Genomic prediction of crossbred performance

Abstract Text: The majority of the commercial slaughter pigs are crossbred animals. However, breeding efforts have been mainly focused on increasing genetic progress of purebred populations. The aim of this work is to evaluate different strategies to improve genomic prediction of crossbred performance taking into account the breed origin of alleles in crossbred populations (breed-specific effects). Previous work showed that marker effects estimated in one breed cannot predict performance in another breed (across-breed prediction). This might be due to breed-specific effects caused by differences in linkage disequilibrium between the marker and the QTL, as well as differences in allele frequencies and in genetic background of the breeds. For prediction of crossbred performance, marker effects estimated in single-breed data showed some predictive value but training on crossbred data achieved higher accuracies, although the breed origin of alleles was ignored. In this study, prediction accuracies of breeding values from a traditional genomic selection model (GS) were compared with prediction accuracies of breeding values from a model that accounts for breed-specific effects (BS). The population evaluated consisted of a two-way (Large White and Landrace) crossbred population. As both parents of all crossbred animals were known, the breed origin of alleles was easily determined after phasing of the data. The trait evaluated was gestation length (GL), for which a genetic correlation between purebred and crossbred performance (r_pc) of 0.90 was estimated. Prediction accuracy of BS breeding values was slightly greater than prediction accuracy of GS breeding values (0.53 and 0.52, respectively). Additional benefits of BS over GS are expected for traits with lower r_pc and when crosses of more distant purebred populations are evaluated. As a step further, a method based on long-range phasing for determining the breed origin of alleles in three-way crossbred data was developed. In a simulation study, the accuracy of breed of origin assignment was determined for 400 three-way crossbred animals with 95% correct assignments, 3% unassigned and < 2% incorrect assignments. Application of this method to real data, including 14,000 genotyped purebred animals and 1700 genotyped three-way crossbred animals, achieved 93% assignments of breed of origin of alleles without using pedigree information. Genotypic data from purebred animals was required to define the haplotypes of the three breeds contributing to the crossbreds. Currently, analyses are underway to use this breed origin information of the three-way crossbred population to estimate breed-specific effects for genomic prediction.

Keywords: Pigs, Crossbred performance, breed-specific effects