DDEL-32. An atlas of ex vivo drug sensitivity profiles in 666 clinical glioblastoma samples revealed distinct survival-associated networks

Abstract Ex vivo drug testing enables direct assessment of glioblastoma physiology by evaluating cellular responses to therapeutic agents. In this study, we analyzed the ex vivo drug sensitivity of 666 clinical glioblastoma samples (isocitrate dehydrogenase wild-type) to construct a comprehensive atlas and identify survival-associated network clustering. The study population included samples from 146 patients in the 3D-PREDICT Registry Clinical Study (NCT03561207), which provided both ex vivo drug sensitivity profiles and survival outcomes, and 520 patients who underwent 3D Predict™ Glioma clinical testing (drug sensitivity profiles only), representing study and commercial cohorts, respectively. Eleven drugs were tested, including DNA-damaging agents (lomustine, carboplatin, temozolomide, procarbazine, irinotecan, etoposide) and targeted agents (abemaciclib, dabrafenib, osimertinib, rucaparib, trametinib). The study cohort included 90 newly diagnosed glioblastomas (n-glioblastoma) and 56 recurrent glioblastomas (r-glioblastoma), and their drug sensitivity profiles were largely similar. Both n- and r-glioblastomas were highly refractory to lomustine, with fewer than 4% of samples exhibiting sensitivity, whereas over 90% of samples were sensitive to procarbazine and rucaparib. Temozolomide-sensitive tumors showed broader drug sensitivity compared to temozolomide-resistant tumors. Distinct clustering of drug sensitivity profiles emerged within both the study and commercial cohorts, with temozolomide, irinotecan, and carboplatin forming central nodes in the network. Notably, co-clustering of sensitivity was associated with survival outcomes; for instance, patients whose tumors exhibited co-sensitivity to temozolomide and carboplatin had significantly improved survival compared to those sensitive only to temozolomide (median overall survival 16.4 vs. 11.8 months, p=0.025). Longitudinal sampling revealed dynamic changes in drug sensitivity, reflecting evolutionary tumor biology. This ex vivo drug sensitivity atlas reveals distinct, non-random survival-associated clustering patterns that reflect underlying glioblastoma cellular physiologies and may inform future clinical trial designs.