Sharing comprehensible physiological models
As our knowledge of physiological systems increases, computational models developed to help explore and understand these systems are becoming better approximations of the real world. Inevitably this leads to more complex models which are difficult to share with colleagues in a useful manner. We present here some recent work aimed at helping scientists share their computational models in a way that ensures they are not only reproducible, but also comprehensible.
Standards to encode mathematical models have developed, such as CellML for differential algebraic models and FieldML for spatial finite element models. Using such standards, it is possible for scientists to share unambiguous descriptions of their mathematical models. Similarly, the Simulation Experiment Description Markup Language (SED-ML) has been developed under the Computational Modelling in Biology Network (COMBINE) to enable unambiguous simulation experiment definitions to be shared. Together with models encoded in a standard format, SED-ML enables scientists to share reproducible simulation experiments independently of any individual software tool.
Repositories exist for scientists to deposit their models and simulation descriptions and share them with the community. For example, the Physiome Model Repository (PMR) currently has over 600 models encoded in the CellML format and many of them have associated SED-ML describing simulation experiments from the source literature.
For scientists to confidently reuse computational models, they must be able to comprehend them. This includes, for example, the biological system represented by the model, the context in which it has been validated, any limitations or restrictions on the use of the model, etc. Having access to the simulation experiments is a good first step in gaining such comprehension, allowing scientists to “play” with the model and observe its response to certain perturbations. Beyond this, scientists currently must refer to the source literature to discover what further knowledge is available for the model, a time consuming and often frustrating task. To address this, we use semantic annotation of the models and simulation experiments to encode this knowledge in a computable format. The addition of this knowledge to repositories such as PMR enabled the development of tools which aid scientists in comprehending the model without the need for scientists to independently mine the literature.