Cancer researchers have spent decades building genomic atlases of human tumors and using laboratory-grown cancer cell lines to uncover genes that drive cancer growth. Together, these efforts have transformed drug discovery and helped identify many of today’s targeted therapies.
But an important gap has remained.
Many cancer cell lines represent only a narrow slice of the diversity seen in patients, lack matched clinical information, and gradually adapt to laboratory culture. Meanwhile, patient-derived organoids—three-dimensional miniature tumors grown directly from patient tissue—better preserve the biology of human cancers but have been too small, inconsistently characterized, or technically difficult to use for large-scale functional studies.
Now, researchers at the Wellcome Sanger Institute and collaborators report what they say is a major step toward closing that gap.
Writing in Nature, the team describes a publicly available biobank of 256 clinically annotated organoids derived from colorectal, esophageal, gastric, pancreatic and ovarian cancers. They also performed genome-wide CRISPR-Cas9 screens across 162 of the models, creating the first large-scale cancer dependency map generated in patient-derived organoids.
The resource combines whole-genome sequencing, transcriptomics, clinical annotation, and functional CRISPR screening, allowing investigators to connect gene dependencies with patient characteristics, tumor evolution, and therapeutic response.
“We know that cell lines don’t capture all of the heterogeneity that cancer has,” said first author Carmen Herranz-Ors, PhD. “Our goal was to fill those gaps by generating models directly from patients while also including detailed clinical metadata and matched normal tissue, so researchers can perform much more accurate analyses.”
Building a better research model
Cancer cell lines remain indispensable research tools, but Herranz-Ors stressed that they cannot fully represent the complexity of human tumors. Some cancer subtypes are poorly represented or missing entirely, and because cell lines must adapt to grow on flat plastic surfaces, they can gradually lose characteristics found in the tumors from which they originated.
Organoids offer an important complement.
Rather than growing as two-dimensional sheets of cells, organoids maintain a three-dimensional architecture that more closely resembles tumors. More importantly, because they are derived directly from patient tissue, they preserve considerably more of the biological diversity present across patients.
“The three-dimensional structure is helpful, but for us the biggest advantage is that they come directly from patients,” Herranz-Ors said. “That allows us to capture much more heterogeneity, and because we also have the patients’ clinical information and matched normal tissue, we can interpret the genomic data much more accurately.”
Unlike many previous collections, every organoid in the new biobank underwent extensive molecular characterization, including whole-genome and transcriptome sequencing. The researchers also generated matched normal DNA from each patient, enabling more reliable identification of tumor-specific mutations.
The organoids and associated datasets—including genomic, transcriptomic, clinical and CRISPR screening data—have been deposited in public repositories to encourage broad use by the research community.
“I think the organoid biobank is a very good resource for scientists who want to test new hypotheses,” Herranz-Ors said. “Researchers can first explore the data, identify models that fit their questions, and then obtain those organoids to perform additional experiments.”
Bringing functional genomics to organoids
The study’s second major advance was demonstrating that organoids can support genome-wide CRISPR screening at a scale comparable to traditional cancer cell lines. That achievement required overcoming significant technical hurdles.
Whole-genome CRISPR screens require enormous numbers of cells—roughly 100 million cells per organoid line—and organoids are substantially more difficult to culture and manipulate than conventional cell lines.
To make the screens feasible, the investigators developed suspension culture methods and designed a more compact CRISPR guide RNA library that reduced the number of cells needed while maintaining genome-wide coverage.
The resulting dependency map identified approximately 1,700 associations between gene dependencies and genomic or clinical characteristics. Many confirmed established cancer biology, providing confidence in the approach, while others highlighted vulnerabilities that had been difficult to detect using cell lines alone.
For example, organoids displayed greater dependence on cholesterol metabolism than conventional cell lines. “We expected that because organoids are probably closer to what happens in real tumors,” Herranz-Ors said. “But we hadn’t been able to demonstrate it before at this level.”
KRAS differences could inform therapy
Among the study’s most clinically relevant observations were differences among KRAS mutations.
Although KRAS-mutant tumors are often considered together, the researchers found that individual KRAS variants exhibit distinct biological behavior.
Organoids carrying KRAS G12 mutations remained more dependent on both KRAS and upstream EGFR signaling than tumors harboring KRAS Q61 mutations, suggesting that patients with different KRAS mutations may respond differently to emerging targeted therapies.
Because the biobank includes many more patient-derived models than previous collections, the researchers were able to distinguish these allele-specific dependencies with greater statistical confidence.
“This could become very useful as pharmaceutical companies continue developing KRAS inhibitors,” Herranz-Ors said. “If we know certain mutations respond differently, that can help identify which patients are most likely to benefit from particular combinations.”
Following tumors as they evolve
The biobank also includes matched organoids collected before and after treatment from the same patients, offering an opportunity to study how tumors evolve under therapeutic pressure.
Although resistance mechanisms were not the primary focus of the current study, the researchers demonstrated that CRISPR screening of paired organoids could uncover new vulnerabilities that emerge after treatment failure, potentially identifying therapeutic opportunities for resistant disease.
“Our goal was really to show examples of what researchers can do with this resource,” Herranz-Ors said. “You can investigate how tumors become resistant and identify new targets that appear after treatment.”
A resource designed for the community
Harranz-Ors repeatedly returned to one point: the paper is less about individual discoveries than about providing a foundation for many future ones.
“We really want people to use the data,” she said. “Everything is public—the genomic data, transcriptomics, clinical information, CRISPR data—and the models themselves can be obtained. We hope this becomes a resource that helps accelerate discoveries across cancer research.”
The authors emphasize that organoids are not intended to replace traditional cell lines, which remain essential for many types of cancer research. Instead, the new resource expands the experimental toolbox by providing patient-derived models that capture tumor diversity while supporting the same kinds of large-scale functional studies that have helped define precision oncology over the past decade.
Note: In related papers published in the same issue of Nature, researchers describe complementary advances in next-generation cancer models. One study, led by researchers at the Broad Institute of MIT and Harvard, presents a dependency map enhanced with next-generation 3D cancer models. A second paper describes the Human Cancer Models Initiative (HCMI) organoid collection as an international flagship resource for cancer research.
