January 15, 2009 (Vol. 29, No. 2)

Kristen Zanella

SimBiology Software Provides a Platform for Model Creation, Stimulation, and Analysis

Modeling has long played a key role in the drug discovery process, from isolating drug targets to screening and pharmacokinetics. Many pharmaceutical companies are currently facing patent expirations without replacement candidates in the pipeline or are struggling with high candidate attrition. Leaner pipelines result in necessary cost decreases, which can often lead to staff reductions. These struggles underscore the need for successful modeling in pharmaceutical R&D.

As a result of this elevated intensity, the needs of pharma modelers have also drastically increased. In general, pharma modelers face two main tasks: first, to complete their projects quickly and second, to communicate their results to their peers in the scientific community and to corporate managers.

Modelers spend a great deal of time integrating various tools into their workflow to create a piecemeal solution that fits their needs. They use one tool to model data, another to perform simulation, and perhaps yet another tool for analysis. Once they have results, researchers will spend an even larger amount of time transferring the results into a format that scientists or executives will understand. This workflow is inefficient and an expensive waste of time.

Creation of MathWorks’  SimBiology software was driven by customer demand to provide a modeling environment that minimizes this inefficiency. Modelers can perform their entire project workflow in a single environment, from creating a model to simulation and analysis. The interface is graphical so modelers have a pathway diagram to easily communicate with scientists and management, yet the underlying technology is mathematically strong to meet modelers’ needs.

Method of Operation

SimBiology offers several different modes of interaction, all of which are fully integrated. The SimBiology desktop is a customizable graphical user interface (GUI) that offers a point-and-click environment for modeling. Modelers can move windows around the desktop using drag and drop or menu selections.

Within the desktop, the block diagram editor allows users to work with the graphical view of their model in a drag-and-drop environment. This includes a library of predefined blocks users can use to create models. Lastly, for modelers who prefer to automate their modeling and analysis, all functionality can also be accessed from the command line.

Constructing Models

SimBiology models consist of compartments, species, parameters, reactions, events, rules, kinetic laws, and units. Users are presented with a variety of ways to create models such as dragging and dropping predefined blocks in the block diagram editor, entering equations and relevant parameters into the GUI, or by importing SBML files. Once a model has been created, the appearance of graphical entities can be altered, including replacing the entity with an image, to better visually communicate or organize a model (Figure 1). Any altered entity can be saved into a library for future reuse. Annotations such as references or notes can also be added to any entity by double-clicking on the element in the diagram editor.

Kinetic laws and units characterize the dynamics of the model. Kinetic laws and units are presented as built-in libraries or user-defined laws, and users can also add units. Variants, or alternate values for parameters, are applied to the model if users want to study various values for model components without creating a new model.

The output or response of a model such as concentration as a function of time can be determined by simulating the model with either stochastic or deterministic solvers. Results display in a figure window or can be embedded into the block diagram editor (Figure 1). Simulation events can be applied to the model to define a sudden change in model behavior such as a bolus injection given to a patient.


Figure 1. The block diagram editor allows users to drag and drop models and elements.

Analysis

Once a model is constructed, analyzing that model helps users gain insight into it. Tasks such as sensitivity analysis can be added to a model to isolate those components most likely to have an impact on the desired output. Or users can perform parameter estimation (Figure 2) to fit values of missing parameters to experimental data. Analyis is performed by selecting the relevant task, which is then added to the model browser and is executed by clicking the Run button. Users can also create user-defined tasks and include them in models, or share them with peer modelers for reuse.


Figure 2. Parameter estimation allows unknown parameters in the model tobe fitted to external data.

Communicating Results

Modeling can only be effective if the results can be communicated easily and clearly to the relevant audiences. Converting scripts or results into intuitive graphs or models is time-consuming and sometimes ineffective. Results must be in a format and language the intended recipients can decipher.

In SimBiology, users can automatically generate a customizable HTML report by clicking on model component types such as the graphical layout of the model, graphical output, or tables in the SimBiology report generator. In addition, graphical images of the model can be exported to leverage the visual nature of the model. Images from the model can also be exported as SBML files.

Modelers at pharmaceutical and biotech companies are under a great a deal of pressure to help alleviate some of the challenges that pharmaceutical researchers and developers are experiencing. The sheer nature of biological research is complex, and trying to create a model to represent biological systems is even more so. Strong modeling capabilities can bring significant value and competitive advantage to a pharmaceutical company. With the right software platform, companies will start to see the increase in efficiency that they are currently seeking.

Kristen Zannella ([email protected]) is biotech and pharmaceutical industry marketing manager at The MathWorks. Web: www.mathworks.com.

Previous articleJ.P. Morgan Healthcare Conference—Monday, January 12 Update
Next articleInvestigators Use DNA Repair Patterns to Determine Pancreatic Cancer Risk