پژوهشهای علوم دامی ایران (Jan 2015)
Effect of enlarging the number of reference population cows and imputed markers on reliability of genomic prediction in Jersey breed
Abstract
The objective of this study was to evaluate the potential gain in accuracy of predictions by imputing genotypes from a low-density marker panel (6640 marker with 10 replicates) to a medium-density marker panel in a simulated population of Jersey cattle using FImpute software and also to evaluate the reliability of genomic estimate of breeding values (GEBV) for milk yield (h2=0.40) and fertility rate (h2=0.04) with different number of cows in the reference population. A population of 900 medium-density panel of genotyped proven sires was simulated to quantify the accuracy of imputation of reference cows. To evaluate the reliability of GEBVs of 3000 randomly selected animals of test population, nine scenarios were employed. Average accuracy of imputation for cows was 98.64 percent. Only one percent difference observed in GEBVs of 54k and GEBVs achieved for imputed markers. According to high rate of imputation, however, just 1 percent difference obtained between comparing GEBV of 54k and GEBV achieved of imputed markers. Therefore, despite of little difference and although the genotyping cost of a medium-density panel is two-fold of low-density, using of 50K genotyped proven sires plus all cows imputed to medium-density is recommended. Following this procedure, not only the higher rate of GEBV for both traits would be achieved but also by decreasing the cost of genotyping, more animals of reference population would be used in prediction model. The more the reference animals, the more the pedigree, phenotypic and genotypic data are available to predict marker effects and, therefore, genomic estimated breeding value.
