Detection of port wine imitations by discriminant analysis using free amino acids profiles

The free amino acid content in wines is closely related to quality, authent icity, and the technology of winemaking. Some attempts have been made in th e past to find an unequivocal relationship between free amino acids and the authenticity of wines, but the complexity and diversity of the factors inv olved has made this matter a controversial one. This study investigates the possibility of using the analysis of 18 free amino acid revels to detect i mitations of Port Wine by linear discriminant analysis. The 105 samples ana lyzed were divided into two data sets: the training set, used to build the model (55 Port Wines and 13 Port imitations) and the test set of unknown sa mples (37 wines). Fifty-four of the 55 Part Wines and 12 from 13 Port Imita tions were correctly classified in the training set; the discriminant funct ion derived enabled the correct classification of 97% of the wines from the test set. Considerable discriminating power was obtained from only five Va riables - alanine, arginine, tyrosine, valine, and leucine (error 2.3% for the test set) - after data reduction by step-wise discriminant analysis. It is concluded that free amino acid content may be a good indicator in discr iminating between Port Wine and Port Imitations.