Associação entre caracteres e modelagem de (co)variâncias na seleção de progênies de feijão
Arquivos
Data
2024-02-27
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Federal de Viçosa
Resumo
Em programas de melhoramento, a avaliação de progênies em uma série de experimentos visa ganhos com a seleção simultânea para um conjunto de caracteres de interesse. Esses ganhos dependem da magnitude e do sinal das correlações entre estes caracteres, podendo ser alterados em função dos ciclos de recombinação e dos ambientes em que famílias ou linhagens são avaliadas. Outro ponto a se considerar é que esses experimentos são, em geral, avaliados em diferentes ambientes (safra, ano, local) e delineamentos estatísticos. Assim, o desbalanceamento estatístico/genético é um problema que sempre está presente nestas avaliações, além da ocorrência de (co)variâncias entre os efeitos genéticos e não genéticos quando um conjunto de progênies são avaliadas em ambientes distintos, aumentando o erro e tornando a identificação daquelas superiores mais complexa. O uso da abordagem de modelos mistos torna-se necessária pois permite modelar diferentes estruturas de variâncias e covariâncias, além de modelar os efeitos genéticos e residuais na presença da interação G x A. No capítulo I objetivou-se estimar as correlações entre os caracteres aspecto comercial de grãos (AG), arquitetura de plantas (ARQ) e produtividade de grãos (PROD), considerando experimentos de avaliação de linhagens e de famílias de feijão, oriundas dos ciclos CIII e CIV do programa de seleção recorrente de feijão vermelho da UFV (PSRFV-UFV) em diferentes safras e anos. De forma geral, as estimativas do coeficiente de correlação foram de magnitudes baixas, porém negativas entre AG e PROD e positivas entre ARQ e PROD, porém de maiores magnitudes para famílias comparadas as linhagens e sem diferenças relevantes comparando estas estimativas nas diferentes safras. Assim, o trabalho do melhorista é facilitado uma vez que as associações entre PROD, AG e ARQ não trazem dificuldades ao melhoramento simultâneo desses caracteres neste programa de melhoramento. No capítulo II, os objetivos foram ajustar modelos, considerando diferentes estruturas de matrizes de (co)variâncias para um conjunto de dados de avaliação de famílias e linhagens oriundas do ciclo CIV do PSRFV-UFV; realizar a seleção por meio do índice de seleção FAI-BLUP, utilizando os valores genotípicos obtidos do modelo melhor ajustado, a fim de dar continuidade ao PSRFV-UFV. O modelo assumindo matrizes de (co)variâncias não estruturadas para os efeitos genéticos foi o que melhor se ajustou ao conjunto de dados. As estimativas de correlações genotípicas entre o desempenho dos genótipos (famílias e linhagens) nos pares de safras foram principalmente de magnitude moderada a alta, para ARQ e AG, corroborando com a fração simples da interação G×A predominando para estes caracteres. Em relação à PROD, estas correlações foram de baixa magnitude, o que corrobora com a predominância da fração complexa da interação G×A para esta característica. Utilizando os valores genotípicos preditos (BLUPs) a partir do modelo melhor ajustado, observou-se estimativas de acurácia seletiva e de comunalidade de maior magnitude se comparadas às obtidas com o modelo mais simples. Concluiu-se que a modelagem das estruturas de (co)variâncias para os efeitos genéticos e não genéticos é estratégia eficiente na análise de dados de experimentos que envolvem a avaliação de famílias ou linhagens em diferentes ambientes. Vinte famílias e 20 linhagens foram selecionadas visando a recombinação e a composição de futuros ensaios de valor de cultivo e uso (VCU), respectivamente. Palavras-chave: Correlações. Phaseolus vulgaris L.. Modelos mistos. Delineamentos experimentais. Feijão-comum.
In breeding programs, the evaluation of progênies across a series of experiments aims to gain from simultaneous selection for a set of traits of interest. These gains depend on the magnitude and sign of the correlations between these traits, which can be altered by recombination cycles and environments in which families or lines are evaluated. Another point to consider is that these experiments are generally evaluated in different environments (crop season, year, location) and statistical designs. Thus, statistical/genetic imbalance is a problem that is always present in these evaluations, in addition to the occurrence of covariance between genetic and non-genetic effects when a set of progenies are evaluated in different environments, increasing the error and making the identification of superior progenies more complex. The use of the mixed model approach becomes necessary as it allows modeling different structures of variance and covariances, as well as modeling genetic and residual effects in the presence of the genotype by environment interaction (G x E). Chapter I aimed to estimate the correlations between the commercial grain aspect (GA), plant architecture (PA) and grain yield (GY) traits, considering an experiment to evaluate red bean lines and families from cycles CIII and CIV of UFV's recurrent bean selection program in different crop seasons and years. In general, the correlation estimates were low, being negative for GA and GY and positive between PA and GY, with higher magnitudes for families compared to lines and no relevant differences when comparing crop seasons. It is concluded that the breeder's work is facilitated since there are no undesirable pleiotropic effects or the presence of linked genes in these traits, indicating that the associations between GY, GA, and PA do not pose difficulties for the simultaneous improvement of these traits in this breeding program. In Chapter II, the objectives were to fit a covariance structure model to the evaluation data set of families and lines from the CIV; to make a comparison between genetic and non-genetic parameters with the simplest model; to carry out selection using the FAI-BLUP selection index, using the genotypic values obtained from the best adjusted model, in order to continue the red bean improvement program. The unstructured covariance model for genetic effects was the best fit for the data set. An increase in heritability and accuracy estimates for both family and lines data, with the most accurate estimates being obtained with the best-fitting model. The genetic correlations between crops season were mainly of moderate to high magnitude, especially for PA and GA, with the simple fraction of the G×E interaction predominating. Regarding GY, the correlations were of low magnitude between environments, confirming the predominance of the complex fraction of the G×E interaction. Using the best linear predicted genotypic values (BLUPs) of the best- fit model confirms the increase in selective accuracy and communality values compared to the BLUPs of the simplest model, which demonstrates that it is more efficient. It was concluded that modeling the (co)variance structures for genetic and non-genetic effects is an efficient strategy in experiments involving the evaluation of families or lines in different environments. Through this approach, it was possible to select 20 best families for the fifth cycle of recurrent selection recombination and to select 20 best lines with the greatest potential for future Cultivation and Use Value tests. Keywords: Correlations. Phaseolus vulgaris L.. Mixed models. Experimental designs. Common bean.
In breeding programs, the evaluation of progênies across a series of experiments aims to gain from simultaneous selection for a set of traits of interest. These gains depend on the magnitude and sign of the correlations between these traits, which can be altered by recombination cycles and environments in which families or lines are evaluated. Another point to consider is that these experiments are generally evaluated in different environments (crop season, year, location) and statistical designs. Thus, statistical/genetic imbalance is a problem that is always present in these evaluations, in addition to the occurrence of covariance between genetic and non-genetic effects when a set of progenies are evaluated in different environments, increasing the error and making the identification of superior progenies more complex. The use of the mixed model approach becomes necessary as it allows modeling different structures of variance and covariances, as well as modeling genetic and residual effects in the presence of the genotype by environment interaction (G x E). Chapter I aimed to estimate the correlations between the commercial grain aspect (GA), plant architecture (PA) and grain yield (GY) traits, considering an experiment to evaluate red bean lines and families from cycles CIII and CIV of UFV's recurrent bean selection program in different crop seasons and years. In general, the correlation estimates were low, being negative for GA and GY and positive between PA and GY, with higher magnitudes for families compared to lines and no relevant differences when comparing crop seasons. It is concluded that the breeder's work is facilitated since there are no undesirable pleiotropic effects or the presence of linked genes in these traits, indicating that the associations between GY, GA, and PA do not pose difficulties for the simultaneous improvement of these traits in this breeding program. In Chapter II, the objectives were to fit a covariance structure model to the evaluation data set of families and lines from the CIV; to make a comparison between genetic and non-genetic parameters with the simplest model; to carry out selection using the FAI-BLUP selection index, using the genotypic values obtained from the best adjusted model, in order to continue the red bean improvement program. The unstructured covariance model for genetic effects was the best fit for the data set. An increase in heritability and accuracy estimates for both family and lines data, with the most accurate estimates being obtained with the best-fitting model. The genetic correlations between crops season were mainly of moderate to high magnitude, especially for PA and GA, with the simple fraction of the G×E interaction predominating. Regarding GY, the correlations were of low magnitude between environments, confirming the predominance of the complex fraction of the G×E interaction. Using the best linear predicted genotypic values (BLUPs) of the best- fit model confirms the increase in selective accuracy and communality values compared to the BLUPs of the simplest model, which demonstrates that it is more efficient. It was concluded that modeling the (co)variance structures for genetic and non-genetic effects is an efficient strategy in experiments involving the evaluation of families or lines in different environments. Through this approach, it was possible to select 20 best families for the fifth cycle of recurrent selection recombination and to select 20 best lines with the greatest potential for future Cultivation and Use Value tests. Keywords: Correlations. Phaseolus vulgaris L.. Mixed models. Experimental designs. Common bean.
Descrição
Palavras-chave
Phaseolus vulgaris, Feijão-comum - Melhoramento genético, Feijão-comum - Seleção, Correlação (Estatística)
Citação
ASSUNÇÃO NETO, Wilson Vitorino. Associação entre caracteres e modelagem de (co)variâncias na seleção de progênies de feijão. 2024. 87 f. Tese (Doutorado em Genética e Melhoramento) - Universidade Federal de Viçosa, Viçosa. 2024.