Inclusion of covariates in the animal model for the genetic evaluation of sheep for ultrasound measurement of the Longissimus thoracis et lumborum muscle area

The use of real-time ultrasonography for the evaluation of carcasses of live animals is subject to human errors. This may result in the wrong interpretation of images and affect the quality of information. In this study, we aimed to estimate variance components, heritability and genetic trend for loin eye area (LEA) measured using ultrasonography in Santa Inês sheep, with inclusion of different covariates. A total of 15 different models were analyzed including the direct additive genetic effect as random, the effects of contemporary group, year of birth and age class as fixed, and four covariates: age of the animal on the day of collection (AG), body weight (BW), loin eye depth (LED), and loin eye length (LEL). The variance components were estimated via Bayesian inference in single-trait analysis, using an animal model. The models were compared using the Bayes factor and the theoretical accuracy of estimated breeding values (EBVs). The heritability estimates ranged from 0.13 to 0.31. The EBVs were more accurate when the age and the body weight of the animal were considered as covariates in the models, with the accuracies increasing from 0.001 to 0.020. The average annual genetic change over the experimental period was 0.22 cm² per year for LEA. Nevertheless, the direct selection for LEA based on EBVs has not been performed by breeders in the farms where the data were collected, which indicates that the genetic gain for LEA is due to the correlated response in this trait for selection on BW and body conformation. The best estimates of heritability and genetic trend for LEA in Santa Inês sheep were obtained using models that included the animal body weight as a covariate.