Bayesian random regression threshold models for genetic evaluation of pregnancy probability in Red Sindhi heifers

Abstract

We proposed a Bayesian random regression threshold model for genetic evaluation of pregnancy probability (PP) in Red Sindhi heifers over different months. Since this breed was recently introduced in Brazil, the age of 14 months usually preconized for Nellore cattle may not reflect the reality of the fertility indicator. In this context, the pregnancy success was evaluated at other ages aiming to understand its genetic variability pattern over time. A total of 4828 phenotyped heifers belong to 657 contemporary groups were used for the analysis. The estimated connectedness was equal to 99.75% considering 6189 individuals in the final pedigree. The random regression threshold models were implemented by combining second, third and fourth order Legendre polynomials to describe the average, the additive genetic and the permanent environmental curves. Additionally, the heterogeneity of the residual variance was also tested here. Based on DIC (deviance information criterion) and posterior model probabilities, the fourth order Legendre polynomials (LEG_4441) for the average, the additive genetic and the permanent environmental effects, assuming homogeneity of the residual variance, outperformed the simplest models. Thus, the fitting quality compensated the increased in the model complexity. The negative genetic correlations between PP at 15 months with PP at latter months indicates that rankings of animals for selection would be few similar in advanced ages. The high values for heritability (varying from 0.32 to 0.45) suggest that PP, mainly until 19 months, can be used as selection criterion for reproductive female performance in breeding programs for Red Sindhi cattle in Brazil. These practical results were only obtained due to the advantages of the proposed random regression threshold models to describe PP over different months.