Strategies to implement genomic selection in Montana composite beef cattle
Carregando...
Arquivos
Data
2021-01-06
Autores
Kluska, Sabrina [UNESP]
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Estadual Paulista (Unesp)
Resumo
A busca por novas características e métodos capazes de incrementar o ganho genético em programas de melhoramento é constante. Com o advento da seleção genômica um grande progresso genético tem sido observado. Entretanto, a seleção genômica em animais cruzados, compostos ou populações multirraciais ainda é pouco difundida. Desta forma, os objetivos do presente estudo foram: i) Investigar o impacto do uso de diferentes matrizes de parentesco na avaliação genética de bovinos Montana; ii) Investigar o efeito do uso de metafundadores (MFs) e grupos de pais desconhecidos (UPG) na avaliação genética de bovinos Montana. Foram utilizados registros de 680.551 animais no pedigree, dos quais, 1899 foram genotipados com painéis de diferentes densidades e posteriormente imputados para o painel da Neogen GeneSeek® Genomic Profiler (GGP) com aproximadamente 30.000 SNPs. Informações fenotípicas de circunferência escrotal aos 12 meses de idade (SC12), ganho de peso pós desmame (PWG), peso à desmama (WW) e peso ao nascimento (BW) foram utilizadas. Quatro matrizes de parentesco distintas, e um modelo unicaracterística, foram utilizadas: 1) Matriz de parentesco aditivo baseada no pedigree (A); 2) Matriz de parentesco genômico, construída como no ssGBLUP default (G_1); 3) Matriz de parentesco genômico, centrada com base nas frequências alélicas dos grupos de tipo biológico ou de componentes principais (G_2); 4) Matriz de parentesco genômico, centrada e escalada com base nas frequências alélicas dos grupos de tipo biológico ou de componentes principais (G_3). Além disso, metafundadores e grupos de pais desconhecidos foram utilizados, os quais foram empregados em modelos multicaracterística BLUP e ssGBLUP. Foram testados modelos com quatro ou dez MFs e UPGs. UPGs foram adicionados na matriz H no ssGBLUP (ssGBLUP_UPG) ou apenas nas matrizes A^(-1) e A_22^(-1) (ssGBLUP_UPGA). Para a validação dos resultados, o método baseado em estatísticas de regressão linear (LR), e 436 animais com fenótipos omitidos, foram utilizados. Ajustes nas matrizes de parentesco genômico não foram capazes de captar maior proporção de variância genética aditiva em relação a variância fenotípica, ou seja, produzir maior herdabilidade. A adição da informação genômica, no modelo, em ambos os estudos, foi capaz de incrementar a estabilidade dos GEBVs. Contudo, a estabilidade dos GEBVs foi superior quando o modelo unicaracterística foi utilizado. Todos os parâmetros de comparação utilizados (estabilidade dos (G)EBVs, acurácia dos modelos, acurácia BIF, dispersão, viés, média dos (G)EBVs e correlação de Spearman) não indicaram nenhuma diferença significativa nas predições quando as matrizes de relacionamento genômico foram ajustadas com base em tipos biológicos ou grupos de componentes principais. As correlações de Spearman dos valores genéticos, entre os modelos baseados em pedigree e genômicos foram baixas, indicando mudanças no ranking dos animais quando a seleção é praticada com estes modelos. Entretanto, quando comparados os modelos genômicos, com G ajustada ou não, a correlação entre os GEBV foi alta, indicando pouca ou nenhuma mudança na classificação dos candidatos a seleção quando a seleção é realizada com base em qualquer um dos modelos genômicos. No geral a inclusão de UPGs nos modelos produziu estabilidade e dispersão similar aos demais modelos genômicos, entretanto, tendências genéticas viesadas foram observadas quando estes foram incluídos somente nas matrizes de parentesco baseado no pedigree. A inclusão de metafundadores no modelo não foi capaz de provocar mudanças consideráveis na estabilidade dos (G)EBVs em avaliações subsequentes, exceto para PWG, e dispersão dos modelos. Entretanto, os modelos com metafundadores, quatro ou dez, produziram um viés menor do que os demais modelos genômicos, e semelhante ao BLUP baseado no pedigree. Estes resultados indicam que o uso de metafundadores pode reduzir o viés das avaliações genômicas em bovinos Montana ao mesmo nível dos modelos baseados em pedigree, com acurácia levemente menor e correlação ou estabilidade dos (G)EBV similar ao ssGBLUP default.
The search for new traits and methods able to increase the genetic gain in breeding programs is constant. With the advent of genomic selection, a great genetic progress has been observed, however the genomic selection in crossed animals, composite or multiracial populations is still not widespread. Thus, the aim of this study were: i) Investigate the impact of the use of different relationship matrices in the genetic evaluation of Montana cattle; ii) Investigate the effect of using metafounders (MFs) and unknown parents groups (UPG) on the genetic evaluation of Montana cattle. Records of 680,551 animals in the pedigree were available, of which 1,899 were genotyped with panels of different densities and subsequently imputed to the Neogen GeneSeek® Genomic Profiler (GGP) panel with around 30,000 SNPs. Phenotypic records of scrotal circumference at 12 months of age (SC12), postweaning weight gain (PWG), weaning weight (WW) and birth weight (BW) were available. Four distinct relationship matrices, and a single-trait model, were used: 1) Additive relationship matrix based on the pedigree (A); 2) Genomic relationship matrix, built as in the default ssGBLUP (G1); 3) Genomic relationship matrix, centered based on the specific-allele frequencies of the groups biological type groups or principal components (G2); 4) Genomic relationship matrix, centered and scaled based on the specific-allele frequencies of the groups of biological type or principal components (G3). In addition, metafounders and unknown parent groups were implemented in a multi-trait model at BLUP and ssGBLUP. Models with four or ten MFs and UPGs were tested. UPGs were added in the H matrix in ssGBLUP (ssGBLUP_UPG) or only in A^(-1) and A_22^(-1) (ssGBLUP_UPGA) matrices. For the validation, the method based on linear regression statistics (LR), and 436 animals with omitted phenotypes, were used. Adjustments in the genomic relationship matrices were not able to capture a greater proportion of additive genetic variance in relation to phenotypic variance, that is, to produce greater heritability. The addition of genomic information, in the model, in both studies, was able to increase the stability of the (G)EBVs. However, the stability of the (G)EBVs was higher when the single-trait model was used. All the comparison parameters used (stability of (G)EBVs, accuracy of models, BIF accuracy, dispersion, bias, mean of (G)EBVs and Spearman correlation) did not indicate any significant difference in predictions when the genomic relationship matrices were adjusted based on biological types or groups of principal components. Spearman's correlations of genetic values between pedigree-based and genomic models were low, indicating changes in the ranking of animals when selection is practiced with these models. However, when comparing the genomic models, with G adjusted or not, the correlation between the GEBVs was high, indicating little or no change in the classification of candidates for selection when the selection is made based on any of the genomic models. Overall, the inclusion of UPGs in the models produced a stability and dispersion similar to the other genomic models, however, biased genetic tendencies were observed when UPGs were considered only in the pedigree-based matrices. The inclusion of metafounders in the model was not able to produce considerable changes in the stability of (G)EBVs in two subsequent evaluations, except for PWG and dispersion of the models. However, the models with metafounders, either four or ten, produced a lower bias than the other genomic models, and similar to the BLUP based on the pedigree. These results indicate that the use of metafounders can reduce the bias of genomic predictions in Montana cattle to the same level as the pedigree-based models, with slightly less accuracy and correlation or stability of (G)EBV similar to the ssGBLUP default.
The search for new traits and methods able to increase the genetic gain in breeding programs is constant. With the advent of genomic selection, a great genetic progress has been observed, however the genomic selection in crossed animals, composite or multiracial populations is still not widespread. Thus, the aim of this study were: i) Investigate the impact of the use of different relationship matrices in the genetic evaluation of Montana cattle; ii) Investigate the effect of using metafounders (MFs) and unknown parents groups (UPG) on the genetic evaluation of Montana cattle. Records of 680,551 animals in the pedigree were available, of which 1,899 were genotyped with panels of different densities and subsequently imputed to the Neogen GeneSeek® Genomic Profiler (GGP) panel with around 30,000 SNPs. Phenotypic records of scrotal circumference at 12 months of age (SC12), postweaning weight gain (PWG), weaning weight (WW) and birth weight (BW) were available. Four distinct relationship matrices, and a single-trait model, were used: 1) Additive relationship matrix based on the pedigree (A); 2) Genomic relationship matrix, built as in the default ssGBLUP (G1); 3) Genomic relationship matrix, centered based on the specific-allele frequencies of the groups biological type groups or principal components (G2); 4) Genomic relationship matrix, centered and scaled based on the specific-allele frequencies of the groups of biological type or principal components (G3). In addition, metafounders and unknown parent groups were implemented in a multi-trait model at BLUP and ssGBLUP. Models with four or ten MFs and UPGs were tested. UPGs were added in the H matrix in ssGBLUP (ssGBLUP_UPG) or only in A^(-1) and A_22^(-1) (ssGBLUP_UPGA) matrices. For the validation, the method based on linear regression statistics (LR), and 436 animals with omitted phenotypes, were used. Adjustments in the genomic relationship matrices were not able to capture a greater proportion of additive genetic variance in relation to phenotypic variance, that is, to produce greater heritability. The addition of genomic information, in the model, in both studies, was able to increase the stability of the (G)EBVs. However, the stability of the (G)EBVs was higher when the single-trait model was used. All the comparison parameters used (stability of (G)EBVs, accuracy of models, BIF accuracy, dispersion, bias, mean of (G)EBVs and Spearman correlation) did not indicate any significant difference in predictions when the genomic relationship matrices were adjusted based on biological types or groups of principal components. Spearman's correlations of genetic values between pedigree-based and genomic models were low, indicating changes in the ranking of animals when selection is practiced with these models. However, when comparing the genomic models, with G adjusted or not, the correlation between the GEBVs was high, indicating little or no change in the classification of candidates for selection when the selection is made based on any of the genomic models. Overall, the inclusion of UPGs in the models produced a stability and dispersion similar to the other genomic models, however, biased genetic tendencies were observed when UPGs were considered only in the pedigree-based matrices. The inclusion of metafounders in the model was not able to produce considerable changes in the stability of (G)EBVs in two subsequent evaluations, except for PWG and dispersion of the models. However, the models with metafounders, either four or ten, produced a lower bias than the other genomic models, and similar to the BLUP based on the pedigree. These results indicate that the use of metafounders can reduce the bias of genomic predictions in Montana cattle to the same level as the pedigree-based models, with slightly less accuracy and correlation or stability of (G)EBV similar to the ssGBLUP default.
Descrição
Palavras-chave
BLUP genômico de passo-único, Grupos genéticos, Matriz de parentesco genômico, Metafundadores, Seleção genômica