Abstract
Respiratory infections rank fourth in the global economic burden of disease. Lower respiratory tract infections are the leading cause of death in low-income countries. The rapid identification of pathogens causing lower respiratory tract infections to help guide the use of antibiotics can reduce the mortality of patients with lower respiratory tract infections. Single-cell Raman spectroscopy is a "whole biological fingerprint" technique that can be used to identify microbial samples. It has the advantages of no marking and fast and non-destructive testing. In this study, single-cell Raman spectroscopy was used to collect spectral data of six respiratory tract pathogen isolates. The T-distributed stochastic neighbor embedding (t-SNE) isolation analysis algorithm was used to compare the differences between the six respiratory tract pathogens. The eXtreme Gradient Boosting (XGBoost) algorithm was used to establish a Raman phenotype database model. The classification accuracy of the isolated samples was 93-100%, and the classification accuracy of the clinical samples was more than 80%. Combined with heavy water labeling technology, the drug resistance of respiratory tract pathogens was determined. The study showed that single-cell Raman spectroscopy-D(2)O (SCRS-D(2)O) labeling could rapidly identify the drug resistance of respiratory tract pathogens within 2 h.