Re: [SOSlib-discuss] [SOSlib-devel] Steady State detection in sbmlsolver

[Rainer M. <ra...@tb...>]

Hi Thorsten,

Thorsten Lenser wrote:
 > Hello everyone,
 >
 > I am using the odeSolver in an evolutionary algorithm that evolves SBML
 > models. Many of the tasks that I want my networks to perform rely on 
some
 > forms of switching between steady-states. Unfortunately, steady state
 > detection is quite provisional in the odeSolver, especially since one 
cannot
 > specify the threshold at which a steady state is reached. Are there 
any plans
 > to improve this for the next release?


It would be easy to provide an optional threshold for steady state 
detection! Thanks for the suggestion, I will definitely include this 
option and will let you know as soon as it is in the CVS. (BTW, I would 
currently suggest using the CVS version. It is quite stable and has a 
lot of improvements and bug fixes and might even be a bit faster).

We keep looking for a general and global solver that could calculate all 
steady states but have no concrete plans, currently.

There exists some test code interfacing SUNDIALS KINSOL, which could 
detect local steady states, close to the initial conditions.
It actually works but not very well. If you are interested in that, you 
could check out latest version from sourceforge CVS and take a look at 
the file findRoot.c in the examples folder. You could manually include 
this example program in the Makefile.am (before running configure) or 
compile it by hand. Only use it with SBML level 2 (otherwise it 
segfaults as it doesn't do the internal conversion).

It uses functions defined in the rootFinder.c file in the src directory.

Maybe a better usage of KINSOL could provide a better functionality to 
detect steady states without integrating or during integrating. The 
provisional way of detecting steady states during integration might 
however still be faster and really find the steady state of an 
integration (and not a close steady state that can't be reached with the 
set initial conditions). But I have never gone into the details of KINSOL.

 > Apart from that, thanks for providing this great tool!

Very welcome. Thanks for using it and for the compliments :-)

Rainer