IBD sharing patterns as intra-breed admixture indicators in small ruminants

The study addresses the challenge of detecting recent admixture and crossbreeding in small ruminant breeds, which threatens livestock diversity and the integrity of locally adapted genetic architectures. Conventional population structure methods (e.g., STRUCTURE/ADMIXTURE/PCAdapt) infer admixture from differences in allele frequency across populations and can fail when source breeds are absent from the dataset, a common situation in developing regions. The hypothesis is that more admixed (less genetically isolated) breeds exhibit more fragmented genomes and therefore share fewer and shorter identity-by-descent (IBD) segments within breed. The purpose is to evaluate whether intra-breed IBD sharing patterns can serve as robust indicators of admixture level, potentially independent of comprehensive knowledge of the surrounding genetic landscape, thereby enabling efficient identification of breeds endangered by crossbreeding.

The authors review threats to livestock genetic diversity from indiscriminate crossbreeding, particularly in developing countries, leading to genetic homogenization and disruption of co-adapted gene complexes. They outline commonly used methods for inferring population structure and admixture (STRUCTURE, ADMIXTURE, PCAdapt), noting limitations when key source populations are missing from analyses. They highlight the underused potential of haplotype information and IBD tracts to capture recent shared ancestry and admixture history. Prior work shows that hybridization generates mosaics of local ancestry and recombination breaks down ancestral haploblocks, shortening IBD tracts over generations. IBD sharing has been successfully applied to infer demographic history, migration, and introgression in plants and animals. This background motivates using intra-breed IBD segment number and length as indicators of admixture-related genome fragmentation.

Datasets: The study compiled SNP array genotypes for 111 goat breeds (2501 individuals; AdaptMap Caprine SNP50 BeadChip, 53,547 SNPs; post-filter 50,220 SNPs) from Europe (40), Africa (52), and Asia (19), and 156 sheep breeds (3304 individuals) from Europe (73/74), Africa (33), and Asia (50/49), primarily Ovine SNP50 BeadChip (French breeds down-sampled from Ovine HD to SNP50 coordinates). Quality control with PLINK v1.07 excluded SNPs with call rate ≤97% and MAF ≤1%, and individuals with ≥10% missing genotypes. Breeds with <8 individuals were excluded; for large breeds, up to 30 individuals were retained. FIS per population was estimated with Hierfstat. IBD detection: BEAGLE 4.1 performed phasing and IBD detection. Candidate IBD segments required LOD ≥4, length ≥0.5 cM (Beagle assumes 1 cM/Mb), and ibdtrim=40. An in-house Python 3 script computed per-breed mean IBD segment length and count shared among individuals of the same breed. Admixture and genetic integrity: LD-pruned SNPs (PLINK -indep, 50-SNP windows, 5-SNP step, VIF threshold 2; R^2>0.5) were used in ADMIXTURE. For K from 2 up to the number of breeds in each dataset, 10 runs per K were summarized with CLUMPAK. At K equal to the number of breeds, each breed’s main-cluster membership proportion (Q) defined its genetic integrity level. Breeds were classified as admixed if Q≤0.85 and slightly admixed if Q>0.85 (threshold near the third quartile of Q distribution). PCA with PCAdapt visualized structure. Genetic originality (AV index): Using the ADIV R package, the AV (average) distinctiveness index was computed per breed from pairwise Reynolds genetic distances, reflecting average dissimilarity to all other breeds (values near 1 indicate high originality/isolation). NeighborNet graphs from SplitsTree visualized breed originality. Statistical analyses: Student’s t-tests compared mean IBD length and count between admixed and slightly admixed groups. Spearman correlations related IBD metrics to genetic integrity level and AV index. ANOVA with Tukey post hoc tests compared datasets by region (Africa, Asia, Europe). Linear Models (LM) and Generalized Additive Models (GAM; mgcv) modeled genetic integrity level and AV index as responses with IBD count or length as predictors; model performance compared by chi-square tests and assessed via adjusted R^2. chromoMap visualized genomic distribution of shared IBD segments.

• Proportion admixed: 72.30% of goat breeds and 62.18% of sheep breeds were classified as admixed; all assessed Spanish sheep breeds and 95% of North/West African goat breeds were admixed.
• Regional differences:
  • Genetic integrity (Q): Goats showed a trend toward higher integrity in Europe (mean 0.75, SD 0.19) than Africa (mean 0.66, SD 0.21; p=0.07). Sheep had higher integrity in Europe (mean 0.81, SD 0.20) than Asia or Africa (~0.65; ANOVA/Tukey p<0.01).
  • AV index (originality): Goats mean AV 0.092 (SD 0.05), higher in Europe (mean 0.11, SD 0.06) than Africa (mean 0.07, SD 0.04; p<0.001). Sheep mean AV 0.097 (SD 0.06), higher in Europe (mean 0.11, SD 0.07) than Asia (mean 0.07, SD 0.03; p<0.01).
  • IBD metrics: Goats averaged 13.30 shared IBD segments per breed (SD 19.91), mean length 5.4 Mb (SD 4.91). Sheep averaged 8.57 segments (SD 12.84), mean length 7.41 Mb (SD 5.67). African goat breeds had significantly fewer (mean 5.62) and shorter IBD segments (mean length 2.94 Mb) than Asian (19.00; 8.27 Mb) and European breeds (20.59; 7.64 Mb) (ANOVA/Tukey p<0.01). European sheep had longer IBD segments (mean 9.77 Mb) than Asian (5.00 Mb) and African (5.85 Mb) breeds (p<0.001).
• Admixed vs slightly admixed:
  • Goats: Admixed breeds had fewer IBD segments (mean 6.82) than slightly admixed (mean 30.82; p<2.2×10^-16) and shorter segments (mean 3.98 Mb vs 9.78 Mb; p=3.66×10^-12).
  • Sheep: Admixed breeds had fewer IBD segments (mean 4.03 vs 16.03; p=9.514×10^-5) and shorter segments (mean 5.28 Mb vs 10.92 Mb; p=9.514×10^-5). In African sheep, IBD count difference was not significant.
• Correlations:
  • Across goats and sheep, AV index correlated strongly with IBD count and length (Spearman ~0.75). Correlations between genetic integrity level and IBD metrics were moderate (~0.5) and weakest/non-significant in African datasets (especially African sheep).
• Modeling genetic integrity from IBD:
  • Goats: LM performed best; adjusted R^2=0.236 for IBD count and 0.307 for IBD length (both p<0.0001). Including dataset origin did not improve models.
  • Sheep: GAM outperformed LM. With dataset origin included, adjusted R^2=0.43 (IBD count) and 0.45 (IBD length); strong nonlinearity (edf>2), indicating threshold/plateau behavior; African dataset showed weakest relationships (IBD length not significant at alpha 0.05 in Africa).
• Modeling AV index from IBD (GAM):
  • Goats: Adjusted R^2=0.77 for IBD count (0.80 with dataset origin); 0.58 for IBD length (0.62 with dataset origin). Strong, often nonlinear relationships; Asia showed linear/quasi-linear patterns.
  • Sheep: Adjusted R^2=0.59 for IBD count (0.66 with dataset origin); 0.56 for IBD length (0.59 with dataset origin).
• Case study (Irish goats): Old Irish Goat (OIG) vs crossbreds (OIGx). OIG had higher genetic integrity (Q=0.94 vs 0.45), more shared IBD segments (46.8 vs 9.8) and longer segments (12.3 Mb vs 2.1 Mb). Shared IBD maps highlighted loss/fragmentation around adaptive loci (e.g., GSTCD and HERC6 regions) in crossbreds.

Findings support the hypothesis that intra-breed IBD sharing patterns reflect admixture levels: breeds with less admixture (greater genetic isolation/originality) share more and longer IBD segments, indicating less genome fragmentation. This relationship holds broadly across goats and sheep and is especially strong when predicting a breed’s genetic originality (AV index), which captures isolation relative to others, rather than solely intra-breed similarity. Regional heterogeneity was evident, notably weaker or non-significant associations in African datasets, suggesting complex demographic and management histories (e.g., extensive recent crossbreeding initiatives, varying selection and inbreeding practices) that modulate IBD signals. The results demonstrate that IBD-based indicators can identify crossbreeding impacts without requiring all potential source breeds in the dataset, overcoming a major limitation of frequency-based admixture methods. However, interpreting IBD patterns requires consideration of alternative forces shaping haplotypes and linkage (e.g., artificial selection, inbreeding, bottlenecks, drift, domestication history) and the recency of events to which IBD is most sensitive (approximately past 100 generations).

IBD sharing patterns within breeds are informative indicators of admixture and genetic isolation in small ruminants. Admixed breeds exhibit fewer and shorter shared IBD segments, while genetically isolated/original breeds show more numerous and longer segments. Because IBD-based indicators do not depend on including all potential source breeds in analyses, they offer a practical tool to detect breeds endangered by crossbreeding and to guide conservation actions aimed at preserving locally adapted genetic architectures. Future work should integrate controlled experimental and theoretical studies to disentangle the effects of selection, inbreeding, demography, and timing on IBD patterns, refine regional models, and validate thresholds for operational monitoring.

• Regional variability: Relationships between IBD metrics and admixture/originality were weaker or non-significant in some African datasets (e.g., African sheep), indicating contextual complexity and potential confounders.
• Model limitations: Genetic integrity predictions showed modest explanatory power in goats (adjusted R^2 ~0.24–0.31) and nonlinear threshold/plateau behaviors in sheep, suggesting unmodeled factors.
• Confounding processes: IBD patterns can be influenced by artificial selection, inbreeding, demographic history (bottlenecks, drift), and domestication processes, complicating attribution solely to admixture.
• Temporal sensitivity: IBD primarily reflects recent history (roughly the past 100 generations); older admixture events may be less detectable.
• Dataset constraints: Uneven sample sizes, exclusion of breeds with <8 individuals, and reliance on SNP50 arrays may limit resolution; some comparisons were not tested due to group size imbalance. Assumptions in IBD detection (e.g., constant 1 cM/Mb recombination rate) may introduce approximation errors.