Métodos paramétricos e não paramétricos para a predição de valores genéticos genômicos de características de importância econômica em suínos
Carregando...
Data
2019-11-29
Autores
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Estadual Paulista (Unesp)
Resumo
A seleção genômica tem sido usada em vários programas de melhoramento de plantas e animais proporcionando ganhos na acurácia de seleção quando comparado a seleção tradicional. Contudo, é importante ressaltar que a superioridade da seleção genômica depende de vários fatores tais como a metodologia de predição dos valores genéticos genômicos. Além do desafio de encontrar a melhor metodologia para a aplicação da seleção genômica, o alto custo de implementação também representa uma dificuldade na sua ampla utilização nos programas de melhoramento animal, principalmente em suínos e aves. Os objetivos deste trabalho foram: (i) avaliar a capacidade de predição de quatro diferentes métodos de seleção genômica para características reprodutivas e produtivas; (ii) utilizar o método “Random Forest” para selecionar conjuntos de marcadores SNPs mais relevantes na explicação dos fenótipos para as características estudadas e verificar o impacto do uso desses conjuntos de SNPs na acurácia dos valores genômicos preditos para duas características de importância econômica na suinocultura. No estudo foi utilizado um arquivo de pedigree com 879.965 animais distribuídos em 13 gerações e um arquivo de fenótipos composto por 73.439 observações de espessura de gordura (EG) e 69.505 registros de número de leitões nascidos vivos por leitegada (NV). Além disso, informações de 969 animais de uma linhagem fêmea tipo Landrace genotipados com um painel customizado com 57.692 marcadores distribuídos por todo o genoma foram utilizadas. A seleção genômica foi realizada aplicando os métodos lineares GBLUP – “Genomic Estimated Breeding Value” e “Single-step” GBLUP e os não lineares brnn - “Bayesian Regularized Neural Network” e snnR – “Sparse Neural Networks”. Para verificar o impacto do uso de conjuntos de SNPs na habilidade de predição da seleção genômica foram avaliados 5 cenários: 1 - 3.000 SNPs mais importantes para EG; 2 - 3.000 SNPs mais importantes para NV; 3- 3.000 SNPs mais importantes para as duas características em conjunto; 4 - 3.000 SNPs selecionados aleatoriamente; e 5 - todos os SNPs. O método “Random Forest” foi utilizado para seleção dos SNPs dos três primeiros cenários e os valores genéticos genômicos foram preditos utilizando o método “Single-step” GBLUP. No estudo das metodologias de seleção genômica verificou-se que as acurácias de predições foram melhores para a característica EG em relação a característica NV. Também se observou que o método que apresentou a melhor predição dos valores genéticos genômicos na população estudada para ambas as características foi “single-step” GBLUP quando comparado ao método GBLUP e os métodos de redes neurais snnR e brnn. Na avaliação da predição dos valores genéticos genômicos utilizando conjuntos de SNPs identificados com o método de aprendizagem de máquina, demonstrou-se que esse método pode ser utilizado para selecionar eficientemente conjuntos de marcadores para formarem painéis de genotipagem de baixa densidade e assim reduzir os custos da implementação da seleção genômica. Isto porque, o conjunto de marcadores selecionados com o mesmo apresentou habilidade de predição semelhante quando comparado a habilidade de predição com a informação de todos os SNPs na seleção genômica apenas para a característica número de leitões nascidos vivos.
Genomic selection has been used in several plant and animal breeding programs providing selection accuracy gains when compared to traditional genetic analysis. However, it is important to emphasize the superiority of genomic selection depends on several factors, such as the methodology for predicting genomic values. In addition to the challenge of finding the best model for application of genomic selection, the high cost of implementation also makes difficult the use of genomic selection in breeding programs of all animal species. Therefore, the aims of this study were: (i) evaluate the prediction ability of four different genomic selection methods using reproductive and productive traits; (ii) use Random Forest analysis to select the most important markers for studied traits and to verify the impact of using these subsets of SNPs on the accuracy of predicted genomic values for two economically important traits in pig production. The pedigree files contained 879,965 individuals, which spanned up to 13 generations and phenotype file composed of 73,439 backfat thickness observations and 69,505 litter records. A total of 969 animals of a Landrace female line were genotyped using a custom Illumina chip consisting of 57,692 SNPs distributed throughout the genome. Genomic selection was performed by applying the linear methods GBLUP (Genomic Estimated Breeding Value) and Single-step GBLUP and the nonlinear methods brnn (Bayesian Regularized Neural Network) and snnR (“Sparse Neural Networks). The prediction capacity of each method was tested by Pearson correlation, mean square measure error and the regression coefficient. To verify the impact of the use of SNP sets on the prediction accuracy of genomic selection, 5 scenarios were evaluated: 1 - 3,000 most important SNPs for backfat thickness; 2 - 3,000 most important SNPs to born alive piglets; 3 - 3,000 most important SNPs for both traits together; 4 - 3,000 randomly selected SNPs; and 5 all SNPs. The Random Forest method was used to select SNPs from the first three scenarios and genomic selection was performed using the GBLUP single-step method. In the study of genomic selection methodologies, it was found that the prediction accuracy was better for the backfat thickness trait in relation to the number of piglets born alive. The results also show that the method that presented the best prediction of genomic genetic values in the population studied for the backfat thickness and number of piglets born alive was single-step GBLUP when compared to GBLUP and neural network methods snnR and brnn. In assessing the prediction of genomic genetic values using subsets of SNP selected with the machine learning method, it has been shown that the Random Forest method can be used to efficiently select subset marker to form low density genotyping panels and thus reduce the costs of implementing genomic selection. This is because the set of markers selected from using machine learning method presented similar prediction ability when compared the prediction ability with the information of all SNPs in the genomic selection for the trait number of born alive piglets.
Genomic selection has been used in several plant and animal breeding programs providing selection accuracy gains when compared to traditional genetic analysis. However, it is important to emphasize the superiority of genomic selection depends on several factors, such as the methodology for predicting genomic values. In addition to the challenge of finding the best model for application of genomic selection, the high cost of implementation also makes difficult the use of genomic selection in breeding programs of all animal species. Therefore, the aims of this study were: (i) evaluate the prediction ability of four different genomic selection methods using reproductive and productive traits; (ii) use Random Forest analysis to select the most important markers for studied traits and to verify the impact of using these subsets of SNPs on the accuracy of predicted genomic values for two economically important traits in pig production. The pedigree files contained 879,965 individuals, which spanned up to 13 generations and phenotype file composed of 73,439 backfat thickness observations and 69,505 litter records. A total of 969 animals of a Landrace female line were genotyped using a custom Illumina chip consisting of 57,692 SNPs distributed throughout the genome. Genomic selection was performed by applying the linear methods GBLUP (Genomic Estimated Breeding Value) and Single-step GBLUP and the nonlinear methods brnn (Bayesian Regularized Neural Network) and snnR (“Sparse Neural Networks). The prediction capacity of each method was tested by Pearson correlation, mean square measure error and the regression coefficient. To verify the impact of the use of SNP sets on the prediction accuracy of genomic selection, 5 scenarios were evaluated: 1 - 3,000 most important SNPs for backfat thickness; 2 - 3,000 most important SNPs to born alive piglets; 3 - 3,000 most important SNPs for both traits together; 4 - 3,000 randomly selected SNPs; and 5 all SNPs. The Random Forest method was used to select SNPs from the first three scenarios and genomic selection was performed using the GBLUP single-step method. In the study of genomic selection methodologies, it was found that the prediction accuracy was better for the backfat thickness trait in relation to the number of piglets born alive. The results also show that the method that presented the best prediction of genomic genetic values in the population studied for the backfat thickness and number of piglets born alive was single-step GBLUP when compared to GBLUP and neural network methods snnR and brnn. In assessing the prediction of genomic genetic values using subsets of SNP selected with the machine learning method, it has been shown that the Random Forest method can be used to efficiently select subset marker to form low density genotyping panels and thus reduce the costs of implementing genomic selection. This is because the set of markers selected from using machine learning method presented similar prediction ability when compared the prediction ability with the information of all SNPs in the genomic selection for the trait number of born alive piglets.
Descrição
Palavras-chave
Brnn, Espessura de gordura, Número de leitões nascidos vivos, “Random Forest”, SnnR