Proteomics-guided Biomarker Discovery, Validation, and Pathway Perturbation in Infection-related Acute Decompensation of Cirrhosis.

Abstract

Background & aims: Inappropriate treatment of infections fuels drug resistance, organ failures, and costs in cirrhosis. We explored proteomics to improve infection diagnosis and management in acutely decompensated (AD) cirrhosis.

Methods: We enrolled 391 patients with AD cirrhosis (92% males, median-age: 41 years), 84 in the discovery cohort (54 infected, 30 non-infected), 147 in the validation cohort I (106 infected, 41 non-infected), and 160 in the validation cohort II (108 infected, 52 non-infected). High-throughput proteomics identified biomarkers in the discovery cohort, validated through enzyme-linked immunoassay in subsequent cohorts. A model for infection was evaluated through discrimination, calibration, and decision curves and was externally validated.

Results: Infected patients exhibited higher leucocyte counts, procalcitonin, organ failures, Model for End-stage Liver Disease scores, and 30-day mortality (P < .001 each). Proteomics identified 516 proteins, 27 upregulated and 38 downregulated, in infections. LGALS3BP, PLTP, CFP, and GPX3 were independently linked to infections (adjusting for severity and systemic inflammatory response syndrome), with composite area under the receiver operating characteristic curve (AUC) of 0.854 (95% confidence interval [CI], 0.787-0.922) in validation cohort I. A PACIFY model (LGALS3BP + procalcitonin + CLIF-COF + lactate) predicted infections with AUC of 0.965 (95% CI, 0.933-0.997) and 0.906 (95% CI, 0.860-0.952) in validation cohorts I and II, outperforming procalcitonin, systemic inflammatory response syndrome, white blood cell, neutrophil-to-lymphocyte ratio, neutrophil %, and composite models (P < .001). The model demonstrated fair calibration, with decision curves indicating a net benefit of the model in treating infections and reducing unnecessary antimicrobial use. Consistent findings were observed on external validation (AUC, 0.949; 95% CI, 0.916-0.982), re-enforcing the accuracy and clinical utility of the model. A deployable app was developed for infection risk estimation, enhancing practical applicability. Impaired phagocytosis, complement functions, hypocoagulation, hypofibrinolysis, dysregulated carbohydrate metabolism, autophagy, heightened cell death, and proteolysis were key perturbed pathways in infections.

Conclusion: The study identifies novel protein signatures and pathways linked with infections in AD cirrhosis. A biomarker-guided treatment of infections can limit unnecessary antimicrobial use and the burden of drug resistance in cirrhosis.