Integrative Transcriptomic and Radiogenomic Analysis Identifies a MAPT-PC-GLUD1 Low-Signature State in Glioblastoma

Glioblastoma (GBM) is characterized by marked molecular, cellular, and spatial heterogeneity, which contributes to therapeutic resistance and poor clinical outcomes. Although metabolic adaptation and ribosome biogenesis are both central to GBM biology, it remains unclear whether a compact transcriptional state linking these programs can be connected to immune contexture, cellular heterogeneity, and MRI-visible imaging phenotypes. We developed an integrative multi-omics and radiogenomic framework to identify a biologically interpretable gene signature associated with GBM heterogeneity and prognosis. Candidate genes were screened from public GBM transcriptomic datasets by differential expression analysis, weighted gene co-expression network analysis, and machine learning-based feature selection. Associations with survival, clinicopathological variables, immune characteristics, genomic features, methylation, and single-cell expression patterns were systematically evaluated. MRI-based radiomic features from 92 patients with matched transcriptomic data were used to construct exploratory predictive models, and voxel-wise maps were generated to visualize imaging-derived intratumoral heterogeneity. We identified a three-gene signature composed of MAPT, PC, and GLUD1. Lower expression of this signature defined a low-signature state associated with worse survival, unfavorable molecular features, higher tumor mutational burden, and a more immunoregulatory tumor microenvironment. Single-cell analysis revealed cell-type-specific expression heterogeneity, whereas radiomic models showed moderate performance in estimating gene-associated imaging signatures. Exploratory pharmacogenomic and molecular docking analyses suggested candidate compounds for further investigation. Together, these findings define a MAPT-PC-GLUD1 low-signature state that links metabolic-cellular programs with immune regulation, malignant progression, and MRI-derived spatial heterogeneity. Rather than serving merely as a prognostic model, this signature provides a biologically interpretable radiogenomic axis for future translational studies in GBM.