Abstract 3673: Identifying druggable target in medulloblastoma

Abstract Background: Approximately 15, 000 children in the United States are diagnosed with cancer annually. While pediatric cancers boast a 10-year survival rate of approximately 80%, they present unique challenges in drug discovery due to the low frequency of recurrent mutations and limited therapeutic targets. Brain cancers in children, including medulloblastoma, remain the leading cause of cancer-related deaths in this age group. Certain subtypes, such as group 3 medulloblastoma, are still associated with poor outcomes despite advances in therapy. Current treatment strategies, while effective at increasing survival, frequently lead to severe long-term toxicities, emphasizing the critical need to identify drug targets that can reduce toxicity and enhance outcomes. Method: Our lab has previously demonstrated that AI-based models can successfully facilitate drug discovery in common adult cancers by identifying and prioritizing key druggable molecular vulnerabilities (Zhu Y et al, Probabilistic graph-based model uncovers previously unseen druggable vulnerabilities in major solid cancers, bioRxiv 2024.06.04). Building on this success, we applied this approach to medulloblastoma by analyzing 491 whole-genome sequencing and 167 RNA sequencing to identify potential druggable targets, and systemically evaluated the predictions. Results: A total of 120 genes were identified as significantly recurrently altered in the cohort, including TERT, PTCH1, DDX3X, KMT2D, CTNNB1, and KMT2C. These genes were used to seed our AI-based model to generate patient network and predict molecular vulnerabilities. Predictions were validated using DepMap medulloblastoma cell lines, confirming 15 of the top 50 ranked genes as dependencies, despite no cell line data being used as input or for model training. Unsupervised clustering of the patient data revealed four distinct molecular clusters consistent with the established medulloblastoma subtypes. Differential expression analysis and elastic net regression were employed to identify representative subtype-specific genes. By combining DNA and RNA data, we constructed subtype-specific molecular networks and identified dependencies unique to each subtype. Conclusions: This study is the first to apply an AI-based model to identify druggable targets in medulloblastoma. By amplifying weak signals within cancer molecular networks, the model successfully prioritized key vulnerabilities and revealed subtype-specific therapeutic opportunities. These findings provide a framework for advancing precision oncology in pediatric brain cancers, with the potential to reduce treatment-related toxicity and improve outcomes for medulloblastoma patients. Citation Format: Dongxue Mao, Li Zhao, Ying Zhu, Chunjie Jiang, Kyle Smith, Paul Northcott, Bissan Al-Lazikani. Identifying druggable target in medulloblastoma. [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 3673.