Meeting Minutes 15 January 2007

Meeting Minutes 15 January 2007
Present: Matt Halstead, Tommy Yu, Poul Nielsen, Mike Cooling, Peter Hunter, Edmund Crampin, Andrew Miller

Poul noted that David Nickerson (Andre), Peter and Poul are organising a developers meeting. The week before easter (two days around April 5th, 2007) was suggested for the date, and there has been no objection so far. The aim would be to get the Auckland, Kyoto, Sydney, Washington, and Oxford groups together, to examine each other's progress and to allow each group to see tools developed by the other groups. There is a need to look for funding for this, although there are several potential sources.

Poul also noted that Alan Garny is putting a paper together, covering CellML tools and the specifcation work. The plan is for the paper to include a 'future developments' section, into which both the Auckland and Oxford groups will have input.

Matt thinks that Mike Dunstan, Gareth, and Tommy should meet, to discuss getting site_tools operational on the site, so that Tommy can use this to deploy and changes he makes.

There was also a discussion about what Tommy's priorities should be. Peter was keen for both internal and external researchers to be able to use the CellML tools ASAP to start fixing any problems (such as missing required constraints). The code is already in Plone to allow this, but Tommy and Gareth need to sort out the process for granting access. Matt was keen for this to interface with the LDAP for within the Institute. Peter suggested that Edmund, Holger Schmid, and himself would be a good trial group for within the institute, with David Nickerson (Andre) and Penny Noble being the 'test cases' from outside of the institute.

There was also some discussion of whether we need a publish and review process (perhaps using Plone's existing workflow). Matt thinks that at the moment, most people would simply want to submit their work and publish it, but a two-stage process could become useful if less trusted people submit models.

Andrew has now got the PCM (CellML API) <=> XPCOM bridge working (with no CORBA in the middle), which also required getting the CellML API to compile with MSVC8. He has also implemented:
1) Support for saving created CellML models to a file.
2) The ability to export integration results in a comma-separated value (CSV) formatted file.
3) The ability to pass model URLs on the command line.
He has also upgraded to the latest Mozilla trunk. During this process, some of the necessary patches had been broken by other changes in Mozilla. In order to reduce the risk of this happening again, Andrew has submitted that patches onto bugzilla and tried to get them reviewed (so hopefully they will make it into the Mozilla trunk). He also fixed a Mozilla regression which broke canvas in XUL documents, and has submitted a patch for this.
The Windows and Mac builds are currently not working, so Andrew is working on fixing this.

Mike asked about what can be done to help merge CellML models when two different variables (in each original model) represent the same biological entity. In Mike's example, a 'Conservation of Receptors' model was merged with a 'Conservation of phospholipase C' model. This model demonstrates two different cases, both of which cannot be dealt with simply by using connections:
1) The concentration of a reactant has a variable in both models, but different equations act on that concentration. In this case, you could formulate the model so that each produces 'contribution to the rate' variables, which are summed to produce the rate of change of the concentration. However, it appears that doing this for every variable would have adverse side-effects on model readability.
2) The same reaction is represented in both models, and there is a need to choose one reaction or the other.
Edmund noted that combining components works well for electrophysiological models, but these complicating factors arise in biochemical reactions.

Poul and Matt suggested that the model could be broken up into smaller components (one per species, and one per reaction), and these can be re-assembled to produce the final model. Matt noted that this could potentially be addressed as a best-practice recommendation.

Peter asked if we could have a meeting to deal with the datatypes at some point. No specific date has been set yet.

Mike also asked whether CellML would ever get 'classes', i.e. templates which can be used to create new components, such as for the Michaelis-Menton equations. The consensus was that it wasn't really needed, because this can generally be done with imports. However, it might be required if we were to have new components created on-the-fly (although vectors might be a better way to deal with this).