Genomic Evaluation for a Crossbreeding System Implementing Breed-of-Origin for Targeted Markers

The genome in crossbred animals is a mosaic of genomic regions inherited from the different parental breeds. We previously showed that effects of haplotypes strongly associated with crossbred performance are different depending upon from which parental breed they are inherited, however, the majority of the genomic regions are not or only weakly associated with crossbred performance. Therefore, our objective was to develop a model that distinguishes between selected single nucleotide polymorphisms (SNP) strongly associated with crossbred performance and all remaining SNP. For the selected SNP, breed-specific allele effects were fitted whereas for the remaining SNP it was assumed that effects are the same across breeds (SEL-BOA model). We used data from three purebred populations; S, LR, and LW, and the corresponding crossbred population. We selected SNP that explained together either 5 or 10% of the total crossbred genetic variance for average daily gain in each breed of origin. The model was compared to a model where all SNP-alleles were allowed to have different effects for crossbred performance depending upon the breed of origin (BOA model) and to a model where all SNP-alleles had the same effect for crossbred performance across breeds (G model). Across the models, the heritability for crossbred performance was very similar with values of 0.29–0.30. With the SEL-BOA models, in general, the purebred-crossbred genetic correlation (r(pc)) for the selected SNP was larger than for the non-selected SNP. For breed LR, the r(pc) for selected SNP and non-selected SNP estimated with the SEL-BOA 5% and SEL-BOA 10% were very different compared to the r(pc) estimated with the G or BOA model. For breeds S and LW, there was not a big discrepancy for the r(pc) estimated with the SEL-BOA models and with the G or BOA model. The BOA model calculates more accurate breeding values of purebred animals for crossbred performance than the G model when r(pc) differs (≈10%) between the G and the BOA model. Superiority of the SEL-BOA model compared to the BOA model was only observed for SEL-BOA 10% and when r(pc) for the selected and non-selected SNP differed both (≈20%) from the r(pc) estimated by the G or BOA model.