SPECTROSCOPIC METHODS AND MULTIVARIATE CLASSIFICATION APPLIED IN THE DIFFERENTIATION OF PATHOGENIC MICRO-ORGANISMS
FT-IR, Fluorescence, Multivariate Analysis, Cryptococcus, Klebsiella sp. and Escherichia coli.
This study demonstrates the development of multivariate classification methods, allied to spectroscopic techniques, such as spectroscopy in the medium infrared region and molecular fluorescence, in the detection of pathogenic microorganisms: fungi and bacteria. The first studies looked for the differentiation of Cryptococcus neoformans and Cryptococcus gatti. These fungi are the etiological agents of cryptococcosis, whose adequate treatment depends on the rapid and correct detection and differentiation of the species. This determination is currently made by classical and molecular techniques and is mostly laborious and expensive. As an alternative method to discriminate C. gattii and C. neoformans, the attenuated total reflectance average infrared spectroscopy, combined with multivariate classification techniques (PCA-LDA / QDA, GA-LDA / QDA, SPA-LDA / QDA ), in which the GA-QDA model obtained sensitivity in the classes C. neoformans and C. gatti of 84.4% and 89.3%, respectively, using only 17 wave numbers. Then, using fluorescence emission and excitation matrix (EEM) spectroscopy, combined with multivariate classification methods (UPCA-LDA / QDA, UGA-LDA / QDA, USPA-LDA / QDA, PARAFAC / PLS-DA, nPLS - DA). The most satisfactory model was UGA-LDA, which used only 5 wavelengths, and presented a sensitivity of 88.9% in calibration and 100.0% of prediction for both species, results that are comparable to routine biological tests. The last study, aimed at the differentiation of sensitive and multiresistant bacteria of the genus Klebsiella sp. and Escherichia coli. Through molecular fluorescence spectroscopy and multivariate classification methods: 2D-LDA, 2D-PCA-LDA, 2D-PCA-QDA and 2D-PCA-SVM. Among these, the models that had the best performance for both bacterial genera were the 2D-PCA-LDA and 2D-PCA-SVM with classification accuracy rates above 93%. Compared to classical methods, the methodologies proposed in these studies demonstrate an innovative, faster and cheaper alternative for the identification of pathogenic microorganisms, such as fungi and bacteria, opening up the possibility of being applied in routine diagnostic laboratories.