Métodos estatísticos na seleção genômica ampla para curvas de crescimento em animais

• Autor(a) principal: Rocha, Gilson Silvério da
• Data de Publicação: 2011
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: LOCUS Repositório Institucional da UFV
• Texto Completo: http://locus.ufv.br/handle/123456789/4041
• topic: Marcadores SNPs; Dados longitudinais; Suínos; SNPs markers; Longitudinal data; Pigs; CNPQ::CIENCIAS AGRARIAS

The main contribution of molecular genetics to the benefit of applied genetic breeding is the direct use of the DNA data in genomic selection, allowing high selective efficiency and speed in the acquisition of genetic gains in selection and low costs. A practical and consistent way of analyzing the productive efficiency of beef animals subjected to selection is through the study of growth curves, as these represent a longitudinal trajectory of the weights of the animals in function of time. Thus, firstly, growth models (non-linear models) are adjusted to the weight-age data of each animal submitted to selection and the parameters estimated as phenotypes are considered. This procedure permits to determine genetic parameter estimates for any growth trajectory point, and to understand the genetic architecture of the entire trajectory, since all the weighing information is condensed by these few biologically interpretable parameters. The parameters estimated from the growth models are used to predict the Genomic Breeding Value (GBV) by means of specific statistical methods for the Genome Wide Selection (GWS). The general objective of this work was to apply statistical methods used in the Genome Wide Selection, mainly RRBLUP/ GWS and the Bayesian LASSO on the study of animal growth curves, considering as phenotypic variables the estimates of the parameters of non-linear regression models. The specific objectives were: to estimate the genomic breeding values for each individual evaluated; to estimate the effect of SNP markers and to identify those with the greatest effects; to select, via grouping techniques, groups of individuals genetically superior, in relation to the growth curve; and to validate all the methodology used via simulation study and apply it to real data of an F2 population of swine originated from the cross of two males from the naturalized Brazilian race Piau with 18 females of a commercial line (Landrace × Large White × Pietrain).The results indicated that the Genome Wide Selection statistical methods were efficient in studying the growth curves, considering simulated and real swine weight-age data. GWS presented high accuracy in the selection of the growth curve trajectory, allowing the detection of the QTLs (Quantitative Trait Loci) for the curve parameters of the individuals studied. In the absence of genes of significant effect, the methods RR-BLUP/GWS and Bayesian LASSO showed similar results but the latter showed more efficiency when the halothane gene, characterized as of significant effect, was included as a marker in the analyses.