Rapid and accurate identification of marine microbes with single-cell Raman spectroscopy.

Rapid and accurate identification of individual microorganisms, such as pathogenic or unculturable microbes, is significant in microbiology. In this work, rapid identification of marine microorganisms by single-cell Raman spectroscopy (scRS) using one-dimensional convolutional neural networks (1DCNN) was explored. Here, single-cell Raman spectra of ten species of marine actinomycetes, two species of non-marine actinomycetes and E. coli (as a reference) were individually collected. Several common classification algorithms in chemometrics, including linear discriminant analysis with principal component analysis and a support vector machine, were applied to evaluate the 1DCNN performance based on the raw and pre-processed Raman spectra. 1DCNN showed superior performance on the raw data in terms of its accuracy and recall rate compared with other classification algorithms. Our investigation demonstrated that the scRS-integrating advanced 1DCNN classification algorithm provided a rapid and accurate approach for identifying individual microorganisms without time-consuming cell culture and sophisticated or specific techniques, which could be a useful methodology for discriminating the microbes that cannot be cultured under normal conditions, especially for 'biological risk'-related emergencies.