3. Introduction to (ODE) Modelling Best Practices

In this part of the computational physiology module we demonstrate some modelling practices that we recommend you follow when developing any mathematical model. The key principles are modularity, reuse, and reproducibility - we demonstrate those here using the tool OpenCOR and the CellML and SED-ML formats, but the principles are valid across the spectrum of computational physiology. It is also important to be able to share and collaborate with other scientists and we demonstrate this using the Physiome Model Repository (PMR).

Here are a couple of links in case you need a refer back to some of the material covered earlier in the module.

3.1. Reproducible and reusable model descriptions

In this section we introduce some of the key concepts and best practices for ensuring your mathematical models are reproducible and reusable.

3.1.1. Units

The first concept is that everything in your model should have clearly defined physical dimensions (or at least a clear statement that the quantity has no physical dimension). This ensures that when a scientist reuses a model they are able to use software tools to check for dimensional consistency and even potentially automatically convert quantities as required.

  1. Take a look through the section A model of ion channel gating and current: Introducing CellML units for a discussion and some examples demonstrating the use of units in CellML and OpenCOR.

3.1.2. Modularity

The second concept is that of modularity. By defining your mathematical model in a modular fashion with clearly defined interfaces you ensure that other scientists (and even yourself!) are able to make sure of the modules in future novel work.

  1. The section A model of the potassium channel: Introducing CellML components and connections provides an introduction to the syntax for defining modules in CellML and the mechanism for defining the communication between the modules.
  2. A model of the sodium channel: Introducing CellML encapsulation and interfaces then introduces the concept of hierarchically grouping modules as an aide to reuse and abstraction, as well as definition of module interfaces.

3.1.3. Reuse

The final key concept is that of reusing existing modules, defined using the previous concepts.

  1. A model of the nerve action potential: Introducing CellML imports introduces the mechanism by which existing modules can be reused in CellML models.

When describing your model it is useful to separate the mathematical model from its instaniation with a specific set of parameter values. This allows the model to be reused with different parameter sets for different purposes, including different simulation protocols, etc. Furthermore, rather than continually redefining the same set if units, it is better practice to simply define your units once and reuse their definition as required. Potentially even defining a standard set of units for a research group or collaboration to use and storing that in the repository for everyone to make use of.

  1. In A model of the cardiac action potential: Importing units and parameters we demonstrate the separation of model parameters and reuse of a common units definition.

3.2. Collaboration, versioning, discovery

One of the main reasons to encode your mathematical models following the principles described in the previous sections is to make the model available for future scientists (again, including yourself!) to utilise the model in novel investigations. The Physiome Model Repository (PMR) has been developed to provide scientists with a suitable repository for sharing their work with the community. Full details are provided in the PMR documentation, but here we highlight some PMR capabilities.

  1. We have previously introduced the various ways to load models from PMR into OpenCOR. In The Physiome Model Repository and the link to bioinformatics we illustrate some of the extra information that is avaialble via the PMR web interface.
  2. Choosing your favourite area of physiology, see if you can discover an interesting model in PMR and load it into OpenCOR to explore. Do you find this a more comprehensive way to understand the model than only reading the source literature?

The primary unit of information in PMR is the workspace. Each workspace is a version controlled repository which is permanently and persistently available, with each resource in the workspace assigned a unique identifier for each version of the resource submitted to the repository.

  1. In Using PMR with OpenCOR we present OpenCOR’s capabilties for directly working with workspaces in PMR.
  2. Compare the method described in step 8 with the previous method scientists had to follow in order to get models into PMR. Do you see why most scientists would prefer to email their models to Andre and let him add them to the repository?

Extra for experts: if you look through the PMR documentation you will be able to find information on how to submit your workspaces to make them public, share them directly with collaborators, and make exposures for specific revisions of the workspace.

3.3. A bond graph-based method for representing physiology

A new field of research that has emerged recently is the application of bond graph theory to the field of systems biology and physiology. This is providing the theoretical framework to guide the development of core modules that can be arbitrarily combined to capture a wide range of physiological phenomena. Combining this framework with CellML has started to provide an extremely powerful technology for both defining core modules spanning the required physics and the methods for integrating these modules into models of physiological systems.

As this is a new and rapidly evolving field of research, the documentation is still being developed. This section of the DTP Computational Physiology module will be largely be taught using powerpoint slides that will be made available on the day.

Some references that are relevant to this work are given here for convenience.

  • Paynter H. Analysis and Design of Engineering Systems (MIT, Cambridge, Mass., 1961).
  • Oster G, Perelson A, and Katchalsky A. 1971. Network thermodynamics. Nature (Lond.). 234:393 (link).
  • Gawthrop PJ and Crampin EJ. Energy based analysis of biochemical cycles using bond graphs. Proc. R. Soc. A 470:20140459, 2014 (link).
  • Gawthrop PJ and Crampin EJ. Modular bond-graph modelling and analysis of biomolecularsystems. IET Systems Biology, 2015 (link).
  • Gawthrop PJ, Cursons J and Crampin EJ. Hierarchical bond graph modelling of biochemical networks. Proc. R. Soc A, 471(2184), 2015 (link).
  • Gawthrop PJ, Siekmann I, Kameneva T, Saha S, Ibbotson MR and Crampin EJ. The energetic cost of the action potential: bond graph modelling of electrochemical energy transduction in excitable membranes. arXiv:1512.00956 (link).

3.3.1. Example models

A collection of models is available for this section of the module. The following steps can be used to create your own copy of these models for use in the tutorial.

  1. Go to http://teaching.physiomeproject.org/workspace/464 in your web browser.
  2. Make sure you are logged in to the teaching instance of PMR.
  3. Click on Fork in the workspace menu, followed by the Fork button. This will create your own private copy of the workspace containing these example models.
  4. In OpenCOR, make sure you are logged into the teaching instance of PMR.
  5. You might need to click the Reload button (green arrow) to update the window to show your copy of the Example bond graph models workspace.
  6. Right-click on the example bond graph models workspace folder and Clone... the workspace (as per step 5).

You will now have a copy of the example models on your computer.