Breeding Value Estimation for Fat Percentage Using Dense Markers on Bos taurus Autosome 14

Abstract

Prediction of breeding values using whole-genome dense marker maps for genomic selection has become feasible with the advances in DNA chip technology and the discovery of thousands of single nucleotide polymorphisms in genome-sequencing projects. The objective of this study was to compare the accuracy of predicted breeding values from genomic selection (GS), selection without genetic marker information (BLUP), and gene-assisted selection (GEN) on real dairy cattle data for 1 chromosome. Estimated breeding values of 1,300 bulls for fat percentage, based on daughter performance records, were obtained from the national genetic evaluation and used as phenotypic data. All bulls were genotyped for 32 genetic markers on chromosome 14, of which 1 marker was the causative mutation in a gene with a large effect on fat percentage. In GS, the data were analyzed with a multiple quantitative trait loci (QTL) model with haplotype effects for each marker bracket and a polygenic effect. Identical-by-descent probabilities based on linkage and linkage disequilibrium information were used to model the covariances between haplotypes. A Bayesian method using Gibbs sampling was used to predict the presence of a putative QTL and the effects of the haplotypes in each marker bracket. In BLUP, the haplotype effects were removed from the model, whereas in GEN, the haplotype effects were replaced by the effect of the genotype at the known causative mutation. The breeding values from the national genetic evaluation were treated as true breeding values because of their high accuracy and were used to compute the accuracy of prediction for GS, BLUP, and GEN. The allele substitution effect for the causative mutation, obtained from GEN, was 0.35% fat. The accuracy of the predicted breeding values for GS (0.75) was as high as for GEN (0.75) and higher than for BLUP (0.51). When some markers close to the QTL were omitted from the model, the accuracy of prediction was only slightly lower, around 0.72. The removal of all markers within 8 cM from the QTL reduced the accuracy to 0.64, which was still much higher than BLUP. It is concluded that, when applied to 1 chromosome and if genetic markers close to the QTL are available, the presented model for GS is as accurate as GEN.