Effect of genetic architecture on accuracy of multi breed genomic prediction
Objectives were to investigate effect of genetic architecture and including random across and within breed effects in GBLUP on accuracy of multi breed genomic prediction. High-density genotypes and imputed synonymous, missense and premature stop codon mutations using sequence data were available for 3000 Holstein Friesians and 3000 Jerseys. Phenotypes of traits with different genetic architectures, regarding allele frequency spectra and number of breed specific QTL, were simulated by sampling 100 QTL from a mutation class. Accuracies of genomic breeding values were estimated using GBLUP including random across and within breed effects. Increase in accuracy by adding individuals of another breed to the reference population and accuracy of across breed genomic prediction was low. Genetic architecture influenced accuracies; accuracies reduced when QTL allele frequencies were lower and QTL were more breed specific. Including a random within breed effect did not affect accuracies.