Multi-Omics and Machine Learning Analyses Reveal PIK3CG, PRKCD, and TRIM22 as Potential Markers of Poor Prognosis and Immune Activation in Glioblastoma

Background: Glioblastoma (GBM) is one of the most aggressive brain tumors with a poor prognosis despite current treatment modalities. This study aimed to identify genes whose high expression is paradoxically associated with both poor survival and enhanced immune activity, as potential targets for combination chemotherapeutic and immunotherapeutic strategies. Methods: Transcriptomic data from patients with central nervous system World Health Organization (WHO) grade IV gliomas (based on the 2016 WHO classification) were analyzed, using datasets from The Cancer Genome Atlas (525 cases), the Chinese Glioma Genome Atlas (250 cases), and the Genotype-Tissue Expression (1,152 normal samples). We initially screened 12,041 genes, prioritizing those showing a paradoxical association with prognosis and immune activation. Key genes were selected through rank statistics, machine-learning-based survival modeling, and pathway network analysis. Further subgroup validation was performed using only isocitrate dehydrogenase (IDH)-wildtype GBM cases, in line with the 2021 WHO classification. Results: Among the 12,041 candidate genes analyzed, phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit gamma (PIK3CG), protein kinase C delta type (PRKCD), and tripartite motif-containing protein 22 (TRIM22) were identified as key biomarkers whose elevated expression was significantly associated with poorer overall and disease-specific survival in IDH-wildtype GBM. These genes also correlated with enhanced immune activity, including increased tumor-infiltrating lymphocytes and elevated expression of programmed death-ligand 1. Pathway network analysis revealed indirect associations with critical immune markers such as CD8A and CD4, suggesting potential immunomodulatory functions. Additionally, differential gene expression and disease ontology analyses demonstrated their relevance across various cancer types. Drug sensitivity profiling using the Genomics of Drug Sensitivity in Cancer database identified AGI-6780, linsitinib, and Nutlin-3a as potential therapeutic agents targeting these genes. Conclusion: This study identifies PIK3CG, PRKCD, and TRIM22 as potential biomarkers and therapeutic targets in IDH-wildtype GBM. Their paradoxical association with poor survival and immune activation may inform personalized treatment strategies that combine conventional chemotherapy with immune-based therapies. While our findings are robust across both mixed and IDH-wildtype-focused cohorts, further mechanistic validation is warranted.