Assessment of Genomic Prediction Accuracy for Meat Quality Traits using Various SNP Densities in Hanwoo Cattle

Abstract

<p>The availability of genome-wide single nucleotide polymorphism (SNP) panels has enabled the implementation of genomic prediction in many livestock species. Genomic prediction is widely applied to estimate genomic breeding values (GBV) since it was first proposed by Meuwissen in 2001. In beef cattle, genomic prediction has promising benefits for the improvement of carcass traits such as meat quality, because estimated breeding values can be obtained without sacrificing the selection candidates. The accuracy of genomic prediction mainly depends on the size and the diversity of the reference population, heritability of trait and the linkage disequilibrium between SNP and QTL. With whole-genome sequence (WGS) data, it is assumed that the causal mutations responsible for trait variation are included in the data, and therefore, the accuracy of prediction is expected to be improved compared to common SNP panels. The objective of this study was to examine the effect of various SNP densities (50K, HD and WGS) on genomic prediction accuracy for meat quality traits in Hanwoo beef cattle. Genomic and phenotypic data from 2,110 animals were used to predict genomic estimated breeding values (GBV) for marbling score (MS), meat texture (MT) and meat colour (MC). The 2110 Hanwoo steers were divided into 10 folds cross-validation using random sampling of individuals. Each of the fold (n=211, 10%) was used as validation dataset whereas the rest of the animals (n=1899, 90%) were used as a reference population. The WGS data (~15 million SNPs) was imputed from the 50K SNP chip to 777K, followed by an imputation step up to the whole-genome sequence level. The accuracy of imputation for WGS was on average 78% for SNPs with a MAF >0.01. The genomic best linear unbiased prediction model was used to predict the GBV for each trait fitting either of the genomic relationship matrices from the 50k, HD, and WGS data. Then the accuracy of GBV was assessed using the Pearson’s correlation between GBV and corrected phenotypic value divided by the square root of heritability. The estimated genomic prediction accuracies for MS, MT, and MC were 0.45, 0.39 and 0.29, respectively using either WGS or HD SNP panel However, the 50K SNP panel yielded slightly higher prediction accuracies for MS (0.46) and MC (0.31) traits than the other panels. The prediction accuracy of MT (0.39) was similar for all SNP densities. The result showed that the high-density SNPs (WGS and HD) did not improve the genomic prediction accuracy for all studied traits.</p>