Genomic Selection in New Zealand Sheep Using a Mixed-breed Training Set

Genomic selection, allowing improved selection decisions by using predictions based on many SNP markers, is available within the New Zealand sheep breeding industry. The current implementation uses a mixed-breed training set based on the predominant Romney breed and two breeds, Coopworth and Perendale, developed around 60 years ago by crossing Romneys with Border Leicester and Cheviot, respectively. Predictions, using GBLUP and fitting six principal components to account for breed structure, have moderate (0.24-0.77) realized accuracy in each of these breeds and in composites containing more than 30% of these breeds for most of the traits routinely evaluated. Useful predictions have also been obtained for traits currently being investigated for use in the industry, such as methane emissions and stayability (e.g. realized accuracies exceed 0.3 for some breeds and traits). Predictions are developed using medium density (~50k) SNP chips or a matching subset from high density (~600k) SNP chips. Prediction in young animals is based on low density (~5k) SNP chips using imputation to medium density. Current research priorities include finding optimal methods for imputation to high density or sequence-level variants, use of genotyping by sequencing, discovery of SNPs with important effects and expanding the application of genomic selection to other breeds that are important to the New Zealand sheep industry.