Meeting Minutes 30th October 2006
Meeting Minutes 30th October 2006
Present: Sarala Dissanayake, Edmund Crampin, Poul Nielsen, Shane Blackett, Peter Hunter, Andrew Miller
Poul thinks that we could move discussions online, although he is not sure if teleconferencing is the right way. It will expand the number of people who are aware of what happens in the meetings.
Peter thinks that we can have discussions online, but also believes that face-to-face meetings are important to getting issues resolved.
Shane agrees that it would be useful to expand the number of people, but thinks the Campfire based approach is not a good one, because it is too closed.
IRC is an option, but Poul didn't like it because you can't see the logs of what has happened when you aren't there.
Andrew will look at what approaches we could use to put Jabber/XMPP or IRC based chat logs onto a Plone site (ideally with real-time updates).
Sarala is back from Japan. She has tried to be involved in the SBGN specification, but it seems that SBGN is just an attempt to push Kitano's visualisation approach, and she found they were not receptive to her ideas. Kitano is not keen on combining entity and process diagrams together, because SBML cannot properly represent these (e.g. enzymes which have multiple states are always represented as multiple entities in SBML and Kitano's visualisation, whereas Sarala's visualisation can have one entity in the visualisation representing mutliple states, with modifiers acting on it). Kitano was not receptive to the idea of having multiple levels of detail, with details insignificant at a certain scale collapsed off the diagram. Kohn (based in Washington) has his own graphical format which does collapse multiple states.
Sarala also talked someone who is working with the BioPAX group. They are keen to use Sarala's visualisation and integrate it into Cytoscape (they are working on a BioPAX plugin for Cytoscape).
At the systems biology conference, Sarala met with Jim Schaff, who is working on Virtual Cell. He is keen to use the CellML API to add CellML support to virtual cell.
Peter noted that the European Physiome Project are adopting the CellML and FieldML as their preferred formats. NIH also has working groups looking at this.
He also met with the authors of CasePath, from Case Western Reserve University, which is a metabolic modelling tool which integrates well with databases. It might be possible to integrate CasePath with some of the CellML tools.
Andrew has got the snapshot build process for PCEnv producing working on Linux and Win32 binaries. Andrew has been working on processing feedback on issues on the mailing list.
Alan Garny and David Nickerson are concerned about the architecture of having the backend run in a separate process. Peter noted that it is not necessarily a problem now for users, and so it wouldn't be worth changing it, given that it will allow other tools to work with it. Shane thinks we should wait until we have more than one tool which needs to interact before we put any more work into this (Peter suggested CMISS could be made to use CellML CORBA Server, but Shane would prefer to use MPI as part of a parallelisation effort).
Andrew also noted that he has now separated out the code in the CCGS which computes the procedural steps (and dependencies between them) from the model, from the code which generates the C code. This gives us several benefits:
Peter noted that he talked with Claev Moller, from MathWorks, about the possibility of making MATLAB able to read CellML. Peter thinks that, as a first step, we can write CellML => MATLAB code generation through a modification of CCGS.
Andrew asked for an opinion on what the priorities for adding features to CCGS should be. The consensus was (higher priorities first):
Poul thinks that we could move discussions online, although he is not sure if teleconferencing is the right way. It will expand the number of people who are aware of what happens in the meetings.
Peter thinks that we can have discussions online, but also believes that face-to-face meetings are important to getting issues resolved.
Shane agrees that it would be useful to expand the number of people, but thinks the Campfire based approach is not a good one, because it is too closed.
IRC is an option, but Poul didn't like it because you can't see the logs of what has happened when you aren't there.
Andrew will look at what approaches we could use to put Jabber/XMPP or IRC based chat logs onto a Plone site (ideally with real-time updates).
Sarala is back from Japan. She has tried to be involved in the SBGN specification, but it seems that SBGN is just an attempt to push Kitano's visualisation approach, and she found they were not receptive to her ideas. Kitano is not keen on combining entity and process diagrams together, because SBML cannot properly represent these (e.g. enzymes which have multiple states are always represented as multiple entities in SBML and Kitano's visualisation, whereas Sarala's visualisation can have one entity in the visualisation representing mutliple states, with modifiers acting on it). Kitano was not receptive to the idea of having multiple levels of detail, with details insignificant at a certain scale collapsed off the diagram. Kohn (based in Washington) has his own graphical format which does collapse multiple states.
Sarala also talked someone who is working with the BioPAX group. They are keen to use Sarala's visualisation and integrate it into Cytoscape (they are working on a BioPAX plugin for Cytoscape).
At the systems biology conference, Sarala met with Jim Schaff, who is working on Virtual Cell. He is keen to use the CellML API to add CellML support to virtual cell.
Peter noted that the European Physiome Project are adopting the CellML and FieldML as their preferred formats. NIH also has working groups looking at this.
He also met with the authors of CasePath, from Case Western Reserve University, which is a metabolic modelling tool which integrates well with databases. It might be possible to integrate CasePath with some of the CellML tools.
Andrew has got the snapshot build process for PCEnv producing working on Linux and Win32 binaries. Andrew has been working on processing feedback on issues on the mailing list.
Alan Garny and David Nickerson are concerned about the architecture of having the backend run in a separate process. Peter noted that it is not necessarily a problem now for users, and so it wouldn't be worth changing it, given that it will allow other tools to work with it. Shane thinks we should wait until we have more than one tool which needs to interact before we put any more work into this (Peter suggested CMISS could be made to use CellML CORBA Server, but Shane would prefer to use MPI as part of a parallelisation effort).
Andrew also noted that he has now separated out the code in the CCGS which computes the procedural steps (and dependencies between them) from the model, from the code which generates the C code. This gives us several benefits:
- It is easier to do certain optimisations. This has been used to make computing the Jacobian more efficient.
- It will be easier to support additional languages.
- It will be easier to add an API allowing CellML models to be made part of other models (e.g. to embed CellML models into CMISS).
- The code is easier to read.
Peter noted that he talked with Claev Moller, from MathWorks, about the possibility of making MATLAB able to read CellML. Peter thinks that, as a first step, we can write CellML => MATLAB code generation through a modification of CCGS.
Andrew asked for an opinion on what the priorities for adding features to CCGS should be. The consensus was (higher priorities first):
- Make an official release of what we have now, instead of just snapshot releases.
- Try to improve integration performance, by using CVODE from the SUNDIALS project.
- Investigate the possibility of getting Mac OSX support - Intel only to start with.
- Get editing support for MathML and the CellML structure working.
- Add CellML Metadata support to the backend, and editing support for this to the UI.