Habilidade de predição de painéis de SNP customizados com diferentes densidades usando o método de passo único genômico BLUP em uma população de gado de corte
Carregando...
Data
2021-12-15
Autores
Rodríguez-Neira, Juan Diego
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Estadual Paulista (Unesp)
Resumo
A implementação da seleção genômica (SG) nos programas de melhoramento de bovinos de corte no Brasil tem sido suportado por o incremento do progresso genético das populações de bovinos Nelore. No entanto, um desafio de custo-benefício surge com base na quantidade e qualidade dos marcadores SNPs usados para predições acuradas dos valores genômicos (GEBVs). Portanto, o objetivo deste estudo é avaliar a habilidade de predição de marcadores SNPs em dados simulados e reais sobre características de alta e baixa herdabilidade por meio do método de passo único GBLUP, empregando painéis customizados de baixa e moderada densidade. No capítulo 2, oito réplicas de 18.000 animais genotipados foram simulados com painéis de 335k SNPs, dos quais 6.000 foram a população de validação para predizer os GEBVs. Critérios de alta (H-I) e baixa (L-I) informatividade além da localização equidistante (E-D) dos marcadores foram usados para customizar quatro densidades (35k, 16k, 4k e 2k) de painéis. Assim, em conjunto com o painel de 335k SNPs, treze cenários foram avaliados. Adicionalmente, os painéis customizados foram imputados a 335k SNPs e a acurácia da imputação foi estimada. Em geral, pelo menos 5% dos marcadores H-I ou E-D do painel de 335k SNPs alcançaram predições genômicas confiáveis e imputações acuradas. No capítulo 3, registros de 945 animais genotipados com alta densidade (HD) nascidos entre 1974 e 2018 para peso aos 210 (W210) e 450 (W450) dias de idade e o efeito materno para W210 (MW210) foram usados, dos quais 247 animais nascido despois do 2008 foram a população de validação. Cinco painéis de diferentes densidades (40k, 20k, 10k, 5k e 2k) foram customizados pelos critérios H-I e E-D e o painel HD foi usado como o cenário desejável. As predições dos GEBVs e a acurácia BIF (Beef Improvement Federation) foram obtidas com os programas da família BLUPF90. O método de regressão linear foi usado para avaliar a habilidade de predição, inflação e viés dos GEBVs para cada painel customizado. Uma superestimação da acurácia BIF foi observada quando diminuiu a densidade dos painéis. Para todas as características, a habilidade de predição apresentou aumento com altas densidades e foi similar para painéis customizados com densidades superiores a 10k SNPs, mesmo assim, a inflação foi baixa com altas densidades e o MW210 presentou as maiores inflações. O viés foi susceptível a superestimação dos GEBVs quando a densidade dos painéis customizados diminuiu. Em geral, a acurácia BIF e o nível do viés foram sensíveis aos painéis customizados de baixa densidade enquanto a habilidade de predição com pelo menos 5.000 H-I ou E-D marcadores a partir do painel HD, apresentou predições acuradas e com menos viés. Esses resultados indicaram que o desenvolvimento de painéis customizados de baixa densidade poderia ser uma abordagem viável para SG utilizando o método ssGBLUP e com melhor custo-benefício nos programas de melhoramento de bovinos de corte.
The implementation of genomic selection (GS) in Brazilian beef cattle breeding programs has been supported by the increase in genetic progress of Nellore cattle populations. However, from the fundamental principles of GS, a cost-benefit challenge arises based on the quantity and quality of SNP markers used for the prediction of reliable genomic values (GEBVs). Therefore, the objective of this study is to evaluate the prediction ability of SNP markers in simulated and real data on high and low heritability traits through of single-step GBLUP method, employing lower and moderate density customized arrays. In chapter 2, eight replicates of 18,000 genotyped animals were simulated with 335k SNPs array, of which 6,000 were the validation population to predict GEBVs. Four densities of customized SNPs arrays (35k, 16k, 4k and 2k) created from high (H-I) and low (L-I) informativeness and evenly distanced SNP markers (E-D) of SNPs, were the criteria to marker select. Thus, in conjunction with the 335k SNPs array, thirteen scenarios were evaluated. In addition, the customized arrays were imputed to 335k SNPs and the accuracy of imputation was estimated. Overall, at least 5% H-I or E-D markers from 335k array achieved reliable genomic predictions and accurate imputations. In chapter 3, records of 945 genotyped animal with highdensity (HD) born between 1974 and 2018 for weight to 210 (W210) and 450 (W450) days of age and maternal effect of W210 (MW210) were used, of which 247 animals born after 2008 were the validation population. Five density arrays were customized (40k, 20k, 10k, 5k and 2k) by H-I and E-D criteria and the HD array was used as desirable scenario. The GEBV predictions and accuracy BIF (beef improvement federation) were obtained with BLUPF90 family programs. The linear regression method was used to evaluate the prediction ability, inflation, and bias of GEBV of each customized array. An overestimation of BIF accuracy was observed when the density arrays decreased. For all traits, the prediction ability increased with higher densities and was similar for customized arrays above 10k, as well, the inflation was low with higher densities and the MW210 effect displayed the higher inflations. The bias was susceptible to overestimation of GEBVs when the density customized arrays decrease. Overall, the BIF accuracy and the level of bias were sensible to of low-density customized arrays while the prediction ability at least 5,000 H-I or E-D markers from HD array, displayed accurate and less biased predictions. These results indicated that the development of low-density customized arrays might be an approach feasible to GS under ssGBLUP method and cost-effective in beef cattle breeding programs.
The implementation of genomic selection (GS) in Brazilian beef cattle breeding programs has been supported by the increase in genetic progress of Nellore cattle populations. However, from the fundamental principles of GS, a cost-benefit challenge arises based on the quantity and quality of SNP markers used for the prediction of reliable genomic values (GEBVs). Therefore, the objective of this study is to evaluate the prediction ability of SNP markers in simulated and real data on high and low heritability traits through of single-step GBLUP method, employing lower and moderate density customized arrays. In chapter 2, eight replicates of 18,000 genotyped animals were simulated with 335k SNPs array, of which 6,000 were the validation population to predict GEBVs. Four densities of customized SNPs arrays (35k, 16k, 4k and 2k) created from high (H-I) and low (L-I) informativeness and evenly distanced SNP markers (E-D) of SNPs, were the criteria to marker select. Thus, in conjunction with the 335k SNPs array, thirteen scenarios were evaluated. In addition, the customized arrays were imputed to 335k SNPs and the accuracy of imputation was estimated. Overall, at least 5% H-I or E-D markers from 335k array achieved reliable genomic predictions and accurate imputations. In chapter 3, records of 945 genotyped animal with highdensity (HD) born between 1974 and 2018 for weight to 210 (W210) and 450 (W450) days of age and maternal effect of W210 (MW210) were used, of which 247 animals born after 2008 were the validation population. Five density arrays were customized (40k, 20k, 10k, 5k and 2k) by H-I and E-D criteria and the HD array was used as desirable scenario. The GEBV predictions and accuracy BIF (beef improvement federation) were obtained with BLUPF90 family programs. The linear regression method was used to evaluate the prediction ability, inflation, and bias of GEBV of each customized array. An overestimation of BIF accuracy was observed when the density arrays decreased. For all traits, the prediction ability increased with higher densities and was similar for customized arrays above 10k, as well, the inflation was low with higher densities and the MW210 effect displayed the higher inflations. The bias was susceptible to overestimation of GEBVs when the density customized arrays decrease. Overall, the BIF accuracy and the level of bias were sensible to of low-density customized arrays while the prediction ability at least 5,000 H-I or E-D markers from HD array, displayed accurate and less biased predictions. These results indicated that the development of low-density customized arrays might be an approach feasible to GS under ssGBLUP method and cost-effective in beef cattle breeding programs.
Descrição
Palavras-chave
Acurácia da predição, Simulação, Acurácia da imputação, Bovinos de corte, Genômica