Record values of other observables

[jonrkarr]

• Constant species
• Parameters
• Species and parameters set via assignment rules
• Reaction fluxes
• Compartment sizes

[jmrohwer]

Assignment rules can now be tracked. See PySCeS/pysces#57 (comment). Fixed in latest PySCeS Release (1.0.2).

[jmrohwer]

This continues the discussion from here where certain model outputs are not recorded.

The unrecorded variables are all assignment rules. In fact they are recorded by PySCeS, so it is a wrapper issue. Staying with Biomodels 297. I've had a brief look at the wrapper code. The wrapper records the simulation results here:

Biosimulators_PySCeS/biosimulators_pysces/core.py

Line 158 in 2d97929

results = model.data_sim.getAllSimData(lbls=False)

@luciansmith I have verified that this array contains all the simulation data including the assignment rules, in fact it is the same data as in the model.sim DataFrame (refer here), just presented as a numpy array with or without labels.

It therefore has to do with filtering of the model attributes which as far as I can see happens here:

Biosimulators_PySCeS/biosimulators_pysces/core.py

Lines 309 to 310 in 2d97929

	dynamic_ids = ['Time'] + list(model.species) + list(model.reactions)
	fixed_ids = (set(model.parameters) | set(model.__compartments__.keys())).difference(set(model.__rules__.keys()))

The rules are not part dynamic_ids nor fixed_ids as far as I can tell. I know too little about the wrapper implementation in general to be able to tell how this is generally done, or what would need to change for them to be recorded.

The assignment and rate rules are in a dictionary mod.__rules__ where the ID of each rule is the key. For each key (rule) the value is another dictionary with the following keys:

• name (the SBML ID of the rule, this is the same as the key in the top-level dict)
• formula
• type (assignment or rate)
• symbols (list of symbols in the rule)
• code string (formula adapted with class references so it can be executed in PySCeS)

Hopefully this will give you sufficient info to track this down further. Unfortunately I don't have time in the coming weeks to work on this myself.

[luciansmith]

Thanks for the analysis! Now that the pipeline works again, I can tackle this.

[luciansmith]

And, this now works! And the PySCeS wrapper can now handle 711 of the biomodels, waaaay up from the previous 241. That section was indeed key! So thanks.

[jmrohwer]

I'm happy you were able to track this down 😃