Meeting Minutes 2008-01-16

Present: Randall Britten, Edmund Crampin, Michael Cooling, James Lawson, Peter Hunter, Justin Marsh, Andrew Miller, Poul Nielsen, Peter Schmiedeskamp, Anil Wipat, Tommy Yu

Updates from last meeting

  • Randall will add a note on next week's agenda to invite Poul to give his input on this matter (what to do with the 11 models in the repository with reaction elements).

    • That was done, and Poul has not followed up with Catherine.

      • Action item: Poul will follow up with Catherine on this matter.


Poul: Should CellML aim to represent reaction only models?

  • The language is incomplete, however this can be done with metadata, but the terms still need to be formulated, with Sarala putting in a lot of work in this area.

    • Poul confirmed that the original intention behind the CellML specification was that reaction-only models lacking kinetics should be able to be represented. It was noted that now that reactions have been removed from the CellML core (with Sarala working on how to represent them as RDF metadata), this means that only components and RDF will remain in these models. Poul thinks that it is important that we can start with non-mathematical CellML models and build them up by adding kinetics.

Catherine (comment prior to meeting start): CellML Workshop 2008 Update

  • There are enough speakers, will be getting in touch with them on what specific topics they will be covering.

  • Will be meeting with Peter and Poul to decide on the programme and themes.

PMR2 update

  • Git view is fully operational in Plone 3

  • Current task is on working on the process to expose (publish) a model.

    • Reason for calling this process "expose" is to prevent confusion, as published model is already referring to a published paper.

    • Exposing a model involves taking a single changeset of a particular model, parsing and caching all the files, and extracting the metadata to put them into an index. The curator will then curate that model and apply curation flags to it, and "expose" it onto the main listing of all models that casual users of the repository will first see.

    • Neal wonders what kind of permanent URI will be given to these revealed models. He suggests that LSID (Life Science Identifiers) could be used in this case.

  • Poul would like to know the timeframe what this prototype will be done, and the estimation Tommy gave is one to two weeks before the workshop.

  • Randall suggests that if we think Git/Mercurial is completely suitable as a backend for CellML model development, we could start creating such a repository and store models into there, and when PMR2 is fully operational those models then can be "expose" for public consumption.

CellML 1.2 Specification

  • Andrew noted that he has almost finished the model interpretation section of the initial draft. He is focusing on getting a document which describes the same things as CellML 1.1 in precise, purely normative terms as a basis for further development of the specification by the community. Some more effort is still required to fully describe the structure of models, and to check to ensure that we haven't lost anything from the CellML 1.1 specification.

  • A point was raised on the mailing list regarding how much the new specification needs to be compatible with CellML 1.1.

    • Andrew pointed out that the CellML API is a 1:1 mapping of the specification to the software

    • Peter pointed out that we'll have tools that will do conversion, not to mention there are still not too many software tools out there that uses the reference CellML API

    • Also it is better to get things right before too many people start to depend on the API.

  • Poul: description for mathematics in CellML is declarative

    • There has been discussion on the mailing list about whether connections should continue to have directions like in and out. Poul noted that when CellML was originally defined, it was intended to be a declarative language, but some concepts, such as the directionality of connections between variables, which aren't meaningful in a declarative language, were added.
    • Peter points out that it is also meaningless in a biological sense.

    • Andrew suggested that the specification could be defined so that directionality does not alter the interpretation of the model by software, but rather, is provided as a convenience facility for use by users. For example, the model could define, in one component, A=2B, and in another, A=3C, with A connected, and regardless of the direction of the interface, which equation is used to work out A and which is used to solve for another variable depends only on what initial_values and other equations are provided

Clearing of Mike Cooling's moderation flag

  • Mailing list policies dictate that new users joining will have a moderation flag set, to prevent spam or flame-wars that could be started by random new users. Moderation flag will be cleared after the team agrees.

  • Suggestion was to allow people of ABI, that participate in CellML meetings, to skip that process. However this would make the mailing list rather Auckland-centric.

  • Resolution was to leave this as it is. Michael Cooling will have his moderation flag cleared as the team agreed.

System of equations solver

  • Randall noted that there had been discussion on tracker item 192 about what needs to be done to allow systems of equations to be solved. Randall had suggested simply giving all algebraic equations to a solver without any analysis, but Andrew feels that this would seriously degrade performance even for models which don't require systems to be solved. Andrew has suggested as an alternative that we could cap the maximum number of free variables for us to identify a system to keep the complexity under control.

  • Tracker item 192 is being used to discusses this.

James on CellML 1.2 specification vs SMBL 3 specification

  • It would be good to find out what SMBL 3 is trying to do.

    • Action item: Andrew will take a look