Methodology of chemometric modeling of spectrometric signals in the analysis of complex samples

A generalized algorithm of the multivariate simulation of spectrometric data is considered for solving typical analytical problems, like the determination of the concentration of a particular analyte and the assignment of a sample to one of predefined classes. In particular, we considered preliminary data processing, exploratory analysis, optimization of a chemometric model, calculation of performance characteristics, transfer of the model to other spectrometers, and automation of chemometric processing for the routine analysis of samples. To illustrate the potential of the method, we selected a system of bovine and porcine heparin, mixtures of soy and sunflower lecithin, and a set of red and white wine samples as test samples. Partial least squares and discriminant analysis were used as chemometric methods. We used proton nuclear magnetic resonance (1H NMR) to record signals. Using the MATLAB environment, chemometric programs were developed for automated data processing in the context of problems under consideration and for the transfer of multivariate models to other spectrometers. Based on the results obtained, a methodology is proposed for the multivariate analysis of spectrometric data, which can be used in the analysis of various types of matrices and spectrometric signals.