Identification and validation of novel candidate genes with diagnostic value for sepsis via weighted gene...

Sepsis is a systemic inflammatory response caused by a variety of causes, which is characterized by high morbidity and mortality. Our work aimed to screen the candidate genes of sepsis and evaluate their diagnostic value using bioinformatics tools.

Multiple GEO datasets were integrated. GSE9960 and GSE28750 were merged as the training set, with other datasets used as validation sets for immune infiltration, prognosis, ROC analysis, subtype analysis, and single-cell analysis. Sepsis-associated genes were identified via weighted gene co-expression network analysis (WGCNA) and protein-protein interaction (PPI) network analysis. Functional enrichment was explored using GSVA and GSEA. Diagnostic potential was evaluated by ROC curves. Correlation between gene expression and immune cells was analyzed by Pearson correlation, and gene expression was validated by qRT-PCR.

WGCNA identified 1,463 sepsis-associated genes, which were enriched in biosynthesis/metabolism and immune-related signaling pathways. Five hub genes (CDK1, CCNB1, CCNA2, AURKB and BUB1) were screened via PPI network, all highly expressed in sepsis. Higher AURKB or BUB1 expression correlated with shorter overall survival. scRNA-seq revealed broad expression of these genes in pediatric immune cells (B cells, monocytes, T cells), but restricted to T cells in adult samples. A logistic regression model combining the five genes showed preliminary discriminative potential for distinguishing sepsis from healthy controls in the training set (AUC = 0.747) and an independent validation cohort (GSE65682, AUC = 0.799). In exploratory analyses, the model also demonstrated potential for differentiating septic shock from cardiogenic shock (AUC = 0.743) and from non-septic shock (AUC = 0.766). These genes exhibited significant correlations with immune cell infiltration in sepsis.

In this study, we identified five sepsis-associated genes (CDK1, CCNB1, CCNA2, AURKB, BUB1). A logistic regression model based on these five genes demonstrated improved and consistent diagnostic performance for sepsis, and these genes also correlated with immune cell infiltration in sepsis. However, as the current results are correlational and do not establish a causal role, they should be interpreted as hypothesis-generating rather than conclusive, warranting further validation in independent prospective cohorts.