Glioblastoma

Abstract

Glioblastoma (GBM) is the most common and aggressive malignant brain tumor in adults, with a five-year survival rate of approximately 7%. Despite maximal surgical resection followed by radiotherapy and alkylating chemotherapy, most patients experience recurrence, reflecting the tumor’s marked biological heterogeneity and resistance to treatment. This review examines how cellular plasticity within cancer stem-like populations, epigenetic regulation, and microenvironmental signaling contribute to therapeutic failure in GBM, and how emerging AI-based imaging approaches may complement existing strategies.

Single-cell and functional studies support a state-based model of glioblastoma stem cells (GSCs), in which stem-like properties are dynamically acquired and lost in response to microenvironmental cues and therapeutic pressure. Hypoxic and perivascular niches stabilize these states, while epigenetic programs involving regulators such as EZH2, BMI1, and histone deacetylases shape transcriptional responses that promote stress tolerance and persistence. Developmental pathways, including Notch, further modulate these programs in a context-dependent manner. Together, these mechanisms help explain how GBM evades eradication and re-establishes disease after treatment.

Clinically, recent advances such as Tumor Treating Fields have produced modest survival gains, and molecular biomarkers enable limited stratification. Radiomic and machine learning approaches extend personalization by extracting quantitative features from routine magnetic resonance imaging (MRI), but remain constrained by cohort size, imaging variability, and interpretability.

Integrating insights from stem cell biology, epigenetic regulation, and AI-assisted diagnostics reframes GBM as a dynamic, state-driven disease. Progress is therefore likely to depend on strategies that address plasticity and heterogeneity while combining biological and computational approaches to guide more adaptive and individualized care.