Desenvolvimento de metodologias analíticas verdes para detecção de adulterantes em leite em pó caprino

Abstract

Goat milk is considered a safe food, capable of producing metabolic and/or physiological effects that can be beneficial to health. Currently, the global consumption of goat milk and its derivatives has been increasing due to the fact that this food presents unique characteristics when compared to cow milk, such as low allergenicity, high digestibility, and functional properties. From an economic point of view, goat milk has a higher commercial value than cow milk, and several studies point out that adulteration by adding cow milk to goat milk is frequent. In this context, food quality control policies are important to ensure the safe consumption of food and prevent financial losses and harm to consumers' health. Therefore, this study aimed to develop new analytical methodologies based on NIR, RAMAN, and chemometric tools for the quality control of goat milk, in order to obtain fast, non-invasive, and non-destructive analyses, aligned with the principles of Green Chemistry. In the first experiment, NIR spectra were acquired from a total of 348 samples, aiming to classify pure goat milk when adulterants such as starch, soy extract, dairy compound, and cow milk were present. Considering the NIR data, baseline offset correction combined with Savitzky-Golay smoothing with 21 points yielded the best results for the models built (DD-SIMCA and OC-PLS). The authentication of powdered goat milk using these data resulted in sensitivity values for the training and test sets of 97.5% and 100%, respectively, despite the specificity in the test being 83.9% using DD-SIMCA. For the models using OC-PLS, the values obtained were 97.5%, 100%, and 95.1%, respectively, for the same performance metrics evaluated. On the other hand, the second experiment proceeded with the acquisition of RAMAN spectra from 83 samples, of which 67 were adulterated only with cow milk, also aiming to classify the matrix regarding adulteration. The RAMAN data, in which three different validation strategies were used, showed that the results for the DD-SIMCA models in terms of sensitivity and specificity were all 100%, except for the third approach with the KS algorithm, which showed a value of 98% for the test set. In all approaches, the best results were obtained using the Savitzky-Golay first derivative with a second-order polynomial and a window of 25 points, along with MSC. As a result, by authenticating powdered milk for the presence of fraudulent adulterants, producers can ensure that consumers are enjoying a safe and high-quality product. Thus, the proposed methodology can contribute to consumer safety and ensure regulatory compliance.