Seleção genômica para características longitudinais de bovinos da raça Holandesa

• Autor(a) principal: Scalez, Daiane Cristina Becker [UNESP]
• Data de Publicação: 2016
• Tipo de documento: Tese
• Idioma: por
• Título da fonte: Repositório Institucional da UNESP
• Texto Completo: http://hdl.handle.net/11449/138211
• topic: Bayesian LASSO; Random regression models; Nonlinear models; RRBLUP; LASSO Bayesiano; RR-BLUP; Modelos de regressão aleatória; Modelos não lineares

Longitudinal traits are of great importance for animal production, however, when it comes to genomic selection several methods are being evaluated for specific traits, such as milk yield in specific ages or total yield. Thus, applying these methods to complete trajectories of the phenotype of interest over time can be an important tool for helping to decide the best time to select the animals. In this study, milk yield, fat and protein percentage, and somatic cell score data of dairy cattle were used, in a two-stage procedure: i) where the nonlinear Wood (1967), Cobby & Le Du (1978) and Wilmink (1987) were adjusted; ii) the model that provided the best fit was the Wilmink model (1987), whose parameter estimates were used as phenotypes later. In addition to the non-linear models, it was also used a random regression model, considering quadratic Legendre polynomials to model the additive genetic and permanent environment effects. In the second stage different methods for estimation of genomic breeding values (GEBV) were applied: RR-BLUP and LASSO, which considered as phenotypes the estimated parameters previously obtained and the random regression coefficients. The correspondence between phenotypes and GEBV was verified by through cross validation using the k-fold method. For all traits, the genomic prediction accuracies of Wilmink’s model parameters were low when RR-BLUP was used. Furthermore, the accuracies for parameter c showed values close to zero for most traits. The prediction accuracies for random regression coefficients were moderate to high, above 0.67, and demonstrated that the use of random regression coefficients can be a good alternative for genomic prediction using longitudinal data.