Grape Must Profiling And Cultivar Discrimination Based On Amino Acid Composition And General Discriminant Analysis With Best Subset

The present study aimed to elucidate the amino acid profile of a number of grapevine cultivars relevant to the South African wine industry using 738 grape must samples obtained during the 2016 and 2017 harvests. Proline and arginine were found to be the most abundant amino acids, with an average of 697.69 mg/L for proline (range 33.22-3445.43 mg/L) and 388.35 mg/L for arginine (range 13.56-1616.56 mg/L) across all vintages, regions, and cultivars. At the other extreme, ornithine (2.01 mg/L), glycine (3.28 mg/L), methionine (3.64 mg/L) and lysine (3.91 mg/L) were found to have the lowest concentrations, both in terms of the overall average, as well as per cultivar. Furthermore, the data were used to demonstrate how characteristic the amino acid profile is of a particular group (red or white) or cultivar. Cultivars were predicted based on their average amino acid concentrations using general discriminant analysis (GDA) and the best subset principle. For white musts, Chardonnay showed the highest prediction accuracy (100%), and Pinotage (75%) for red cultivars. Overall, the white cultivars included in this study were more accurately distinguished from one another (75.6%) compared to the red (60.1%). This predictive ability was subsequently compared to the accuracy of predicting cultivars based on only the arginine and proline concentrationsas well as the ratio between the two. The use of only these amino acids as well as the addition of the proline/arginine ratio as a predictor variable did not offer satisfactory discriminatory power between either white or red cultivars