Integrated bioinformatics and machine learning, research on specific biomarkers for large-artery atherosclerosis stroke.

Large-artery atherosclerosis stroke (LAA) is the main subtype of ischemic stroke. Currently, the diagnosis of LAA is confirmed through magnetic resonance imaging. At present, numerous substances are extensively studied as biomarkers, including metabolites, inflammatory factors, RNA, and more. In this study, we incorporated LAA peripheral blood data from various datasets and employed methods such as differential analysis, GSEA, WGCNA, PPI and machine learning to identify multiple molecules uniquely expressed in LAA. We validated these characteristic genes through qRT-PCR using clinical blood samples. We identified 131 differentially expressed genes in peripheral blood samples of stroke patients through differential analysis, and found through functional enrichment analysis that they are associated with inflammation related signaling pathways, Lysosome, Phagosome, and other related reactions. WGCNA identified the modules most relevant to large-artery atherosclerosis stroke, obtained 50 key genes by taking the intersection, and then used random forest to determine feature genes. The reliability of the results was verified by qRT-PCR. Finally, we constructed lncRNA-mRNA network to evaluate the interaction between lncRNA and mRNA. Our research suggested that PYGL, CLEC4D, LAMP2, and IRS2 can serve as biomarkers for diagnosing LAA, and Lnc-STARD3NL-4:1 may be a key lncRNA RNA regulating multiple LAA related factors.