Bayesian analysis on the growth traits in nelore mocho cattle raised in cerrado biome

This paper aims to estimate the (co) variances components on the genetic parameters for growth traits in Nelore Mocho cattle, raised in Cerrado biome, using Bayesian statistics. The data has been obtained from 48,063 cattle, provenient of Nelore Mocho herd, born from 1987 to 2009, participant of Breeder and Research Nacional Association’s (ANCP) Nelore Brazil Program. It has been evaluated the standard weight in 120 (P120), 210 (P210), 365 (P365), 450 (P450) years old, the weight gain from the birth to the weaning (GPND), and the weight gain from weaning to yearling (GPDS). The estimates on (co) variances components and genetic parameters have employed univariate analyses considering the Bayesian statistics animal model by the MTGSAM (Multiple Trait using Gibbs Sampler under Animal Model). The estimated values for direct heritability ranged from medium to high magnitude in standard weights (with range of 0.23 to 0.33) and the weight gain (0.18 for GPND and 0.24 for GPDS), for the creates and recreates phase, respectively. These results indicate that there is genetic variability in Nelore Mocho herd for growth traits, showing effective genetic growth when using the direct selection of these features in genetic improvement programs.


Introduction
The Brazilian beef cattle industry, an outstanding highlight in agribusiness, with national effective of approximately 212 million cattle, ranks third in importance. Most of Brazilian cattle is produced in the 1 Faculdade de Medicina Veterinária da Universidade Federal de Uberlândia 2 Associação Nacional de Criadores e Pesquisadores Cerrado, where herding consists of zebu cattle well adapted to tropical conditions, the Nelore and Nelore Mocho that account for almost 90% of zebu (Bos indicus) present in Brazil (ABIEC, 2012).
Despite this favorable scenario, the economic agents involved in cattle meat production and marketing acknowledge the need to improve productivity levels; therefore, increasing the country's competitiveness in the international market. However, there are few studies addressing estimates of genetic parameters for growth traits in Nelore Mocho cattle (SOUZA et al., 2004;LIMA et al., 2005;MALHADO et al., 2005;SANTOS et al. 2005), as well as those focusing on cattle in the Cerrado.
Traditionally, the Restricted Maximum Likelihood (REML) method is the most widely adopted in genetic evaluation programs. Opposing this trend, several studies employed new methods for obtaining estimates of (co) variance components and genetic parameters (FARIA et al, 2007).
Among these, there is the Bayesian analysis, made possible by Gibbs Sampling from the methods of Markov Chain Monte Carlo (MCMC), which considers the population parameters as random variables and enables integrating prior probabilities (a priori) in the data, resulting in posterior probabilities (a posteriori) and consequently better accuracy (VAN TASSEL & VAN VLECK, 1996).
From the knowledge on estimates of variance components and genetic parameters for growth traits, new selection indices will be used to identify superior genotypes in different production systems, accelerating genetic progress of the cattle herds of Nelore Mocho breed. Thus, this study aims to estimate the components of (co) variance and genetic parameters for growth traits in Nelore Mocho cattle bred in the Cerrado biome by using Bayesian statistics. Faria, Guimarães, Gomes, Miguel, Lemos, Pereira, Lôbo Bayesian analysis on the growth traits… In the formation of contemporary groups for the growth traits, the study has considered grouping the animals born on the same farm, the same year same birth season, same sex and same management batch. The effect of birth season resulted into four classes division: animals born between February and April, from May to July, August to October, and from November to January, according to the distribution of births that commonly occur in this biome.
In the database formatting was considered the elimination of groups with less than four animals. Beyond this restriction, outliers were also removed, which corresponded to animals with measurements of 3.5 standard deviations from the median of contemporary group. For all growth traits were assumed as fixed effects the contemporary groups (CG) and the age class at calving (CIVP), while the latter was divided into: less than or equal to three years, three to four years; four to five years; five to six years, six to ten and greater than ten.
Note that while carrying on the Bayesian statistics all the effects are considered random effects.
In implementing the Gibbs sampling, we considered the size of the initial chain of 300,000 cycles, so that the first 50,000 cycles were discarded, and sampled every 1,000 cycles; thus, totaling 250 samples. The convergence criterion adopted was 10 -9 .
To verify the accuracy of the estimates was used the Monte Carlo method to assess error, obtained by the square root of the value found in the division of the variance of each feature by the number of samples (VAN TASSEL & VAN VLECK, 1996).
The mean indexes of Nelore Mocho breed, the mean values, standard deviation, coefficient of variation, minimum and maximum values for growth traits have been measured and are presented on Table 1.
The result obtained for P120 (126 ± 20.11 kg) was similar to those found in the literature for Nelore, 126 ± 20.11 kg by GARNERO et al. The results observed in this study for the characteristic P365 (238 ± 38.64 kg) has also been similar to those found by Payá et al. (2007) and   The Table 3 shows the estimates of (co) variance components and The mean estimate of direct heritability for P365 was 0.28, and accompanied the median values obtained for the characteristic P120 and P210, unlike maternal heritability that showed lower magnitude, with range of 0.04 a 0.12. The estimated value for direct heritability P365 was Faria, Guimarães, Gomes, Miguel, Lemos, Pereira, Lôbo Bayesian analysis on the growth traits… lower than that estimated by Faria et al. (2007). These authors, conducting studies using the Bayesian analysis and Nelore animals found results of 0.49. Previous studies by Gonçalves et al. (2011) and Silva et al. (2013) have also estimated, for Nelore cattle, higher heritability than those found in this study for P365. The mean values estimated for direct heritability of P450 had high magnitude, 0.33, the same has been found by Faria et al. (2007) and Sena et al. (2013).  With regard to the characteristics GPND and GPDS, the direct heritability estimates were of medium magnitude, 0.18 and 0.24, respectively ( Table 4). The direct heritability estimate for GPND was similar than observed by Cardoso et al. (2001) who used data on weight gain from birth to weaning in Angus breed animals obtaining a result of 0.25. It did not occur with the direct heritability estimated by Cardoso et al. (2004) for post-weaning weight gain of Angus cattle whose value was 0.20. The results obtained by Laureano et al. (2011) compared the estimates of maternal heritability of GPND and GPDS, and direct heritability to GPDS were lower than those obtained in this study, being 0.06, 0.04 and 0.23, respectively. That same author has obtained a result of 0.21 of direct estimate of heritability, higher than that found on this study. Albuquerque and Meyer (2001) reported that with the animal aging the maternal additive genetic effect decreases, which it has also been observed in this study on the analysis of the maternal heritability that decreased with increasing age of animals. However, estimates of maternal Faria, Guimarães, Gomes, Miguel, Lemos, Pereira, Lôbo Bayesian analysis on the growth traits… heritability for growth traits evaluated in the growing phase were not zero, which proves the influence of maternal effect on the phenotype of growth characteristics evaluated annually and yearling and therefore, there is need to include this effect in models of genetic analysis.

As shown in
The average estimates for components of covariance between direct and maternal effects were negative for all growth traits evaluated, as well as the genetic correlation between direct and maternal additive effects, indicating antagonism between direct and maternal additive genetic effect.