Identification of Novel Biomarkers for Ischemic Stroke Through Integrated Bioinformatics Analysis and Machine Learning

Abstract

Ischemic stroke leads to permanent damage to the affected brain tissue, with strict time constraints for effective treatment. Predictive biomarkers demonstrate great potential in the clinical diagnosis of ischemic stroke, significantly enhancing the accuracy of early identification, thereby enabling clinicians to intervene promptly and reduce patient disability and mortality rates. Furthermore, the application of predictive biomarkers facilitates the development of personalized treatment plans tailored to the specific conditions of individual patients, optimizing treatment outcomes and improving prognoses. Bioinformatics technologies based on high-throughput data provide a crucial foundation for comprehensively understanding the biological characteristics of ischemic stroke and discovering effective predictive targets. In this study, we evaluated gene expression data from ischemic stroke patients retrieved from the Gene Expression Omnibus (GEO) database, conducting differential expression analysis and functional analysis. Through weighted gene co-expression network analysis (WGCNA), we characterized gene modules associated with ischemic stroke. To screen candidate core genes, three machine learning algorithms were applied, including Least Absolute Shrinkage and Selection Operator (LASSO), random forest (RF), and support vector machine-recursive feature elimination (SVM-RFE), ultimately identifying five candidate core genes: MBOAT2, CKAP4, FAF1, CLEC4D, and VIM. Subsequent validation was performed using an external dataset. Additionally, the immune infiltration landscape of ischemic stroke was mapped using the CIBERSORT method, investigating the relationship between candidate core genes and immune cells in the pathogenesis of ischemic stroke, as well as the key pathways associated with the core genes. Finally, the key gene VIM was further identified and preliminarily validated through four machine learning algorithms, including generalized linear model (GLM), Extreme Gradient Boosting (XGBoost), RF, and SVM-RFE. This study contributes to advancing our understanding of biomarkers for ischemic stroke and provides a reference for the prediction and diagnosis of ischemic stroke.