Using phenotypes and genotypes of three-breed cross to improve breeding value estimation of purebred animals

Breeding programs for pigs and poultry are currently based on genomic prediction (GP) of breeding values that mainly use training populations of purebred animals (PB), while the aim is to improve performance of crossbred animals (CB), which are usually a three-breed cross. The success of such
an approach therefore depends on purebred-crossbred genetic correlation (rpc). In our studied pig population, estimates of rpc for daily gain (DG), back fat thickness (BF) and loin depth (LD), were 0.61, 0.75 and 0.82, respectively. The magnitudes of these rpc indicate that genetic progress is being realized in crossbreds kept under commercial conditions by selecting on performance of
purebreds at the nucleus level. However, including phenotypes and genotypes of CB in the training population, will lead to higher genetic progress for these traits in CB. When using CB for GP, effects of single nucleotide polymorphisms (SNPs) may be breed-specific because linkage disequilibrium (LD) patterns between a SNP and a quantitative trait locus (QTL) and allele frequencies and effects of a QTL may differ between parental breeds. Estimation of breed-specific effects of alleles in a CB population, requires knowing their breed-of-origin. We investigated the performance of assigning breed-of origin of alleles in real data of three-breed cross pigs (finishers). The approach consists of three steps : 1)
phase the haplotypes of PB and CB, 2) determine unique haplotypes among the PB, and 3) assign the breedof-origin for each allele carried on haplotypes of CB. Genotypic and phenotypic data was available for 14,187 PB from the three parental breeds, and 1,723 finishers. On average 93.0% of the alleles of a finisher were assigned a breed-of-origin without using pedigree, and 94.6% using pedigree information. Using these derived breed origin of alleles, we built three partial genomic relationship matrices, one for each breed.
These matrices are used to estimate three separate genetic values of a finisher, one from each breed origin. This model will be tested by predicting breed specific breeding values for DG, BF, and LD.