Exploiting sequence variants for genomic prediction in Australian sheep using Bayesian models

Abstract

The accuracy of genomic predictions could be potentially improved by creating competitively priced low to medium density custom SNP chips, that include sequence SNPs strongly associated with a range of economically important traits. The SheepGenomesDB and Australia Sheep CRC have recently completed whole-genome sequencing of 726 sheep, enabling the imputation of approximately 46,000 Australian sheep of multiple breeds and crosses that were previously genotyped with lower density SNP chips. Subsets of these sheep are recorded for a range of growth and meat quality traits. We used this dataset to discover putative causal SNPs associated with these traits and then combined these SNPs with the 50k SNP chip genotypes for Bayesian genomic prediction. The genomic predictions were validated in purebred Merino and Border Leicester × Merino crossbreds. On average there was a 5% increase in the accuracy of genomic breeding values by adding the top sequence SNPs to the 50k SNP genotypes compared to using only the 50k genotypes.