Metabolomics analysis identifies differential metabolites and potential diagnostic biomarkers among pediatric sepsis subtypes

Abstract

Background Sepsis in children can be caused by a variety of pathogens, with bacteria and viruses being the most common. This study used metabolomics to identify differences in metabolic profiles and potential biomarkers among pathogens causing pediatric sepsis. Methods Serum metabolomic profiles of pediatric bacterial and viral sepsis were obtained from the MetaboLights database (MTBLS563). Principal component analysis (PCA), partial least squares discriminant analysis (PLS-DA), and orthogonal PLS-DA were employed to explore metabolic distinctions. Differential expression metabolites (DEMs) were identified using the Wilcoxon rank-sum test and variable importance in projection (VIP) scores. Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment, receiver operating characteristic (ROC) analysis, Extreme Gradient Boosting (XGBoost) modeling, and Shapley Additive exPlanations (SHAP) analysis were conducted to determine diagnostic metabolites and evaluate model performance. Results PCA and PLS-DA revealed distinct metabolic profiles among bacterial pediatric sepsis (PBID_PS), viral pediatric sepsis (VID_PS), and healthy controls. Fourteen differential metabolites were identified, primarily enriched in nitrogen metabolism, arginine biosynthesis, and the metabolism of alanine, aspartate, and glutamate. Among them, choline, glutamate, and glutamine exhibited strong discriminatory ability between PBID_PS and VID_PS. XGBoost and SHAP analyses confirmed these metabolites as key diagnostic indicators, achieving excellent predictive performance and revealing distinct metabolic reprogramming underlying different etiologies of pediatric sepsis. Conclusion Metabolomic profiling revealed distinct metabolic signatures between bacterial and viral pediatric sepsis, with glutamate, glutamine, and choline serving as potential biomarkers.