225 Genomic diversity of Katahdin sheep and impacts on genomic prediction

Abstract

Abstract In the late 1950s, Katahdin hair sheep were developed as a composite breed of medium size and moderate prolificacy, with potential to express resistance to gastrointestinal nematodes. With its increasing popularity and the recent adoption of genomic prediction in the Katahdin breeding program, evaluating the genetic diversity in this breed is necessary. Prediction accuracies rely on the strength of genetic relationships. That strength often increases with selection using genomically-enhanced estimated breeding values; however, it carries the risk of decreasing diversity. We aim to characterize the genetic diversity present in the breed and the subsequent impact on the accuracies of genomic predictions. Genotypes on Katahdin sheep from member flocks in the National Sheep Improvement Program were used. Data from two medium density 50k bead chips and one high density 600k bead chip were merged and limited to shared autosomal single nucleotide polymorphisms (SNP). After quality control, there were 9,705 animals and 31,984 SNP remaining. Only for analyses of inbreeding and heterozygosity, linkage disequilibrium (LD) pruning was performed; variants with an LD score exceeding 0.8 were removed with 29,904 SNP remaining. To determine the extent of population structure, principal component (PC) analysis was conducted with PC one and PC two explaining 4.9% and 3.8% of the variation in the population, respectively. When examining subpopulations, a single ancestral population explained 99.9% of the genetic variation among animals. When clustered by identity-by-state, all animals were placed in a single cluster. Inbreeding coefficients were estimated with four methods: method of moments, variance-standardized relationship minus one, excess homozygosity, and correlation between uniting gametes. Across methods, average inbreeding coefficients ranged from 1.14% to 1.20% with Pearson and Spearman correlations ranging from -0.51 to 0.91 (Table 1). Historical effective population sizes (Ne) were estimated for earlier generations, and a linear regression fit to predict more recent Ne. At the founding of the breed, historical Ne was estimated to be 310, and the current Ne was estimated to be 150. Effects of Ne on the accuracy of genomic predictions were examined for reference population sizes ranging from 1k to 25k for traits of low (0.1), moderate (0.3), and high (0.5) heritabilities. The corresponding number of effective chromosomal segments were 2NeL, where L was the length of the genome in morgans (26M). Accuracies of genomic predictions increased as the reference population size grew, with only marginal gains once the population exceeded 15k. Overall, there were no detectable genetic subpopulations in the Katahdin breed. Current levels of inbreeding were low. With a small Ne the breed can achieve high accuracy for genomic predictions by aiming for a reference population of 15k animals. Negative impacts on genetic diversity due to genomic selection are not of immediate concern.