A vasculogenic mimicry-related prognostic signature for medulloblastoma based on machine learning and immune microenvironment analysis

Vasculogenic mimicry (VM) is a tumor vascularization mechanism independent of endothelial cells and has been associated with aggressiveness and poor prognosis in various solid tumors. However, the clinical significance of VM in medulloblastoma (MB) and its relationship with the tumor immune microenvironment (TIME) remain largely unexplored. This study aimed to identify VM-related genes in MB, construct and validate a prognostic risk model based on these genes, and investigate their association with immune infiltration. The GSE50161 and GSE85217 datasets were downloaded from the GEO database. Core VM-related differentially expressed genes (DEGs) were identified through differential expression analysis, intersection with VM-related gene sets, and protein-protein interaction (PPI) network screening. Prognostic genes were selected using univariate Cox regression, LASSO-Cox regression, and random survival forest (RSF) analyses, followed by multivariate Cox regression to construct the risk model. Model performance was evaluated using Kaplan-Meier survival curves and time-dependent ROC curves. Single-sample gene set enrichment analysis (ssGSEA) was employed to estimate immune cell infiltration, and correlations between VM transcriptional activity and immune features were assessed. A total of 60 core VM-related DEGs were identified, which were enriched in hypoxia, epithelial-mesenchymal transition, and multiple tumor-associated signaling pathways. Intersection of the three prognostic screening methods yielded 13 shared prognostic genes, from which a four-gene risk model comprising MSH6, CASP6, NRP1, and FOXM1 was constructed. Patients in the high-risk group exhibited significantly worse overall survival ( P = 0.00019), with AUC values of 0.741, 0.736, and 0.728 at 1, 3, and 5 years, respectively. Multivariate Cox regression confirmed the risk score as an independent prognostic factor (HR > 1, 95%CI > 1). Subgroup analysis demonstrated significant prognostic value in the SHH and Group 4 molecular subgroups. Immune infiltration analysis revealed distinct immune cell enrichment patterns between risk groups, and the independent VM score showed significant correlations with multiple immune cell types, particularly natural killer cells ( R = 0.731) and MDSCs ( R = 0.722). All four prognostic genes were positively correlated with the independent VM score.