Background: Glioblastoma is an aggressive and heterogeneous brain tumor, posing significant treatment challenges. DNA methylation profiling enables a detailed characterization of tumor heterogeneity. This study applies methylation-based deconvolution to identify glioblastoma cellular composition and its association with patient outcomes. Methods: Reference atlas with DNA methylation signatures from brain-tumor-relevant cell-types including in vitro-derived human neural progenitors and oligodendrocytes-lineage precursors, was used to deconvolve 263 adult glioblastoma, IDH-Wildtype (IDHwt) samples. Tumor purity was evaluated using RF_Purify, and Kaplan-Meier survival analysis was performed with clinical data from 115 TCGA adult glioblastoma, IDHwt patients. Results: Deconvolution revealed distinct glioblastoma cellular compositions consistent with single-cell RNA sequencing studies. Purity analysis distinguished neoplastic (~70%) from non-neoplastic components. Neoplastic components were primarily oligodendrocyte-lineage cells (69%), including oligodendrocyte-like (42%) and oligodendrocyte precursor-like (27%) populations, along with astrocyte-like (20%) and mesenchymal stem cell-like (11%) signatures. The non-neoplastic fraction comprised macrophages, vascular endothelial cells, and immune cells. Higher oligodendrocyte-like proportions correlated with poorer survival (MST: 11.9 months; p<0.011), while higher astrocyte-like proportions were linked to better survival (MST: 17 months; p<0.046). The astrocyte-to-oligodendrocyte ratio emerged as a strong prognostic marker (MST: 16.2 vs. 9 months; p<0.007). Conclusions: Methylation-based deconvolution effectively characterizes glioblastoma heterogeneity, identifying oligodendrocyte-like predominance and its negative prognostic impact. In contrast, astrocyte-like populations are associated with improved outcomes. The astrocyte-to-oligodendrocyte ratio represents a promising biomarker for patient stratification, potentially guiding personalized treatment strategies.