عنوان مقاله [English]
Background and objectives: One major component in designing breeding programs is accurate assessment of breeding values. In most countries, selection decisions in animal breeding are usually based on predicted genetic values obtained using best linear unbiased predictors (BLUP). The first applications of BLUP and mixed linear models are associated with assuming homogeneous variance components. However, numerous studies have showed that there exists heterogeneity variance for milk yield and genetic and residual variances are not constant between herd and different production level. The level of herd production, management and nutritional practices and breeding strategies are heterogeneous variance factors. The problem of heterogeneous variances in genetic evaluation of dairy cows is that above average animals in the more variable herds may be over evaluated and larger proportion of animals is selected from these herds than when variance is homogeneous. When heterogeneity is not properly taken into account, differences in variance components can result in biased estimates of breeding values, in-correct ranking of sire and dams elite and reduced genetic progress. In evaluation of dairy cows, a simple method to reduce heterogeneous phenotypic variance, is pre-adjustment of records prior to fitting the animal model. The objectives of this study were to investigate the heterogeneity variance components for milk test-day records as well as the impact of data pre-adjustment method on on genetic parameters, breeding values and re-ranking of the top sires and dams in the Iranian Holstein cows.
Materials and methods: In this research, heterogeneity of variance components for milk production trait was investigated by using 1843985 test-day records of first parity of Iranian Holstein cows. Collected records were including 199353 cows in 983 herds that were collected by the Animal Breeding Center and Promotion of Animal Products of Iran. According to the average of herd-year production three groups (high, medium and low) were identified. The heterogeneity of variances was studied using the Bartlett’s test. A pre-correction method was used to correct heterogeneity of variance. Variance components, genetic correlations, heritabilities as well as animals’ breeding values (BVs) were estimated by ASreml program, using animal model. Spearman's rank correlations and re-ranking of top sire and dam before and after data pre-correction were estimated. The number of best sires and cows removed from top animal list after adjustment the heterogeneity of variance was obtained in order to investigate the benefit of this adjustment.
Results: The results of Leven’s test for milk test-day records before data pre-correction were significant (P<0.001) which indicated the heterogenous of variances. Data correction did not result in homogeneity of variance across the herd-year classes but the heterogeneity of variance was decreased by 25%. Applying of the pre-correction method resulted in slightly higher heritabilities so that estimates of total heritabilities were 0.319±0.007 and 0.351 ±0.009 before and after adjustment for heterogeneity of variance, respectively. Additive genetic correlations for milk test-day records were estimated in the range of 0.345 to 0.985 before and 0.403 to 0.996 after pre-correction, respectively. Adjustment of data had a considerable effects on the top 1% of animals and caused that 14% and 20% of top sires and dams, respectively, to be excluded from selection when compared to the homogenous variance scenario. Substantially different in re-ranking of top animals occurred and the best of sire and dam cows had rank 8 and 23 after adjustment of heterogeneity of variance
Conclusion: The results showed that variance components of test-day records based on different production levels in Holstein cow population are not homogeneous and heterogeneity of variance may influence the ranking and genetic evaluation of top animals.