ABI Meeting Minutes 2008-05-21

Present: Catherine Lloyd, Peter Hunter, James Lawson, Tommy Yu, Andrew Miller, Edmund Crampin, Randall Britten, Mike Cooling, Poul Nielsen, Justin Marsh

Review of last week's action items:

Catherine has given Abhishek a demonstration of the xCellerator pathway model creation software. Catherine, Peter and Abhishek plan to have a meeting to follow this up.

2.)  Andrew checked for items in the old cellml.org PCNG tracker and found that there were some items that need to be archived, but not necessarily migrated to the new Physiome tracker. A static archive of these old tracker issues will be created before the PCNG tracker is completely removed.

3.)  Tommy has redirected www.cellml.org/tracker to the new Physiome tracker.

4.)  James hasn't been in touch with Martin Fink regarding his research into ontological annotation of elements in CellML, but Peter is following up on this work. This continues to be high priority and under heavy discussion.
  • The group discussed representation of ion channels within currently available ontologies.
  • Peter noted that the Gene Ontology community has not really had much input from ion channel experts, and remains predominantly specific to protein interactions and genomic annotation.
  • Peter wondered how the "GO term" for ion channels correspond to consensus classifications by ion channel physiologists. This is something we should check out.
  • Poul noted that we should be able to leverage multiple existing ontologies within the CellMLBio ontology to describe a single term. That is, collation of information about entities from multiple sources.
  • Peter mentioned www.ionchannels.org as a possible source of relevant information or links to information.
5.)  James forwarded to the team@cellml.org list the Le Novere et al. draft paper describing the development and usage of ontologies within systems biology modeling. There is an open invitation to contribute but none of us really have significant experience with any of the ontologies described within (SBO, KiSAO, TeDDy.)

6.)  Andrew has been in touch with Kevin Burrage regarding setting up a teleconference meeting to discuss the possibility of including capabilities for stochastic modeling within CellML 1.2 - Kevin is currently in the UK so time differences are an issue here.
  • The meeting is currently set for 7:30pm, Friday 23rd May, NZ time. Unfortunately this is not a practical time for many of the members of the group who wish to participate in the discussion.
  • James noted that he was intending to listen in but probably wouldn't have a lot to contribute, and would be happy to get a recording of the conversation.
  • We were keen to use the BestGRID teleconferencing system for this meeting but Kevin does not have easy access to this system on his end.
  • Randall noted that he isn't sure how one might record a Skype session; Mike suggested that he could probably set up a system to record the audio output.
  • Randall suggested using some kind of commercial provider to set up and record the meeting.
  • Peter agreed this is a reasonable idea in the short term but is keen to look into other options for future conference calling.
  • If Andrew can reschedule the meeting to a better time, this would be helpful. If not, Poul and Peter suggested that he could just talk to Kevin and discuss the main issues and we could hopefully have a larger meeting in the future.

Agenda for this week:

Petri net modeling - meeting with Valeriy Vyatkin from software engineering
  • Valeriy works on tools for analysis of large Petri nets.
  • Unfortunately these tools are only really designed for analysis of elementary discrete Petri nets, whereas analysis of the kinds of Petri nets used by computational biologists is not really tractable by the kind of software Valeriy works on.

James - Paper about CellML model repository

  • James suggested that a paper be written describing the CellML model repository, in the vein of the Le Novere et al. 2006 paper "BioModels Database: a free, centralized database of curated, published, quantitative kinetic models of biochemical and cellular systems."

  • He suggests two main components to this paper:
    • A generalised overview of the repository describing its goals and functionality
    • A discussion of the issues facing the effort to create such a repository, including curation, workflows, community management etc.
  • There is consensus from the group that this is a good idea.
  • Edmund suggested PLoS Computational Biology or Experimental Physiology as potential journals to try and publish this in.
Action item 1.)  James will set up a folder on cellml.org and start the process of compiling some notes on the paper.

Mike / James / Randall - Curation, information contained in documentation and version histories
  • Mike had a fairly lengthy chat with James and Randall yesterday (Tuesday 20/05/08) about the utility of models in the repository to his work in their current curation state.
  • He was specifically referring to the Pandit 2001 model, but the issues he raised apply more generally.
  • The current latest version (07) of this model has a curation rating of "three stars" as it has been thoroughly curated and has had input from Frank Sacshe, a domain expert. A number of corrections have been made to the original model, as described in the Pandit 2001 publication.
  • What the model does not have, however, is associated full documentation of precisely how and why the model is different from the original model.
  • There is some version history associated with the model by means of metadata, but a full description of the corrections is not available from the CellML repository.
  • This is partly to do with the fact that COR was used to curate this model, and deletes metadata from CellML files if they are downloaded from the repository, modified and then saved, and partly due to holes in curation workflow best practices.
    • N.B. Alan Garny has just released a version of COR which does not delete metadata from CellML files, which will improve matters enormously.
  • James emphasised that where the workflows are properly in place, we will be able to build up this information as users and curators work on models. It is very difficult to use the modification note system retroactively, however, so a different approach will have to be taken to represent historical changes to models.
  • Even if all this information were to be present in the modification history metadata, Mike suggested that the model documentation should display it prominently so that a modeler immediately knows why the model they have downloaded is different from that described in the publication.
    • Mike would like to know, "what the changes from the quoted resource (i.e. paper in most cases) to the current downloaded model are, who made them, and an explanation. E.g. "Tau_b includes a negative V term as there was a mistake in the paper [j.lawson@auckland.ac.nz]"
  • A repository user needs to be provided with information on what the model does 'out-of-the-box' - for example which specific figure(s) from the publication it is parameterised to reproduce when simulated.
  • Mike feels strongly that without this information, a model in the repository is of little use to a modeler.
  • The group discussed whether adding a modification history note when uploading a new version of an already existing model should in fact be compulsory.
    • Mike noted that this may not be sufficient as people could simply put useless information into this field.
    • This useless information will, however, be associated with that person's name, and would affect their reputation.
    • Tommy has made the requirement of modification history notes a compulsory feature of updating models in PMR2.
    • Peter suggested that a curation flag could be used to denote that a model has a decent modification history.
      • N.B. this pends the development of the curation flag system, which will be implemented in PMR2.
    • Further to this, Randall raised the point that there is no requirement for or even mention of information about a model's modification history within the current level / star based curation scheme.
  • Mike feels that even a modification history based on users making descriptive notes of what they have changed between versions of a model may not be sufficiently useful or usable to people who want to use repository models for research.
    • A user would have to trawl through the modification history notes to figure out how the model differs from the uncurated / uncorrected model described in the original publication. This would be tedious and time consuming.
    • He believes that it is necessary to also have a succinct, up-to-date summary of the differences between the latest version of a model and the model as described by the publication on which it is based.
    • Randall argued that the kind of summary Mike is talking about might not actually be so different from the collated modification history entries, as it would eventually become threadlike and protracted anyway. Mike disagreed that this would necessarily be the case.
    • James suggested that we can try to put as much of this information as possible into the model documentation, by way of a compromise.
  • Edmund suggested that the repository could allow users to submit comments on models. These comment threads would have to be carefully moderated.

Enable moderated comments about models on repository

  • This possibility has previously been raised but not seriously discussed.
  • Plone allows administrators to 'Allow discussion on this item,' which in turn allows logged-in users to post comments about content.
  • We do have the capability to discuss models within the tracker, which has additional features that might be useful here.
  • A link to the relevant tracker item for each model could be placed in a prominent position on each model page.
  • However the group consensus was that having to use the tracker would put up a barrier to users reading and adding to discussion of models.
  • The possibility of model wikis was also raised (and has been previously discussed by the group,) but the Plone 'allow discussion' feature is probably the best way to allow users to provide feedback and commetary on models.
  • Randall suggested that an RSS feature allowing curators and users to subscribe to discussion feeds would be very useful. Plone may allow for this but this needs to be checked on.
  • The group decided to turn on the Plone's 'allow discussion on this item' as the default setting for model pages.
Action item 2.)  Tommy to implement the 'allow discussion' feature on model pages.

James - Put up CellML presentations and posters on cellml.org

  • James has put the *.ppt file of his Friday morning presentation to the ABI on CellML and synthetic biology into his cellml.org user directory. He noted that a number of people have been giving presentations on CellML recently and suggested that we should really be making these available online.
  • This is becoming fairly common practice for research groups.
  • Copyright issues might have to be cleared up before files are made available on the net if the talks were given in a private or casual setting.
  • Mike and Peter suggested that this might be something for the Auckland Bioengineering Institute to consider as a wider issue.
Action item 3.)  James will look into setting up a location on cellml.org for posters and presentations to be stored and start collecting them.

PCEnv / CellML API progress update

  • Andrew has been working on the non-linear algebraic solver and described progress on the current system for finding systems of equations. See Physiome Tracker Item 192 for further information.
  • Justin has been working on resolving bugs labelled as blockers for the release of PCEnv 0.4
    • He estimates that he has about another 1.5 weeks of work to do on this.
    • See Physiome Tracker Item 382 for further information.

PMR2 progress update

  • Tommy has figured out how to use the Zope catalog and has searching working cleanly in the prototype.
  • The front-end needs tidying up but the prototype is almost ready.
  • He has created a VMware image for PMR2.
  • The next step in the project is to distribute the VMware image to people such as Catherine, Mike, James and Jonna for initial testing.
    • Randall suggested that the image could be put on the ABI LAN - it would be good if people could try out the collaboration functionality. We could use Amazon hosting - this has the blessing of the ABI IT staff.
    • He is currently working on a way to migrate the current PMR links into PMR2.
  • Once the prototype is finished and tested, PMR2 needs to be built. This shouldn't take too long - 3 months was given as a minimum time-frame estimate.

CellML 1.2 specification progress update

  • The upcoming discussion on the possibility for stochastic modeling in CellML 1.2 with Kevin Burrage is the only news on this front.

Additional notes from Peter

  • Peter will be going away soon, and had a few issues to raise.
  • He asked whether anyone has experience or knowledge of dynamic SVG or tools for working with it. Dynamic SVG allows animation and could be used to represent ion transport across membranes etc.
    • Nobody was familiar with this.
  • Peter has received a reply from Grace Peng, who is organising the panel on model sharing at the Society on Industrial and Applied Mathematics Conference on Life Sciences meeting in Montreal in August. See the wiki page for this panel for details on its purpose. (James previously provided Peter with information on policies of major journals on deposition of computational models in databases, which he fowarded to Grace Peng.)
    • Grace is looking for a good example of work in which the CellML model repository had been used as a data resource for research.
    • Catherine suggest Frank Sachse's work on combining his fibroblast electrophysiology model with the Pandit et al. 2001 model in the repository.