🧬 biology

Divergence of detachment forces in the finite Voronoi model

This paper identifies a critical time-step dependence in the finite Voronoi model caused by diverging detachment forces during tissue fracture, proposes a regularization method to correct this unphysical behavior, and demonstrates that proper calibration of near-detachment mechanics is essential for accurately simulating nonconfluent tissue dynamics.

Original authors: Wei Wang, Brian A. Camley

Published 2026-08-05

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Wei Wang, Brian A. Camley

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ⚕️ This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer

Imagine a bustling city made entirely of living, breathing cells. Sometimes, these cells pack together so tightly they form a solid, seamless wall, like a mosaic where every tile touches its neighbors. Other times, they pull apart, leaving gaps and forming wandering groups that drift through the body. This dance between staying together and falling apart is crucial for life; it's how our bodies heal wounds, how embryos grow, and unfortunately, how cancer spreads. To understand this, scientists use computer models—digital sandboxes where they can watch cells move, push, and pull without needing a microscope. One popular type of model treats cells like bubbles in a foam, using a mathematical grid called a "Voronoi diagram" to decide who touches whom. It's a clever trick that works great when cells are packed tight, but what happens when they start to drift apart? That's the question a team of researchers at Johns Hopkins University set out to answer, and they found a surprising glitch in the math that changes everything we thought we knew about how tissues break.

The researchers, Wei Wang and Brian A. Camley, were studying a specific version of these models called the "finite Voronoi model." Think of this model as a game where every cell is a circle that can't grow bigger than a certain size. When two cells touch, they flatten against each other, creating a straight edge, while the rest of their body remains a perfect curve. This setup is great for simulating tissues that are starting to break up, like a cluster of cancer cells trying to escape a tumor. However, the team discovered that when they tried to simulate these cells actually separating, the computer code started acting weird.

In their simulations, they noticed something strange: the speed at which they ran the computer mattered more than the physics of the cells themselves. If they ran the simulation with a "slow motion" setting (using very small time steps), the cells refused to let go of each other, no matter how hard they pulled. If they ran it faster (larger time steps), the cells would suddenly snap apart. It was as if the cells were glued together by the speed of the clock rather than by any real biological force. The team realized this wasn't just a computer bug; it was a fundamental flaw in the geometry of the model.

Here is the core of the problem: In this specific model, as two cells pull apart, the point where they touch gets smaller and smaller. The math that calculates the force needed to pull them apart relies on the size of that tiny contact point. As the contact point shrinks toward zero, the math says the force needed to separate them shoots up to infinity. In the real world, cells don't need infinite strength to pull apart; they just stretch and let go. But in the computer model, the force becomes so huge that the simulation gets stuck. The only reason the cells ever separated in the original code was that the computer took big enough "steps" in time to accidentally jump over that infinite force wall. It was like trying to walk through a door that gets infinitely narrow; if you take small steps, you get stuck, but if you take a giant leap, you might accidentally land on the other side.

To fix this, the authors introduced a "regularization," which is a fancy word for a simple rule to stop the math from breaking. They decided that once the contact between cells gets smaller than a tiny, specific threshold (about 0.45 units in their simulation), the force stops increasing and stays constant. This is like saying, "Okay, the contact is so small now that it's effectively gone, so let's stop pretending it's holding them together with infinite strength." With this fix, the model stopped depending on how fast the computer was running, and the cells could detach in a way that made physical sense.

But the team didn't stop there. They wanted to know if their fixed model actually matched reality. So, they compared it to a different, more complex model called the "deformable polygon model," where cells aren't forced to be perfect circles with straight edges; they can squish and stretch like real jelly. They found that simply fixing the math wasn't enough; they also had to tune the model carefully. Depending on how they tuned the "stickiness" and the size of the cells, the results changed dramatically.

In some settings, making cells more "fluid-like" (a state where they move around easily) made the tissue break apart faster. In other settings, it made the tissue hold together tighter. This means that the way you set up the rules for how cells pull apart can completely flip the outcome of the simulation. The researchers showed that without carefully calibrating these detachment forces, you can't trust the model to tell you whether a tissue will stay together or fall apart.

The paper concludes that for scientists studying how tissues fracture or how cells separate, using this finite Voronoi model requires a very careful setup. You can't just plug in numbers and hit "run." You have to fix the infinite force glitch and then choose the right way to match the model to real-world physics. If you do, the model becomes a powerful tool for understanding how tissues break. If you don't, you might just be watching a computer glitch play out, thinking it's a biological discovery. The authors have even released their own software package, called PyAFV, to help others run these simulations correctly, ensuring that future studies on tissue fracture are built on solid ground rather than mathematical quicksand.

Technical Summary: Divergence of Detachment Forces in the Finite Voronoi Model

Problem Statement

The finite Voronoi (FV) model is a computational framework designed to simulate nonconfluent tissues, where cells or cell clusters are separated by gaps, extending the capabilities of conventional Voronoi models that assume confluent (space-filling) monolayers. While the FV model offers a computationally efficient alternative to phase-field or deformable polygon models for studying cohesive clusters and tissue rupture, this paper identifies a critical pathological behavior: the unphysical divergence of detachment forces.

When the line tension on cell-medium interfaces ( $\lambda^{(n)}$ ) exceeds the tension on cell-cell contacts ( $\lambda^{(c)}$ ), the model predicts that the force required to separate two cells diverges as the contact length approaches zero. This divergence arises from the geometric constraint that cell boundaries are composed of straight Voronoi edges and fixed-radius circular arcs. Consequently, numerical simulations of tissue fracture exhibit a strong, unphysical dependence on the integration time step ( $\Delta t$ ). In the limit of small $\Delta t$ , the numerical evolution fails to "skip" the force divergence, causing cells to appear artificially adherent and preventing cluster rupture, whereas larger time steps allow unphysical detachment.

Methodology

The authors employ a multi-pronged analytical and numerical approach to diagnose and resolve this issue:

Analytical Derivation: The authors analyze a simplified system of a cell doublet (two identical cells). By deriving the interaction force as a function of the separation distance, they demonstrate that the force diverges as the cell centers approach the maximum separation distance ( $d \to 2\ell$ ). This divergence is traced to the derivative of the non-contacting perimeter with respect to the separation parameter, which becomes singular due to the specific geometry of the outer vertices where straight edges meet circular arcs.
Regularization: To address the divergence, the authors introduce a cutoff parameter $\delta$ . This modifies the calculation of the outer vertices by bounding the vanishing denominator in the force derivative, effectively truncating the force-distance curve before it diverges. This introduces a new microscopic length scale representing a threshold contact length below which the strict Voronoi shape assumption is considered invalid.
Calibration against Deformable Polygon (DP) Models: To determine appropriate values for the regularization parameters (specifically the maximum radius $\ell$ $ℓ$ and the cutoff $\delta$ $δ$ ), the authors compare the FV model against a Deformable Polygon (DP) model. The DP model shares the same energy function (area elasticity, perimeter elasticity, and line tensions) but allows cell boundaries to deform continuously without the geometric constraints of the Voronoi tessellation.
- Strategy 1: The FV model's maximum radius $\ell$ is set to the energy-minimizing radius $\ell_0$ derived from the DP model, and a fixed cutoff $\delta$ is applied.
- Strategy 2: The FV model is calibrated such that its detachment force matches the detachment force of the DP model, requiring an adjustment of $\delta$ for different parameter sets.
Simulation of Fracture Transitions: Using the regularized and calibrated models, the authors simulate the fracture–no-fracture transition in nonconfluent tissues. They measure the median survival time ( $t_{1/2}$ ) of cell clusters as a function of preferred perimeter ( $P_0$ ), tension difference ( $\Lambda$ ), and cell motility ( $v_0$ ).

Key Contributions

Identification of a Geometric Pathology: The paper demonstrates that the divergence of detachment forces in the FV model is an intrinsic property of its geometric assumptions, not a numerical artifact of the integration scheme. This explains why previous simulations showed time-step-dependent fracture behaviors.
Regularization Scheme: The authors propose a simple, physically motivated regularization (Eq. 10) that bounds the detachment force, rendering the model numerically stable and independent of the time step.
Calibration Framework: Two distinct calibration strategies are presented to align the FV model with the more physically realistic DP model. The paper highlights that the choice of calibration (specifically how the maximum radius $\ell$ and cutoff $\delta$ are set) fundamentally alters the predicted mechanics of cell separation.
Code Availability: The authors provide the PyAFV Python package to facilitate the reproduction of these results and future applications of the regularized model.

Key Results

Time-Step Dependence: In the unregularized FV model, decreasing the time step $\Delta t$ systematically suppresses cluster rupture, leading to an artificial increase in the median survival time $t_{1/2}$ . This confirms that fracture events in the original model were driven by numerical "skipping" of the force divergence.
Force Divergence: Analytical results confirm that for $\Lambda > 0$ , the detachment force diverges as the contact length vanishes. The regularization successfully removes this divergence, resulting in finite detachment forces.
Impact of Calibration on Phase Diagrams: The choice of calibration strategy qualitatively changes the fracture–no-fracture phase diagram:
- Uncalibrated/Unregularized: Increasing the preferred perimeter $P_0$ (making cells more fluid-like) facilitates fracture, consistent with previous unregularized studies.
- Strategy 1 (Fixed $\delta$ ): After calibrating $\ell$ to the energy-minimizing value $\ell_0$ , the trend reverses: increasing $P_0$ makes clusters harder to rupture.
- Strategy 2 (Matching Detachment Forces): This strategy yields detachment forces that are generally higher than Strategy 1, making clusters even more resistant to fracture. The dependence of fracture times on $P_0$ becomes weak.
Role of Detachment Mechanics: The study concludes that the fracture–no-fracture transition in nonconfluent tissues is governed by near-detachment mechanics. The sign of the transition's dependence on cell shape ( $P_0$ ) is not an intrinsic property of the tissue but is determined by how the detachment forces are regularized and calibrated.

Significance and Claims

The authors claim that for studies focused on fracture or intercellular adhesion in nonconfluent monolayers, a physically motivated calibration of near-detachment mechanics in the finite Voronoi model is essential. Without regularization and proper calibration, the model produces spurious results where physical outcomes (such as whether a tissue fractures) depend on arbitrary numerical choices like the time step.

The paper asserts that the divergence of detachment forces is a fundamental limitation of the strict Voronoi geometry when applied to nonconfluent systems with differing interface tensions. By introducing the cutoff and providing calibration strategies, the authors restore the FV model as a reliable tool for simulating tissue rupture. They emphasize that the choice of regularization parameters can qualitatively alter phase diagrams, suggesting that previous interpretations of the clustered-to-dispersed transition in similar models may need re-evaluation depending on the specific calibration used. The work does not claim to solve all issues in tissue modeling but provides a necessary correction to make the FV model physically consistent for detachment events.