Seleção e associação genômica ampla para características de crescimento e escores visuais em bovinos da raça Hereford e Braford.

• Ano de defesa: 2017
• Autor(a) principal: CAMPOS, G. S.
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Palavras-chave em Português: Validação cruzada; Predição genômica; Modelo Bayesiano; Bovino; Gado de Corte; Genoma; Método Estatístico
• Link de acesso: http://www.alice.cnptia.embrapa.br/alice/handle/doc/1099079

This thesis was structured in three chapters. In the first one, the genetic parameters were estimated using linear and threshold models for the traits of visual scores and also the cross validation and multinomial regression were used for validation of the models. There was no difference in the parameter estimation when the scores had normal distribution, such as for conformation, precocity, musculature and size. Heritability values (h2) ranged from 0.18 to 0.26 with the linear model and from 0.19 to 0.29 with the threshold. However, when the score had no normal distribution, such as navel, there were advantages in using the threshold model, with a h2 value of 0.42 and a linear model of 0.22. The second study aimed to evaluate the accuracy genomic predictions using different methods, for growth traits and visual scores obtained at weaning and yearling in cattle of the Hereford and Braford breeds. Phenotype data 126,290 animals belonging to the Delta G Connection breeding program and a set of 3,552 genotyped animals were used. The GBLUP, BayesB and BayesC methods were tested and higher accuracy were obtained with Bayesian methods. For the growth traits, greater gain in accuracy compared to the traditional method (BLUP) was with the BayesB methodology for birth weight (BW) of 23.8%, and for the visual scores it was for size at the yearling (SY), of 29.8% with the BayesB and BayesC methods. For the approaches combining all sources of information, greater gains were obtained with the single-step ssGBLUP methodology. Among all the characteristics, for weaning measures, the average gain was 40.7% for the weaning measures and 36.7% for yearling. Lower prediction accuracy was observed in the groups containing only Hereford cattle, indicating that the training set composed of the majority of Braford animals will not estimate accurate predictions for the Hereford in the validation set. The third study aimed to perform a genome wide association study (GWAS) using Bayesian methodology to identify the most representative genomic regions and SNPs associated with growth traits. It was selected the most representative windows and the SNPs that explained more than 20% of the genetic variance estimated for the traits studied. After this selection, the most informative SNPs regarding parameters, model frequency (MF), t-like (TL), linkage disequilibrium (DL) and minor allele frequency (MAF) were used in a panel of low density. Reduced panel accuracy was estimated from crossvalidation, using k-means and random clustering methods. Higher accuracy estimates were obtained for weaning characteristics. Greater gains in accuracy can be obtained if more animals are genotyped and phenotyped. These panels may be useful for future studies related to fine mapping for the discovery of causal variants and are an interesting alternative for reducing the costs of genotyping and implementation of genomic selection.