Identification and validation of potential shared diagnostic markers for sepsis-induced ARDS and cardiomyopathy via WGCNA and machine learning

Jing Wei¹Baoyue Huang^1*Kunlin Hu¹Xiong Bin^1,2Shulin Xiang^1*

• ¹The People’s Hospital of Guangxi Zhuang Autonomous Region, Nanning, China
• ²Guangxi Clinical Research Center Construction Project for Critical Treatment of Major Communicable Diseases, Nanning, China

Background: Sepsis frequently results in complications such as acute respiratory distress syndrome (ARDS) and cardiomyopathy. This study aims to identify common diagnostic markers and elucidate the underlying mechanisms of these sepsis-induced complications. Methods: We obtained datasets related to ARDS and sepsis-induced cardiomyopathy (SIC) from the GEO database and applied weighted gene co-expression network analysis (WGCNA) to identify differentially expressed genes (DEGs), which were integrated with key module genes. Feature genes were selected using support vector machine-recursive feature elimination (SVM-RFE) and random forest (RF) algorithms. An artificial neural network (ANN) model was constructed and its diagnostic performance was evaluated using receiver operating characteristic (ROC) curves. Machine learning algorithms effectively identified key hub genes associated with sepsis-induced ARDS and cardiomyopathy, with their robustness validated through ROC analysis. A cellular model of sepsis-induced lung injury was employed to examine hub gene expression. Additionally, we investigated inflammation and immune responses by characterizing immune landscapes using CIBERSORT and performing correlation analyses among feature genes, immune infiltration, and clinical characteristics. Finally, potential small-molecule compounds were identified from the PubChem database. Results: Five key genes—LCN2, AIF1L, STAT3, SOCS3 and SDHD—were identified. SOCS3 showed strong diagnostic potential with gene set enrichment analysis (GSEA) highlighting its role in biological processes and immune responses. SOCS3 expression correlated strongly with immune cells. Dexamethasone, resveratrol and curcumin were identified as potential SOCS3-targeting drugs. Conclusion: Five genes were identified as diagnostic biomarkers for sepsis-induced ARDS and cardiomyopathy, with SOCS3 serving as a key hub gene and potential therapeutic target.