[pydstool-users] Assistance with large scale model re-definition

[Amogh J. <ja...@vt...>]

Hello,

I am currently working on a relatively large systems biology model of 25
variables. I am running into problems in PyDSTool, which I attribute to my
inexperience with python and PyDSTool. I have a list of questions, and
though I have tried reading the documentation and answers on the internet,
I still have many questions on the "right way" of modeling in PyDSTool.

1. *Model redefinition: *
In order to compare my model output with experimental conditions, I need to
continuously redefine the varspecs dictionary in order to set one or more
derivatives to '0' or other custom expressions (to simulate gene knockouts,
for example). I have managed to do this currently by creating a function
that generates "special" varspecs dictionaries by comparing the user
defined expression(s) with a reference dictionary containing the "basal"
reactions, and then creating a different model for each user-defined
condition. [I maintain a list of these conditions.]
I find that this method is not very efficient. Even if I pickle some of
these recurring model instances, I believe there must be a better/more
efficient method of dynamically creating models. I would greatly appreciate
any tips or links to discussions regarding model construction.

2. *Parameter estimation:*
I have tried reading the ParamEst documentation and have tried to take
apart the pest_test1.py script to carry out a least squares parameter
estimation. However, I am finding it hard to wrap my head around the
concepts of model Interfaces, and so on. Could you point me to a
step-by-step breakdown of the working of ParamEst? (I gave up after
creating functions that used the set() function to 'manually' pass
parameter combinations to calculate the residuals at each iteration in my
data fitting function)

Also, for my system of 25 equations, computing the trajectory for
trange=[0,1000] takes ~5s. Is this normal?

Thank you very much for your time, and thanks for the great tool!

Cheers!

Amogh Jalihal
Graduate Student
Genetics, Bioinformatics and Computational Biology
Virginia Tech