draft `Model` to MTK.___System conversion. #40

paulflang · 2021-04-14T19:19:18Z

reactionsystem.jl should probably eventially be moved to the ModelingToolkit.jl package (to get rid of all the ModelingToolkit/Catalyst/Symbolics dependencies here). Other things to clarify:

Model needs to have localParameters promoted.
We need to figure out how to make kineticLaws extensive (see SBML.jl and SciML integration #31 )
We need to figure out how to extract forward and reverse kineticLaws from Reactions where reversible=true.
We need to decide on the type of Reaction.kinetic_math. This will allow to convert it to the right MTK symbols.

anandijain · 2021-04-14T20:05:39Z

src/reactionsystem.jl

+end
+
+""" ODESystem constructor """
+function ModelingToolkit.ODESystem(sbmlfile::String)


eventually need to fix the kwargs, so that @named mymodel = ODESystem(sbmlfile) will work

where can I find documentation of the @named macro. I've never used it.

help?> @named Rewrite @named y = foo(x) to y = foo(x; name=:y).

I think you basically just need to change this to
function ModelingToolkit.ODESystem(sbmlfile::String; kwargs...)
and then pass them to the actual constructor

anandijain · 2021-04-14T23:45:33Z

src/reactionsystem.jl

+
+""" Expand reversible reactions to two reactions """
+function expand_reversible(model)
+    model  # Todo: convert all Reactions that are `reversible=true` to a forward and reverse reaction with `reversible=false`.


Do we need to do this if X + Y ↔ XY is valid catalyst? https://catalyst.sciml.ai/stable/tutorials/basics/#Combining-several-reactions-in-one-line

Alternatively, it could be something that Catalyst needs. It seems like a useful transform to have.

Maybe worth pinging Sam about

The way as SBML would describe the kineticLaw for a reversible reaction is X + Y ↔ XY is kAs*X*Y-kDi*XY. Catalyst however needs kAs*X*Y, kDi*XY, that is two kineticLaws. I mean this is a simple example, but somehow we need to find a way that generally works in splitting a combined kineticLaw in a forward and reverse kineticLaw. Any suggestions how to do this in a clean way?

exaexa · 2021-04-15T18:23:39Z

So I finally got to reading this, quick comments:

reactionsystem.jl should probably eventially be moved to the ModelingToolkit.jl package (to get rid of all the ModelingToolkit/Catalyst/Symbolics dependencies here).

Yeah, I'm not sure that others would really welcome pulling in the extra dependencies. Also, the code is very ODE specific.

We need to figure out how to extract forward and reverse kineticLaws from Reactions where reversible=true.

Does this mean that there are cases that reaction has 2 different math rules that get triggered by changing the reaction direction?

We need to decide on the type of Reaction.kinetic_math. This will allow to convert it to the right MTK symbols.

I'll keep it at SBML.Math, it is the best representation of how libsbml "parses" the MathML. Also, writing a translator from it to any other representation is super easy, I can probably write it. Can you send me an example of the MTK symbols you want to have at the end of the conversion?

paulflang · 2021-04-15T19:04:52Z

Yeah, I'm not sure that others would really welcome pulling in the extra dependencies. Also, the code is very ODE specific.

Yeah, I just thought for developing purposes it is easier if it is in the same repo. But can be moved now or whenever we feel like things have stabilised. What do you think?

Does this mean that there are cases that reaction has 2 different math rules that get triggered by changing the reaction direction?

Not quite, if I understand you correctly. It just means that that you have a single math rule for the reaction rate which can be positive or negative, depending on whether the forward or reverse reaction is dominating. Consider the reaction X + Y ↔ XY where SBML would set reversible=true. This reaction would not have seperate kineticLaw for forward and reverse reaction. Rather it would have a single rate law kAs*X*Y-kDi*XY. However, for conversion into a ReactionSystem we need kAs*X*Y and kDi*XY as two seperate kineticLaws.

Also, writing a translator from it to any other representation is super easy, I can probably write it. Can you send me an example of the MTK symbols you want to have at the end of the conversion?

I think SBML.jl should be agnostic of MTK (in the long run). It would make more sense to convert to Symbolics.jl symbols, I believe.
Long story short, everything that is a <ci>x</ci> should become Num(Variable(Symbol(x))), <times/> should become the Julia * operator, plus/> the Julia + operator, etc. And then it should be straightforward for me to convert that to the corresponding MTK types.
But @anandijain knows more about that.

paulflang · 2021-04-15T19:11:12Z

For an XML example of bidirectional reaction you can look at #39

anandijain · 2021-04-15T19:13:29Z

Yeah to that very last point. You can quickly see how MathML does this lowering here https://github.com/SciML/MathML.jl/blob/main/src/maps.jl. Basically I just string match the name of the first element of the "apply" XML node and return the equivalent Julia function

anandijain · 2021-04-21T22:01:31Z

Why was this closed?

laurentheirendt · 2021-04-22T06:26:00Z

it was probably automatically closed when I merged #39 ...

paulflang · 2021-04-22T17:48:27Z

@fbergmann or @skeating: When an SBML reaction is reversible, it still only contains a single kineticLaw. Is there a function in libSBML that separates the single kineticLaw into a forward and reverse kineticLaw?

anandijain · 2021-04-22T21:48:38Z

src/reactionsystem.jl

+    end
+    subsdict = _get_substitutions(model)
+    # PL: Todo: @Anand: can you convert kinetic_math to Symbolic expression. Perhaps it would actually better if kinetic Math would be a Symbolics.jl expression rather than of type `Math`? But Mirek wants `Math`, I think.
+    kl = substitute(reaction.kinetic_math, subsdict)  # PL: Todo: might need conversion of kinetic_math to Symbolic MTK expression


IIRC @exaexa was taking this Math* sublanguage conversion. But I can get started on it for sure

Manifest.toml

… to ODEProblem

…ing}

… is nothing but throw warning.

… ReactionSystem.

…simulation results of pure ReactionSystem will be therefore be incorrect

ChrisRackauckas · 2021-06-14T10:49:56Z

src/reactionsystem.jl

+""" ODEProblem constructor """
+function ModelingToolkit.ODEProblem(model::Model,tspan;kwargs...)  # PL: Todo: add u0 and parameters argument
+    odesys = ODESystem(model)
+    ODEProblem(odesys, [], tspan; kwargs...)
+end
+
+""" ODEProblem constructor """
+function ModelingToolkit.ODEProblem(sbmlfile::String,tspan;kwargs...)  # PL: Todo: add u0 and parameters argument
+    odesys = ODESystem(sbmlfile)
+    ODEProblem(odesys, [], tspan; kwargs...)
+end


I would assume it would be best to have the user explicitly construct the ODESystem -> ODEProblem. This is one of those shorthands that causes doc problems.

Looks fishy for sure, and from the user perspective the code with the shorthand is almost as long as without the shorthand.

exaexa · 2021-06-17T16:11:27Z

This is now partly merged and partly going to https://github.com/paulflang/SBMLToolkit.jl . Thanks!

Update Project.toml

draft Model to MTK.___System conversion.

cfe431e

anandijain reviewed Apr 14, 2021

View reviewed changes

laurentheirendt marked this pull request as draft April 16, 2021 07:22

laurentheirendt closed this Apr 16, 2021

laurentheirendt deleted the branch LCSB-BioCore:master April 16, 2021 07:55

laurentheirendt mentioned this pull request Apr 22, 2021

Draft for converting Model to ModelingToolkit types #32

Closed

laurentheirendt reopened this Apr 22, 2021

paulflang added 5 commits April 22, 2021 20:28

add kwargs to constructors of MTK types (to support @nAmed)

fd6ad63

throw error when model contains reversible reactions

32f5e4d

add initialConcentration to species

5f94736

convert initial_concentrations to initial_amounts

d188def

draft to_extensive_math function

402acbc

anandijain reviewed Apr 22, 2021

View reviewed changes

add test case for to_initial_amounts

6558506

anandijain reviewed Apr 23, 2021

View reviewed changes

Manifest.toml Outdated Show resolved Hide resolved

paulflang added 7 commits April 23, 2021 21:29

add test for mtk_reaktion() and -getsubstitution()

de7d16b

untrack Manifest.toml

bf6b080

add test case for u0map

f837186

add test for get_paramap()

f9221d1

add tests for creating ReactionSystem

09a7228

add tests for ODESystem and ODEProblem constructors

8e6a03c

add conversion_options argument to readSBML()

7e58ca6

paulflang added 7 commits June 6, 2021 14:39

add tests for conversion of ReactionSystem to ODESystem and ODESystem…

778ba4a

… to ODEProblem

remove compartments with no size from parameters

48db58b

fix simulation for with ceil or factorial in kineticLaw

3600346

parse SBML Booleans to SBML.MathVal{Bool} instead of SBML.MathVal{Str…

9932b3b

…ing}

add floor as supported SBML function.

7e11c2e

use initial_concentration as initial_amount when Compartment.size…

d714fb5

… is nothing but throw warning.

throw error when attempting to convert a model without reactions to a…

ea55421

… ReactionSystem.

This was referenced Jun 7, 2021

SBML.jl and SciML integration #31

Closed

add support for running SBML converters #55

Merged

paulflang added 5 commits June 7, 2021 22:57

set version to 0.4.1

b11f448

add independent variable to the states

4169c47

make all species function of time to get structural_simplify working

87766de

add tests for structural_simplify

f460328

raise warning when species are boundaryCondition or constant and …

59fd135

…simulation results of pure ReactionSystem will be therefore be incorrect

ChrisRackauckas reviewed Jun 14, 2021

View reviewed changes

This was referenced Jun 14, 2021

add reversible flag #57

Merged

add initial concentration, fix certainty of units #58

Merged

support relational operators in math #59

Merged

paulflang added 6 commits June 14, 2021 15:25

fix handling of hOSU species

f726cdb

sync with upstream master

3c9d313

sync upstream master

67a8f84

fix initialConcentration parsing

bbe19eb

fix hOSU handling

4ae0332

fix tests after merging upstream master

1070601

anandijain mentioned this pull request Jun 17, 2021

remove String dispatches for MTK constructors SciML/SBMLToolkit.jl#2

Merged

exaexa closed this Jun 17, 2021

paulflang referenced this pull request Jun 17, 2021

Merge pull request #70 from anandijain/patch-5

646b2d4

Update Project.toml

paulflang mentioned this pull request Jun 17, 2021

Require the SBML version with Symbolics support SciML/SBMLToolkit.jl#3

Closed

paulflang mentioned this pull request Jul 10, 2021

piecewise handling fails #112

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

draft `Model` to MTK.___System conversion. #40

draft `Model` to MTK.___System conversion. #40

paulflang commented Apr 14, 2021

anandijain Apr 14, 2021

paulflang Apr 14, 2021

anandijain Apr 14, 2021

anandijain Apr 14, 2021

paulflang Apr 15, 2021

anandijain Apr 14, 2021

anandijain Apr 14, 2021

paulflang Apr 15, 2021

exaexa commented Apr 15, 2021

paulflang commented Apr 15, 2021

paulflang commented Apr 15, 2021

anandijain commented Apr 15, 2021

anandijain commented Apr 21, 2021

laurentheirendt commented Apr 22, 2021

paulflang commented Apr 22, 2021

anandijain Apr 22, 2021

ChrisRackauckas Jun 14, 2021

exaexa Jun 14, 2021

exaexa commented Jun 17, 2021

draft Model to MTK.___System conversion. #40

draft Model to MTK.___System conversion. #40

Conversation

paulflang commented Apr 14, 2021

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

exaexa commented Apr 15, 2021

paulflang commented Apr 15, 2021

paulflang commented Apr 15, 2021

anandijain commented Apr 15, 2021

anandijain commented Apr 21, 2021

laurentheirendt commented Apr 22, 2021

paulflang commented Apr 22, 2021

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

exaexa commented Jun 17, 2021

draft `Model` to MTK.___System conversion. #40

draft `Model` to MTK.___System conversion. #40