Practical approaches in detection of food fraud with multivariate calibration

Background and objective: In the realm of analytical chemistry, multivariate calibration involves creating mathematical models that connect diverse instrumental signals with analyte concentrations. This approach provides a mean to quantitatively analyze complex mixtures, particularly in multicomponent systems. To address food adulteration concerns, this paper explores the application of Raman spectroscopy and Partial Least Squares Regression (PLSR) using the MVC1 software. The main objective is to demonstrate the software's efficiency in quantifying the adulteration of hazelnut oil in extra virgin olive oil (EVOO).
Materials and methods: The analysis leverages the MVC1 software, a valuable tool for multivariate linear and nonlinear calibrations. One-leave-out cross-validation and the Durbin-Watson statistical test are employed to determine the optimal number of PLS factors and identify outliers. Statistical parameters including RMSEP, %REP, R², and explained variance are used to evaluate the calibration model's performance. Key figures of merit including sensitivity, analytical sensitivity, LOD, and LOQ, are computed to assess the analytical technique's precision and reliability.
Results and conclusion: The study effectively quantifies the percentage of adulteration in EVOO by hazelnut oil, a pressing concern in food authenticity and safety. The results demonstrate the MVC1 software's capability in establishing reliable calibration models. By achieving a balance between sensitivity and analytical sensitivity, the model accurately predicts analyte concentrations. It also sets robust detection and quantitation limits, ensuring precise analysis. This research showcases the practical application of advanced analytical techniques and software tools to address real-world problems, contributing to the authenticity and purity of food products in the market.