Abstract A031: A single-cell standard-of-care atlas of glioblastoma

Abstract Glioblastoma (GBM) remains the most lethal brain tumor in adults. Despite aggressive standard-of-care (SOC) treatment - combining surgical resection, radiation therapy (RT), and temozolomide (TMZ) chemotherapy, and often lomustine in recurrent cases - median survival remains <20 months. In spite of numerous clinical trials, no major advances were obtained in the past 20 years. Three major challenges limit treatment success: (1) its infiltrative nature makes complete surgical removal mostly impossible without harming healthy tissue; (2) RT and TMZ often fail, with >50% of patients experiencing disease progression even during therapy, revealing the presence of intrinsically treatment-resistant cells; and (3) its high inter- and intratumoral heterogeneity, driven by complex genomic alterations and the presence of a multitude of single-cell states, in addition to significant plasticity to adapt to new therapeutic perturbations. Despite these obstacles, ∼5% of patients show an exceptional response to SOC, and similar responsive populations are also found when treating with experimental treatments. This suggests that treatment inefficacy may be due to imprecise patient selection rather than treatment failure. To solve this issue, the field is moving towards functional precision oncology (FPO), which involves the testing of live tumor samples against various therapeutic options outside the body (“ex vivo”) to observe functional responses. While current FPO approaches mostly measure cytotoxicity, they lack the detail needed to capture non-lethal molecular changes or distinguish between sensitive and resistant single cells. Comprehensive single-cell therapeutic atlases integrating cytotoxicity with omics data could fill this gap, but such resources are still unavailable for GBM and other diseases. In this study, we created a single-cell therapeutic atlas for glioblastoma, integrating cytotoxicity assays with single-cell RNA-seq to uncover therapeutic signatures driving diverse responses to SOC therapies - RT, TMZ, combined RT and TMZ, and lomustine. This atlas includes >60 patient-derived glioblastoma cell lines (PDCLs) from which >280 on-treatment samples were profiled at single-cell resolution, resulting in >1M of single-cell profiles. First, by comparing PDCL profiles with a large, curated single-cell atlas of GBM patient samples (GBMap), we confirm that PDCLs accurately reflect the phenotypic and functional diversity seen in primary tumor tissue. Next, we applied a semi-supervised machine learning approach to identify therapeutic metaprograms across PDCLs and treatments, in addition to exploring how gene regulatory networks shift in response to treatment while inferring therapeutic trajectories. Finally, we used Random Forest models to integrate genetic background, cytotoxic responses, metaprograms, and patient outcomes. Together, these findings reveal robust therapeutic molecular signatures in PDCLs after SOC treatment, offering predictive insights into patient responses. Citation Format: Yanti De Visser, Carmen Bravo González-Blas, Lou Van Eupen, Marleen Derweduwe, Annelies Claeys, Frederik De Smet. A single-cell standard-of-care atlas of glioblastoma [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Functional and Genomic Precision Medicine in Cancer: Different Perspectives, Common Goals; 2025 Mar 11-13; Boston, MA. Philadelphia (PA): AACR; Cancer Res 2025;85(5 Suppl):Abstract nr A031.