目的：构建m6A相关胶质母细胞瘤（glioblastoma，GBM）预后模型，识别受m6A调控的潜在预后生物 标志物。方法：从 TCGA 和 GEO 中分别收集了 GBM 的表达、临床表型和生存数据，m6A2Target 中获取 m6A调控因子。获得m6A调控因子与TCGA-GBM基因交集。根据TCGA-GBM表达，计算正常-癌症间差 异的基因。基于差异的m6A调控因子进行一致性聚类，识别亚型，构建亚型间差异的预后相关基因模型，并 进行验证。结果：从m6A2Target中获取了21个m6A调控因子，与TCGA-GBM包含的基因交集有19个。 其中6个m6A调控因子在GBM显著上调。基于差异的m6A调控因子的表达进行一致性聚类，得到cluster1/2共2个亚型。且cluster2亚型中患者的免疫评分以及基质评分均显著高于cluster1，但cluster1的整体 肿瘤纯度显著高于cluster2。生存分析发现，cluster2生存预后更差。cluster1和cluster2之间显著差异表达 的基因共6591个，单因素cox回归和多因素cox回归筛选到171个显著预后相关基因。LASSO-cox回归构 建预后风险模型显示在训练集（TCGA-GBM）及测试集（GSE121720）中均具有较好的模型效能。结论：多 个m6A因子在GBM患者中存在显著差异，且基于m6A分型的2组患者存在预后不同，并基于m6A的组间 差异构建了预后模型可指导患者的5年生存率。

To construct an m6A-related prognostic model for glioblastoma (GBM) and identify potential prognostic biomarkers regulated by m6A. Methods: Expression, clinical phenotype, and survival data of GBM were collected from TCGA and GEO, respectively, and 21 m6A regulatory factors were obtained from m6A2Target, of which 19 intersected with genes in TCGA-GBM. First, genes with differential expression between normal and cancer tissues were calculated based on TCGA-GBM expression data. Then, consistency clustering was performed based on differentially expressed m6A regulatory factors to identify subtypes, and a prognostic model based on differentially expressed genes between subtypes was constructed and validated using a test set. Results: We obtained 21 m6A regulatory factors from m6A2Target, of which 19 intersected with the genes included in TCGA-GBM. Among them, 6 m6A regulatory factors showed significantly higher expression in GBM. Based on the differential expression of m6A regulatory factors, two subtypes, cluster1 and cluster2, were identified through consensus clustering. Furthermore, patients in cluster2 exhibited higher immune scores and stromal scores, while cluster1 showed significantly higher overall tumor purity. Survival analysis revealed poorer prognosis in cluster2. A total of 6591 genes showed significant differential expression between cluster1 and cluster2. Through univariate and multivariate Cox regression, 171 significantly prognostic-related genes were selected. The LASSO-Cox regression constructed a prognostic risk model that demonstrated good performance in both the training set (TCGA-GBM) and the testing set (GSE121720). Conclusion: Multiple m6A factors show significant differences in GBM patients, and two groups of patients based on m6A typing have different prognoses. A prognostic model based on intergroup differences in m6A can guide patients' 5-year survival rate.