# Research article proteomics-based plasma biomarkers for predicting CRKP infection in ICU sepsis patients

## TL;DR

This study identifies plasma proteins that can help quickly distinguish between two types of K. pneumoniae infections in ICU patients, potentially improving treatment decisions.

## Contribution

The paper introduces a novel proteomics-based diagnostic model using plasma biomarkers to differentiate CRKP and CSKP infections rapidly.

## Key findings

- 28 differentially expressed proteins were identified, with 16 upregulated and 12 downregulated.
- A diagnostic model using 10 proteins, including PLXNB1 and S100A1, achieved high accuracy (AUC > 0.90) in distinguishing CRKP and CSKP infections.
- A simplified two-protein model also showed excellent performance and good calibration in both training and testing cohorts.

## Abstract

Early differentiation between carbapenem-resistant Klebsiella pneumoniae (CRKP) and carbapenem-sensitive K. pneumoniae (CSKP) infections is critical due to limited treatment options and high mortality associated with CRKP. Current diagnostic methods are slow and insufficient for timely clinical decision-making, especially in ICU settings. Identifying reliable biomarkers for rapid discrimination is urgently needed.

We performed plasma proteomic profiling of ICU sepsis patients infected with CRKP or CSKP using data-independent acquisition (DIA) mass spectrometry. Significantly differentially expressed proteins (DEPs) underwent Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG), and Disease Ontology (DO) functional annotation and enrichment analyses. Hub proteins were identified through protein-protein interaction network analysis. Protein biomarkers for constructing a diagnostic model by logistic regression analysis were further selected by XGboost and Lasso. The model was then evaluated for discrimination, calibration, and clinical utility by area under curve (AUC), the Hosmer–Lemeshow goodness-of-fit test and calibration curve, and decision curve, respectively.

A total of 1,432 proteins and 13,482 peptides were identified in the plasma samples. Among these, 28 DEPs were detected, including 16 upregulated and 12 downregulated proteins. Functional enrichment analysis indicated that these DEPs were primarily associated with neural and cardiovascular pathways. Using a combination of XGBoost and LASSO algorithms, 10 protein biomarkers were selected to construct a diagnostic model. The proteins of the optimal diagnostic model included PLXNB1 and S100A1. Notably, a simplified two-protein model demonstrated excellent diagnostic accuracy with an AUC exceeding 0.90 in both training and testing cohorts. The Hosmer–Lemeshow goodness-of-fit test yielded p-values of 0.825 and 0.295 in the training and testing sets, respectively, indicating good model calibration.

PLXNB1 and S100A1 serve as promising plasma biomarkers for early, non-culture-based differentiation of CRKP and CSKP infections. Their integration into clinical workflows could improve rapid diagnosis and guide targeted therapy in critically ill sepsis patients.