Avaliação genética em rebanhos pastorais com incerteza de paternidade

Abstract

Many sheep flocks in the world are managed collectively, in community way, in which animals from different owners graze together. These conditions make it difficult or impossible to identify the paternity of offspring, as several rams and ewes can mate with no control or livestock record. This practice prevents the obtaining of genetic evaluations using animal model. The aim of this study is to evaluate the efficiency of the use of two methods, the hierarchical animal model (MH) and the average numerator of the relationship matrix (NMMP), in the estimation of genetic parameters and breeding values of sheep in situation with paternity uncertainty in which there is mating in the presence of multiple sires. The methods were compared in two cases, one with a simulated trait with 7.5 kg2 phenotypic variance, mean 14.5 kg, and heritability of 0.30, so that the true genetic parameters were known and one with real data, with weights at birth and at weaning of Santa Inês breed from Embrapa Goats and Sheep. In both cases, the datas were analyzed in four ways: 1) with full pedigree, as available (MPC); 2) ignoring the full knowledge of parents (MDP); 3) using multiple sires and the average numerator of the relationship matrix (NMMP); and 4) using multiple sires and hierarchical model (MH). Analyses were performed with Bayesian procedures with the technique of Markov Chain Monte Carlo (MCMC). For all models, MCMC chains of 1,100,000 cycles, after burn-in period of 400,000 cycles, with sampling period every 10 cycles were used. The NMMP and MH models were compared by the Deviance Information Criteria (DIC) and Pseudo Bayes Factor (PBF). The differences in the estimates of the breeding values and in the ranking of animals between the models were verified by correlations (Pearson and Spearman). For both the simulated data as to the real data, MH model was slightly higher than NMMP model, according to DIC and PBF criteria. However, no significant differences were observed between these two models, for the estimation of genetic parameters and the ranking and identification of the best rams. In the case of real data, depending on the structure of the data available, models with multiple sires were not efficient in estimating the breeding values for maternal effects, although there have been no different estimates between models for the genetic parameters of this effect ((co) variance and heritability). This may have occurred because the heritability estimated for the maternal effect was practically zero. The results confirmed that the genetic evaluation models with use of multiple sires, who consider paternity uncertain, are efficient to estimate genetic parameters and rank the best rams in sheep flocks with community pastoral characteristics. Despite the best fit of the data by MH, both models were similar to the estimates and may be considered in genetic evaluations.