Higher-order data analysis to leverage the performance of food quality control procedures

GOICOECHEA, HÉCTOR C

Congreso; Conferentia Chemometrica 2023; 2023

The potential demonstrated using chemometrics in analytical chemistry has been escorted by a tireless pursuit of new advantages and benefits of multidimensional data analysis. Recent investigations in multi-way data analysis have revealed that second- and higher-order models can profitably exploit the second-order advantage, albeit additional benefits in third- and higher-order models are still under discussion. However, more theoretical and practical investigations should be conducted to clarify the basic theory of multi-way methods to explore the essence of higher-order methodologies. [1,2].Fluorescence excitation-emission matrix (EEM) spectroscopy coupled with multi-way analysis is a powerful tool for the analysis of fluorophore mixtures or complex systems because of its straightforwardness, selectivity, and high sensitivity. When combined with an extra instrumental or experimental mode, it can render appealing outcomes in terms of analytical performance. This work is devoted to demonstrating that third-order data analysis can be conveniently utilized in pursuing leveraging the analytical performance of the methods, in the field of food quality control. First, to demonstrate the enhancement of the analytical performance of a method with quantitative aims, a four-way multivariate calibration method was developed for the simultaneous determination of 5 pesticides (thiabendazole, carbendazim, pirimiphos-methyl, imidacloprid and clothianidin) in citrus. Third-order data were acquired by registering the photo-induced fluorescence of the analytes as EEM at different times of UV irradiation in organized media.On the other hand, the effect of increasing the number of instrumental modes for classification analysis was also evaluated. Here, aiming at discriminating virgin olive oils from extra virgin olive oils, third-order data analysis was accomplished on data generated by the EEM monitoring of the thermal degradation of olive oil. In both cases, different data arrays (first- (in classification), second- and third-order data and data fusion) were built and subjected to several chemometric models to evaluate the properties and advantages of each data structure. Accordingly, PARAFAC, MCR-ALS and PLS-based modelling were used in each case. Quantitative results were evaluated through the predictive performance and detection capabilities, whereas classification results were evaluated through global indices, such as average sensitivity, non-error rate, and average precision. The results revealed different degrees of improvement by the inclusion of an additional mode to the data structure. The obtained results shed light on the fact that the use of higher-order data is an attractive approach to be explored in the classification field, particularly, in the study of samples with very similar spectral profiles, for which no evident classification patterns are observed.