Abstract
Sepsis is a systemic inflammatory response syndrome caused by an infection featuring high morbidity and mortality due to complex mechanisms underlying immune dysfunction. In this study, based on the sepsis transcriptome profiles from the GEO datasets (GSE65682, GSE28750, GSE95233, and GSE167363), we used the machine learning method and other computational algorithms, such as differential gene expression analysis, weighted gene coexpression network analyses (WGCNA), and the building of PPI networks to identify four hub genes (DDX24, GZMM, KCNA3, and NCL). The quantitative reverse transcription PCR performed preliminary validation that all four hub genes were significantly downregulated in patients with sepsis. DDX24 had the highest diagnostic performance (AUC > 0.8) for discriminating patients from normal subjects. GZMM was found to be significantly related to the prognoses of patients as well as APACHE II scores, and the downregulated expression pattern might represent T cell and NK cell exhaustion. Analysis based on single-cell RNA sequencing showed that DDX24 and GZMM were mainly expressed in T cells and NK cells, and the expression trends strongly correlate with patient survival. Functional enrichment analysis suggested that the hub genes likely participate in regulation of immune responses, especially those pertaining to T cells. Drug prediction found 25 candidate drugs that will serve as new therapeutic targets for precision medicine to treat sepsis. Overall, the multifaceted study shed light on key roles played by these hub genes (especially DDX24 and GZMM) in the development of sepsis and will be useful references in diagnosing patients and estimating prognosis.