RE: What do SBML libraries need to focus on next?
21 Jul '05 17:52
Hyss certainly looks different from the normal run of the mill
simulators that abound.
To be honest I wish I understood the theory behind stochastic simulation
better, there are a number of things which nag me and I've asked around
but no one has given me a good answer. Three things in particular:
1. What are the rules regarding the use of different rate laws in
2. I've seen a few hybrid schemes for continuous/discrete simulators but
I've seen little theory to back them up. In particular the question of
how to mesh a variable step size continuous solver with a variable step
size stochastic solver does not appear to have been satisfactorily
answered (or has it?).
3. As we all appreciate here, systems biology is just simply building a
model and hitting the run button, it is also understanding the model
(amongst other things) and for that we have a great variety of tools and
particularly theory to help us. My impression is that stochastic models
(including hybrid ones, and deterministic hybrid) do not have the same
degree of assistance, therefore what can one do to dissect a stochastic
model in the same manner we do with a continuous model?
From: Howard Salis [mailto:firstname.lastname@example.org]
Sent: Thursday, July 21, 2005 3:04 PM
To: SBML Discussion List
Subject: Re: [sbml-discuss] What do SBML libraries need to focus on
Well, since you're asking, check out http://hysss.sourceforge.net. It's
GPL'd, of course.
Though, it's not complete and we'll soon be adding yet another (very
innovative) hybrid method which utilizes a new type of probabilistic
steady state approximation.
The (slightly crude, but very capable) GUI is written in Matlab. I
obviously don't spend a lot of time on the GUI.
Herbert Sauro wrote:
>This remark is to the whole community, how many ode/gillespie wrappers
>do we really need? Are there now almost 100 listed on the sbml site?
>Instead of writing yet another sbml/ode wrapper isn't it time we
>actually did something truly innovative?
>From: Michael Hucka [mailto:email@example.com]
>Sent: Thursday, July 21, 2005 1:20 PM
>To: SBML Discussion List
>Cc: LibSBML Discussion List
>Subject: Re: [sbml-discuss] What do SBML libraries need to focus on
>>>>>>On 21 Jul 2005, Stefan Hoops <firstname.lastname@example.org> wrote:
> shoops> What I definitely exclude are the following points:
> shoops> - Computing higher-level information about a model.
> shoops> - Solvers (mentioned by Howard Salis)
> shoops> This can and should be provided in additional shoops>
> libraries or tools by whoever wants to shoops> contribute.
>Agreed. There is currently no plan to roll solvers into libSBML. If
>we (SBML Team) do ever work on solution engines, I forsee them being
>separate packages or libraries; it makes more sense that way.