Integrated ex vivo screening and transcriptomic profiling to prioritize drug combinations for rare cancers

Discovering effective drug combinations requires testing many dose combinations across a diverse panel of tumor models. This approach is limited in rare cancers by a scarcity of cell lines and representative high-throughput models that would make exhaustive screening feasible and predictive. Patient-derived xenograft (PDX) models are genomically representative but too low-throughput for this purpose. Culturing PDX cells ex vivo in three-dimensional (3D) matrices offers a genomically representative and clinically relevant platform for preclinical drug testing, capturing the microenvironmental cues that shape in vivo drug response while requiring only limited tissue per assay. Toward this end, we designed and validated an experimental-computational framework, "ex vivo assessment of combination therapies" (EXACT), to enable drug combination discovery in rare tumors. Using PDX models of malignant peripheral nerve sheath tumors (MPNST), we built a platform to culture PDX cells ex vivo over multiple days, monitoring drug sensitivity and measuring transcriptomic responses to treatment. Computational analysis of these transcriptomic responses then identifies which compensatory pathway creates a unique vulnerability to a second drug. EXACT thus offers a biologically informed, scalable approach for prioritizing drug combinations in rare tumors, nominating drugs alongside biological rationale. Using this methodology, we identified a MEK inhibitor plus HDAC inhibitor combination with enhanced activity in vitro and in vivo, forming the basis of an active clinical trial. This platform could be adapted for real-time use with primary patient specimens, enabling personalized therapeutic discovery. SignificanceEXACT integrates PDX-derived 3D drug screening with biologically informed computational analysis to identify and explain effective combinations, providing a scalable strategy for therapeutic discovery in rare cancers such as MPNST.