Identification of protein biomarkers to differentiate between gram-negative and gram-positive infections in adults suspected of sepsis

Abstract

Background: Sepsis is mostly caused by bacterial infections and requires a prompt diagnosis. There is a need for improved diagnostics by differentiating between gram-negative and gram-positive bacterial infections.

Methods: The plasma levels of 285 unique proteins in patients with gram-negative infection (n = 154), gram-positive infection (n = 92), and in healthy controls (n = 35) were quantified using proximity extension assay. Three machine learning algorithms; random forest, recursive feature elimination, and adaptive least absolute shrinkage and selection operator (Lasso) were employed to identify discriminative proteins, with their effectiveness assessed using accuracy metrics. The selected proteins were further evaluated for their ability to differentiate between gram-negative and gram-positive infections through logistic regression and area under the receiver operating characteristic curve.

Results: We identified 55 discriminative proteins differentiating between gram-negative and gram-positive infections using the Lasso, the best performing algorithm. The discriminative proteins achieved AUROC values of 0.69 for gram-negative infections and 0.66 for gram-positive infections, both compared to the remaining groups, and 0.58 for differentiating between the two infection groups. Comparative statistical analysis revealed no significant differences in protein expression between gram-negative and gram-positive patients.

Conclusions: We identified 55 proteins with some discriminative potential between gram-negative and gram-positive infections. However, the overall predictive performance was low and did not exceed that of established single biomarkers. These findings highlight the challenges of applying a multimarker approach in infection classification and emphasize the need for further studies using larger and more diverse cohorts, as well as broader analytical methods, to explore their potential clinical utility.

Clinical trial: Not applicable.

Supplementary Information: The online version contains supplementary material available at 10.1186/s12879-025-11973-5.

Keywords: Biomarkers; Diagnostics; Gram-negative bacteria; Gram-positive bacteria; Machine learning; Proteomics; Sepsis.

Fig. 1

Workflow of processing data. Preprocessing (green boxes), selection, and evaluation of model algorithms (blue boxes). RF, random forest; Lasso, least absolute shrinkage and selection operator; RFE-LR, recursive feature elimination-logistic regression

Fig. 2

Evaluation of model performance. Accuracies of the three models were determined using a hold-out test on datasets with missing data values below LOD ranging from < 5% to < 80%. RFE-LR (white bar), RF (light grey bar) and Lasso (dark grey bar)

Fig. 3

Distribution of samples from patients using 285 proteins. A) Principal component analysis (PCA) plot B) t-distributed stochastic neighbor embedding (tSNE) plot. Yellow dots: samples from patients with gram-negative infection, blue dots: samples from patients with gram-positive infection, pink dots: samples from healthy control individuals

Fig. 4

The predictive power of the 55 discriminative proteins in classifying the groups of participants. Blue dashed line: patients with gram-negative infection, yellow dashed line: patients with gram-positive infection, green dashed line: healthy control individuals, red line: direct comparison between patients with gram-positive and gram-negative infections, excluding healthy controls

Fig. 5

Distribution of samples of patients using the 55 discriminative proteins. A) Principal component analysis (PCA) plot B) t-distributed stochastic neighbor embedding (tSNE) plot. Yellow dots: samples from patients with gram-negative infection, blue dots: samples from patients with gram-positive infection, pink dots: healthy control individuals

Fig. 6

Detailed analysis of selected discriminative proteins. Skewness (A) and kurtosis (B) of 55 discriminative proteins selected by Lasso (patients with gram-positive infection; blue line, patients with gram-negative infection; red line). (C) receiver operating characteristic analysis showing specificity and sensitivity of the set of discriminative proteins after excluding skewed proteins. (D) violin plot of five skewed discriminative proteins in patients with gram-negative and gram-positive infection, as assessed by a student t-test with Benjamini-Hochberg adjustment. MFAP5; microfibrillar-associated protein 5, TNF; tumor necrosis factor, TNFRSF13B; TNF receptor superfamily member 13B, ADA; Adenosine Deaminase, CD8A; T-Cell surface glycoprotein CD8 alpha chain