Machine learning assisted Raman spectroscopy to discern the markers associated with colistin Resistance

Raman spectroscopy (RS) is rapidly becoming a key analytical tool for a wide range of microbiology applications. In combination with diverse machine learning methods, RS has demonstrated potentials to be translated in form of a culture-free, rapid, and objective tool for identifying antimicrobial resistance (AMR). Colistin is regarded as the final line of defense antibiotic for treating infections caused by gram-negative bacteria. In this study, we have employed a combinatorial approach of machine learning and RS to identify a novel spectral marker associated with phosphoethanolamine modification in lipid A moiety of colistin resistant gram-negative Escherichia coli. The visible spectral fingerprints of this marker have been validated by partial least square regression and discriminant analysis. The origin of the spectral feature has been confirmed by hyperspectral imaging and K-means clustering of a single bacterial cell. The chemical structure of the modified lipid A moiety has been verified by gold standard MALDI-TOF mass spectrometry. Our findings support futuristic applicability of this spectroscopic marker in objectively identifying colistin-sensitive and resistant strains.