ABI Meeting Minutes 2008-07-02

Present: Andrew Miller, Mike Cooling, James Lawson, Geoff Nunns, Tommy Yu, Dougal Cowan, Justin Marsh, Catherine Lloyd, Randall Britten

Apologies: Poul Nielsen, Peter Hunter, Edmund Crampin


New member: Dougal Cowan has recently started work at the ABI as a technical writer. We will be collaborating with Dougal to have documentation on things CellML written.


Last week's action items


1.)  Justin created Tracker Item 1145 to discuss the workaround required for installing PCEnv on Windows machines without admin privileges.

2.)  Poul has looked into possible sources of funding for the ICBME meeting in Singapore this December.

3.)  Tommy is waiting on IT to help him out with providing an externally accessible instance of the PMR2 prototype

Action item 1.)  Moved forward to next week: Tommy will sort this out with IT when they have time.

4.)  Mike has added his thoughts on the PMR2 prototype to Tracker item 248.

Action item 2.)  Moved forward to next week: James and Catherine still to document their experiences.

5.)  Randall posted the PCEnv 0.5 roadmap to the CellML Discussion mailing list

6.)  Still to do.

Action item 3.)  Moved forward to next week: Andrew to create tracker item on updating code export in API to allow code to be generated for models containing DAE systems.


Agenda for this week


Catherine - Collaboration with Dan Beard's group: 'Biological Components Databank'
  • Dan Beard's group has built a proof-of-concept, wiki-based 'Biological Components Databank'
  • They appear to have a very similar goal to the CellML project in building a library of downloadable modules of code that represent biological components
  • They have put up one CellML model so far - of the enzyme citrate synthase
  • Catherine looked at this model, which is coded using reaction elements
    • She recoded it without the reaction elements, but it was underconstrained
  • Catherine noted that the model appears to have been generated somehow, rather than coded by hand
    • Would be good to ask Dan's group how they generated the CellML, and whether they intend to use this process for more CellML models
    • Andrew wondered if they have a primary format that their models are coded in that they are translating into CellML
    • James suggested that they might simply have the parameters and equations for the models they want to represent
  • Randall is keen for us to collaborate with Dan's group further on this, as we have such similar objectives
Action item 4.)  Catherine will email Dan Beard and request more information on their project

Catherine - ICBME 2008: December 3rd-6th, Singapore
  • Catherine talked to Poul regarding our plans for this conference.
  • It will be expensive to send people to the conference, so:
    • The 'workshop' proposal is probably no longer viable
    • We will probaly send 2-3 people who will present multiple talks / posters each
    • Attendees are allowed to give two presentations with each registration
  • The current proposed list of subjects and authors is:
    • a.) general overview of CellML project (specifications, metadata, Physiome Project) - Poul
    • b.) tools overview - Alan / Andrew / Justin
    • c.) model repository (assuming live PMR2) - Tommy / Randall
    • d.) model development / curation / best practice guidelines - Catherine / James
    • e.) modularity reuse - Mike / Andre
    • f.) visualisation of CellML models and linkage to ontologies - Sarala
  • Abstract submission closes 30th July
  • Mike is not sure whether he will still be with the CellML team in December - he may be taking a post-doctoral position at MIT
  • Randall will not be able to attend
  • Catherine: Sarala said she will probably have finished her PhD by December, and may have some funding to attend the conference using her PhD travel stipend
Action item 5.)  Catherine will talk to Poul further regarding ICME2008 and continue working on the relevant administrative issues.

PMR2 progress update
  • Tommy has received some good feedback on the PMR2 prototype over the last week from Andre
    • James and Randall noted that some of Andre's ideas are more relevant to medium term goals for PMR2
Action item 6.)  Randall will ask Andre to summarise his impression of the PMR2 prototype
  • Tommy has been working on making the prototype software secure
  • He plans to get the PMR2 prototype put up on a VPS, pending input from ABI IT
  • Randall has been testing the prototype - in particular he has been working with CellML 1.1 imports and interfacing with the model files directly through Mercurial
  • Andrew asked if Tommy could link him to the test site
Action item 7.)  Tommy to send the link to the test site to team@cellml.org
  • James has been testing out the utility of Workspaces as a kind of Desktop-like 'working directory'
  • Unfortunately the PMR2 prototype software is not able to handle PCEnv sessions with embedded xul files because of the way that they are referred to in the session file RDF (as a literal rather than a URIref)
  • This is discussed in Tracker item 1079.
  • Andrew noted that fixing this issue would break the session files we currently have.
  • James would prefer this to be done once PMR2 is up, as many of the session files will have to be recreated then anyway.

PCEnv progress update
  • Justin has been looking into getting PCEnv running on the Mac again (Mac development has been stalled for some time)
  • The machine he is working on is running MacOS 10.2, but XULrunner has dependencies that mean it can only be built on MacOS 10.4+
    • Justin has asked IT to upgrade the machine to MacOS 10.5
  • In the mean time, he has been working on some of the smaller bug fixes
  • As a side note, Catherine intends to requisition a Mac laptop, so having PCEnv working on the Mac is important to her

CellML API progress update
  • Andrew is continuing his work on creating validation rules against the CellML Specification Service.
  • He now has most of the rules defined, and is currently working on validation of MathML
  • Randall has created a test-case for the DAE solver which allows the introduction of stiffness
    • The current approach does not alter step-size dynamically in response to stiffness. See Tracker item 417.

CellML 1.2 specification progress update
  • Andrew is continuing his research into the lambda calculus-based typing system
  • He is working on a fairly informal document laying out the broad picture before he gets into defining specific rules

Update on repository contents
  • James has been working on the Sachse 2008 session file
Discussion of unit inconsistencies:
  • Catherine asked about the issue he encountered with unit scaling that made the model give different results in PCEnv / COR and Jsim
  • James had to insert a dimensionless scaling factor into an equation which Jsim inserts automatically when the model is run. This caused the model to work properly in PCEnv and COR but not Jsim
  • After consulting with Andrew, giving the scaling factor the appropriate units solved the problem
  • The following example was considered:
    • X has units mM, Y has units nM
    • model contains equation: X = Y
    • following best practice, the modeller should insert a scaling factor "1nM / 1mM"
    • the equation would then be X = (1 nM / 1 mM) * Y
  • The following table represents the behaviour of Jsim, COR and PCEnv with respect to unit scaling:
 Autoconverts   |
Warns user   |
Assigns as-is
Jsim  
               X  
PCEnv                  X
COR  
                X 
  • Currently PCEnv and COR conform to the CellML specification, whereas Jsim does not
  • Andrew and Mike were concerned that Jsim alters a modeller's code without asking first - there may be instances where such an inconsistency is actually intentional.

  • Randall sees why Jsim's approach seems correct at first, but after discussion, agreed with Andrew and Mike. Randall expressed this as 'x=y is a mathematical statement, not a physical measurement statement'.

  • There was consensus among the group that COR's approach is the best. PCEnv should warn users where such an inconsistency exists
  • James would like this issue to be documented so that future modellers are not confused
Action item 8.)  Andrew to create a tracker item to develop some documentation discussing and comparing treatment of unit inconsistencies between CellML editing tools

Action item 9.)  Andrew will create a tracker item to address this issue within the API's validation service so that the user is warned of the inconsistency.
  • Dougal should be cc'd to the tracker item
  • Geoff now realises the significance of COR's warnings about unit inconsistencies, and will have to go through the models he has authored and fix them

  • Geoff has added the Yamaguchi 1996 lumped parameter model to the repository.
    • It recreates the published results for the simulation, except in the case of FTI_end, FLA_max and ATP, because CellML cannot compute finite integrals or maximums of functions. These values are instead estimated based on graphs, but must be re-estimated for different initial conditions.
    • Andrew noted that the new typing system he is proposing for CellML 1.2 should be able to handle this kind of problem
    • Randall suggested that FieldML might also present a solution
Action item 10.)  Randall to follow up on possibility of leveraging FieldML for representing maxima/minima and finite integrals
  • Geoff is currently coding up two models from ModelDB
    • One of these requires the DAE solver to solve a steady-state approximation, which then informs the system of ODEs
    • James noted that this approach is very common for this kind of model
  • Catherine has put up a model for Kyung-Rok Kwag, who is doing a masters under Marc Jacobs
  • She has also been working on a private, pre-publication model, and the Saucerman & McCulloch 2003 model

Catherine - PCEnv sessions
  • Catherine was curious about exactly what would be involved in learning to create PCEnv sessions for models at a similar level to those James makes
  • James said that actually creating sessions is fairly simple - uploading them onto the repository entails more work, and alteration of URLs within the session etc. Embedding Javascript in SVG files for use in sessions is not so easy.
  • Justin is keen to create a tool that allows users to associate particular functionality with elements of SVG images, but this would probably be a fairly sizeable undertaking.
  • James noted that creating sessions for every model that gets put up is not currently viable as sessions are version-specific, which means the session has to be altered each time a new version of the model is created.

    • PMR2 will address this by supporting consistent relative URLs