sbml-qual Mailing List for Systems Biology Markup Language (SBML)

Hi Claudine and Sarah,

Sorry that I missed your presentation at COMBINE.  I think the future directions that you describe below are important ones, and I have some comments on them.
> 
> It became clear whilst working on the recent qual paper that having a common description of the model was not sufficient to guarantee identical simulation results. It is also necessary to precisely establish such things as update mechanism, initial condition(s), input nodes behaviours or even perturbations (LOF, GOF, or subtler ones).
> Thus the ability to produce a SED-ML file to describe what should be done with a qual model would be advantageous in standardising our use of logical models.
> 
Some of this may be good for SED-ML such as update synchronous or asynchronous update mechanism, but I think some things will require a tighter definition of the semantics is some of the places that I highlighted in early comments.
> 
> 1. identify aspects of logical models that are possibly not yet expressed in qual and thus highlight the way forward for the specification
> 
There are two issues with qual which limit my ability to use it.

1) The inability for qual and core to easily interact. 

I would like to be able to use qual to represent the logical portion of the model, but have it capable of interacting with the quantitive portion expressed using core.  SBML L3V2 will likely remove some of the barriers to this in that it will allow qualitative species to appear in math expressions as well as on the left-hand side of initial assignments, event assignments, rules, etc.  As for appearing in math, the interpretation is pretty straightforward as it is just the value of the species current level.  Assignments are trickier as what should be done when an expression assigns a value outside the allowed levels?

One special case where qual could be very helpful are Boolean variables (i.e., one with only two levels).  It would be nice to special case these so they can appear for example as operands to logical operators.  I'm currently faking Boolean variables using parameters, but it requires awkward conversions (added automatically) such as X==1 to convert the variable to a Boolean and piecewise to convert an assignment from a Boolean expression into a real result.  If qual could add real Boolean variables to SBML modeling, I believe this would be a nice benefit.  

2) The inability to express quantitative values such as delay or probability of a transition.

Right now I model this by using events rather than transitions.  I'm not sure the best way to deal with this as we would likely not want to simply extend transitions to include all the components of events which would be a bit redundant.  Maybe the best solution is once qualitative species can occur in event assignments then events can be used in the place of transitions when quantitative information is needed.  This is definitely though something worth experimenting with and discussing further.

> 2. establish a common means of outputting results to allow easy comparison
> 
I think this can be accomplished by simply outputting traces in CSV format where time is simply given as an integer representing the simulation step rather than absolute time.  

> 3. establish what information a simulation needs to ensure the results of a simulation are identical.
> 
This is a challenging problem since even if one agrees to synchronous or asynchronous simulation as the update scheme, qual allows for non-determinism.  Indeed, this is a critical and important aspect of qual.  Therefore, for testing one likely needs to store a state graph in which the nodes are the states with the levels of the QS and the transitions are the possible next states.  To do comparison then to a simulation, one would need to walk the graph starting in the initial state to show that the simulation produced is one that is allowed by the state graph.

> The last two points would involve creating SED-ML files and thus establish how SED-ML needs to be adjusted to accommodate description of logical model simulation. For now, it seems that most requirements may be covered by specifying relevant terms in KISAO. In the coming weeks, we will make a proposal, to be discussed.
> 
You are correct that at a minimum a KISAO term is need to express what type of update to use (synchronous or asynchronous).  I don't think there are other things that you want to express in SED-ML.  One could perhaps indicate how to resolve non-determinisim but I think that is a pretty odd thing to put into SED-ML.  It would be like telling an SSA simulator what random numbers to draw.  It would make it reproducible but it is very artificial.  Better to work out a scheme of comparison like the state graph one that I described above.

Cheers,

Chris

Dear all,

It became clear whilst working on the recent qual paper that having a common description of the model was not sufficient to guarantee identical simulation results. It is also necessary to precisely establish such things as update mechanism, initial condition(s), input nodes behaviours or even perturbations (LOF, GOF, or subtler ones).
Thus the ability to produce a SED-ML file to describe what should be done with a qual model would be advantageous in standardising our use of logical models.

There were a couple of sessions at COMBINE regarding the qual package and how logical modelling may interact with SED-ML.

No real conclusions were drawn, other than the fact that the issue of simulation description (SED-ML) needs to be addressed. Nobody present had any major requests for extension of the features currently covered by Version 1 of the qual specification.

It was proposed that a way forward would be to start to develop a test-suite of toy models with expected simulation results to allow compliance of different software to be tested. This test-suite would also need to include the necessary SED-ML files for each case. Since some people already have test models this Defining elementary models
for this test-suite and specifying simulation parameters would enable the following:

1. identify aspects of logical models that are possibly not yet expressed in qual and thus highlight the way forward for the specification

2. establish a common means of outputting results to allow easy comparison

3. establish what information a simulation needs to ensure the results of a simulation are identical.

The last two points would involve creating SED-ML files and thus establish how SED-ML needs to be adjusted to accommodate description of logical model simulation. For now, it seems that most requirements may be covered by specifying relevant terms in KISAO. In the coming weeks, we will make a proposal, to be discussed.

Best
Claudine & Sarah

----------------------------------------------
Claudine Chaouiya
Instituto Gulbenkian de Ciência - IGC
Rua da Quinta Grande, 6
P-2780-156 Oeiras - PORTUGAL

http://compbio.igc.gulbenkian.pt/nmd/
Phone:	(351) 214 46 45 01 
Email: cha...@ig...

Grr.  Google Plus's interface for groups is more frustrating than I realized.  So instead of sending me your profiles, can you do this?  Add COMBINE to your circles, and I'll use that to add you back to COMBINE.  Here's the COMBINE page:

https://plus.google.com/u/0/b/115309720146468337003/115309720146468337003

MH

For the SBML qual session later today, scheduled approx 16:00 Paris time, I will try to run a Google Hangout.  To connect people, I need to have you in a circle or add you explicitly.  Can you all please send me your Google Plus profile URL?  It will look something like this:

  https://plus.google.com/113119278868160414201

Thanks,
MH

Hi members of the qual mailing list

http://co.mbine.org/events/COMBINE_2013

During the COMBINE meeting there is a breakout session on the
Qualitative Modeling Package scheduled for Tues Sept 17 in the 16.00 -
17.30 (CEST) timeslot.

If you are attending COMBINE and want to be in this session but it
conflicts with another session you wish to attend can you let us know.
We cannot guarantee to change the session but it may be possible if it
affects too many people.

Also, if you are not attending COMBINE but would like to attend this
breakout session remotely then can you also let us know. Note remote
connection for breakouts is likely to be via Skype on someone's laptop;
so we cannot guarantee amazing quality but we can try and connect people
if they would like to attend remotely.

Thanks

Claudine & Sarah