Stuck at "Computing IO-equations" about structuralidentifiability.jl HOT 9 CLOSED

@LinneaFranssen , thanks for the example!
Global identifiability is a much more challenging computation, so it may be stuck. Your system does not look too hard, I will check what causes the problem.

For the initial conditions, this tool can only include them into the local identifiability analysis. I suggest you to try this webapp (the same algorithm has also implementations in maple and julia)

from structuralidentifiability.jl.

Dear @pogudingleb,

thank you very much for the quick response and for looking into this!

I had tried that app actually but it kept disconnecting quickly. I will try again. :)

from structuralidentifiability.jl.

Just an update: it still happens with this code

odeMyModel = @ODEmodel(
    Complex'(t) = 1/C*(((2*kon*free_receptor(t)*Drug(t)-koff*Complex(t))*C) - (ke_Complex*Complex(t)*C) - ((kon_2*free_receptor(t)*Complex(t)-2*koff*Complex_2(t))*C)),
    Complex_2'(t) = 1/C*(((kon_2*free_receptor(t)*Complex(t)-2*koff*Complex_2(t))*C) - (ke_Complex_2*Complex_2(t)*C)),
    Drug'(t) = 1/C*(-(ke_Drug*Drug(t)*C) - ((2*kon*free_receptor(t)*Drug(t)-koff*Complex(t))*C) + (Fa*ka*Drug_SC(t))),
    free_receptor'(t) = 1/C*(-((2*kon*free_receptor(t)*Drug(t)-koff*Complex(t))*C) + (66/2500*C) - (kdeg_free_receptor*free_receptor(t)*C) - ((kon_2*free_receptor(t)*Complex(t)-2*koff*Complex_2(t))*C)) + u_free_receptor(t),
	Drug_SC'(t) = -(Fa*ka*Drug_SC(t)) - ((1-Fa)*ka*Drug_SC(t)) + u_drug_SC(t),
    y1(t) = Drug(t),                                   # output free Drug
    y2(t) = free_receptor(t) + Complex(t) + 2*Complex_2(t)      # output total receptor
)

 SI_output = assess_identifiability(odeMyModel,0.99)

and Julia crashes out silently without error messages - even from the REPL.

from structuralidentifiability.jl.

@LinneaFranssen
Just two clarify, is the second system the same as the first one?

from structuralidentifiability.jl.

from structuralidentifiability.jl.

Very interesting case! The model is in some way special, and I did not optimise the code for models of this sort thinking that this does not happen in practice. I have an idea how to speed up the computation, I will try it over the weekend.
I also did the computation a bit differently (by changing manually some things in the code) and I think that all the parameters in your original model are globally identifiable.

from structuralidentifiability.jl.

Sorry, taking longer to fix this. Hope to do this by the end of the year)

from structuralidentifiability.jl.

@LinneaFranssen
Sorry, this has not been fixed last year as hoped, but I did not forget and hope to get to this asap

from structuralidentifiability.jl.

@LinneaFranssen
It took a bit longer than I anticipated, but it works now! The version on GitHub (can be installed by Pkg.add("https://github.com/SciML/StructuralIdentifiability.jl"); it will be in the release 0.4.9) computes both models in a couple of minutes on my laptop. Let me know if you encounter any further issues.

The model turned out to be very interesting algorithmically, I wonder if it is published somewhere?

from structuralidentifiability.jl.