Machine learning analysis of magnetic covalent organic framework based heterostructures extracted intracellular metabolic fingerprint for direct hypervirulent Klebsiella pneumoniae prediction.
Journal:
Talanta
PMID:
40158311
Abstract
Hypervirulent Klebsiella pneumoniae (hvKP), known for its high virulence and epidemic potential, has emerged as a significant global public health threat. Therefore, improving the identification of hvKP and enabling earlier and faster detection in the community to support subsequent effective treatment and prevention of hvKP is an urgent issue. In this study, we introduce a new approach utilizing magnetic covalent organic framework based heterostructures (denoted FeO@COF@Au) for the analysis of intracellular metabolites from bacterial cells, facilitating the rapid diagnosis of hvKP. Importantly, intracellular metabolites were extracted from bacterial cells using cold methanol to preserve their abundance and stability, and their metabolite fingerprints were rapidly obtained by FeO@COF@Au. Using this method, we effectively extracted intracellular metabolic fingerprints from 136 clinical K. pneumoniae isolates collected from patients. Machine learning analysis of these fingerprint variations successfully distinguished hypervirulent K. pneumoniae from classical strains (cKP), achieving an area under the curve (AUC) of 1.00 in both the training and testing sets based on 359 m/z features. This strategy shows great potential for the rapid diagnosis of hvKP and could significantly improve its management.