Virtual Cell Biologically Oriented Interface

Purpose

The Virtual Cell, designed to be accessible to the experimental biologist, is a fully modular computational framework that provides a general approach to modeling the spatially organized and interdependent chemical events that underlie dynamic cellular processes. A Java biological interface allows experimentalists to input specific geometry, in the form of experimental images, specify compartment topology, define and assign species, chemical reactions and transport kinetics, and computational mesh.

The user interface for the Virtual Cell allows full interaction from the WWW. This interface facilitates the creation and analysis of abstract models, generation of specific simulations, control and monitoring of simulations, and data analysis. This interface aims to present the physiology of chemical and structural phenomena and mechanisms in a natural and consistent manner.

Current Implementation

The current Virtual Cell application utilizes a newly designed system-level architecture based on unified modeling language (UML). The system is decomposed into an application framework and system services. The application framework is the Modeling Framework. The system services are the Database Service, the Simulation Control Service (which encapsulates the Simulation Library), and the Simulation Data Service. The architecture is designed such that the location of the user interface and the corresponding back end services are transparent to the majority of the application. The typical configuration is a Java applet running in a WWW browser, with the Database, Simulation Control, and Simulation Data services executing on a remote machine (WWW server). Alternatively, the software may be executed as a standalone application on a local machine with the requirement that the Java Runtime Environment and a C++ compiler are installed.

Physiological Models

The physiology section allows for the specification of cellular structures (topology) and molecular species (with default initial concentrations and diffusion rates) within the cellular model under study. The model is further defined by a collection of biochemical reactions acting on molecular species within the cellular structures. Biochemical reactions are objects that represent complete descriptions of reaction stoichiometry and kinetics. Reactions can be either a collection of related reaction steps occurring at or near a single structure, or trans-membrane fluxes. Membranes are automatically generated when a structure is enclosed within another structure.

Geometric Models

A geometric model represents the morphometry of a particular cell or a portion of a cell. Such a model is required to represent the cellular system. The geometry may be captured by a variety of imaging modalities, or it may be analytically defined for very regular structures or symmetric cells. Internal structures that are included in the physiological model, but are not spatially resolved, may be mapped continuously as a volume fraction of the enclosing compartment. The geometry may be specified as compartmental, 1-D, 2-D or 3-D segmented images. Segmentation is necessary to define discrete regions that represent the anatomical features of interest. Membranes are implicitly represented as the boundary between dissimilar compartments. A compartmental model represents a single point simulation based on the defined physiological model, the surface to volume ratio and volume fractions. Compartmental models are useful in calculating quick approximate solutions.

Applications

The regions defined in the geometry model are mapped to structures defined in the physiology model. Compartmental models represent a single point simulation and multiple regions within such a model are still mapped to a single compartment. The simulation context allows for the specification of spatially accurate initial conditions and boundary conditions for each species of each structure. It is also a means of isolating simulation specific assumptions from the rest of the model. Electrical mapping is also defined within the application component of the software.

Simulations

The Equation Viewer displays the equations generated as a result of mapping the physiological model to either a cellular geometry model (spatial simulation) or a single point approximation (compartmental model). The parameter values may be substituted (and the expression simplified), or left in their symbolic representation.

Single Point Approximation

The Compartmental Simulation component executes a compartmental (single point) simulation based on the defined physiological model and the geometric assumptions entered in the Structure mapping (surface to volume ratios and volume fractions). This results in a set of nonlinear ordinary differential equations that typically are solved in seconds. This allows an interactive, though manual, modification of parameters and a quick determination of the effect over time. Once the simulation is complete, each species can be viewed easily, and includes local sensitivity analysis. The local sensitivity analysis is computed at the nominal trajectory of the ODE system and performs a direct solution for all sensitivities of a single parameter. It uses exact Jacobian's and Rate Sensitivities (Symbolic) and displays the normalized (log) sensitivities as a function of time. For small parameter changes, the state trajectory is extrapolated.

Spatial Simulation

Partial differential equations that correspond to diffusive species and ordinary differential equations for non-diffusive species are generated for the solution of a complete spatial simulation. The equations are sent to a remote server where an executable is generated, run and the results are collected and stored. The simulation data server coordinates the client access to the server-sided simulation for display and analysis. The analysis capability available includes graphing the spatial distribution of a species as a line scan and graphing a time series at a single point.

Storage

The current implementation allows whole physiological models, geometric models, and simulation contexts to be stored and retrieved. The simulation context is stored such that it encapsulates all of the information to reproduce and describe a particular spatial simulation. There is, however, no ability to querying the stored models for specific attributes. The models are currently stored intact using Java's Object Serialization capability. Experimental segmented images are uploaded to the image database. Within the database the user can assign spatial scales and identify regions based on anatomical features.

Data Export , Analysis, and Interoperability

The simulation results are in the form of multivariate, time series data that can have up to 3 spatial dimensions. To allow flexibility in analysis, exporting facilities have been introduced. Data may be exported as spreadsheet data (comma separated value .csv), image files (GIF), or movies (Quicktime or Animated GIFs). The capability for exporting models from the Virtual Cell format to SMBL Levels 1 and 2 is also currently available and work is being done to accommodate the translation into CellML.

Equations from the Virtual Cell can be exported into MATLAB (version 5 or 6) and reports can be generated via .pdf, .rtf or .html.