Fluorescent Machine Learning Aided Classification of Pathogenic Bacteria Using the Excitation Emission Matrix

Abstract

The rapid identification of pathogenic bacterial strains is becoming a challenging task as it causes many hospital associated infections. Many have reported on the use of fluorescence spectroscopy as an alternative to characterize bacteria. As bacteria possess intrinsic fluorophores, attempts were made to characterize eight strains using excitation emission matrix (EEM) measurements. From the results of parallel factor analysis (PARAFAC), four fluorophores, tryptophan, anthranilic acid, nicotinamide adenine dinucleotide, and flavin adenine dinucleotide, were identified with varying distributions. The data obtained from PARAFAC analysis were subjected to hierarchical cluster analysis (HCA) and linear discriminant analysis (LDA). This study demonstrates the potential of EEM technique to classify bacteria with 100% accuracy.