Classificação de genótipos de trigo usando espectroscopia de infravermelho próximo e quimiometria

Abstract

Wheat, Triticum aestivum L., is one of the most important cereal crops, accounting for about 30% of world production. The interactions between wheat genotype and environmental conditions define the quality of the grain and consequently its industrial use. Each type of industrial application requires wheat flour with specific physicochemical and rheological characteristics. Near infrared spectroscopy (NIRS) has among its main applications numerous uses for the analysis of wheat quality. One of the advantages of NIRS is the low time of analysis and simple sample preparation, the spectra obtained contains chemical information that may be related to the properties of wheat flour. Chemometrics methods aid in the interpretation of instrumental analytical data, such as spectral data. Thus, the overall objective of the project is to employ near infrared spectroscopy combined with chemometrics to discriminate wheat genotypes. A total of 180 samples (8 genotypes, 17 crop cities and 2 harvests) were analyzed by the IAPAR-Londrina Institute of Agronomic Institute’s Winter Cereals program. NIR spectra were collected from wheat flours extracted in an experimental mill at the Laboratory of Plant Physiology of IAPAR. The spectra were pretreated and chemometrics methods (PCA and PLS-DA) were used to classify the samples. The use of the NIRS combined with the PCA exploratory analysis, and by the HCA and k-means cluster analyzes indicate that the groups are formed by the genotypes; the place of cultivation and the crop do not present an important effect in the formation of these groups. Through the PCA analysis, a separation of the eight wheat genotypes in three groups was observed. These genotypes separated mainly in the characteristic bands of moisture, protein and ashes, presenting higher humidity and lower content of ashes and proteins than the other genotypes. The hardness of the wheat genotypes was an important factor for the formation of these groups, where the soft genotypes severely separated from those of hard texture, semi hard and very hard. The PLS-DA model using the NIR spectra also assertively classified the samples with mean values of 95.83% for sensitivity and 99.53% for selectivity. Through this model it was also possible to notice a clear separation of classes LD121102, LD132210 and LD131102. Thus, it was possible to classify the wheat samples through the NIR spectra in tandem with the chemometrics methods PCA, HCA and PLS-DA.