The CellML Model Repository

Catherine M. Lloyd 0 James R. Lawson 0 Peter J. Hunter 0 Poul F. Nielsen 0 Associate Editor: Jonathan Wren 0 Auckland Bioengineering Institute, The University of Auckland , Auckland 1010, New Zealand Summary: The CellML Model Repository provides free access to over 330 biological models. The vast majority of these models are derived from published, peer-reviewed papers. Model curation is an important and ongoing process to ensure the CellML model is able to accurately reproduce the published results. As the CellML community grows, and more people add their models to the repository, model annotation will become increasingly important to facilitate data searches and information retrieval. Availability: The CellML Model Repository is publicly accessible at http://www.cellml.org/models Contact: The Author 2008. Published by Oxford University Press. All rights reserved. For Permissions, please email: D o w n l o a d e d f r o m h t t p : / / b i o i n f o r m a t i c s . o x f o r d j o u r n a .l s o r g / b y g u e s t o n O c t o b e r 2 6 , 2 0 1 4 1 INTRODUCTION High throughput experimental techniques have led to the population of web-accessible databases with vast amounts of biological data. Mathematical models of biological systems are playing an essential role in the interpretation of this data. The scientific community now faces the challenge of the mathematical models themselves becoming increasingly complex and numerous. There is a need for centralized databases to store all these models in standard formats to make them easily accessible and reusable by the research community. Publishing the models in a standard format, concurrent with the submission of a written paper, will eliminate many of the errors introduced into the model during the publication process. Here we introduce the CellML Model Repository (http://www.cellml.org/models) and discuss it as a solution to these challenges. The BioModels database (Le Novere et al., 2006) is a similar effort, containing biochemical pathway models that have been described in peer-review publications, expressed in SBML (Hucka et al., 2003). Similarly, JWS Online (Olivier and Snoep, 2004) is a repository of kinetic models describing biological systems, and ModelDB (Hines et al., 2004) is a database which stores published models in the field of computational neuroscience. CellML (Lloyd et al., 2004) and the CellML Model Repository are part of the IUPS Physiome Project (Hunter and Nielsen, 2005) effort to create a virtual physiological human. The explicit representation of modularity, together with the flexible nature of the CellML language which allows the description of a diverse range of cellular and subcellular systems, are two essential features of CellML with regards to its role in the Physiome Project. Initially the CellML Model Repository started out as a set of examples to illustrate how the language could be applied to describe various biological processes, and to test its features as the language evolved. Later, once the CellML 1.0 specification was stabilized, the CellML repository became a collection of CellML descriptions of models drawn from peer-reviewed journal publications. The CellML Model Repository has since undergone significant growth, with over 330 freely available, quantitative models of biological processes taken from the peer-reviewed literature. In contrast with other databases, such as BioModels, JWS and ModelDB, which focus on specific areas such as systems biology pathway models or computational neuroscience, the CellML Model Repository contains models describing a wide range of biological processes, including: signal transduction pathways, metabolic pathways, electrophysiology, immunology, the cell cycle, muscle contraction and mechanical models and constitutive laws. This wide scope exemplifies CellMLs ability to describe much of the biochemistry, electrophysiology and mechanics of the intracellular environment. Lumped parameter models dealing with systems physiology (e.g. blood pressure control, fluid retention, electrolyte balance, endocrine function, etc.) are also within the scope of CellML. MODEL CURATION Currently, of the 330 models in the CellML Model Repository, approximately half have been curated to some degree. A star system signifies the curation status of a CellML model. No stars indicate the model has yet to be curated (level 0); one star denotes the CellML model is consistent with the published paper (level 1); two stars imply the CellML model has been checked for typographical errors, unit consistency, completeness (i.e. there are no missing parameters or equations), overconstraints and finally, and arguably most importantly, the CellML model is capable of reproducing the published results (level 2). If a CellML model has three stars it is known to satisfy physical constraints such as conservation of mass, momentum, charge, etc. At this level the curation is conducted by a domain expert (level 3). From experience, we have found that levels 1 and 2 can be mutually exclusive. Frequently, the errors introduced into the model during the publication process require us to correct minor typographical errors or unit inconsistencies, and/or contact the original model author to request missing parameter values or equations. The process of model curation involves the following sequence of actions:  The CellML model is loaded into an editing and simulation environment such as the Physiome CellML Environment (PCEnv) or Cellular Open Resource (COR). Any obvious typographical errors and unit inconsistencies are corrected, which is facilitated by a series of error messages and validation prompts generated by the software, and the rendering of the MathML equations in an easily readable format.  Assuming the model is able to be run, we then compare the simulation output with the results in the published paper this typically involves comparing the graphical results with the published figures.  If we cannot get the CellML model to run, or the simulation output disagrees with the published results, we then attempt to contact the original model author(s) and seek their advice and, where possible, obtain the original model code, which may be in a wide range of different programming languages. We aim to complete the curation of all the models in the CellML repository, ideally to the level that they replicate the results in the published paper (level 2), however we acknowledge this will not be possible for all models. According to the dynamic, growing nature of the CellML Model Repository, we have designed it with the concept of community curation in mind, so that groups of expert modellers with vested interests in particular models are able to collaborate on their curation. MODEL ANNOTATION Metadata, the extra information associated with a model, are embedded in CellML using the W3C approved RDF standard. In order for a CellML model to be committed to the repository, at the very leas (...truncated)