This study aimed to analyze the gene expression characteristics of sepsis and search for potential biomarkers involved in the pathogenesis. The data on sepsis were obtained from the GEO database according to the keyword "sepsis". The CIBERSORT algorithm was applied to determine the immune cell. WGCNA package were applied to build a weighted gene network. Then, a topological overlap matrix was created and dynamic hybrid cutting was applied to categorizing the genes with identical expression patterns. Component analysis of each module was implemented according to module eigengenes. In order to detect the important modules, the connections among the immune infiltration of Mφ and the modules were computed by Pearson's tests. PPI network was made using the STRING database and cytoHubba was applied to find hub genes. A total of 760 sepsis samples as well as 42 healthy control samples were involved. A total of 14 gene modules were generated. Thereinto, the correlations of the yellow (includes 168 hub genes) and blue (includes 526 hub genes) modules with Mφ0 were 0.39 and -0.42, while with Mφ1 were 0.3 and -0.4. 916 up-regulated and 454 down-regulated DEGs were found in the sepsis group. 451 intersected genes were obtained after the intersecting of DEGs and the hub genes from blue and yellow modules. Subsequent GO enrichment analysis suggested that 451 overlapping genes were enriched in "T cell activation", "lymphocyte differentiation" and "T cell differentiation" for biological process. Besides, KEGG enrichment analysis showed that "Human T-cell leukemia virus 1 infection" and "Th17 cell differentiation" were the most enriched pathways. In PPI network, UTP6, RRS1, RRP1B, DDX18, and DDX24 were identified as hub genes. ROC analysis showed the AUC values of these genes were all greater than 0.95. UTP6, RRS1, RRP1B, DDX18, and DDX24 participate in the pathogenetic process of sepsis through regulating the activation and differentiation of lymphocytes. Besides, these five genes could be used for diagnosing sepsis.