Machine Learning and DIA Proteomics Reveal New Insights into Carbapenem Resistance Mechanisms in Klebsiella pneumoniae

The emergence of Carbapenem-resistant Klebsiella pneumoniae (CRKP) represents a major public health concern, primarily driven by its ability to evade a wide range of antibiotics. Despite extensive genomic studies, proteomic insights into antibiotic resistance mechanisms remain scarce. Here, we employed a Data-Independent Acquisition (DIA)-based quantitative proteomics approach to investigate proteomic differences between 78 CRKP and 18 Carbapenem-sensitive K. pneumoniae (CSKP) clinical isolates. A total of 3380 proteins were identified, with 946 showing significant differential expression. CRKP isolates exhibited increased expression of efflux pumps, beta-lactamases, and transcriptional regulators, while proteins associated with transport were enriched in CSKP isolates. To validate our findings, a quantitative proteomics analysis in an independent cohort of 10 CRKP and 11 CSKP isolates was performed. The key biomarkers identified via machine learning in the discovery cohort, including aldehyde dehydrogenase (KPN_03361), acyltransferase (KPN_02072), uncharacterized protein (YjeJ), plasmid partition protein B (ParaB), HTH-type transcriptional activator (RhaR), and beta-lactamase (Bla), were evaluated. They collectively achieved AUC > 0.7 in the validation cohort, confirming their discriminatory capacity as diagnostic markers. These findings provide novel insights into the molecular mechanisms of antibiotic resistance and identify promising biomarkers for diagnosing carbapenem-resistant K. pneumoniae, offering potential avenues for therapeutic intervention.