International team creates 700 patient-derived cancer models to accelerate drug development

For decades, cancer researchers have faced a fundamental problem: the laboratory models they use to test new drugs don't always match the tumors they're trying to treat. A patient's cancer is a unique thing—shaped by their own genetics, their environment, their history. But the cell lines scientists have relied on, many developed in the 1950s and grown as flat layers in petri dishes, capture only a fraction of that complexity. An international consortium has now spent a decade trying to fix this gap.

Led by researchers at MIT's Koch Institute, the Broad Institute, Dana-Farber Cancer Institute, and the National Cancer Institute, along with dozens of partner institutions, the team has created nearly 700 new cancer models derived directly from patient tumors. These models represent 25 different cancer types and are now available to researchers worldwide. The work, published today in Nature, is the result of a systematic effort to build what the field has long needed: a library of tumor samples diverse enough to reflect the actual genetic and molecular landscape of human cancer.

The project began in 2016, following the completion of the Cancer Genome Atlas—a massive effort to sequence cancer cells from thousands of patients. That work revealed something sobering: the roughly 1,000 patient-derived cancer cell lines that existed at the time didn't capture the full diversity of tumor genetics. Most came from European and Southeast Asian patients. Rare cancers were barely represented. Jesse Boehm, a senior researcher on the project, put it plainly: "We realized that a thousand wasn't enough, that the international community needed to invest in many more thousands to represent all cancers, all genotypes, all ethnicities."

The consortium collected more than 2,700 tumor samples from hospitals in the United States, the United Kingdom, and the Netherlands—patients who consented to have their cells used for research. The samples included common cancers like lung, liver, and pancreatic tumors, but also about 150 rare types: gallbladder cancers, small intestine tumors, and others that rarely make it into research collections. Converting these raw samples into usable models proved technically demanding. The researchers developed specialized culturing techniques, growing cells in custom growth media and three-dimensional scaffolds. About one-third of the samples successfully became stable cell lines or organoids—three-dimensional structures of cells embedded in a gelatin-like matrix that more closely mimic the architecture of actual tumors.

Once established, which could take up to a year per model, each was rigorously characterized. Researchers sequenced the genomic DNA, analyzed RNA expression patterns, and examined epigenomic modifications to ensure the laboratory models matched the original tumors. Every model now comes with extensive data: the patient's inherited genetic mutations, information about treatments they received, and detailed molecular profiles. All of this has been deposited at the American Type Culture Collection, a nonprofit distributor, making the resource freely available to the global research community.

The practical payoff is already visible. In a companion paper also published today, Broad Institute researchers profiled more than 300 of the new models using high-throughput sequencing and CRISPR screens—a technique that systematically disables genes to identify which ones cancer cells depend on for survival. These vulnerability maps have been added to the Cancer Dependency Map, a growing database now containing information on more than 2,000 cancer types. This kind of large-scale screening, once impossible with limited models, could accelerate the identification of new drug targets and the validation of experimental treatments.

Mushriq Al-Jazrawe, scientific director of the High Throughput Sciences platform at the Koch Institute and a lead author, emphasized the scale of the undertaking: "A resource of this scale depends on the kind of systematic effort that often happens behind the scenes." The work involved hundreds of scientists across dozens of institutions, each contributing technical expertise and careful stewardship of precious patient samples.

Boehm and his colleagues are clear that this is not an endpoint. The formal HCMI project is winding down, but researchers plan to continue developing models from additional patient samples, with particular focus on pediatric cancers and rare tumors still underrepresented in the collection. "We now have about 2,000, but if we really want to represent all humans with cancer in our preclinical research, more work is needed," Boehm said. "I think this will hopefully be not the end, but the beginning."