Identification of RING finger protein-related biomarkers in sepsis based on explainable machine learning: A comprehensive analysis.

OBJECTIVETo systematically identify RING finger protein (RNF) biomarkers with diagnostic value in sepsis and to explore their potential as therapeutic targets.

METHODSSepsis peripheral blood transcriptome datasets from Gene Expression Omnibus (GEO) were integrated. Sepsis-associated RNF genes were identified via differential expression analysis and weighted gene co-expression network analysis (WGCNA). Feature genes were selected using LASSO, SVM-RFE, and SHAP. Summary-data-based Mendelian randomization (SMR) assessed causality, and diagnostic performance was validated by receiver operating characteristic (ROC) curves in three independent cohorts. Single-cell RNA sequencing resolved cellular localization and perturbations. DrugReflector deep learning predicted repurposable drugs, molecular docking evaluated binding affinity, and PheWAS assessed target safety.

RESULTSA total of 13 sepsis-related RNF genes were identified, and three feature genes (ZFP36L2, RNF125, RNF175) were further selected. SMR showed that only elevated RNF175 expression significantly increased sepsis risk (OR = 1.267, 95% CI 1.002-1.603, P = 0.048). The diagnostic AUCs of RNF175 were 0.931, 0.776, and 0.835 in the training and two validation cohorts, respectively. Single-cell atlas revealed predominant RNF175 expression in T cells with significant upregulation in sepsis (P < 0.001). Deep learning predicted tadalafil as the top candidate, with a molecular docking binding energy of -9.072 kcal/mol for RNF175. PheWAS found no significant phenotype associations, indicating low off-target risk.

CONCLUSIONSRNF175 is an early sepsis biomarker with both diagnostic value and causal relevance, potentially acting through T-cell modulation. Tadalafil may be a pathway-targeted repurposable drug. This study provides new perspectives for precision medicine in sepsis.