This study presents a tentative analysis of the lipid composition of 47 legume samples, encompassing species such as Phaseolus spp., Vicia spp., Pisum spp., and Lathyrus spp. Lipid extraction and GC/MS (gas chromatography with mass spectrometric detection) analysis were conducted, followed by multivariate statistical methods for data interpretation. Hierarchical Cluster Analysis (HCA) revealed two major clusters, distinguishing beans and snap beans (Phaseolus spp.) from faba beans (Vicia faba), peas (Pisum sativum), and grass peas (Lathyrus sativus). Principal Component Analysis (PCA) yielded 2D and 3D score plots, effectively discriminating legume species. Linear Discriminant Analysis (LDA) achieved a 100% accurate classification of the training set and a 90% accuracy of the test set. The lipid-based fingerprinting elucidated compounds crucial for discrimination. Both PCA and LDA biplots highlighted squalene and fatty acid methyl esters (FAMEs) of 9,12,15-octadecatrienoic acid (C18:3) and 5,11,14,17-eicosatetraenoic acid (C20:4) as influential in the clustering of beans and snap beans. Unique compounds, including 13-docosenoic acid (C22:1) and γ-tocopherol, O-methyl-, characterized grass pea samples. Faba bean samples were discriminated by FAMEs of heneicosanoic acid (C21:0) and oxiraneoctanoic acid, 3-octyl- (C18-ox). However, C18-ox was also found in pea samples, but in significantly lower amounts. This research demonstrates the efficacy of lipid analysis coupled with multivariate statistics for accurate differentiation and classification of legumes, according to their botanical origins.
Keywords: GC/MS analysis; food authentication; legumes; lipid profiles; multivariate statistics.