Virtual Cell modelling and simulation software environment

Abstract

The Virtual Cell (VCell; http://vcell.org/) is a problem solving environment, built on a central database, for analysis, modelling and simulation of cell biological processes. VCell integrates a growing range of molecular mechanisms, including reaction kinetics, diffusion, flow, membrane transport, lateral membrane diffusion and electrophysiology, and can associate these with geometries derived from experimental microscope images. It has been developed and deployed as a web-based, distributed, client-server system, with more than a thousand world-wide users. VCell provides a separation of layers (core technologies and abstractions) representing biological models, physical mechanisms, geometry, mathematical models and numerical methods. This separation clarifies the impact of modelling decisions, assumptions and approximations. The result is a physically consistent, mathematically rigorous, spatial modelling and simulation framework. Users create biological models and VCell will automatically (i) generate the appropriate mathematical encoding for running a simulation and (ii) generate and compile the appropriate computer code. Both deterministic and stochastic algorithms are supported for describing and running non-spatial simulations; a full partial differential equation solver using the finite volume numerical algorithm is available for reaction-diffusion-advection simulations in complex cell geometries including 3D geometries derived from microscope images. Using the VCell database, models and model components can be reused and updated, as well as privately shared among collaborating groups, or published. Exchange of models with other tools is possible via import/export of SBML, CellML and MatLab formats. Furthermore, curation of models is facilitated by external database binding mechanisms for unique identification of components and by standardised annotations compliant with the MIRIAM standard. VCell is now open source, with its native model encoding language (VCML) being a public specification, which stands as the basis for a new generation of more customised, experiment-centric modelling tools using a new plug-in based platform.

Figure 1

Decomposing the modeling process in VCell: the relationship between the principal components of the BioModel workspace are shown. A single Physiology can be used in different contexts, as described by one or more Applications. Every Application is translated into a unique Math Description, which in turn is the basis for running one or more different Simulations for each of the Applications.

Figure 2

Automatic import of reactions into VCell from the KEGG database.

Figure 3

The current VCell platform. The overall architecture of the web-based distributed system is shown, including the software components developed in house (white and blue) and third-party middleware (yellow). All the components to the left of the separator line are deployed as a single Java application on client (user’s) computers.

Figure 4

A time sequence illustrating membrane diffusion of a molecular species (D=10μm2/s) on a 3-D cellular membrane after “bleaching” in a prototypical FRAP experiment. Each frame is labeled with time in seconds from beginning of experiment. Time 0.0 shows an evenly distributed concentration of a diffusible molecule (in red, high concentration) followed by a “bleach” event at time 2.0 creating a localized concentration deficit (in blue, low concentration) and subsequent diffusion at times 2.5–4.0 from un-bleached membrane to bleached area causing a concentration gradient (green, yellow) to form. In addition to illustrating the results of the VCell lateral diffusion algorithm, this also exemplifies the data display and export features.

Figure 5

Stochastic simulations with the Virtual Cell: a VCell model of an enzymatic reaction (left); a deterministic solution for the enzyme-substrate complex ES is superimposed over ten stochastic trajectories (right); histogram of the number of copies at t = 4.0 obtained from 1000 trials.