Discovering effective drug combinations typically requires screening many dose combinations across diverse tumor models. For rare cancers, this exhaustive experimental approach is often infeasible because representative cell lines and high-throughput models are scarce. Patient-derived xenograft (PDX) models more faithfully reflect tumor genomics and biology but are too low-throughput to support broad combination testing. The authors frame these constraints as the motivation for a more efficient, biologically informed strategy that uses limited tissue while preserving clinically relevant context.
To address these limitations the study introduces EXACT (ex vivo assessment of combination therapies), an integrated experimental-computational pipeline. EXACT combines short-term culture of PDX-derived tumor cells in three-dimensional matrices with systematic drug sensitivity assays and transcriptomic profiling after treatment. The central rationale is that measuring gene-expression changes in response to a first-line drug can reveal activated compensatory pathways; computational analysis of those changes can nominate a second agent to target the emergent vulnerability. This approach prioritizes combinations with mechanistic justification rather than relying solely on large combinatorial screens.
The experimental component of EXACT uses PDX-derived cells cultured ex vivo in three-dimensional matrices over multiple days. Culturing in 3D aims to preserve microenvironmental cues that shape in vivo drug responses while requiring only limited amounts of tissue per assay—an important consideration for rare tumors. The authors validated the platform in terms of feasibility for sustained culture, monitoring of drug sensitivity over time, and compatibility with downstream transcriptomic assays. The ex vivo format is positioned as a middle ground between high-throughput cell line screens and low-throughput in vivo PDX studies.
A key element of EXACT is measuring transcriptomic responses to treatment and applying computational analyses to those data. By sequencing RNA following exposure to a primary drug, the pipeline identifies pathways that are upregulated or otherwise altered as compensatory responses. Computational prioritization then predicts which secondary targets, if inhibited, would exploit the induced vulnerability and potentially produce synergistic or enhanced therapeutic effects. The method therefore links phenotype (drug sensitivity) with mechanism (transcriptomic adaptation) to inform combination selection.
The authors applied EXACT to patient-derived xenograft models of malignant peripheral nerve sheath tumors (MPNST), a rare and clinically challenging malignancy. Using the ex vivo 3D culture platform, they tracked drug responses over days and obtained transcriptomic profiles after single-agent treatments. Computational interrogation of these treatment-induced transcriptomes identified candidate compensatory pathways and matched potential secondary drugs to those pathways. The work demonstrates the platform’s capacity to generate actionable combination hypotheses from limited PDX material.
Using the EXACT pipeline the team prioritized a combination of a MEK inhibitor plus an HDAC inhibitor. This combination exhibited enhanced activity in both in vitro ex vivo assays and in vivo experiments described by the authors. The reported preclinical efficacy of this combination informed the initiation of an active clinical trial. The source reports the combination selection and its translational progression but does not list specific proprietary drug names or dosing regimens in the summary; details and experimental parameters are available in the article’s methods and supplementary materials.
The authors emphasize that EXACT is scalable and adaptable for near–real-time use with primary patient specimens. Because the ex vivo PDX-derived culture requires limited tissue and couples directly to transcriptomic readouts, the pipeline could be used to generate personalized therapeutic hypotheses for patients with rare tumors where conventional model systems are lacking. EXACT therefore has potential as a platform for prioritized, biologically rational combination discovery and for supporting trial design in rare cancer settings.
The authors made associated data and computational code available through public repositories referenced in the source. Competing interests disclosed include industry affiliations and consulting relationships for some investigators. Funding sources reported include an American Cancer Society Research Professor Award, the CDMRP/Neurofibromatosis Research Program (NFRP) grant, and support from the Gilbert Family Foundation. The source indicates that all relevant experimental details, datasets, and analytical pipelines are described in the full preprint and supplementary materials.