From phylogeny to photonics: A smart spectroscopic approach assisted with neural networks for rapid typing of Enterobacteriaceae isolates from chicken meat

Abstract

Bacteria are ubiquitous organisms; while many are harmless, some can cause serious disease outbreaks. Nearly a quarter of the global food supply is ruined by microbial activity, such as spoilage microorganisms or disease-causing pathogens, leading to substantial economic losses and human safety problems. Most mishaps associated with bacterial contamination could be avoided if a 100 % accurate, easy-to-use, rapid detector/identifier of bacteria was available. Compared to the time-consuming and complicated conventional molecular characterization methods, laser-induced breakdown spectroscopy (LIBS) combined with appropriate chemometric analysis provides a potential approach for the rapid and reliable characterization of bacterial species. This study identified isolated bacterial contaminants from store-bought chicken fillets using molecular characterization of the polymerase chain reaction (PCR) products of small subunit 16S rRNA sequences. Eight different strains were identified, all belonging to the family Enterobacteriaceae. Phylogenetic relationships were investigated using a constructed phylogenetic tree. In this study, nanoparticle-enhanced laser-induced breakdown spectroscopy (NELIBS) was used to differentiate bacterial samples based on their elemental composition. Spectral data were analyzed using principal component analysis (PCA) to explore the variance between the pure bacterial samples. With a cumulative variance between 95 % and 97 %, sample groups at closely related taxonomic levels (genus, species, and subspecies) could be discriminated. The robustness of the LIBS characterization was tested in the presence of a second bacterial species in the ablated specimen using mixtures of bacterial cultures at different concentrations. Using artificial neural networks (ANN), the spectral data were classified with an accuracy of up to 100 %.