Synergies of Simulation Model Documentation and Simulation Experiment Generation

Wilsdorf, Pia and Ruscheinski, Andreas and Uhrmacher, Adelinde M. (2018) Synergies of Simulation Model Documentation and Simulation Experiment Generation. In: 16th International Conference on Computational Methods in Systems Biology (CMSB 2018), 12-14 Sep 2018, Brno, Czech Republic. Poster.

Full text not available from this repository.


Developing simulation models implies phases of successive model refinement, executing diverse simulation experiments, and careful documentation. We identify possible synergies between the generation and execution of simulation experiments and simulation model documentations, and further determine basic requirements for exploiting these synergies based on a first proof of concept solution. We propose a template-based approach that extracts information from a given model documentation. Thereby, templates are blueprints for different simulation experiment types used in the model development process, such as optimization, sensitivity analysis, or statistical model checking. The templates are then adapted to the concrete simulation model and study based on the information provided in the model documentation. Templates are defined according to the syntax of a target experiment specification language, which is the Simulation Experiment Specification on a Scala Layer (SESSL) language in our case. Each template comprises immutable sections describing the basic structure of the experiment type, as well as mutable sections, i.e., template variables that serve as placeholders to be filled with concrete data. To complete those experiment templates, a vocabulary is introduced that specifies the expected content and, optionally, default values for each template variable. A basic vocabulary describes the typical constituents of every simulation experiment specification, such as a reference to the model, the initial model state, parameter values, a reference to the simulation algorithm to execute the model, etc. This basic vocabulary needs to be complemented for specific experiment types, e.g., information specific to sensitivity analysis. To automatically fill-in the template variables with actual values, our approach exploits simulation model documentations which may appear in various formats such as verbal narratives, tabular representations, or formal languages. However, often a mixture of different formats is used that may also include links to external data sources or ontologies. In particular, the less formal the presentation is, the less information can be extracted automatically. Thus, depending on how the model is documented, different methods are required to determine the necessary information. Furthermore, in addition to simply completing a given template, the documentation can be used to derive suitable templates, e.g., based on documented hypotheses. We demonstrate the feasibility of our approach by using a simulation model of the binding kinetics of EGF to cellular transmembrane receptors of human foreskin fibroblast cells, and its corresponding documentations, i.e., behavioral hypotheses described in temporal logic, and parameter values presented in tables. Whereas the former is used to generate statistical model checking experiments, the latter results in a set of sensitivity analysis experiments. Our case study reveals interdependencies between simulation model documentation and the possibilities, and challenges for generating simulation experiments.

Item Type: Conference or Workshop Item (Poster)
Projects: GrEASE