If you’re looking to identify a new therapeutic compound, then setting up a drug screen is an obvious step. But are you aware of how low clinical trial success rates can be, especially in oncology? The figure is around 3–5% (depending on the dataset), meaning many promising candidates fail along the way (1).
There’s no doubt that a high attrition rate (the proportion of drug candidates that fail and drop out during development) is undesirable and will cost you in both time and money. So what can be done to maximize your chances of success?
In this article, we discuss screening model options, explain why many are turning to patient-derived organoids for drug development, and offer advice on using them in your research.
Choosing the Right Model
If you want to develop a successful screen, you need to understand why so many fail. One pitfall of traditional screening models is that they do not appropriately reflect the biology, physiology, and heterogeneity of the disease.
Take colorectal cancer (CRC), for example. It is not one disease, but at least four, each with its own molecular fingerprint, ranging from a subtype found in 37% of patients to one found in just 13% (2). No two tumors are heterogeneous in quite the same way either; the amount of internal variation within a single patient’s tumor can look completely different from the next patient’s (3).
Before you start planning your screen, consider which model fits your needs. Let’s start by comparing some of the most common models, and the pros and considerations for each.
Weighing up Common Models for Drug Screening
Issues with Immortal Cells
If you’ve done cell culture before, you’re probably very familiar with immortalized cells and how relatively cheap, easy, and quick to grow they are, making them ideally suited to rapid screening.
However, they notoriously lack important disease-relevant mutations, have high genetic drift over passages, and offer low heterogeneity (4).
While immortalized cell lines are well-suited to high-throughput drug screens, their low biological relevance can lead to high attrition rates, which we want to avoid.
To improve your odds, choose a model that more closely reflects the disease you’re studying.
Getting Closer with Patient-derived Cells
Patient-derived cells (PDCs), as their name suggests, come from patient tumor tissue and are cultured as flat, 2D monolayers. This offers a step up from immortalized cells in terms of disease relevance, and they are still relatively cost-effective and easy to culture.
However, growing in a monolayer means they are, by their very nature, two-dimensional, and so don’t accurately represent the 3D tumor environment, which can impact drug responses: PDCs and 3D organoids derived from the same patient show different drug responses, despite having the same underlying genetics (5).
More Relevant: Patient-derived Xenografts
To get a more three-dimensional take on disease, you can transplant patient-derived tissue into animal models. With patient-derived xenografts (PDXs), fresh patient tumor tissue is implanted into immunocompromised mice and allowed to grow before further passaging to expand the number of tumor-bearing mice, since a single engraftment doesn’t provide enough matched tumors for a drug study.
These PDX models retain tumor heterogeneity and genomic stability even with passaging, making them highly representative of the original tumor. However, a big drawback is that these models can take 4 to 8 months to develop.
When you factor in the difficulty of scaling these models, their incompatibility with high-throughput screening, and variable engraftment success rates (10–70%), PDX models become less appealing overall (4).
So you need a balance among ease of use, cost, and compatibility with high-throughput screening. Let’s look at another patient-derived option you should consider: PDOs.
The Promise of Patient-derived Organoids for Drug Development
If you are looking for a compromise between cost-effective, speedy immortalized cell lines and the more biologically relevant PDX models, PDOs may be the right choice (Figure 1). Like PDX models, PDOs are developed directly from patient samples and retain important disease features.
Figure 1. Key features of PDOs as a model for drug screening.
Unlike PDXs, PDOs are cultured as 3D organoids in vitro rather than in vivo, reducing development time and cost. They also don’t require you to have any specialized facilities beyond a standard cell culture setup.
PDOs offer you an added benefit: they have higher generation rates than PDXs (70-100% vs 10-70%) (4), and because you grow them in dishes or well plates, you can handle and screen them at scale, making them well-suited to high-throughput screening.
Struggling to find patients with the genotypes or mutations you need? With PDXs, you may need to wait months or even years to get patient samples with the right mutations and diversity for a meaningful screen. Because organoids are grown in culture, you can genetically manipulate them to introduce the mutations or genotypes you are interested in, saving precious time developing your screen (6).
And if you want to investigate the mechanism of action once you’ve got your lead compound(s), you can again use genetic manipulation, for example, using knockdowns to show changes in drug susceptibility (6).
What You Need to Know When Using Patient-derived Organoids
PDOs may sound like the perfect choice for your screen, but what do you need to know to use them successfully?
Before you warm up your media, you need to think about where you will source your organoids, ensuring they are representative of the disease of interest, that they include specific genotypes relevant to your question, and whether you want matched normal tissue (keep reading to find out why).
Sourcing Your Organoids
If you have access to patient samples representative of the disease you wish to study and practical experience in developing organoids, deriving your own models is a viable, cost-effective option, especially since the time from biopsy to data analysis can be as short as three weeks (7).
However, if you are not able or willing to develop your own models in-house, you can purchase pre-validated, quality-controlled, and characterized models, allowing you to run assays within days of thawing.
This approach offers you access to models of specific genotypes sourced from many patients, providing greater heterogeneity that better reflects the disease.
Why You Need to Think About Genotype
Many diseases are heterogeneous, and their drug response often depends on the mutations they carry. Therefore, you should ensure the models used in your screen include any relevant mutations.
For example, if you are interested in CRCs resistant to EGFR or MEK pathway inhibitors, ensure your panel includes KRAS or NRAS mutations.
Bigger Does Not Necessarily Mean Representative
A small screening panel may be unrepresentative because of its size, but ensuring your panel reflects the disease accurately takes more than running a larger screen; it requires checking that mutation frequencies match those seen in the patient population.
Take the time to make sure your panel reflects what you see in the patient population you are interested in, in terms of mutation types and frequencies, and you’ll boost your chances of identifying a viable lead.
The Importance of Matched Samples
Matched PDOs means having organoids from both the tumor and normal tissue (of the same type) from the same patient. Matched samples improve your understanding of compound cytotoxicity.
A compound you test only on tumor samples may look incredibly promising. Still, if it kills healthy matched cells, it’s likely highly cytotoxic and therefore lacks specificity (we want to target the disease, not kill healthy cells). However, if you see large differences in IC50 values (the drug concentration needed to inhibit growth by 50%), that’s a good sign that your candidate is specifically targeting the tumor.
Considerations When Using PDOs for Lead Optimization in Drug Development
So your screen has turned up a handful of promising leads, but how do you decide which one to take forward? Choosing the wrong candidates can be costly and set your research back, so evaluate them carefully before moving to in vivo studies.
Build Certainty with a Wider Screen
A smaller screening panel is great for testing many different compounds, but once you’ve narrowed it down to just a few, you want to be able to test a larger panel to see which genotypes respond and which don’t, helping you spot the patients most likely to benefit, and any resistance, early on.
Understand How Your Candidates Work
Once you’ve found your lead candidate, understanding how it works can give you greater confidence that it’s a viable therapy. Because PDOs can be genetically manipulated, you can knock down the candidate’s target to confirm it works as intended, or even introduce mutations to see how resistance might emerge (6).
Best of all, you can do this in weeks, allowing you to drop weaker candidates quickly, before you’ve invested in costly in vivo studies.
But before you invest time and money into an organoid-based screen, you should be sure it’s the right approach. The following study shows how a PDO screen was used to identify a lead compound, assess how well it inhibits tumor growth compared with the current standard-of-care treatment, evaluate tumor specificity, and uncover its mechanism of action.
Patient-derived Organoid Screening for CRC Drug Candidates
CRC is the third most common cancer in the world (8), and of new diagnoses, 20% have metastatic disease, with another quarter developing metastases later. Put together, that’s nearly half of all CRC patients facing metastatic disease, a stage that’s hard to treat and where effective new therapies could make a real difference.
To make this even more urgent, metastatic CRC (mCRC) is largely incurable. However, about half of patients with unresectable tumors show activation in the EGFR pathway, making them candidates for cetuximab, a monoclonal antibody that blocks EGFR signaling and is the current standard of care for these tumors.
Unfortunately, this treatment is not effective for a subset of patients. Those with KRAS or NRAS mutations currently have no targeted treatment available (9).
To identify effective treatments for KRAS-mutant CRC, researchers put a panel of CRC PDOs to work, screening for potential therapeutic candidates from a library of bispecific antibodies.
Each antibody in the library had two arms: one that binds a cancer stem cell surface marker (LGR4, LGR5, ZNRF3, or RNF43) and another that binds a growth factor receptor (EGFR or HER3).
After screening over 500 bispecific antibodies across three CRC organoid models, 28 candidates advanced to a second screen. This was against a larger panel of 25 different CRC models that captures most known CRC genetic alterations (10).
The organoids showed a mutational profile similar to real colorectal tumors. Their genotypes overlapped closely with TCGA (The Cancer Genome Atlas) data across the four pathways driving most colorectal cancers: WNT, RTK/RAS, TP53, and TGF-β (Figure 2).
A mutational profile that reflects the real disease means it is broadly representative of the patient population, so how the organoids respond to a drug is more likely to reflect how patients would.
Figure 2. Genetic alterations in the HUB Organoids CRC biobank closely match TCGA-reported CRC data across the four main driver pathways (WNT, RTK/RAS, TP53, TGF-β). Adapted from Herpers et al. (10).
This second round of bispecific antibody screening identified a promising lead candidate (that bound both EGFR and LGR5), with IC50 values as low as 0.08 to 3.09 µg/mL in KRAS mutant tumor organoids. This showed that low doses inhibited tumor organoid growth and meant it was a promising candidate.
It was even more promising when the researchers found that healthy organoids from the same patients were largely unaffected, even at doses roughly 100-fold higher than those that inhibited the tumors. This showed that the lead candidate acted specifically on cancer cells (Figure 3) (10).
Figure 3. Efficacy and specificity of lead candidate compared with cetuximab. The table shows IC50 values across matched tumor and normal colon organoids. The graph shows the impact of dose on tumor size across matched tumor and healthy colon organoids. Adapted from Herpers et al. (10).
And when they compared this candidate to the standard-of-care treatment, cetuximab, it was better at inhibiting CRC organoid growth, including in KRAS-mutant organoids (10).
Determining the Lead Candidate’s Mechanism of Action in KRAS-mutant CRC
The HUB Organoid models played a central role in screening, but their usefulness did not end there. Because these models are relatively straightforward to genetically manipulate, researchers could confirm how the lead candidate works. Using shRNA knockdown of LGR5 decreased sensitivity to the bispecific antibody, showing that targeting LGR5 was critical (Figure 4A).
However, targeting LGR5 or EGFR separately produced little effect, and even when two separate antibodies (one against each target) were combined, only moderate inhibition was seen (Figure 4B). This meant that dual binding specifically produced the desired anti-tumor effect.
Figure 4. Uncovering the mechanism of action of the lead antibody. (A) LGR5-knockdown organoids (shLGR5) showed reduced sensitivity to the antibody. (B) Monospecific antibodies targeting LGR5 or EGFR alone, or in combination, showed only moderate growth inhibition compared with the lead candidate. Adapted from Herpers et al. (10).
Want the full picture? The complete case study, including the original screening data and imaging, is available here.
The Power of PDO-based Screening in Drug Development
This study shows what a well-designed organoid screen can achieve: identifying a promising lead candidate, confirming its tumor specificity using matched normal-tissue organoids, discovering its mechanism of action through genetic manipulation, and benchmarking it against the standard-of-care treatment.
And to further show the power of PDOs, the lead candidate in this case study, now known as petosemtamab, has since progressed to two Phase 3 trials in mCRC, which began in 2026 (11, 12).
References
- Wong, C.H., Siah, K.W. and Lo, A.W. (2019) Estimation of clinical trial success rates and related parameter. Biostatistics, 20(2), pp. 273–286.
- Guinney, J., Dienstmann, R., Wang, X., et al. (2015) The consensus molecular subtypes of colorectal cancer. Nature Medicine, 21(11), pp. 1350–1356.
- Suzuki, Y., Ng, S.B., Chua, C., et al. (2016) Multiregion ultra-deep sequencing reveals early intermixing and variable levels of intratumoral heterogeneity in colorectal cancer. Molecular Oncology, 10(9), pp. 1391–1403.
- Fan, H., Demirci, U. and Chen, P. (2019) Emerging organoid models: leaping forward in cancer research. Journal of Hematology & Oncology, 12, p. 142.
- Jabs, J., Zickgraf, F.M., Park, J., et al. (2017) Screening drug effects in patient-derived cancer cells links organoid responses to genome alterations, Molecular Systems Biology, 13(11), p. 955.
- Verissimo, C.S., Overmeer, R.M., Ponsioen, B., et al. (2016) Targeting mutant RAS in patient-derived colorectal cancer organoids by combinatorial drug screening. eLife, 5, e18489.
- Martini, G., Belli, V., Napolitano, S., et al. (2023) Establishment of patient-derived tumor organoids to functionally inform treatment decisions in metastatic colorectal cancer. ESMO Open, 8(3), 101198.
- Bray, F., Laversanne, M., Sung, H., et. al. (2024) Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA: A Cancer Journal for Clinicians, 74(3), pp. 229–263.
- Biller, L.H. and Schrag, D. (2021) Diagnosis and treatment of metastatic colorectal cancer: a review. JAMA, 325(7), pp. 669–685.
- Herpers, B., Eppink, B., James, M.I., et al. (2022) Functional patient-derived organoid screenings identify MCLA-158 as a therapeutic EGFR × LGR5 bispecific antibody with efficacy in epithelial tumors. Nature Cancer, 3(4), pp. 418–436.
- ClinicalTrials.gov (2026a) Study of petosemtamab plus chemotherapy versus cetuximab plus chemotherapy in RAS and BRAF wild-type, unresectable or metastatic, left-sided colorectal cancer (LiGeR-CRC1). Identifier: NCT07702032. https://clinicaltrials.gov/study/NCT07702032 (Accessed 17 September 2026)
- ClinicalTrials.gov (2026b) Study of petosemtamab plus chemotherapy versus cetuximab or bevacizumab plus chemotherapy in RAS and BRAF wild type, recurrent, unresectable or metastatic colorectal cancer (LiGeR-CRC2). Identifier: NCT07775287. https://clinicaltrials.gov/study/NCT07775287 (Accessed 17 September 2026)
