Re: [cellml-discussion] ABI CellML meeting minutes 2009-09-30
Hi Dagmar, Andre, all: On VCell side, there is nothing new since August, as we have been tied up with a few major grant submissions (last one, and biggest one, went out the door yesterday...) We did complete in early summer the pure Java library to handle SED-ML elements, as I mentioned in Auckland that we would... The link to the VCell solver is also done, but has the obvious limitations of SBML/VCML incompatibilities. (btw, a lot of these should go away soon due to a parallel development effort) I'm not sure whether the library has been updated to match the latest changes in SED-ML over the summer, and whether it was posted to our Wiki or elsewhere. I am cc'ing Dan, who is the main MIASE-related developer on our team, and should be able to give more detailed answers. Ion From: Dagmar Koehn [dagmar.ko...@uni-rostock.de] Sent: Monday, October 26, 2009 6:44 AM To: CellML Discussion List Cc: Moraru,Ion; Richard Adams Subject: Re: [cellml-discussion] ABI CellML meeting minutes 2009-09-30 Hej Dandre :-), some statements inline. David Nickerson wrote: Hi Dagmar, Thanks for your comments! I'll just make a couple of notes inline below... * Andre said that SED-ML does not yet support everything that we need to do for simulation and graphing. I agree that SED-ML still is at quite an early stage and might not cover everything needed. However, even if the structure of SED-ML does not offer particular constructs for some of your needs, you are allowed (by definition of the SED-ML schema) to attach annotations to *any* SED-ML element. Thus, you could extend SED-ML towards supporting whatever additional needs you might have for information to put in the simulation description file. Do you think that would be sufficient? As Nicolas mentioned, that would actually be a nice benchmark for us to see in what way SED-ML needs to be extended (certainly by what you would put in the annotations often). I agree. The way forward here is to start using SED-ML with CellML models and see how things work out and what possible extensions are required, and also whether these can be incorporated through the existing annotation mechanism. I know that at least myself and maybe others have volunteered to do this at some point, but I have yet to sit down and do it :) We are also hoping that a core SED-ML supporting library or tool might become available that we could work with in order to implement support for CellML. From the Waiheke meeting, Ion offered to make the VCell code available but I haven't had a chance to follow up on this. I cc:ed Ion as I must confess I didn't follow developments on VCell neither. Maybe he could state on how things evolved. Additionally, Richard Adams at Edinburgh University did some work on Java library development for SED-ML. It is all at: http://miase.svn.sourceforge.net/viewvc/miase/sed-ml/jlibsedml/ There also is an online SED-ML file validator at: http://www.bioinformatics.ed.ac.uk:8080/SBSVisualWebApp/html/sedml/ Richard said, it is all still preliminary, but we were discussing on continuing the development, so maybe it would be a good idea to get in contact with him - so as not to do things twice... I also cc:ed Richard :-) * Andrew said that we are still trying to convince the SED-ML group to separate out graphing and simulation. I can say that SED-ML is *not* about graphing at all - in the sense that I understand graphing. SED-ML should provide the description of the data that is used to create the output, and also how these data relate, e.g. for a 2 dimensional plot you would have to specify what to plot against what (x and y axes). Let me cite Nicolas again from an earlier mail: E.g. we can create a report {time, var1, var2}, but some information will only emerge if we plot var1 versus var2. In some sense the relationship between var1 and var2 representation is part of the post-processing. * Andre said that you might want to change some of the graphing metadata to get different graphs from the same simulation, or vice versa. If we are talking about running one simulation and creating a number of different graphs (say, many 2D plots) from that simulation, this is already possible in SED-ML right now. All you have to do is to define a number of (what we call) data generators, referring to the variables/parameters you want to use for your output. Then you can define as many outputs as you want, referring to the same or different data generators and to the same of different simulation setups. If you look at the example given in the publication of CMSB 2008, on page 9/10 (http://www.springerlink.com/content/n67n137071431xt7/), we defined a data generator called time and we use it to create 2 different curves from one single simulation (only the x axis is varying here, using 2 different models). If we, however, are talking about producing the same graph one
Re: [cellml-discussion] ABI CellML meeting minutes 2009-09-30
I cc:ed Ion as I must confess I didn't follow developments on VCell neither. Maybe he could state on how things evolved. Additionally, Richard Adams at Edinburgh University did some work on Java library development for SED-ML. It is all at: http://miase.svn.sourceforge.net/viewvc/miase/sed-ml/jlibsedml/ There also is an online SED-ML file validator at: http://www.bioinformatics.ed.ac.uk:8080/SBSVisualWebApp/html/sedml/ Richard said, it is all still preliminary, but we were discussing on continuing the development, so maybe it would be a good idea to get in contact with him - so as not to do things twice... I also cc:ed Richard :-) thanks Dagmar (and Richard for the correction), if I get a chance I'll be having a look - although my Java is not particularly strong. SED-ML would allow you [being in Auckland] to send me [being in Rostock] the *description* of what you did, so that I can launch a simulation tool and repeat the steps you did. If you want to reuse the simulation data (results), you probably want to use efforts such as SBRML from Pedro Mendes' group to do so (http://www.comp-sys-bio.org/tiki-index.php?page=SBRML). You could include a reference to a result data set encoded in SBRML in your SED-ML file (as a note only, currently) to refer to the result data from your simulation description. ... if that is what you want ... :-) yep. it will certainly be interesting to see how this evolves - whether people actually want to share results or if you always want to re-run the simulation in your own tool :) Cheers, Andre. ___ cellml-discussion mailing list cellml-discussion@cellml.org http://www.cellml.org/mailman/listinfo/cellml-discussion
Re: [cellml-discussion] ABI CellML meeting minutes 2009-09-30
Hi Dagmar, Thanks for your comments! I'll just make a couple of notes inline below... * Andre said that SED-ML does not yet support everything that we need to do for simulation and graphing. I agree that SED-ML still is at quite an early stage and might not cover everything needed. However, even if the structure of SED-ML does not offer particular constructs for some of your needs, you are allowed (by definition of the SED-ML schema) to attach annotations to *any* SED-ML element. Thus, you could extend SED-ML towards supporting whatever additional needs you might have for information to put in the simulation description file. Do you think that would be sufficient? As Nicolas mentioned, that would actually be a nice benchmark for us to see in what way SED-ML needs to be extended (certainly by what you would put in the annotations often). I agree. The way forward here is to start using SED-ML with CellML models and see how things work out and what possible extensions are required, and also whether these can be incorporated through the existing annotation mechanism. I know that at least myself and maybe others have volunteered to do this at some point, but I have yet to sit down and do it :) We are also hoping that a core SED-ML supporting library or tool might become available that we could work with in order to implement support for CellML. From the Waiheke meeting, Ion offered to make the VCell code available but I haven't had a chance to follow up on this. * Andrew said that we are still trying to convince the SED-ML group to separate out graphing and simulation. I can say that SED-ML is *not* about graphing at all - in the sense that I understand graphing. SED-ML should provide the description of the data that is used to create the output, and also how these data relate, e.g. for a 2 dimensional plot you would have to specify what to plot against what (x and y axes). Let me cite Nicolas again from an earlier mail: E.g. we can create a report {time, var1, var2}, but some information will only emerge if we plot var1 versus var2. In some sense the relationship between var1 and var2 representation is part of the post-processing. related to the point below, but what I think this is hinting at is that we prefer to see a distinction between describing and performing a simulation and processing/extracting/manipulating/graphing the resultant data. I don't see this as a major sticking point at present but rather something that will evolve over time if we can get the SBML and CellML communities using a common simulation experiment description framework. * Andre said that you might want to change some of the graphing metadata to get different graphs from the same simulation, or vice versa. If we are talking about running one simulation and creating a number of different graphs (say, many 2D plots) from that simulation, this is already possible in SED-ML right now. All you have to do is to define a number of (what we call) data generators, referring to the variables/parameters you want to use for your output. Then you can define as many outputs as you want, referring to the same or different data generators and to the same of different simulation setups. If you look at the example given in the publication of CMSB 2008, on page 9/10 (http://www.springerlink.com/content/n67n137071431xt7/), we defined a data generator called time and we use it to create 2 different curves from one single simulation (only the x axis is varying here, using 2 different models). If we, however, are talking about producing the same graph one time with a red line, one time with a green line - those things are not part of SED-ML core information, and they should go to the above mentioned annotations. no, what this is addressing that modellers should be able to reuse different parts of the simulation experiment description without needing to redefine it. For example, here in Auckland I define an experiment using the Hodgkin-Huxley model with a certain stimulus protocol and produce some action potentials. Now you over in Rostock want to use that data to produce some current-voltage data. It would be great if you were able to make use of my simulation description and its resultant data without having to redefine the simulation description locally or perhaps even re-running the simulation. I'm not sure if this is already something that SED-ML can include, but it is certainly something that we'd be keen to see in the future. In terms of actually defining the presentation of specific outputs (lines on a graph, surfaces, points, etc), this again is something that we are very keen on in order to be able to unambiguously and completely describe outputs from a simulation experiment. This is not a small problem, although the current CellML graphing metadata begins to address this in regard to specific types of outputs. As above, we don't see this as a major issue blocking the
Re: [cellml-discussion] ABI CellML meeting minutes 2009-09-30
Thanks Dagmar. We are committed to SED-ML! Your comments are very helpful. Peter -- Sent using BlackBerry - Original Message - From: cellml-discussion-boun...@cellml.org cellml-discussion-boun...@cellml.org To: CellML Discussion List cellml-discussion@cellml.org Sent: Fri Oct 09 19:41:19 2009 Subject: Re: [cellml-discussion] ABI CellML meeting minutes 2009-09-30 Dear all, I'd like to comment on the last meeting minutes, particularly on the metadata specification comments made by Andre: * Andre said that SED-ML does not yet support everything that we need to do for simulation and graphing. I agree that SED-ML still is at quite an early stage and might not cover everything needed. However, even if the structure of SED-ML does not offer particular constructs for some of your needs, you are allowed (by definition of the SED-ML schema) to attach annotations to *any* SED-ML element. Thus, you could extend SED-ML towards supporting whatever additional needs you might have for information to put in the simulation description file. Do you think that would be sufficient? As Nicolas mentioned, that would actually be a nice benchmark for us to see in what way SED-ML needs to be extended (certainly by what you would put in the annotations often). * Andrew said that we are still trying to convince the SED-ML group to separate out graphing and simulation. I can say that SED-ML is *not* about graphing at all - in the sense that I understand graphing. SED-ML should provide the description of the data that is used to create the output, and also how these data relate, e.g. for a 2 dimensional plot you would have to specify what to plot against what (x and y axes). Let me cite Nicolas again from an earlier mail: E.g. we can create a report {time, var1, var2}, but some information will only emerge if we plot var1 versus var2. In some sense the relationship between var1 and var2 representation is part of the post-processing. * Andre said that you might want to change some of the graphing metadata to get different graphs from the same simulation, or vice versa. If we are talking about running one simulation and creating a number of different graphs (say, many 2D plots) from that simulation, this is already possible in SED-ML right now. All you have to do is to define a number of (what we call) data generators, referring to the variables/parameters you want to use for your output. Then you can define as many outputs as you want, referring to the same or different data generators and to the same of different simulation setups. If you look at the example given in the publication of CMSB 2008, on page 9/10 (http://www.springerlink.com/content/n67n137071431xt7/), we defined a data generator called time and we use it to create 2 different curves from one single simulation (only the x axis is varying here, using 2 different models). If we, however, are talking about producing the same graph one time with a red line, one time with a green line - those things are not part of SED-ML core information, and they should go to the above mentioned annotations. * Peter said that we want to encourage cooperation, and use accepted standards if they exist. I cannot tell you how much we would like to see CellML using SED-ML :-) I know that we do progress pretty slowly, and I am very sorry for that. * Andre said that the minimal information standard should be out soon. Maybe a word on that: We are currently (and have been for a while... slow again, I know) working on writing up the MIASE guidelines. In my opinion, we do make good progress and I think that we have come to quite a good consensus already. So, there is hope that soon won't take too long. I am happy to answer all the questions you might have! Many greetings from Rostock, Dagmar Dougal Cowan wrote: I have put the minutes from this Wednesday's meeting up at: http://www.cellml.org/community/meeting/minutes/2009/10.07 Dougal __ Information from ESET Smart Security, version of virus signature database 4488 (20091007) __ The message was checked by ESET Smart Security. http://www.eset.com ___ cellml-discussion mailing list cellml-discussion@cellml.org http://www.cellml.org/mailman/listinfo/cellml-discussion ___ cellml-discussion mailing list cellml-discussion@cellml.org http://www.cellml.org/mailman/listinfo/cellml-discussion ___ cellml-discussion mailing list cellml-discussion@cellml.org http://www.cellml.org/mailman/listinfo/cellml-discussion