Meeting Minutes 28 March 2007
Meeting Minutes 28 March 2007
Present: James Lawson, Tommy Yu, Poul Nielsen, Andrew Miller
Poul noted that the workshop was next week. He will be sending out an e-mail soon to remind at the institute people that the level 5 seminar room won't be available for other users. He hasn't been in touch with Andre in the last week, but the most recent estimate he heard was that there will be around 30 people attending. Most of these people will arrive on Tuesday. Funding is still an issue, although we have some funding from NZIMA and the Maurice Wilkins CMB CoRE.
Poul has read the minutes from last week. He suggested that if James doesn't understand something from cardiac biology, he ask the relevant person, or if he doesn't know who to ask, try Peter or himself, rather than spend too much time trying to work out things himself.
Tommy noted that he has needed to rewrite a lot of code to get the meta-data processing working. One of the major problems he encountered was that some models referenced multiple papers. Andrew thinks that each model should only reference one model paper. The problem is that papers which describe things other than the model paper are being referenced in the same way as model papers (for example, papers describing the biology which is being modelled). Poul and Andrew agreed that there needs to be some way to separate the 'primary reference' about the paper from 'secondary references'. However, Poul thinks that it might be adequate to merely separate references into a primary and secondary groups, while Andrew would prefer a system where there is an RDF resource about the biology or other aspects underlying the model (such as earlier models, experimental methods, and so on). The reasoning behind this is that, in some cases, there might be several references to the same information (especially in the case of earlier models, where it might be possible to just refer to the earlier CellML model by URL). Also, if we group the references together now, it will be hard to separate them later, while if we categorise them now, it will be easy to add any new categories which may arise later. Andrew will send a proposal to the mailing list about this.
Tommy also noted that Matt had a proposal underway to change the system by which URLs for CellML models are generated, which allows for stable URLs which don't change as new versions come out. Andrew thinks that any such system should add on to the current system (through the use of redirects, or giving the same model multiple URLs), and not replace it. Tommy will contact Matt to try to get the details of his proposal.
James noted that, for some models, there are multiple versions showing up on the main list or models, and for other models, there is only one version showing up on the main list of models, with version links down the bottom. Tommy clarified that all variants show up on the main page, but corresponding versions only show up at the bottom of each page.
James has been continuing to work on the 37 cardiac models from Penny Noble et. al. in Oxford. Most worked without modification in PCEnv. Nine of those which didn't work have since been fixed by James. James is updating the PCEnv curation information in the model repository to reflect this.
James asked if he should be writing documentation on models that lack it. Poul thinks he should be. James also wishes that he could edit the 'temporary documentation' in the CellML models through the repository front-end. The documentation is currently in docbook format, so there is no easy way to do this. However, Poul thinks this would be a good opportunity to review how metadata is represented. Andrew noted that it would probably be a long way off before we can generate sufficient human-readable data directly from the metadata (we at least need Sarala's work, to generate the figures). However, Poul thinks that we can, in the mean time, write documentation in a more controlled format, using a specified set of key-words, to make a later transition to using metadata easier. Tommy will try to arrange a meeting with Tingting and Matt to discuss how to achieve this.
James asked if we were using the wiki to discuss CellML issues. Poul thinks that we should discuss this at the workshop, so will ask Andre if we can find a slot for to allocate for this.
Andrew has written two proposed best current practice documents. The first describes how to include external code in a CellML model. Poul thinks that we need to think more about whether external code which has multiple outputs does this through a vector or a list (so as not to constrain data-type issues in any future version of CellML, as all entries in the vector might by defined to have the same type). However, the consensus was that this doesn't really matter, because if more data-types were supported in a later CellML specification, external code would have to be revised to reflect this anyway. Matt and Andre are the only other people to express a view on this. Andrew thinks that perhaps a face-to-face meeting before the CellML workshop might be a good idea, but we need to see what time Andre will be here, and whether Matt will be able to make the same time.
With regards to the top-level mathematics operator BCP proposal, Poul thinks that we should also require that good patterns write mathematics so that there is only one interpretation of the mathematics (without requiring additional constraints from the pattern definition). In particular, he was concerned that Pattern 4 allows for the possibility that the reset rules are never considered, as long as the rate condition always holds true. The consensus was that we should add in a general guideline that there be only one interpretation for each pattern, and that we change to use piecewise, with the rate equation in the otherwise, and the reset rules in piece elements.
Andrew asked about what we should do to reference tool specific 'sessions' from the repository. As attachments are difficult to implement given the current repository architecture, the consensus was that we should just add fields providing URL references. Another benefit of this is that we can let tool developers store sessions on their own system, and therefore take responsibility for change control. Andrew was concerned that as only one session can be open at a time in PCEnv, it would be good to still allow users to open individual models. Tommy suggested that there be a list of sessions associated with each model, which are displayed alongside the 'Open in PCEnv' link, and the user can either open a session or the model. Poul thinks we should also get Peter's feedback on this, when he is back.
Poul noted that the workshop was next week. He will be sending out an e-mail soon to remind at the institute people that the level 5 seminar room won't be available for other users. He hasn't been in touch with Andre in the last week, but the most recent estimate he heard was that there will be around 30 people attending. Most of these people will arrive on Tuesday. Funding is still an issue, although we have some funding from NZIMA and the Maurice Wilkins CMB CoRE.
Poul has read the minutes from last week. He suggested that if James doesn't understand something from cardiac biology, he ask the relevant person, or if he doesn't know who to ask, try Peter or himself, rather than spend too much time trying to work out things himself.
Tommy noted that he has needed to rewrite a lot of code to get the meta-data processing working. One of the major problems he encountered was that some models referenced multiple papers. Andrew thinks that each model should only reference one model paper. The problem is that papers which describe things other than the model paper are being referenced in the same way as model papers (for example, papers describing the biology which is being modelled). Poul and Andrew agreed that there needs to be some way to separate the 'primary reference' about the paper from 'secondary references'. However, Poul thinks that it might be adequate to merely separate references into a primary and secondary groups, while Andrew would prefer a system where there is an RDF resource about the biology or other aspects underlying the model (such as earlier models, experimental methods, and so on). The reasoning behind this is that, in some cases, there might be several references to the same information (especially in the case of earlier models, where it might be possible to just refer to the earlier CellML model by URL). Also, if we group the references together now, it will be hard to separate them later, while if we categorise them now, it will be easy to add any new categories which may arise later. Andrew will send a proposal to the mailing list about this.
Tommy also noted that Matt had a proposal underway to change the system by which URLs for CellML models are generated, which allows for stable URLs which don't change as new versions come out. Andrew thinks that any such system should add on to the current system (through the use of redirects, or giving the same model multiple URLs), and not replace it. Tommy will contact Matt to try to get the details of his proposal.
James noted that, for some models, there are multiple versions showing up on the main list or models, and for other models, there is only one version showing up on the main list of models, with version links down the bottom. Tommy clarified that all variants show up on the main page, but corresponding versions only show up at the bottom of each page.
James has been continuing to work on the 37 cardiac models from Penny Noble et. al. in Oxford. Most worked without modification in PCEnv. Nine of those which didn't work have since been fixed by James. James is updating the PCEnv curation information in the model repository to reflect this.
James asked if he should be writing documentation on models that lack it. Poul thinks he should be. James also wishes that he could edit the 'temporary documentation' in the CellML models through the repository front-end. The documentation is currently in docbook format, so there is no easy way to do this. However, Poul thinks this would be a good opportunity to review how metadata is represented. Andrew noted that it would probably be a long way off before we can generate sufficient human-readable data directly from the metadata (we at least need Sarala's work, to generate the figures). However, Poul thinks that we can, in the mean time, write documentation in a more controlled format, using a specified set of key-words, to make a later transition to using metadata easier. Tommy will try to arrange a meeting with Tingting and Matt to discuss how to achieve this.
James asked if we were using the wiki to discuss CellML issues. Poul thinks that we should discuss this at the workshop, so will ask Andre if we can find a slot for to allocate for this.
Andrew has written two proposed best current practice documents. The first describes how to include external code in a CellML model. Poul thinks that we need to think more about whether external code which has multiple outputs does this through a vector or a list (so as not to constrain data-type issues in any future version of CellML, as all entries in the vector might by defined to have the same type). However, the consensus was that this doesn't really matter, because if more data-types were supported in a later CellML specification, external code would have to be revised to reflect this anyway. Matt and Andre are the only other people to express a view on this. Andrew thinks that perhaps a face-to-face meeting before the CellML workshop might be a good idea, but we need to see what time Andre will be here, and whether Matt will be able to make the same time.
With regards to the top-level mathematics operator BCP proposal, Poul thinks that we should also require that good patterns write mathematics so that there is only one interpretation of the mathematics (without requiring additional constraints from the pattern definition). In particular, he was concerned that Pattern 4 allows for the possibility that the reset rules are never considered, as long as the rate condition always holds true. The consensus was that we should add in a general guideline that there be only one interpretation for each pattern, and that we change to use piecewise, with the rate equation in the otherwise, and the reset rules in piece elements.
Andrew asked about what we should do to reference tool specific 'sessions' from the repository. As attachments are difficult to implement given the current repository architecture, the consensus was that we should just add fields providing URL references. Another benefit of this is that we can let tool developers store sessions on their own system, and therefore take responsibility for change control. Andrew was concerned that as only one session can be open at a time in PCEnv, it would be good to still allow users to open individual models. Tommy suggested that there be a list of sessions associated with each model, which are displayed alongside the 'Open in PCEnv' link, and the user can either open a session or the model. Poul thinks we should also get Peter's feedback on this, when he is back.