Proteomics-based machine learning model for predicting secondary infection in HBV-related liver failure - PubMed
2 months ago
- #Proteomics
- #Machine Learning
- #Liver Failure
- Proteomics-based machine learning model developed for predicting secondary infection (SI) in HBV-related liver failure patients.
- Study involved 114 patients in discovery cohort and 60 each in two validation cohorts using untargeted proteomics.
- Identified SI-related proteins through feature selection and logistic regression modeling, with external validation via targeted proteomics and ELISA.
- Inflammatory and coagulation pathway dysregulation strongly linked to SI risk.
- Final model includes Lysozyme (LYZ), Calmodulin 1 (CALM1), Serpin Family D Member 1 (SERPIND1), Dermatopontin (DPT), total bilirubin, and AST, showing excellent discrimination (AUROC 0.980 in discovery, 0.873 in validation).
- Model outperforms traditional markers like CRP, WBC, and NE%, and predicts 28-day mortality better than CLIF-C ACLF and MELD scores.
- ELISA-based validation in cohort 2 achieved AUROC of 0.883, confirming model reliability.
- Proteomics-derived model supports early clinical intervention for high-risk patients.