Dependency graph for models with custom propensities

[ahellander]

There are two ways that immediately comes to mind for how to do this:

1.) Use regular expressions and try to analyze the string propensity function to determine what species it contains.

2.) Explicitly evaluate the propensities to test if they change value when a species changes copy number. (For each propensity, test changing all species in turn)

[briandrawert]

A third way:
By default make the dependency for a custom propensity dense ones (most inefficient, but always correct). Allow a parameter to the Reaction object that specifies the dependancies: e.g.
A M-M propensity [http://en.wikipedia.org/wiki/Michaelis%E2%80%93Menten_kinetics] would then be:
Reaction(reactants={S:1}, products={P:1}, dependancies=[S,E], propensity_function="k1_E_S/(Km+S)/vol")

[briandrawert]

Note, there is a related "bug" in that pyurdme always adds the 'vol' (divides if order 2, multiply if order 0) to custom propensities. See pyurdme.py line 1970.

I have made this a separate issue: #68

[ahellander]

Yeah, that is an option too. However, I think we push this for 1.1 since it becomes "only" a performance issue for custom propensities. Let's add this to the list of enhancements.