LAUSR

This study investigated the efficiency of genomic prediction using an admixed reference population comprising 3 Yorkshire populations with different genetic backgrounds. In total, 2,084 and 1,388 individuals with growth and reproduction records, respectively, were genotyped with a PorcineSNP80 marker panel. The corrected phenotypic values derived from conventional EBV of each population were taken as response variables. Three approaches, that is, a linear genomic BLUP (GBLUP) model, a Bayesian mixture model (BayesR), and single-step GBLUP (ssGBLUP), were implemented to predict genomic breeding values. Our results indicated that the accuracy of genomic prediction was increased by enlarging the reference population by admixing different populations. However, the improvement was lower than expected, because the relationships among individuals of different populations were not strong enough. Among the 3 approaches, for reproduction and growth traits, ssGBLUP produced 30 to approximately 38% and 23 to 31%, respectively, higher accuracy than GBLUP. And the ssGBLUP produced 28 to approximately 38% and 18 to approximately 31% higher accuracy than BayesR. In addition, ssGBLUP also yielded lower bias. In most situations, BayesR performed comparably to GBLUP for most traits. Our results indicated ssGBLUP using an admixed reference population is also meaningful for national joint genetic evaluation of Chinese pig breeding.