Objective To screen key genes related to the diagnosis and prognosis of sepsis using bioinformatics methods. Methods A retrospective study was conducted on 90 sepsis patients admitted to the surgical intensive care unit (ICU) of Zhongshan Hospital, Fudan University from August 2022 to January 2023, as well as 30 control patients in ICU during the same period. The sepsis group was divided into a death subgroup (n=36) and a survival subgroup (n=54) based on the outcome, and peripheral blood mononuclear cells were collected for RNA sequencing. Linear models for microarray data (Limma) and weighted gene co-expression network analysis (WGCNA) were used to screen differentially expressed genes and module genes, combined with LASSO regression and random forest model for feature gene screening, candidate genes were determined. Nomogram and ROC curves to evaluate the diagnostic and prognostic value of candidate genes were constructed. The differential expression of candidate genes between the sepsis group and the control group was verified using external datasets and RT-qPCR. Results SEMA4F and PQLC3 were identified as candidate genes, and a nomogram for sepsis diagnosis and prognosis prediction was successfully constructed. The ROC curve showed that the AUC of the predictive efficacy of genes SEMA4F, PQLC3, and their combination for sepsis were 0.830, 0.926, and 0.930, respectively. The AUC of SEMA4F, PQLC3, SOFA score, and their combined predictive power for sepsis prognosis were 0.744, 0.768, 0.759, and 0.832, respectively. The AUC of the two genes in the diagnosis and prognosis prediction of sepsis validated by external datasets was greater than 0.588. The RT-qPCR results showed that there were statistically significant differences in the expression levels of the two genes among the control group, survival subgroup, and death subgroup (P＜0.05). Conclusions The SEMA4F and PQLC3 genes can serve as potential molecular markers for the diagnosis and prognosis of sepsis, and help improve the predictive value of SOFA score.