# Comparing tumour-model preparations

Source: https://onco.cc/terms/functional-model-selection/  
OnCo record `functional-model-selection` (Term). Data CC BY-NC 4.0, attribute "Data from OnCo (onco.cc)"; commercial use needs a licence.

## TL;DR

Choose a preparation by the biology and time the question requires, then verify its actual components and endpoint.

## Summary

Conventional xenografts, epithelial organoids and short-term fragments make different trade-offs. Xenografts add a living host with mouse support and exposure. Organoids expand selected populations in supplied culture conditions. Fragments can retain local neighbours briefly without perfused circulation. This comparison describes preparations and decision windows; the existing model-context framework owns the broader question of intended-use validation.

## Fields

- Kind: Term
- Last checked: 2026-11-13

## Notes

- A conventional xenograft implants viable tumour material into a host; an epithelial organoid grows selected cells in supplied matrix and medium; a fragment or slice retains a cut piece of local tissue. Ladan used 300 micrometre (0.3 millimetre) breast slices. None supplies complete human drug metabolism, healthy-organ toxicity or recruitment of new immune cells from human circulation, and every preparation can fail. Fixed archival tissue cannot supply the living input. Whether frozen material is usable must be shown for the particular protocol; Ladan's assay used fresh tissue.
- Throughput and repetition depend on how material can be expanded. A xenograft can be passaged into further mice, but each drug comparison needs suitable animals and controls. An established organoid line can be expanded and passaged to compare a drug panel; Sachs reported retention of many source features during extended passaging, not unlimited fidelity. A cut piece has a finite culture window and a limited number of matched conditions; it cannot simply be expanded while retaining the same original neighbourhood. Ladan's group described viability for up to a week. These are practical preparation contrasts, not a universal numerical throughput ranking or an assumption that every organoid survives freezing. Sources: https://doi.org/10.1016/j.cell.2017.11.010 ; https://pmc.ncbi.nlm.nih.gov/articles/PMC10542784/
- The useful trade-off is preserved biology against time and repeatability. A characterised PDX supports growth under whole-animal dosing, usually on a months-long establishment clock. Expanded organoids support repeated direct tumour-cell experiments and panels, often over weeks; conventional epithelial cultures usually lack immune cells, while Neal's specialised method retained them. A short-term fragment may preserve resident neighbours for a local-interaction question over days, but their function must be demonstrated. No preparation is universally more predictive because it is more complex. Sources: https://doi.org/10.1016/j.cell.2018.11.021 ; https://pmc.ncbi.nlm.nih.gov/articles/PMC4167608/
- The clocks and take rates below belong to different studies. Hidalgo's review described four to eight months to a study-ready PDX. Izumchenko reported 578/1,163 establishment and mean 16 weeks from engraftment to drug exposure, with tumour-type variation: 85 percent in colorectal specimens and 30 percent in breast specimens in that series. Cartry reported a 61.5 percent core-biopsy take rate and six-week median, range four to ten weeks, to a 25-drug colorectal organoid panel. Boileve reported 54/87 establishment (62 percent) and mean 6.8 weeks to a pancreatic chemogram. Ladan used three-day exposures and stated that results can be known in seven to fourteen days; this is a planned window, not an observed turnaround distribution. There were 23/30 usable tests from tumour-containing biopsies. The PDX paper reports 49 percent for 578/1,163; direct count arithmetic is approximately 49.7 percent.
- Ladan's breast-slice pilot found 15/20 category matches (75 percent; reported confidence interval 53 to 89 percent), against a pre-set target of at least 55 percent, between the ex vivo anthracycline assay and MRI response after four courses of anthracycline-containing treatment. Both systems used sensitive, intermediate and resistant categories, and neither observed a resistant tumour. Of five disagreements, three were ex vivo sensitive with intermediate MRI response and two the reverse. Among the 23 successful laboratory tests, six were sensitive and 17 intermediate; among the 20 MRI comparisons, four were sensitive and 16 intermediate. This leaves resistance detection untested rather than established. Source: https://pmc.ncbi.nlm.nih.gov/articles/PMC10542784/
- Other agreement measures answer different questions. Vlachogiannis reported 88 percent positive and 100 percent negative predictive values in 21 drug-response comparisons in heavily pretreated gastrointestinal cancers. Boileve reported 83.3 percent sensitivity and 92.9 percent specificity in 34 evaluable pretreated pancreatic patients. Izumchenko compared 129 treatments across 92 patients: 112/129 agreement grouped stable disease with partial and complete responses, while exact-category agreement was 91/129. These endpoints and selected denominators cannot be ranked as head-to-head accuracy or taken as proof that model-guided care improves survival.
- Voabil's primary abstract reports that immune-cell reactivation after ex vivo PD-1 blockade in patient-derived tumour fragments predicted clinical response in that study, with perturbation analyses identifying tumour-resident T cells as a key component. This supports a scoped fragment/immune-interaction example, not evidence that choosing treatment from a fragment test improves outcomes. A near-term decision may rule out establishing a new conventional PDX, but does not guarantee a usable fragment; immune presence, movement, recognition and killing remain distinct measurements. Source: https://doi.org/10.1038/s41591-021-01398-3
- Dijkstra's colorectal and lung organoid study used master and working biobanks, cryopreserved cultures and experiments below passage 30. That demonstrates a specified bank-and-recovery workflow for those established organoid cultures. It does not show that every organoid line or fresh tumour fragment can be frozen, recovered and retested with unchanged properties. Source: https://pmc.ncbi.nlm.nih.gov/articles/PMC6558289/
- In a wholly fictional development project, D0 supplies one verified malignant preparation. Drug A asks about a direct repair perturbation: a tumour-only organoid supplies a viable-cell count after 48 hours, with division and death measured separately. The invented exposure series is 0.1, 0.3 and 1 micromoles per litre with three wells per condition and vehicle, positive and mechanism controls. Antibody B instead needs compatible tumour and immune populations over a stipulated seven days, with tumour-only, immune-only, matched antibody or carrier, and recognition controls. B's concentration and schedule are missing. The conventional graft's mouse exposure, schedule, assessment day and controls are also missing, so it remains an incomplete proposal. These conditions are not doses established as safe or achievable in people.
- A separate fictional three-week decision exercise names real drugs but invents every result. A tumour-only organoid can test a direct olaparib response using an attributed living-cell count after five days; it cannot supply the T cells needed for the pembrolizumab question. Fresh fragments retaining functional resident T cells or a characterised compatible co-culture can address a two-day activation question, with attributed tumour-cell death and recognition controls still needed. A conventional graft cannot restore an intact human immune system and its establishment may miss the deadline. Predicting RNA after BRCA1 loss is a computational hypothesis, not a performed drug exposure. Confirm the actual assay clock: a colorectal six-week median does not forecast the timing for this fictional breast specimen. No choice supplies complete human metabolism, healthy-organ toxicity or every original tumour population.

## Sources

- PDX review: https://pmc.ncbi.nlm.nih.gov/articles/PMC4167608/
- PDX primary denominators and response definitions: https://pmc.ncbi.nlm.nih.gov/articles/PMC5834154/
- Cartry colorectal organoids: https://doi.org/10.1186/s13046-023-02853-4
- Boileve pancreatic organoids: https://doi.org/10.1053/j.gastro.2024.05.032
- Ladan breast slices: https://pmc.ncbi.nlm.nih.gov/articles/PMC10542784/
- Vlachogiannis gastrointestinal organoids: https://pmc.ncbi.nlm.nih.gov/articles/PMC6112415/
- Voabil et al. 2021: tumour fragments and PD-1 blockade: https://doi.org/10.1038/s41591-021-01398-3
- Sachs et al. 2018: long-term breast organoid culture: https://doi.org/10.1016/j.cell.2017.11.010
- Neal et al. 2018: specialised immune-containing organoids: https://doi.org/10.1016/j.cell.2018.11.021
- DeRose et al. 2011: graft species compartments: https://pmc.ncbi.nlm.nih.gov/articles/PMC3553601/

## Connected records

- technologies: [Functional (ex vivo) drug testing](https://onco.cc/technologies/functional-drug-testing/), [Patient-derived organoids](https://onco.cc/technologies/organoids/), [Patient-derived xenografts](https://onco.cc/technologies/pdx-models/)

---
JSON: https://onco.cc/api/v1/entities/functional-model-selection.json