Identification of Cuproptosis-Related Gene in Sepsis by Machine-Learning

Sepsis is a serious public issue that affects millions of people. Cuproptosis, a newly form of cell death, has been linked to the course of certain illnesses. Therefore, the goal of our study was to investigate clusters of sepsis associated with cuproptosis and to develop a prediction model. Using GSE33341 dataset, the expression profiles of immune- and cuproptosis-related genes (CRGs) in sepsis patients were examined. We investigated molecular clusters based on CRGs and the accompanying immune cell infiltration. The weighted gene co-expression network analysis (WGCNA) method was used to identify differentially expressed genes in a specific cluster. Four machine learning methods were then contrasted in order to determine the best machine model. A nomogram, calibration curve, decision curve analysis were used to validate predictive efficiency. Immune infiltration analysis revealed significant immunological heterogeneity among the CRGs. The random forest machine model indicated the best discriminative performance, with comparatively smaller residual and root mean square errors and a larger area under the curve (AUC = 0.983). A final 5-gene random forest model was created, and two external validation datasets demonstrated that it performed satisfactorily (AUC = 0.964 and 0.851). The accuracy of predicting sepsis was also demonstrated by the nomogram, calibration curve, and decision curve analysis. Further investigation indicated a substantial relationship between age and SAR1B expression. This study identified the impact of cuproptosis on sepsis for the first time, and further characterized the underlying molecular pathways causing sepsis heterogeneity.