Why A Repository

Modelers are producing more and more complex models. Unless these models are sufficiently characterized and made available to the research community their reuse will be minimal, and reproducing simulation experiments incorporating them will prove problematic. Consensus on the content and form of experiment recipes that combine models and simulations will encourage model sharing and facilitate reuse.

A set of guidelines specifying the Minimum Information About a Simulation Experiment (MIASE) proposes a common set of information necessary to reproduce simulation experiments that incorporate quantitative models. We have instantiated these guidelines in a web-based content management system. With RepoSE you can create Simulation and Experiment Descriptions, enrich them with experimental data and annotate them with domain meta-information to facilitate classification, searching and cross referencing.

SED-ML

One instantiation of the MIASE guidelines is SED-ML - an XML schema, instances of which are recipes describing the combination of models and simulations into reproducible experiments. In particular, SED-ML describes five components essential to compose a simulation experiment description, i.e.:

• simulations - a description of the simulations' method, type and algorithm (KiSAO)
• models - a description of the models' location, language and modifications
• tasks - the glue that combine models and simulations into experiments
• data generators - how to present the simulations' results, e.g. 2D graph
• outputs - how to transform raw simulation output into numerical or graphical results

Insuring Reproducable Results

The Simulation & Experiment Description Meta Language (SED-ML) is a means - like a recipe - to describe the combination of simulations and models in reproducible experiments.  We have built a web-based simulation and experiment description repository based on SED-ML. Implemented as a Plone add-on product, our repository allows researchers to:

• create simulation & experiment descriptions (SEDs) compliant with SED-ML
• annotate simulations & experiments with domain meta-information from various ontologies
• search for simulations & experiments based on a variety of criteria, e.g. algorithm type
• collaborate with colleagues by allowing them to modify your simulations & experiments
• curate simulations & experiments to build a collection of verified and approved SEDs
• export simulations & experiments in SED-ML compliant XML

In contrast to EBI's existing BioModels database, which hosts only biochemical models written in SBML, our repository will uniquely record any simulation experiment, including those written in \Cpp, thus making the tool generally applicable to the types of simulation models used within the NCSB.

We have extended the SED-ML standard to accommodate source-code models, and enriched the output types with descriptive text, images and animations - which makes the repository useful as a laboratory notebook.

Results

• We can easily realise almost any collaboration scheme imaginable, e.g. research groups can have their own repositories where only curated and approved simulations & experiments are visible to the public, while works in progress are accessible to group members, and colleagues can collaborate on individual simulations and experiments.
• We can annotate simulations & experiments with domain meta-information, e.g. species name or algorithm type. In addition, any number of user defined tags can be applied to your simulations & experiments, all of which serve as search criteria to help you find and correlate information quickly and efficiently.
• We believe the synergy achieved by paring SED-ML's precise standard with Plone's superior content management functionality will help foster and improve model reuse within the NCSB.

Getting Started