Single step evaluations using haplotype segments

Abstract

The low reliability of genomic evaluations in some breeds may be improved by regression on ancestral haplotypes instead of all available markers. The aim of this paper was to examine the use of haplotypes in constructing genomic relationships for single step evaluations. BayesB model was fitted to all markers simultaneously to estimate marker effects, using deregressed proofs (DRP) of reference bulls as data.  Phased 5-marker haplo-blocks were constructed from the markers with the highest absolute effect size and 4 adjacent markers. Haplo-blocks and their estimated variances were then used to compute genomic relationships. Evaluations used cow DRP of milk and protein as data, weighted by effective record contribution. Estimated genomic breeding values (GEBV) were validated from regression coefficients in a linear model. The validation reliabilities for milk over marker-based methods were improved by up to 4% with 1500 segments and 40% weight on the pedigree. Reliabilities were smaller for protein but comparable when the weight on pedigree was 40%. Inflation levels for both traits were higher with haplo-blocks than individual markers. Haplo-blocks appeared to be beneficial and indicate a need to further assess the optimal number of haplotypes and the weight on pedigree for single step evaluations.