Meeting Minutes 2 May 2007

Meeting Minutes 2 May 2007
Present: Poul Nielsen, Tommy Yu, James Lawson, Andrew Miller, Matt Halstead (via teleconference), Jonna Terkildsen

Matt is still working on the Physiome Project RDF metadata processing. He is having problems getting the CellML repository working with Plone 2.9.6, which will allow him to use some Plone 3 features.

Matt has helped Peter to prepare an application for a grid computing grant. This application has used CellML in its examples.

Matt has also been working on the CellML ontologies.

He has been working with Andre to get proposals from the CellML workshop onto the CellML website.

Poul noted that Dave Brooks has been looking at existing field technologies that could become standardised as 'FieldML'.

Tommy: Has finished his fixes to the RDF library, and has been looking at how he can correct the current malformed XML. He plans to get his code onto the repository soon, as it is currently creating broken models.

Matt has suggested that Tommy should write formal tests for his code.

James has been in contact with Penny Noble about some models which are giving inconsistent results between COR and PCEnv. The Noble '98 model is showing particularly significant differences. James had to make some minor changes to the model to get it to run in PCEnv, and has sent this model back to Penny Noble to see if it produces the same or different results.

Matt suggests 'Related models' section of the repository show versions and variants separately, as it seems confusing as it is now. Poul suggested a hierarchical view.

Penny Noble also has been supplying models with no metadata. Matt suggested that Tommy could implement more metadata editing (e.g. revision history editing) to allow Penny to input this data.

James has started looking at the signal transduction models in the CellML repository, to see how he can fix them. He is also looking for components which could be pulled out of models to allow for model reuse. Poul noted that there were two ways to address re-use: taking existing models and identifying reusable components, or starting from first principles and building components based on common formulae (as in early versions of SBML).

Andrew noted that there was some similarity between this and the ideas proposed in Dan Beard's thread, although to be useful we would need to take the information from it and put it into metadata associated with a CellML model.

Jonna noted that she had seen significant differences between integrators, and some integrators seem to go very slow (and appear to freeze) when they hit certain points. There was some discussion of whether we can stop the integration, or somehow set some type of minimum step size after which the integrator will just skip to the next step even if it doesn't converge. At current, all the integrators allow the step size to fall to the epsilon value, and then give an error. Poul thinks that some way to debug this is needed. Andrew suggested allowing CSV data to be exported even if the run doesn't complete, so that these problems can be diagnosed (perhaps as well as a count of how many steps have been taken).

Jonna also suggested that there be some documentation of the strengths and weaknesses of each algorithm in PCEnv.

Andrew has got PCEnv working on Windows Vista (it was broken due to a range of issues with the way Vista handles environment variables, and the new security model in Vista). He has made a number of improvements of error messages in PCEnv.

Andrew has also generated a very large model of yeast transcriptional regulation (with no maths, only component structure). It is creating problems for the current tools because of its size. It can be loaded into the CellML API, but it is slow to do anything with it.

Poul noted that the W3C have been reviewing content MathML, in preparation for MathML 3.0. Andrew noted that we might have trouble getting the internal discussions, or participating in the discussion, because the University is not a W3C consortium member.

Andrew also noted that there has been some work towards fixing MathML rendering in Mozilla, although it is not yet complete, which will restore the math viewing functionality in PCEnv when complete.