Landscape of Metabolic Fingerprinting for Diagnosis and Risk Stratification of Sepsis

BACKGROUND: Sepsis and septic shock, a subset of sepsis with higher risk stratification, are hallmarked by high mortality rates and necessitated early and accurate biomarkers. METHODS: Untargeted metabolomic analysis was performed to compare the metabolic features between the sepsis and control systemic inflammatory response syndrome (SIRS) groups in discovery cohort, and potential metabolic biomarkers were selected and quantified using multiple reaction monitoring based target metabolite detection method. RESULTS: Differentially expressed metabolites including 46 metabolites in positive electrospray ionization (ESI) ion mode, 22 metabolites in negative ESI ion mode, and 4 metabolites with dual mode between sepsis and SIRS were identified and revealed. Metabolites 5-Oxoproline, L-Kynurenine and Leukotriene D4 were selected based on least absolute shrinkage and selection operator regularization logistic regression and differential expressed between sepsis and septic shock group in the training and test cohorts. Respective risk scores for sepsis and septic shock based on a 3-metabolite fingerprint classifier were established to distinguish sepsis from SIRS, septic shock from sepsis. Significant relationship between developed sepsis risk scores, septic shock risk scores and Sequential (sepsis-related) Organ Failure Assessment (SOFA), procalcitonin (PCT) and lactic acid were observed. CONCLUSIONS: Collectively, our findings demonstrated that the characteristics of plasma metabolites not only manifest phenotypic variation in sepsis onset and risk stratification of sepsis but also enable individualized treatment and improve current therapeutic strategies.