Heterogeneidade de variância e parâmetros genéticos para produção de leite em bovinos, mediante inferência bayesiana

Abstract

Considering the intense use of the artificial insemination technique and the use of the mixed models methodology, the dairy breeders can be evaluated through the production records of their daughters distributed under different environmental conditions and in different generations, with the importance of including the effect of variance heterogeneity in genetic evaluations. The objective was to verify the presence of variance heterogeneity for milk production up to 305 days of lactation (PL305) in the Brazilian Girolando, Gyr and Holstein herds and how this heterogeneity may interfere in the classification of the genetic values of the breeders of each breed. A total of 3,235, 2,884 and 11,178 milk production records were used up to 305 days of lactation of progenies from 73, 79 and 304 Girolando, Gyr and Holstein breeders, respectively, from 2000 to 2015, from the Progeny Test of the Girolando Breed conducted by the Genetic Breeding Program of the Girolando Breed (PMGG). The standard deviation (SD) classes were formed by standardizing means of standard deviations of the herd-year classes, and the low DP class consisted of values of standard deviations equal to and less than zero and the class of high DP with positive deviation values. The months of childbirth were grouped in two seasons, the dry season from April to September and rain from October to March and used in the group formation of contemporaries by joining the classes of herd, year and season of calving. In the evaluation model were included the fixed effects of group of contemporaries, covariates of cow's age at calving (linear and quadratic effects) and degree of heterozygosity (linear effect), genetic additive random effects of animal, permanent environment of the animal and residue. The analyzes to obtain components of variance and genetic values of the individuals for the milk production were obtained in two situations, in the first one disregarding differences between the variances between the classes of phenotypic standard deviations (general analysis) and in the second situation, the data were analyzed considering the milk production in each class of DP as distinct characteristics, obtaining the components of (co) variances and genetic values in each class. Components of variances were estimated by Bayesian inference using the Gibbs sampler to obtain a posteriori means. Different heritability coefficients were observed, higher for the high DP class, in the Girolando breed (0.27) and in the Gyr breed (0.34), and the same did not occur in the Holstein breed, with the same coefficient being low and high DP class (0.25). In the present study, high genetic correlations were found between the low and high DP classes (0.90, 0.90 and 0.95) for the Girolando, Gyr and Holstein breeds, respectively. Correlations of order (spearman) found considering all bulls were high for the three races analyzed (equal to or above 0.88). The presence of the heterogeneity of variance verified in this study being mainly of genetic nature did not affect the genetic evaluation of the breeders of the three breeds studied.