Menu
▾
▴
sed-ml-discuss — Migrating to google groups: https://groups.google.com/d/forum/sed-ml-discuss
This list is closed, nobody may subscribe to it.
| 2008 |
Jan
(1) |
Feb
(35) |
Mar
(41) |
Apr
(4) |
May
(19) |
Jun
(26) |
Jul
(3) |
Aug
(2) |
Sep
(2) |
Oct
(1) |
Nov
|
Dec
(3) |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 2009 |
Jan
(49) |
Feb
(15) |
Mar
(17) |
Apr
(7) |
May
(26) |
Jun
(1) |
Jul
(5) |
Aug
|
Sep
|
Oct
(1) |
Nov
|
Dec
|
| 2010 |
Jan
|
Feb
(1) |
Mar
(29) |
Apr
(4) |
May
(31) |
Jun
(46) |
Jul
|
Aug
(5) |
Sep
(3) |
Oct
(2) |
Nov
(15) |
Dec
|
| 2011 |
Jan
(8) |
Feb
(1) |
Mar
(6) |
Apr
(10) |
May
(17) |
Jun
(23) |
Jul
(5) |
Aug
(3) |
Sep
(28) |
Oct
(41) |
Nov
(20) |
Dec
(1) |
| 2012 |
Jan
(20) |
Feb
(15) |
Mar
(1) |
Apr
(1) |
May
(8) |
Jun
(3) |
Jul
(9) |
Aug
(10) |
Sep
(1) |
Oct
(2) |
Nov
(5) |
Dec
(8) |
| 2013 |
Jan
(2) |
Feb
(1) |
Mar
|
Apr
(16) |
May
(13) |
Jun
(6) |
Jul
(1) |
Aug
(2) |
Sep
(3) |
Oct
(2) |
Nov
(6) |
Dec
(2) |
| 2014 |
Jan
(4) |
Feb
(5) |
Mar
(15) |
Apr
(16) |
May
|
Jun
(6) |
Jul
(3) |
Aug
(2) |
Sep
(1) |
Oct
|
Nov
(13) |
Dec
(8) |
| 2015 |
Jan
(7) |
Feb
|
Mar
(3) |
Apr
|
May
(6) |
Jun
(24) |
Jul
(3) |
Aug
(10) |
Sep
(36) |
Oct
(3) |
Nov
|
Dec
(39) |
| 2016 |
Jan
(9) |
Feb
(38) |
Mar
(25) |
Apr
(3) |
May
(12) |
Jun
(5) |
Jul
(40) |
Aug
(13) |
Sep
(4) |
Oct
|
Nov
|
Dec
(2) |
| 2017 |
Jan
|
Feb
|
Mar
|
Apr
(2) |
May
(29) |
Jun
(26) |
Jul
(12) |
Aug
|
Sep
|
Oct
(4) |
Nov
|
Dec
|
|
From: Nicolas Le N. <n.l...@gm...> - 2015-09-17 16:17:42
|
Indeed. This would require two things: 1- a way to call a DataSource from the parameter attribute of a change class. 2- a way to change variable at runtime. I thing 2) can be solved with a boolean attribute "runTime" on the class Change. If true, the changes are made during the simulation (synchronisation algorithms are a complete different can of worm). If false, the changes are only performed at initial time. To be honest, until we have that, no hybrid modelling is feasible with SED-ML, neither Karr et al, the dynamic FBA of Covert et al, or our biochemical-electrophysiological models. 1) will be interesting ... I believe what you want to do is part of a vast ensemble of models using matrices of values instead of computed values. They are used a lot in other fields, such as ecology, population dynamics etc. We are still catching up with formalism that differ significantly from the ODE+fixed initial conditions paradigm. All that said, it is very likely that no-one could read back your model and simulation description, and run simulations out of the box. Because no software would understand what you would come up with to solve the problem. So I am almost tempted to say use placeholders for the evolving parameters, provide a NuML file with your values and put a text in the metadata file of the COMBINE archive describing how to generate the correct values at the correct time. Then, the users who really want to use your model just write a python script and create piecewise simulations (run, collect result, modify parameter value, run, append collected results etc.) On 17/09/15 16:45, Ally Hume wrote: > Hi everyone, > > I've looked at the L1V3 spec and can't see how I would get time series data into a parameter of an SBML model. > > The only way I know to represent that data in SBML is as a parameter with an associated piecewise assignment function that implements the interpolation. I can't see how L1V3 supports the creation of this piecewise function from the data. I can't see how I would generate the XML for this piecewise function and specify that it must be inserted into the model. So I assume I am talking about entirely new features here, or am I missing something? > > Ally > > > Ally Hume > Researcher, SynthSys, The University of Edinburgh > Software and Data Architect, EPCC, The University of Edinburgh > > > > On 17 Sep 2015, at 09:16, Felix Winter <wi...@ka...> wrote: > >> Hi everyone, >> >> If I am not mistaken the current draft for SED-ML L1V3 does only cover part of >> what Ally was proposing. While it provides a link to data (in the NuML format) >> and a way of using it in the SED-ML file itself there is currently no way to >> describe an interpolation algorithm (or any other pre-processing). >> >> If the pre-processing is not described inside of SED-ML it is quite difficult to >> describe anything slightly more complex than a linear interpolation. For a cubic >> spline interpolation we would still need to pre-process the data and use an SBML >> construct to describe the equation. While just listing (and adressing) the >> individual points of data may be fine for tasks such as parameter value >> estimation this might not suffice if the data is to be used as a continuous >> input into the model. >> >> I would therefore second the motion to use the COMBINE meeting to think beyond >> the current draft and see whether there is a feasable way to describe the way >> the data is used inside a model using SED-ML. >> >> As one of the persons who do regularly not attend the COMBINE and the HARMONY >> meetings but is still interested in SED-ML I would be really grateful if someone >> can share any progress made with the mailing-list. >> >> Thanks, >> Felix >> >>> On September 17, 2015 at 1:56 AM David Nickerson <dav...@gm...> >>> wrote: >>> >>> >>> Hi Ally, Herb, everyone, >>> >>> We don't have a specific agendas set for any of the SED-ML breakout >>> sessions at the COMBINE meeting and the current session titles reflect >>> topics that we believe to be useful topics to at least kick off the >>> sessions. >>> >>> Just over a year ago, the SED-ML editors released a draft for the next >>> version of SED-ML which is available here: >>> http://co.mbine.org/specifications/sed-ml.level-1.version-3-draft-1. >>> The main change in this specification is the addition of data to >>> SED-ML and being able to use data to control the simulation >>> experiment, model variables, etc. The additions to the specification >>> represented the consensus achieved during much discussion over the >>> preceding HARMONY and COMBINE meetings (and various google docs, etc). >>> As far as I am aware, there is currently only one implementation that >>> supports the L1V3 draft (Frank Bergmann's .NET framework) and while >>> some of us hoped to get our own implementations up and running at the >>> HARMONY meeting earlier this year, a second implementation has yet to >>> appear. And thus the L1V3 draft is still a draft :) >>> >>> For the COMBINE breakout on data with SED-ML, it would be good to >>> discuss both the potential implementation of the L1V3 specification in >>> code as well as producing some examples or use-cases (to also help >>> address the issue of examples that Lucian has raised on this list >>> recently). So yes - the topics you mentioned Ally would be ideal to >>> discuss. The main background reading would be the L1V3 draft spec >>> (http://co.mbine.org/specifications/sed-ml.level-1.version-3-draft-1) >>> which would also lead you to the NuML specification: >>> https://github.com/NuML/NuML/blob/master/numl-spec-l1v1.pdf. Since the >>> use of data in SED-ML is dependent on NuML, we will also likely be >>> discussing the state of NuML and where that project is heading. >>> >>> >>> Cheers, >>> David. >>> >>> >>> >>> On Thu, Sep 17, 2015 at 4:39 AM, Herbert Sauro <hs...@gm...> wrote: >>>> We'd be looking for the same things. >>>> >>>> Herbert >>>> >>>> On Wed, Sep 16, 2015 at 7:05 AM, Ally Hume <a....@ed...> wrote: >>>>> >>>>> Hi, >>>>> >>>>> I notice that there will be a session at Combine entitled "Using data with >>>>> SED-ML". I was wondering if somebody could tell me a more detailed >>>>> description of what may be discussed here? I've looked in the archives for >>>>> recent posts to this list but nothing seems to cover this topic. >>>>> >>>>> I've recently been looking at adding external time course data to SBML >>>>> models. In our use case this is temperature data but it could obviously be >>>>> many other measurements as well. I've written code to add external data to >>>>> SBML models as a parameter with an associated assignment rule that defines >>>>> the data using a piecewise function [1]. This works for our purposes but >>>>> does not seem like the cleanest way to do things. >>>>> >>>>> SED-ML looks like a place where this could be done a lot better. >>>>> >>>>> Adopting similar strategies to those already used in SED-ML it would be >>>>> nice to include: >>>>> - references to datasets (similar to references to models) >>>>> - link between entities in these datasets and parameters in a model >>>>> - specification of an interpolation algorithm (ontology-based like >>>>> simulation algorithms) >>>>> >>>>> This would be a considerable improvement over my current approach of >>>>> writing this into SBML where the interpolation algorithm and original >>>>> dataset are effectively lost (although I could add annotations to document >>>>> such details). >>>>> >>>>> There could also be an argument made for some basic data pre-processing >>>>> possibly similar to the post-processing of result data already in SED-ML. >>>>> >>>>> I'm sure there will be alternative ideas but I just wondering if this is >>>>> the type of thing that will be in the session. I will be attending Combine >>>>> but I don't want to wait for the session and discover it's actually on >>>>> something totally different, or discover there is actually a lot of >>>>> background documentation that I could have read in advance. >>>>> >>>>> Thank you, >>>>> >>>>> Ally >>>>> >>>>> [1] https://github.com/allyhume/SBMLDataTools >>>>> >>>>> >>>>> Ally Hume >>>>> Researcher, SynthSys, The University of Edinburgh >>>>> Software and Data Architect, EPCC, The University of Edinburgh >>>>> >>>>> >>>>> -- >>>>> The University of Edinburgh is a charitable body, registered in >>>>> Scotland, with registration number SC005336. >>>>> >>>>> >>>>> >>>>> ------------------------------------------------------------------------------ >>>>> Monitor Your Dynamic Infrastructure at Any Scale With Datadog! >>>>> Get real-time metrics from all of your servers, apps and tools >>>>> in one place. >>>>> SourceForge users - Click here to start your Free Trial of Datadog now! >>>>> http://pubads.g.doubleclick.net/gampad/clk?id=241902991&iu=/4140 >>>>> _______________________________________________ >>>>> SED-ML-discuss mailing list >>>>> SED...@li... >>>>> https://lists.sourceforge.net/lists/listinfo/sed-ml-discuss >>>> >>>> >>>> >>> >>> >>> >>> -- >>> >>> >>> David Nickerson >>> about.me/david.nickerson >>> >> >>> _______________________________________________ >>> SED-ML-discuss mailing list >>> SED...@li... >>> https://lists.sourceforge.net/lists/listinfo/sed-ml-discuss >> Felix Winter >> Rostock, Germany >> >> http://orcid.org/0000-0003-2987-6797 >> >> ------------------------------------------------------------------------------ >> Monitor Your Dynamic Infrastructure at Any Scale With Datadog! >> Get real-time metrics from all of your servers, apps and tools >> in one place. >> SourceForge users - Click here to start your Free Trial of Datadog now! >> http://pubads.g.doubleclick.net/gampad/clk?id=241902991&iu=/4140 >> _______________________________________________ >> SED-ML-discuss mailing list >> SED...@li... >> https://lists.sourceforge.net/lists/listinfo/sed-ml-discuss >> > > -- Nicolas LE NOVERE, Babraham Institute, Babraham Campus Cambridge, CB22 3AT Tel: +441223496433, Mob:+447833147074, twitter:@lenovere, Skype:n.lenovere n.l...@gm..., http://orcid.org/0000-0002-6309-7327 http://lenoverelab.org/, http://lenoverelab.org/perso/lenov/ |
|
From: Ally H. <a....@ed...> - 2015-09-17 15:45:22
|
Hi everyone, I've looked at the L1V3 spec and can't see how I would get time series data into a parameter of an SBML model. The only way I know to represent that data in SBML is as a parameter with an associated piecewise assignment function that implements the interpolation. I can't see how L1V3 supports the creation of this piecewise function from the data. I can't see how I would generate the XML for this piecewise function and specify that it must be inserted into the model. So I assume I am talking about entirely new features here, or am I missing something? Ally Ally Hume Researcher, SynthSys, The University of Edinburgh Software and Data Architect, EPCC, The University of Edinburgh On 17 Sep 2015, at 09:16, Felix Winter <wi...@ka...> wrote: > Hi everyone, > > If I am not mistaken the current draft for SED-ML L1V3 does only cover part of > what Ally was proposing. While it provides a link to data (in the NuML format) > and a way of using it in the SED-ML file itself there is currently no way to > describe an interpolation algorithm (or any other pre-processing). > > If the pre-processing is not described inside of SED-ML it is quite difficult to > describe anything slightly more complex than a linear interpolation. For a cubic > spline interpolation we would still need to pre-process the data and use an SBML > construct to describe the equation. While just listing (and adressing) the > individual points of data may be fine for tasks such as parameter value > estimation this might not suffice if the data is to be used as a continuous > input into the model. > > I would therefore second the motion to use the COMBINE meeting to think beyond > the current draft and see whether there is a feasable way to describe the way > the data is used inside a model using SED-ML. > > As one of the persons who do regularly not attend the COMBINE and the HARMONY > meetings but is still interested in SED-ML I would be really grateful if someone > can share any progress made with the mailing-list. > > Thanks, > Felix > >> On September 17, 2015 at 1:56 AM David Nickerson <dav...@gm...> >> wrote: >> >> >> Hi Ally, Herb, everyone, >> >> We don't have a specific agendas set for any of the SED-ML breakout >> sessions at the COMBINE meeting and the current session titles reflect >> topics that we believe to be useful topics to at least kick off the >> sessions. >> >> Just over a year ago, the SED-ML editors released a draft for the next >> version of SED-ML which is available here: >> http://co.mbine.org/specifications/sed-ml.level-1.version-3-draft-1. >> The main change in this specification is the addition of data to >> SED-ML and being able to use data to control the simulation >> experiment, model variables, etc. The additions to the specification >> represented the consensus achieved during much discussion over the >> preceding HARMONY and COMBINE meetings (and various google docs, etc). >> As far as I am aware, there is currently only one implementation that >> supports the L1V3 draft (Frank Bergmann's .NET framework) and while >> some of us hoped to get our own implementations up and running at the >> HARMONY meeting earlier this year, a second implementation has yet to >> appear. And thus the L1V3 draft is still a draft :) >> >> For the COMBINE breakout on data with SED-ML, it would be good to >> discuss both the potential implementation of the L1V3 specification in >> code as well as producing some examples or use-cases (to also help >> address the issue of examples that Lucian has raised on this list >> recently). So yes - the topics you mentioned Ally would be ideal to >> discuss. The main background reading would be the L1V3 draft spec >> (http://co.mbine.org/specifications/sed-ml.level-1.version-3-draft-1) >> which would also lead you to the NuML specification: >> https://github.com/NuML/NuML/blob/master/numl-spec-l1v1.pdf. Since the >> use of data in SED-ML is dependent on NuML, we will also likely be >> discussing the state of NuML and where that project is heading. >> >> >> Cheers, >> David. >> >> >> >> On Thu, Sep 17, 2015 at 4:39 AM, Herbert Sauro <hs...@gm...> wrote: >>> We'd be looking for the same things. >>> >>> Herbert >>> >>> On Wed, Sep 16, 2015 at 7:05 AM, Ally Hume <a....@ed...> wrote: >>>> >>>> Hi, >>>> >>>> I notice that there will be a session at Combine entitled "Using data with >>>> SED-ML". I was wondering if somebody could tell me a more detailed >>>> description of what may be discussed here? I've looked in the archives for >>>> recent posts to this list but nothing seems to cover this topic. >>>> >>>> I've recently been looking at adding external time course data to SBML >>>> models. In our use case this is temperature data but it could obviously be >>>> many other measurements as well. I've written code to add external data to >>>> SBML models as a parameter with an associated assignment rule that defines >>>> the data using a piecewise function [1]. This works for our purposes but >>>> does not seem like the cleanest way to do things. >>>> >>>> SED-ML looks like a place where this could be done a lot better. >>>> >>>> Adopting similar strategies to those already used in SED-ML it would be >>>> nice to include: >>>> - references to datasets (similar to references to models) >>>> - link between entities in these datasets and parameters in a model >>>> - specification of an interpolation algorithm (ontology-based like >>>> simulation algorithms) >>>> >>>> This would be a considerable improvement over my current approach of >>>> writing this into SBML where the interpolation algorithm and original >>>> dataset are effectively lost (although I could add annotations to document >>>> such details). >>>> >>>> There could also be an argument made for some basic data pre-processing >>>> possibly similar to the post-processing of result data already in SED-ML. >>>> >>>> I'm sure there will be alternative ideas but I just wondering if this is >>>> the type of thing that will be in the session. I will be attending Combine >>>> but I don't want to wait for the session and discover it's actually on >>>> something totally different, or discover there is actually a lot of >>>> background documentation that I could have read in advance. >>>> >>>> Thank you, >>>> >>>> Ally >>>> >>>> [1] https://github.com/allyhume/SBMLDataTools >>>> >>>> >>>> Ally Hume >>>> Researcher, SynthSys, The University of Edinburgh >>>> Software and Data Architect, EPCC, The University of Edinburgh >>>> >>>> >>>> -- >>>> The University of Edinburgh is a charitable body, registered in >>>> Scotland, with registration number SC005336. >>>> >>>> >>>> >>>> ------------------------------------------------------------------------------ >>>> Monitor Your Dynamic Infrastructure at Any Scale With Datadog! >>>> Get real-time metrics from all of your servers, apps and tools >>>> in one place. >>>> SourceForge users - Click here to start your Free Trial of Datadog now! >>>> http://pubads.g.doubleclick.net/gampad/clk?id=241902991&iu=/4140 >>>> _______________________________________________ >>>> SED-ML-discuss mailing list >>>> SED...@li... >>>> https://lists.sourceforge.net/lists/listinfo/sed-ml-discuss >>> >>> >>> >> >> >> >> -- >> >> >> David Nickerson >> about.me/david.nickerson >> > >> _______________________________________________ >> SED-ML-discuss mailing list >> SED...@li... >> https://lists.sourceforge.net/lists/listinfo/sed-ml-discuss > Felix Winter > Rostock, Germany > > http://orcid.org/0000-0003-2987-6797 > > ------------------------------------------------------------------------------ > Monitor Your Dynamic Infrastructure at Any Scale With Datadog! > Get real-time metrics from all of your servers, apps and tools > in one place. > SourceForge users - Click here to start your Free Trial of Datadog now! > http://pubads.g.doubleclick.net/gampad/clk?id=241902991&iu=/4140 > _______________________________________________ > SED-ML-discuss mailing list > SED...@li... > https://lists.sourceforge.net/lists/listinfo/sed-ml-discuss > -- The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336. |
|
From: Nicolas Le N. <n.l...@gm...> - 2015-09-17 14:20:04
|
NuML has been developed as part of SBRML. The idea was to have an XML representation of multi-dimensional arrays. Developed by the same people also developing SBML, SED-ML CellML, SBGN-ML etc., it was kind of homogeneous. The grand idea was that at some point we would have a set of non-overlapping coherent set of XML languages, all using similar mechanisms for mapping, metadata etc. And all encapsulated in a COMBINE archive. So we could use NuML to store lists of values for SBML distrib, output of simulations, experimental measurements to use for parameter estimation etc. That said, I am sure one could do that with several formats around. On 17/09/15 15:03, Frank Bergmann wrote: > Hello Ally, > > the reason is basically, that there was a community vote many years back, where the decision to use NUML rather than any other format. The primary reason is that this way the mapping could be done once, that people implementing support would have to implement. Internally tools could support as many other formats as required / useful for them. > > best > Frank > >> On Sep 17, 2015, at 3:09 PM, Ally Hume <a....@ed...> wrote: >> >> Hi David, everyone, >> >> Thanks for pointing me to these documents (they don't seem to be linked to from the sed-ml.org which is why I only read L1V2 initially). >> >> While I'm can understand that it's a drag when new voices come along querying old decisions I'm going to do that anyway! I apologise if this is derailing so please bat me away if this as ready been endlessly discussed. >> >> Essentially my question is why do we need NuML? Would HDF5 not provide all (and more) of the functionality? Domain specific references to ontology terms could be added to the HDF5 metadata and then it looks like it would provide everything in NuML. HDF5 has widespread adoption, many tools and APIs, and is a far more compact and efficient format for multidimensional data than NuML which looks to be extremely verbose. >> >> It would seem to me to be far simpler to adopt HDF5 and simply add any domain specific metadata we decide is required. Clearly there is an challenge in getting the required implementations for any specifications so where possibly we should be adopting existing standard already accepted and implemented elsewhere. >> >> Regards, >> >> Ally >> >> >> Ally Hume >> Researcher, SynthSys, The University of Edinburgh >> Software and Data Architect, EPCC, The University of Edinburgh >> >> >> >> On 17 Sep 2015, at 00:56, David Nickerson <dav...@gm...> wrote: >> >>> Hi Ally, Herb, everyone, >>> >>> We don't have a specific agendas set for any of the SED-ML breakout >>> sessions at the COMBINE meeting and the current session titles reflect >>> topics that we believe to be useful topics to at least kick off the >>> sessions. >>> >>> Just over a year ago, the SED-ML editors released a draft for the next >>> version of SED-ML which is available here: >>> http://co.mbine.org/specifications/sed-ml.level-1.version-3-draft-1. >>> The main change in this specification is the addition of data to >>> SED-ML and being able to use data to control the simulation >>> experiment, model variables, etc. The additions to the specification >>> represented the consensus achieved during much discussion over the >>> preceding HARMONY and COMBINE meetings (and various google docs, etc). >>> As far as I am aware, there is currently only one implementation that >>> supports the L1V3 draft (Frank Bergmann's .NET framework) and while >>> some of us hoped to get our own implementations up and running at the >>> HARMONY meeting earlier this year, a second implementation has yet to >>> appear. And thus the L1V3 draft is still a draft :) >>> >>> For the COMBINE breakout on data with SED-ML, it would be good to >>> discuss both the potential implementation of the L1V3 specification in >>> code as well as producing some examples or use-cases (to also help >>> address the issue of examples that Lucian has raised on this list >>> recently). So yes - the topics you mentioned Ally would be ideal to >>> discuss. The main background reading would be the L1V3 draft spec >>> (http://co.mbine.org/specifications/sed-ml.level-1.version-3-draft-1) >>> which would also lead you to the NuML specification: >>> https://github.com/NuML/NuML/blob/master/numl-spec-l1v1.pdf. Since the >>> use of data in SED-ML is dependent on NuML, we will also likely be >>> discussing the state of NuML and where that project is heading. >>> >>> >>> Cheers, >>> David. >>> >>> >>> >>> On Thu, Sep 17, 2015 at 4:39 AM, Herbert Sauro <hs...@gm...> wrote: >>>> We'd be looking for the same things. >>>> >>>> Herbert >>>> >>>> On Wed, Sep 16, 2015 at 7:05 AM, Ally Hume <a....@ed...> wrote: >>>>> >>>>> Hi, >>>>> >>>>> I notice that there will be a session at Combine entitled "Using data with >>>>> SED-ML". I was wondering if somebody could tell me a more detailed >>>>> description of what may be discussed here? I've looked in the archives for >>>>> recent posts to this list but nothing seems to cover this topic. >>>>> >>>>> I've recently been looking at adding external time course data to SBML >>>>> models. In our use case this is temperature data but it could obviously be >>>>> many other measurements as well. I've written code to add external data to >>>>> SBML models as a parameter with an associated assignment rule that defines >>>>> the data using a piecewise function [1]. This works for our purposes but >>>>> does not seem like the cleanest way to do things. >>>>> >>>>> SED-ML looks like a place where this could be done a lot better. >>>>> >>>>> Adopting similar strategies to those already used in SED-ML it would be >>>>> nice to include: >>>>> - references to datasets (similar to references to models) >>>>> - link between entities in these datasets and parameters in a model >>>>> - specification of an interpolation algorithm (ontology-based like >>>>> simulation algorithms) >>>>> >>>>> This would be a considerable improvement over my current approach of >>>>> writing this into SBML where the interpolation algorithm and original >>>>> dataset are effectively lost (although I could add annotations to document >>>>> such details). >>>>> >>>>> There could also be an argument made for some basic data pre-processing >>>>> possibly similar to the post-processing of result data already in SED-ML. >>>>> >>>>> I'm sure there will be alternative ideas but I just wondering if this is >>>>> the type of thing that will be in the session. I will be attending Combine >>>>> but I don't want to wait for the session and discover it's actually on >>>>> something totally different, or discover there is actually a lot of >>>>> background documentation that I could have read in advance. >>>>> >>>>> Thank you, >>>>> >>>>> Ally >>>>> >>>>> [1] https://github.com/allyhume/SBMLDataTools >>>>> >>>>> >>>>> Ally Hume >>>>> Researcher, SynthSys, The University of Edinburgh >>>>> Software and Data Architect, EPCC, The University of Edinburgh >>>>> >>>>> >>>>> -- >>>>> The University of Edinburgh is a charitable body, registered in >>>>> Scotland, with registration number SC005336. >>>>> >>>>> >>>>> >>>>> ------------------------------------------------------------------------------ >>>>> Monitor Your Dynamic Infrastructure at Any Scale With Datadog! >>>>> Get real-time metrics from all of your servers, apps and tools >>>>> in one place. >>>>> SourceForge users - Click here to start your Free Trial of Datadog now! >>>>> http://pubads.g.doubleclick.net/gampad/clk?id=241902991&iu=/4140 >>>>> _______________________________________________ >>>>> SED-ML-discuss mailing list >>>>> SED...@li... >>>>> https://lists.sourceforge.net/lists/listinfo/sed-ml-discuss >>>> >>>> >>>> >>>> ------------------------------------------------------------------------------ >>>> Monitor Your Dynamic Infrastructure at Any Scale With Datadog! >>>> Get real-time metrics from all of your servers, apps and tools >>>> in one place. >>>> SourceForge users - Click here to start your Free Trial of Datadog now! >>>> http://pubads.g.doubleclick.net/gampad/clk?id=241902991&iu=/4140 >>>> _______________________________________________ >>>> SED-ML-discuss mailing list >>>> SED...@li... >>>> https://lists.sourceforge.net/lists/listinfo/sed-ml-discuss >>>> >>> >>> >>> >>> -- >>> >>> >>> David Nickerson >>> about.me/david.nickerson >>> >>> ------------------------------------------------------------------------------ >>> Monitor Your Dynamic Infrastructure at Any Scale With Datadog! >>> Get real-time metrics from all of your servers, apps and tools >>> in one place. >>> SourceForge users - Click here to start your Free Trial of Datadog now! >>> http://pubads.g.doubleclick.net/gampad/clk?id=241902991&iu=/4140 >>> _______________________________________________ >>> SED-ML-discuss mailing list >>> SED...@li... >>> https://lists.sourceforge.net/lists/listinfo/sed-ml-discuss >>> >> >> >> -- >> The University of Edinburgh is a charitable body, registered in >> Scotland, with registration number SC005336. >> >> >> ------------------------------------------------------------------------------ >> Monitor Your Dynamic Infrastructure at Any Scale With Datadog! >> Get real-time metrics from all of your servers, apps and tools >> in one place. >> SourceForge users - Click here to start your Free Trial of Datadog now! >> http://pubads.g.doubleclick.net/gampad/clk?id=241902991&iu=/4140 >> _______________________________________________ >> SED-ML-discuss mailing list >> SED...@li... >> https://lists.sourceforge.net/lists/listinfo/sed-ml-discuss > > > ------------------------------------------------------------------------------ > Monitor Your Dynamic Infrastructure at Any Scale With Datadog! > Get real-time metrics from all of your servers, apps and tools > in one place. > SourceForge users - Click here to start your Free Trial of Datadog now! > http://pubads.g.doubleclick.net/gampad/clk?id=241902991&iu=/4140 > _______________________________________________ > SED-ML-discuss mailing list > SED...@li... > https://lists.sourceforge.net/lists/listinfo/sed-ml-discuss > -- Nicolas LE NOVERE, Babraham Institute, Babraham Campus Cambridge, CB22 3AT Tel: +441223496433, Mob:+447833147074, twitter:@lenovere, Skype:n.lenovere n.l...@gm..., http://orcid.org/0000-0002-6309-7327 http://lenoverelab.org/, http://lenoverelab.org/perso/lenov/ |
|
From: Frank B. <fbe...@ca...> - 2015-09-17 14:03:28
|
Hello Ally, the reason is basically, that there was a community vote many years back, where the decision to use NUML rather than any other format. The primary reason is that this way the mapping could be done once, that people implementing support would have to implement. Internally tools could support as many other formats as required / useful for them. best Frank > On Sep 17, 2015, at 3:09 PM, Ally Hume <a....@ed...> wrote: > > Hi David, everyone, > > Thanks for pointing me to these documents (they don't seem to be linked to from the sed-ml.org which is why I only read L1V2 initially). > > While I'm can understand that it's a drag when new voices come along querying old decisions I'm going to do that anyway! I apologise if this is derailing so please bat me away if this as ready been endlessly discussed. > > Essentially my question is why do we need NuML? Would HDF5 not provide all (and more) of the functionality? Domain specific references to ontology terms could be added to the HDF5 metadata and then it looks like it would provide everything in NuML. HDF5 has widespread adoption, many tools and APIs, and is a far more compact and efficient format for multidimensional data than NuML which looks to be extremely verbose. > > It would seem to me to be far simpler to adopt HDF5 and simply add any domain specific metadata we decide is required. Clearly there is an challenge in getting the required implementations for any specifications so where possibly we should be adopting existing standard already accepted and implemented elsewhere. > > Regards, > > Ally > > > Ally Hume > Researcher, SynthSys, The University of Edinburgh > Software and Data Architect, EPCC, The University of Edinburgh > > > > On 17 Sep 2015, at 00:56, David Nickerson <dav...@gm...> wrote: > >> Hi Ally, Herb, everyone, >> >> We don't have a specific agendas set for any of the SED-ML breakout >> sessions at the COMBINE meeting and the current session titles reflect >> topics that we believe to be useful topics to at least kick off the >> sessions. >> >> Just over a year ago, the SED-ML editors released a draft for the next >> version of SED-ML which is available here: >> http://co.mbine.org/specifications/sed-ml.level-1.version-3-draft-1. >> The main change in this specification is the addition of data to >> SED-ML and being able to use data to control the simulation >> experiment, model variables, etc. The additions to the specification >> represented the consensus achieved during much discussion over the >> preceding HARMONY and COMBINE meetings (and various google docs, etc). >> As far as I am aware, there is currently only one implementation that >> supports the L1V3 draft (Frank Bergmann's .NET framework) and while >> some of us hoped to get our own implementations up and running at the >> HARMONY meeting earlier this year, a second implementation has yet to >> appear. And thus the L1V3 draft is still a draft :) >> >> For the COMBINE breakout on data with SED-ML, it would be good to >> discuss both the potential implementation of the L1V3 specification in >> code as well as producing some examples or use-cases (to also help >> address the issue of examples that Lucian has raised on this list >> recently). So yes - the topics you mentioned Ally would be ideal to >> discuss. The main background reading would be the L1V3 draft spec >> (http://co.mbine.org/specifications/sed-ml.level-1.version-3-draft-1) >> which would also lead you to the NuML specification: >> https://github.com/NuML/NuML/blob/master/numl-spec-l1v1.pdf. Since the >> use of data in SED-ML is dependent on NuML, we will also likely be >> discussing the state of NuML and where that project is heading. >> >> >> Cheers, >> David. >> >> >> >> On Thu, Sep 17, 2015 at 4:39 AM, Herbert Sauro <hs...@gm...> wrote: >>> We'd be looking for the same things. >>> >>> Herbert >>> >>> On Wed, Sep 16, 2015 at 7:05 AM, Ally Hume <a....@ed...> wrote: >>>> >>>> Hi, >>>> >>>> I notice that there will be a session at Combine entitled "Using data with >>>> SED-ML". I was wondering if somebody could tell me a more detailed >>>> description of what may be discussed here? I've looked in the archives for >>>> recent posts to this list but nothing seems to cover this topic. >>>> >>>> I've recently been looking at adding external time course data to SBML >>>> models. In our use case this is temperature data but it could obviously be >>>> many other measurements as well. I've written code to add external data to >>>> SBML models as a parameter with an associated assignment rule that defines >>>> the data using a piecewise function [1]. This works for our purposes but >>>> does not seem like the cleanest way to do things. >>>> >>>> SED-ML looks like a place where this could be done a lot better. >>>> >>>> Adopting similar strategies to those already used in SED-ML it would be >>>> nice to include: >>>> - references to datasets (similar to references to models) >>>> - link between entities in these datasets and parameters in a model >>>> - specification of an interpolation algorithm (ontology-based like >>>> simulation algorithms) >>>> >>>> This would be a considerable improvement over my current approach of >>>> writing this into SBML where the interpolation algorithm and original >>>> dataset are effectively lost (although I could add annotations to document >>>> such details). >>>> >>>> There could also be an argument made for some basic data pre-processing >>>> possibly similar to the post-processing of result data already in SED-ML. >>>> >>>> I'm sure there will be alternative ideas but I just wondering if this is >>>> the type of thing that will be in the session. I will be attending Combine >>>> but I don't want to wait for the session and discover it's actually on >>>> something totally different, or discover there is actually a lot of >>>> background documentation that I could have read in advance. >>>> >>>> Thank you, >>>> >>>> Ally >>>> >>>> [1] https://github.com/allyhume/SBMLDataTools >>>> >>>> >>>> Ally Hume >>>> Researcher, SynthSys, The University of Edinburgh >>>> Software and Data Architect, EPCC, The University of Edinburgh >>>> >>>> >>>> -- >>>> The University of Edinburgh is a charitable body, registered in >>>> Scotland, with registration number SC005336. >>>> >>>> >>>> >>>> ------------------------------------------------------------------------------ >>>> Monitor Your Dynamic Infrastructure at Any Scale With Datadog! >>>> Get real-time metrics from all of your servers, apps and tools >>>> in one place. >>>> SourceForge users - Click here to start your Free Trial of Datadog now! >>>> http://pubads.g.doubleclick.net/gampad/clk?id=241902991&iu=/4140 >>>> _______________________________________________ >>>> SED-ML-discuss mailing list >>>> SED...@li... >>>> https://lists.sourceforge.net/lists/listinfo/sed-ml-discuss >>> >>> >>> >>> ------------------------------------------------------------------------------ >>> Monitor Your Dynamic Infrastructure at Any Scale With Datadog! >>> Get real-time metrics from all of your servers, apps and tools >>> in one place. >>> SourceForge users - Click here to start your Free Trial of Datadog now! >>> http://pubads.g.doubleclick.net/gampad/clk?id=241902991&iu=/4140 >>> _______________________________________________ >>> SED-ML-discuss mailing list >>> SED...@li... >>> https://lists.sourceforge.net/lists/listinfo/sed-ml-discuss >>> >> >> >> >> -- >> >> >> David Nickerson >> about.me/david.nickerson >> >> ------------------------------------------------------------------------------ >> Monitor Your Dynamic Infrastructure at Any Scale With Datadog! >> Get real-time metrics from all of your servers, apps and tools >> in one place. >> SourceForge users - Click here to start your Free Trial of Datadog now! >> http://pubads.g.doubleclick.net/gampad/clk?id=241902991&iu=/4140 >> _______________________________________________ >> SED-ML-discuss mailing list >> SED...@li... >> https://lists.sourceforge.net/lists/listinfo/sed-ml-discuss >> > > > -- > The University of Edinburgh is a charitable body, registered in > Scotland, with registration number SC005336. > > > ------------------------------------------------------------------------------ > Monitor Your Dynamic Infrastructure at Any Scale With Datadog! > Get real-time metrics from all of your servers, apps and tools > in one place. > SourceForge users - Click here to start your Free Trial of Datadog now! > http://pubads.g.doubleclick.net/gampad/clk?id=241902991&iu=/4140 > _______________________________________________ > SED-ML-discuss mailing list > SED...@li... > https://lists.sourceforge.net/lists/listinfo/sed-ml-discuss |
|
From: Frank B. <fbe...@ca...> - 2015-09-17 14:00:18
|
> SED-ML editors, should we create includethis draft in > http://co.mbine.org/standards/specifications/ > > And say that people can start experimenting with it? > > Basically, how stable is this draft? > it is very stable, it resulted from discussions that happened two years ago, and have not changed much since. primarily because other implementations were missing … Frank > On 17/09/15 14:09, Ally Hume wrote: >> Hi David, everyone, >> >> Thanks for pointing me to these documents (they don't seem to be linked to from the sed-ml.org which is why I only read L1V2 initially). >> >> While I'm can understand that it's a drag when new voices come along querying old decisions I'm going to do that anyway! I apologise if this is derailing so please bat me away if this as ready been endlessly discussed. >> >> Essentially my question is why do we need NuML? Would HDF5 not provide all (and more) of the functionality? Domain specific references to ontology terms could be added to the HDF5 metadata and then it looks like it would provide everything in NuML. HDF5 has widespread adoption, many tools and APIs, and is a far more compact and efficient format for multidimensional data than NuML which looks to be extremely verbose. >> >> It would seem to me to be far simpler to adopt HDF5 and simply add any domain specific metadata we decide is required. Clearly there is an challenge in getting the required implementations for any specifications so where possibly we should be adopting existing standard already accepted and implemented elsewhere. >> >> Regards, >> >> Ally >> >> >> Ally Hume >> Researcher, SynthSys, The University of Edinburgh >> Software and Data Architect, EPCC, The University of Edinburgh >> >> >> >> On 17 Sep 2015, at 00:56, David Nickerson <dav...@gm...> wrote: >> >>> Hi Ally, Herb, everyone, >>> >>> We don't have a specific agendas set for any of the SED-ML breakout >>> sessions at the COMBINE meeting and the current session titles reflect >>> topics that we believe to be useful topics to at least kick off the >>> sessions. >>> >>> Just over a year ago, the SED-ML editors released a draft for the next >>> version of SED-ML which is available here: >>> http://co.mbine.org/specifications/sed-ml.level-1.version-3-draft-1. >>> The main change in this specification is the addition of data to >>> SED-ML and being able to use data to control the simulation >>> experiment, model variables, etc. The additions to the specification >>> represented the consensus achieved during much discussion over the >>> preceding HARMONY and COMBINE meetings (and various google docs, etc). >>> As far as I am aware, there is currently only one implementation that >>> supports the L1V3 draft (Frank Bergmann's .NET framework) and while >>> some of us hoped to get our own implementations up and running at the >>> HARMONY meeting earlier this year, a second implementation has yet to >>> appear. And thus the L1V3 draft is still a draft :) >>> >>> For the COMBINE breakout on data with SED-ML, it would be good to >>> discuss both the potential implementation of the L1V3 specification in >>> code as well as producing some examples or use-cases (to also help >>> address the issue of examples that Lucian has raised on this list >>> recently). So yes - the topics you mentioned Ally would be ideal to >>> discuss. The main background reading would be the L1V3 draft spec >>> (http://co.mbine.org/specifications/sed-ml.level-1.version-3-draft-1) >>> which would also lead you to the NuML specification: >>> https://github.com/NuML/NuML/blob/master/numl-spec-l1v1.pdf. Since the >>> use of data in SED-ML is dependent on NuML, we will also likely be >>> discussing the state of NuML and where that project is heading. >>> >>> >>> Cheers, >>> David. >>> >>> >>> >>> On Thu, Sep 17, 2015 at 4:39 AM, Herbert Sauro <hs...@gm...> wrote: >>>> We'd be looking for the same things. >>>> >>>> Herbert >>>> >>>> On Wed, Sep 16, 2015 at 7:05 AM, Ally Hume <a....@ed...> wrote: >>>>> >>>>> Hi, >>>>> >>>>> I notice that there will be a session at Combine entitled "Using data with >>>>> SED-ML". I was wondering if somebody could tell me a more detailed >>>>> description of what may be discussed here? I've looked in the archives for >>>>> recent posts to this list but nothing seems to cover this topic. >>>>> >>>>> I've recently been looking at adding external time course data to SBML >>>>> models. In our use case this is temperature data but it could obviously be >>>>> many other measurements as well. I've written code to add external data to >>>>> SBML models as a parameter with an associated assignment rule that defines >>>>> the data using a piecewise function [1]. This works for our purposes but >>>>> does not seem like the cleanest way to do things. >>>>> >>>>> SED-ML looks like a place where this could be done a lot better. >>>>> >>>>> Adopting similar strategies to those already used in SED-ML it would be >>>>> nice to include: >>>>> - references to datasets (similar to references to models) >>>>> - link between entities in these datasets and parameters in a model >>>>> - specification of an interpolation algorithm (ontology-based like >>>>> simulation algorithms) >>>>> >>>>> This would be a considerable improvement over my current approach of >>>>> writing this into SBML where the interpolation algorithm and original >>>>> dataset are effectively lost (although I could add annotations to document >>>>> such details). >>>>> >>>>> There could also be an argument made for some basic data pre-processing >>>>> possibly similar to the post-processing of result data already in SED-ML. >>>>> >>>>> I'm sure there will be alternative ideas but I just wondering if this is >>>>> the type of thing that will be in the session. I will be attending Combine >>>>> but I don't want to wait for the session and discover it's actually on >>>>> something totally different, or discover there is actually a lot of >>>>> background documentation that I could have read in advance. >>>>> >>>>> Thank you, >>>>> >>>>> Ally >>>>> >>>>> [1] https://github.com/allyhume/SBMLDataTools >>>>> >>>>> >>>>> Ally Hume >>>>> Researcher, SynthSys, The University of Edinburgh >>>>> Software and Data Architect, EPCC, The University of Edinburgh >>>>> >>>>> >>>>> -- >>>>> The University of Edinburgh is a charitable body, registered in >>>>> Scotland, with registration number SC005336. >>>>> >>>>> >>>>> >>>>> ------------------------------------------------------------------------------ >>>>> Monitor Your Dynamic Infrastructure at Any Scale With Datadog! >>>>> Get real-time metrics from all of your servers, apps and tools >>>>> in one place. >>>>> SourceForge users - Click here to start your Free Trial of Datadog now! >>>>> http://pubads.g.doubleclick.net/gampad/clk?id=241902991&iu=/4140 >>>>> _______________________________________________ >>>>> SED-ML-discuss mailing list >>>>> SED...@li... >>>>> https://lists.sourceforge.net/lists/listinfo/sed-ml-discuss >>>> >>>> >>>> >>>> ------------------------------------------------------------------------------ >>>> Monitor Your Dynamic Infrastructure at Any Scale With Datadog! >>>> Get real-time metrics from all of your servers, apps and tools >>>> in one place. >>>> SourceForge users - Click here to start your Free Trial of Datadog now! >>>> http://pubads.g.doubleclick.net/gampad/clk?id=241902991&iu=/4140 >>>> _______________________________________________ >>>> SED-ML-discuss mailing list >>>> SED...@li... >>>> https://lists.sourceforge.net/lists/listinfo/sed-ml-discuss >>>> >>> >>> >>> >>> -- >>> >>> >>> David Nickerson >>> about.me/david.nickerson >>> >>> ------------------------------------------------------------------------------ >>> Monitor Your Dynamic Infrastructure at Any Scale With Datadog! >>> Get real-time metrics from all of your servers, apps and tools >>> in one place. >>> SourceForge users - Click here to start your Free Trial of Datadog now! >>> http://pubads.g.doubleclick.net/gampad/clk?id=241902991&iu=/4140 >>> _______________________________________________ >>> SED-ML-discuss mailing list >>> SED...@li... >>> https://lists.sourceforge.net/lists/listinfo/sed-ml-discuss >>> >> >> > > > -- > Nicolas LE NOVERE, Babraham Institute, Babraham Campus Cambridge, CB22 3AT > Tel: +441223496433, Mob:+447833147074, twitter:@lenovere, Skype:n.lenovere > n.l...@gm..., http://orcid.org/0000-0002-6309-7327 > http://lenoverelab.org/, http://lenoverelab.org/perso/lenov/ > > ------------------------------------------------------------------------------ > Monitor Your Dynamic Infrastructure at Any Scale With Datadog! > Get real-time metrics from all of your servers, apps and tools > in one place. > SourceForge users - Click here to start your Free Trial of Datadog now! > http://pubads.g.doubleclick.net/gampad/clk?id=241902991&iu=/4140 > _______________________________________________ > SED-ML-discuss mailing list > SED...@li... > https://lists.sourceforge.net/lists/listinfo/sed-ml-discuss |
|
From: Nicolas Le N. <n.l...@gm...> - 2015-09-17 13:14:14
|
SED-ML editors, should we create includethis draft in http://co.mbine.org/standards/specifications/ And say that people can start experimenting with it? Basically, how stable is this draft? On 17/09/15 14:09, Ally Hume wrote: > Hi David, everyone, > > Thanks for pointing me to these documents (they don't seem to be linked to from the sed-ml.org which is why I only read L1V2 initially). > > While I'm can understand that it's a drag when new voices come along querying old decisions I'm going to do that anyway! I apologise if this is derailing so please bat me away if this as ready been endlessly discussed. > > Essentially my question is why do we need NuML? Would HDF5 not provide all (and more) of the functionality? Domain specific references to ontology terms could be added to the HDF5 metadata and then it looks like it would provide everything in NuML. HDF5 has widespread adoption, many tools and APIs, and is a far more compact and efficient format for multidimensional data than NuML which looks to be extremely verbose. > > It would seem to me to be far simpler to adopt HDF5 and simply add any domain specific metadata we decide is required. Clearly there is an challenge in getting the required implementations for any specifications so where possibly we should be adopting existing standard already accepted and implemented elsewhere. > > Regards, > > Ally > > > Ally Hume > Researcher, SynthSys, The University of Edinburgh > Software and Data Architect, EPCC, The University of Edinburgh > > > > On 17 Sep 2015, at 00:56, David Nickerson <dav...@gm...> wrote: > >> Hi Ally, Herb, everyone, >> >> We don't have a specific agendas set for any of the SED-ML breakout >> sessions at the COMBINE meeting and the current session titles reflect >> topics that we believe to be useful topics to at least kick off the >> sessions. >> >> Just over a year ago, the SED-ML editors released a draft for the next >> version of SED-ML which is available here: >> http://co.mbine.org/specifications/sed-ml.level-1.version-3-draft-1. >> The main change in this specification is the addition of data to >> SED-ML and being able to use data to control the simulation >> experiment, model variables, etc. The additions to the specification >> represented the consensus achieved during much discussion over the >> preceding HARMONY and COMBINE meetings (and various google docs, etc). >> As far as I am aware, there is currently only one implementation that >> supports the L1V3 draft (Frank Bergmann's .NET framework) and while >> some of us hoped to get our own implementations up and running at the >> HARMONY meeting earlier this year, a second implementation has yet to >> appear. And thus the L1V3 draft is still a draft :) >> >> For the COMBINE breakout on data with SED-ML, it would be good to >> discuss both the potential implementation of the L1V3 specification in >> code as well as producing some examples or use-cases (to also help >> address the issue of examples that Lucian has raised on this list >> recently). So yes - the topics you mentioned Ally would be ideal to >> discuss. The main background reading would be the L1V3 draft spec >> (http://co.mbine.org/specifications/sed-ml.level-1.version-3-draft-1) >> which would also lead you to the NuML specification: >> https://github.com/NuML/NuML/blob/master/numl-spec-l1v1.pdf. Since the >> use of data in SED-ML is dependent on NuML, we will also likely be >> discussing the state of NuML and where that project is heading. >> >> >> Cheers, >> David. >> >> >> >> On Thu, Sep 17, 2015 at 4:39 AM, Herbert Sauro <hs...@gm...> wrote: >>> We'd be looking for the same things. >>> >>> Herbert >>> >>> On Wed, Sep 16, 2015 at 7:05 AM, Ally Hume <a....@ed...> wrote: >>>> >>>> Hi, >>>> >>>> I notice that there will be a session at Combine entitled "Using data with >>>> SED-ML". I was wondering if somebody could tell me a more detailed >>>> description of what may be discussed here? I've looked in the archives for >>>> recent posts to this list but nothing seems to cover this topic. >>>> >>>> I've recently been looking at adding external time course data to SBML >>>> models. In our use case this is temperature data but it could obviously be >>>> many other measurements as well. I've written code to add external data to >>>> SBML models as a parameter with an associated assignment rule that defines >>>> the data using a piecewise function [1]. This works for our purposes but >>>> does not seem like the cleanest way to do things. >>>> >>>> SED-ML looks like a place where this could be done a lot better. >>>> >>>> Adopting similar strategies to those already used in SED-ML it would be >>>> nice to include: >>>> - references to datasets (similar to references to models) >>>> - link between entities in these datasets and parameters in a model >>>> - specification of an interpolation algorithm (ontology-based like >>>> simulation algorithms) >>>> >>>> This would be a considerable improvement over my current approach of >>>> writing this into SBML where the interpolation algorithm and original >>>> dataset are effectively lost (although I could add annotations to document >>>> such details). >>>> >>>> There could also be an argument made for some basic data pre-processing >>>> possibly similar to the post-processing of result data already in SED-ML. >>>> >>>> I'm sure there will be alternative ideas but I just wondering if this is >>>> the type of thing that will be in the session. I will be attending Combine >>>> but I don't want to wait for the session and discover it's actually on >>>> something totally different, or discover there is actually a lot of >>>> background documentation that I could have read in advance. >>>> >>>> Thank you, >>>> >>>> Ally >>>> >>>> [1] https://github.com/allyhume/SBMLDataTools >>>> >>>> >>>> Ally Hume >>>> Researcher, SynthSys, The University of Edinburgh >>>> Software and Data Architect, EPCC, The University of Edinburgh >>>> >>>> >>>> -- >>>> The University of Edinburgh is a charitable body, registered in >>>> Scotland, with registration number SC005336. >>>> >>>> >>>> >>>> ------------------------------------------------------------------------------ >>>> Monitor Your Dynamic Infrastructure at Any Scale With Datadog! >>>> Get real-time metrics from all of your servers, apps and tools >>>> in one place. >>>> SourceForge users - Click here to start your Free Trial of Datadog now! >>>> http://pubads.g.doubleclick.net/gampad/clk?id=241902991&iu=/4140 >>>> _______________________________________________ >>>> SED-ML-discuss mailing list >>>> SED...@li... >>>> https://lists.sourceforge.net/lists/listinfo/sed-ml-discuss >>> >>> >>> >>> ------------------------------------------------------------------------------ >>> Monitor Your Dynamic Infrastructure at Any Scale With Datadog! >>> Get real-time metrics from all of your servers, apps and tools >>> in one place. >>> SourceForge users - Click here to start your Free Trial of Datadog now! >>> http://pubads.g.doubleclick.net/gampad/clk?id=241902991&iu=/4140 >>> _______________________________________________ >>> SED-ML-discuss mailing list >>> SED...@li... >>> https://lists.sourceforge.net/lists/listinfo/sed-ml-discuss >>> >> >> >> >> -- >> >> >> David Nickerson >> about.me/david.nickerson >> >> ------------------------------------------------------------------------------ >> Monitor Your Dynamic Infrastructure at Any Scale With Datadog! >> Get real-time metrics from all of your servers, apps and tools >> in one place. >> SourceForge users - Click here to start your Free Trial of Datadog now! >> http://pubads.g.doubleclick.net/gampad/clk?id=241902991&iu=/4140 >> _______________________________________________ >> SED-ML-discuss mailing list >> SED...@li... >> https://lists.sourceforge.net/lists/listinfo/sed-ml-discuss >> > > -- Nicolas LE NOVERE, Babraham Institute, Babraham Campus Cambridge, CB22 3AT Tel: +441223496433, Mob:+447833147074, twitter:@lenovere, Skype:n.lenovere n.l...@gm..., http://orcid.org/0000-0002-6309-7327 http://lenoverelab.org/, http://lenoverelab.org/perso/lenov/ |
|
From: Ally H. <a....@ed...> - 2015-09-17 13:08:51
|
Hi David, everyone, Thanks for pointing me to these documents (they don't seem to be linked to from the sed-ml.org which is why I only read L1V2 initially). While I'm can understand that it's a drag when new voices come along querying old decisions I'm going to do that anyway! I apologise if this is derailing so please bat me away if this as ready been endlessly discussed. Essentially my question is why do we need NuML? Would HDF5 not provide all (and more) of the functionality? Domain specific references to ontology terms could be added to the HDF5 metadata and then it looks like it would provide everything in NuML. HDF5 has widespread adoption, many tools and APIs, and is a far more compact and efficient format for multidimensional data than NuML which looks to be extremely verbose. It would seem to me to be far simpler to adopt HDF5 and simply add any domain specific metadata we decide is required. Clearly there is an challenge in getting the required implementations for any specifications so where possibly we should be adopting existing standard already accepted and implemented elsewhere. Regards, Ally Ally Hume Researcher, SynthSys, The University of Edinburgh Software and Data Architect, EPCC, The University of Edinburgh On 17 Sep 2015, at 00:56, David Nickerson <dav...@gm...> wrote: > Hi Ally, Herb, everyone, > > We don't have a specific agendas set for any of the SED-ML breakout > sessions at the COMBINE meeting and the current session titles reflect > topics that we believe to be useful topics to at least kick off the > sessions. > > Just over a year ago, the SED-ML editors released a draft for the next > version of SED-ML which is available here: > http://co.mbine.org/specifications/sed-ml.level-1.version-3-draft-1. > The main change in this specification is the addition of data to > SED-ML and being able to use data to control the simulation > experiment, model variables, etc. The additions to the specification > represented the consensus achieved during much discussion over the > preceding HARMONY and COMBINE meetings (and various google docs, etc). > As far as I am aware, there is currently only one implementation that > supports the L1V3 draft (Frank Bergmann's .NET framework) and while > some of us hoped to get our own implementations up and running at the > HARMONY meeting earlier this year, a second implementation has yet to > appear. And thus the L1V3 draft is still a draft :) > > For the COMBINE breakout on data with SED-ML, it would be good to > discuss both the potential implementation of the L1V3 specification in > code as well as producing some examples or use-cases (to also help > address the issue of examples that Lucian has raised on this list > recently). So yes - the topics you mentioned Ally would be ideal to > discuss. The main background reading would be the L1V3 draft spec > (http://co.mbine.org/specifications/sed-ml.level-1.version-3-draft-1) > which would also lead you to the NuML specification: > https://github.com/NuML/NuML/blob/master/numl-spec-l1v1.pdf. Since the > use of data in SED-ML is dependent on NuML, we will also likely be > discussing the state of NuML and where that project is heading. > > > Cheers, > David. > > > > On Thu, Sep 17, 2015 at 4:39 AM, Herbert Sauro <hs...@gm...> wrote: >> We'd be looking for the same things. >> >> Herbert >> >> On Wed, Sep 16, 2015 at 7:05 AM, Ally Hume <a....@ed...> wrote: >>> >>> Hi, >>> >>> I notice that there will be a session at Combine entitled "Using data with >>> SED-ML". I was wondering if somebody could tell me a more detailed >>> description of what may be discussed here? I've looked in the archives for >>> recent posts to this list but nothing seems to cover this topic. >>> >>> I've recently been looking at adding external time course data to SBML >>> models. In our use case this is temperature data but it could obviously be >>> many other measurements as well. I've written code to add external data to >>> SBML models as a parameter with an associated assignment rule that defines >>> the data using a piecewise function [1]. This works for our purposes but >>> does not seem like the cleanest way to do things. >>> >>> SED-ML looks like a place where this could be done a lot better. >>> >>> Adopting similar strategies to those already used in SED-ML it would be >>> nice to include: >>> - references to datasets (similar to references to models) >>> - link between entities in these datasets and parameters in a model >>> - specification of an interpolation algorithm (ontology-based like >>> simulation algorithms) >>> >>> This would be a considerable improvement over my current approach of >>> writing this into SBML where the interpolation algorithm and original >>> dataset are effectively lost (although I could add annotations to document >>> such details). >>> >>> There could also be an argument made for some basic data pre-processing >>> possibly similar to the post-processing of result data already in SED-ML. >>> >>> I'm sure there will be alternative ideas but I just wondering if this is >>> the type of thing that will be in the session. I will be attending Combine >>> but I don't want to wait for the session and discover it's actually on >>> something totally different, or discover there is actually a lot of >>> background documentation that I could have read in advance. >>> >>> Thank you, >>> >>> Ally >>> >>> [1] https://github.com/allyhume/SBMLDataTools >>> >>> >>> Ally Hume >>> Researcher, SynthSys, The University of Edinburgh >>> Software and Data Architect, EPCC, The University of Edinburgh >>> >>> >>> -- >>> The University of Edinburgh is a charitable body, registered in >>> Scotland, with registration number SC005336. >>> >>> >>> >>> ------------------------------------------------------------------------------ >>> Monitor Your Dynamic Infrastructure at Any Scale With Datadog! >>> Get real-time metrics from all of your servers, apps and tools >>> in one place. >>> SourceForge users - Click here to start your Free Trial of Datadog now! >>> http://pubads.g.doubleclick.net/gampad/clk?id=241902991&iu=/4140 >>> _______________________________________________ >>> SED-ML-discuss mailing list >>> SED...@li... >>> https://lists.sourceforge.net/lists/listinfo/sed-ml-discuss >> >> >> >> ------------------------------------------------------------------------------ >> Monitor Your Dynamic Infrastructure at Any Scale With Datadog! >> Get real-time metrics from all of your servers, apps and tools >> in one place. >> SourceForge users - Click here to start your Free Trial of Datadog now! >> http://pubads.g.doubleclick.net/gampad/clk?id=241902991&iu=/4140 >> _______________________________________________ >> SED-ML-discuss mailing list >> SED...@li... >> https://lists.sourceforge.net/lists/listinfo/sed-ml-discuss >> > > > > -- > > > David Nickerson > about.me/david.nickerson > > ------------------------------------------------------------------------------ > Monitor Your Dynamic Infrastructure at Any Scale With Datadog! > Get real-time metrics from all of your servers, apps and tools > in one place. > SourceForge users - Click here to start your Free Trial of Datadog now! > http://pubads.g.doubleclick.net/gampad/clk?id=241902991&iu=/4140 > _______________________________________________ > SED-ML-discuss mailing list > SED...@li... > https://lists.sourceforge.net/lists/listinfo/sed-ml-discuss > -- The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336. |
|
From: Felix W. <wi...@ka...> - 2015-09-17 08:16:34
|
Hi everyone, If I am not mistaken the current draft for SED-ML L1V3 does only cover part of what Ally was proposing. While it provides a link to data (in the NuML format) and a way of using it in the SED-ML file itself there is currently no way to describe an interpolation algorithm (or any other pre-processing). If the pre-processing is not described inside of SED-ML it is quite difficult to describe anything slightly more complex than a linear interpolation. For a cubic spline interpolation we would still need to pre-process the data and use an SBML construct to describe the equation. While just listing (and adressing) the individual points of data may be fine for tasks such as parameter value estimation this might not suffice if the data is to be used as a continuous input into the model. I would therefore second the motion to use the COMBINE meeting to think beyond the current draft and see whether there is a feasable way to describe the way the data is used inside a model using SED-ML. As one of the persons who do regularly not attend the COMBINE and the HARMONY meetings but is still interested in SED-ML I would be really grateful if someone can share any progress made with the mailing-list. Thanks, Felix > On September 17, 2015 at 1:56 AM David Nickerson <dav...@gm...> > wrote: > > > Hi Ally, Herb, everyone, > > We don't have a specific agendas set for any of the SED-ML breakout > sessions at the COMBINE meeting and the current session titles reflect > topics that we believe to be useful topics to at least kick off the > sessions. > > Just over a year ago, the SED-ML editors released a draft for the next > version of SED-ML which is available here: > http://co.mbine.org/specifications/sed-ml.level-1.version-3-draft-1. > The main change in this specification is the addition of data to > SED-ML and being able to use data to control the simulation > experiment, model variables, etc. The additions to the specification > represented the consensus achieved during much discussion over the > preceding HARMONY and COMBINE meetings (and various google docs, etc). > As far as I am aware, there is currently only one implementation that > supports the L1V3 draft (Frank Bergmann's .NET framework) and while > some of us hoped to get our own implementations up and running at the > HARMONY meeting earlier this year, a second implementation has yet to > appear. And thus the L1V3 draft is still a draft :) > > For the COMBINE breakout on data with SED-ML, it would be good to > discuss both the potential implementation of the L1V3 specification in > code as well as producing some examples or use-cases (to also help > address the issue of examples that Lucian has raised on this list > recently). So yes - the topics you mentioned Ally would be ideal to > discuss. The main background reading would be the L1V3 draft spec > (http://co.mbine.org/specifications/sed-ml.level-1.version-3-draft-1) > which would also lead you to the NuML specification: > https://github.com/NuML/NuML/blob/master/numl-spec-l1v1.pdf. Since the > use of data in SED-ML is dependent on NuML, we will also likely be > discussing the state of NuML and where that project is heading. > > > Cheers, > David. > > > > On Thu, Sep 17, 2015 at 4:39 AM, Herbert Sauro <hs...@gm...> wrote: > > We'd be looking for the same things. > > > > Herbert > > > > On Wed, Sep 16, 2015 at 7:05 AM, Ally Hume <a....@ed...> wrote: > >> > >> Hi, > >> > >> I notice that there will be a session at Combine entitled "Using data with > >> SED-ML". I was wondering if somebody could tell me a more detailed > >> description of what may be discussed here? I've looked in the archives for > >> recent posts to this list but nothing seems to cover this topic. > >> > >> I've recently been looking at adding external time course data to SBML > >> models. In our use case this is temperature data but it could obviously be > >> many other measurements as well. I've written code to add external data to > >> SBML models as a parameter with an associated assignment rule that defines > >> the data using a piecewise function [1]. This works for our purposes but > >> does not seem like the cleanest way to do things. > >> > >> SED-ML looks like a place where this could be done a lot better. > >> > >> Adopting similar strategies to those already used in SED-ML it would be > >> nice to include: > >> - references to datasets (similar to references to models) > >> - link between entities in these datasets and parameters in a model > >> - specification of an interpolation algorithm (ontology-based like > >> simulation algorithms) > >> > >> This would be a considerable improvement over my current approach of > >> writing this into SBML where the interpolation algorithm and original > >> dataset are effectively lost (although I could add annotations to document > >> such details). > >> > >> There could also be an argument made for some basic data pre-processing > >> possibly similar to the post-processing of result data already in SED-ML. > >> > >> I'm sure there will be alternative ideas but I just wondering if this is > >> the type of thing that will be in the session. I will be attending Combine > >> but I don't want to wait for the session and discover it's actually on > >> something totally different, or discover there is actually a lot of > >> background documentation that I could have read in advance. > >> > >> Thank you, > >> > >> Ally > >> > >> [1] https://github.com/allyhume/SBMLDataTools > >> > >> > >> Ally Hume > >> Researcher, SynthSys, The University of Edinburgh > >> Software and Data Architect, EPCC, The University of Edinburgh > >> > >> > >> -- > >> The University of Edinburgh is a charitable body, registered in > >> Scotland, with registration number SC005336. > >> > >> > >> > >> ------------------------------------------------------------------------------ > >> Monitor Your Dynamic Infrastructure at Any Scale With Datadog! > >> Get real-time metrics from all of your servers, apps and tools > >> in one place. > >> SourceForge users - Click here to start your Free Trial of Datadog now! > >> http://pubads.g.doubleclick.net/gampad/clk?id=241902991&iu=/4140 > >> _______________________________________________ > >> SED-ML-discuss mailing list > >> SED...@li... > >> https://lists.sourceforge.net/lists/listinfo/sed-ml-discuss > > > > > > > > > > -- > > > David Nickerson > about.me/david.nickerson > > _______________________________________________ > SED-ML-discuss mailing list > SED...@li... > https://lists.sourceforge.net/lists/listinfo/sed-ml-discuss Felix Winter Rostock, Germany http://orcid.org/0000-0003-2987-6797 |
|
From: Frank B. <fbe...@ca...> - 2015-09-17 06:49:44
|
I’m afraid I’m away currently on a conference, with limited internet access. I can look into getting limited support into the C++ library, such that the direct elements discussed in the spec draft could be set. However, any of the advanced processing like implemented in the .net library (which is actually resolving and mapping the actual data) would not be available in this version. Cheers Frank > On Sep 17, 2015, at 6:43 AM, David Nickerson <dav...@gm...> wrote: > > On Thu, Sep 17, 2015 at 3:37 PM, Herbert Sauro <hs...@gm...> wrote: >> In that case what we would really like is an update to Frank's C/C++ library >> together with Python bindings. I think that would help a lot of people. >> Maybe this is something that can be discussed? Looks like the spec is in >> place but we need developer help. > > yep - exactly correct, that is likely to be the most useful path > forward. And certainly a key part of the discussion. As it will be > with most discussions regarding tool support for new SED-ML > features... > > > Cheers, > David. > > ------------------------------------------------------------------------------ > Monitor Your Dynamic Infrastructure at Any Scale With Datadog! > Get real-time metrics from all of your servers, apps and tools > in one place. > SourceForge users - Click here to start your Free Trial of Datadog now! > http://pubads.g.doubleclick.net/gampad/clk?id=241902991&iu=/4140 > _______________________________________________ > SED-ML-discuss mailing list > SED...@li... > https://lists.sourceforge.net/lists/listinfo/sed-ml-discuss |
|
From: David N. <dav...@gm...> - 2015-09-17 04:43:40
|
On Thu, Sep 17, 2015 at 3:37 PM, Herbert Sauro <hs...@gm...> wrote: > In that case what we would really like is an update to Frank's C/C++ library > together with Python bindings. I think that would help a lot of people. > Maybe this is something that can be discussed? Looks like the spec is in > place but we need developer help. yep - exactly correct, that is likely to be the most useful path forward. And certainly a key part of the discussion. As it will be with most discussions regarding tool support for new SED-ML features... Cheers, David. |
|
From: Herbert S. <hs...@gm...> - 2015-09-17 03:37:16
|
In that case what we would really like is an update to Frank's C/C++ library together with Python bindings. I think that would help a lot of people. Maybe this is something that can be discussed? Looks like the spec is in place but we need developer help. Herbert On Wed, Sep 16, 2015 at 4:56 PM, David Nickerson <dav...@gm...> wrote: > Hi Ally, Herb, everyone, > > We don't have a specific agendas set for any of the SED-ML breakout > sessions at the COMBINE meeting and the current session titles reflect > topics that we believe to be useful topics to at least kick off the > sessions. > > Just over a year ago, the SED-ML editors released a draft for the next > version of SED-ML which is available here: > http://co.mbine.org/specifications/sed-ml.level-1.version-3-draft-1. > The main change in this specification is the addition of data to > SED-ML and being able to use data to control the simulation > experiment, model variables, etc. The additions to the specification > represented the consensus achieved during much discussion over the > preceding HARMONY and COMBINE meetings (and various google docs, etc). > As far as I am aware, there is currently only one implementation that > supports the L1V3 draft (Frank Bergmann's .NET framework) and while > some of us hoped to get our own implementations up and running at the > HARMONY meeting earlier this year, a second implementation has yet to > appear. And thus the L1V3 draft is still a draft :) > > For the COMBINE breakout on data with SED-ML, it would be good to > discuss both the potential implementation of the L1V3 specification in > code as well as producing some examples or use-cases (to also help > address the issue of examples that Lucian has raised on this list > recently). So yes - the topics you mentioned Ally would be ideal to > discuss. The main background reading would be the L1V3 draft spec > (http://co.mbine.org/specifications/sed-ml.level-1.version-3-draft-1) > which would also lead you to the NuML specification: > https://github.com/NuML/NuML/blob/master/numl-spec-l1v1.pdf. Since the > use of data in SED-ML is dependent on NuML, we will also likely be > discussing the state of NuML and where that project is heading. > > > Cheers, > David. > > > > On Thu, Sep 17, 2015 at 4:39 AM, Herbert Sauro <hs...@gm...> wrote: > > We'd be looking for the same things. > > > > Herbert > > > > On Wed, Sep 16, 2015 at 7:05 AM, Ally Hume <a....@ed...> wrote: > >> > >> Hi, > >> > >> I notice that there will be a session at Combine entitled "Using data > with > >> SED-ML". I was wondering if somebody could tell me a more detailed > >> description of what may be discussed here? I've looked in the archives > for > >> recent posts to this list but nothing seems to cover this topic. > >> > >> I've recently been looking at adding external time course data to SBML > >> models. In our use case this is temperature data but it could obviously > be > >> many other measurements as well. I've written code to add external data > to > >> SBML models as a parameter with an associated assignment rule that > defines > >> the data using a piecewise function [1]. This works for our purposes but > >> does not seem like the cleanest way to do things. > >> > >> SED-ML looks like a place where this could be done a lot better. > >> > >> Adopting similar strategies to those already used in SED-ML it would be > >> nice to include: > >> - references to datasets (similar to references to models) > >> - link between entities in these datasets and parameters in a model > >> - specification of an interpolation algorithm (ontology-based like > >> simulation algorithms) > >> > >> This would be a considerable improvement over my current approach of > >> writing this into SBML where the interpolation algorithm and original > >> dataset are effectively lost (although I could add annotations to > document > >> such details). > >> > >> There could also be an argument made for some basic data pre-processing > >> possibly similar to the post-processing of result data already in > SED-ML. > >> > >> I'm sure there will be alternative ideas but I just wondering if this is > >> the type of thing that will be in the session. I will be attending > Combine > >> but I don't want to wait for the session and discover it's actually on > >> something totally different, or discover there is actually a lot of > >> background documentation that I could have read in advance. > >> > >> Thank you, > >> > >> Ally > >> > >> [1] https://github.com/allyhume/SBMLDataTools > >> > >> > >> Ally Hume > >> Researcher, SynthSys, The University of Edinburgh > >> Software and Data Architect, EPCC, The University of Edinburgh > >> > >> > >> -- > >> The University of Edinburgh is a charitable body, registered in > >> Scotland, with registration number SC005336. > >> > >> > >> > >> > ------------------------------------------------------------------------------ > >> Monitor Your Dynamic Infrastructure at Any Scale With Datadog! > >> Get real-time metrics from all of your servers, apps and tools > >> in one place. > >> SourceForge users - Click here to start your Free Trial of Datadog now! > >> http://pubads.g.doubleclick.net/gampad/clk?id=241902991&iu=/4140 > >> _______________________________________________ > >> SED-ML-discuss mailing list > >> SED...@li... > >> https://lists.sourceforge.net/lists/listinfo/sed-ml-discuss > > > > > > > > > ------------------------------------------------------------------------------ > > Monitor Your Dynamic Infrastructure at Any Scale With Datadog! > > Get real-time metrics from all of your servers, apps and tools > > in one place. > > SourceForge users - Click here to start your Free Trial of Datadog now! > > http://pubads.g.doubleclick.net/gampad/clk?id=241902991&iu=/4140 > > _______________________________________________ > > SED-ML-discuss mailing list > > SED...@li... > > https://lists.sourceforge.net/lists/listinfo/sed-ml-discuss > > > > > > -- > > > David Nickerson > about.me/david.nickerson > > > ------------------------------------------------------------------------------ > Monitor Your Dynamic Infrastructure at Any Scale With Datadog! > Get real-time metrics from all of your servers, apps and tools > in one place. > SourceForge users - Click here to start your Free Trial of Datadog now! > http://pubads.g.doubleclick.net/gampad/clk?id=241902991&iu=/4140 > _______________________________________________ > SED-ML-discuss mailing list > SED...@li... > https://lists.sourceforge.net/lists/listinfo/sed-ml-discuss > |
|
From: David N. <dav...@gm...> - 2015-09-16 23:56:37
|
Hi Ally, Herb, everyone, We don't have a specific agendas set for any of the SED-ML breakout sessions at the COMBINE meeting and the current session titles reflect topics that we believe to be useful topics to at least kick off the sessions. Just over a year ago, the SED-ML editors released a draft for the next version of SED-ML which is available here: http://co.mbine.org/specifications/sed-ml.level-1.version-3-draft-1. The main change in this specification is the addition of data to SED-ML and being able to use data to control the simulation experiment, model variables, etc. The additions to the specification represented the consensus achieved during much discussion over the preceding HARMONY and COMBINE meetings (and various google docs, etc). As far as I am aware, there is currently only one implementation that supports the L1V3 draft (Frank Bergmann's .NET framework) and while some of us hoped to get our own implementations up and running at the HARMONY meeting earlier this year, a second implementation has yet to appear. And thus the L1V3 draft is still a draft :) For the COMBINE breakout on data with SED-ML, it would be good to discuss both the potential implementation of the L1V3 specification in code as well as producing some examples or use-cases (to also help address the issue of examples that Lucian has raised on this list recently). So yes - the topics you mentioned Ally would be ideal to discuss. The main background reading would be the L1V3 draft spec (http://co.mbine.org/specifications/sed-ml.level-1.version-3-draft-1) which would also lead you to the NuML specification: https://github.com/NuML/NuML/blob/master/numl-spec-l1v1.pdf. Since the use of data in SED-ML is dependent on NuML, we will also likely be discussing the state of NuML and where that project is heading. Cheers, David. On Thu, Sep 17, 2015 at 4:39 AM, Herbert Sauro <hs...@gm...> wrote: > We'd be looking for the same things. > > Herbert > > On Wed, Sep 16, 2015 at 7:05 AM, Ally Hume <a....@ed...> wrote: >> >> Hi, >> >> I notice that there will be a session at Combine entitled "Using data with >> SED-ML". I was wondering if somebody could tell me a more detailed >> description of what may be discussed here? I've looked in the archives for >> recent posts to this list but nothing seems to cover this topic. >> >> I've recently been looking at adding external time course data to SBML >> models. In our use case this is temperature data but it could obviously be >> many other measurements as well. I've written code to add external data to >> SBML models as a parameter with an associated assignment rule that defines >> the data using a piecewise function [1]. This works for our purposes but >> does not seem like the cleanest way to do things. >> >> SED-ML looks like a place where this could be done a lot better. >> >> Adopting similar strategies to those already used in SED-ML it would be >> nice to include: >> - references to datasets (similar to references to models) >> - link between entities in these datasets and parameters in a model >> - specification of an interpolation algorithm (ontology-based like >> simulation algorithms) >> >> This would be a considerable improvement over my current approach of >> writing this into SBML where the interpolation algorithm and original >> dataset are effectively lost (although I could add annotations to document >> such details). >> >> There could also be an argument made for some basic data pre-processing >> possibly similar to the post-processing of result data already in SED-ML. >> >> I'm sure there will be alternative ideas but I just wondering if this is >> the type of thing that will be in the session. I will be attending Combine >> but I don't want to wait for the session and discover it's actually on >> something totally different, or discover there is actually a lot of >> background documentation that I could have read in advance. >> >> Thank you, >> >> Ally >> >> [1] https://github.com/allyhume/SBMLDataTools >> >> >> Ally Hume >> Researcher, SynthSys, The University of Edinburgh >> Software and Data Architect, EPCC, The University of Edinburgh >> >> >> -- >> The University of Edinburgh is a charitable body, registered in >> Scotland, with registration number SC005336. >> >> >> >> ------------------------------------------------------------------------------ >> Monitor Your Dynamic Infrastructure at Any Scale With Datadog! >> Get real-time metrics from all of your servers, apps and tools >> in one place. >> SourceForge users - Click here to start your Free Trial of Datadog now! >> http://pubads.g.doubleclick.net/gampad/clk?id=241902991&iu=/4140 >> _______________________________________________ >> SED-ML-discuss mailing list >> SED...@li... >> https://lists.sourceforge.net/lists/listinfo/sed-ml-discuss > > > > ------------------------------------------------------------------------------ > Monitor Your Dynamic Infrastructure at Any Scale With Datadog! > Get real-time metrics from all of your servers, apps and tools > in one place. > SourceForge users - Click here to start your Free Trial of Datadog now! > http://pubads.g.doubleclick.net/gampad/clk?id=241902991&iu=/4140 > _______________________________________________ > SED-ML-discuss mailing list > SED...@li... > https://lists.sourceforge.net/lists/listinfo/sed-ml-discuss > -- David Nickerson about.me/david.nickerson |
|
From: Herbert S. <hs...@gm...> - 2015-09-16 16:39:30
|
We'd be looking for the same things. Herbert On Wed, Sep 16, 2015 at 7:05 AM, Ally Hume <a....@ed...> wrote: > Hi, > > I notice that there will be a session at Combine entitled "Using data with > SED-ML". I was wondering if somebody could tell me a more detailed > description of what may be discussed here? I've looked in the archives for > recent posts to this list but nothing seems to cover this topic. > > I've recently been looking at adding external time course data to SBML > models. In our use case this is temperature data but it could obviously be > many other measurements as well. I've written code to add external data to > SBML models as a parameter with an associated assignment rule that defines > the data using a piecewise function [1]. This works for our purposes but > does not seem like the cleanest way to do things. > > SED-ML looks like a place where this could be done a lot better. > > Adopting similar strategies to those already used in SED-ML it would be > nice to include: > - references to datasets (similar to references to models) > - link between entities in these datasets and parameters in a model > - specification of an interpolation algorithm (ontology-based like > simulation algorithms) > > This would be a considerable improvement over my current approach of > writing this into SBML where the interpolation algorithm and original > dataset are effectively lost (although I could add annotations to document > such details). > > There could also be an argument made for some basic data pre-processing > possibly similar to the post-processing of result data already in SED-ML. > > I'm sure there will be alternative ideas but I just wondering if this is > the type of thing that will be in the session. I will be attending Combine > but I don't want to wait for the session and discover it's actually on > something totally different, or discover there is actually a lot of > background documentation that I could have read in advance. > > Thank you, > > Ally > > [1] https://github.com/allyhume/SBMLDataTools > > > Ally Hume > Researcher, SynthSys, The University of Edinburgh > Software and Data Architect, EPCC, The University of Edinburgh > > > -- > The University of Edinburgh is a charitable body, registered in > Scotland, with registration number SC005336. > > > > ------------------------------------------------------------------------------ > Monitor Your Dynamic Infrastructure at Any Scale With Datadog! > Get real-time metrics from all of your servers, apps and tools > in one place. > SourceForge users - Click here to start your Free Trial of Datadog now! > http://pubads.g.doubleclick.net/gampad/clk?id=241902991&iu=/4140 > _______________________________________________ > SED-ML-discuss mailing list > SED...@li... > https://lists.sourceforge.net/lists/listinfo/sed-ml-discuss > |
|
From: Ally H. <a....@ed...> - 2015-09-16 14:05:46
|
Hi, I notice that there will be a session at Combine entitled "Using data with SED-ML". I was wondering if somebody could tell me a more detailed description of what may be discussed here? I've looked in the archives for recent posts to this list but nothing seems to cover this topic. I've recently been looking at adding external time course data to SBML models. In our use case this is temperature data but it could obviously be many other measurements as well. I've written code to add external data to SBML models as a parameter with an associated assignment rule that defines the data using a piecewise function [1]. This works for our purposes but does not seem like the cleanest way to do things. SED-ML looks like a place where this could be done a lot better. Adopting similar strategies to those already used in SED-ML it would be nice to include: - references to datasets (similar to references to models) - link between entities in these datasets and parameters in a model - specification of an interpolation algorithm (ontology-based like simulation algorithms) This would be a considerable improvement over my current approach of writing this into SBML where the interpolation algorithm and original dataset are effectively lost (although I could add annotations to document such details). There could also be an argument made for some basic data pre-processing possibly similar to the post-processing of result data already in SED-ML. I'm sure there will be alternative ideas but I just wondering if this is the type of thing that will be in the session. I will be attending Combine but I don't want to wait for the session and discover it's actually on something totally different, or discover there is actually a lot of background documentation that I could have read in advance. Thank you, Ally [1] https://github.com/allyhume/SBMLDataTools Ally Hume Researcher, SynthSys, The University of Edinburgh Software and Data Architect, EPCC, The University of Edinburgh -- The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336. |
|
From: Alan G. <ala...@in...> - 2015-09-11 21:39:39
|
> From: Anna Zhukova [mailto:ann...@in...] > On 09/09/2015 22:00, Alan Garny wrote: > >> From: Pedro Mendes [mailto:ped...@ma...] > >> On 09/03/2015 12:56 PM, Nicolas Le Novere wrote: > >> >> On 03/09/15 12:41, Alan Garny wrote: > >>>> From the SED-ML L1V2 specifications, it would seem that a > >>>> simulation > >>>>> can have only one algorithm (section 2.4.3). However, what if I > >>>>> need to specify two algorithms? Indeed, some DAE models may > >>>>> require solving a system of non-linear algebraic equations, in > >>>>> which case we might want to be able to use two solvers such as IDA > >>>>> <http://computation.llnl.gov/casc/sundials/description/description > >>>>> .h > >>>>> tml#descr_ida> > >>>>> > >>> and KINSOL > >>>>> <http://computation.llnl.gov/casc/sundials/description/description > >>>>> .h > >>>>> tml#descr_kinsol> > >>>>> > >>> from the SUNDIALS library. I guess that for now, I could specify IDA > >>>>> as the simulation algorithm and refer to KINSOL in an annotation > >>>>> in the SED-ML simulation object? > >>> If solving a systems of equation requires several algorithms working > >>> together, you would have to specify how they work together. Listing > >>> the algorithms is not sufficient. If the details of the hybrid > >>> method is described somewhere, I would advocate the creation of a > >>> new KiSAO term, that would identify this specific combination of > >>> solvers. We already have such terms in the ontology. > >> I would even say that such cases where two algorithms are used, are > >> themselves a new algorithm and need to be described as such. For > >> example, LSODA uses several algorithms to multiply matrices with > >> vectors, and to factorize matrices (and so does CVODE), but we do not > >> describe them as being combinations of those other algorithms. Thus > >> there should be a specific entry to refer to the actual algorithm > >> (which no doubt will use > > other > >> algorithms itself). > >> > >> It should not be a problem to add more KiSAO entries to new > >> algorithms or even variations of existing ones. This way we keep an > >> ontology of what > > they > >> are. (the only problem would be lack of humanpower to keep KiSAO > >> updated). > > FWIW, regarding CVODE (and IDA, to some extent), I have recently asked > > for some new KiSAO ids to be added (see > > http://sourceforge.net/p/kisao/feature-requests/12/). I have also > > asked for one for "step size" (see > > http://sourceforge.net/p/kisao/feature-requests/13/). > Thanks for submitting terms to KiSAO. They have now been added. If more > terms are needed, please, do not hesitate to submit them to > https://sourceforge.net/p/kisao/feature-requests/. Thanks a lot Anna, that's very useful. Alan |
|
From: Anna Z. <ann...@in...> - 2015-09-10 15:03:15
|
On 09/09/2015 22:00, Alan Garny wrote: >> From: Pedro Mendes [mailto:ped...@ma...] >> On 09/03/2015 12:56 PM, Nicolas Le Novere wrote: >> >> On 03/09/15 12:41, Alan Garny wrote: >>>> From the SED-ML L1V2 specifications, it would seem that a simulation >>>>> can have only one algorithm (section 2.4.3). However, what if I need >>>>> to specify two algorithms? Indeed, some DAE models may require >>>>> solving a system of non-linear algebraic equations, in which case we >>>>> might want to be able to use two solvers such as IDA >>>>> <http://computation.llnl.gov/casc/sundials/description/description.h >>>>> tml#descr_ida> >>>>> >>> and KINSOL >>>>> <http://computation.llnl.gov/casc/sundials/description/description.h >>>>> tml#descr_kinsol> >>>>> >>> from the SUNDIALS library. I guess that for now, I could specify IDA >>>>> as the simulation algorithm and refer to KINSOL in an annotation in >>>>> the SED-ML simulation object? >>> If solving a systems of equation requires several algorithms working >>> together, you would have to specify how they work together. Listing >>> the algorithms is not sufficient. If the details of the hybrid method >>> is described somewhere, I would advocate the creation of a new KiSAO >>> term, that would identify this specific combination of solvers. We >>> already have such terms in the ontology. >> I would even say that such cases where two algorithms are used, are >> themselves a new algorithm and need to be described as such. For example, >> LSODA uses several algorithms to multiply matrices with vectors, and to >> factorize matrices (and so does CVODE), but we do not describe them as >> being combinations of those other algorithms. Thus there should be a >> specific entry to refer to the actual algorithm (which no doubt will use > other >> algorithms itself). >> >> It should not be a problem to add more KiSAO entries to new algorithms or >> even variations of existing ones. This way we keep an ontology of what > they >> are. (the only problem would be lack of humanpower to keep KiSAO >> updated). > FWIW, regarding CVODE (and IDA, to some extent), I have recently asked for > some new KiSAO ids to be added (see > http://sourceforge.net/p/kisao/feature-requests/12/). I have also asked for > one for "step size" (see > http://sourceforge.net/p/kisao/feature-requests/13/). Thanks for submitting terms to KiSAO. They have now been added. If more terms are needed, please, do not hesitate to submit them to https://sourceforge.net/p/kisao/feature-requests/. Cheers, Anna |
|
From: Alan G. <ala...@in...> - 2015-09-09 20:00:15
|
> From: Pedro Mendes [mailto:ped...@ma...] > On 09/03/2015 12:56 PM, Nicolas Le Novere wrote: > >> On 03/09/15 12:41, Alan Garny wrote: > >> From the SED-ML L1V2 specifications, it would seem that a simulation > >>> can have only one algorithm (section 2.4.3). However, what if I need > >>> to specify two algorithms? Indeed, some DAE models may require > >>> solving a system of non-linear algebraic equations, in which case we > >>> might want to be able to use two solvers such as IDA > >>> <http://computation.llnl.gov/casc/sundials/description/description.h > >>> tml#descr_ida> > >>> > > and KINSOL > >>> <http://computation.llnl.gov/casc/sundials/description/description.h > >>> tml#descr_kinsol> > >>> > > from the SUNDIALS library. I guess that for now, I could specify IDA > >>> as the simulation algorithm and refer to KINSOL in an annotation in > >>> the SED-ML simulation object? > > > If solving a systems of equation requires several algorithms working > > together, you would have to specify how they work together. Listing > > the algorithms is not sufficient. If the details of the hybrid method > > is described somewhere, I would advocate the creation of a new KiSAO > > term, that would identify this specific combination of solvers. We > > already have such terms in the ontology. > > I would even say that such cases where two algorithms are used, are > themselves a new algorithm and need to be described as such. For example, > LSODA uses several algorithms to multiply matrices with vectors, and to > factorize matrices (and so does CVODE), but we do not describe them as > being combinations of those other algorithms. Thus there should be a > specific entry to refer to the actual algorithm (which no doubt will use other > algorithms itself). > > It should not be a problem to add more KiSAO entries to new algorithms or > even variations of existing ones. This way we keep an ontology of what they > are. (the only problem would be lack of humanpower to keep KiSAO > updated). FWIW, regarding CVODE (and IDA, to some extent), I have recently asked for some new KiSAO ids to be added (see http://sourceforge.net/p/kisao/feature-requests/12/). I have also asked for one for "step size" (see http://sourceforge.net/p/kisao/feature-requests/13/). Alan |
|
From: Pedro M. <ped...@ma...> - 2015-09-09 10:05:12
|
On 09/03/2015 12:56 PM, Nicolas Le Novere wrote: >> On 03/09/15 12:41, Alan Garny wrote: >> From the SED-ML L1V2 specifications, it would seem that a >> simulation >>> can have only one algorithm (section 2.4.3). However, what if I >>> need to specify two algorithms? Indeed, some DAE models may >>> require solving a system of non-linear algebraic equations, in >>> which case we might want to be able to use two solvers such as >>> IDA >>> <http://computation.llnl.gov/casc/sundials/description/description.html#descr_ida> >>> > and KINSOL >>> <http://computation.llnl.gov/casc/sundials/description/description.html#descr_kinsol> >>> > from the SUNDIALS library. I guess that for now, I could specify IDA >>> as the simulation algorithm and refer to KINSOL in an annotation >>> in the SED-ML simulation object? > If solving a systems of equation requires several algorithms working > together, you would have to specify how they work together. Listing > the algorithms is not sufficient. If the details of the hybrid method > is described somewhere, I would advocate the creation of a new KiSAO > term, that would identify this specific combination of solvers. We > already have such terms in the ontology. I would even say that such cases where two algorithms are used, are themselves a new algorithm and need to be described as such. For example, LSODA uses several algorithms to multiply matrices with vectors, and to factorize matrices (and so does CVODE), but we do not describe them as being combinations of those other algorithms. Thus there should be a specific entry to refer to the actual algorithm (which no doubt will use other algorithms itself). It should not be a problem to add more KiSAO entries to new algorithms or even variations of existing ones. This way we keep an ontology of what they are. (the only problem would be lack of humanpower to keep KiSAO updated). Pedro -- Pedro Mendes Professor of Computational Systems Biology http://www.comp-sys-bio.org School of Computer Science Manchester Centre for Integrative Systems Biology University of Manchester Manchester Institute of Biotechnology 131 Princess Street Manchester, M1 7DN, U.K. |
|
From: Chris J. M. <my...@ec...> - 2015-09-09 03:14:37
|
Hi, I’m pleased to announce the agenda for COMBINE 2015 is now available here: http://co.mbine.org/events/COMBINE_2015/agenda <http://co.mbine.org/events/COMBINE_2015/agenda> The list of attendees is also up-to-date, so please verify that your name is listed. If it is not and you think you have registered, please contact me. If you have not registered, please do so as soon as you can, and no later than October 1st. http://co.mbine.org/events/COMBINE_2015/attendees <http://co.mbine.org/events/COMBINE_2015/attendees> Finally, and most IMPORTANTLY, the room block will expire on Monday September 14th. They will neither guarantee rooms or the reduced conference price after that date. So, if you have not reserved your room yet, please do so. Look forward to welcoming many of you to Salt Lake City next month. Chris |
|
From: David N. <dav...@gm...> - 2015-09-03 20:10:55
|
---------- Forwarded message ---------- From: Michael Hucka <mh...@ca...> Date: Fri, Sep 4, 2015 at 6:20 AM Subject: [sysbio] Special issue of JIB for standards in systems biology To: sy...@ca... For a long time, since the demise of Nature Precedings, we did not have a good venue to publish the specification documents for SBML and other community standards. Thanks to the efforts of the Journal for Integrative Bioinformatics, and particularly Falk Schreiber, we now do. A special issue of the Journal of Integrative Bioinformatics devoted to community standards has just been published: http://journal.imbio.de/archive.php?pindex=12&iid=27 The purpose is to provide a citable reference for the standards specifications, as well as promote them and make them more visible. The special issue is anticipated to be an annual event. JIB is open access, and indexed/tracked/covered by PubMed, DBLP Computer Science Bibliography, and Google Scholar. A big thanks to the JIB team for their efforts! Mike -- David Nickerson about.me/david.nickerson |
|
From: Felix W. <wi...@ka...> - 2015-09-03 13:59:54
|
Hi Alan, > > So, am I to understand that the step size to use with the Euler forward > method should be 0.05 in this case (i.e. 200/4000)? That's the only thing I > can think of, but I hope to be wrong since depending on the model we want to > integrate, 0.05 might be too big for the Euler forward method. In other > words, although we might want results at 0.05 intervals, we might need to > integrate using 0.01 for example. So, how could I do that? Should I use the > KiSAO id for maximum step size and therefore have something like: > As far as I understand the spec the parameters in the uniformTimeCourse element are not to be interpreted as the step size, but only as the output interval. I don't think there is currently a way to set the step size in SED-ML. > > <uniformTimeCourse id="simulation" initialTime="0" > outputStartTime="0" outputEndTime="200" numberOfPoints="4000"> > > <algorithm kisaoID="KISAO:0000030"> > > <listOfAlgorithmParameters> > > <algorithmParameter kisaoID="KISAO:0000467" > value="0.01"/> > > </listOfAlgorithmParameters> > > </algorithm> > > </uniformTimeCourse> > > > > This doesn't seem right to me (one-step methods don't have a "maximum step > size", only a "step size"), but who knows. > For a uniformTimeCourse simulation one would usually use an integration algorithm that automatically adapts its step size, but if you really want to use the Euler method, maybe your solution is the way to go. Best regards, Felix Felix Winter Rostock, Germany http://orcid.org/0000-0003-2987-6797 |
|
From: Alan G. <ala...@in...> - 2015-09-03 13:52:32
|
Hi Felix,
> > So, am I to understand that the step size to use with the Euler
> > forward method should be 0.05 in this case (i.e. 200/4000)? That's the
> > only thing I can think of, but I hope to be wrong since depending on
> > the model we want to integrate, 0.05 might be too big for the Euler
> > forward method. In other words, although we might want results at 0.05
> > intervals, we might need to integrate using 0.01 for example. So, how
> > could I do that? Should I use the KiSAO id for maximum step size and
> > therefore have something like:
> As far as I understand the spec the parameters in the uniformTimeCourse
> element are not to be interpreted as the step size, but only as the output
> interval.
Ok, that's how I understand it too (and how I have been using it so far).
> I don't think there is currently a way to set the step size in SED-ML.
Strange if that's really the case.
> > <uniformTimeCourse id="simulation" initialTime="0"
> > outputStartTime="0" outputEndTime="200" numberOfPoints="4000">
> >
> > <algorithm kisaoID="KISAO:0000030">
> >
> > <listOfAlgorithmParameters>
> >
> > <algorithmParameter kisaoID="KISAO:0000467"
> > value="0.01"/>
> >
> > </listOfAlgorithmParameters>
> >
> > </algorithm>
> >
> > </uniformTimeCourse>
> >
> >
> >
> > This doesn't seem right to me (one-step methods don't have a "maximum
> > step size", only a "step size"), but who knows.
> For a uniformTimeCourse simulation one would usually use an integration
> algorithm that automatically adapts its step size, but if you really want to
> use the Euler method, maybe your solution is the way to go.
The sole fact that there is a KiSAO id for the Euler forward method tells me that there should also be a KiSAO term for "step size". Ok, I am going to make an 'official' request for it on the KiSAO website ("maximum step size" really isn't suitable in my view).
Alan
|
|
From: Alan G. <ala...@in...> - 2015-09-03 13:12:52
|
Hi, Another question, if you don't mind. I want to specify the step size to use for a one-step method such as the Euler forward method <http://bioportal.bioontology.org/ontologies/KISAO/?p=classes&conceptid=http %3A%2F%2Fwww.biomodels.net%2Fkisao%2FKISAO%23KISAO_0000030> . However, there doesn't seem to be a KiSAO id for "step size" (while there is one for <http://bioportal.bioontology.org/ontologies/KISAO/?p=classes&conceptid=http %3A%2F%2Fwww.biomodels.net%2Fkisao%2FKISAO%23KISAO_0000467&jump_to_nav=true> "maximum step size", for example). Considering that it's such a basic algorithm parameter, I imagine there must be a reason for it and that I am missing it? In fact, googling for "KISAO:0000030" (i.e. the KiSAO id for the Euler forward method), I found things like: <uniformTimeCourse id="simulation" initialTime="0" outputStartTime="0" outputEndTime="200" numberOfPoints="4000"> <algorithm kisaoID="KISAO:0000030"/> </uniformTimeCourse> So, am I to understand that the step size to use with the Euler forward method should be 0.05 in this case (i.e. 200/4000)? That's the only thing I can think of, but I hope to be wrong since depending on the model we want to integrate, 0.05 might be too big for the Euler forward method. In other words, although we might want results at 0.05 intervals, we might need to integrate using 0.01 for example. So, how could I do that? Should I use the KiSAO id for maximum step size and therefore have something like: <uniformTimeCourse id="simulation" initialTime="0" outputStartTime="0" outputEndTime="200" numberOfPoints="4000"> <algorithm kisaoID="KISAO:0000030"> <listOfAlgorithmParameters> <algorithmParameter kisaoID="KISAO:0000467" value="0.01"/> </listOfAlgorithmParameters> </algorithm> </uniformTimeCourse> This doesn't seem right to me (one-step methods don't have a "maximum step size", only a "step size"), but who knows. Cheers, Alan. |
|
From: Alan G. <ala...@in...> - 2015-09-03 12:15:43
|
> From: Nicolas Le Novere [mailto:n.l...@gm...] > On 03/09/15 12:41, Alan Garny wrote: > > From the SED-ML L1V2 specifications, it would seem that a simulation > > can have only one algorithm (section 2.4.3). However, what if I need > > to specify two algorithms? Indeed, some DAE models may require solving > > a system of non-linear algebraic equations, in which case we might > > want to be able to use two solvers such as IDA > > <http://computation.llnl.gov/casc/sundials/description/description.htm > > l#descr_ida> > > and KINSOL > > <http://computation.llnl.gov/casc/sundials/description/description.htm > > l#descr_kinsol> from the SUNDIALS library. I guess that for now, I > > could specify IDA as the simulation algorithm and refer to KINSOL in > > an annotation in the SED-ML simulation object? > If solving a systems of equation requires several algorithms working > together, you would have to specify how they work together. Listing the > algorithms is not sufficient. If the details of the hybrid method is described > somewhere, I would advocate the creation of a new KiSAO term, that would > identify this specific combination of solvers. We already have such terms in > the ontology. There is no such description, so I guess I might have to stick to the annotation for now. > > Another question I have is about algorithm parameters that don't have > > a corresponding KiSAO id. For example, for CVODE > > <http://computation.llnl.gov/casc/sundials/description/description.htm > > l#descr_cvode>, I have seen KiSAO ids for CVODE itself, as well as for > > the maximum step size, the maximum number of steps, relative tolerance > > and absolute tolerance, but not for the type of integration method > > (i.e. > > BDF or Adams-Moulton) and a few other parameters that can be set in > > CVODE. I understand that it's more of a KiSAO question, but until > > those KiSAO ids become available, what would be the best way to 'set' > > those parameters in a SED-ML file? Once again, I am thinking of > > referring to them in various annotations in the SED-ML algorithm > > object, but is that the recommended approach? > > > > I am not happy with that approach since it will effectively render my > > SED-ML file unusable by others, but I can't think of any other way at > > this stage. > > Maybe the solution is to extend the corresponding KiSAO term? > > http://sourceforge.net/p/kisao/feature-requests/new/ Ok, I am going to ask for them to be added. Thanks, Alan |
|
From: Nicolas Le N. <n.l...@gm...> - 2015-09-03 11:57:07
|
On 03/09/15 12:41, Alan Garny wrote: > From the SED-ML L1V2 specifications, it would seem that a simulation > can have only one algorithm (section 2.4.3). However, what if I need > to specify two algorithms? Indeed, some DAE models may require > solving a system of non-linear algebraic equations, in which case we > might want to be able to use two solvers such as IDA > <http://computation.llnl.gov/casc/sundials/description/description.html#descr_ida> > and KINSOL > <http://computation.llnl.gov/casc/sundials/description/description.html#descr_kinsol> > from the SUNDIALS library. I guess that for now, I could specify IDA > as the simulation algorithm and refer to KINSOL in an annotation in > the SED-ML simulation object? If solving a systems of equation requires several algorithms working together, you would have to specify how they work together. Listing the algorithms is not sufficient. If the details of the hybrid method is described somewhere, I would advocate the creation of a new KiSAO term, that would identify this specific combination of solvers. We already have such terms in the ontology. > Another question I have is about algorithm parameters that don’t have > a corresponding KiSAO id. For example, for CVODE > <http://computation.llnl.gov/casc/sundials/description/description.html#descr_cvode>, > I have seen KiSAO ids for CVODE itself, as well as for the maximum > step size, the maximum number of steps, relative tolerance and > absolute tolerance, but not for the type of integration method (i.e. > BDF or Adams-Moulton) and a few other parameters that can be set in > CVODE. I understand that it’s more of a KiSAO question, but until > those KiSAO ids become available, what would be the best way to ‘set’ > those parameters in a SED-ML file? Once again, I am thinking of > referring to them in various annotations in the SED-ML algorithm > object, but is that the recommended approach? > > I am not happy with that approach since it will effectively render my > SED-ML file unusable by others, but I can’t think of any other way at > this stage… Maybe the solution is to extend the corresponding KiSAO term? http://sourceforge.net/p/kisao/feature-requests/new/ -- Nicolas LE NOVERE, Babraham Institute, Babraham Campus Cambridge, CB22 3AT Tel: +441223496433, Mob:+447833147074, twitter:@lenovere, Skype:n.lenovere n.l...@gm..., http://orcid.org/0000-0002-6309-7327 http://lenoverelab.org/, http://lenoverelab.org/perso/lenov/ |
14 messages has been excluded from this view by a project administrator.
