Question about behavior

[danielriosgarza]

Dear Devlin,

When I try to use Macaw, most of my reactions are flagged as "always blocked." I am questioning if this could be related to an unexpected behaviour of the code:

import cobra
from macaw.main import run_all_tests
model = cobra.io.read_sbml_model("test_model.sbml")
model.optimize()

model.optimize() <Solution 62.084 at 0x7fae212c7da0>

Then, after:

(test_results, edge_list) = run_all_tests(model)

Starting dead-end test...
 - Found 766 dead-end metabolites.
 - Found 674 reactions incapable of sustaining steady-state fluxes in either direction due to these dead-ends.
 - Found 90 reversible reactions that can only carry steady-state fluxes in a single direction due to dead-ends.
Starting duplicate test...
 - Skipping redox duplicates because no redox_pairs and/or proton_ids were provided.
 - Found 4 reactions that were some type of duplicate:
   - 0 were completely identical to at least one other reaction.
   - 2 involve the same metabolites but go in the opposite direction or have the opposite reversibility as at least one other reaction.
   - 4 involve the same metabolites but with different coefficients as at least one other reaction.
Skipping diphosphate test because IDs for mono- and diphosphate ions were not provided.
Starting loop test...
/home/drgarza/metabolic_modelling/lib/python3.12/site-packages/macaw/flux_based.py:78: FutureWarning: DataFrame.applymap has been deprecated. Use DataFrame.map instead.
  solutions = solutions.applymap(lambda x: 0 if abs(x) < zero_thresh else x)
 - Found 212 reactions involved in infinite loops.
Starting dilution test...

 - Found 106 metabolites for which adding a dilution constraint prevented fluxes through all associated reactions.
 - Found 224 reactions that are blocked by one or more dilution constraints.
1105 of the 1105 reactions in the given GSMM were flagged by at least one of the tests.

and,

model.optimize()

now returns,

<Solution 0.000 at 0x7fae1ef64d10>

Is this the expected behavior? Could the fact that my source model changes during the execution cause such a high number of "always blocked" reactions?

[Devlin-Moyer]

I believe I have found and fixed the bug responsible for this in the dead-end test; can you either send me your test model so I can see if the problem is fixed or update your installation of MACAW and test it yourself?

The fact that the model is edited by the tests is intended behavior, since I am trying to identify all cases like
a --> b <--> c --> d
(where a and b and c are metabolites and the arrows are reactions) where the reaction between b and c is effectively irreversible (at steady state), since b can only be produced and c can only be consumed, and edit the reaction between b and c to be b --> c. Unfortunately, the dead-end test sometimes seems to edit the directions and/or reversibilities of other reactions, and I'll find some of the reactions that are being wrongly edited, change something that stops them from being edited, and then later find out that that change made a different group of reactions start being wrongly edited, so I seem to have missed the actual underlying problem before and hope that I've got it this time.

[danielriosgarza]

Hi Devlin,

Thank you for the reply. I tested again, but the problem seems to persist.

I have attached a test model. I am interested in doing this for a large batch of models similar to this one. If you can make it work, it would be greatly appreciated.

Best regards,
Daniel

test_model.zip

[Devlin-Moyer]

It looks like the problem was that your model had some reactions that were irreversible but could only carry flux in the reverse direction, like so:
a <-- b
which caused problems with the dead-end test, because Cobrapy would put b in Reaction.products even though b is actually a reactant in that reaction, and the dead-end test assumes that reactions can always produce their products. I have edited the dead-end test so that it looks for and flips all such reactions (i.e. turns them from a <-- b into b --> a) before starting, and this seems to have resolved the problem. The number of dead-end metabolites in test_model.sbml has dropped from 766 to 624, and the number of dead-end reactions (which are the "always blocked" reactions) has dropped from 674 to 511. I have confirmed that all 511 of those reactions are incapable of carrying fluxes in either direction before MACAW starts editing things.