A rapid dicrimination of wheat, walnut and hazelnut flour samples using chemometric algorithms on GC/MS data

Abstract

© 2019, Springer Science+Business Media, LLC, part of Springer Nature. There is a worldwide growing trend in developing methods for determining authenticity and detecting adulteration in food products. In this study an approach utilizing gas chromatography—mass spectrometry (GS/MS) combined with chemometric multivariate data analysis was proposed in order to determine discrimination and classification possibilities of flour samples produced from 16 genotypes of wheat, 9 genotypes of hazelnut, and 8 genotypes of walnut, grown in the Vojvodina region, Republic of Serbia. Plant samples were milled into flour, lipid fraction was extracted with n-hexane and derivatized using a 0.2 M TMSH solution and analyzed on a GC/MS device. Molecular ions of eluting lipid components were selected, isolated from total ion current chromatograms and their peaks employed in further data processing. Unsupervised exploratory data analysis techniques: principal component analysis (PCA), expression heat mapping, hierarchical cluster analysis (HCA) and principal coordinate analysis (PCoA) were used to select the most important variables, and to explore their potential in discrimination of investigated flour samples according to belonging botanical origin. PCA and heat maps demonstrated that molecular ions of methyl esters of four fatty acids: 270 (palmitic), 294 (linoleic), 296 (oleic), and 298 (stearic), were most discriminative variables, HCA and PCoA showed a clear and strong separations between groups of analyzed samples. A support vector machine (SVM) algorithm was employed in order to classify samples in three groups. The performance of the classification SVM model was excellent, achieving high coefficient of determination of 98.6, with only 1 value being misclassified.