arxiv: 2604.24673 · v1 · submitted 2026-04-27 · 🧬 q-bio.CB

Recognition: unknown

Quantifying the effect of phenotype on clustering behaviour in melanoma: from monoculture to co-culture

Helen Byrne, Nathan Schofield, Richard White, Ruth Baker

Authors on Pith no claims yet

Pith reviewed 2026-05-07 17:09 UTC · model grok-4.3

classification 🧬 q-bio.CB

keywords melanomacell clusteringcoagulationproliferationphenotypic heterogeneityODE modelBayesian inferencein vitro

0 comments

The pith

A model of melanoma cell clustering shows invasive phenotypes coagulate nearly three times faster than proliferative ones.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper builds a system of ordinary differential equations that tracks cluster formation through coagulation, fragmentation, and cell proliferation. It fits the model to time-lapse microscopy data from two melanoma cell lines using Bayesian methods, revealing clear differences between phenotypes. Invasive cells merge into clusters at much higher rates, while proliferative cells divide slightly faster. When the same equations are applied to mixed co-cultures, clustering follows a hybrid pattern dominated by the invasive rate but with higher overall division. These findings matter because clusters can seed metastasis, so identifying which cell behaviors drive them points to possible ways to slow early tumor spread.

Core claim

Invasive melanoma cells in monoculture exhibit nearly threefold higher coagulation rates than proliferative cells, whereas proliferative cells display slightly higher proliferation rates. When applied to co-culture data, the model predicts hybrid coagulation behaviour of the clusters influenced by both proliferative and invasive cells but dominated by the invasive cells, and an elevated proliferation rate, suggesting a mutually beneficial effect of phenotypic heterogeneity on cell proliferation.

What carries the argument

A system of ordinary differential equations within a coagulation-fragmentation-proliferation framework that describes the time evolution of cluster size distributions, fitted to experimental counts via Bayesian inference.

If this is right

Co-culture clusters exhibit coagulation rates intermediate between the two monocultures but pulled toward the invasive value.
Proliferation rates rise in mixed cultures relative to either monoculture alone.
The fitted rate differences match known distinctions in gene expression between the two phenotypes.
Phenotypic mixing produces both faster merging and faster growth, consistent with cooperative effects inside tumors.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

Disrupting the high-coagulation behavior of invasive cells could slow cluster formation more efficiently than uniform treatments across all cells.
Varying the ratio of the two phenotypes in experiments would test whether the model correctly predicts how mixing changes overall growth speed.
Applying the same equations to three-dimensional culture data would show whether the rate differences survive outside flat surfaces.

Load-bearing premise

The assumption that coagulation, fragmentation, and proliferation are the main processes shaping cluster sizes, without needing extra terms for cell motility or changes in adhesion.

What would settle it

If measured coagulation rates between pairs of invasive cells do not come out close to three times those of proliferative cells, or if observed co-culture cluster sizes deviate systematically from the model's hybrid predictions.

read the original abstract

Melanoma is an aggressive form of skin cancer. Survival rates are excellent if it is detected early but fall markedly if it metastasises. A key step in early tumour progression is the formation of cell clusters, which can promote metastasis. However, the mechanisms driving cell clustering, and the role of phenotypic heterogeneity in the dynamics of these clusters, remain poorly understood. In this work, we propose a system of ordinary differential equations that models cluster formation dynamics within a coagulation-fragmentation-proliferation framework. Using Bayesian inference, we fit this model to in vitro time-lapse microscopy data from two melanoma phenotypes-proliferative and invasive-to uncover the predominant mechanisms driving cluster formation and how these differ between phenotypes. Additionally, we provide preliminary insights into how clustering behaviour in co-cultures contrasts with that observed in monocultures. The model quantifies phenotypic differences in clustering dynamics: invasive cells in monoculture exhibit nearly threefold higher coagulation rates than proliferative cells, whereas proliferative cells display slightly higher proliferation rates. These differences align with known gene expression profiles. When applied to co-culture data, the model predicts hybrid coagulation behaviour of the clusters influenced by both proliferative and invasive cells but dominated by the invasive cells, and an elevated proliferation rate, suggesting a mutually beneficial effect of phenotypic heterogeneity on cell proliferation.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper fits an ODE coagulation-fragmentation-proliferation model to melanoma time-lapse data and reports invasive cells coagulate nearly three times faster than proliferative cells, with co-cultures showing hybrid clustering dominated by the invasive phenotype plus elevated proliferation.

read the letter

The main thing to know is that this work takes a standard coagulation-fragmentation-proliferation ODE framework, fits it via Bayesian inference to monoculture data from two melanoma phenotypes, and extracts concrete rate differences: invasive cells show roughly threefold higher coagulation while proliferative cells show modestly higher proliferation. They then reuse those parameters on co-culture data to predict a hybrid coagulation pattern dominated by the invasive cells and an overall rise in proliferation, which they link to known gene expression differences. That produces new quantitative comparisons and a mixed-culture prediction not in the prior literature they cite. The approach is straightforward and the numbers are testable in follow-up experiments. The modeling step is the clear limitation. The ODEs omit explicit motility, adhesion variation, or microenvironment terms, so any such effects in the microscopy would be absorbed into the fitted coagulation rates and reduce mechanistic clarity. The abstract gives no sign of posterior predictive checks on full cluster-size histograms, sensitivity tests on fragmentation assumptions, or comparison to a spatial or agent-based alternative, which leaves the quantitative claims plausible but not yet tightly verified. Readers working on mathematical models of tumor progression and metastasis would get the most from the rate estimates and the co-culture extension. The paper is clear enough on its own terms and the claims are falsifiable, so it deserves a serious referee even if revisions will be needed for stronger validation.

Referee Report

3 major / 2 minor

Summary. The manuscript develops a coagulation-fragmentation-proliferation ODE model for melanoma cell cluster dynamics. Bayesian inference is used to fit coagulation and proliferation rates separately to time-lapse microscopy data from proliferative and invasive monocultures. The resulting parameters are then applied to co-culture data to predict hybrid clustering behaviour (invasive-dominated coagulation) and elevated proliferation, with claims that these differences align with known gene-expression profiles.

Significance. If the ODE framework is shown to be adequate, the work supplies quantitative, phenotype-specific rate estimates (invasive coagulation nearly 3x proliferative) and preliminary co-culture predictions that could inform metastasis studies. The Bayesian fitting approach and direct use of in vitro time-lapse data are strengths; the alignment with independent biological knowledge adds credibility. Significance is currently limited by incomplete verification of model assumptions and predictive performance.

major comments (3)

[Modeling and inference sections] Modeling and inference sections: the Bayesian fits to monoculture data are presented without reported posterior predictive checks on full cluster-size histograms, cross-validation, or sensitivity analysis to fragmentation assumptions. Because the central claims rest on the quantitative rate differences extracted from these fits, the absence of these diagnostics leaves the adequacy of the effective ODE description unverified.
[Co-culture results] Co-culture results: predictions of hybrid coagulation behaviour and elevated proliferation are generated by reusing parameters fitted exclusively to monoculture data rather than by joint fitting or direct comparison against co-culture observations. This reuse weakens the claim that the model successfully predicts co-culture dynamics.
[ODE framework assumptions] ODE framework assumptions: the coagulation-fragmentation-proliferation system is taken as capturing the dominant mechanisms, yet no model-comparison tests (e.g., against a spatial PDE or agent-based alternative) are shown to rule out substantial contributions from cell motility or phenotype-dependent adhesion visible in the microscopy. If such processes are absorbed into the fitted coagulation rates, the mechanistic interpretability of the reported threefold difference is compromised.

minor comments (2)

[Abstract] Abstract: the phrase 'hybrid coagulation behaviour of the clusters' is used without a quantitative definition or reference to the specific predicted cluster-size distributions; a brief numerical illustration would improve clarity.
[Notation] Notation: ensure that the symbols for cluster number densities n_k(t), coagulation kernel K(i,j), and fragmentation function F(k) are introduced once and used consistently in all equations and figure captions.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for their constructive and detailed comments, which have prompted us to strengthen the validation and clarify the scope of our model. We address each major comment below and have incorporated revisions where feasible given the available data and experimental design.

read point-by-point responses

Referee: Modeling and inference sections: the Bayesian fits to monoculture data are presented without reported posterior predictive checks on full cluster-size histograms, cross-validation, or sensitivity analysis to fragmentation assumptions. Because the central claims rest on the quantitative rate differences extracted from these fits, the absence of these diagnostics leaves the adequacy of the effective ODE description unverified.

Authors: We agree that these diagnostics are essential to verify the model. In the revised manuscript we have added posterior predictive checks that overlay simulated cluster-size histograms on the experimental data for both phenotypes. We also performed a sensitivity analysis on the fragmentation rate parameter, confirming that the reported nearly threefold difference in coagulation rates is robust across reasonable fragmentation values. Cross-validation was constrained by the limited number of independent time-lapse experiments, but we included a temporal hold-out validation (fitting to early time points and predicting later ones) whose results are now shown in the supplement. These additions directly address the adequacy of the ODE description. revision: yes
Referee: Co-culture results: predictions of hybrid coagulation behaviour and elevated proliferation are generated by reusing parameters fitted exclusively to monoculture data rather than by joint fitting or direct comparison against co-culture observations. This reuse weakens the claim that the model successfully predicts co-culture dynamics.

Authors: We acknowledge that parameter reuse constitutes an out-of-sample prediction rather than a joint fit, which limits the strength of the validation claim. Phenotype-specific tracking is not feasible in the current co-culture imaging data without additional labeling, precluding joint fitting. We have now added a direct comparison of the monoculture-parameter predictions against the observed co-culture cluster-size histograms, demonstrating reasonable agreement especially for the invasive-dominated regime. The text has been revised to present this explicitly as a predictive test and to moderate the language around mutual benefit while retaining the biological interpretation. revision: partial
Referee: ODE framework assumptions: the coagulation-fragmentation-proliferation system is taken as capturing the dominant mechanisms, yet no model-comparison tests (e.g., against a spatial PDE or agent-based alternative) are shown to rule out substantial contributions from cell motility or phenotype-dependent adhesion visible in the microscopy. If such processes are absorbed into the fitted coagulation rates, the mechanistic interpretability of the reported threefold difference is compromised.

Authors: We agree that motility and adhesion may contribute to clustering and that the fitted coagulation rates are effective (net) rates that could absorb these processes. The ODE model is intended as a population-level description rather than a fully mechanistic spatial model. We have expanded the discussion to explicitly state the mean-field assumptions, note the potential role of spatial effects visible in the microscopy, and explain that a systematic comparison to PDE or agent-based alternatives was beyond the present scope owing to data resolution and computational demands. We retain the interpretation of the threefold difference as a phenotype-specific effective rate while acknowledging the limitation on mechanistic decomposition. revision: partial

Circularity Check

0 steps flagged

No significant circularity in derivation chain.

full rationale

The paper constructs an ODE coagulation-fragmentation-proliferation model, fits its rate parameters via Bayesian inference to monoculture time-lapse data for each phenotype separately, and then re-uses the resulting parameter values to generate predictions on separate co-culture data. This is ordinary model calibration followed by application to new observations; the co-culture outputs are not forced by construction from the monoculture fits, nor does any equation reduce to a tautology. No self-citation is invoked to justify uniqueness of the framework, no ansatz is smuggled, and no known empirical pattern is merely renamed. The derivation chain from model equations through inference to phenotype-specific rates and co-culture predictions remains independent of its own inputs.

Axiom & Free-Parameter Ledger

4 free parameters · 1 axioms · 0 invented entities

The central claims rest on three fitted rate parameters per phenotype plus the assumption that the chosen ODE framework is sufficient; no new entities are postulated.

free parameters (4)

coagulation rate (invasive)
Fitted via Bayesian inference to monoculture time-lapse data; reported as nearly threefold higher than proliferative.
coagulation rate (proliferative)
Fitted via Bayesian inference to monoculture time-lapse data.
proliferation rate (proliferative)
Fitted via Bayesian inference; reported as slightly higher than invasive.
proliferation rate (invasive)
Fitted via Bayesian inference.

axioms (1)

domain assumption Cluster formation dynamics are adequately described by a coagulation-fragmentation-proliferation ODE system
Invoked as the modeling framework in the abstract.

pith-pipeline@v0.9.0 · 5534 in / 1341 out tokens · 50779 ms · 2026-05-07T17:09:17.307960+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

2 extracted references · 2 canonical work pages

[1]

Computational Statistics 15(3):337–354

https://doi.org/10.1038/s41556-021-00740-8 Berkhof J, Van Mechelen I, Hoijtink H (2000) Posterior predictive checks: principles and discussion. Computational Statistics 15(3):337–354. https://doi.org/10.1007/ s001800000038 Campbell NR, Rao A, Hunter MV, et al (2021) Cooperation between melanoma cell states promotes metastasis through heterotypic cluster f...

work page doi:10.1038/s41556-021-00740-8 2000
[2]

Genes & Development 33(19- 20):1295–1318

https://doi.org/10.1016/j.celrep.2015.09.037 Rambow F, Marine JC, Goding CR (2019) Melanoma plasticity and phenotypic diversity: therapeutic barriers and opportunities. Genes & Development 33(19- 20):1295–1318. https://doi.org/10.1101/gad.329771.119 Raue A, Karlsson J, Saccomani MP, et al (2014) Comparison of approaches for param- eter identifiability ana...

work page doi:10.1016/j.celrep.2015.09.037 2015