.0800.1961.7272.3KLA260.2820.265-0.060-0.0660.2001.7273KOC200.3010.287-0.045-0.0490.2171.7879KRK230.2730.273-0.0020.0010.2011.7879.1KVK240.2770.238-0.135-0.1630.1781.6667.1MKN270.3140.310-0.016-0.0140.2291.8990.5NDZ280.2570.249-0.038-0.0310.1801.7676.5SKZ250.2740.259-0.05-0.0580.1941.7475.5ZOM290.3280.299-0.088-0.0980.221.8586n sample size, Ho observed heterozygosity, He expected heterozygosity, Ppoly proportion of polymorphic loci, FIS_L mean locus-specific inbreeding coefficient, FIS_G global inbreeding coefficient calculated as 1 − Ho/He, MAF minor allele frequency, AR allelic richness Genetic diversity parameters estimated across fifteen Anatolian sheep populations n sample size, Ho observed heterozygosity, He expected heterozygosity, Ppoly proportion of polymorphic loci, FIS_L mean locus-specific inbreeding coefficient, FIS_G global inbreeding coefficient calculated as 1 − Ho/He, MAF minor allele frequency, AR allelic richness The inbreeding coefficients also differed among populations. Positive FIS estimates were observed in AWS and HAM, whereas most of the remaining populations showed negative or near-zero values at both the locus-specific and global levels. KRK was characterized by values close to zero, indicating little deviation between observed and expected heterozygosity. In contrast, KVK showed the most pronounced negative inbreeding estimates among the studied groups. The diversity indices demonstrated substantial heterogeneity among the fifteen Anatolian sheep populations in terms of within-population genomic variation. Analysis of genetic differentiation The pairwise FST matrix revealed clear variation in the degree of genetic differentiation among the sampled sheep populations (Fig. 2). Overall, CMS and AWS exhibited the highest levels of differentiation relative to the remaining groups, whereas several other populations showed comparatively lower pairwise differentiation. Among the closest relationships, the lowest differentiation was observed between GKD and MKN, followed by other relatively weakly differentiated pairs involving HAM, KOC, KRK, AKN, and ZOM. In contrast, KVK showed consistently elevated differentiation from most other populations, particularly in comparisons involving AWS and CMS. Intermediate levels of differentiation were observed among the remaining population pairs, indicating a heterogeneous but structured pattern of genomic divergence across the dataset. Collectively, the pairwise FST results demonstrated substantial among-population variation in genetic differentiation across the fifteen Anatolian sheep populations. Pairwise FST matrix showing genetic differentiation among fifteen sheep breeds Analysis of PCA PCA revealed a structured distribution of the fifteen Anatolian sheep populations in multivariate space (Fig. 3). The first two principal components captured only a small proportion of the total genomic variation, yet they clearly differentiated several population groups. AWS and CMS were positioned distinctly from most other populations along the positive side of the first principal component. In contrast, GKD, MKN, and ZOM were located on the negative side of the same axis. NDZ also occupied a negative position, although it was separated from GKD, MKN, and ZOM by its relative placement on the second component. Several populations, including AKN, HAM, KLA, KOC, KRK, HMN, ESM, and SKZ, clustered more closely in the central region of the PCA space, indicating relatively similar multivariate positions. Within this central assemblage, ESM, HMN, SKZ, and KVK were shifted toward the positive end of the second principal component, with KVK showing the greatest upward displacement along this axis. By contrast, AWS, CMS, NDZ, GKD, MKN, and ZOM were positioned on the negative side of the second component. The centroid relationships further reflected this overall configuration. The closest group proximities were observed among several central populations, particularly involving ESM, SKZ, HMN, KLA, HAM, AKN, KOC, and KRK. In contrast, the greatest separations involved AWS and CMS relative to multiple other groups, and ZOM relative to the positively displaced central clusters. The PCA pattern demonstrated clear among-population structuring, with both compact central clustering and distinct peripheral separation of several populations. PCA plot of genetic variation among Anatolian sheep populations Analysis of phylogenetic structure The phylogenetic analysis revealed substantial heterogeneity among the fifteen Anatolian sheep populations in both within-group and between-group branch-length structure (Fig. 4). Group-specific patristic distance summaries showed clear differences in the extent of within-population phylogenetic dispersion. CMS and AWS exhibited comparatively lower within-group patristic distances, whereas GKD, MKN, HAM, KOC, KRK, and ZOM showed comparatively greater internal branch-length divergence. Intermediate within-group patterns were observed for AKN, ESM, HMN, KLA, KVK, NDZ, and SKZ. Assessment of group monophyly further distinguished the populations. AKN, AWS, HMN, KLA, KOC, KVK, SKZ, and ZOM formed monophyletic groupings in the inferred tree, whereas CMS, ESM, GKD, HAM, KRK, MKN, and NDZ did not form exclusive clades. The mean patristic distances among groups also indicated marked variation in phylogenetic proximity across the dataset. Comparatively closer relationships were observed among several pairs involving AWS and CMS, as well as among AKN, CMS, KLA, NDZ, and SKZ. In contrast, larger between-group distances were particularly associated with GKD, MKN, and ZOM compared with other populations. The phylogenetic structure showed a differentiated pattern across populations, combining variation in within-group branch-length dispersion, monophyletic clustering, and between-group patristic divergence. Circular phylogeny depicting between- and within-population divergence in Anatolian sheep Analysis of genetic structure The sNMF analysis identified K = 3 as the most supported clustering solution, as indicated by the lowest cross-entropy value among the tested models. This result suggests that the genetic structure of the fifteen Anatolian sheep populations is best explained by three major ancestry components rather than by a single homogeneous gene pool or by completely isolated breed-specific clusters. Therefore, K = 3 was considered the most informative level for interpreting the broad historical and genomic structure of the sampled populations. At K = 3, the inferred ancestry components showed uneven but biologically meaningful distributions across the fifteen Anatolian sheep populations (Fig. 5). The first ancestry component, represented by the red cluster in Fig. 5, was predominant in several populations and may reflect ancestry related to the earlier dispersal wave of thin-tailed sheep from the domestication center. This component was particularly evident in populations such as GKD, MKN, NDZ, and ZOM, although each of these populations also showed variable levels of admixture with the other inferred components. a. Population structure of fifteen Anatolian sheep populations at K = 3. b. Cross-entropy analysis for determining the optimal K The second ancestry component, represented by the green cluster in Fig. 5, was most clearly observed in AWS and CMS, with only minor contributions from the remaining components. This pattern may correspond to the later expansion of fat-tailed sheep lineages, which are historically important in the Near East, southeastern Anatolia, and adjacent regions. The strong contribution of this component in AWS is consistent with the known distribution of fat-tailed sheep in the southeastern part of Türkiye and the broader Fertile Crescent region. The presence of this component in CMS may indicate historical introgression or shared ancestry with southern and southeastern sheep genetic backgrounds. The third ancestry component was strongly represented in ESM, HMN, KLA, KOC, KRK, KVK, and SKZ, although the degree of admixture with the other two components varied among these populations. This third cluster is unlikely to represent a completely independent domestication event. Instead, it most probably reflects a secondary Anatolian admixed ancestry that emerged through historical miscegenation between the earlier thin-tailed ancestry, the later fat-tailed ancestry, and regionally adapted local populations. The distribution of this component across western, central, northern, and some eastern Anatolian breeds suggests that it may have been shaped by long-term gene flow, transhumant pastoralism, local breed formation, and historical animal exchange routes within Anatolia. AKN and HAM showed more admixed ancestry profiles. AKN displayed ancestry mainly partitioned between the second and third clusters, with a smaller contribution from the first cluster. HAM also showed a mixed pattern, with major contributions from the second and third components and a lower contribution from the first. These results indicate that some Anatolian sheep populations retain composite genomic backgrounds, probably reflecting multiple layers of historical gene flow rather than a single ancestral origin. Overall, the K = 3 structure pattern indicates that Anatolian sheep populations preserve signatures of three broad genomic components: an ancestry related to the first thin-tailed sheep dispersal, an ancestry associated with the later expansion of fat-tailed sheep, and a third admixed Anatolian component shaped by local miscegenation, regional adaptation, and historical exchange among sheep populations. Thus, the observed structure reflects both ancient demographic processes and more recent breed formation within Turkey. Analysis of molecular variance (AMOVA) The AMOVA showed that the total genetic variation in the sampled sheep populations was partitioned between among-group and within-group components (Table 2). Most of the molecular variance was attributable to within-group differences, although a considerable proportion was explained by variation among groups. The variance component associated with among-group differentiation was significant, confirming the presence of genetic structuring across the sampled populations. Taken together, these results indicate that genetic diversity was predominantly maintained within groups, with a substantial and statistically supported component of differentiation among groups. Table 2Analysis of AMOVA for sampled sheep populationsSourceSSDdfMSDVariance component (σ²)% of total varianceP-valueAmong groups4.829140.34490.01351137.87< 0.001Within groups7.6243440.02220.02216462.13—Total12.4533580.03480.035675100.00—SSD Sum of squares for deviation, df degrees of freedom, MSD mean square deviation Analysis of AMOVA for sampled sheep populations SSD Sum of squares for deviation, df degrees of freedom, MSD mean square deviation Analysis of ROH The class-wise distribution of ROH counts varied markedly among the studied groups (Table 3). In most groups, ROH were primarily concentrated in the 0–6 Mb and 6–12 Mb classes, whereas the representation of longer segments declined progressively across the 12–24 Mb, 24–48 Mb, and > 48 Mb categories. This pattern was particularly evident in AKN, ESM, GKD, HMN, KOC, KRK, MKN, and ZOM, in which ROH were mainly restricted to the two shortest classes, with limited representation in the intermediate classes and no detectable segments in the longest category. A similar distribution was also observed in NDZ, although this group showed a broader extension into the 12–24 Mb and 24–48 Mb classes. Table 3Class-wise mean ROH count by groupGroup0–66–1212–2424–48> 48TotalAKN3.2501.8000.2000.00000.0005.2500AWS33.00024.80011.6005.41000.68875.4980CMS21.40019.60011.3003.00000.36855.6680ESM8.8204.1400.2730.09090.00013.3239GKD1.5300.5790.1050.00000.0002.2140HAM11.8005.0003.9601.24000.56022.5600HMN3.0500.9501.0500.00000.0005.0500KLA4.8504.0001.8100.19200.00010.8520KOC0.8000.3000.1000.00000.0001.2000KRK11.1003.8700.3480.00000.00015.3180KVK0.5830.1670.04170.12500.0000.9167MKN3.5601.2600.3700.07410.0005.2641NDZ19.60013.6004.5700.46400.00038.2340SKZ8.7205.4402.8000.76000.16017.8800ZOM1.0300.8280.1030.00000.0001.9610 Class-wise mean ROH count by group KLA and SKZ exhibited ROH across a wider range of size classes, with most segments still concentrated in the 0–6 Mb, 6–12 Mb, and 12–24 Mb categories. By contrast, AWS, CMS, and HAM showed the broadest class-wise distribution, with ROH represented across the 0–6 Mb, 6–12 Mb, 12–24 Mb, and 24–48 Mb classes, and additional segments observed in the > 48 Mb category. The > 48 Mb class was absent in most groups, detected only in AWS, CMS, HAM, and SKZ. The ROH count profile demonstrated substantial heterogeneity among the groups in the relative representation of the 0–6 Mb, 6–12 Mb, 12–24 Mb, 24–48 Mb, and > 48 Mb classes (Fig. 6). Violin plots showing the distribution of group mean runs ROH counts FROH results The class-wise percentage contribution to total FROH varied substantially among the studied groups (Table 4) (Fig. 7). In AKN, ESM, GKD, KOC, KRK, KVK, MKN, and ZOM, the largest contribution to total FROH was derived from the shortest ROH class, indicating a predominance of short homozygous segments in the genomic inbreeding profile of these groups. Similarly, HMN showed a major contribution from the shortest class, although the 12–24 Mb class also accounted for a considerable proportion. In NDZ, the 6–12 Mb class represented the largest share of total FROH, followed by the shortest class and the 12–24 Mb class. Table 4Class-wise percentage contribution to total FROHGroup0–66–1212–2424–48> 48AKN53.4041.804.820.000.00AWS21.2023.7024.6024.406.03CMS18.1030.7032.3015.803.05ESM53.8036.606.453.230.00GKD63.1030.206.680.000.00HAM32.7019.8031.3013.203.04HMN45.5023.2031.300.000.00KLA27.1036.5032.204.190.00KOC63.0026.4010.600.000.00KRK63.2028.907.910.000.00KVK62.8014.107.1416.000.00MKN48.2033.7014.403.670.00NDZ34.5041.1022.601.750.00SKZ31.5034.3023.408.552.33ZOM46.0047.206.760.000.00 Class-wise percentage contribution to total FROH Violin plots of class-wise percent contributions to the mean FROH A different distribution pattern was observed in AWS, CMS, HAM, KLA, and SKZ, where medium- and long-length ROH classes contributed more prominently to total FROH. In AWS, the contribution was distributed broadly across all ROH classes, including the longest category. CMS and KLA were characterized by major contributions from the 6–12 Mb and 12–24 Mb classes, whereas HAM showed a substantial contribution from the 12–24 Mb class in addition to the shorter categories. SKZ also showed contributions across all length classes, with the distribution extending into the longest category. By contrast, the > 48 Mb class was absent in most groups, detected only in AWS, CMS, HAM, and SKZ. The class-wise composition of total FROH demonstrated marked heterogeneity among the groups in the relative contribution of short, intermediate, and long ROH segments. Analysis of gene flow Estimates of gene flow revealed marked variation in migration intensity among the fifteen Anatolian sheep populations (Fig. 8). Overall, GKD, MKN, HAM, KRK, and KOC showed comparatively higher levels of gene flow, whereas AWS and CMS exhibited the lowest mean migration estimates among the studied groups. Intermediate patterns were observed in AKN, ZOM, KLA, NDZ, HMN, SKZ, ESM, and KVK. At the pairwise level, the strongest gene flow was observed among several central population combinations, particularly those involving GKD, MKN, HAM, KRK, and KOC. Additional high-migration relationships were also evident between MKN and ZOM. In contrast, the weakest pairwise gene flow estimates were concentrated mainly in comparisons involving CMS and AWS, particularly with KVK, HMN, ESM, SKZ, KLA, and ZOM. The gene flow pattern showed substantial heterogeneity across the dataset, with clear differences in both population-specific and pairwise migration profiles among the Anatolian sheep populations. Network of gene flow among fifteen Anatolian sheep populations Discussion Analysis of genetic diversity The diversity estimates obtained across the fifteen Anatolian sheep populations indicate a clearly heterogeneous genomic landscape, with substantial between-breed differences in heterozygosity, minor allele frequency, allelic richness, proportion of polymorphic loci, and inbreeding coefficients. Such heterogeneity is expected in indigenous sheep resources shaped by contrasting breeding histories, effective population sizes, geographic isolation, and variable intensities of human-mediated selection. Comparable genome-wide studies in Turkish, Greek, and Ethiopian sheep have likewise shown that local breeds rarely form a uniform diversity spectrum; rather, they tend to comprise populations with moderate-to-high variation alongside others showing narrower genomic backgrounds [14, 49, 50]. Within this framework, GKD, MKN, and ZOM emerged as the most diversity-rich populations in the present dataset, as they consistently ranked higher across several complementary indices, including heterozygosity, minor allele frequency, allelic richness, and the proportion of polymorphic loci. Interpreted jointly, these patterns suggest that these populations have retained broader standing variation and may have experienced weaker recent erosion of genomic diversity than lower-diversity groups. This interpretation is consistent with population-genomic theory because both heterozygosity and allelic richness capture different but complementary aspects of diversity, with allelic richness being particularly informative for preserving long-term evolutionary potential and low-frequency variants [51]. By contrast, AWS and CMS displayed comparatively reduced diversity across several parameters, indicating a more restricted within-population genomic base. A pattern of this kind is commonly associated with stronger genetic drift, founder effects, reduced effective population size, or sustained propagation from a relatively narrow breeding pool. Similar interpretations have been proposed in other livestock populations, where declining diversity indices have been linked to demographic contraction and increased vulnerability to future loss of adaptive flexibility [51, 52]. An additional point of importance is that observed and expected heterozygosity were broadly concordant across most breeds, suggesting that, for many populations, current diversity levels are not dominated by extreme departures from Hardy–Weinberg expectations. However, the breed-specific FIS profiles demonstrate that the distribution of this diversity is not uniform. Positive FIS estimates in AWS and HAM indicate relative heterozygote deficiency, whereas most of the remaining populations showed negative or near-zero values. In population-genetic terms, positive FIS values are generally interpreted as signatures of heterozygote deficit, while negative values indicate heterozygote excess relative to Hardy–Weinberg expectations. These deviations may arise through different mechanisms, including non-random mating, population substructure, recent drift, or admixture. They therefore should be interpreted as indicators of breeding structure rather than as direct proof of inbreeding alone [14, 53]. The AWS pattern is particularly notable because its reduced diversity metrics coincide with positive FIS values. Together, these results point to a population that is not only genetically narrower than several other Anatolian breeds but also characterized by an internal structure consistent with a heterozygote deficit. This combination may reflect a more restricted mating network, local subdivision, or the cumulative effects of drift over time. HAM also deserves attention for the same reason, although its overall diversity profile was not as reduced as AWS’s. In contrast, KRK displayed FIS values close to zero, implying little deviation between observed and expected heterozygosity and therefore a comparatively balanced internal structure. At the opposite end, KVK showed the most pronounced negative inbreeding coefficients in the dataset, suggesting marked heterozygote excess, a pattern often seen in populations with recent admixture, broader gene flow, or weaker recent inbreeding signals [14, 50]. Importantly, the present results reinforce that genomic diversity should not be inferred from a single parameter alone. Heterozygosity, minor allele frequency, allelic richness, and polymorphic locus proportion describe related but non-identical dimensions of variation, and their combined interpretation provides a more biologically informative view of breed status. This is especially relevant in conservation genomics, because populations can differ not only in the amount of variation they harbor, but also in the extent to which rare alleles and segregating loci have been retained. The conservation literature has repeatedly emphasized that long-term adaptive capacity depends on maintaining broad allelic diversity rather than focusing narrowly on immediate performance-oriented breeding outcomes [51, 52]. The diversity profile of Anatolian sheep revealed here supports a differentiated interpretation rather than a single general conclusion. Some populations, particularly GKD, MKN, and ZOM, appear to retain relatively broad genomic variation and may therefore represent valuable reservoirs of standing diversity. Others, especially AWS and CMS, show evidence of a narrower diversity base and may warrant closer monitoring in future conservation and breeding programs. From a practical perspective, these findings suggest that management priorities should be population-specific: breeds with richer diversity should be preserved as reservoirs of adaptive and evolutionary potential, whereas populations with reduced diversity or heterozygote deficiency may benefit from breeding designs that limit further loss of variation while avoiding indiscriminate introgression. Similar conclusions have been drawn in recent sheep genomic studies, which underscore that sustainable improvement and conservation require breed-by-breed management grounded in actual genomic diversity patterns rather than assumptions based only on geography or phenotype [49, 50, 54]. Genetic differentiation The pairwise FST landscape indicates that genetic differentiation among the fifteen Anatolian sheep populations is clearly structured, but far from uniform. Rather than supporting a simple model of discrete, evenly spaced breeds, the data reveal a mosaic of weakly, moderately, and more clearly partitioned genomic backgrounds. This pattern is consistent with genome-wide studies in sheep showing that local populations often retain signals of both historical connectivity and breed-specific divergence, with the observed level of differentiation reflecting the combined effects of shared ancestry, drift, reproductive isolation, and regional breeding history [50, 55]. A central result of the present analysis is that CMS and AWS occupy the upper end of the differentiation spectrum, showing the strongest genomic separation relative to many of the remaining populations. In population-genetic terms, this suggests that these two groups represent comparatively distinct gene pools within the broader Anatolian sheep framework. Such elevated differentiation is biologically plausible in indigenous sheep systems, where some breeds remain more isolated due to geographic constraints, directional breeding, limited ram exchange, or maintenance within relatively closed reproductive networks. Similar breed-specific distinctiveness has been described in other regional sheep datasets, where certain local breeds stand apart from neighboring populations despite broader background relatedness across the species [54, 56]. At the opposite end of the spectrum, the lowest differentiation was observed between GKD and MKN, with other comparisons showing relatively weak differentiation involving HAM, KOC, KRK, AKN, and ZOM. These low-FST relationships suggest that not all recognized populations in Anatolia are separated by strong genomic boundaries. Instead, some appear to retain a substantial shared background, which may reflect historical gene flow, incomplete lineage separation, parallel processes of breed formation, or long-standing regional connectivity. Kijas, Townley [55] similarly showed that sheep breeds do not always form fully isolated genetic entities, and that genome-wide differentiation frequently follows gradients of relatedness rather than rigid categorical boundaries. The present findings therefore support a view of Anatolian sheep as a heterogeneous yet connected metapopulation landscape, in which some breeds have remained relatively close, while others have undergone greater divergence. This interpretation is also consistent with recent evidence from Greek indigenous sheep, in which pairwise differentiation varied substantially among breeds, and the genomic structure reflected both local distinctiveness and broader historical admixture across the eastern Mediterranean [50]. In that context, the Anatolian pattern observed here appears especially informative because it captures both ends of the differentiation continuum within a single regional dataset: very close population pairs on one side and markedly separated groups such as CMS- and AWS-related comparisons on the other. Another notable aspect of the results is KVK’s position, which consistently showed higher differentiation than most other populations, particularly in comparisons involving AWS and CMS. This suggests that KVK also occupies a relatively distinct genomic position, even though it does not define the upper boundary of divergence as CMS or AWS do. The repeated appearance of KVK in higher-differentiation contrasts implies that its breed history may have involved stronger reproductive isolation, a different ancestral contribution, or more limited recent gene flow with many other Anatolian populations. Similar cases have been reported in local sheep breeds from Central and Eastern Europe, where one or a few populations emerge as genetically more distinct from the regional background despite belonging to the same broad geographic system [56]. Importantly, the intermediate FST values observed across many of the remaining breed pairs argue against overinterpreting Anatolian sheep structure as either fully homogeneous or fully fragmented. Instead, the data point to graduated genomic divergence, with populations distributed across a continuum of relatedness. This kind of structure is particularly common in sheep, whose domestication history, transhumance routes, local adaptation, and repeated human-mediated movement have produced overlapping but non-identical genomic backgrounds across regions. Global and transregional sheep genomic studies have repeatedly shown that modern breed structure often reflects layers of ancient dispersal superimposed on more recent regional isolation and targeted breeding [54, 55]. From a biological and management perspective, these results have two major implications. First, highly differentiated populations, especially CMS and AWS, may represent relatively distinct reservoirs of breed-specific genomic variation and therefore deserve particular attention in conservation and breed characterization efforts. Second, weakly differentiated populations should not be assumed redundant, as low pairwise divergence can coexist with valuable local adaptive variation, especially in indigenous systems shaped by complex demographic histories. As emphasized in recent studies of sheep diversity, the practical value of FST lies not in applying a universal cutoff but in understanding the relational structure it reveals among breeds and the historical processes that shape that variation [50, 56]. The pairwise differentiation profile supports the conclusion that Anatolian sheep populations form a genomically stratified but internally uneven system. Some populations remain relatively close, consistent with shared ancestry or historical connectivity, whereas others—most notably CMS, AWS, and to a lesser extent KVK—show clearer divergence from the broader Anatolian background. Thus, the emerging picture is not one of simple breed discreteness, but of a complex regional network shaped by unequal divergence, partial connectivity, and breed-specific histories. This interpretation is fully compatible with contemporary genome-wide evidence showing that indigenous sheep resources are best understood as dynamic, historically layered populations rather than as uniformly separated breed blocks [50, 54, 55]. Analysis of PCA The PCA results indicate that the genomic structure of the fifteen Anatolian sheep populations is organized, but not discretely partitioned. The first two components explained only a limited fraction of total genomic variance. Yet, they still resolved a biologically meaningful configuration composed of a compact central cluster and several clearly displaced peripheral populations. In the present dataset, AWS and CMS occupied the most distinct positions on the positive side of PC1, whereas GKD, MKN, and ZOM were shifted toward the negative side of the same axis; NDZ also fell on the negative side but remained separated from those populations along PC2. By contrast, AKN, HAM, KLA, KOC, KRK, HMN, ESM, and SKZ were concentrated in the central PCA space, indicating greater multivariate affinity, whereas KVK showed an additional upward displacement along PC2. Taken together, this pattern supports the interpretation of Anatolian sheep as a structured genomic continuum with unequal breed separation, rather than a system of fully isolated breed blocks. Such a combination of central overlap and peripheral divergence is well aligned with genome-wide studies in sheep, which show that local breeds often retain partial connectivity despite clear population structuring [55, 57]. A major feature of the ordination is the pronounced separation of AWS and CMS from most other populations. Their displacement from the central assemblage suggests that they carry comparatively distinct genomic backgrounds, likely reflecting greater historical divergence, more limited gene flow, or breed-specific demographic trajectories. This interpretation is consistent with broader ovine population-genomic evidence showing that PCA often isolates certain breeds at the edges of multivariate space when they have experienced stronger reproductive separation or regionally specific breeding histories. Kijas, Townley [55] reported that sheep generally show low-to-moderate overall differentiation but still cluster by breed history and geographic origin. In contrast, analyses of Croatian sheep similarly revealed that some populations are clearly separated from others despite broader background relatedness across the species. Equally informative is the compact central clustering of AKN, HAM, KLA, KOC, KRK, HMN, ESM, and SKZ, with centroid proximities also supporting close relationships among several of these populations. This concentration in the middle of the PCA space suggests that a substantial subset of Anatolian sheep populations shares relatively similar multivariate genomic profiles, compatible with either historical gene flow, shared ancestry, parallel breed formation, or incomplete differentiation. Comparable PCA outcomes have been reported in other regional sheep datasets, in which neighboring or historically connected breeds showed broad overlap rather than sharp separation. For example, Drzaic, Curik [57] emphasized that Croatian native breeds exhibit complex structure shaped by geography and gene flow rather than simple all-or-none partitioning, and Michailidou, Kyritsi [50] likewise reported that indigenous Greek sheep can display both distinct breed signals and broader interconnectedness across the eastern Mediterranean. The positions of GKD, MKN, ZOM, and NDZ further show that differentiation in this dataset is not restricted to a single contrast between a central core and two outliers. Instead, the PCA suggests multiple axes of genomic structuring. GKD, MKN, and ZOM were jointly displaced on the negative side of PC1, whereas NDZ occupied a related but distinguishable position because of its placement on PC2; at the same time, KVK, together with ESM, HMN, and SKZ, was shifted upward on the second axis, with KVK showing the strongest positive displacement. This layered arrangement indicates that PC2 captures an additional dimension of breed differentiation superimposed on the broader separation captured by PC1. Similar multi-axis patterns have been observed in genomic studies of sheep from southeastern Europe, where PCA did not separate breeds along a single linear gradient but instead reflected overlapping historical processes, including regional isolation, admixture, and breed-specific drift [50, 58]. From a broader perspective, these findings reinforce the view that Anatolia represents a historically layered ovine gene pool, shaped by both shared ancestry and uneven divergence. PCA did not indicate genomic homogeneity, but neither did it support complete fragmentation. Rather, it revealed a mosaic in which some populations remain tightly grouped, whereas others occupy distinct peripheral positions. This interpretation closely matches global and regional evidence showing that sheep breeds often preserve signals of ancient dispersal, subsequent regional isolation, and partial admixture, producing PCA landscapes with both overlap and separation [50, 55]. Accordingly, the present PCA should be read as evidence of stratified but continuous population structure: AWS and CMS emerge as the clearest peripheral groups, GKD–MKN–ZOM–NDZ define a second displaced configuration, and several central populations retain close multivariate affinity. In this sense, the Anatolian sheep gene pool appears not as a set of fully discrete genomic units, but as a regionally structured system with both compact clustering and breed-specific divergence. Phylogenetic structure The phylogenetic reconstruction shows that Anatolian sheep populations are neither uniformly discrete nor phylogenetically diffuse. Instead, they form a heterogeneous lineage mosaic defined by three concurrent features: variation in within-population branch-length dispersion, incomplete versus complete monophyly, and uneven between-population patristic divergence. In the present dataset, AWS and CMS showed comparatively lower within-group patristic distances, whereas GKD, MKN, HAM, KOC, KRK, and ZOM displayed broader internal branch-length dispersion. At the same time, AKN, AWS, HMN, KLA, KOC, KVK, SKZ, and ZOM formed monophyletic clusters, whereas CMS, ESM, GKD, HAM, KRK, MKN, and NDZ did not. This combination indicates that phylogenetic cohesion in Anatolian sheep is breed-specific rather than generalizable across the full panel. Comparable genome-wide studies in sheep have likewise shown that some breeds form compact and well-resolved lineages. In contrast, others remain only partially separated because of shared ancestry, regional connectivity, or historical admixture [50, 55]. A key result is that monophyly and internal compactness did not fully coincide. AWS combined low within-group patristic dispersion with monophyletic clustering, consistent with a relatively cohesive lineage. CMS, however, also showed low internal distances but failed to form an exclusive clade, indicating that branch-length compactness alone does not guarantee phylogenetic distinctness. Conversely, GKD, MKN, HAM, KOC, KRK, and ZOM exhibited broader internal dispersion, yet not all of them behaved similarly with respect to monophyly. This is biologically important because it suggests that Anatolian breeds differ not only in how distinct they are from one another, but also in how internally uniform their lineages remain. Similar decoupling between breed clustering and internal heterogeneity has been reported in regional genomic analyses of sheep, where demographic history, breeding structure, and variable introgression produce mixed phylogenetic signals rather than clear breed boundaries [58, 59]. The pattern of between-group patristic distances further strengthens this interpretation. Several of the more closely related phylogenetic relationships involved AWS and CMS, as well as comparisons among AKN, CMS, KLA, NDZ, and SKZ. In contrast, larger between-group distances were particularly associated with GKD, MKN, and ZOM relative to other populations. Thus, Anatolian sheep do not occupy a single phylogenetic continuum of equal spacing; rather, they are distributed across an uneven relational landscape in which some breeds remain phylogenetically proximate while others are consistently more offset. This kind of asymmetric structure is fully compatible with broader sheep population-genomic evidence showing that domestic breeds often preserve both signals of historical connectedness and localized divergence within the same regional system [2, 54]. Taken together, these findings support a mosaic model of Anatolian sheep phylogeny. Some populations behave as relatively coherent genealogical units, whereas others appear embedded in a more diffuse network of shared ancestry. The main message is therefore not simple breed discreteness, but unequal lineage resolution across populations. This interpretation is especially relevant for conservation and breed management, because populations that are both phylogenetically cohesive and relatively distinct may warrant treatment as more clearly delimited genomic units. In contrast, non-monophyletic and internally dispersed populations may require management strategies that explicitly account for internal heterogeneity and historical connectivity. Recent genomic studies in sheep have similarly emphasized that conservation priorities should be based on the actual structure of breed relationships rather than on breed labels alone [50, 60]. Genetic structure and historical interpretation of the K = 3 pattern The sNMF analysis supported K = 3 as the most informative clustering level, as indicated by the lowest cross-entropy value. This pattern is biologically meaningful because previous studies of sheep populations have shown that low-K structure analyses can capture major historical ancestry components, particularly the differentiation between fat-tailed and thin-tailed sheep. Ciani, Mastrangelo [61] reported that, at K = 2, K = 3, and K = 4, one ADMIXTURE component consistently corresponded to fat-tailed sheep, supporting the interpretation that one of the ancestry components observed in the present study may reflect fat-tailed sheep ancestry. Similarly, Tapio, Ozerov [62] reported that Bayesian clustering analysis revealed three broad genetic clusters in northern Eurasian sheep, namely Nordic, Composite, and Fat-tailed clusters, and that additional substructure within the fat-tailed cluster was detectable at K = 3. Therefore, the K = 3 solution in the present study is consistent with previous evidence that three-component models can reveal biologically meaningful historical structure in sheep populations. In this framework, the first component, represented by the red cluster in Fig. 5, may reflect an older ancestry associated with the initial dispersal of thin-tailed sheep from the domestication center. Yurtman, Özer [63] showed that Neolithic sheep from Anatolia had a complex demographic history since domestication, indicating that modern Anatolian sheep may retain multiple ancestral layers rather than a single homogeneous origin. This interpretation is also compatible with the central role of Anatolia and the Near East in the early domestication of sheep and its subsequent dispersal. Thus, the red component may represent an ancient thin-tailed ancestry layer that has been partially retained in some modern Anatolian sheep populations. The second component, represented by the green cluster, may correspond to the later expansion of fat-tailed sheep. Moradi, Nejati-Javaremi [64] identified genomic regions associated with fat deposition in thin- and fat-tailed sheep breeds and suggested that selection for fat-tail size occurred several thousand years after domestication. More recently, Jin, Liu [65] reported that fat-tailed sheep began to diverge from thin-tailed sheep approximately 4.4–7.5 thousand years ago, supporting the view that fat-tailed sheep represent a later differentiated lineage rather than the earliest domesticated sheep form. These findings support the interpretation that the strong contribution of this component in AWS reflects a later fat-tailed sheep ancestry associated with the Near East, eastern Anatolia, southeastern Anatolia, and neighboring regions. The third component most likely does not represent a third independent domestication event, but rather a secondary Anatolian admixed ancestry arising from historical miscegenation among earlier thin-tailed sheep, later fat-tailed sheep, and locally adapted regional populations. This interpretation is supported by several genome-wide studies showing that K = 3 often captures not only deep ancestral divergence but also regional admixture and gene-flow patterns in sheep. For example, Amane, Belay [66] reported the lowest cross-validation error at K = 3 in Sudanese and continental African sheep datasets, where the inferred clusters were broadly associated with geographic distribution, tail morphotype, and population type. Likewise, Abied, Ahbara [67] found the lowest ADMIXTURE CV error at K = 3, separating Sudanese thin-tailed Desert sheep, Chinese thin-tailed sheep, and Chinese fat-tailed Tan sheep into three major gene pools. These studies indicate that a K = 3 pattern can reflect a biologically meaningful combination of ancestry, morphology, geography, and historical gene flow. The presence of the third component in several western, central, northern, and eastern Anatolian breeds suggests that long-term gene flow, transhumant pastoralism, local breeding practices, and animal exchange routes within Anatolia shaped this ancestry. Michailidou, Kyritsi [50] reported comparable patterns of admixture and historical gene flow in indigenous Greek sheep breeds, including genetic introgression from both Anatolia and Eastern Europe. In addition, recent archaeogenomic evidence from Anatolian sheep indicates that fat-tailed breeds, mostly originating from eastern Turkey, show greater affinity to Iron Age Anatolian sheep. In contrast, thin-tailed breeds from western Turkey retain more Neolithic-like ancestry. This supports the hypothesis that modern Anatolian sheep populations were not formed by a single linear ancestry, but by repeated historical layers of migration, introgression, and regional breed formation. Thus, the K = 3 solution can be interpreted as a biologically meaningful representation of the historical formation of Anatolian sheep: an older, thin-tailed ancestry; a later, fat-tailed ancestry; and a secondary, admixed Anatolian ancestry. This interpretation explains why some populations show a dominant single ancestry component, whereas others display more heterogeneous genomic profiles. Therefore, Anatolian sheep should be considered a structured yet historically interconnected genomic system shaped by early dispersal from the domestication center, subsequent expansion of fat-tailed sheep, and long-term admixture across Turkey’s diverse ecological landscapes. Analysis of AMOVA The AMOVA results show that a dual pattern of diversity retention and population differentiation shapes the genomic architecture of the sampled Anatolian sheep populations. Most molecular variation was maintained within groups, yet the among-group component was also significant, demonstrating that these populations cannot be treated as a single, weakly structured gene pool. Instead, the data support a model of substantial internal diversity superimposed on statistically supported intergroup stratification. This interpretation is consistent with genome-wide studies in sheep, in which the largest share of variation was typically observed within populations. At the same time, significant among-population variance still revealed biologically meaningful subdivision [68, 69]. What makes the present pattern important is not simply that within-group variation predominates—that is common in livestock and especially in sheep—but that the between-group signal remains strong enough to be statistically detected despite that predominance. In practical terms, this means that Anatolian sheep retain considerable diversity within groups while simultaneously preserving non-random differences in allele frequencies among groups. Edea, Dessie [68] reported a similar general principle in Ethiopian sheep, in which most genomic variation resided within populations, yet AMOVA still detected significant structure among them. The same broader logic has also been observed in geographically structured sheep datasets, where spatial or regional grouping captured a significant, though smaller, fraction of total variance, confirming that population subdivision persists even when within-population diversity remains high [69]. The biological implication is clear: Anatolian sheep are neither genomically homogeneous nor fragmented into fully isolated units. Rather, they represent a structured diversity system in which a broad reservoir of variation is preserved within groups. At the same time, regional or population-level history has generated a measurable layer of differentiation among them. This interpretation aligns well with recent comparative analyses showing that native sheep breeds often combine substantial internal diversity with population structure shaped by development history, geography, admixture, and local breeding practices [54, 56]. From a conservation perspective, this AMOVA pattern is especially informative. Because most variation is retained within groups, maintaining within-group effective diversity remains essential. At the same time, the significant among-group component indicates that each group also contributes non-redundant genomic information to the broader Anatolian sheep gene pool. As a result, conservation priorities should not focus solely on preserving population labels or on maximizing pooled heterozygosity, but on protecting both internal variation and between-group distinctiveness. Similar conclusions have been emphasized in modern sheep genomic studies, which argue that native breeds should be managed as reservoirs of both adaptive within-population diversity and historically structured between-population differentiation [54, 56]. Overall, the AMOVA results support the robust conclusion that the sampled Anatolian sheep populations form a genetically structured yet internally variable system. The dominant share of within-group diversity rules out a simplistic bottlenecked interpretation across the dataset as a whole. In contrast, the significant among-group variance confirms that differentiation is real and biologically relevant. In high-impact terms, the key message is that Anatolian sheep diversity is organized, not diffuse; shared, but not homogeneous. That combination is precisely what makes these populations important as genomic resources for both conservation and future breeding strategies. ROH landscape and breed-specific autozygosity The ROH results reveal a sharply heterogeneous autozygosity landscape across the Anatolian sheep panel, indicating that these populations have not experienced a shared or uniform demographic trajectory. Instead, the class-wise ROH distributions point to breed-specific combinations of historical relatedness, recent inbreeding, drift, and management structure. ROH are informative precisely because their length distribution carries a temporal signal: shorter segments generally reflect older shared ancestry, whereas long tracts are more consistent with recent parental relatedness and more recent autozygosity [70, 71]. Against that framework, the dominant signal in the present dataset is the prevalence of short ROH classes. In most groups, ROH were concentrated in the 0–6 Mb and 6–12 Mb categories, while counts declined progressively with increasing class length. This pattern was especially clear in AKN, ESM, GKD, HMN, KOC, KRK, MKN, and ZOM, where ROH were largely confined to the shortest categories and the longest class was absent. Biologically, such a profile is more compatible with older background autozygosity and accumulated demographic history than with intense, very recent inbreeding. Recent sheep studies have reported the same general logic: when short ROH dominate and long tracts are rare, the signal typically reflects older inbreeding or long-term population processes rather than strong contemporary consanguinity [72, 73]. At the other end of the spectrum, AWS, CMS, and HAM showed the broadest ROH class distribution, spanning the 0–6 Mb, 6–12 Mb, 12–24 Mb, and 24–48 Mb classes, with additional segments also extending into the > 48 Mb category. This is the most consequential result of the analysis. The presence of very long ROH is widely interpreted as evidence that recombination has not yet fragmented recently inherited identical-by-descent chromosomal segments, making such tracts especially informative for recent inbreeding or recent mating among related individuals [71, 74]. In practical terms, AWS, CMS, and HAM appear to carry not only a legacy of older homozygosity but also a more recent layer of genomic consolidation than most of the other populations in the panel. Among the remaining breeds, KLA and SKZ occupied an intermediate position. Their ROH were distributed across a wider size range than in the short-ROH-dominated groups, but the bulk of segments remained concentrated in the 0–6 Mb, 6–12 Mb, and 12–24 Mb classes. This suggests a mixed demographic signal: broader homozygosity than in the low-ROH groups, but without the same degree of consistent long-tract enrichment seen in AWS, CMS, and HAM. Notably, SKZ also showed detectable ROH in the > 48 Mb class, which distinguishes it from most breeds, even though its overall profile was not as expanded as that of the three most ROH-broadened populations. Similar breed-level contrasts in both ROH burden and ROH length spectra have been documented in Mediterranean and Italian sheep, where local management and demographic history generated markedly different homozygosity profiles among breeds [75, 76]. These contrasts matter because ROH burden alone is not enough; the length composition of ROH is often more biologically revealing than the total count. Peripolli, Munari [70] emphasized that ROH patterns should be interpreted as the combined genomic outcome of inbreeding, drift, bottlenecks, selection, and breeding management rather than as a simple pedigree proxy. In the present study, that principle is especially relevant: several breeds exhibited ROH, but only a limited subset carried the full spectrum of segment lengths, from short to very long. This indicates that Anatolian sheep populations differ not merely in the extent of homozygosity they harbor, but in the temporal depth and likely causes of that homozygosity. Overall, the ROH landscape supports a strong biological conclusion: the Anatolian sheep panel is not genomically uniform in its autozygosity profile. Most breeds are characterized primarily by short ROH, consistent with older demographic processes. In contrast, AWS, CMS, and HAM stand out for their broader ROH spectra and the inclusion of very long segments, indicating stronger recent autozygosity. KLA and SKZ appear intermediate, while the remaining breeds largely retain a short-ROH-dominated pattern. In high-impact terms, the key message is that ROH in these populations capture multiple demographic timescales simultaneously, revealing a system shaped by uneven recent inbreeding pressure superimposed on deeper historical structure, as also observed in genomic studies of sheep from Europe, North Africa, and West Asia [72, 76, 77]. FROH and inbreeding dynamics The class-specific partitioning of total FROH provides a more informative view of genomic inbreeding than overall FROH alone because the length distribution of ROH reflects the temporal structure of autozygosity. Peripolli, Munari [70] emphasized that ROH-based approaches are especially valuable in livestock because they not only quantify genomic inbreeding but also help distinguish different demographic and breeding processes. Likewise, Ceballos, Joshi [78] explained that the number and length of ROH reflect population history, with longer tracts generally indicating more recent identity by descent. In contrast, shorter tracts are more consistent with older shared ancestry. In the present study, AKN, ESM, GKD, KOC, KRK, KVK, MKN, and ZOM showed the largest contribution to total FROH from the shortest ROH class. HMN displayed a broadly similar pattern, although the 12–24 Mb class also contributed appreciably, whereas NDZ differed slightly, showing its highest contribution in the 6–12 Mb class. These class distributions suggest that genomic inbreeding in these populations is shaped mainly by older background autozygosity, accumulated relatedness, and long-term drift rather than by very recent close inbreeding. This interpretation is consistent with findings from sheep populations, in which short and intermediate ROH were linked primarily to older demographic structures and cross-generational inbreeding patterns rather than to acute contemporary consanguinity [79]. A contrasting pattern emerged in AWS, CMS, HAM, KLA, and SKZ, where medium- and long-length ROH contributed more strongly to total FROH. AWS exhibited contributions from all ROH classes, including the longest category, indicating a more layered inbreeding profile shaped by both historical and recent processes. CMS and KLA were dominated by the 6–12 Mb and 12–24 Mb classes, suggesting that their genomic inbreeding burden is concentrated in intermediate- to long tracts rather than in short ancestral segments alone. HAM also showed a substantial contribution from the 12–24 Mb class, while SKZ presented a broad class distribution extending into the longest ROH category. As Peripolli, Munari [70] noted, the relative contribution of ROH classes is critical for distinguishing ancient background homozygosity from more recent co-ancestry in livestock genomes. Particularly noteworthy is the fact that the > 48 Mb class was detected only in AWS, CMS, HAM, and SKZ. Stoffel, Johnston [80] showed that long ROH are a strong genomic manifestation of recent inbreeding and are closely connected to the genetic architecture of inbreeding depression. More recently, Kyriazis, Robinson [74] argued that long ROH are among the most reliable genomic indicators of reduced fitness and conservation risk. Therefore, the restriction of the longest ROH class to only a subset of populations suggests that recent close relatedness is not uniform across Anatolian sheep but is concentrated in specific groups. Taken together, these findings show that genomic inbreeding in Anatolian sheep does not follow a single common ROH profile. Instead, the breeds differ markedly in the temporal composition of autozygosity. Short ROH, which mainly dominate some populations, reflect an older background of inbreeding, whereas others carry a stronger burden of intermediate and long ROH, pointing to more recent co-ancestry. This heterogeneity indicates that genomic risk should be interpreted on a breed-specific basis and that conservation or mating-management strategies should be tailored to the ROH architecture of each population rather than applied uniformly across all breeds [70, 79]. Patterns of gene flow in Anatolian sheep The gene-flow analysis revealed a markedly uneven migration landscape across the fifteen Anatolian sheep populations, indicating that a single demographic process cannot explain contemporary breed relationships. Instead, the results support a mosaic pattern in which some populations remain strongly connected through historical or ongoing exchange, whereas others appear substantially more isolated. This interpretation is consistent with the broader history of domestic sheep, a species shaped by extensive dispersal and recurrent introgression under human management [55, 81]. In Anatolia, previous studies have likewise shown that breed structure reflects geography, breeding history, and repeated movement of animals across production systems rather than strict long-term reproductive isolation alone. A notable outcome of the present study is the comparatively elevated gene flow detected in GKD, MKN, HAM, KRK, and KOC. This pattern suggests that these populations have either retained broader historical connectivity or have been more exposed to ram exchange, animal trade, transhumant movement, or other forms of mediated admixture. Kijas, Townley [55] showed that sheep breeds often display weak phylogeographic partitioning because breed formation has frequently involved recent and repeated gene exchange. Similarly, Meadows, Li [81] reported unusually high interregional maternal introgression in sheep, reinforcing the central role of migration in sheep demographic history. Within Türkiye, migration signals among native breeds have also been reported, indicating that present-day relationships reflect both local continuity and secondary contact among populations [11, 82]. The particularly strong pairwise migration signals involving GKD, MKN, HAM, KRK, and KOC further indicate that connectivity is not only population-specific but also highly relationship-dependent. In livestock, gene flow is rarely symmetric or geographically simple; rather, it is often structured by breeding preferences, market networks, replacement of breeding males, and movement of flocks under traditional systems. Comparable heterogeneous migration profiles have been reported in Mediterranean and Eurasian sheep datasets, in which some breeds occupy central positions within gene-flow networks, while others remain peripheral despite geographic proximity [50, 83]. The high migration rate observed between MKN and ZOM is also noteworthy, as it suggests substantial genetic exchange despite the retention of breed identity. In domestic sheep, visible breed distinctiveness does not necessarily imply deep reproductive isolation, and differentiated populations may still share considerable genomic background through recurrent introgression [13, 14, 82]. In contrast, AWS and CMS displayed the lowest mean migration estimates, and the weakest pairwise gene flow was concentrated in comparisons involving these populations. Reduced gene flow may reflect geographic separation, localized selection objectives, founder effects, or more closed breeding structures. Similar patterns in sheep have been associated with clearer breed separation and, in some cases, increased conservation concern [11, 84]. The repeated appearance of weak migration links between CMS or AWS and populations such as KVK, HMN, ESM, SKZ, KLA, and ZOM suggests that this isolation is systematic rather than incidental. These findings indicate that Anatolian sheep diversity is maintained through an unequal balance between connectivity and separation, with important implications for conservation and breeding management [11, 14, 50]. Conclusions This study provides a genome-wide perspective on the population structure and genomic diversity of fifteen Anatolian sheep populations. It demonstrates that these genetic resources are both highly diverse and unevenly structured. Although most variation was distributed within populations, multiple complementary analyses revealed clear differences among populations in genetic diversity, differentiation, admixture, homozygosity patterns, and gene flow. Some populations retained comparatively broad genomic variation and stronger connectivity, whereas others showed signs of greater isolation, reduced migration, or elevated homozygosity. These findings indicate that Anatolian sheep do not represent a homogeneous genetic pool, but rather a complex mosaic of lineages shaped by distinct demographic histories and varying levels of genomic exchange. This heterogeneity has direct implications for conservation planning and breed management. Population-specific strategies will therefore be essential to preserve unique genomic backgrounds, limit further erosion of diversity, and support the sustainable use of Anatolian sheep genetic resources under future breeding and conservation frameworks. Abbreviations Akkaraman Analysis of molecular variance Awassi Çukurova Meat Sheep Degrees of freedom Eşme Inbreeding coefficient within populations Genomic inbreeding coefficient based on runs of homozygosity Fixation index Gökçeada Hamdani Expected heterozygosity Hemşin Observed heterozygosity Kangal Akkaraman Koçeri Karayaka Kıvırcık Minor allele frequency Mean square deviation Norduz Principal component analysis Runs of homozygosity Sakız Sparse non-negative matrix factorization Single-nucleotide polymorphism Sum of squares for deviation Zom Authors’ contributions Mervan Bayraktar: Writing – original draft, Validation, Methodology, Formal analysis, Conceptualization. Murat Durmuş: Data collection, Visualization, Methodology, Data curation. Murat Turan: Data collection, Visualization. Özge Durmuş: Data collection, Visualization.B. Devrim Özcan: Visualization. Nazan Koluman: Visualization. Funding Not supported by any institution. Data availability The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request. Declarations Ethics approval and consent to participate All experimental procedures were reviewed and approved by the Animal Care and Ethics Committee at Cukurova University (Approval No. 2025/25–47). Animals were sourced from both privately owned flocks/commercial farms and the Çukurova University research herd. Written informed consent for sample collection and the use of anonymized genetic data for research was obtained from all private flock/farm owners or their authorized representatives before sampling. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Availability Statement The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.