Meeting Minutes 18 April 2007
Meeting Minutes 18 April 2007
Present: James Lawson, Tommy Yu, Jonna Terkildsen, Ting Ting Chen, Andrew Miller
James has been in contact with Penny Noble. She has sent him all of the models she has been working on. James noted that many of these models are ones that he has already worked on, but some of them are additional variants of the same models, e.g. epicardial, endocardial, and M-cell variants.
Jonna asked if any of these were CellML 1.1 models. Tommy noted that we can't put CellML 1.1 models in the current repository, so we only have 1.0 models at this stage.
James also noted that COR strips out all metadata from the models. He has asked Penny if she has the information for the metadata (such as the original models that she started with, and what changes she has made), but she had effectively no information on this at all. However, James is corresponding with her about what sort of data would be useful in the future, even if she is not putting it into standard CellML metadata herself.
Tommy has been working on getting the CellML metadata editing facilities to use his backend library (as opposed to Carey's code, which would often generate bad XML). This is now mostly working, but there are still a few things that need to be fixed (such as book article references) before it can be put onto the live cellml.org site.
He has also created 'Open in' links for COR and JSim, and has updated this part onto the live site.
Jonna has been having problems with simulating her models where the stimulus was getting skipped over by the integrator due to the stiff nature of the problem. James Lawson has had this problem too. Andrew noted that Andre had also had this problem. The solution at that time was to add a maximum step size. Now that Jonna is aware that the maximum step size can be specified in simulation metadata, this goes part-way to solving the problem. However, this is still not very good when you want to get the model into a 'steady' state under a simulation protocol, because most of the time, hardly anything happens, and the problem is only stiff at a minority of all points in time. There are several options. The ideal option would be to analytically determine, from the mathematics, such 'points of interest', and ensure that the integrator happens to stop on these time-points. The problem is that this is very difficult (indeed, probably impossible) to do in general. We could do it for specific cases that occur commonly in models. Another option would be to provide this information in additional metadata.
Ting Ting has been working on creating a 'sand-box' environment for CellML models, which will ultimately lead to a new CellML repository API. She will initially work on a command-line interface to this. She is currently trying Zope / Python / ZODB, but as part of the project she will evaluate a range of different languages. One of the objectives will be to create a distributed repository, rather than requiring all the models to be held together in one place. In addition, this will ultimately allow for support of CellML 1.1.
Andrew went with Peter Hunter to the mini-workshop on model sharing at the NIH. There was a discussion by Herbert Sauro about SBML, as well as by Peter about CellML. Andrew demonstrated PCEnv, albeit briefly due to limitations on available time.
James and Jonna noted that it was difficult to find connections, because you have to open them all individually. James was keen on adding a new view which still uses the tree layout, but puts connections together with components. Andrew didn't think this was a very good approach, because it duplicated connections in the view, and perhaps more importantly, we could come up with a large number of different views like this, and adding them all in doesn't seem to be very sustainable. Andrew suggested that perhaps a context menu option 'Jump to connection', with a submenu to choose which connection, associated with variables, would be a cleaner and more sustainable way to add this. There was agreement on this. Andrew also noted that we could then add options on the connection / variable mapping to jump back to the variables.
James noted that there were some error messages that popped up on his system when he opens a model but PCEnv was already open. Andrew has seen this before, but it doesn't happen for him, so he can't find the problem. He will look at it on James' system to try to identify the cause of the problem.
James has been in contact with Penny Noble. She has sent him all of the models she has been working on. James noted that many of these models are ones that he has already worked on, but some of them are additional variants of the same models, e.g. epicardial, endocardial, and M-cell variants.
Jonna asked if any of these were CellML 1.1 models. Tommy noted that we can't put CellML 1.1 models in the current repository, so we only have 1.0 models at this stage.
James also noted that COR strips out all metadata from the models. He has asked Penny if she has the information for the metadata (such as the original models that she started with, and what changes she has made), but she had effectively no information on this at all. However, James is corresponding with her about what sort of data would be useful in the future, even if she is not putting it into standard CellML metadata herself.
Tommy has been working on getting the CellML metadata editing facilities to use his backend library (as opposed to Carey's code, which would often generate bad XML). This is now mostly working, but there are still a few things that need to be fixed (such as book article references) before it can be put onto the live cellml.org site.
He has also created 'Open in' links for COR and JSim, and has updated this part onto the live site.
Jonna has been having problems with simulating her models where the stimulus was getting skipped over by the integrator due to the stiff nature of the problem. James Lawson has had this problem too. Andrew noted that Andre had also had this problem. The solution at that time was to add a maximum step size. Now that Jonna is aware that the maximum step size can be specified in simulation metadata, this goes part-way to solving the problem. However, this is still not very good when you want to get the model into a 'steady' state under a simulation protocol, because most of the time, hardly anything happens, and the problem is only stiff at a minority of all points in time. There are several options. The ideal option would be to analytically determine, from the mathematics, such 'points of interest', and ensure that the integrator happens to stop on these time-points. The problem is that this is very difficult (indeed, probably impossible) to do in general. We could do it for specific cases that occur commonly in models. Another option would be to provide this information in additional metadata.
Ting Ting has been working on creating a 'sand-box' environment for CellML models, which will ultimately lead to a new CellML repository API. She will initially work on a command-line interface to this. She is currently trying Zope / Python / ZODB, but as part of the project she will evaluate a range of different languages. One of the objectives will be to create a distributed repository, rather than requiring all the models to be held together in one place. In addition, this will ultimately allow for support of CellML 1.1.
Andrew went with Peter Hunter to the mini-workshop on model sharing at the NIH. There was a discussion by Herbert Sauro about SBML, as well as by Peter about CellML. Andrew demonstrated PCEnv, albeit briefly due to limitations on available time.
James and Jonna noted that it was difficult to find connections, because you have to open them all individually. James was keen on adding a new view which still uses the tree layout, but puts connections together with components. Andrew didn't think this was a very good approach, because it duplicated connections in the view, and perhaps more importantly, we could come up with a large number of different views like this, and adding them all in doesn't seem to be very sustainable. Andrew suggested that perhaps a context menu option 'Jump to connection', with a submenu to choose which connection, associated with variables, would be a cleaner and more sustainable way to add this. There was agreement on this. Andrew also noted that we could then add options on the connection / variable mapping to jump back to the variables.
James noted that there were some error messages that popped up on his system when he opens a model but PCEnv was already open. Andrew has seen this before, but it doesn't happen for him, so he can't find the problem. He will look at it on James' system to try to identify the cause of the problem.