ABI Meeting Minutes 2008-04-02

Present: Randall Britten, James Lawson, Justin Marsh, Andrew Miller, Tommy Yu, Abhishek Tiwari, Poul Nielsen, Peter Hunter, Kaihsu Tai, David Nickerson, Mike Cooling

Apologies: Catherine Lloyd

Agenda for this week:

1.) Demonstration of PMR2 Prototype by Tommy

  • Tommy demonstrated the Mercurial-based prototype of his PMR2 software on the projector in the meeting room. This software is essentially a web-based Plone3 front-end to a Mercurial repository.

  • Tommy started by showing the group the listing of example models in his test repository and then used Randall's CellML 1.1 example model 'buckets' to further demonstrate the functionality of the software.

  • Each 'workspace' can have any number of files within it, whether they be CellML files, documentation files, images, PCEnv session files etc. and within this workspace a user is able to use relative URLs to refer to other files.

  • Tommy showed how changes can be made to files, and the system for uploading changed files and entering commit comments or modification histories. Currently, the web client only allows a user to merge an altered file with the file that they originally downloaded and changed. He intends to implement a diff viewer in this system.

  • When a model file has been uploaded it can be 'exposed' - this functionality mainly serves to allow the model to be viewed in the public listing. The model can be given a human-readable tag, associated with a documentation file and curation information is added at this point, although the curation flag system is yet to be developed.

  • Andrew asked about access controls and whether there was currently any access control that would allow, for example, only a specific person or persons to view the model before it is exposed publically.

    • This has not been developed yet. Randall noted that we are really concentrating on getting a CellML 1.1 capable repository out and that this is the kind of feature that Tommy can work on when the software is live.

  • Tommy loaded up PCEnv and entered a URL into the 'File->Open' field. Although this was working for him just previously, unfortunately there were technical issues at the meeting.

  • There was some discussion about the cause of this issue centering on URL setting by the operating system and the 'base URI' written to the XML file. This discussion became quite technical and a break-away session was organised for after the meeting.

  • James asked about the possibility of accessing an build-out implementation of the prototype PMR2 software when he visits Neil Wipat's lab in Newcastle in mid-April. Since he does not have a laptop, it looks like this will not be possible.

2.) Alan - Standardisation of parameter names

  • This item regarding an email sent to Randall by Alan describing sentiment from the Oxford group (particularly Steve Niederer,) that standardisation of parameter and component names in CellML models would make them much easier to work with. For example, instead of calling a component "L_type_calcium_channel" its shorthand name "ICaL" should be used.

  • Poul pointed out that ontological annotation of models in the form of metadata is desired expressly for the purpose of the kind of work Steve is doing.

  • David noted that the Kyoto group is creating and using a constrained vocabulary to annotate models with ontological terms derived from component names.

  • Peter said that we are not sure whether the Kyoto ontology is publically accessible.

  • Andrew recalled that Matt Halstead has a Protégé file that describes a constrained vocabulary generated by Mikey several years ago that is generated from and is used to annotate all the models in the repository. While this will now be out of date, it would certainly be useful to retrieve.

  • Matt was working on a way to represent this Protégé information in RDF, but apparently never completed it. He did plan, however, that this information be initially represented within the CellML files using the cmeta:bioentity tag.

  • Peter would like this information to be retrieved so that we can recommence work on the constrained vocabulary.

Action Item 1.) James will email Matt and ask for the Protégé files.

  • Peter noted that we should be adding database references into the constrained vocabulary - for example, SwissProt, PDB etc.

  • Andrew suggested two phases to this process: firstly, creating unique terms for each entity, secondly annotating these terms against databases.

  • James noted that the Biomodels Database would be a good model, as they have done a significant amount of work in this area. He believes that now the models in the repository are around 50% curated, now would be a reasonable time to start work in earnest on getting the models annotated with ontologies and database references.

  • David thinks that Biomodels use a "BQS-describe" RDF system to link entities to databases, registries etc.

  • Poul asked if anyone knew whether Biomodels uses SBO (Systems Biology Ontology) to annotate their models.

  • Peter suggested that we should be linking into SBO if possible, and we should be taking cues from the Bio-ontologies group. It is unlikely that any single ontology already in use will be able to meet our needs but we should certainly be able to incorporate other ontologies into our own constrained vocabulary. In particular, ontologies such as SBO will not be able to be used to effectively annotate electrophysiological models.

Action Item 2.) James will do some research into how we should approach this work and come up with a proposal suggesting the direction we should be taking.

3.) David - Graphing and simulation metadata

  • David presented his work on creating a framework for describing a developing reference implementations of computational models at the recent CellML workshop. The graphing and simulation metadata specifications need a lot of work; he is keen to build up momentum on these projects again and was keen to know who would be interested.

  • Justin and Andrew expressed interest in developing the specs. James is also keen to contribute some ideas and act as a sounding-board for proposals.

  • Andrew mentioned the email that the CellML discussion list received from the MAISE people (MAISE is like a MIRIAM proposal for simulation and graphing of models).

  • David noted that they don't seem to have been interested in the commentary on their efforts that we replied to their email with.

    • Andrew suggested to them that they keep their graphing and simulation metadata standards separate, and that they should not have time as the compulsory variable to integrate over.

  • Poul suggested that we simply push the point and email them again restating our concerns and suggesting that we would be keen to join their effort if they are addressed.

    • Andrew would like someone else to send the email, to make it clear that it is not just one person who is concerned.

  • Peter agreed that we should follow this approach, but emphasised that we should keep working as we have been under the assumption that we will continue to develop our own graphing and simulation metadata specifications.

Action Item 3.) We will have a meeting to discuss the future of the graphing and simulation metadata specifications while David is still here [post-script: meeting to be at 11am this Friday.

4.) PCEnv / CellML API progress update

  • Justin has been conducting a complexity analysis of the non-linear solver algorithm and the time required to perform the necessary computations should be reasonable - in the order of tens of minutes.

  • Limit for size of sets of equations to about 20. Greater sizes than this cause computation time required to be unreasonably long.

  • Peter is worried that tens of minutes will be too long. We need the solver to be able to work over seconds, not minutes. Justin should drop this code and go back to the idea of simply designating equations as part of systems of non-linear equations.

  • Poul asked what kind of information would be required to do this and how it might be done.

    • Andrew and Justin suggested that it would be relatively simple - a list of the equations involved and the variables affected would be sufficient. It could be done using cmeta:ID tags.

  • Poul asked about identification of the type of non-linear algebraic system and whether it would be costly in terms of computation.

    • Andrew thinks that it should be computationally efficient to identify the type of mathematics automatically.

  • Peter is of the strong opinion that we need a plan with concrete milestones, names and dates for the COR / PCEnv merger.

  • Randall believes the first major objective to achieve here will be the incorporation of the COR-style textual representation of CellML within PCEnv.

  • Alan Garny has put together a potential layout for COR/PCEnv representing each module which will need to be developed - this should be referenced in the planning document.

  • Peter asked about the Mozilla MathML representation bug - it has now been fixed in the Linux build, and MathML is represented in Windows in Greek symbols. This may just require a look at the fonts used, however.

  • James suggested that a new release 0.3.2 be made including the MathML representation fix - the group seemed to agree.

  • Peter asked whether the bugs in Andrew's code for the SVG right click menu that brings up annotations for the icon clicked on should be fixed in a 0.3.2 release.

    • Randall suggested that this might be getting towards a 0.4 release.

  • It would be good to at least make a snapshot of PCEnv with the MathML representation fix in it.

  • Randall asked whether the validation service Andrew has been working on should be available for the 0.4 release. This is unlikely.

5.) Kaihsu Tai, Oxford University - proposal for a systems biology dictionary

  • Kaihsu is visiting from Oxford and attended last week's CellML workshop.

  • He is interested in creating a kind of dictionary of motifs or functional models in systems biology by mining the CellML repository, and discussed his idea with the CellML group.

  • Peter noted that in the electrophysiology models, functional modules are really the membrane channels, but that the signal transduction models would be more complex.

  • Poul suggested that templates for specific kinds of systems could be generated, and that these might be useful in model creation as well as decomposition of models into components.

    • Mike mentioned the symmetry between modules in his analysis of control systems in signalling networks that he studied for his PhD as an example of this. He suggested that these kinds of motifs derived from analysis of biological signalling networks might be useful to feed into the synthetic biology community.

  • Poul noted that the CellML repository will be ripe for this kind of effort once the CellML 1.1 capable PMR2 repository is live.

  • Andrew noted that CellML 1.2 will allow for much more generic representations of systems; this may help.

  • Andrew gave an explanation of pi calculus to Peter.