Integrative transcriptomics, machine learning, and molecular docking derive a DAM-like macrophage signature for risk stratification and therapeutic nomination in glioblastoma.

Glioblastoma presents a significant challenge for drug development due to its complex immunosuppressive microenvironment. Identifying candidate therapeutic targets within this landscape requires advanced computational approaches. Employing an integrative transcriptomics strategy combining single-cell RNA sequencing and bulk transcriptomics, we identified a distinct Disease-Associated Microglia (DAM)-like macrophage subpopulation as a putative contributor to therapeutic resistance. Using machine learning-based feature selection via LASSO, we identified a core stromal remodeling signature comprising MMP9, MMP14, and CCL2 that links metabolic reprogramming to an immune-excluded phenotype. Our analysis suggests that this specific metabolic state is associated with elevated secretion of matrix-modifying enzymes, which may contribute to restricting T cell access. The signature was validated against four alternative prognostic models and showed favorable generalization in an independent external cohort. An individualized nomogram integrating the risk score with clinical variables was constructed for personalized survival prediction. Computational pharmacogenomic profiling suggests that this signature may predict intrinsic resistance to temozolomide but high sensitivity to CSF1R and TGF-β pathway inhibitors, and molecular docking simulations provide structural support for candidate drug-target interactions. These in silico findings identify pexidartinib and galunisertib as candidate agents for further preclinical evaluation targeting the DAM-like metabolic state. This study provides a computational framework for dissecting macrophage heterogeneity and nominating candidate therapeutic strategies that require further experimental validation in glioblastoma.