An international team of scientists, funded by the US National Institutes of Health (NIH), has released a new collection of 665 laboratory cancer models, covering 25 different cancer types. These models were grown from tumor tissue donated by 2,780 patients, as part of a decade-long research program called the Human Cancer Models Initiative (HCMI). The collection is now available to researchers worldwide.
Why this matters
Cancer models, living tumor tissue grown in a lab, are one of the main tools scientists use to study how cancer works. They let researchers observe how a tumor grows, test how it responds to different drugs, and see how it might eventually develop resistance to treatment, all without directly experimenting on patients. This new library gives scientists a much wider set of models to work with, which means more chances to spot patterns, test new drug ideas, and understand why certain cancers are so hard to treat.
Why it’s better than what came before: Most existing cancer model collections have a major weakness: they only represent a small slice of how diverse cancer really is in real patients. Two tumors can carry the same diagnosis on paper, “breast cancer,” say, and still behave completely differently depending on a person’s genetics, cancer subtype, or treatment history. Older model libraries often missed that variety. This new collection was specifically designed to close that gap: it includes rare cancers, brain tumors (historically hard to model), and samples from patients with different genetic backgrounds, not just the “typical” cases that used to dominate research collections.
How they built it, and the numbers behind it
Tumor samples for the Living Library of cancer came from 2,780 patient donors, collected through the Human Cancer Models Initiative, an international effort that has run for over a decade. The final library contains 665 laboratory models across 25 cancer types. It includes 3D organoids, small, lab-grown clusters of cells that mimic the structure of a real organ or tumor far better than flat, single-layer cell cultures, as well as models built from brain tumor cells, known as neurospheres.
153 of the models represent rare cancers, and 71 come from patients of non-European ancestry, both groups historically underrepresented in cancer research. 522 models come with detailed clinical data attached: information about the patient’s treatment history and outcomes, which lets researchers connect a model’s biology to how the real tumor actually behaved in the clinic.
To check how accurately the lab models matched the original tumors, researchers compared 421 matched tumor-model pairs and found 97.8% agreement in genetic alterations, 95% concordance in epigenetic features, and 92% similarity in RNA-expression patterns.
In plain terms, even after growing in a lab dish for a long time, these models still closely resemble the real tumors they came from.
The rest, why it’s exciting, and what it isn’t
Researchers can search the whole collection online by cancer type, prior treatment, and other patient characteristics, through the HCMI’s searchable catalog.
NIH Director Jay Bhattacharya called the release “a major win for cancer research,” noting on X that earlier model collections only captured a fraction of cancer’s biological diversity. Scientists have already started using the models, for example, to study the genetic factors behind glioblastoma’s resistance to chemotherapy.
It’s worth being clear about what this is, and what it isn’t: this is not a new drug or treatment, and nobody’s cancer care changes today. What it is, though, is a shared research tool, a more realistic foundation scientists can use to figure out why some tumors respond to treatment and others don’t, hunt for new drug targets, and test therapies in the lab before they ever reach a clinical trial.
Read more about NIH’s funding opportunities on OncoDaily:
NIH Launches New Funding Opportunity to Advance AI Research Infrastructure