Proteomics-based machine learning model for predicting secondary infection in HBV-related liver failure
Feixiang Xiong1, Jianming Zheng2, Jiajia Chen3
1National Center for Infectious Diseases, Beijing Ditan Hospital, Capital Medical University, Beijing, China.
Nature Communications
|March 3, 2026
Summary
A new plasma proteomics model accurately predicts secondary infections in Hepatitis B liver failure patients. This tool aids early intervention, improving outcomes and outperforming existing risk scores.
Area of Science:
- Hepatology
- Proteomics
- Infectious Diseases
Background:
- Patients with Hepatitis B Virus-related liver failure face high risks of secondary infections.
- Early detection tools for secondary infections in these patients are currently limited.
Purpose of the Study:
- To develop and validate a plasma proteomics-based model for early risk assessment of secondary infections.
- To identify key protein biomarkers associated with secondary infection risk.
Main Methods:
- Prospective multicenter study with discovery and validation cohorts.
- Untargeted and targeted proteomics, Minimum Redundancy Maximum Relevance feature selection, logistic regression, and ELISA.
- Model performance evaluated using AUROC, compared against CRP, WBC, NE%, CLIF-C ACLF, and MELD scores.
Main Results:
- Proteomics identified dysregulation in inflammatory and coagulation pathways linked to secondary infections.
- A model comprising LYZ, CALM1, SERPIND1, DPT, total bilirubin, and AST demonstrated high predictive accuracy (AUROC 0.980 discovery, 0.873 validation).
- The model outperformed established clinical markers and scores in predicting secondary infections and 28-day mortality.
Conclusions:
- A plasma proteomics-derived model reliably identifies patients with Hepatitis B liver failure at high risk for secondary infections.
- This model supports timely clinical intervention, potentially improving patient outcomes.
- ELISA validation confirmed the model's robustness and potential for clinical application.
