Population Structure in a Thai Multibreed Dairy Cattle Population

Accounting for population structure is important to help identify SNP associated with production traits in domestic animals particularly in multibreed populations. Models used to identify relevant SNP in multibreed populations utilize genetic groups usually constructed based on expected breed fractions. However, these groups may not appropriately account for structural differences due to SNP allelic frequencies. Thus, the objectives of this study were to construct genetic groups using SNP marker information, obtain genetic distances between genetic groups, and determine the correspondence between SNP-based and breed-fraction based genetic groups. The study included 2,661 cattle (89 bulls and 2,572 cows) from 304 farms located in Central, Northern, Northeastern, and Southern regions of Thailand, with complete pedigree that had been genotyped with GeneSeek Genomic Profiler 9K. Only SNP with minor allele frequency higher than 0.01, call rate larger than 90%, P-value of Hardy-Weinberg equilibrium lower than 0.0001, and r² value of linkage disequilibrium lower than 0.2 were included in this study (n = 5,005). A principal component analysis was used to obtain eigenvectors that were subsequently utilized to assign animals to genetic groups using a k-means clustering algorithm. Computations were performed using the discriminant analysis of principal component (DAPC) program of R-package adegenet. The optimum number of genetic groups in this population based on the lowest Bayesian Information Criterion (BIC) value was 28. Genetic distances among these SNP-based genetic groups were estimated using Nei’s genetic distance. The DAPC scatterplot of the first and second principal components showed four genetic groups clearly separated, and 24 genetic groups were very close to each other forming a “super cluster”. Conversely, Nei’s genetic distances among the 28 groups revealed 3 clusters, one containing group 23, a second one including groups 1, 2, 12, 20, 21,25, and 27, and a third cluster with the remaining groups. There was almost no correspondence (r = 0.00025) between breed composition of animals and their allocation to SNP-based genetic groups. In fact, SNP-based genetic groups contained animals of a wide range of Holstein fractions, and animals with Holstein fractions above 90% were represented in all SNP-based genetic groups. Thus, the DAPC algorithm was effective at identifying structural differences among animals based on gene frequencies regardless of their breed origin. However, genetic distances between these groups showed a different clustering pattern compared to the one obtained with the DAPC scatterplot of the first and second principal components.