Early oncology drug discovery is defined by uncertainty. Tumors are heterogeneous, evolve under selective pressure, and interact dynamically with their microenvironment. Against this complexity, researchers must make high stakes decisions about targets, modalities, dosing strategies, and indications long before clinical data are available. Increasingly, the success of these decisions depends on the biological relevance of the preclinical models used to generate them.
Patient-derived model systems have emerged as central tools in this decision-making process. In particular, patient-derived xenografts (PDXs) and PDX-derived organoids (PDXOs) offer complementary insights into therapeutic behavior when thoughtfully applied. Understanding how, when, and why to deploy these models is a critical scientific consideration in early oncology programs.
Key scientific considerations in early oncology drug discovery
At the earliest stages of discovery, researchers seek to answer three foundational questions:
- Is the biology relevant to human disease?
- Does the therapeutic engage its intended mechanism in a patient-like context?
- Can early signals of efficacy, resistance, or toxicity be detected before clinical testing?
Traditional cell lines, while valuable for mechanistic studies, often fail to recapitulate the genetic diversity, architecture, and evolutionary pressures present in patient tumors. As a result, they may overestimate efficacy or miss resistance mechanisms that later emerge in the clinic. Patient-derived models attempt to bridge this translational gap by preserving key features of the original tumor, enabling more informed early decisions.
Choosing the right patient-derived models
PDXs are established by engrafting fresh patient tumor tissue into immunocompromised mice. These models retain the histopathology, genomic landscape, and intra-tumoral heterogeneity of the donor tumor across passages. Numerous studies have demonstrated concordance between patient responses and PDX treatment outcomes, supporting their use as translationally relevant efficacy models.
However, PDX studies are resource intensive and low throughput, which can limit their use in early-stage screening or rapid hypothesis testing. This has driven growing interest in PDXOs, three-dimensional cultures generated directly from PDX tissue that preserve tumor-intrinsic characteristics while enabling scalable experimentation.
Critically, comparative analyses have shown biological equivalence between PDXs and their matched PDXOs across multiple dimensions, including morphology, gene expression, mutational profiles, and drug response patterns. When derived and maintained appropriately, PDXOs mirror the therapeutic sensitivities observed in vivo, making them powerful tools for early decision-making.
Rather than competing systems, PDXs and PDXOs represent different points along a continuum of biological complexity and experimental control.
Measuring efficacy, mechanism, and risk
Model selection should be guided by the specific scientific question at hand:
- Therapeutic activity and ranking: PDXOs enable medium- to high-throughput testing across diverse genetic backgrounds, supporting early compound prioritization and structure–activity relationship studies.
- Mechanism and resistance: Organoid systems allow controlled perturbation and longitudinal analysis to uncover pathway dependencies or adaptive resistance mechanisms.
- In vivo validation: PDXs remain essential for assessing pharmacokinetics, pharmacodynamics, tumor growth inhibition, and systemic tolerability in a whole-organism context.
Importantly, correlations between patient tumors, matched PDXs, and PDXOs have shown that drug responses observed in organoids frequently align with both in vivo PDX outcomes and clinical behavior. This triangulation strengthens confidence in early signals and reduces reliance on any single experimental system.
From a safety perspective, while neither PDXs nor PDXOs fully model immune-mediated toxicities, their use can still inform therapeutic index by revealing on-target effects across molecularly defined tumor subtypes.
Early model decisions shape clinical outcomes
Decisions made during discovery shape the entire development trajectory. Selecting models that inadequately represent patient biology can lead to false positives, advancing compounds that ultimately fail, or false negatives that prematurely eliminate promising candidates.
Integrating PDX and PDXO data early in the discovery process allows researchers to move beyond average responses and interrogate heterogeneity at the level that matters most: the patient. By revealing which tumor subsets are genuinely responsive or resistant, these models support biomarker hypotheses grounded in human-derived biology rather than surrogate systems. Early exposure to resistance mechanisms also provides an opportunity to refine therapeutic strategies before clinical testing, rather than reacting to failure in the clinic.
Crucially, these insights inform indication selection and trial design, helping teams align preclinical evidence with realistic clinical hypotheses. The result is a tighter connection between discovery biology and development strategy, reducing uncertainty as programs advance toward first-in-human studies.
New directions in oncology model development
Oncology model development is increasingly shifting toward patient-centric and integrative strategies that reflect the biological complexity seen in the clinic. Growing PDX and PDXO libraries now capture greater tumor diversity, including rare indications, treatment-refractory disease, and more representative patient populations. At the same time, researchers are adopting multi-model workflows that deliberately combine in vitro, ex vivo, and in vivo systems to interrogate complementary aspects of tumor biology rather than relying on any single model in isolation.
These approaches are further strengthened by advances in data integration, where functional responses observed in patient-derived models are analyzed alongside genomic, transcriptomic, and spatial data to support systems-level decision making. These capabilities are enabling more personalized research strategies, using patient-derived models not only to assess drug activity but also to explore individualized therapeutic hypotheses and precision medicine applications. As a result, the field is moving away from model selection as a logistical consideration and toward model strategy as a scientific discipline in its own right.
Early oncology discovery increasingly demands models that can inform complex biological and translational decisions with greater fidelity. PDX and PDXO, when used in concert, provide a powerful framework for understanding therapeutic behavior in clinically relevant contexts. By aligning model choice with scientific intent, researchers can reduce uncertainty, improve decision quality, and ultimately increase the likelihood that early discoveries translate into meaningful clinical benefit.
