Live model · simplified hypothesis model, ported from gbr_topo.py
Groove Versus Field, Live
Running in your browser · exploratory, not calibrated
A simplified agent-based model — adhesion, repulsion, neighbour alignment, free-edge migration, sensitivity adaptation — mechanistically inspired by selected prior studies and one preprint, ported to run live. Change a cue below and watch the tissue reorganize in real time.
What this is, and isn't. This page is not a finalized PhD membrane design — it explores one relatively stable question underneath it: how physical or topographical cues on a biomaterial surface might guide gingival cell organization, collective migration, and soft-tissue closure. Early GBR membrane exposure is one possible clinical motivation, but the polymer or biomaterial composition, layer architecture, groove geometry and dimensions, fabrication technique, cell model, and any antibacterial or osteogenic components are all still open. The bioelectric field explored here is primarily a comparison tool inspired by the Cohen Lab's electrotaxis work, not a component already committed to a final design. Treat everything below as a hypothesis-generating exploration of one sub-problem, not a prediction of which final membrane will work.
Collective closure, in real time
t = 0 steps · N = 0 cells
Cell agents (colour = health)Groove axisField direction (biases electrotactic migration inward)Wound region (only this fixed circle is measured)Reference annulus (distant, same global cues — not cue-free)
Open wound area (occupancy proxy)100%
Distant reference gap (same cues)0%
Radial order (global)+0.00
Radial order (fixed peri-wound annulus)+0.00
Groove alignment (fixed peri-wound annulus)+0.00
Stress burden (time-avg.)0.0%
Comparison scenarios
Groove topography
00.30.825 µm
Depth is used here as a phenomenological proxy for groove-guidance strength. Groove width is not modeled, and the monotonic depth–torque relationship is not fitted to Doll et al. or to a specific GBR membrane material.
Bioelectric field
00.51248
Cell division
Model interpretation
What supports the model — and what remains open.
This isn't a black box, but it also isn't a validated replica: each component below is motivated by a specific prior study or preprint, this page approximates it with a much simpler discrete-agent model, and every place the model stays qualitative is named, not hidden.
02 · Evidence base
Candidate model components, motivated by prior studies and one preprint
Collective, correlated motion
Found that epithelial cells don't move as independent particles, but the specific pattern is history-dependent, not just size-dependent: tissues that expanded from a smaller starting size develop tissue-spanning vortices in their interior from feedback between a cell's own heading and its neighbours' motion, while tissues that started at (or reached) the same size without that expansion history instead show different density profiles and arrested proliferation. So "cells move collectively" is the qualitative takeaway, not a claim that a tissue of any given size shows this vortex behaviour — it depends on how the tissue got there, which this page's fixed initial disc doesn't model at all. Their model is also a continuum active-polar-fluid description; this page instead uses discrete self-propelled agents with pairwise forces, so the citation supports the biological motivation, not a shared implementation — each agent here pushes/pulls on its neighbours and blends its direction with theirs, which is the qualitative behaviour, not the same mathematics.Heinrich et al. 2020 — eLife (Cohen lab)
Field-directed collective steering
Showed that an applied electric field doesn't just redirect isolated cells — it can steer the collective migration of a whole coupled epithelial sheet, turning a moving group the way you'd steer a single cell. That's the empirical basis for testing a field as a guidance cue here at all: an external cue really can redirect an already-moving, mechanically-coupled tissue, not just individual stragglers.Cohen, Nelson & Maharbiz 2014 — Nat. Mater.
"Cellular cruise control"
Showed a tissue actively damps its own sensitivity to a steady external command the longer that command stays constant, the way cruise control settles onto one speed and stops "pushing." β-adaptation on the field torque is this page's model of that damping. The Pulsed toggle explores whether periodic off-phases let sensitivity recover under the model's own assumed adaptation/recovery dynamics — it is not an independent empirical test of cruise control. The specific adaptation rate and the 20-steps-on/20-off duty cycle used here are modeling choices inspired by that finding, not values taken from the paper — so a pulsed condition outperforming a continuous one on this page demonstrates the model's own built-in assumption, not a replication of the Breinyn result.Breinyn et al. 2024 — bioRxiv preprint, not yet peer-reviewed as of this writing
Contact guidance on microgrooved surfaces (this page's core thesis axis)
Studied human gingival fibroblasts on microgrooved Ti6Al4V ELI surfaces fabricated by UV lithography and wet etching. The strongest early alignment and spreading were observed for nominal mask widths of 7.5 and 10 µm, which produced measured groove widths of approximately 11.5 and 13.9 µm and depths of 1.6 and 2.4 µm; a broader favorable range of 7.5–20 µm nominal width and 1.6–3.2 µm depth was also discussed, with alignment decreasing for larger structures. These effects were most evident after 8 hours, and most of the statistical differences diminished (not disappeared — oriented cells were still visible on structured surfaces) after 72 hours, once cells had filled the surface. This is direct, gingival-cell-specific evidence that contact guidance from microgrooves is real and worth modeling — it does not quantitatively calibrate this model's groove dimensions or guidance strength for an as-yet-undetermined GBR membrane material (see the depth control's own note), and is the closest empirical anchor here to the actual PhD thesis axis (biomaterial topography), which the electrotaxis literature above doesn't cover on its own.Doll, Husari, Ahrens, Spindler, Guber & Steinberg 2021, "Enhancing the soft-tissue integration of dental implant abutments — in vitro study to reveal an optimized microgroove surface design to maximize spreading and alignment of human gingival fibroblasts,"J. Biomed. Mater. Res. B 109(11), 1768–1776 · DOI 10.1002/jbm.b.34836
Stress burden
Showed that over-driving a strongly adherent epithelial sheet damages it, not just speeds it up — pulling on it faster than its cell–cell junctions can accommodate. So sustained, high-force herding here carries a real stress cost (the Stress burden metric — the time-averaged fraction of agents over an arbitrary tension threshold at each step, not a measured biological damage readout), not just a speed benefit. The specific threshold value is this model's own parameter choice, not one derived from the paper — Shim et al. supports the general "over-driving costs something" idea, not this exact number.Shim, Devenport & Cohen 2021 — PNAS
The one thing added on top of that base model: a groove-guidance torque, standing in for contact guidance from the membrane's microgrooves (see the Doll et al. entry above for direct gingival-fibroblast evidence of that effect).
01 · Design logic
Why the experiment is built this way
Why circular
A round exposure has no privileged direction — every point on the rim sees the same geometry, so any directional closure bias — favoring one side over another — has to come from the cue being tested (groove or field), not from a straight edge or a corner nudging cells one way for free. It does not mean the model has zero closure bias with no cue active: the free-edge polarization already gives every condition, cue or no cue, some baseline inward pull from open space alone (see the caveats below). What circularity controls for is asymmetry, not the presence of closure itself. It also gives "aligned," "perpendicular," and "inward" one clean, unambiguous meaning at every point on the boundary.
Why the field is radial
The field always points from a cell's current spot straight at the wound centre — a convergent field, not a fixed lab-frame direction. That's an idealised, fully-circular extension of a real precedent, not a literal replay of it: Zajdel, Shim & Cohen (2021, Biosens. Bioelectron., "Come together") closed a linear 1.5mm gap between two keratinocyte sheets with a field converging on that gap. It is a different mechanism from the Cohen Lab's SCHEEPDOG platform (Zajdel et al. 2020, Cell Syst.), which steers tissue with one spatially-uniform field direction, switched over time across four electrodes — not a field that re-aims itself per cell, per position, all the way around a circle.
Why aligned vs. perpendicular grooves
A groove is a nematic cue: it locks a cell's long axis onto a line but doesn't pick which way along it to crawl — unlike the field, which is polar and does pick a direction. "Radial" and "tangential" are both idealized, wound-centred patterns — useful for asking a clean mechanistic question, but not something you could cut into a membrane before knowing exactly where the exposure will be. A third, "parallel" option is included precisely because it's the manufacturable alternative: one fixed groove direction across the whole surface, regardless of where a wound later opens.
03 · Interpretation limits
Read with these caveats
Confinement, an edge-torque fix, and pre-equilibration
A confining force at the tissue's edge makes this a genuine closure assay, not an unconfined spread. Free-edge polarisation now fires only where open space faces the tissue interior (the wound), so the outer rim no longer fights the wall as a false second wound edge. The tissue is also seeded as a full, wound-free disc, given a brief cue-free pre-equilibration (30 steps, field/groove/division off, self-propulsion and noise still on) to settle the seeding jitter, and only then cut open at the wound — so the timed run doesn't start by conflating "the wound closing" with "the freshly-seeded lattice settling."
Proliferation is off
Turning it on reproduces the uncontrolled growth seen earlier (~1,000 → 12,000+ cells in 800 steps), capped here at 3,000 for performance.
Speed isn't calibrated
The playback pace was chosen to be watchable, not to match any measured migration rate — real epithelial fronts move far slower.
Dimensionless units, no material fixed
A qualitative, mechanistic model — not fitted to gingival keratinocytes, MDCK cells, or any specific candidate biomaterial surface. PCL–HTCC, PCL/PLGA and similar are candidate examples under consideration, not a confirmed thesis material.
Occupancy proxy, not real space
"Open wound area" and "distant reference gap" both work by sampling whether a grid point falls inside any cell's drawn footprint — an oriented ellipse, matching how a cell is actually rendered, not a circle, so a cell pointing across a gap covers less of it than one pointing along its length. Because footprints are agent-sized, not a continuous field, even a mechanically settled, fully confluent monolayer reads as partly "open" — this is a sampling artifact of the occupancy test, not real physical space between cells. Closure values are sensitive to the assumed agent footprint size (currently 0.62 × 0.372 units, chosen to match the drawn cell) — no footprint-sensitivity analysis has been run, so treat exact closure percentages as this-model's-scale numbers, not calibrated measurements.
Reference ring is same-condition, not cue-free
The distant reference ring exists to make that sampling artifact visible: it is not cue-free or "undisturbed" tissue — it sits far from the wound but is still exposed to whatever groove or field cue is currently selected globally, so read it as "how open does normal, same-condition tissue look under this occupancy test," not as a zero baseline. The wound-closure percentage itself is calibrated separately: it's measured relative to the actual sampled open area right after the wound is cut (not a theoretical circle area), so closure still starts at 0% regardless of this baseline artifact.
Front-zone metrics, and closure that can go negative
Radial order and groove alignment are reported for a fixed peri-wound annulus (within 3 units of the wound boundary on either side) rather than averaged over the whole tissue — this is a geometric ring at a constant radius, not a dynamically tracked advancing front, so treat it as "activity near the wound," not literal front-line tracking. Radial order runs +1 (inward) to −1 (outward); groove alignment is nematic (+1 parallel, −1 perpendicular to the groove axis) and says nothing by itself about whether closure succeeded. Open wound area and closure can now go past 100%/below 0% respectively if a wound genuinely widens (e.g. under strong tension-driven retraction), rather than being clamped to look like it never gets worse than the starting point.
Quantitative comparison (same seed)
Runs C1–C5 head-to-head from the identical, pre-equilibrated starting tissue (current seed shown above), division and pulsing off, for a fixed number of steps. This is one run, not averaged across multiple seeds — read it as a hypothesis-generating comparison, not a statistically powered result.
Click "Run all conditions" to compare C1–C5 quantitatively at the current seed.