Estratégias de seleção de linhagens de soja quanto a modelagem estatística e definição de ideótipo
Loading...
Date
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Universidade Federal de Viçosa
Abstract
Em um programada de melhoramento genético, a escolha de moledos estatísticos que forneçam valores genotípicos acurados é de extrema importância. Não obstante, índices que fazem uso da várias características de interesse agronômico para a seleção de genótipos superiores também são uma ferramenta importante para a obtenção de incremento em tais características. Este trabalho foi dividido em dois capítulos e objetivou: (1) Comparar modelos estatísticos, associados às estratégias de análises citadas, que favoreçam a maior eficiência na predição de valores genéticos de linhagens para características agronômicas de interesse; e (2) Comparar diferentes cenários para construção do ideótipo com o uso do MGIDI para a seleção de linhagens superiores de soja. Para tanto, ensaios foram implantados em Capinópolis nos anos agrícolaS de 2020/21 e 2021/22 e em Madre de Deus nos anos agrícolas de 20129/20 e 2021/22. As características avaliadas para os ensaios conduzidos em Capinópolis foram: produção (g), altura de planta (cm) e ciclo (dias), já em Madre de Deus a característica avaliada foi produção (g). No primeiro capítulo três modelos foram testados: (1) Delineamento em Blocos Aumentados (DBA); (2) Análise Espacial autorregressiva de duas dimensões (AR1 x AR1); e (3) Analise Espacial autorregressiva de duas dimensões com adição de bloco. Os componentes de variância foram obtidos pela metodologia de modelos mistos Restricted Maximum Likelihood (REML) e os valores genotípicos foram preditos utilizando Best Linear Unbased Predidction (BLUP). Analisou-se cada ambiente separadamente para cada variável. Para comparar os três modelos foram utilizadas a herdabilidade proposta por Cullis (h 2g ), a Acurácia com base na herdabilidade de Cullis (r ĝg ), Akaike Information Criterion (AIC) e Bayesian Information Criteria (BIC). Concluiu-se que os modelos que melhor se ajustam são os que consideram a correlação espacial entre as parcelas. Para três ambientes é preferencial a utilização da análise espacial sem a adição de blocos e para um abiente é preferível o modelo que abrange a análise espacial com a adição de blocos. De posse dos valores genétipos obtidos por meio dos modelos mistos, o MGIDI foi aplicado em diferentes cenários para verificação da melhor estratégia de definição do ideótipo. O ideotipo foi desenhado em seis diferentescenários: (1) considerando apenas o sentido; (2) considerando o sentido + peso; (3) considerando valores da média da melhor testemunha; (4) considerando valores da média da melhor testemunha + peso; (5) considerando valor de uma cultivar alvo para a região e (6) considerando valor de uma cultivar alvo para a região + peso. Os cenários foram comparados por meio do ganho de seleção (GS), do ganho de seleção ajustado (GSA) e do índice de coincidência (IC).As análses demonstraram que o cenário um é o que consegue maior ganho genético total quando consideradas todas as características simultaneamente, no entanto, o cenário cinco é o que consegue maiores ganhos de seleção para produção. Tendo em vista que não foram obtidos ganhos expressivos para ciclo, que a altura de planta da população selecionada já atende ao ideótipo requerido e que o incremento da produção é o maior objetivo de um programa de melhoramento, para este conjunto de dados, a melhor construção do ideótipo foi a utilizada no cenário cinco. Portanto, os genótipos selecionados foram: MDC1312S7, MDC139S3, MDC25101S2, MDC110S2, MDC292S7, MDC137S3, MDC141S5, MDC246S3, MDC13S2, MDC241S6, MDC82S2, MDC241S5, MDC248S1, MDC15S5, MDC141S6, MDC141S2, MDC81S5, MDC91S4 e MDC255S1. Keyword(s): Glycine max (L.) Merr.; Glycine max (L.) Merr.; Índice multi característico; MGIDI; Análise espacial; Modelos mistos; Delineamento aumentado
In an breeding program, the choice of statistical models that provide accurate genotypic values is extremely important. However, indices that make use of various characteristics of agronomic interest for the selection of superior genotypes are also an important tool for obtaining an increase in such characteristics. This work was divided into two chapters and aimed to: (1) Compare statistical models, associated with the aforementioned analysis strategies, which favor greater efficiency in predicting genetic values of lines for agronomic characteristics of interest and (2) Compare different scenarios for construction ideotype using MGIDI to select superior soybean lines. For tabto, Cultivation Value and Use (VCU) trials were implemented in Capinópolis in the agricultural years 2020/21 and 2021/22 and in Madre de Deus in the agricultural years 20129/20 and 2021/22. The characteristics evaluated for the trials conducted in Capinópolis were: productivity (g), plant height (cm) and cycle (days), while in Madre de Deus the characteristic evaluated was productivity (g). In the first chapter, three models were tested: (1) Augmented Block Design (DBA); (2) Two-dimensional autoregressive Spatial Analysis (AR1 x AR1) and (3) Two-dimensional autoregressive Spatial Analysis with block addition. The variance components were obtained using the Restricted Maximum Likelihood (REML) and the genotypic values were predicted using Best Linear Unbased Predidction (BLUP). Each environment was analyzed separately for each variable. To compare the three models, the heritability proposed by Cullis (h 2 g), Accuracy based on the heritability of Cullis (rĝg), Akaike Information Criterion (AIC) and Bayesian Information Criteria (BIC) were used. It was concluded that the models that best fit are those that consider the spatial correlation between plots. For three environments, the use of spatial analysis without the addition of blocks is preferred and for one environment, the model that covers spatial analysis with the addition of blocks is preferable. With the genotype values obtained through the mixed models, the MGIDI was applied in different scenarios to verify the best strategy for defining the ideotype. The ideotype was designed in six different scenarios: (1) considering only the meaning; (2) considering meaning + weight; (3) considering average values of the best witness; (4) considering average values of the best witness + weight; (5) considering the value of a target cultivar for the region and (6) considering the valueof a target cultivar for the region + weight. The scenarios were compared using the selection gain (GS), the adjusted selection gain (GSA) and the coincidence index (CI). The analyzes demonstrated that scenario one is the one that achieves the greatest total genetic gain when considering all characteristics simultaneously, however, scenario five is the one that achieves the greatest selection gains for productivity. Considering that no significant gains were obtained for the cycle, that the plant height of the selected population already meets the required ideotype and that increasing production is the main objective of a breeding program, for this set of data, the best construction of the ideotype was used in scenario five. Therefore, the selected genotypes were: MDC1312S7, MDC139S3, MDC25101S2, MDC110S2, MDC292S7, MDC137S3, MDC141S5, MDC241S5, MDC15S5, MDC246S3, MDC13S2, MDC241S6, MDC82S2, MDC248S1, 41S6, MDC141S2, MDC81S5, MDC91S4 and MDC255S1. Keyword(s): Glycine max (L.) Merr.; Multi-characteristic index; MGIDI; Spatial analysis; Mixed models; Augmented delineation
In an breeding program, the choice of statistical models that provide accurate genotypic values is extremely important. However, indices that make use of various characteristics of agronomic interest for the selection of superior genotypes are also an important tool for obtaining an increase in such characteristics. This work was divided into two chapters and aimed to: (1) Compare statistical models, associated with the aforementioned analysis strategies, which favor greater efficiency in predicting genetic values of lines for agronomic characteristics of interest and (2) Compare different scenarios for construction ideotype using MGIDI to select superior soybean lines. For tabto, Cultivation Value and Use (VCU) trials were implemented in Capinópolis in the agricultural years 2020/21 and 2021/22 and in Madre de Deus in the agricultural years 20129/20 and 2021/22. The characteristics evaluated for the trials conducted in Capinópolis were: productivity (g), plant height (cm) and cycle (days), while in Madre de Deus the characteristic evaluated was productivity (g). In the first chapter, three models were tested: (1) Augmented Block Design (DBA); (2) Two-dimensional autoregressive Spatial Analysis (AR1 x AR1) and (3) Two-dimensional autoregressive Spatial Analysis with block addition. The variance components were obtained using the Restricted Maximum Likelihood (REML) and the genotypic values were predicted using Best Linear Unbased Predidction (BLUP). Each environment was analyzed separately for each variable. To compare the three models, the heritability proposed by Cullis (h 2 g), Accuracy based on the heritability of Cullis (rĝg), Akaike Information Criterion (AIC) and Bayesian Information Criteria (BIC) were used. It was concluded that the models that best fit are those that consider the spatial correlation between plots. For three environments, the use of spatial analysis without the addition of blocks is preferred and for one environment, the model that covers spatial analysis with the addition of blocks is preferable. With the genotype values obtained through the mixed models, the MGIDI was applied in different scenarios to verify the best strategy for defining the ideotype. The ideotype was designed in six different scenarios: (1) considering only the meaning; (2) considering meaning + weight; (3) considering average values of the best witness; (4) considering average values of the best witness + weight; (5) considering the value of a target cultivar for the region and (6) considering the valueof a target cultivar for the region + weight. The scenarios were compared using the selection gain (GS), the adjusted selection gain (GSA) and the coincidence index (CI). The analyzes demonstrated that scenario one is the one that achieves the greatest total genetic gain when considering all characteristics simultaneously, however, scenario five is the one that achieves the greatest selection gains for productivity. Considering that no significant gains were obtained for the cycle, that the plant height of the selected population already meets the required ideotype and that increasing production is the main objective of a breeding program, for this set of data, the best construction of the ideotype was used in scenario five. Therefore, the selected genotypes were: MDC1312S7, MDC139S3, MDC25101S2, MDC110S2, MDC292S7, MDC137S3, MDC141S5, MDC241S5, MDC15S5, MDC246S3, MDC13S2, MDC241S6, MDC82S2, MDC248S1, 41S6, MDC141S2, MDC81S5, MDC91S4 and MDC255S1. Keyword(s): Glycine max (L.) Merr.; Multi-characteristic index; MGIDI; Spatial analysis; Mixed models; Augmented delineation
Description
Citation
RODRIGUES, Fernanda Cupertino. Estratégias de seleção de linhagens de soja quanto a modelagem estatística e definição de ideótipo. 2024. 84 f. Tese (Doutorado em Genética e Melhoramento) - Universidade Federal de Viçosa, Viçosa. 2024.
