Predicting Treatment Outcomes in Glioblastoma: A Risk Score Model for TMZ Resistance and Immune Checkpoint Inhibition

Abstract

Glioblastoma (GBM) presents significant therapeutic challenges due to its invasive nature and resistance to standard chemotherapy, i.e., temozolomide (TMZ). This study aimed to identify gene signatures that predict poor TMZ response and high PD-L1/PD-1 tumor expression, and explore potential sensitivity to alternative drugs. We analyzed The Cancer Genome Atlas (TCGA) biopsy data to identify differentially expressed genes (DEGs) linked to these characteristics. Among 33 upregulated DEGs, 5 were significantly correlated with overall survival. A risk score model was built using these 5 DEGs, classifying patients into low-, medium-, and high-risk groups. We assessed immune cell infiltration, immunosuppressive mediators, and epithelial-mesenchymal transition (EMT) markers in each group using correlation analysis, Gene Set Enrichment Analysis (GSEA), and machine learning. The model demonstrated strong predictive power, with high-risk patients exhibiting poorer survival and increased immune infiltration. GSEA revealed upregulation of immune and EMT-related pathways in high-risk patients. Our analyses suggest that high-risk patients may exhibit limited response to PD-1 inhibitors, but could show sensitivity to etoposide and paclitaxel. This risk score model provides a valuable tool for guiding therapeutic decisions and identifying alternative chemotherapy options to enable the development of personalized and cost-effective treatments for GBM patients.