Seleção genômica para precocidade sexual de machos da raça Nelore
Carregando...
Data
2021-08-10
Autores
Felipe, Ester Ferreira [UNESP]
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Estadual Paulista (Unesp)
Resumo
O objetivo deste trabalho foi avaliar a aplicação de diferentes modelos e testar diferentes cenários de treinamento e validação para a predição de valores genômicos para Idade à Puberdade de Machos (IPM), em bovinos da raça Nelore. O conjunto de dados utilizado pertence às fazendas participantes do Programa Nelore Brasil, da Associação Nacional de Criadores e Pesquisadores (ANCP), sendo dados de 11.095 animais fenotipados para IPM e 37.146 animais genotipados com o painel CLARIFIDE® Nelore 3.0 contendo 27.821 marcadores do tipo single nucleotide polymorphisms (SNP). Os efeitos dos marcadores foram estimados a partir dos dados genômicos, considerando diferentes distribuições a priori para os efeitos e variâncias dos SNPs. Os modelos estudados foram: Bayes A, Bayes B, Bayes C, Bayesian LASSO, GBLUP (Genomic Best Linear Unbiased Prediction) e ssGBLUP (single-step Genomic Best Linear Unbiased Prediction). A habilidade de predição das diferentes metodologias foi comparada testando diferentes cenários de treinamento e validação e por meio das correlações entre os pseudo-fenótipos (EBV (valor genético estimado) e Y* (fenótipo ajustado para os efeitos fixos)) e o valor genômico direto predito (DGV). Este trabalho teve como objetivo avaliar a aplicação dos modelos Bayesianos, GBLUP e ssGBLUP e testar diferentes cenários de treinamento e validação. Não houve diferença na habilidade de predição e sobretudo no viés, entre os modelos bayesianos, e estes são mais vantajosos para realizar a seleção genômica para IPM quando comparados ao GBLUP, sendo menos viesados e possuindo maior habilidade de predição. Não houve diferença na habilidade de predição entre os modelos bayesianos e o ssGBLUP, entretanto os modelos Bayes A, Bayes B e Bayes C apresentaram DGVs menos viesados. A metodologia mais adequada para predizer os valores genômicos da IPM foi a validação cruzada.
The objective of this work was to evaluate the application of different models and test different training and validation scenarios for the prediction of genomic values for Age at Puberty of Males (IPM) in Nellore cattle. The data set used belongs to farms participating in the Nelore Brazil Program, of the National Association of Breeders and Researchers (ANCP), with data from 11,095 animals phenotyped for IPM and 37,146 animals genotyped with the CLARIFIDE® Nellore 3.0 panel containing 27,821 single-type markers nucleotide polymorphisms (SNP). The effects of the markers were estimated from genomic data, considering different a priori distributions for the effects and variances of the SNPs. The models studied were: Bayes A, Bayes B, Bayes C, Bayesian LASSO, GBLUP (Genomic Best Linear Unbiased Prediction) and ssGBLUP (single-step Genomic Best Linear Unbiased Prediction). The predictive ability of the different methodologies was compared by testing different training and validation scenarios and through the correlations between the pseudo-phenotypes (EBV (estimated breeding value) and Y* (phenotype adjusted for fixed effects)) and the direct genomic value predicted (DGV). This work aimed to evaluate the application of Bayesian, GBLUP and ssGBLUP models and test different training and validation scenarios. There was no difference in prediction ability, and especially in bias, between Bayesian models, and these are more advantageous to perform genomic selection for IPM when compared to GBLUP, being less biased and having greater predictive ability. There was no difference in prediction ability between the Bayesian models and the ssGBLUP, however the Bayes A, Bayes B and Bayes C models presented less biased DGVs. The most adequate methodology to predict IPM genomic values was cross-validation.
The objective of this work was to evaluate the application of different models and test different training and validation scenarios for the prediction of genomic values for Age at Puberty of Males (IPM) in Nellore cattle. The data set used belongs to farms participating in the Nelore Brazil Program, of the National Association of Breeders and Researchers (ANCP), with data from 11,095 animals phenotyped for IPM and 37,146 animals genotyped with the CLARIFIDE® Nellore 3.0 panel containing 27,821 single-type markers nucleotide polymorphisms (SNP). The effects of the markers were estimated from genomic data, considering different a priori distributions for the effects and variances of the SNPs. The models studied were: Bayes A, Bayes B, Bayes C, Bayesian LASSO, GBLUP (Genomic Best Linear Unbiased Prediction) and ssGBLUP (single-step Genomic Best Linear Unbiased Prediction). The predictive ability of the different methodologies was compared by testing different training and validation scenarios and through the correlations between the pseudo-phenotypes (EBV (estimated breeding value) and Y* (phenotype adjusted for fixed effects)) and the direct genomic value predicted (DGV). This work aimed to evaluate the application of Bayesian, GBLUP and ssGBLUP models and test different training and validation scenarios. There was no difference in prediction ability, and especially in bias, between Bayesian models, and these are more advantageous to perform genomic selection for IPM when compared to GBLUP, being less biased and having greater predictive ability. There was no difference in prediction ability between the Bayesian models and the ssGBLUP, however the Bayes A, Bayes B and Bayes C models presented less biased DGVs. The most adequate methodology to predict IPM genomic values was cross-validation.
Descrição
Palavras-chave
Genética animal, Melhoramento genético, Puberdade, Bovino de corte, Touro