A chemometric approach to the detection of milk adulteration based on protein profiles determined by capillary electrophoresis

The general objective of this study was to utilize chemometrics in the interpretation of capillary electrophoresis milk protein profiles, for the detection of pasteurized milk adulteration with rehydrated milk powder or a rehydrated dairy-based milk substitute. The specific objectives were 1) to collect quantitative data on major casein and whey proteins in authentic and adulterated milks in a single CE analysis; and 2) to apply a pattern recognition procedure, Soft Independent Modeling of Class Analogies (SIMCA), on collected CE protein data, for the development of a statistical model useful in the detection of pasteurized milk adulteration. Authentic samples were fresh milk collected from various farms over a period of six months. Adulterated samples were authentic fresh milk partially or totally substituted with rehydrated milk powder or a rehydrated commercial milk substitute at different levels. Quantitative protein data obtained by capillary free zone electrophoresis for beta-lactoglobulin, alpha-lactalbumin, beta-casein, and alpha-casein of 86 samples, authentic and adulterated samples, were used as a training set to build a SIMCA multivariate statistical model. The detection of sample outliers was useful for the elimination of unusual samples and optimization of the multivariate model. From the 35 commercial pasteurized milks tested, which were treated as unknowns, a total of 14 samples (40%) were not assigned to the authentic or fresh milk group, meaning that these samples had some type of adulteration at the levels included in the training set (> 15%). Decision-making on detecting adulteration of unknown commercial pasteurized milk samples was eased since predictions were based on statistical probabilities.