An agent-based model of wound healing in fibroblast scratch assays
- 1University of Michigan
Abstract
Wound healing assays are widely used to study fibroblast migration and proliferation in the lab. Although traditional scratch assays provide valuable insight, they can be time-intensive, inflexible, and difficult to reproduce. Computational models provide an efficient and reproducible way to study fibroblast behavior in silico. We developed an agent-based simulation of fibroblast scratch assays that models migration, proliferation, contact inhibition, cell-cell mechanical interactions, and the extracellular matrix (ECM) through a particle representation of fibroblasts and the ECM. To validate our model, we performed two tests: a leave-one-out test, where model parameters were calibrated to 6 imaging time series and tested on the remaining one, and a prediction test, where model parameters were calibrated to the first 24 hours of data and the remaining 66 hours of data were predicted. In our leave-one-out test, our model achieved a mean normalized root mean squared error (mean NRMSE) of 3.34%, and a Pearsons R of > 0.98 across all tested wound geometries time series. In the prediction test, the model predicted the remaining 66 hours of wound closure with an NRMSE of 2.08%. These results demonstrate that experimentally observed wound healing behavior of fibroblasts across different geometries can be modeled by local, agent-based rules. Our model provides an efficient platform for the experimentation and in silico study of fibroblast behavior.
Related
- Proprietary Processing of Placental Tissue Allografts Influences Wound Healing Trajectories in a Rodent Excisional Skin Wound Model
- Implications for Faster Healing with Fewer Applications: Sustained Graft Persistence and Favorable Wound-Healing Biology of a Proprietary Processed Placental Tissue Allograft
- CsGATA5‐CsTRYS/CsTDC2 module coordinates tryptophan‐mediated wound healing in postharvest citrus fruit
- CD30 ligand promotes skin wound healing through keratinocyte signaling and Vγ6+ γδ T-cell responses
- Programmable DNA-Silk hydrogel with temporal mechanical reinforcement for mesenchymal stem cell differentiation and accelerated wound healing
BibTeX
@misc{sumanaweera2026agentbased,
title = {{An agent-based model of wound healing in fibroblast scratch assays}},
author = {R. T. K. Sumanaweera and C. Stansbury and W. Meixner and J. Cwycyshyn and l. Muir and I. Rajapakse},
howpublished = {bioRxiv},
year = {2026},
month = oct,
doi = {10.64898/2026.10.15.759485},
url = {https://www.biorxiv.org/content/10.64898/2026.10.15.759485v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - An agent-based model of wound healing in fibroblast scratch assays
AU - R. T. K. Sumanaweera
AU - C. Stansbury
AU - W. Meixner
AU - J. Cwycyshyn
AU - l. Muir
AU - I. Rajapakse
PY - 2026
DA - 2026/10/15
PB - bioRxiv
DO - 10.64898/2026.10.15.759485
UR - https://www.biorxiv.org/content/10.64898/2026.10.15.759485v1
AB - Wound healing assays are widely used to study fibroblast migration and proliferation in the lab. Although traditional scratch assays provide valuable insight, they can be time-intensive, inflexible, and difficult to reproduce. Computational models provide an efficient and reproducible way to study fibroblast behavior in silico. We developed an agent-based simulation of fibroblast scratch assays that models migration, proliferation, contact inhibition, cell-cell mechanical interactions, and the extracellular matrix (ECM) through a particle representation of fibroblasts and the ECM. To validate our model, we performed two tests: a leave-one-out test, where model parameters were calibrated to 6 imaging time series and tested on the remaining one, and a prediction test, where model parameters were calibrated to the first 24 hours of data and the remaining 66 hours of data were predicted. In our leave-one-out test, our model achieved a mean normalized root mean squared error (mean NRMSE) of 3.34%, and a Pearsons R of > 0.98 across all tested wound geometries time series. In the prediction test, the model predicted the remaining 66 hours of wound closure with an NRMSE of 2.08%. These results demonstrate that experimentally observed wound healing behavior of fibroblasts across different geometries can be modeled by local, agent-based rules. Our model provides an efficient platform for the experimentation and in silico study of fibroblast behavior.
ER -