Photonic Data Science

Classification of Raman spectra recorded from single cells is commonly applied to bacteria that exhibit small sizes of approximately 1 to 2 μm. Here, we study the possibility to adopt this classification approach to filamentous bacteria of the genus Streptomyces. The hyphae can reach extensive lengths of up to 35 μm, which can correspond to a single cell identified in light microscopy. The classification of Raman bulk spectra will be demonstrated. Here, ultraviolet resonance Raman (UV RR) spectroscopy is chosen to classify six Streptomyces species by the application of a tree-like classifier. For each knot of the hierarchical classifier, estimated classification accuracies of over 94% are accomplished. In contrast to the classification of bulk spectra, the classification of single-cell spectra requires a homogenous substance distribution within the cell. Consequently, the bacterial cell chemistry can be represented by one individual spectrum. This requirement is not fulfilled when different spectra are processed from different locations within the cell. Bacteria of the investigated genus Streptomyces exhibit, besides the normal bacterial spectra, lipid-rich spectra. The occurrence of lipid enrichment depends on culture age and nutrition availability. With this study, we investigate the cell substance distribution, especially of lipid-rich fractions. The classification utilizing a tree-like classifier is also applied to the Streptomyces single-cell spectra, resulting in classification accuracies between 80 and 93% for the investigated Streptomyces species.