Plasma proteomics identifies molecular subtypes in sepsis.

Summary

In a prospective multi-center cohort (n=333), LC–MS/MS plasma proteomics identified four sepsis subtypes with distinct clinical severity and immune features. One cluster had 100% mortality, while others differed in adaptive vs acute inflammatory signatures and immunoglobulin levels. A machine-learning classifier using 10 proteins plus Ig quantities accurately assigned patients to subtypes for potential trial enrichment.

Key Findings

• Four proteome-defined sepsis subtypes were identified, spanning a severity gradient; one cluster showed 100% mortality.
• Subtypes exhibited distinct immune signatures: adaptive immunity activation with elevated immunoglobulins vs acute inflammation with lowest Ig levels, corroborated by orthogonal assays.
• A random-forest classifier using 10 proteins plus immunoglobulin quantities accurately assigned patients to clusters 1–3, enabling potential diagnostic implementation.

Clinical Implications

Proteomic subtyping could enable early stratification, personalized immunomodulation, and predictive enrichment in sepsis trials. A pragmatic 10-protein+Ig panel may be adapted to routine diagnostics after validation.

Limitations

• Generalizability beyond the studied cohort and settings is unproven; no interventional validation
• Classifier optimized for clusters 1–3; longitudinal dynamics and external clinical implementation remain to be tested

Future Directions

Validate the 10-protein+Ig panel across diverse cohorts, develop a clinical-grade assay, and design subtype-enriched interventional trials.