Fast discrimination of bacteria using a filter paper-based SERS platform and PLS-DA with uncertainty estimation.

Abstract

Rapid and reliable identification of bacteria is an important issue in food, medical, forensic, and environmental sciences; however, conventional procedures are time-consuming and often require extensive financial and human resources. Herein, we present a label-free method for bacterial discrimination using surface-enhanced Raman spectroscopy (SERS) and partial least squares discriminant analysis (PLS-DA). Filter paper decorated with gold nanoparticles was fabricated by the dip-coating method and it was utilized as a flexible and highly efficient SERS substrate. Suspensions of bacterial samples from three genera and six species were directly deposited on the filter paper-based SERS substrates before measurements. PLS-DA was successfully employed as a multivariate supervised model to classify and identify bacteria with efficiency, sensitivity, and specificity rates of 100% for all test samples. Variable importance in projection was associated with the presence/absence of some purine metabolites, whereas confidence intervals for each sample in the PLS-DA model were calculated using a resampling bootstrap procedure. Additionally, a potential new species of bacteria was analyzed by the proposed method and the result was in agreement with that obtained via 16S rRNA gene sequence analysis, thereby indicating that the SERS/PLS-DA approach has the potential to be a valuable tool for the discovery of novel bacteria. Graphical abstract This paper describes the discrimination of bacteria at the genus and species levels, after minimal sample preparation, using paper-based SERS substrates and PLS-DA with uncertainty estimation.