A modeling and simulation language for biological cells with coupled mechanical and chemical processes

Biological cells are the prototypical example of active matter. Cells sense and respond to mechanical, chemical and electrical environmental stimuli with a range of behaviors, including dynamic changes in morphology and mechanical properties, chemical uptake and secretion, cell differentiation, proliferation, death, or migration. Modeling and simulation of such dynamic phenomena poses a number of computational challenges. A modeling language to describe cellular dynamics must be able to naturally represent complex intra and extra-cellular spatial structures, and coupled mechanical, chemical and electrical processes. In order to be useful to domain experts, a modeling language should be based on concepts, terms and principles native to the problem domain. A compiler must then be able to generate an executable model from this physically motivated description. Finally, an executable model must efficiently calculate the time evolution of such dynamic and inhomogeneous phenomena. We present a spatial hybrid systems modeling language, compiler and mesh-free Lagrangian based simulation engine which will enable domain experts to define models using natural, biologically motivated constructs and to simulate time evolution of coupled cellular, mechanical and chemical processes acting on a time varying number of cells and their environment.

💡 Research Summary

The paper tackles the formidable challenge of modeling and simulating biological cells, which are quintessential examples of active matter. Cells constantly sense mechanical, chemical, and electrical cues from their fluidic environment and respond with a rich repertoire of behaviors—morphological remodeling, changes in mechanical stiffness, secretion and uptake of signaling molecules, differentiation, proliferation, apoptosis, and migration. Capturing these intertwined processes in a single computational framework requires a language that can (i) express complex intra‑ and extracellular spatial structures, (ii) naturally encode coupled mechanical, chemical, and electrical phenomena, (iii) be written using terminology familiar to biologists, (iv) be compiled into an efficient executable, and (v) run on a simulation engine capable of handling a time‑varying number of cells and a deformable environment.

To meet these requirements, the authors introduce Mechanica Modeling Language (MML) together with a dedicated compiler and a mesh‑free Lagrangian simulation engine (the “Mechanica” platform). MML is built around two fundamental abstractions: objects and processes. Objects can be primitive (scalars, integers, booleans) or composite types specially designed for mesoscopic biology—particles, links, concentrations, amounts, spatial regions, and spatial fields. Each object carries a type‑defined state space, an identity, and optional attributes such as position, mass, volume, or orientation. Particles represent “parcels” of material rather than individual atoms; they may be point‑like or have complex shapes. Links connect two or more particles, generalizing bonded interactions from molecular dynamics (MD). Unlike fixed MD bonds, links in MML can be created, broken, and assigned arbitrary user‑defined force functions, enabling everything from simple Hookean springs to sophisticated dipole, quadrupole, or octopole interactions.

Processes are the dynamic engines of the language and come in two flavors. Continuous processes describe rate‑based transformations akin to chemical reaction kinetics. They accept a set of input variables (e.g., concentrations attached to particles) and produce outputs at a user‑specified rate. Because concentrations are attached to moving particles, advection is automatically handled: as a particle translates, its chemical content moves with it, and concentration values are automatically scaled by the particle’s volume. Discrete processes mirror SBML events: they are triggered by Boolean predicates, have an associated probability (default = 1), and execute atomically, instantly updating variable values. Crucially, discrete processes can also create or destroy objects, thereby modeling cell division, apoptosis, or the emergence of new structures.

MML’s syntax draws from a functional subset of Microsoft TypeScript, incorporates scene‑description constructs from POVRay for spatial definitions, and uses pattern‑matching rules inspired by BioNetGen, OCaml, and Mathematica. This hybrid design lets modelers write concise, declarative specifications: for example, a link definition link(a,b){ -k*(1-dist(a,b)) } creates a Hookean spring with rest length 1 between particles a and b. Material regions (MaterialRegion) define bounded volumes (e.g., a sphere) and can be automatically filled with particles of a given type using a fill function. When a region is homogeneous, attached concentration or amount variables are uniform throughout; when the region deforms, volume‑preserving links can enforce incompressibility.

The compiler parses MML scripts, builds an intermediate representation, and generates code for the underlying simulation engine. The engine employs a mesh‑free Lagrangian particle method, allowing forces, chemical reactions, diffusion, and volume conservation to be evaluated concurrently without constructing a global mesh. This approach sidesteps the scalability limits of traditional MD (fixed particle count, limited force libraries) and of stochastic particle simulators such as MCell, Smoldyn, or ML‑Space (which typically assume static geometry and lack mechanical interactions).

The authors systematically compare MML/Mechanica with existing approaches:

• SBML – excellent for well‑stirred reaction networks but limited spatial support and no mechanics.
• Petri Nets / Colored Petri Nets – capture concurrency and discrete events but lack dynamic geometry and force modeling.
• Traditional MD (e.g., LAMMPS) – atomistic detail, fixed particle numbers, predefined force fields; unsuitable for mesoscopic cell scales.
• Reactive MD (ReaxFF, ReaDDy, SRSim) – add bond formation/breakage and reactions but still limited to simple particle attributes and often require hand‑coded C++/Java force definitions.
• Stochastic particle simulators (MCell, Smoldyn, ML‑Space) – handle reaction‑diffusion in complex geometries but cannot model mechanical forces or dynamic morphology.
• tDPD (transport Dissipative Particle Dynamics) – merges particle and continuum descriptions, supports chemical concentrations, but lacks a dedicated modeling language and discrete event handling.

Mechanica builds on the strengths of tDPD (particle‑based transport, continuous fields) while filling the gaps: it provides a high‑level declarative language, supports both continuous reactions and discrete events, allows dynamic creation/destruction of particles, and integrates mechanical links with user‑defined force laws.

Key contributions of the work are:

1. Domain‑specific language that lets biologists express mechanical, chemical, and electrical concepts directly, without low‑level programming.
2. Unified object‑process framework that simultaneously handles continuous reaction‑diffusion and discrete event dynamics.
3. Mesh‑free Lagrangian simulation engine capable of efficiently evolving large, deformable particle ensembles with attached fields.
4. Comprehensive comparison demonstrating functional advantages over SBML, Petri Nets, MD, reactive MD, stochastic particle simulators, and tDPD.

The paper concludes that the Mechanica platform enables researchers to construct mechanistic, multiscale models of cells and tissues that were previously infeasible. By providing a natural, biologically motivated modeling language coupled with a performant simulation backend, the work opens the door to systematic exploration of coupled mechano‑chemical phenomena across scales—from intracellular signaling to tissue morphogenesis.