Meeting minutes, 18th of July 2007

Meeting minutes, 18th of July 2007

Present: Sarala Dissanayake, Catherine Lloyd, Randall Britten, James Lawson, Poul Nielsen, Andrew Miller, Matt Halstead (speakerphone)

  • Marc Jacobs discussed the CellML tracker at the PI meeting. The IT group feels that the CellML group did not discuss this with them. Randall is unsure why this is, as they were CCd. Andrew noted that it would only take him about 10 minutes work to do the final step of the deployment, but he can't because he doesn't have access to the server to do it, and so we need to rely on the IT staff to do this.

  • James summarised the e-mails which had been on the list.

  • James has been involved in an e-mail exchange with Alan Garny where Alan Garny noted that COR puts out points on a regular grid while PCEnv puts out points according to the integrator's adaptive step size, which makes it easier to ensure that the integrator doesn't skip over stimuli. Andrew hadn't seen the mail (James will forward this to him). Andrew noted that we used to put points on a regular grid, but after much discussion decided that this was a bad idea as we didn't always represent detail in certain areas. However, after discussion with Mike Cooling, there is a tracker item that has been open for some time to add some extra integrator parameters that ensures the integrator stops on certain boundaries.

  • Alan has asked about editing environments for XUL. Matt thinks that the issue is that he is used to RAD environments. Andrew noted that there were some efforts like this for XUL, but they were at a fledgling stage at the time. Andrew noted that Alan Garny will probably not want to be changing the top-level layout of PCEnv, but rather will probably be making new types of controls which slot into this layout, so at this stage such tools probably wouldn't be much use anyway. Matt suggests that we seek to get Alan Garny to focus on specific things that he would need to change in PCEnv when evaluating this.

  • Alan had also discussed stimulus protocols on the list. Catherine noted that some papers have complex stimulus protocols. It was noted that these can be represented in MathML just like any other stimulus. The group decided that for now there was no further action required, as the discussion on the mailing list was resolving the issues.

  • James noted that sometimes, stimulus protocols are not described in the paper. They are sometimes in supplementary code or can be obtained from the author. However, this means that the model with the stimulus protocols doesn't match the paper. Andrew suggested that there be several versions of the 1.0 model, one which represents the paper and one which works. Poul and Matt also suggested creating 1.1 models, even if we can't put them into the repository yet.

  • James had a meeting with Sarala discussing what Sarala is doing with BioPAX, with a view to possibly annotating all models that way. He will meet with Matt and Sarala tomorrow about this, and will look at creating an example model with BioPAX annotations.

  • It was agreed that we need to make it clearer on the CellML site that reaction is deprecated. Matt suggested annotating the 1.1 specification to note this. Andrew doesn't think that this is a good idea, because we shouldn't change an existing frozen specification even to 'annotate' it. Andrew suggested a new version. There was some discussion about what it should be called - Andrew doesn't like 1.2 because we have historically changed the namespace with version changes at this level, which wouldn't make sense here. Andrew suggested 1.1.1, with the namespace remaining unchanged. Poul suggested that someone create a draft and formally propose it so that we can get feedback from the mailing list.

  • There was some discussion about how we could represent discrete delays in CellML. There were three options that emerged: writing delays as a convolution of the dirac delta function with the delayed variable, changing CellML to always represent variables that change with time as functions (so that delays could be incorporated into the function argument), or introducing a new operator to describe a delayed variable. None of the options will magically make delays work with existing integrator software. Poul is in favour of the convolution of the dirac delta function with the variable, because it can be represented using existing MathML markup and is an incremental change (as opposed to changing how most CellML is written for a minority of models). Andrew thinks that the new operator will be quite a clean approach from a model representation and implementation standpoint (although there are numerical issues if we can't predict exactly which historical values we need ahead of time as well as with boundary conditions). However, he also noted that the functional representation may have benefits for a PDE representation. James thinks that the extra verboseness of writing variables as functions may be sufficiently overcome if software can hide it when not required.