Explorando a variação genética de características contínuas e binárias de bovinos leiteiros sob abordagem Bayesiana

Abstract

The economic efficiency in dairy cattle is associated to the animal’s productive and reproductive capacity. This way, to reduce the losses in the production system and to increase the profit; selection criteria should include production and milk composition traits plus traits related to animal health and fertility. However, since health and reproduction traits are generally associated to non-normal or unknown distributions, the use of threshold model under Bayesian approach is adequate to the genetic evaluation of such phenotypes. Therefore, the aim in this study was to estimate variance components, genetic parameters and to predict breeding values for continuous and discrete distributed traits of Holstein cattle by using linear and threshold models under Bayesian approach. A total of 64,657 records for milk yield (MY, kg/day), casein content (CP, g/100g milk), subclinical mastitis incidence (SMI) and precocity (PREC) of 5,439 first to seventh parity cows, with days in milk between 5 and 365, and belonging to four farms were used. Genetic analysis was based on a linear model (for MY and CN continuous traits) and on a threshold model (for SMI and PREC binary traits), which included the effects of contemporary groups, days in milk, lactation number, and the additive genetic, permanent environmental and residual effects. Variance components, breeding values as well as other unknown parameters of the model, were obtained under non-informative and informative priors, by using Gibbs sampler to obtain the marginal posterior distributions. The heritability posterior means for MY, CP, SMI and PREC, with non-informative prior were 0.142, 0.399, 0.134 e 0.377, respectively. For informative prior, these values were 0.131, 0.271, 0.134 e 0.436. Significative genetic gains were only observed for CP and PREC, with values equal to 0.014 g/100g of milk and 0.032 unit by year, respectively. These results highlight that linear and threshold models are adequate to estimate the variance components and genetic parameters of the studied traits.