ABI Meeting Minutes 2008-02-20

Present: Justin Marsh, Randall Britten, Andrew Miller, Tommy Yu, Catherine Lloyd, Peter Hunter, James Lawson

Apologies: Poul Nielsen

Agenda for this week:

Agenda Item 1 - Catherine, Randall: "Stimulus protocol: user control vs. hard coding"

  • Catherine and Randall raised the issue of how much of a model should be hard coded and how much should be left to the discretion of the user.

  • Currently, curation protocol dictates a model should be coded such that it can be run as-is from the repository in PCEnv, COR or JSim. This currently precludes the possibility of having models that have undefined variables that can be defined by the user. The example of the Irvine et al. 1999 model was raised, which can be run using a range of initial values for the variable defining the voltage at which the membrane is held. At present, it would be necessary to upload thirteen variants of the model to represent this range of values.

  • James suggested the idea of including simulation notes accessible by PCEnv/COR/Jsim etc. with the model to define appropriate ranges for undefined variables.

  • Randall noted that we should consider the role of the model repository and whether the emphasis on configuring a model so that it runs in the CellML simulation environments is correct.

  • Using importing has been discussed in the past as a solution to these issues. A core model could leave all variables with defined initial conditions free, and import these values from a secondary parameters file.

  • It was proposed that the repository might include the core model and also a simulatable implementation of that model, perhaps in the form of a PCEnv session or similar, which assembles the required combination of CellML files to allow simulation.

  • Randall noted that we should not wait until the CellML 1.1 capable PMR2 repository is online to begin working on examples of 1.1 models, where appropriate, to the general agreement of those present.

  • Peter enquired as to the feasibility of creating a tool that is able to flatted CellML 1.1 models into 1.0 form. This has been discussed before and it is not such much the CellML code as the metadata that will be difficult to address.

Agenda Item 2 - Peter: "Multiscale Demo Ideas"

  • Peter has secured a significant grant involving CellML from the European Network of Excellence

  • As part of this relationship, Peter will be attending a meeting in Brussels in June to discuss the software and technical requirements of members of the Network. As such he has been considering what we could provide in the way of demonstration of CellML and related technologies.

  • Specifically, Peter discussed opportunities for further collaboration with Bernard DeBono from the EBI in Cambridge

    • His group has been primarily concerned with genomic and proteomic databases, and is behind the popular Reactome database of protein-protein interactions networks.

    • They have developed an enXML (an XML exchange protocol for EnCore - a project within the ENFIN network of Excellence) to represent and handle this information.

    • Peter is interested in linking CellML to these databases as part of a multiscale modelling framework - i.e. proteins described within CellML models could be linked using RDF to their corresponding entries in databases such as Reactome and the PDB.

    • He envisions being able to click on a protein in an SVG diagram embedded in a PCEnv session. This would bring up a drop-down menu which would give the user options including plotting the concentration or current vs time trace in the graph above (already available,) or opening a browser page directed at the entry for that protein within an online database.

      • He suggests the Ten Tusscher 2004 model as a good example to work on, as it is a popular model in the cardiac electrophysiology modelling field.

Agenda Item 3 - Randall: "Sharing Andrew's time between specification work and PCEnv"

  • Randall notes that while Justin has been learning quickly, Andrew still knows PCEnv the best, and expertise would be very helpful to the project. The request here is that Andrew spend ~50% of his 10 hours a week on PCEnv.

  • The group agreed. Motion passed.

Agenda Item 4 - James: "PCEnv feature request: that axis percentage normalisation be able to be used by sessions"

  • James noted that the axis percentage normalisation function in PCEnv is very useful for creating sessions because it means that traces with different scales can be put on the same graph, thus making better use of limited real estate in the output display. It would also be useful to be able to use the graph gridlines feature.

  • Andrew said that this would require this feature to be defined in the (draft) Graphing and Metadata Specification which PCEnv sessions refer to.

  • David Nickerson has been working on this specification but no additions have been made to it since the middle of last year.

  • Action Item 1.)  Justin will begin the process of addressing this issue by writing a proposal to update the Graphing and Metadata Specification.

Action Item 5 - Randall: "Models in repository that differ from original publication"

  • Tracker item 344 discusses the Youm et al. 2006 model: http://www.cellml.org/models/youm_kim_han_kim_joo_leem_goto_noma_earm_2006_version01

  • Alberto Corrias from the National University of Singapore posted a tracker item expressing concern that the model on the repository is different from the published model.

  • Honours student Nandita Carvalho worked on implementing this model in MATLAB, and ended up having to heavily rework it due to serious inconsistencies and omissions in the paper's description of the model. Thus the model on the site is Catherine's CellML translation of Nandita's MATLAB model.

  • The group agreed that in most cases the repository should contain the original version of the model, whether it works or not. This is useful even in the case that the model is non-functional because it demonstrates this fact. In this case, however, it would be a signficant amount of work to code the original version of the model, and the consensus was that Catherine should not be required to do this if she did not wish to.

  • David Nickerson has expressed his dissatisfaction with this scenario on the tracker - "I must say this is quite a disconcerting issue if the model in the model repository is not actually an implementation of the purported original publication."

Agenda Item 6 - Justin: "PCEnv progress update"

  • Justin has been working on and has fixed a bug relating to the Linux build crashing when loading some models with piecewise equations

Agenda Item 7 - Tommy: "PMR2 progress update"

  • Tommy has been working on 'exposing' models, and allowing models to be viewed in a similar manner to the current repository.

Agenda Item 8 - Andrew: CellML specification update

  • Andrew has completed a mapping between the original and new normative CellML 1.1 specifications

  • He has also been researching possibilities for typing in CellML 1.2, including how OpenMath treats this issue.

  • OpenMath has two typing systems, STS (Simple Typing System) and ECC (a typing system based off the extended calculus of constructions). Andrew noted that it was difficult to express certain types of structures, such as sets, in ECC, and another more recent lambda calculus, called OCC (Open Calculus of Constructions), which came out more recently than the OpenMath type support work, might provide a better basis for a CellML type system.

Further Items:

  • John Davidson has coded up a model in both CellML (using Jsim) and CMISS, but while it works in Jsim, it does not work in PCEnv or COR. PCEnv gives an 'overconstrained' error.

    • Action item 2.)  Catherine will work with John to help fix the CellML model, and then put it up on the repository.

  • Firefox 3 beta has fixed MathML rendering. If we use the new version of Mozilla, we can hopefully get this functionality within PCEnv. Peter said that it would be good to achieve this by the time of the CellML workshop.

  • Regarding an action item from a previous meeting, Peter has contacted Mike Hucka to investigate the possibility of joint-funding for a SBML <-> CellML translation project. Mike Hucka agrees this is a priority and arrangements are progressing.

  • Catherine wants Andrew to make a document describing what is new between CellML 1.0 and 1.1, and 1.1 and 1.2 - this would help people to get involved in the discussions regarding these features.

    • Action Item 3.)  Andrew said that while the 1.2 features obviously aren't fully defined yet, he will be happy to do this.

  • James enquired as to Andrew's arrangements for attending the synthetic biology standards workshop arranged by Herbert Sauro's lab at the University of Washington in the US.

    • Andrew could present some of Mike Cooling's recent work on creating CellML models of iGEM Biobrick-based models. Mike is concerned about intellectual property (IP) arrangements, since he intends to publish this work. Peter doesn't see IP as an issue here.

    • Andrew suggested that he could use a more standard example such as the Ten Tusscher 2004 model, and decompose it into modular elements. Obviously this would not be addressing synthetic biology directly, but the implications would hopefully be clear.

    • Action Item 4.)  Peter will talk to Mike regarding IP