SciML
Conversion to non-linear system does not handle presence of differential equations:
This code:
@parameters v1 K1 v2=1.0 K2=0.4 d
@variables t Y(t)
@species X(t)
eq = Differential(Y) ~ v1*Y/(Y+K1) - d*Y
rx1 = @reaction v1*$Y/($Y+K1), 0 --> X
rx2 = @reaction d, X --> 0
@named rn = ReactionSystem([rx1, rx2, eq], t)
ns = convert(NonlinearSystem, rn);
println.(ns.eqs)
returns
0 ~ (v1*Y(t)) / (K1 + Y(t)) - d*X(t)
Differential(Y(t)) ~ (v1*Y(t)) / (K1 + Y(t)) - d*Y(t)
However, the Differential(Y(t)) should be set to 0.
We don't handle conversion of non-reactions currently, but assume the user has input them as they desired in their model. So that would need to be added by someone, presumably it can reuse what is done in MTK.
It feels fairly obvious that if someone has made a mixed ReactionSystem/Differential equation ODE, that if they want to convert it to a NonlinearSystem they want those derivatives to be 0 as well. I think we should make this the case (ideally put a warning until we have fixed this), and then maybe put in a kwarg to disable this automatic conversion.
Yes, I agree. Feel free to PR if you want to add that. You can probably reuse what is in MTK for the conversion to make sure the derivatives are on the correct side and such.