ABI CellML meeting minutes 2010-03-17

Present: Randall Britten, Alan Garny, David "Andre" Nickerson, Peter Hunter, Poul Nielsen, Tommy Yu, Andrew Miller, Justin Marsh, Geoff Nunns, Catherine Lloyd

Last week's action items:

1) Dougal to check the browser usage for non-auckland visitors to PMR2.

Firefox was found to be the most common browser, but not by a large margin.

2) Justin to build a snapshot of OpenCell; Geoff to test it on Windows & Linux, Catherine to test it on Mac.

Done.

3) Justin to re-enable COR view on the trunk

Done.

4) Andrew will contact Alan with information on what benchmarks we need regarding speed from Oxford; Alan will find out the information from the servers in Oxford.

Done.

This week's agenda:

1. (Lucian) In certain 1.1 models there are inconsistencies between the capitalisation of the model names and the import declarations. Case differences are not picked up when OpenCell is running on Windows, but they have to be the same under Linux and Mac (example: some of the Guyton models). Question: do we want to enforce the same case under Windows too?

Catherine raised the issue of incorrect capitalization breaking 1.1 models.
There was a discussion of Windows vs Linux file URIs.
Andrew noted that we needed a way to determine which of the many possible file URIs on Windows is the canonical form that we could check was used in URIs. A Mercurial repository would be a possible source for this information.
Randall said that modellers can already test for these problems by uploading a model to PMR2 and running it from there.
Andrew said that we could make a CellML link checker as part of PMR2, which Tommy confirmed.
Geoff suggested that people who don't follow the meetings would be unaware of this issue, and it is therefore important to provide tests and warning messages.
There was an extensive discussion about how to validate URIs, and how to provide warning messages.

2. OpenCell development - update

Justin has made release candidate 3, enabled the COR view on the trunk, and fixed a graphing issue, tracker item 2481.
Alan has been working on a "Qt port", which currently exists as an interface mock-up similar to YAPE.
Alan intends to add raw XML editing via Scintilla, a graph plotting demo, and basic metadata editing functionality to this mock-up.
Peter asked if the mock-up runs on all 3 platforms, Alan confirms that it runs perfectly on all.
Peter asked what downsides Alan has found to Qt C++; Alan says that there is a lot to learn about Qt, and that because it is C++ the compile times could become long.
Alan reported that all the things OpenCell needs are supported by Qt, such as SVG, JavaScript, graphing, Scintilla XML editing, Python scripting etc.
There was a discussion about PyQt vs Qt.
Previous discussion on the topic of alternative platforms for OpenCell can be found in tracker item 1763.
Alan reported that his current prototype has recieved positive feedback.
Andrew reminded us that he is concerned that c++ is not ideal for writing UI logic, and is a step backwards from Javascript. Randall reiterated that Alan's work using Qt does not imply that we will adopt it, but serves to allow us to gain experience with it so that we are better informed to make the choice of the new development technology.

3. CellML API

1. Release process for the API

Andre asked whether there would be announcements for the CellML API release candidates.
After some discussion it was agreed that the policy will be to announce the API release candidates with OpenCell release candidate announcements since the two are currently released as part of the same process.

2. General

Andrew has now produced a Windows installer, and has given it to Alan and Andre to try. There still appear to be a few minor adjustments needed for it to work correctly.
Alan has tried Andrew's installer in a clean Windows 7 VM - the API can be installed, but requires a conflict to be fixed. The OpenCell build failed. Further attempts to install the API have also failed.
Andrew mentioned that Andre's attempts to install the API failed due to permissions settings.
Randall reported that the OpenCMISS project has put Windows CellML support on hold due to these API issues. What is needed is an installer that works for the typical OpenCMISS developer, who already has their own development environment installed.
Randall noted that the OpenCMISS developers have put support for CellML in OpenCMISS from Windows on hold pending the availability of an installer, and suggested that this be a higher priority than support from a clean install. Andrew noted that the installer has been developed to support both cases, and that it would actually be faster to continue to fix problems with both use cases in parallel (Java support on Windows, OpenCMISS support, allowing other developers to test on Windows). There was consensus for this approach.

4. cellml.org - update

Catherine has hosted the EVO recordings from the CellML workshop on BeSTGRID.

5. Repository contributions - update

Catherine has been working on the Decker et al. 2009 cardiac model, but this has a calcium description in the MATLAB that can't be translated directly into CellML. She is trying a workaround from the Hund and Rudy 2004 model and also Keith Decker has offered to look at the CellML file.
There was a discussion about model optimisation.
Geoff has been working on a Conradie et al. immunology/cell cycle model, but this requires events (to describe cell division).
Geoff has been working on two more models from Dan Beard's lab which should be finished today.
There was an extensive discussion about how to implement events, piecewise statements, and other workarounds.

6. PMR2 development - update

Tommy has been working on license and citation support. The license part is mostly done; any number of licenses are supported, and the infrastructure to use licensing is in place. It is possible to extract the license terms from the CellML metadata.

7. CellML specifications

Andrew has implemented full unit support in ModML. Errors will be given if the units are incorrect.
Andrew noted that ModML Units produces errors when there are units mismatches, but due to the run-time nature of error, it is sometimes difficult to pinpoint the underlying cause of the error. Andrew is looking into ways to improve this situation.

8. Metadata specifications (general, citations, graphing, simulation, other)

Catherine is going to send a fully biologically annotated model (Goldbeter 1991) to Sarala and the saint team. This has been annotated with MIRIAM and SBO terms.
There was a discussion about annotating this model.
Catherine said that the model is not structured properly for component annotation, and that only variables have been annotated.