The models include organoids and other laboratory-grown systems engineered to preserve important characteristics of the tumors from which they originated. Organoids are three-dimensional cellular structures that can reproduce aspects of the architecture and cellular composition of an organ. In brain cancer research, for example, HCMI models include neurosphere-like clusters with properties associated with neural stem cells. Unlike many traditional cancer cell lines, which can acquire extensive genetic changes after years of laboratory growth, these models are intended to remain closely connected to the biology of individual patients’ tumors.

That connection is supported by a large body of molecular and clinical information. HCMI has developed models spanning common cancer subtypes and including many of the major genetic alterations that define those diseases. Researchers can associate each model with genomic, transcriptomic, epigenomic, and clinical data, allowing them to examine cancer at several biological levels at once. Genomic data reveal mutations and chromosomal changes, transcriptomic data show which genes are active, and epigenomic data describe chemical and structural mechanisms that regulate gene activity without changing the underlying DNA sequence.

A central question in cancer modeling is whether a tumor retains its defining features after being removed from the body and cultured for extended periods. To address this issue, HCMI scientists compared 421 original tumors with their paired laboratory models. The results showed a striking degree of similarity. The models agreed with their parent tumors on 97.8% of assessed genetic alterations, showed 95% concordance in epigenetic features, and retained 92% similarity in RNA expression patterns. These findings indicate that the culture conditions used to grow the models did not substantially erase the molecular identity of the cancers.

The preservation of these features is particularly important for investigating treatment resistance, one of the most persistent challenges in oncology. A tumor may fail to respond to an initial therapy, or it may adapt during treatment and return in a more aggressive form. HCMI models can help scientists trace the genetic and molecular events behind that process. In glioblastoma models, researchers identified inherited abnormalities, acquired amplification of cancer-promoting genes, and mutational signatures associated with previous exposure to temozolomide, a chemotherapy drug commonly used against the disease. Such signatures are molecular records of DNA damage and repair processes that can reveal how a tumor has responded to therapy in the past.

Because the models retain characteristics of individual tumors, they may also support more precise studies of drug sensitivity. Researchers can expose different models to the same treatment and compare the results with the genetic alterations present in each cancer. This approach may help identify vulnerabilities that are shared across patients, as well as weaknesses restricted to a smaller molecular subgroup. In the longer term, these experiments could improve the design of targeted therapies and help explain why a drug benefits some patients but not others.

The collection also broadens the range of cancers available for laboratory investigation. Of the models, 153 represent rare cancers, which are often difficult to study because patient numbers are low and biological samples are scarce. The resource includes 71 models from people of non-European ancestry, an important step toward addressing the underrepresentation of diverse populations in biomedical research. Although the collection does not capture every aspect of tumor biology—particularly the full influence of a patient’s immune system, blood supply, and surrounding tissues—it provides a more diverse foundation for research than many existing model repositories.

More than 500 models, specifically 522, are accompanied by detailed clinical information. Scientists can search the HCMI catalog according to cancer type, treatment history, demographic characteristics, and other features. The models are being distributed to the broader research community through a partnership with the American Type Culture Collection, allowing laboratories around the world to work with standardized biological material rather than relying on isolated, difficult-to-reproduce samples. The associated data are intended to make experiments more comparable and to accelerate the movement of discoveries from basic research toward preclinical testing.

The HCMI compendium represents the culmination of a decade-long effort involving the National Cancer Institute and collaborators in the United States and internationally, including research groups at the Wellcome Sanger Institute, the Broad Institute of MIT and Harvard, and Cold Spring Harbor Laboratory. Published in Nature, the work establishes a large-scale framework for connecting patient tumors to experimentally tractable models. By combining preserved tumor biology with detailed molecular and clinical records, the initiative gives cancer researchers a powerful way to investigate tumor evolution, uncover mechanisms of resistance, and search for therapeutic vulnerabilities that could ultimately improve precision oncology.

Subject of Research: Patient-derived cancer models, tumor biology, cancer evolution, treatment resistance, organoids, and precision oncology

News Publication Date: 5-Aug-2026

Web References:
https://www.cancer.gov/ccg/research/functional-genomics/hcmi
https://www.atcc.org/hcmi
https://hcmi-searchable-catalog.nci.nih.gov/
https://www.cancer.gov/

References:
Nature article, DOI: 10.1038/s41586-026-10806-y

Keywords: Human Cancer Models Initiative, HCMI, cancer research, patient-derived tumor models, organoids, tumor biology, cancer genomics, epigenomics, transcriptomics, glioblastoma, temozolomide resistance, precision oncology, drug sensitivity, rare cancers, National Cancer Institute, NIH

The study, published August 5, 2026, in Nature, integrates the 3D models into DepMap’s existing collection of more than 1,000 2D cancer models. Researchers found that the two model systems often captured different genetic dependencies, meaning that neither could provide a complete picture of cancer biology on its own. The 3D models also represented cancer subtypes that are difficult or impossible to maintain as conventional cell lines, including certain brain and gastrointestinal tumors. The findings suggest that the next generation of cancer research may depend on using multiple model formats rather than selecting a single “best” system.

Cancer dependencies are genes or molecular pathways that tumor cells require for growth, proliferation, or survival. In principle, disabling such a gene can selectively damage cancer cells while leaving healthy cells less affected. DepMap scientists systematically measure these dependencies by perturbing genes, often with CRISPR-based methods, and then observing whether cancer cells continue to grow. When these results are combined with information about mutations, gene expression, and cellular state, researchers can identify relationships between a tumor’s molecular features and its potential drug vulnerabilities.

Traditional cancer cell lines have been indispensable for this work because they are relatively easy to grow, manipulate, and screen at large scale. However, they are usually adapted to flat plastic surfaces and nutrient-rich laboratory media. During that adaptation, some features of the original tumor can disappear. Organoids and spheroids offer a different environment: cells grow in three dimensions, often within a gel matrix or as suspended clusters, allowing them to establish cell-cell contacts and structural states that more closely resemble those found in patient tumors. Brain tumor models may grow as neurospheres, while many gastrointestinal and pancreatic models are maintained as patient-derived organoids.

The Broad-led team found that 3D models retained mutations and cellular programs observed in human tumors but absent from the available 2D lines. These differences were linked to new genetic dependencies that could not be detected through traditional screening alone. The result is a more diverse dependency landscape, potentially expanding the number of drug targets available for cancers that have been poorly represented in standard laboratory systems. The researchers describe the expanded collection as a step toward a more complete map of the biological requirements that drive different forms of cancer.

One particularly striking example involved glioblastoma, an aggressive and frequently lethal brain cancer. Three-dimensional glioblastoma models lacking the tumor-suppressor gene CDKN2A were substantially more sensitive to suppression of CDK6 than models retaining an intact copy of the gene. CDKN2A normally helps restrain cell-cycle progression, while CDK6 promotes the transition toward DNA replication and cell division. The result indicates that CDKN2A loss may serve as a biomarker for identifying glioblastoma patients who could be more likely to respond to existing CDK6-targeting drugs, although the finding will require validation in additional models and clinical studies.

The researchers also uncovered a vulnerability in pancreatic and other gastrointestinal organoids. Some of these models maintained a gene-expression program previously associated with a clinically important pancreatic cancer subtype. That program was largely lost in conventional cell lines, making it difficult to study using standard approaches. Organoids carrying the transcriptional state depended on several genes involved in WNT signaling, a pathway that regulates cell identity, tissue organization, and stem-cell behavior. The result points to a possible therapeutic opportunity that is specific to a cancer state preserved in 3D culture and may help explain why certain tumors behave differently in patients than in conventional laboratory experiments.

The study further showed that experimental conditions can shape dependency measurements in distinct ways. Genes involved in cell adhesion and cytoskeletal organization were particularly sensitive to whether cells grew in two or three dimensions. By contrast, genes involved in lipid metabolism were influenced more strongly by the composition of the culture medium, regardless of the physical format. Changing either the growth geometry or the nutrients available to cells altered the genes on which they depended. These observations provide a technical warning for cancer researchers: a dependency observed in one culture system may reflect not only tumor genetics but also the artificial conditions used to maintain the cells.

The expanded DepMap also revealed why 2D models remain valuable. Some breast cancer cell lines contained tumor markers that were not found in any of the 3D models examined, showing that organoids do not automatically reproduce every clinically relevant feature. Instead, the two systems offer complementary views of cancer biology. By combining their data, researchers can compare dependencies across genetic backgrounds, tissue architectures, and cellular states, improving the chances of distinguishing broadly useful drug targets from vulnerabilities that emerge only in particular experimental environments. The new resource is now available through the DepMap portal, where scientists can use it to investigate cancer mechanisms, prioritize therapeutic targets, and choose model systems more strategically.

Subject of Research: Cells

Article Title: A dependency map enhanced with next-generation 3D cancer models

News Publication Date: August 5, 2026

Web References: Cancer Dependency Map portal; Nature article

References: Neiswender JV, Maffa S, Brenan L, et al. “A dependency map enhanced with next-generation 3D cancer models.” Nature. August 5, 2026. DOI: 10.1038/s41586-026-10843-7.

Keywords: Cancer Dependency Map, DepMap, organoids, spheroids, 3D cancer models, cancer dependencies, glioblastoma, CDKN2A, CDK6, WNT signaling, pancreatic cancer, gastrointestinal cancer, precision medicine, drug targets, Broad Institute