Comparative assessment of methods for estimating genomic relationships and their use in predictions in an admixed population

Abstract

Different genomic relationship (G) estimators were compared within and across populations in an admixed population. By assessing relationship coefficients separately for different populations, this study found that scaling G with current data allele frequencies across breeds increased coefficients for individuals in distant related populations, when compared to using breed allele means calculated from breed proportions. The latter however shifted most relationships towards zero or less. The predictions of direct estimated genomic values (DGV) were unaffected regardless of allele frequencies used. Relationship coefficients that combine genomic information and polygenic variation from the pedigree slightly increased the validation reliability of DGV in the current population.