Abstract
This integrated study combines bioinformatics, machine learning, and Mendelian randomization (MR) to discover and validate molecular biomarkers for sepsis diagnosis. Methods include differential expression analysis, weighted gene co-expression network analysis (WGCNA) for identifying sepsis-related modules and hub genes, and functional enrichment analyses to explore the roles of hub genes. Machine learning algorithms identify 3 diagnostic genes - CD177, LDHA, and MCEMP1 - consistently highly expressed in sepsis patients. The nomogram model effectively predicts sepsis risk, supported by receiver operator characteristic (ROC) curves. Correlations between diagnostic genes and immune cell infiltration are observed. MR analysis reveals a positive causal relationship between MCEMP1 and sepsis risk. In conclusion, this study presents potential sepsis diagnostic biomarkers, highlighting the genetic association of MCEMP1 with sepsis for insights into early diagnosis.