The accuracy of genomic predictions for milk related traits in Najdi cattle breed

The objective of this study was to evaluate the performance of genomic selection for milk yield, fat percent and protein percent in Najdi cattle breed from station and Agmari herds using different statistical models. Traditional estimated breeding values obtained with a random regression model using pedigree and phenotypic information for each trait from 1990 to 2016 were used as response variables. Predictability was evaluated by using a 10 replicated Training-Testing scheme and considering four models including genomic best linear unbiased prediction scaled with observed allele frequency (GBLUP), and allele frequency of 0.5 (G05BLUP), BayesA and BayesB. The results showed that GBLUP had better performance than G05BLUP for milk yield (0.411 vs 0.385), but performance of G05BLUP was better for fat percent (0.257 vs 0.302) and protein percent (0.363 vs 0.388). The accuracy of breeding values of milk yield decrease to 0.371 and 0.353, and accuracy of fat percent increased to 0.329 and 0.314 with BayesA and BayesB, respectively. Bayesian methods had same accuracy as GBLUPs for protein percent. Across traits and methods BayesA with 0.14 for protein percent, and G05BLUP with 0.71 for milk yield had smallest and highest bias as deviate from 1, respectively. Accuracy of prediction using Agmariherds in addition station herd increased 0.01 to 0.09 depending on method and trait, but also bias were more than before. In conclusion, accuracy of genomic prediction of milk traits in Najdi cattle breed are moderate but suitable considering small size of population which makes genomic selection feasible in this breed.