Identification of Key Genes Influencing Carotid Atherosclerotic Plaque Stability Through Integrated Bioinformatics and Machine Learning Approaches: Implications for Immune Infiltration Dynamics

BackgroundThis investigation aims to identify pivotal genetic determinants influencing carotid atherosclerotic plaque stability and characterize their associations with immune cell infiltration patterns, particularly focusing on correlations between key genes and immune cell surface markers. MethodsThree microarray datasets (GSE28829, GSE43292, GSE163154) were retrieved from the NCBI Gene Expression Omnibus (GEO). Differentially expressed genes (DEGs) associated with AS progression were identified using the Limma package in R. Subsequent functional characterization employed Gene Ontology (GO)/Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses and protein-protein interaction (PPI) network construction. Machine learning algorithms - random forest (RF), least absolute shrinkage and selection operator (LASSO) regression, and support vector machine-recursive feature elimination (SVM-RFE) - were implemented for hub gene identification. Validation procedures included stratified box plots, receiver operating characteristic (ROC) curve analysis, and logistic regression calibration using external datasets. Immune cell infiltration profiles were quantified through CIBERSORT algorithm, with subsequent correlation analysis between hub genes and differentially expressed immune cell markers. ResultsComparative analysis revealed 559 and 1,357 DEGs in GSE28829 and GSE43292 respectively. The Gene Ontology (GO)/Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis demonstrated significant involvement of DEGs in immune-related pathways, while PPI network analysis highlighted predominant inflammatory-immune interactions. Machine learning convergence identified four hub genes: CCL21, RAB23, SIGLEC1, and NRP1. The predictive performance of these key genes was confirmed through ROC curves and calibration curves of logistic regression models. The CIBERSORT algorithm was used to determine the proportion of immune cells in early-stage and late-stage samples, revealing differences in the distribution of five immune cell types. Additionally, the correlation between the expression levels of four key genes and immune cell surface markers was analyzed, resulting in the identification of six gene-immune marker correlation pairs with a correlation coefficient of |R|>0.6. ConclusionThis study identified four key genes (CCL21, RAB23, SIGLEC1, and NRP1) using ML methods. Specific immune cells exhibited varying infiltration proportions between early-stage and late-stage samples, including naive B cells, memory B cells, CD8+ T cells, M0 macrophages, and M2 macrophages. The common cell surface markers of CD8+ T cells, B cells, and macrophages were identified. A correlation analysis of the expression of key genes and immune cell markers was conducted, demonstrating a high correlation between the expression of six gene pairs (SIGLEC1-CD68, SIGLEC1-CSF1R, SIGLEC1-MRC1, SIGLEC1-CD163, RAB23-CSF1R, and RAB23-CD68). These findings may provide a theoretical foundation for further research into the immune mechanisms involved in AS development.