WUILLOUD Rodolfo German
congresos y reuniones científicas
Tracing the geographical origin of mendoza (argentina) grape seeds by icp-ms multielemental analysis and advanced chemometrics techniques
CANIZO, BRENDA; ESCUDERO, LETICIA B.; PEREZ, MARIA B.; PELLERANO, ROBERTO G.; WUILLOUD, RODOLFO G.
Simposio; 14th RIO SYMPOSIUM ON ATOMIC SPECTROMETRY; 2017
Universidade do Espírito Santo (UFES)
The identification of the geographical origin of wine grapes is of great interest in today's globalized trade, because it is directly related with wine provenance. Regional classification using mineral profile with chemometrics has been worldwide explored for wines; however, the classification of grapes according to their geographical origin has not been extensively studied. As trace elements in grapes are mainly located in their seeds , the aim of this work was to investigate the potential use of several multivariate statistics tools combined with multielemental data obtained by ICP-MS analysis of grape seeds as an original source for the geographical differentiation of grapes cultivated in nearby locations. Grapes were collected from vineyards located in 5 winemaking regions of Mendoza. Seeds were separated from the rest of the whole bunches, washed with Milli-Q water, lyophilized, pulverized, and finally acid digested. All samples were characterized by 29 descriptors: Ag, As, Ce, Co, Cs, Cu, Eu, Fe, Ga, Gd, La, Lu, Mn, Mo, Nb, Nd, Ni, Pr, Rb, Sm, Te, Ti, Tl, Tm, U, V, Y, Zn, and Zr. Basic chemometrical characterization was made by PCA. The results showed that only the samples from one region could be clearly differentiated, but other samples with different geographical origin could not be solved by this method. This suggested the need of using more complex chemometric analysis. Five chemometric models were selected and tested, i.e. LDA, PLS-DA, k-NN, SVM, and RF, and displayed different degrees of success in the prediction of test samples. The order of successful identification rates was as follows: RF>SVM >k-NN> LDA>PLS-DA. The use of nonlinear methods resolved the classification problem, being RF the ideal model for discriminating the grape seed samples according to their geographical origins, with an overall classification accuracy of 98.3%. The first intraregional classification of grape seeds from Mendoza province, Argentina, was achieved in this work. RF method is proved to be a promising tool in classification analysis and quality control of raw materials used in wine-making industry.References:1 Rogiers, S. Y., Greer, D. H., Hatfield, J. M., Orchard, B. A., & Keller, M. (2006). Mineral sinks within ripening grape berries (Vitis vinifera L.). Vitis - Journal of Grapevine Research, 45(3), 115?123.2 Mironeasa, S., Leahu, A., & Codin, G. (2010). Grape Seed : physico-chemical , structural characteristics and oil content, 16(1), 1?6.