Biophysical Modeling: Methods

Our efforts span multiple methodologies:

Agent-based models and software

We build discrete, agent-based models to capture the microscopic dynamics of cellular components, incorporating steric interactions, binding–unbinding kinetics, and active force generation. We then run large-scale simulations of these particle ensembles efficiently on high-performance computing architectures (see aLENS).

Explicit coarse-graining

We place strong emphasis on systematically coarse-graining microscopic models into continuum descriptions through kinetic and statistical-mechanical approaches, providing a rigorous link between microscopic mechanisms and system-scale behavior.

Continuum models and mixed formulations

We design advanced numerical solvers for our continuum models (PDEs) that can handle complex geometries, viscoelastic materials, moving interfaces, and strongly nonlinear constitutive laws, such as those arising in the cell cytoskeleton. We also develop mixed Eulerian–Lagrangian formulations to study systems where slender, flexible filaments (e.g., microtubules) deform, interact, and are transported within a surrounding fluid environment (see SkellySim).

Data-driven models and inference

We develop statistical and machine-learning methods to infer biophysical models from noisy data and to construct efficient, physics-aware surrogate operators that mitigate the computational cost of large-scale simulations. We also leverage such techniques to infer parameters and validate hypotheses in our existing models. By combining computational modeling with data-driven methods, we aim to bridge the gap between theory and experiment in biophysical systems.