Machine learning identifies KRT8 dysregulation and endothelial remodeling in Moyamoya disease

Moyamoya disease (MMD) is a rare occlusive cerebrovascular disease, and its pathological mechanism remains unclear at present. The abnormal vascular remodeling may be involved in vascular endothelial cells. In this study, RNA seq was performed on the superficial temporal arteries of 10 patients with MMD. Integrated analysis was conducted by combining validation set with training set. Key genes were identified through differential analysis and WGCNA. The functions of potential biomarkers were explored by methods such as correlation analysis, KEGG analysis, PPI network, and tube formation experiments. Integrated analysis of three cohorts (43 MMD vs. 26 controls) identified 19 shared DEGs, including upregulated KRT8/KRT18 and downregulated NT5C2 (P < 0.001). Enrichment revealed dysregulation in circadian rhythm, calcium signaling, and metabolic pathways (P < 0.05). Immune infiltration showed elevated pro-inflammatory cells (neutrophils, M1 macrophages) and reduced Tregs/NK cells (P < 0.05). Machine learning (SVM-RFE, Boruta, LASSO) prioritized KRT8 as diagnostic markers (AUC > 0.96). KRT8 overexpression enhanced angiogenesis in HBMECs (1.5-fold tube formation, P < 0.01). This omics approach delineates MMD's molecular interplay between inflammation, metabolism, and vascular remodeling. KRT8 may promote vascular remodeling in MMD by regulating the tube-forming ability of endothelial cells. This could be a highlight of therapeutic targets for MMD and shed light on the mechanism research of MMD.