Abstract Glioblastoma (GBM) is an extremely aggressive and treatment-resistant primary brain tumor with a markedly poor prognosis. Although various molecular subtypes have been proposed to improve diagnostic and therapeutic approaches, their translation into clinical practice remains limited due to ambiguous classification criteria resulting from intra-tumoral heterogeneity and low clinical relevance. In this study, RNA-seq data from GBM cell lines in the DepMap database and bulk RNA-seq datasets from The Cancer Genome Atlas (TCGA), the Chinese Glioma Genome Atlas (CGGA), PRJNA1051047, and in-house cohorts were utilized. Furthermore, GeoMx DSP data paired with matched in-house RNA-seq samples were examined to characterize spatial transcriptomic patterns. Using Hallmark single-sample Gene Set Enrichment Analysis (ssGSEA) module scores, GBM cell lines were divided into two new subtypes by Non-negative Matrix Factorization (NMF) consensus clustering. Differentially expressed gene (DEG) signatures derived from these clusters were then used to classify IDH-wildtype GBM samples from four independent cohorts into two groups—designated as CA (cytokine-active) and GA (growth-active) subtypes—via a Bayesian compound covariate prediction (BCCP) model. The CA subtype exhibited poorer survival outcomes and elevated enrichment of tumor-associated pathways, including IL2-STAT5 signaling, apoptosis, and inflammatory response. Notably, TP53 emerged as a prominent upstream regulator in the CA group, consistent with increased ssGSEA module scores of 10 TP53-related pathways in various curated datasets in four patient cohorts. GeoMx DSP analyses further compared TP53 expression and ssGSEA module scores across α-SMA, CD45, and CD31 annotated regions of interest (AOIs) categorized into CA and GA subtypes. While TP53 expression did not significantly differ between the two groups in any AOI category, the Hallmark TP53 signaling module score was specifically elevated in α-SMA AOIs of CA subtype, particularly around perivascular regions. Collectively, two clinically relevant subtypes were identified, with the CA subtype being strongly associated with poor prognosis and significant dysregulation of TP53 downstream pathways. These findings suggest that TP53 may play a potentially important role in the aggressiveness of GBM. Citation Format: Yu Jin Kim, Jeongman Park, Woo Young Kwon, Jaejoon Lim, Sung Hwan Lee. A novel glioblastoma subtype classification using hallmark gene set signatures: Association between poor prognosis and TP53 downstream pathway abstract. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 3886.
Kim et al. (Fri,) studied this question.