Identification of Sepsis-Associated Encephalopathy Biomarkers Through Machine Learning and Bioinformatics Approaches

Abstract

Background Sepsis-associated encephalopathy (SAE) is prevalent in septic patients and presents as a combination of extracranial infection and clinical manifestations of neurological dysfunction. Typical symptoms of the disease include acute cognitive impairment and long-term cognitive decline. It is associated with increased mortality in sepsis. The aim of this study was to identify SAE-related genes and explore their diagnostic value in SAE. Methods We analyzed the existing sepsis-associated encephalopathy datasets GSE198862 and GSE167610, subsequently merged them after batch correction, and reanalyzed the combined dataset. Using Weighted Gene Co-expression Network Analysis (WGCNA), we identified the most crucial gene module. This module was then subjected to various machine learning methods to identify feature genes, and finally, in vitro experiments were conducted to validate the expression of these genes in sepsis-associated encephalopathy. Results In the analysis of the combined GSE198862 and GSE167610 datasets, we identified 138 differentially expressed genes, with 84 genes showing significant upregulation in the non-merged datasets. Notably, the "Coral" module, discovered through WGCNA, contained 728 genes, exhibiting a remarkable overlap with the previously identified differentially expressed genes. Machine learning approaches, including Elastic Net regression, LASSO, random forest, and XGBoost, yielded 5 and 11 marker genes, respectively. These markers, including Lcn2, Atp10d, Rps21, Anax2, Gabarap, S100a11, Pglyrp1, Labm3, and Fkbp4, displayed conspicuous upregulation in a concentration-dependent neurodegenerative disease model. Conclusion This study reveals significant upregulated biomarkers in septic encephalopathy, indicating the core mechanisms associated with the pathogenesis of the condition, which could serve as potential therapeutic targets.