The following are summaries of some discussions that were brought during the recent Multiscale Modeling (MSM) PI Consortium Meeting (2007). The topics relate to sharing models, driving standards in multiscale modeling and the development of reliable and reusable software.

1. We now have some successful model exchanges standards in biology, most notably SBML and CellML. However we still still many journals publishing modeling papers where the model is given in an appendix as a set of mathematical equations (e.g. differential equations). The problem with this approach is that virtually 100% of models published in this manner are incorrectly specified in the paper which means that it is very difficult to reproduce the work of the author, let alone build on their efforts. The work by the biomodels.net group is building a database of models that have been extracted from the literature and carefully curated. This is a huge amount of work, very costly in time and effort. It would be far better if journals could request authors to supply the model in some computer readable form when submitting their paper for review. If accepted, the journal could submit the model to an organization like biomodels.net or host it themselves. Some journals do now accept models as part of the manuscript submission, however this is not enforced. It would be of interest if major organizations such as the NIH, DOE and NSF could stipulate that model developed during the duration of the support should be made available to the community in some computer readable form, preferable submitted to a recognized model repository.

2. SBML was designed to represent cellular models at a single scale, it does not attempt to address multiscale models to any great extent. The reason for this was that at the time of its development, the community did not have have sufficient expertise in multiscale modeling to formulate a workable and acceptable standard. However, since the development of SBML, progress has continued in the modeling field and perhaps now is the time to visit the issue of model sharing and standards in multiscale modeling. A possible use case that could steer this effort would be to develop a sharing language (akin to SBML) that combines intracellular network modeling with multicellular systems. Since some of the theory for multicellular modeling is relatively well developed it wouldn't be too controversial to try and develop in this direction.

3. I believe that the modeling/systems biology community would benefit enormously from the development of specific, well tested and documented code libraries, preferably written in a portable and linkable language, such as C. libSBML is a good example of what can be accomplished and I would consider this as a model for software development in systems biology. Currently we have a lot of reinvention in the community which I think is holding back development and particularly innovation since no one ever gets the time to move beyond the basics, how many ODE simulators do we really need? This approach has and is used by the Department of Energy, the sundials suite being a particularly good example of what can be done, such work as has served the community very well.

Herbert M Sauro
Department of Bioengineering
University of Washington
Seattle