Alternative bayesian models for genetic evaluation of biometrical, physical, and morphological reproductive traits in nelore bulls

Reproductive performance is one of the most important factors for the productive efficiency in beef cattle production. Biometric testicular and physical and morphological traits of the ejaculate are used to evaluate the reproductive performance of bulls. The phenotypic data of semen physical and morphological traits are expressed in percentage or notes, thus the evaluation of such traits through models that assume normal data distribution can be questioned. We aimed to compare the mixed models fitted under alternative and traditional Gaussian distributions for physical and morphological semen traits. Additionally to identify the most suitable model, we aimed also to predict genetic parameters for reproductive traits via Bayesian inference. Phenotypic data of 615 Nelore bulls, aged between 18 and 36 months, were used. The traits sperm motility (MOT), major (MD), minor (MID), and total (TD) sperm defects and percentage of normal spermatozoa were evaluated through Gaussian and Exponential models. For the physical traits expressed in scores, sperm vigor (VIG), and semen mass activity (MASS), the Gaussian and Poisson models were compared. Only Gaussian model was used for genetic parameters estimation of biometric testicular and seminal vesicle traits. The exponential model presented a better fit quality for MD and MID data than Gaussian model. For MASS the best model was Poisson. For all other evaluated traits, the Gaussian model presented the best fit. Heritability estimates were high for testicular biometric traits, ranging from 0.34 to 0.5. However, for the biometric measures of the seminal vesicle the heritabilities were low (0.04 for seminal vesicle length and 0.07 for seminal vesicle width). For the morphological traits of the semen, the heritability estimates were high, ranging from 0.36 to 0.50. For the semen physical traits, the heritability estimates varied widely, from 0.04 for MOT and VIG to 0.57 for MASS. The model assumption influences the bull genetic evaluation for physical and morphological semen traits, resulting in substantial ranking differences. However, the Gaussian model exhibited the best prediction accuracies for all traits.