Genomic Prediction in a Numerically Small Sheep Breed Population Using Imputed Sequence Variants

Abstract

The accuracy of genomic prediction for a numerically small sheep breed was investigated based on a large multi-breed admixed reference set using moderate or high density SNP genotypes, imputed whole genome sequence genotypes or selected sequence variants based on a genome wide association study (GWAS). Reference set with weight and eating quality phenotypes was divided into a GWAS sub set (n=4,000), a training set (n=13,466 to 38,098) and a validation set with data of 143 to 169 purebred Dorper sheep. Genomic BLUP was used to estimate genomic breeding values and prediction accuracy was evaluated in the validation set based on the correlation between GBV and corrected phenotypes. Results showed a prediction accuracy between 20% and 30% based on 50k genotypes across different trait, which increased on average by 2.5% to 7.0% by using HD genotypes or selected sequence variants derived from an independent GWAS.