Fix FAD/Oxygen-Peroxide Duplicate Reactions

[Devlin-Moyer]

Main improvements in this PR:

As proposed in #639:

• Removes MAR00059 for being a less accurate version of MAR01706
• Changes the GPR of MAR01706 to ENSG00000113790 or ENSG00000133835 and adds PMID:19357427 to its references
• Removes MAR01639 for being redundant with MAR01638
• Removes MAR03301 for being redundant with MAR02804
• Removes MAR03306 for being redundant with MAR02808
• Removes MAR03311 for being redundant with MAR02810
• Removes MAR04966 for being redundant with MAR02818
• Removes MAR03364 for being redundant with MAR03276
• Removes MAR03369 for being redundant with MAR03289
• Replaces MAM02630x (O₂) with MAM01802x (FAD) and MAM02041x (H₂O₂) with MAM01803x (FADH₂) in MAR00442, MAR00759, MAR00777, MAR00800, MAR00833, MAR00872, MAR00952, MAR00968, MAR00974, MAR00975, MAR00977, MAR00978, MAR00982, MAR01001, MAR01005, MAR01018, MAR01162, MAR01205, MAR01254, MAR01271, MAR01291, MAR01296, MAR01303, MAR01600, MAR01704, MAR02408, MAR02510, MAR03053, MAR03054, MAR03055, MAR03056, MAR03057, MAR03062, MAR03066, MAR03070, MAR03074, MAR03078, MAR03082, MAR03086, MAR03090, MAR03094, MAR03098, MAR03102, MAR03326, MAR03330, MAR03334, MAR03338, MAR03342, MAR03346, MAR03350, MAR03356, MAR03360, MAR03459, MAR03466, MAR03470, MAR03498, MAR03506, MAR03512, MAR03849, MAR03984, MAR04001, MAR04650, MAR04789, MAR04981, MAR05025, MAR05026, MAR05028, MAR05097, MAR05100, MAR05108, MAR05119, MAR05175, MAR05178, MAR05180, MAR05182, MAR05184, MAR05189, MAR05196, MAR05300, MAR05327, MAR05331, MAR05350, MAR05366, MAR05372, MAR05374, MAR05377, MAR05382, MAR05421, MAR05428, MAR05431, MAR06284, MAR06285, MAR06322, MAR06344, MAR06345, MAR06933, MAR06936, MAR06943, MAR06946, MAR06949, MAR06952, MAR06959, MAR06962, MAR06964, MAR07641, MAR07647, MAR07704, MAR07706, MAR08017, MAR08021, MAR08415, MAR08689, MAR08690, MAR08755, MAR11529, MAR11534, MAR11539, MAR12372, MAR12808, and MAR20066
• Creates new reaction MAR20114 to represent reduction of O₂ by FADH₂: MAM02630x + MAM01803x -> MAM02041x + MAM01802x, GPR: ENSG00000161533 or ENSG00000168306 or ENSG00000087008 or ENSG00000110887 or ENSG00000203797 or ENSG00000007171 or ENSG00000148832 or ENSG00000179761 or ENSG00000158125, reference: PMID:30378035

I hereby confirm that I have:

• Tested my code on my own computer for running the model
• Selected develop as a target branch
• Any removed reactions and metabolites have been moved to the corresponding deprecated identifier lists

[Devlin-Moyer]

@feiranl @haowang-bioinfo when I initially made this PR, it passed all of the metabolic task checks, but when I merged in the latest develop branch just now, it failed the "verification" tasks, but the error message doesn't give me enough details to follow up on what exactly is causing that issue, so I'm not sure how to proceed. (no rush to answer, since I know I've opened a bunch of issues and PRs recently, but just flagging that I don't know how to respond to this test failing)

[haowang-bioinfo]

when I initially made this PR, it passed all of the metabolic task checks, but when I merged in the latest develop branch just now, it failed the "verification" tasks, but the error message doesn't give me enough details to follow up on what exactly is causing that issue, so I'm not sure how to proceed. (no rush to answer, since I know I've opened a bunch of issues and PRs recently, but just flagging that I don't know how to respond to this test failing)

will manually check what specific task(s) are affected by this PR, which appears to involve over amount of changes

[haowang-bioinfo]

manual check located that the step of "reverse MAR00059 and remove 0.5 O2" (commit 9d7b81c) caused the failure in checking verification tasks. Specifically, it allows:

• Rephosphorylation of ATP in cytoplasm, mitochondria, peroxisome, Golgi, ER, nucleus, lysosome
• Rephosphorylation of ATP from protons
• Generation of reduction potential in cytoplasm, mitochondria, and
  peroxisome (FADH2)
• mol ATP phosphorylated per mol glucose consumed exceeds 40
• ATP production from Pi and O2

these tasks suppose to fail for a mass and energy balanced model

[JonathanRob]

Good catch, @haowang-bioinfo.

The removal of the 0.5 O2 in that reaction resulted in it becoming mass-imbalanced, and reversing the reaction led to a thermodynamically infeasible reaction.

For example, a loop could be formed with reaction MAR01706:

Rxn	Equation
MAR01706	3alpha,7alpha-dihydroxy-5beta-cholest-24-enoyl-CoA [x] + H2O [x] ⇒ (24R,25R)3alpha,7alpha,24-trihydroxy-5beta-cholestanoyl-CoA [x]
MAR00059	(24R,25R)3alpha,7alpha,24-trihydroxy-5beta-cholestanoyl-CoA [x] + FAD [x] ⇒ 3alpha,7alpha-dihydroxy-5beta-cholest-24-enoyl-CoA [x] + FADH2 [x]
Net	H20 + FAD -> FADH2
Net (w/ O2)	H20 + FAD -> FADH2 + 0.5 O2

Even without O2, it allows water to reduce FAD, which is incorrect. With O2, it allows the same reduction, but also produces O2 from water, which is also problematic.

@Devlin-Moyer, what was your justification for reversing and removing O2 from MAR00059? It seems to me that neither of those changes should be made.

[Devlin-Moyer]

From #639:

ACOX2 irreversibly oxidizes its substrates (see Uniprot and this paper), so MAR00059 should point in the opposite direction, and it should either not involve O₂ at all, or involve no FAD or FADH₂ and consume 1 O₂ and produce 1 H₂O₂, as in the other three reactions ACOX2 is currently associated with (MAR01296, MAR01704, and MAR04001)

I was just thinking about redox when putting together this issue, and since every other reaction on this list was straightforward redox, it never occurred to me to check the empirical formulas of these metabolites to figure out if the 0.5 O₂ was necessary; I stand by my initial claim that the original form of MAR00059 is not a reaction that ACOX2 catalyzes, but I don't know which gene it should be associated with instead or if there are different modifications that should be made to its stoichiometry

[Devlin-Moyer]

Hmm apparently if you remove the 0.5 O₂ from the original reaction, it exactly matches R04547, but KEGG doesn't associate that with any human enzymes, so do we know if this reaction happens in humans at all? Does removing it create a gap?

[Devlin-Moyer]

I also found (24R,25R)3alpha,7alpha,24-trihydroxy-5beta-cholestanoyl-CoA in the primary bile acid biosynthesis KEGG pathway (near the bottom left, directly above chenodeoxycholate), where it participates in two reactions already represented by MAR01706 and MAR01708.

Now that I look at MAR01706 and MAR00059 (as it originally was) side-by-side:

MAR00059: 3alpha,7alpha-dihydroxy-5beta-cholest-24-enoyl-CoA [x] + FADH2 [x] + 0.5 O2 [x] ⇒ (24R,25R)3alpha,7alpha,24-trihydroxy-5beta-cholestanoyl-CoA [x] + FAD [x]
MAR01706: 3alpha,7alpha-dihydroxy-5beta-cholest-24-enoyl-CoA [x] + H2O [x] ⇒ (24R,25R)3alpha,7alpha,24-trihydroxy-5beta-cholestanoyl-CoA [x]

I think the situation is that MAR00059 is a less-accurate version of MAR01706, maybe added during gap filling at some point before MAR01706 existed, so probably MAR00059 should just be removed altogether as a (near-)duplicate of MAR01706. Let me know if you think that checks out and I can update everything accordingly

[haowang-bioinfo]

Hmm apparently if you remove the 0.5 O2 from the original reaction, it exactly matches R04547, but KEGG doesn't associate that with any human enzymes, so do we know if this reaction happens in humans at all? Does removing it create a gap?

so the change made to MAR00059 is based on KEGG, which actually is known with rich amount of unbalanced reactions. Genes associate with R04547 all appear to be from bacterial kingdom.

Does removing it create a gap? Let me know if you think that checks out

it's confirmed that removal of MAR00059 could fix aforementioned failures (allows water to reduce FAD, produces O2 from water) in verification tasks

[JonathanRob]

Thanks for the thorough update, @Devlin-Moyer. I agree that removing MAR00059 seems like a good solution, and thank you @haowang-bioinfo for verifying.

[haowang-bioinfo]

should the GPR of MAR01706 be modified, and how?

[Devlin-Moyer]

Based on the research I did in #634 and #637, the GPR of MAR01706 should be EHHADH or HSD17B4 (i.e. ENSG00000113790 or ENSG00000133835), so the necessary changes are removing ACAA1, which is a 3-ketoacyl CoA thiolase, which is clearly not what's happening in MAR01706, and changing the "and" to an "or"
I'm currently working on updating #639 to accurately and completely describe the issues with MAR00059, why it should be removed, and can also throw in this change to the GPR of MAR01706

[haowang-bioinfo]

@Devlin-Moyer which Rhea reaction could be associated with MAR01706?

[Devlin-Moyer]

Pretty sure it's 47408/47409

[haowang-bioinfo]

@Devlin-Moyer which Rhea reaction could be associated with MAR01706?

Pretty sure it's 47408/47409

will update Rhea id for MAR01706 in #707