BioModels Workshop 2009: Day 1

BioModels Database Introduction: Nicolas Le Novere

Repository of quantitative models only for the moment: no implicit statement of biochemical accuracy as a consequence of being in the database, but must be of biological interest and only those that have been described in peer-reviewed scientific literature. In terms of curation: model syntax and semantics are checked; and then models are simulated to check the correspondence to the reference; model components are annotated; they improve identification and retrieval. Models are accepted in various formats, and exported in other formats too.

The models come from individuals, existing model repositories, journals, and direct curation from literature by BioModels curators. Within the individuals category, submitters are members of the SBML community and authors. More than 200 journals *advise* deposition, including all PLoS, BMC, and Nature Mol Sys Bio.

BioModels Database Technical Aspects: Chen Li

The infrastructure of the db includes a set of tomcat application server clusters. MySql databases sit behind these server clusters. There is also a mirror site at Caltech. All models in the BioModels database have to pass through the BioModels pipeline: syntax check, consistency check, divergence to either curated or non-curated branch. When a model is submitted, the db parses it and fetches MIRIAM anntoations. It fetches information from GO, UP, ChEBI, and the taxonomy db, and then added into the model. Exports are available in lots of formats: most of the SBML levels, CellML, XPP-Aut, VCell, SciLab, BioPax. For BioPax and VCell they use a Java converter developed in-house; for CellML, SciLab and XPP they use an XSLT, and to build the PDF they use SBML2Latex. There are also SVG, GIF and various other visualizations available. There is also a link to the JWS online simulator.

They also have a Model of the Month, which is available via the web site or via an RSS feed. THey use AJAX for parts of their web interface: to view a models tree that is created based on the GO hierarchy; an internal-only annotation tool; sub-model generation and more. There is also a nice display of the Mathematical equations. They have a set of web services that are publicly accessible. The source code and database schema are available from sourceforge.

BioModels stores the frozen models: the way the models were when the publications were submitted. They need to correspond exactly to how it was published. However, if a modification was created by the authors and then a new paper made, the new version can then go into the database. If the models don't run, they don't reproduce the published results and therefore aren't MIRIAM compliant. Therefore they remain in the non-curated section of the database.

SBML Converters: Nicolas Rodriguez

They have: Scilab, XPP, CellML 1.0, BioPax Level 2, Dot/SVG, Vcell and PDF. For BioPax the original conversion lost a lot of granularity (physical entity -> species, for example). Now, by making use of the MIRIAM annotation, a more precise characterization can be made (e.g. UniProt annotation implies a protein in biopax, which is more specific than physical entity). For CellML, a new conversion from SBML to CellML is being developed by Andrew Miller, but it is still in the early stages. They're waiting for CellML 1.2 + CellML metadata to make the conversion better. The current SVG and GIF exports are not satisfactory, and they're looking for collaboration with other groups or efforts.

Model Curation and Annotation: Lukas Endler

Within the curated branch, models are: checked for MIRIAM compliance, a curation figure is added, model elements are manually added, and they get a BioModels ID. In the non-curated branch they are only slightly edited by curators, and only publication details and creation details are added. For MIRIAM compliance specifically within BioModels (more restrictive than MIRIAM compliance), the models must be: correctly encoded in a standard format (valid SBML), contain a link to a peer-reviewed journal, the creators' contact details, be able to reproduce the results given in the reference publication, and reflect the structure of the processes and formulas described in the reference publication.

The non-curated branch is valid SBML, but not MIRIAM compliant: cannot reproduce results, the models differ in structure from the publication, or it is not a kinetic model. If it is MIRIAM compliant, then it goes into this branch if the models contain kinetics they do not know how to curate yet (boolean models) or some parts are not encoded in SBML (e.g. spatial information). Another reason it would go here if it is MIRIAM compliant is if there is a significant tailback due to insufficient time and workforce, in which case it will be moved into the curated branch as soon as possible.

The curation guidelines are that they should: read the publication; go through the SBML model and compare all the elements (where possible they create reactions out of differential equations, add names to unnamed reactions, rules and events); change names and IDs to correspond to the article; try to reproduce one or two key results of the reference publication and create a curation result (e.g. a figure or table); add notes; move the model to the curated branch for annotation and publication.

http://biomodels.net/meetings/2008/index.html (Yes, it is the 2009 meeting, even though the URL says "2008").

Please note that this post is merely my notes on the presentation. They are not guaranteed to be correct, and unless explicitly stated are not my opinions. They do not reflect the opinions of my employers. Any errors you can happily assume to be mine and no-one else's. I'm happy to correct any errors you may spot - just let me know!