ABI CellML Meeting Minutes, 2nd June 2010

Present: Dougal Cowan, Randall Britten, David "Andre" Nickerson, Mike Cooling, Andrew Miller, Tommy Yu, Justin Marsh, Catherine Lloyd (via Skype), Poul Nielsen (via Skype)

Apologies: Peter Hunter

Last week's action items:

1) Justin will test the Saucerman model and compare with the output of the MATLAB code.

These were not checked, will be carried over to next week.

This week's agenda:

1) (Catherine) Following up on the request for a simple metadata editor... what are our options and how soon can I actually be using a tool (bearing in mind I am not a computer scientist and I want a nice and easy, intuitive tool to use)?

Randall reported that although it had been planned to implement a metadata editor using the JVM/Java/Scala platform, this has been put on hold as Peter has requested that unless other teams can agree to support this platform, we should not start developing with it. Randall will try to start these discussions again this week, after his absence last week.
Catherine really wants a "form" style metadata editor, a way of entering metadata that will eliminate errors and omissions.
Mike suggested we could make an HTML page with some Javascript that spits out appropriately formatted metadata from a form.
There was a discussion about how to quickly implement such a form-style metadata editor.
Randall asked what the shortest path to this editor would be.
Justin mentioned that for a stand-alone metadata editing form, development would be relatively quick and easy using Scala/JVM, probably just 1-3 days. The use of the Eclipse-RCP framework is more appropriate for a full scale project, such as the proposed rewrite of OpenCell, and requires more work.
Tommy suggested a webservice that gets the citation metadata from a database.
Andre said that the citation metadata should just be stored as a PUBMED ID, which can be used to generate the citation information on an exposure page.
Mike mentioned that there are other types of citation which must be supported as well - Tommy said that this is supported via Dublin Core.
Mike said that the current scheme of hand-editing XML is crazy, frustrating and error prone, which Catherine agreed strongly with.
Justin asked if it would be OK for him to spend a day or two working on a simple metadata editor.
Randall suggested initially, we should try using a commercial editor such as Altova XMLSpy (which Matt Halstead also recently recommended to Catherine), and revisit this if that proves not to be helpful.
Mike said that documentation is needed for metadata use and other issues to do with PMR2. This needs to include the relationships between workspaces, exposures, metadata, etc. Without this documentation, PMR2 is not viable without Catherine and Tommy.
There was a discussion about PMR2 documentation. Randall said that this should start as soon as possible.

Action item 1) Mike to generate a starting list/outline for Dougal from which he and Tommy could start a document.

2) (Catherine) PMR2 and metadata rendering: Unless a very strict method of metadata entry into a model is followed PMR2 will not display the metadata. This includes incomplete data sets (for example when it expects a paper citation and there may not be one). This is going to potentially cause big problems with 1.1 libraries. Taking Mike's current examples - a) the metadata sets may be "incomplete" and therefore they won't render, and b) if Mike has the metadata in a completely separate file to the model (which is perfectly acceptable) I haven't got a clue how the model should be annotated as PMR2 looks to annotate a CellML model file itself.

This was covered sufficiently in discussions on item 1.

3) (Catherine) Metadata specification - I believe if we have a clearer defined specification this would help Tommy to implement metadata rendering in PMR2 (Tommy please correct me if I am wrong).

Catherine and Tommy reported that the current metadata is broken - some metadata added to models does not show up in exposures.
Randall asked if Mike would be willing to work on the metadata spec.
Mike said that in two months time, he will be able to start work on the metadata spec.

9) PMR2 development - update

Randall reported that Tommy wants feedback on the staged version of PMR2 0.3.
Catherine said that people over there are keen to see the FieldML example. She asked how close the 0.3 release is. Tommy reported that it is essentially ready for release, although he has been fixing some minor bugs.
Randall tried to clarify that the "FieldML examples" in the staged PMR2 v0.3 are using the cmgui file format, these files will be converted to the FieldML format once a conversion tool is available.
Catherine asked if testing had been done. Tommy said that as long as everything Catherine needs is working, that should be sufficient testing. Andre reported that he has tested embedded workspaces.
There was a discussion about current and future testing of PMR2.
Mike reported that the search in PMR2 is listing expired exposures, and is not listing published exposures.
There was a discussion about how to fix the search function.
Randall said that fixing the search should be the top priority for version 0.4.

Action item 2) Mike to explain the situation to Peter and Poul via email.

10) (Andre) Model provenance manuscript

Randall reported that Andre is suggesting a substantial re-write of the paper. Andre feels that the contents are not worth a paper on their own.
There was a discussion about the research concepts in the paper.
Andrew said that the paper shows increased provenance information after switching to a DVCS for a model repository.
Andre said that there is no proof of increased provenance information. He also said that the statistical analysis of the data was meaningless to him.
There was a discussion about provenance information, and how the information in the paper was gathered.
Mike asked if the paper fills thead requirements for one of the PMR2 related publications set out previously.
There was a discussion about the statistical analysis of the data, and how far the data goes back.
Andrew said that there is sufficient data to show and increase in provenance information, so we don't have to go back further.
Andre suggested that the content of the paper could be condensed into a couple of paragraphs in another paper, and that nobody has yet used the distributed aspects of Mercurial in ways that the paper suggests are some of the advantages of a distributed version control system.
Mike said that the paper is probably OK for a lower tier journal, but would need more information or analysis for a higher tier journal.
Justin said that the paper is also being published to be cited by other papers in the future.
Mike asked if more information could be extracted as to what the changes/provenance information consists of; whether it could be categorised into small fixes, metadata changes, etc. There was a discussion about possible ways to further analyse the data.
Randall suggested that it might not be possible to put such an emphasis on the distributed aspect of the DVCS, as we haven't used it in that way. He said that we have tested it and found it works as a system to co-version groups of associated files.
There was an extensive discussion about what to include in the paper, and whether to take parts of this paper and put them in other PMR2 publications.
There was a discussion about the best order to publish the PMR2 publications in.
Andre suggested that this paper would make more sense in comparison to a more general PMR2 paper or application note.
Mike said that the paper contains a lot of material that belongs in a more general PMR2 paper. He suggested there might be difficulties in publishing more general papers after publishing this one as is.
Randall suggested publishing a more general PMR2 rationale/overview paper first.
Justin asked Tommy if he can forward around his application note; Tommy will do this.
There was a discussion about how to write and who should write the general PMR2 paper.
The blocker was a) not having anyone with time and b) who should be the first author. Mike suggested it should be a Britten paper, for which there was general agreement. Randall agreed to write up at least
headings and solicit contributions from co-authors similar to the Beard et al. 2009 paper that Peter organised.

Action item 3) Randall will flesh out his outline of the PMR2 general paper. This can be used to help work out where the content should be shared out into the various publications.

4) CellML core specs - update

No update.

5)Metadata specifications (general, citations, graphing, simulation, other) - update

It was noted that Mike will start work on the metadata spec in 2 months time.
Andre briefly reported on some of the discussions on the SED-ML mailing list.

6) OpenCell development

Andrew now has a Windows 7 32bit VM, and is installing all the pre-requisites for the build environment on it. Once this is done he will take a snapshot of the VM. He should then be able to test the installer on Windows 7 32bit.
There was a discussion about upgrading Andrew's system.
Justin has been working on code export and IDA support. IDA support is not straightforward, and is currently failing in OpenCell.
There was a discussion about the causes of this failure.
Andrew spent some time looking into how our build environment could operate if we move to Scala on the Java Virtual Machine. He has tried out Artifactory, which is a Maven-compatible repository for storing snapshots and releases (including caches of external dependencies), and has looked into using an existing Maven archetype for setting up Scala projects.

7) CellML API - update

Justin has been trying to package together RunCellML under windows in such a way that it can be used within Frank Bergmann's libsedml tools to perform simulations using CellML models. Given that a static executable does not appear to work, Andre thinks a single package/zip file that contains all the required dll's plus the RunCellML executable should be ok for Frank to make some progress on using SED-ML with CellML models.

8) Repository contributions - update

Catherine has been working on a large, complex rumen model (Dijkstra, Neal, Beever, France, 1992).
Catherine has added some of Mike Cooling's models.
Hanne continues to add SVG sessions. She has finished the electrophysiology section, and is now doing metabolism models.