Diversidade fenotípica e determinação do teor de óleo em pinhão-manso por espectroscopia no infravermelho próximo
Arquivos
Data
2016-04-29
Autores
Moraes, Amanda Carolina Prado de [UNESP]
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Estadual Paulista (Unesp)
Resumo
O objetivo do presente trabalho consistiu em caracterizar a diversidade fenotípica em genótipos de pinhão-manso através de análises multivariadas, além de desenvolver um modelo de calibração para a determinação do teor de óleo por meio da espectroscopia do infravermelho próximo (NIR) visando a seleção de genótipos com alto teor de óleo. Para a análise fenotípica, foram avaliados 11 caracteres agronômicos em 165 genótipos de pinhão-manso pertencentes ao banco de germoplasma in vivo localizado no Departamento de Produção Vegetal da UNESP/FCAV, Jaboticabal, SP. Os caracteres utilizados foram: altura da planta, diâmetro do caule, número de ramificações primárias, comprimento de fruto, largura de fruto, peso de fruto, formato de fruto, comprimento de semente, largura de semente, peso de semente e teor de óleo. Os dados foram submetidos à análise de componentes principais (PCA), análise de agrupamento pelo método hierárquico de Ward e método não hierárquico k-médias. O caráter formato de fruto foi retirado das análises multivariadas por ser de natureza qualitativa. Na PCA, os 4 CPs explicaram 71,62% da variância total. Os caracteres de maior contribuição para a discriminação dos genótipos no CP1 foram peso de semente, largura de fruto, comprimento de fruto e peso de fruto. Destacaram-se 22 genótipos promissores com potencial para serem explorados em programas de melhoramento genético. O dendrograma obtido pela análise de agrupamento de Ward e para a análise de k-médias geraram nove grupos, dos quais 5 agruparam genótipos com características vantajosas. As análises multivariadas permitiram a caracterização dos genótipos, possibilitando a seleção para muitos caracteres avaliados e a indicação de estratégias de ganhos com a seleção, para programas de melhoramento genético. Para a construção da curva de calibração e validação para teor de óleo, foram utilizadas sementes de 174 plantas, que tiveram o óleo extraído pelo método "Soxhlet" para obtenção dos valores de referência, que foram entre 50,27% a 63,21% (m/m) de óleo. Os espectros foram coletados das sementes sem tegumento, intactas, por acessório de rotação, por meio da espectroscopia por infravermelho próximo (NIR) em equipamento modelo Tango, Bruker. Cada espectro foi retirado com uma média de 64 scans com resolução de 16 cm-1. As regiões espectrais selecionadas para a curva de calibração foram de 8992 a 7424 cm-1 e 5336 a 4928 cm-1. Os coeficientes de correlação (R2) da curva construída foram de 73,23% para calibração e de 70,79% para validação, sendo o erro quadrado médio de calibração (RMSEC) de 1,35 e o erro quadrado médio de validação cruzada (RMSECV) de 1,40 entre os dados obtidos pelo método de referência e pelo NIR. Os resultados indicaram que o teor de óleo da espécie Jatropha curcas L. pode ser estimado por NIR, e o método de referência utilizado foi satisfatório, apresentando boa correlação com a informação espectral, tornando a curva construída confiável. Contudo, futuramente é necessária a adição de novos valores para a melhoria da sua robustez e amplitude.
The purpose of this study was to characterize the phenotypic diversity in Jatropha genotypes by multivariate analysis, and developing a calibration model for determination oil content by Near-infrared spectroscopy (NIR) in order to select genotypes with high oil content. For the phenotypic analysis, 11 agronomic characters were evaluated in 165 genotypes of Jatropha belonging to the germplasm bank in vivo located in the Department of Plant Production UNESP / FCAV, Jaboticabal, SP. The characters were: plant height, stem diameter, number of primary branches, fruit length, fruit width, fruit weight, fruit shape, seed length, seed width, seed weight and oil content. The data were submitted to principal component analysis (PCA), cluster analysis by the hierarchical method of Ward and non- hierarchical method of k-means. The character fruit shape was removed from multivariate analysis for being qualitative. In PCA, the 4 PCs explained 71.62% of the total variance. The greatest contribution of the characters to discriminate the genotypes in PC1 were seed weight, fruit width, fruit length and fruit weight. They highlighted 22 promising genotypes with potential to be exploited in breeding programs. The dendrogram obtained by Ward cluster analysis and the k-means analysis generated nine groups, 5 of which grouped genotypes with advantageous features. Multivariate analyzes allowed the characterization of the genotypes, allowing the selection for many traits and indicating gains strategies with selection for genetic improvement programs. For the construction of the calibration and validation curve for oil content of seeds were used 174 plants that had oil extracted by the method "Soxhlet" to obtain reference values which were between 50.27 to 63% 21% (m/m) of oil. The spectra were collected from intact seeds without husk by attachment rotation through near-infrared spectroscopy (NIR) in Tango machine model, Bruker. Each spectrum was taken with an average of 64 scans with a resolution of 16 cm-1. The spectral regions chosen for the calibration curve were 8992-7424 cm-1 and 5336-4928 cm-1. The correlation coefficients (R2) of the curve constructed were 73,23% for calibration and 70,79% for validation, the calibration mean square error (RMSEC) was 1,35 and cross validation mean square error (RMSECV) was 1,40 between the data obtained by the reference method and the NIR. The results indicated that the oil content of Jatropha curcas L. can be estimated by NIR, and the reference method was satisfactory, with good correlation with the spectral information, making the built curve reliable. However, future addition of new value to improve their robustness and amplitude is required.
The purpose of this study was to characterize the phenotypic diversity in Jatropha genotypes by multivariate analysis, and developing a calibration model for determination oil content by Near-infrared spectroscopy (NIR) in order to select genotypes with high oil content. For the phenotypic analysis, 11 agronomic characters were evaluated in 165 genotypes of Jatropha belonging to the germplasm bank in vivo located in the Department of Plant Production UNESP / FCAV, Jaboticabal, SP. The characters were: plant height, stem diameter, number of primary branches, fruit length, fruit width, fruit weight, fruit shape, seed length, seed width, seed weight and oil content. The data were submitted to principal component analysis (PCA), cluster analysis by the hierarchical method of Ward and non- hierarchical method of k-means. The character fruit shape was removed from multivariate analysis for being qualitative. In PCA, the 4 PCs explained 71.62% of the total variance. The greatest contribution of the characters to discriminate the genotypes in PC1 were seed weight, fruit width, fruit length and fruit weight. They highlighted 22 promising genotypes with potential to be exploited in breeding programs. The dendrogram obtained by Ward cluster analysis and the k-means analysis generated nine groups, 5 of which grouped genotypes with advantageous features. Multivariate analyzes allowed the characterization of the genotypes, allowing the selection for many traits and indicating gains strategies with selection for genetic improvement programs. For the construction of the calibration and validation curve for oil content of seeds were used 174 plants that had oil extracted by the method "Soxhlet" to obtain reference values which were between 50.27 to 63% 21% (m/m) of oil. The spectra were collected from intact seeds without husk by attachment rotation through near-infrared spectroscopy (NIR) in Tango machine model, Bruker. Each spectrum was taken with an average of 64 scans with a resolution of 16 cm-1. The spectral regions chosen for the calibration curve were 8992-7424 cm-1 and 5336-4928 cm-1. The correlation coefficients (R2) of the curve constructed were 73,23% for calibration and 70,79% for validation, the calibration mean square error (RMSEC) was 1,35 and cross validation mean square error (RMSECV) was 1,40 between the data obtained by the reference method and the NIR. The results indicated that the oil content of Jatropha curcas L. can be estimated by NIR, and the reference method was satisfactory, with good correlation with the spectral information, making the built curve reliable. However, future addition of new value to improve their robustness and amplitude is required.
Descrição
Palavras-chave
Análise de agrupamento, Análise de componentes principais, Jatropha curcas L., NIR, Cluster analysis, Principal components analysis