Cancer researchers cannot study a cancer they have no laboratory model of, and for many tumor types the available models have been decades old, few in number, or unrepresentative of the patients who actually get the disease.
The National Cancer Institute has now published results from a ten-year effort that roughly doubles the public supply. The Human Cancer Models Initiative released 665 next-generation laboratory models representing 25 types of cancer, derived from 2,780 donors. The compendium describing them was published in Nature.
Keith Ligon of Dana-Farber Cancer Institute, a co-senior author, said the effort doubles the number of in vitro models available for use and includes very rare cancer types for which only one or two prior models existed in the entire scientific community. The resource is public and free to qualified researchers. That is the part with consequences, because a model held by one institution advances one laboratory's work, while a model distributed to any qualified researcher advances the field.
What a Patient-Derived Model Is
The term covers several things, and the differences matter for what a laboratory can learn.
Traditional cancer cell lines are cells grown flat on plastic, often for decades. Many were established in the mid-twentieth century, have drifted genetically from the tumors they came from, and grow as two-dimensional sheets that behave differently from tissue.
The new collection is a mix of three-dimensional organoids and two-dimensional cell lines. Organoids are three-dimensional clusters grown from a patient's own tissue that self-organize and retain much more of the original tumor's architecture and behavior.
The resource goes beyond the living cultures. Many models are linked to matched tissue from the original tumor, detailed clinical information including outcomes and treatment history, and molecular data including DNA and RNA sequencing, transcriptomic profiles, and epigenetic measurements. Of the 665 models, 522 include comprehensive clinical data. Scientists at the Van Andel Institute led the DNA methylation analysis, which assessed how closely the models reflect the epigenetic state of the tumors they came from. That uniformity is a quiet but significant feature: models generated by different laboratories using different methods are difficult to compare, while a standardized set can be used across studies.
Rare Cancers and Underrepresented Patients Gain the Most
The practical benefit is concentrated where the models were created.
Of the 665 models, 153 represent rare cancer types, roughly 23 percent. That share is disproportionate to how research funding usually flows, and it addresses a specific problem. Rare cancers are difficult to study precisely because too few patients exist to support large clinical trials, which makes reliable laboratory models more important rather than less.
A second figure deserves equal attention: 71 models are derived from people of non-European ancestry. Cancer models have historically skewed heavily toward one population, which limits what can be said about how tumors behave and how they respond to drugs in everyone else. The collection also includes pediatric cancers.
The malignancies covered include pancreatic, breast, endometrial, colorectal, bladder, ovarian, head and neck, and lung cancers, among others.
How It Was Built and Who Built It
The scale required an arrangement no single funder could manage.
The initiative was a large-scale collaboration among the National Cancer Institute, Cancer Research UK, the Wellcome Sanger Institute, and the Hubrecht Organoid Technology Foundation. Contributing institutions span the United States, the United Kingdom, Italy and the Netherlands. Cold Spring Harbor Laboratory and Northwell Health led one of the largest teams and contributed more than 150 models, with Hans Clevers of the Hubrecht Institute, a pioneer of organoid development, also participating.
Vincenzo Corbo, whose team at the University of Verona generated and distributed more than 70 pancreatic and colorectal models, said seeing external researchers actively use the distributed models is the ultimate validation of the project's impact.
Approximately 2,800 patients consented to contribute tissue and data about their cancer biology and treatment. Every model in the collection exists because a patient agreed to donate tissue, usually during treatment for a serious illness, with no personal benefit to themselves from doing so.
The initiative's original target was about 1,000 models, so the work is not finished.
What It Means for Patients, Honestly
The benefit here is real and indirect, and overstating it would misrepresent what infrastructure does.
No patient will be treated differently tomorrow because of this release. It is a research resource, not a therapy or a diagnostic. Its value is that experiments which previously could not be run at all now can be, particularly for cancers where the practical alternative was no usable model whatsoever.
The realistic timeline for benefit is long. A drug candidate identified using these models still faces years of preclinical development and clinical testing, with most candidates failing. What the resource changes is the quality and representativeness of the starting point, which is where a great deal of translational failure originates.
For patients, the connected and genuinely actionable idea here is tissue donation. Research biobanking programs at cancer centers ask patients to consent to donation of tissue that would otherwise be discarded after a biopsy or surgery. It carries no additional procedure in most cases, and it is what makes collections like this one possible. Patients interested can ask their oncology team whether a biobanking or research consent program exists at their institution.
The models themselves are distributed to qualified researchers through an established repository rather than by request from individual patients or families, which is standard for resources of this kind.
MedicalDaily will report findings that emerge from the collection as they are published.
Frequently Asked Questions
What was released? 665 patient-derived cancer models representing 25 cancer types, from 2,780 donors, described in a Nature compendium.
What is a patient-derived model? Living cells grown from a patient's tumor, including three-dimensional organoids that retain more of the original tumor's behavior than traditional flat cell lines.
Why does the public supply matter? Models held privately advance one laboratory. Publicly distributed models advance the whole field.
How many are rare cancers? 153 of the 665, alongside 71 derived from people of non-European ancestry and 522 with comprehensive clinical data.
Who built it? The NCI with Cancer Research UK, the Wellcome Sanger Institute and Hubrecht Organoid Technology, plus contributing institutions in four countries.
Does this change my treatment? No. It is a research resource, not a therapy or diagnostic test.
Can I contribute tissue? Ask your oncology team whether your center has a research biobanking or tissue donation consent program.
