PREDICTION OF MIXTURE COMPOSITION BY CHROMATOGRAPHIC CHARACTERIZATION, MULTIVARIATE CLASSIFICATION AND PARTIAL LEAST-SQUARES REGRESSION, A COMPARISON OF METHODS

Elucidation of the mixing proportions of mixtures of natural triacylglycerol oils is of paramount importance in several areas, most notably in quality control in the pharmaceutical, cosmetic and food industries. The detection of adulteration of edible oils, i.e., where high-priced commodity oils are mixed with lower-priced substitutes, is also of primary importance. Three goals were set for this study; to construct a model that detects adulteration, that identifies the components in the mixture and also predicts the levels of the components present in the mixture. This was done by utilizing the fatty acid profiles and the triacylglycerol molecular species profile, analyzed by chromatographic methods, principal components analysis and partial least-squares regression. All three goals of this study were fulfilled.