Repository navigation
A few reactions that look fishy #348
Description
Description of the issue:
-
These two reactions are probably the same, but different stoichiometry. They have the same GPR.
MAR01364, MAR03880:
arachidonate[c] + O2-[c] => 12-peroxy-(5Z,8Z,10E,14Z)-eicosatetraenoate[c] + H+[c]'
3 arachidonate[c] + 5 H+[c] + 5 O2-[c] => 3 12-peroxy-(5Z,8Z,10E,14Z)-eicosatetraenoate[c] + 4 H2O[c]' -
These three reactions doesn't make fully sense to me, should probably be investigated:
MAR03964, MAR07799 and MAR00080 -
These two reactions are most likely the same, but with different stoichiometry (same GPR):
MAR09985, MAR11307:
5-aminolevulinate[c] + 2 H+[c] <=> 2 H+[s] + 5-aminolevulinate[s]'
H+[s] + 5-aminolevulinate[s] => 5-aminolevulinate[c] + H+[c]'
Copied from #346
- These reactions could be joined into a reversible rxn to avoid loops, they have the same GPRs:
'MAR08482': {'3-oxotetradecanoyl-[ACP][c] + H+[c] + NADPH[c] => HMA[c] + NADP+[c]'}
'MAR02358': {'HMA[c] + NADP+[c] => 3-oxotetradecanoyl-[ACP][c] + H+[c] + NADPH[c]'}
Reacted by Hao Wang
Activity
Devlin-Moyer commentedon Jun 17, 2024
MAR09985 and MAR11307 were addressed in #588/#623
MAR02358 was removed for being a duplicate of MAR08482 in #657
MAR01364, MAR03880, MAR03964, MAR07799 and MAR00080 don't appear to have been mentioned in any other issues or pull requests. I'm not sure I fully understand what's going on in MAR01364 or MAR03880, but I agree with your assessment that they seem to represent the same real-world reaction. MAR03964, MAR07799 and MAR00080 appear to represent V-type ATPases.
johan-gson commentedon Jun 17, 2024
Nice that you looked into this! I'm thinking we can maybe drop the issue with the ATP hydrolysis functions and close this, or what do you think? They look a bit similar, but could be 3 real reactions. Nice to have someone else in Boston working on this :)
Devlin-Moyer commentedon Jun 17, 2024
after taking a closer look, MAR03964 seems to be an ATP maintenance reaction rather than a representation of a V-type ATPase (since the proton it produces is still in the cytosol as opposed to the lysosome), and I suspect that MAR07799 is a less accurate version of MAR00080. The GPRs of MAR07799 and MAR00080 involve the same set of genes, but all the genes are just ORed together in the GPR of MAR07799 and grouped into a long nested thing in the GPR of MAR00080 that seems to be representing the existence of alternative isoforms for several subunits of the V-type ATPase complex in humans (source). So there's an argument to be made that MAR07799 should be removed
Devlin-Moyer commentedon Jun 17, 2024
the arachidonate reactions probably merit a separate issue, because there seem to be several similarly suspicious pairs in addition to the one you mentioned that have the peroxide in a slightly different position, e.g. MAR03882 and MAR01355; MAR03884 and MAR01361
johan-gson commentedon Jun 17, 2024
Sounds good, I think you know this far better than I do, so just go ahead!
edkerk commentedon Mar 23, 2026
@Devlin-Moyer Is this Issue then completely resolved, after #829 ?
Devlin-Moyer commentedon Mar 23, 2026
#829 takes care of the second item in the original list, but I don't believe the potential redundancy of
MAR01364andMAR03880has been resolvededkerk commentedon Jul 11, 2026
The MAR01364 / MAR03880 item is now resolved. Both are the same GPR-less arachidonate → 12-peroxy-eicosatetraenoate reaction, but MAR01364 (1:1) is charge-imbalanced (reactants −2, products 0) while MAR03880 is balanced, so MAR01364 is removed in #1040.
While checking, the cause turned out to be class-wide, which is the "separate issue" flagged earlier. The simple hydroperoxide fatty acids are each one H short: the formula is
CnH30O4where the correct HPETE/HPDHA anion isCnH31O4(arachidonateC20H31O2+ O₂ → 12-HPETEC20H31O4balances exactly). Because of this the clean 1:1 dioxygenations do not balance, and inflated-stoichiometry "workaround" reactions (like MAR03880) were added to force balance around the wrong formula.Metabolites that need +1 H (
CnH30O4→CnH31O4):Reactions currently charge-imbalanced because of these formulas (all mass-balanced; charge off by +1 to +4), which would balance once the formulas are corrected:
MAR01355, MAR01358, MAR01361, MAR01363, MAR01405, MAR01410, MAR01411, MAR01412, MAR01413, MAR01414, MAR01416, MAR01417(plus MAR01364, removed here).To resolve the class:
CnH30O4→CnH31O4).parent + O₂ → hydroperoxide.O2-, MAM02631) where molecular O₂ (MAM02630) is the actual lipoxygenase co-substrate; those should be switched to O₂.fix: remove charge-imbalanced duplicate lipoxygenase reactions (#348)
edkerk commentedon Jul 11, 2026
Update: on closer analysis the full formula correction cascades much wider than my previous comment implied. The hydroxy (HETE/HDHA) products are also one H short, so the reduction reactions balance with the current formulas, and correcting the formulas touches ~23 metabolites and ~105 reactions with non-uniform imbalances.
So #1040 now removes all five charge-imbalanced 1:1 duplicates that have balanced multi-stoichiometry counterparts (MAR01364/03880, MAR01355/03882, MAR01358/03878, MAR01361/03884, MAR01405/03261). That resolves the duplicate-reaction subject of this issue while keeping the model balanced.
The remaining formula/balance correction, and the genuinely charge-imbalanced non-duplicate reactions, are split out to #1042.
fix: remove charge-imbalanced duplicate lipoxygenase reactions (#348) (…
edkerk commentedon Sep 27, 2026
Follow-up in #1105 to the reactions removed in #1040: each kept reaction now records the removed duplicate in its
rxnRetired, so the old ids can be traced.A correction to my earlier comment here: the "X-peroxy" metabolites (MAM00302, MAM00335, MAM01214, MAM01246, MAM00275, MAM00303, MAM00372, MAM00404, MAM00418, MAM01198, MAM01215) are not hydroperoxides that are one H short. Their formulas (C20H30O4 and C22H30O4, charge -1) are those of the peroxyl radicals (ROO•) formed in free-radical, non-enzymatic peroxidation of arachidonate and DHA, the route to isoprostanes and neuroprostanes (PMID:2123555, humans in vivo; PMID:10694406, DHA-derived neuroprostanes). Converting them to hydroperoxides, as planned in #1042, would duplicate the existing 8(S)- and 12(S)-HPETE metabolites and unbalance the reactions that consume them. So their formulas stay. MAM00302 and MAM01246 already had peroxyl radical structures; #1104 gives the same kind of structure to the others that have one (MAM00275, MAM00303, MAM00335, MAM00404, MAM00418, MAM01198, MAM01214, MAM01215), replacing hydroperoxide or unrelated SMILES and removing the cross-references of those other compounds (e.g. 8-HPETE CHEBI:146288 on MAM01214, retinyl palmitate ids on MAM00303). MAM00372 has no structure yet. #1042 should be rescoped accordingly.
Implemented in #1105 (retired pointers); structures in #1104.
edkerk commentedon Oct 5, 2026
Fixed in Human 2.1.0.