Forums

Darren, I don't think I fully agree with you here. As I am sure you are
aware there is a very well documented theory on how to deal with
consevation laws in reaction networks, algorithms have been published
etc, and all the main tools already do the analysis.

The first point is, if you are going to put the conservation laws
explicitly in the SBML then you will need more information that a simple
algebraic law. In particular you must indicate a parameter that
represents the total mass of each conservation law because this
parameter is as important as a Km, Vmax etc to the dynamics of the
network, there is a lot of literature on this. As far as I know all the
tools that do conservation analysis internally generate this parameter
and make it available to the user. Secondly given that the theory for
consevation analysis is so well established, we should not use algraic
laws to represent them, as I indicated in an earlier email, it is always
possible to express the conservation laws as simple linear assignment
equations (Sd = C + L0*Si), there isn't any need to represent them as an
algebriac law, this is complete overkill. There are much more important
applications of the algebraic laws that conservation analysis.

If there is demand to store conservation laws in the sbml that I would
recommend a special rule for them (Nick Allen already has something
along these lines). As regards the argument that with out the
conservations laws the system is overdetermined, if you add the
consevation laws to the sbml, then the system is even more
overdetermiend since you're just repeating information that is already
in the raction network.

>I think there is some general confusion about the use of algebraic
rules that needs to be cleared up first. >From the discussion at the
meeting, it seemed that there was a desire to use algebraic rules to
encode >conservation laws. It seems to me that this is inappropriate, as
it makes models over-determined.
>The use-case for this is that certain analysis tools only work if the
stoichiometry matrix for a model is of >full rank. Often such matrices
are not of full rank, and this is due to conservation laws in the
reaction
>network. It seems to me that the proper way to include such laws in the
model is as "constraints". Then for >the majority of tools that don't
care about the rank of the stoichiometry matrix, the conservation laws
can >be largely ignored, but used to provide a diagnostic check that
nothing has gone wrong. For a tool which >_does_ care about the rank of
the stoichiometry matrix, the model needs to be "reduced" down to a full
rank >form. This can be done by converting each constraint to an
algebraic rule, and removing one of the species
>determined by the constraint from the model. The choice of which
species to remove is essentially arbitrary, >but needs to be done in
order to prevent the model from becoming over-determined. A tool could
be written to >do this automatically.

Herbert Sauro

>--
>Darren Wilkinson
>email: darrenjwilkinson@btinternet.com
>home www: http://www.darrenjwilkinson.btinternet.co.uk/
>work www: http://www.staff.ncl.ac.uk/d.j.wilkinson/