MDB-66. High-Throughput Small Molecule Drug Screening, Transcriptomics and Integrative AI Identify Targetable Myc-Dependent Therapeutic Vulnerabilities in High-Risk Medulloblastoma

Abstract INTRODUCTION MYC-amplified Group 3 medulloblastomas (MBGroup3-MYC) do not respond to conventional up-front therapies and are almost universally fatal (5% survival). MYC is not directly targetable. Thus, the systematic development of therapies which target complementary MYC-dependencies offers clear potential to rationally re-design effective therapeutic strategies. METHODS Isogenic cell-based models with regulable shRNA MYC-knockdown were developed in three independent MYC-amplified backgrounds (D283Med; D425Med; HD-MB03). These were utilised to i) define the MYC-dependent transcriptome/proteome in MBGroup3, through analysis in conjunction with our primary tumour cohort, and ii) as the basis of an automated MYC-dependent high-throughput drug screen (HTS; 500 clinically-actionable compounds). Bespoke integrative AI tools and bioinformatic analysis, and subsequent siRNA/pharmacodynamic validation, identified and validated candidate MYC-dependent therapeutic vulnerabilities. Prioritised drugs were advanced to pre-clinical trials across two in vivo MBGroup3 models (GTML/Trp53KI/KI (GEMM; MYCN-driven) and GMYC (GEMM-derived allografts; MYC-driven)) to evaluate candidate efficacy and toxicity. RESULTS shRNA-mediated MYC knockdown could be maintained for 28 days without escape in all models and, as anticipated, was associated with cell cycle abrogation and elevated caspase-3/-7 activity. HTS across all models revealed that MYC expression confers differential dependencies to small molecule inhibitors, with 82 compounds spanning 11 major drug classes demonstrating MYC-dependent activity. AI-based integration of HTS data with model and primary patient MYC-dependent transcriptome data shortlisted multiple druggable molecular targets which, in turn, were independently validated. To date, we have progressed alisertib (AURKAi), volasertib (PLK1i) and prexasertib (CHK1i) to in vivo pre-clinical trials. Promisingly, both alisertib and prexasertib demonstrated efficacy within MBGroup3-MYC models. CONCLUSIONS Integrated human tumour and model-based screening approaches have enabled identification and validation of critical MYC co-dependencies, targetable using established small-molecules, to inhibit MBGroup3 tumour growth. Their further evaluation, including integration into established and/or rationally-designed combinations, will be essential next steps in their (pre-)clinical development.