Dizzy

Category Cross-Omics>Pathway Analysis/Gene Regulatory Networks/Tools and Cross-Omics>Agent-Based Modeling/Simulation/Tools

Abstract Dizzy is a software tool for stochastically and deterministically modeling the spatially homogeneous kinetics of integrated large-scale genetic, metabolic, and signaling networks.

Notable features include a modular simulation framework, reusable modeling elements, complex kinetic rate laws, multi-step reaction processes, steady-state noise estimation, and spatial compartmentalization.

Modular simulation framework --

Dizzy has a modular design in which a simulator is a plug-in that conforms to a well-defined software interface. Each simulator is implemented as a self-contained unit that creates all of the internal data structures it needs to function.

This allows for a variety of simulation techniques to be applied to a single model description, and for the clean separation of the simulation method from the model description.

The model definition is focused on the biochemical semantics of defining chemical species and reactions. The technique and parameters for simulating the model specified in the simulation controller, and do Not require changes to the model.

Dizzy includes both stochastic and deterministic simulators. The stochastic simulators are discrete-event or multiple-event Monte Carlo algorithms. The deterministic simulators model the dynamics as a set of ordinary differential equations (ODEs), which are solved numerically.

One benefit of this modular design is that one may use a deterministic ODE-based solver for optimization and parameter fitting, and switch to a stochastic simulation technique for exploring the stochastic dynamics, once the model parameters have been established.

This modularity also simplifies the task of implementing a new simulator and integrating it into the system.

Simulators available in Dizzy --

Dizzy includes an efficient implementation of a stochastic simulator based on Gillespie’s Direct Method. It uses the Monte Carlo technique to generate an approximate solution of the master equation for chemical kinetics.

In this method, simulation time is advanced in discrete steps, with precisely one reaction occurring at the end of each discrete time-step. Both the time steps and the reaction that occurs are random variables.

Dizzy also implements a stochastic simulator based on Gibson and Bruck’s Next Reaction Method. The computational cost of this Monte Carlo-type method scales logarithmically with the number M of reaction channels, in contrast with the Gillespie algorithm which scales linearly with M.

The manufacturers have implemented a tree traversal technique to analyze a rate expression for a chemical reaction that has a complex kinetic rate law, in order to ascertain the dependence of the rate expression upon the various chemical species in the model.

This permits applying the Gibson-Bruck method to models that implement complex kinetic rate laws.

Two stochastic simulators based on Gillespie’s Accelerated Approximate Method (here referred to as the “Tau-Leap” Method) have been implemented in Dizzy. The Tau-Leap Method is a stochastic process that approximately solves the chemical master equation, based on a controllable, dimensionless error parameter.

Two (2) versions of the Tau-Leap algorithm have been implemented Tau-Leap Complex and Tau-Leap Simple.

The Tau-Leap Simple algorithm is intended for simple models entirely composed of reaction channels with mass-action kinetics.

The Tau-Leap Complex algorithm is a novel adaptation of Tau-Leap that is intended for use with models with complicated (e.g., enzymatic) rate expressions whose partial derivatives are very expensive to evaluate symbolically.

Dizzy also includes a deterministic simulator based on a fifth order Runge-Kutta ODE solver. Step size is adaptively controlled, based on a fourth order error estimation formula. Both relative and absolute error tolerances may be independently specified, as well as the initial step size.

Although Runge-Kutta is Not state-of-the-art for high-accuracy integration, it is particularly useful for models in which a derivative function is discontinuous.

Two additional deterministic simulators have been implemented based upon the odeToJava ODE solver package by Patterson and Spiteri.

This package includes a Dormand-Prince fourth/fifth order solver with adaptive step size control. It also contains a Runge-Kutta implicit-explicit ODE solver that is useful for systems with a high degree of stiffness.

Templates-reusable and hierarchical model elements --

Dizzy’s model definition language permits the definition of reusable, parameterized model elements called templates. This enables the construction of a prepackaged library of templates that can simplify the task constructing a complex model.

Templates are hierarchical in the sense that a template definition may contain a template instance. In such a case, the inner template namespace is nested within the containing template’s namespace.

A file inclusion mechanism allows the separation of template definitions from the model definition. This facilitates the development and re-use of a library of previously defined and curated model elements, such as genes, fractional saturation functions, etc.

Complex kinetic rate laws --

Dizzy enables the creation of reduced stochastic models containing reactions whose propensities may be expressions of arbitrary complexity, representing the average effect of underlying reaction steps that are in quasi-steady-state (QSS).

This permits efficient approximate modeling of enzyme-catalyzed reactions and other processes for which the overall kinetic rate is more complicated than mass-action kinetics.

Multi-step and delayed reaction processes --

Dizzy enables the simulation of complex multi-step processes such as elongation and translocation during transcription or translation, through two (2) methods. One may define a “multi-step” reaction process, or a reaction process with an intrinsic, phenomenological time delay.

Estimation of steady-state stochastic noise --

Dizzy provides a feature for estimating or calculating the steady-state stochastic fluctuations of the species in a biochemical model, requiring only the solution of the deterministic dynamics.

Integrated, graphical, and portable software framework --

The notable software features of Dizzy are: portability across many computer architectures, integration with external software programs, and a graphical user interface (GUI).

Dizzy is implemented in the Java programming language, which enables Dizzy to execute on any computer platform for which a Java 2 Runtime Environment of version 1.4.1 or newer, is available. Dizzy has been optimized for efficient numerical computation in the Java Runtime Environment.

Dizzy is capable of simulating models expressed in the Systems Biology Markup Language (SBML). Dizzy can also export a model into SBML. For optimal performance, a model should be written in the Dizzy model definition language, rather than imported from SBML.

The SBML import feature enables Dizzy to simulate a model constructed with the BioTapestry software program (see G6G Abstract Number 20385).

Three (3) of the simulation algorithms within Dizzy may be invoked through the Systems Biology Workbench (SBW) - (see G6G Abstract Number 20453); the Gibson-Bruck, Gillespie Direct, and Runge-Kutta ODE simulators.

These simulators may also be easily invoked from any SBW-enabled model development platform, such as CellDesigner - (see G6G Abstract Number 20159) or JDesigner.

Dizzy provides a menu-driven graphical user interface. This user interface includes screens for simulation control, model editing, plotting simulation results and browsing/searching the hypertext User manual.

Visualization of a biochemical model in a graphical representation is enabled through a software bridge to the Cytoscape software system (see G6G Abstract Number 20092).

System Requirements

Contact manufacturer.

Manufacturer

Manufacturer Web Site Dizzy

Price Contact manufacturer.

G6G Abstract Number 20645

G6G Manufacturer Number 104243