A machine learning-based approach to prognostic model of sepsis with PANoptosis-related genes and performance of scRNA-seq data to assess prognostic signatures

Abstract Sepsis is a systemic inflammatory response syndrome caused by the invasion of pathogenic microorganisms such as bacteria into the body. PANoptosis is an inflammatory programmed cell death with key characteristics of pyroptosis, apoptosis, and/or neoptosis. At present, there is no strong evidence to suggest that the prognosis of sepsis is closely related to PANoptosis. In this study, 38 key differentially expressed genes(DEGs) were obtained by analyzing DEGs in sepsis microarray data GSE65685 and GSE95233 and crossing them with the PANopotosis gene set. Then, gene features were screened through univariate analysis, lasso regression analysis, and multivariate COX regression analysis to construct a prognosis model consisting of three predictive features: IKBKB, AIM2, and CTSG. We used Kaplan Meier (K-M) survival analysis, receiver operating characteristic (ROC) time curve analysis, internal validation, and principal component analysis to evaluate the performance of the prognostic model. In addition, sepsis patients were divided into high-risk and low-risk groups based on risk scores and gene set enrichment analysis (GSEA) results, and significant differences were found in multiple immune cell functions and immune related KEGG signaling pathways. Subsequently, scRNA seq data and immune cell infiltration analysis showed that the IKBKB and AMI2 genes were highly expressed in all immune cells of sepsis patients, while the CTSG gene was mainly highly expressed in monocytes, neutrophils, NK cells, and proliferative T cells. In the analysis of target genes for 16 immunosuppressive drugs, only CSF3 was highly expressed in high-risk patients, indicating that CSF3 may be the most promising target for treating sepsis.