Modelos de regressão aleatória para avaliação genética da curva de crescimento de ovinos da raça Santa Inês

• Autor(a) principal: Sarmento, José Lindenberg Rocha
• Data de Publicação: 2007
• Tipo de documento: Tese
• Idioma: por
• Título da fonte: LOCUS Repositório Institucional da UFV
• Texto Completo: http://locus.ufv.br/handle/123456789/1418
• topic: Avaliação genética; Ovinos deslanados; Regressão aleatória; Genetic evaluation; Ovine; Random regression; CNPQ::CIENCIAS AGRARIAS::ZOOTECNIA::GENETICA E MELHORAMENTO DOS ANIMAIS DOMESTICOS

Data set of 17,767 records from 4,210 Santa Inês lambs, belonging to Empresa Estadual de Pesquisa Agropecuária da Paraíba (EMEPA-PB) and Empresa Brasileira de Pesquisa Agropecuária (EMBRAPA Caprinos and EMBRAPA Tabuleiros Costeiros), were used to evaluate different random regression models to estimate (co)variance components and predict breeding values to the growth curve, aiming evaluate the importance on including maternal effect on the analysis model (genetic and permanent environment), determine the most suitable modeling for average growth trajectory, fit different structures to pattern residual variance and compare functions with different orders to fit changes on different random effects considered on analysis. Fixed and random regressions were fitted by Legendre polynomials in single-trait analysis with DXMRR software. An improvement was observed to data fitting when maternal effect (genetic or permanent environment) was included in the model. Direct heritability was increased by the maternal effect when it was not considered in analysis model. Permanent maternal environment effect contributed to maternal variance and increased maternal heritability when it was not included in the model. A cubic polynomial function estimated closely values to observed means to fit the average growth curve compared to lower and higher orders, especially in the end of the curve. Linear fitting were better than cubic function to estimate direct heritabilities. Changes on animal ranking based on predicted breeding values, using linear to cubic fitting, were small, however, the difference was higher on magnitude of predicted breeding values. The model was inadequate when considering homogeneity of residual variances. Fitting variance functions with any order was better than obtained by classes, which an order six polynomial provided best fitting among tested structures. Residue fitting intervened on estimates of animal variance (genetic and permanent environment) and at sire ranking based on predicted breeding values. A continuous function with orders three, three, five and three, respectively to direct genetic additive and maternal effect and permanent environment to animal and dam effect, was enough to describe changes on variances related to growth curve of Santa Inês sheep. Estimated heritabilities from birth to 56 days old were lower than maternal, reflecting major contribution from the dam over lambkin s phenotype in this growth stage. From this age, direct heritability became higher until the end of the studied growth period. A random regression models using cubic function to fixed curve, variance function of order six to residue and functions with orders three, three, five and three, respectively to direct genetic additive and maternal effect and permanent environment to animal and dam effect, allowed efficiently describing of variances and covariances related to the growth curve of the studied Santa Inês sheep.