A cancer model is any experimental system designed to mimic some aspect of cancer so researchers can study the disease outside a living patient. These systems range from tumor cells growing in a plastic dish to mice carrying human tumors, three-dimensional miniature organs grown in the lab, and even computer simulations that predict how a tumor will expand and respond to drugs. No single model captures everything about cancer, which is precisely why dozens of different types exist and why choosing the right one for a given question matters enormously for whether a finding ever helps a real person.
Cell Lines and Where Cancer Modeling Began
The story of cancer models starts with cells in a dish. In 1951, researchers took cervical cancer cells from a patient named Henrietta Lacks and, for the first time, managed to keep human-derived cells alive and dividing indefinitely in the lab.1PubMed Central. The haplotype-resolved genome and epigenome of the aneuploid HeLa cancer cell line Those HeLa cells became the foundation of modern cell-line research and are still used today. Since then, hundreds of cancer cell lines have been established from virtually every tumor type. They grow quickly, are cheap to maintain, and let scientists run thousands of experiments in parallel.
The catch is that cell lines drift genetically over time, and different labs working with supposedly the same line can end up studying meaningfully different cells. One analysis of a well-known breast cancer cell line, MCF7, found that only about a third of coding mutations were shared across all substrains. Drug responses diverged just as sharply: among compounds that strongly inhibited growth in at least one substrain, most showed minimal effect on another substrain of the same line.2Disease Models & Mechanisms. The secret lives of cancer cell lines That means two labs could test the same drug on what they believe is the same cell line and reach opposite conclusions. Cell lines remain useful for rapid screening, but researchers now treat results from them as a starting point rather than proof that a drug will work in patients.
Organoids and Three-Dimensional Culture
A major step beyond flat cell-line cultures is the organoid, a tiny three-dimensional structure grown from patient tumor cells that self-organizes into something resembling the original tissue. Unlike traditional cell lines that have been passaged for decades, organoids are typically derived directly from a patient’s biopsy or surgical sample and can faithfully recapitulate the molecular features of the original tumor, including its internal diversity.3PubMed Central. Patient derived organoids in prostate cancer: improving therapeutic efficacy in precision medicine This makes them far more representative of a real tumor than a cell line that has been growing in a lab for years.
Organoids sit in a sweet spot between simplicity and complexity. They are more biologically realistic than a flat layer of cells, yet far faster and cheaper to produce than an animal model. Researchers can grow panels of organoids from many different patients, expose each to a library of drugs, and identify which patients’ tumors respond to which treatments. That capacity has made organoids a central tool in precision oncology, where the goal is to match a specific patient to the therapy most likely to work for their individual tumor.
Animal Models and Their Varieties
When researchers need to study cancer in a living body, with blood supply, immune cells, and organs interacting, they typically turn to mice. But “mouse model” covers several very different approaches, each with its own strengths.
- Patient-derived xenografts (PDX): A piece of a patient’s tumor is surgically implanted into an immunodeficient mouse. The tumor grows in the animal while retaining much of the biology and internal diversity of the original cancer. PDX models are widely used for drug development, co-clinical trials, and personalized medicine because they closely represent the tumor a specific patient is fighting.4PubMed Central. Patient-derived xenograft models, applications and challenges in cancer research
- Genetically engineered mouse models (GEMMs): Instead of transplanting a human tumor, scientists alter the mouse’s own genes so that cancer arises spontaneously. These models are especially valuable for studying how the immune system interacts with a developing tumor, because the cancer grows in a host with a fully functional immune system.5PubMed Central. Genetically engineered mouse models of cancer reveal new insights about the antitumor immune response
- Syngeneic models: Mouse tumor cells are implanted into mice of the same genetic strain, again preserving a working immune system. These are the workhorse of immunotherapy research, because they let scientists test checkpoint inhibitors and other immune-based drugs in a host whose immune cells can actually attack the tumor.6PubMed Central. CD4+ and CD8a+ PET imaging predicts response to novel PD-1 checkpoint inhibitor: studies of Sym021 in syngeneic mouse cancer models
PDX models give you human tumor biology but sacrifice the immune system, since the host mouse must be immunodeficient to avoid rejecting the foreign tissue. GEMMs and syngeneic models keep the immune system intact but use mouse tumors, which are not identical to human cancers. Researchers pick the model that best suits the question they are trying to answer, often using more than one type across the course of a project.
Modeling the Tumor Microenvironment
A tumor is not just a ball of cancer cells. It is embedded in a complex neighborhood of blood vessels, immune cells, connective-tissue cells called fibroblasts, and signaling molecules. This surrounding ecosystem, known as the tumor microenvironment, plays a major role in whether a cancer grows, spreads, or resists treatment. Many promising drugs have failed precisely because they were tested in models that ignored this neighborhood.
To address that gap, engineers have developed three-dimensional platforms that reconstruct elements of the microenvironment outside the body. Using biomaterials and microengineering technologies, these systems can incorporate stromal and immune cells alongside tumor cells in a spatially organized way.7PubMed Central. Engineered 3D ex vivo models to recapitulate the complex stromal and immune interactions within the tumor microenvironment One group, for instance, built a pancreatic cancer organoid model that included both cancer-associated fibroblasts and infiltrating immune cells, demonstrating that the fibroblasts activated in response to the tumor and that immune cell behavior changed depending on the tumor’s presence.8PubMed Central. Development of primary human pancreatic cancer organoids, matched stromal and immune cells and 3D tumor microenvironment models
A newer evolution of this idea is the tumor-microenvironment-on-a-chip, which combines three-dimensional cell culture with tiny fluid channels that simulate blood flow and nutrient delivery. These devices let researchers observe how tumor cells interact with their surroundings under conditions that more closely mirror what happens inside the body.9PubMed Central. Tumor-microenvironment-on-a-chip: the construction and application They are still largely research tools rather than standard clinical instruments, but they represent a significant leap toward models that capture the full complexity of a living tumor.
Computational and Mathematical Models
Not all cancer models involve living cells or animals. Computational models use mathematics and simulation to predict how tumors grow, how blood vessels form around them, and how drugs travel through tissue. These in silico models can test thousands of treatment scenarios in hours rather than the months required by lab experiments.
One multiscale mathematical model of the tumor microenvironment simulated both tumor growth and the development of new blood vessels, then tested various drug regimens on the virtual tumor. The results suggested that frequent, low-dose drug schedules could normalize tumor blood vessels, improve drug delivery, and reduce cancer cell invasion more effectively than conventional high-dose approaches.10PubMed Central. Simulations of tumor growth and treatment response: Benefits of high-frequency, low-dose drug regimens and concurrent vascular normalization Another approach uses patient imaging data to build organ-scale models personalized to an individual’s anatomy. When applied to solid tumor growth, a data-driven simulation method achieved short-to-medium-term prediction errors under one percent and long-term errors under twenty percent, even with very short training periods.11PubMed. Data-Driven Simulation of Fisher-Kolmogorov Tumor Growth Models Using Dynamic Mode Decomposition
Other computational models focus on the physical forces inside a tumor, such as how fluid pressure affects drug distribution. Elevated pressure inside tumors is one reason drugs often fail to penetrate deeply. A model that accounts for vessel network shape, tissue permeability, and the balance between diffusion and convection can help predict where a drug will actually reach and where it will not.12PLOS ONE. Interstitial Fluid Flow and Drug Delivery in Vascularized Tumors: A Computational Model Some researchers also feed micro-imaging data from real tumors into lattice-based models, using machine learning to govern cell behavior and microvascular growth, creating simulations that start from real anatomy rather than theoretical assumptions.13Scientific Reports. Multiscale computational modeling of cancer growth using features derived from microCT images
Why Most Cancer Drugs Still Fail in Humans
Despite all these model types, the track record for translating preclinical success into clinical benefit remains sobering. Over ninety percent of oncology drug candidates that succeed in animal studies go on to fail in human clinical trials.14PubMed Central. Integrating New Approach Methodologies (NAMs) into Preclinical Regulatory Evaluation of Oncology Drugs That gap exists for several compounding reasons. Mice are not tiny humans: differences in body size, metabolism, organ physiology, and the molecular targets themselves all introduce translational uncertainty.15Journal of Nuclear Medicine. Of Mice and Humans: Are They the Same?—Implications in Cancer Translational Research Cell lines, as discussed, accumulate mutations that make them less representative of any real patient’s tumor. Even PDX models, which retain much of the original tumor’s architecture, lose the human immune system and grow in a mouse stroma that may respond differently than human tissue would.
Animal models are also fundamentally limited in their ability to capture the full complexity of human carcinogenesis and disease progression.16PubMed Central. Lost in translation: animal models and clinical trials in cancer treatment A drug that shrinks a mouse tumor growing under the skin might have no effect on the same cancer type growing in a human organ, where the local blood supply, immune landscape, and mechanical environment are all different. Recognizing these limitations is what drives the push toward newer platforms such as organoids, organ-on-chip systems, and computational models, often grouped under the umbrella term “new approach methodologies” or NAMs.
Personalized Cancer Avatars
One of the most promising uses of cancer models is building what researchers call a “cancer avatar” for an individual patient. The idea is straightforward: take tumor cells from a patient, grow them into organoids or implant them into mice, then test a panel of drugs on those models to identify which treatment the patient’s cancer is most likely to respond to before giving the drug to the patient.
In brain cancer, researchers have shown that patient-derived organoids and orthotopic xenografts retain the histological, genetic, and molecular features of the parent tumor, making them clinically relevant avatars for precision oncology in gliomas.17PubMed Central. Patient-derived organoids and orthotopic xenografts of primary and recurrent gliomas represent relevant patient avatars for precision oncology A pilot study in patients with rare thymic cancers took circulating tumor cells from blood draws, grew them into organoids, and tested drug sensitivity to see whether the lab results correlated with how patients actually responded to treatment.18PubMed Central. Personalized cancer avatars for patients with thymic malignancies: A pilot study with circulating tumor cell-derived organoids This kind of work is still early, but it points toward a future where treatment decisions are guided by testing drugs on a patient’s own cells rather than relying solely on population-level statistics.
A practical challenge is speed. Growing a PDX mouse model can take months, often longer than a patient can afford to wait before starting treatment. Organoids are faster, typically ready in weeks, which makes them a more realistic option for real-time clinical decision-making. The tradeoff is that organoids lack the full in vivo environment that a PDX model provides.
Modeling Dormancy and Cancer Recurrence
Some of the most dangerous aspects of cancer happen when the disease appears to be gone. After successful treatment, tiny populations of surviving cancer cells can enter a dormant state, lying quiet and undetectable for months to years before suddenly reactivating and producing a relapse. Understanding and modeling this dormancy is crucial because it is the mechanism behind most cancer deaths from solid tumors.
Mouse models have demonstrated that residual tumor cells surviving targeted therapy can persist in a state of dormancy, with adequate blood supply and no immune pressure against them, and still retain the ability to re-enter the growth cycle and produce recurrent tumors after extended quiet periods.19PubMed Central. Cellular dormancy in minimal residual disease following targeted therapy Other mouse models specifically recapitulate the clinical pattern seen in breast cancer, where patients can remain in apparent remission for years or even decades before a relapse occurs.20Endocrinology. Exploiting Mouse Models to Recapitulate Clinical Tumor Dormancy and Recurrence in Breast Cancer These models are critical because dormant cells cannot be studied in patients. You cannot biopsy what you cannot detect. Animal models provide the only window into what these cells are doing while they wait.
Cancer Stem Cells and Drug Resistance
Closely related to dormancy is the problem of drug resistance, and one leading explanation for why many cancers come back involves a subpopulation called cancer stem cells. These cells, which typically make up a small fraction of a tumor, can self-renew and regenerate the full diversity of the original cancer.21PubMed Central. Cancer stem cells: Role in tumor growth, recurrence, metastasis, and treatment resistance They tend to be more resistant to chemotherapy and radiation than ordinary cancer cells, meaning they can survive treatment and seed a new tumor.
Understanding the mechanisms behind this resistance is an active area of cancer-model research. Various model systems, from cell-line assays to organoids to animal models, are used to identify the molecular pathways that allow cancer stem cells to evade therapy and to test drugs that might specifically target those cells.22PubMed Central. Cancer Stem Cells (CSCs) in Drug Resistance and their Therapeutic Implications in Cancer Treatment If a new drug kills the bulk of a tumor but leaves cancer stem cells intact, the model should reveal that failure before the drug reaches a patient.
Models for Metastasis
Cancer kills most often when it spreads to distant organs, yet modeling metastasis is one of the hardest challenges in the field. A tumor that breaks off and travels to bone, for example, encounters a completely different tissue environment than the organ where it started. Animal models of bone metastasis are considered essential for understanding how cancers colonize skeletal tissue and for developing therapies to prevent or treat those metastases.23PubMed Central. Animal Models of Bone Metastasis Different approaches exist: injecting cancer cells directly into the bloodstream to simulate the spread, implanting them at the original organ site and waiting for spontaneous metastasis, or using genetically engineered models where metastasis arises naturally as the disease progresses.
Each approach captures a different stage of the metastatic process. Direct injection skips the early steps of invasion and entry into the bloodstream, giving researchers a clean look at how cancer cells colonize a distant organ. Spontaneous metastasis models preserve the entire cascade but take longer and are less predictable. The ideal model depends on which part of the metastatic journey the researcher wants to study.
Comparative Oncology and Dogs as Cancer Models
An underappreciated corner of cancer modeling involves pet dogs. Dogs develop many of the same cancer types humans do, including osteosarcoma, lymphoma, melanoma, and bladder cancer, and their tumors arise spontaneously rather than being artificially induced. Because dogs share living environments with their owners and have functioning immune systems, their cancers can provide translational insights that laboratory mouse tumors cannot. The value of canine cancer models has grown substantially, with studies analyzing genetic and clinical similarities between human and dog tumors.24PubMed Central. Comparative oncology: overcoming human cancer through companion animal studies
From a practical standpoint, veterinary clinical trials can serve as an intermediate step between mouse studies and human trials. Dogs with naturally occurring cancer are treated with experimental drugs, providing safety and efficacy data in a large, outbred species whose tumors behave much like human cancers. The dogs benefit from access to cutting-edge treatment, and the data generated help refine the drug before human testing begins.
Living Biobanks and Sharing Resources
Building a cancer model from scratch for every experiment is slow and expensive. That reality has driven the creation of living biobanks: repositories that collect, grow, and store viable tumor organoids and tissues so researchers worldwide can access well-characterized models without starting from a biopsy. These biobanks store organoids from different cancer types, genetic backgrounds, and disease subtypes, creating a library that supports both basic research and drug screening efforts.25PubMed Central. Living biobank-based cancer organoids: prospects and challenges in cancer research
The long-term vision is ambitious. If a biobank holds organoids representing a wide range of genetic profiles for a given cancer type, researchers can screen a new drug against that entire panel and identify which patients, defined by molecular subtype, are most likely to benefit. It turns personalized medicine from a one-patient-at-a-time endeavor into something scalable.
Non-Mammalian Models
Not every cancer model involves a mammal. Zebrafish, for example, are increasingly used because their larvae are transparent, allowing researchers to watch tumor cells invade tissue and recruit new blood vessels in real time under a microscope. Human cancer cells can be injected into zebrafish larvae and observed as they interact with the host vasculature, offering a rapid, visual assay for tumor angiogenesis and drug response.26npj precision oncology. Refined high-content imaging-based phenotypic drug screening in zebrafish xenografts These models are far cheaper and faster than mouse experiments, making them useful for large-scale drug screening where hundreds of compounds need a first pass.
Other non-mammalian organisms used in cancer research include fruit flies, which share many cancer-relevant genetic pathways with humans, and the nematode worm C. elegans, which has been instrumental in understanding basic cell-division and cell-death mechanisms. The 3R principle, which stands for replacement, reduction, and refinement of animal use, has accelerated interest in these alternative organisms as a way to conduct ethical cancer research without relying so heavily on mice.27PubMed Central. Beyond the mouse: 3R-guided alternative animal models transforming cancer research
How Models Fit Together in Drug Development
In practice, cancer models are not used in isolation. A typical drug development pipeline starts broad and narrows. Early screening might involve testing tens of thousands of compounds on cell lines or computational models to identify candidates worth pursuing. Promising compounds then move to organoid panels or organ-on-chip systems to see if they work against more realistic tumor tissue. Compounds that pass those filters advance to animal models, often starting with syngeneic or xenograft mice, for safety and efficacy testing in a whole organism. Only after clearing these stages does a drug enter human clinical trials.
Each model type serves as a filter, catching different kinds of failure. Cell lines catch compounds that have no anti-cancer activity at all. Organoids catch drugs that kill cancer cells in a monolayer but fail against three-dimensional tumor architecture. Animal models catch drugs that are toxic in a living body or that fail to reach the tumor through the bloodstream. Computational models can run in parallel at any stage, predicting optimal dosing schedules or identifying patients whose tumors have the molecular features needed for a drug to work. The ongoing challenge, and the reason the field keeps inventing new model types, is that all of these filters together still miss a large share of the drugs that will ultimately fail in people.