Abstract B034: Identifying intratumoural heterogeneity in glioma at the cohort level via the Minderoo Precision Brain Tumour Programme

Abstract Background Glioblastoma (GBM) is the most common adult primary brain cancer and has an average survival of 14. 6 months. This prognosis is unchanged over the past 20 years. Treatment resistance has been associated with intratumoural heterogeneity but few datasets have explored how heterogeneity affects outcome using multi-omic, multi modal datasets at the cohort level. Methods We use multi-region whole genome sequencing at initial surgery with concurrent whole transcriptome sequencing to integrate genomic variance with cell type deconvolution to define cellular heterogeneity in a clinical context and at the cohort level. Moreover, we explore how heterogeneity in alternate splicing affects cell type composition and survival. We performed paired whole genome sequencing and whole transcriptome sequencing on 319 samples taken at primary debulking surgery where 50 had paired, spatially distinct samples that included 176 GBM, 56 astrocytoma, and 26 oligodendroglioma as well as other rarer brain tumour subtypes. Results GBM cases were 61% male with a median observed survival of 383 days. The astrocytoma cohort comprised 31 high-grade and 25 low-grade tumors with median observed survival of 919 days. The oligodendroglioma cohort comprised 11 high-grade and 15 low-grade tumors with a median observed survival of 880 days. We developed an integrated genome-wide framework to summarize and compare DNA alteration landscapes and RNA expression programs across glioma and within patients. We parsed and quality-filtered small variants, structural variants, copy-number alterations, and gene fusions. Subsequent integration of RNA-sequencing was performed by constructing a cohort-wide variance-stabilized expression matrix. In the GBM cohort, DNA hotspots concentrated at characteristic sites of genome instability such as chromosome 7, 9, 10 and 12 whereas RNA hotspots included developmental and repression of neuronal-regulatory genes. We next defined an end-to-end isoform landscape of glioma by integrating differential transcript usage (DTU) between astrocytoma/oligodendroglioma and GBM alongside within-GBM isoform-state heterogeneity that is decoupled from gene expression. Across 319 sequenced tumors, we tested 16, 260 genes comprising 110, 390 transcripts and identified 2, 878 DTU genes at OFDR 0. 05, including 494 “dosage-independent” DTU genes with weak gene-expression evidence and small gene-level fold changes. Together, these results demonstrate extensive isoform remodelling that distinguishes GBM from lower-grade gliomas and reveal a distinct layer of within-GBM heterogeneity characterized by reproducible isoform-state switching that is frequently independent of gene dosage, nominating candidate biomarkers and mechanistic hypotheses for splicing-driven tumor phenotypes. Conclusion Significant heterogeneity at the level of the genome, transcriptome and spliceosome exists within glioma. Understanding how this heterogeneity affects precision and immune-based therapies may provide a route to improving glioma outcomes. Citation Format: Richard Mair, Matthaios Pitoulias. Identifying intratumoural heterogeneity in glioma at the cohort level via the Minderoo Precision Brain Tumour Programme abstract. In: Proceedings of the AACR Special Conference in Cancer Research: Brain Cancer; 2026 Mar 23-25; Philadelphia, PA. Philadelphia (PA): AACR; Cancer Res 2026;86 (6Suppl): Abstract nr B034.