Multiscale Modeling in MOOSE: Interfaces, Interoperability and Standards

Introduction
Multiscale Object Oriented Simulation Environment (MOOSE) is a general purpose biological
simulator. It runs on multi-core as well as multi-node computer systems while making
load-balancing and messaging transparent to the user. It is multiscale in the sense of
handling simulations from molecular to large network scales, events from microseconds to
days, and in terms of running on hardware scaling from laptops to large clusters. It
provides a Python based interface and can be used synergistically with other libraries
and simulators that use Python. It has a graphical-user interface that allows for easy
plotting, and 3D visualization of complex models and their state.

Multiple scales of modeling
The scales in biology can range from a few molecules bouncing around in a vesicle to
networks of thousands of neurons in brain regions. The times can be anywhere between
microseconds to days (or millenia for evolutionary biologists). Fast solvers have been
implemented/interfaced for reaction-diffusion chemical kinetics (GNU Scientific Library),
stochastic chemical kinetics for small volumes (Gillespie algorithm), spatial Monte Carlo
calculations for individual molecules (Smoldyn) and realistic compartmental modeling
of neurons (Hines’ algorithm). A key area of development in MOOSE is to integrate models
of signaling pathways with compartmental models for studying emergent properties at
the interface between biochemical and electrical signaling. MOOSE presents an intuitive
object-oriented interface to the user, while transparently handling fast calculations with
specialized numerical engines for each level of detail.

Impact on standards
There are multiple standards for model specification at various levels and MOOSE supports
three of them: the GENESIS scripting language, SBML and NeuroML. Moreover, it aims to
support the Network Interchange format for NEuroscience (NineML) as the specification
matures. In the absence of a common framework to combine model components specified
in different formats, the user has to put significant effort in developing composite
models and such models remain non-standard. However, as simulating composite models
out of existing ones becomes easier, it will be important for the community to find a way
to integrate the existing standards for maximum productivity. MOOSE is one of the first
simulators with this cross-scale capability, and provides a key test-bed for implementations
of multiscale model definition standards.