Identification of MMP8, DDX24, RNASE2, and EMB as a novel diagnostic gene panel for sepsis: a transcriptome-based modeling study.

OBJECTIVE: Sepsis is a life-threatening condition with high mortality and complex pathology. Early diagnosis is critical but remains challenging due to a lack of effective biomarkers. This study aims to identify specific diagnostic markers to distinguish sepsis from non-sepsis, and to develop a robust diagnostic model. METHODS: PBMC transcriptome data from our cohort (24 healthy controls, 29 common infections, 51 sepsis patients) were analyzed to identify genes with expression levels increasing or decreasing with infection severity. Low-expression genes were excluded, and candidate markers were evaluated using multiple GEO datasets. Top-performing genes were selected to build a LASSO regression-based diagnostic model. Model performance was assessed by AUC, BSS, ROC and calibration curves, nomogram, DCA, and CIC curves. Functional analyses (GO, KEGG, immune infiltration, GSEA, PPI network) were performed to explore underlying immune mechanisms. RESULTS: A total of 114 genes showing expression changes with infection severity were identified. Four genes-MMP8, DDX24, RNASE2, and EMB-were selected based on diagnostic performance. The resulting model performed well in both our cohort and public datasets (AUC: 0.811-1; BSS: -0.464 - 0.964). In the GSE69686 dataset, the model also showed predictive value for neonatal and pediatric sepsis (AUC: 0.666-0.852; BSS: -1.34 - -0.293). Immune infiltration and GSEA revealed enrichment of these genes in neutrophils, monocytes, and T cells, reflecting key immune features of sepsis. CONCLUSION: The four identified genes-MMP8, DDX24, RNASE2, and EMB-collectively form a diagnostic model that effectively distinguishes sepsis patients from those with non-sepsis individuals.