Use Devlin's algorithm to identify even more cycles

[hgscott]

Started doing it here, but was having a hard time using the code.

[hgscott]

Devlin ran his tests on my KBase model (think it still had the acetate loop issue), and sent me the results: https://github.com/C-CoMP-STC/GEM-mit1002/blob/4d620ff91b5d07d17b00bd21ec6a3e0a4bfe9da8/macaw-results/helen-alteromonas_test-results.csv

[hgscott]

I took the csv file that Devlin sent me, and I plotted the proportion of reactions had any entry other than "ok" for each test.

[hgscott]

Not really sure what to do next, so I read Devlin's paper:

• It is important to note that the tests in MACAW are designed to flag reactions that are worth examining for potential errors as well as reactions that may simply be near erroneous reactions, so the percentages listed here should not be interpreted as the percentage of reactions in each GSMM that contain one or more errors.
• Dilution constraints require that every metabolite should be produced more than it is consumed
• The dilution test produces a list of all reactions found to be blocked by adding a dilution constraints for a single metabolites and an edge list describing a bipartite network of metabolites and reactions
• Many of the reactions flagged by the dilution test are involved in thermodynamically infeasible loops that can perfectly recycle metabolites (frequently cofactors like ubiquinone)

[hgscott]

Don't want to go any further on the analysis on this until I run it on the correct version (ModelSeedPy) of the Amac model.

[hgscott]

Ran on the (nongapfilled) ModelSEEDpy model.

What do the different columns mean?

• dead_end_test
  • one or more metabolite IDs separated by semicolons indicating which dead-end metabolites participate in that reaction.
  • "only when going forwards" or "only when going backwards" if it was a reversible reaction that was not a dead-end but had a reactant or product that could only be consumed by all other reactions it participates in or only be produced by all other reactions it participates in.
• dilution_test
  • "always blocked" if the reaction was incapable of non-zero fluxes regardless of whether or not any dilution constraints were imposed on the model.
  • Devlin's documentation says that the other options are "blocked by dilution" or "unblocked by dilution", but my results all have lists of metabolites
  • It looks like the metabolites listed are all in the reaction (as either reactants or products)
  • So I'm assuming that the reaction was blocked when that metabolite has a dilution reaction
  • Not sure how to tell if any of the metabolite lists came from "unblocked by dilution", but Devlin's documentation says that is rare
• pathway
  • Combines all of the flagged reactions into a network, usually with several distinctly connected components that are not connected to one another.
  • The entry is a discrete integer indicating which connected component each reaction is in.
  • Reactions that were not flagged by any tests or not connected to any other reactions that were flagged by any tests (this only happens with reactions flagged by the dead-end or diphosphate tests, and is generally uncommon) are always assigned a "pathway" of 0.
  • The "loop_test" column, doesn't seem to add things to the pathways. There are multiple reactions marked as "in loop" that are in the 0 pathway.

[hgscott]

How do I use the results?