Comparison of multivariate classification algorithms using EEM fluorescence data to distinguish Cryptococcus neoformans and Cryptococcus gattii pathogenic fungi

Abstract

Cryptococcus neoformans and Cryptococcus gattii are the etiologic agents of cryptococcosis, whose suitable treatment depends on rapid and correct detection and differentiation of the Cryptococcus species. Currently, this identification is made by classical and molecular techniques; however most of them are considered laborious and expensive. As an alternative method to discriminate C. gattii and C. neoformans, excitation-emission matrix (EEM) fluorescence spectroscopy combined with multivariate classification methods, Unfolded Partial Least Squares Discriminant Analysis (UPLS-DA), multiway-Partial Least Squares Discriminant Analysis (nPLS-DA), Parallel Factor Analysis (PARAFAC), Principal Component Analysis (PCA), Successive Projection Algorithm (SPA) and Genetic Algorithm (GA), followed by Linear Discriminant Analysis (LDA) and Quadratic Discriminant Analysis (QDA) was herein investigated. This technique showed to be an innovative and low cost methodology which requires a small sample volume. Among the methods, the most successful model was UGA-LDA, which showed a sensitivity of 88.9% within only 5 selected wavelengths in calibration and 100.0% prediction for both classes of C. neoformans and C. gattii, equaling or surpassing some of the biological tests that are usually carried out to differentiate these fungi.