Forums

I looked for a draft, and I found
http://www.cds.caltech.edu/~afinney/Cox-MDL-for-stochastic-models.pdf .
As I read this, there would only be two additional changes to the spec
needed for stochastic simulation: the discrete attribute for species and
two kinetic laws allowed for reversible reactions. (Please correct me if
I'm wrong.) Additional processing would be up to the simulation
algorithm, and is very simple to do. The changes to the rate constants
are easily computable.

For people uninterested in stochastic simulation, they could just report
an error that the SBML file is for stochastic simulation if they see the
discrete attribute.

I know that these changes can be done via annotation, but I view
annotation as useful for people with specific needs for their software,
and as you mention, this comes up relatively often.

Bill

On Fri, 10 Mar 2006, Herbert Sauro wrote:

> Seems to make sense to me. I would agree that there is a real problem
> with representing stoachstic models in the current SBML, we are
> confronted with this problem almost every week. I know that Mike and
> others wrote a draft for something along these lines, but I suspect
> nothing happened afterwards beause it looked a bit too complicated for
> the rest of us to implement.
>
> Herbert
>
> -----Original Message-----
> From: Stefan Hoops [mailto:shoops@vbi.vt.edu]
> Sent: Friday, March 10, 2006 10:28 AM
> To: sbml-discuss@caltech.edu
> Subject: Re: [sbml-discuss] Multiple kineticLaw Sections Per Reaction
>
> Hello All
>
> I think what we see here is again the problem that SBML originally was
> only developed fro ODE based modelling. We all know that we have moved
> far beyond that however one crucial component of SBML has not kept
> track.
>
> This component in my opinion is the way we specify the kinetic laws. A
> kinetic law for a stochastic model is different from the kinetic law for
> an ODE model. This discussion makes that clear again.
>
> I suggest that we start a discussion on how to deal with this issue.
>
> My propose to create a new object "kinetic law". The real contend will
> be in subclasses derived from this. Two subclasses come immediately into
> my mind these are "ODE kinetic law" and "stochastic kinetic law". The
> ODE kinetic law is what we currently understand under kinetic law. The
> stochastic kinetic law could have a forward and a backward probability.
> Additional subclasses can be envisioned and added.
>
> A reaction may have only one entity of each subclass of the kinetic law.
> The reason for this restriction is that each tool now easily can decide
> which law to use dependent on the type of simulation/analysis intended.
> In addition allowing multiple different subclasses of "kinetic law"
> would allow us to store a stochastic and an "equivalent" ODE model in
> the same file.
>
> What are your thoughts on this?
>
>
> Thanks,
> Stefan
>
> --
> Stefan Hoops, Ph.D.
> Senior Project Associate
> Virginia Bioinformatics Institute - 0477 Virginia Tech Bioinformatics
> Facility I Blacksburg, Va 24061, USA
>
> Phone: (540) 231-1799
> Fax: (540) 231-2606
> Email: shoops@vbi.vt.edu
>

--
"It is horrifying that we have to fight our own government to save the
environment."
-- Ansel Adams