Exploring the ferroptosis-related gene lipocalin 2 as a potential biomarker for sepsis-induced acute respiratory distress syndrome based on machine learning

Abstract

BackgroundSepsis is a major cause of mortality in patients, and ARDS is one of the most common outcomes. The pathophysiology of acute respiratory distress syndrome (ARDS) caused by sepsis is significantly impacted by genes related to ferroptosis. MethodsIn this study, Weighted gene co-expression network analysis (WGCNA), protein-protein interaction (PPI) networks, functional enrichment analysis, and machine learning were employed to identify characterized genes and to construct receiver operating characteristic (ROC) curves. Additionally, DNA methylation levels were quantified and single-cell analysis was conducted. To validate the alterations in the expression of Lipocalin-2 (LCN2) and ferroptosis-related proteins in the in vitro model, Western blotting was carried out, and the changes in intracellular ROS and Fe2+ levels were detected. ResultsA combination of eight machine learning algorithms, including RFE, LASSO, RandomForest, SVM-RFE, GBDT, Bagging, XGBoost, and Boruta, were used with a machine learning model to highlight the significance of LCN2 as a key gene in sepsis-induced ARDS. Analysis of immune cell infiltration showed a positive correlation between neutrophils and LCN2. In a cell model induced by LPS, it was found that Fer-1, a ferroptosis inhibitor, was able to reverse the expression of LCN2. Knocking down LCN2 in BEAS-2B cells reversed the LPS-induced lipid peroxidation, Fe2+ levels, ACSL4, and GPX4 levels, indicating that LCN2, a FRG, plays a crucial role in mediating ferroptosis. ConclusionUpon establishing an FRG model for individuals with sepsis-induced ARDS, we determined that LCN2 could be a dependable marker for predicting survival in these patients. This finding provides a basis for more accurate ARDS diagnosis and the exploration of innovative treatment options.