Reaction equations have wrong cofactors or directions, and some duplicates form loops #1173

Several equations do not match their enzyme or their own cross-references, some reactions run in a direction their chemistry does not allow, and some duplicates form cycles that pFBA uses. A few reactions have no human enzyme. Most of these come from HMR or Recon3D and are older than v2.0.0; some were introduced, or left half-done, by #945, #1028, #1039, #1090, #1105 and #1106.

MAR10461  H2O + NADP+ + indole => 2 H+ + NADPH + indoxyl                     CYP2E1   (a P450 that makes NADPH, without O2)
MAR00814  ... + UDP-N-acetyl-D-galactosamine => ... + V3(NeuAc)2-Gb5Cer + UDP ST8SIA1 (a sialyltransferase with a GalNAc donor)
MAR01389  15(S)-HETrE + H2O <=> 15-HpETrE + 2 H+                             (an oxidation written as a hydrolysis, charge +2)
MAR04575  (S)-dihydroorotate[c] + FMN[c] <=> orotate[c] + FMNH2[c]          DHODH    (a mitochondrial quinone reductase)
MAR05381  PAPS + thioredoxin => PAP + sulfite + oxidized thioredoxin          SULT1A3  (PAPS reductase, EC 1.8.4.8, is bacterial and fungal)

Evidence

Equations rewritten

Directions and spontaneous flags (the rules of #1112)

These directions and flags are older than v2.0.0; #1103, #1105 and #1106 applied the rules of #1112 in part, and #1033 added the Ca2+ channel genes of MAR11311 without the bound.

Reactions added

Reactions removed (deprecated, or merged into the twin in brackets)

Notes and references

data/thermodynamics/loopExceptions.tsv lists the 12 reactions that still carry flux with all boundaries closed, each with its reason. They are ion-coupled transports whose cycles close through free plasma-membrane ion movement, and the transamination triangle MAR05391, MAR09953, MAR03899. No check reads the list yet.

Proposed change

Done in #1177:

1. Change the equations, directions and spontaneous flags in the tables above, and the GPR of the vitamin E thiolase step (HADHA and HADHB, Redefine the biomass reactions MAR13082 and MAR00021 (and review the tasks that depend on them) #1118). Support: per row, see the Support column.
2. Add MAR20221, MAR20224–MAR20228 and the metabolites MAM20094m, MAM20095c, MAM20095e, MAM20095m. No id is reused. Support: per row, see the Support column.
3. Remove the reactions in the "Reactions removed" table; the merged ones are recorded in rxnRetired of their twins. Support: per row, see the Support column.
4. Add the notes and references in the last table, and data/thermodynamics/loopExceptions.tsv. Support: per row, see the Support column.

The same oxidation-as-hydrolysis pattern is on MAR02545, MAR02560, MAR02567, MAR02571, MAR02577, and CYP27A1 side-chain steps are still written as dehydrogenations on MAR01689, MAR01691, MAR01790, MAR01792 and the cytosolic copies MAR01681, MAR01804, MAR01531, MAR01811; these are left for the bile-acid and eicosanoid curation.

Left open:

• The deprecation of the unknown CYP24A1 reaction MAR02767, and MAR02140 into MAR08006 (CYP27B1). These wait for the biomass change of Redefine the biomass reactions MAR13082 and MAR00021 (and review the tasks that depend on them) #1118.
• The ETF shuttle MAR06513 and MAR07789, the only FADH2[r] source of the epoxygenases written with FADH2. These also wait for Redefine the biomass reactions MAR13082 and MAR00021 (and review the tasks that depend on them) #1118.
• The CYP24A1 side-chain cleavage to calcitroic acid: its C4 product is not identified, so it cannot be written balanced (Curate dead-end pathways #1133).

Weakly supported changes

These changes rest on missing evidence more than on measured evidence. Each is made because the previous form describes chemistry or an enzyme that no source shows. The table gives what follows if the previous form is kept.

Counter-arguments considered

• "Blocked reactions do no harm." MAR10461 makes NADPH from indole, and the Golgi ceramidases make a trans fatty acid from an oleoyl ceramide. Both act as soon as their pathway is connected.
• "Salsolinol is made by an (R)-salsolinol synthase, so its formation is not spontaneous." Naoi 2004 and Chen 2018 report an (R)-salsolinol synthase, and RHEA:74851 and EC 3.5.99.12 exist. No human gene is known (K27544 lists a rat gene alone), human samples contain both enantiomers, and the model reactions are not stereospecific.
• "Keep MAR06709 as a spontaneous β-carboline formation." Without an aldehyde the equation describes no reaction: tryptophan gives β-carbolines by condensation with an aldehyde (Herraiz 2003).
• "Deprecating MAR06512 removes FAD uptake." Cells take up riboflavin, not FAD, and the model makes FAD from riboflavin.
• "PAPS reduction may be needed for sulfite." Humans make sulfite by cysteine and sulfide oxidation, which the model has.
• "Remove the ETF shuttle MAR07789 together with MAR00222." It is the only FADH2[r] source of the ER epoxygenase reactions written with FADH2 (MAR00944, MAR00948, MAR00956, MAR03788), and removing it blocks the eicosanoid reactions after them. Those epoxygenases should use NADPH; that rewrite waits for Redefine the biomass reactions MAR13082 and MAR00021 (and review the tasks that depend on them) #1118.
• "Deprecate the other cytosolic neolacto steps, such as MAR00907, with the LacCer duplicates." MAR00907 is one step of a cytosolic copy of the whole neolacto chain (MAR00906–MAR00909), which is the only route to its product and its exit; the chain should move to [g] as a whole.
• "New spontaneous reactions can create energy cycles." With all boundaries closed, ATP maintenance and NAD(P)H, FADH2 and ubiquinol dissipation stay 0, and no reaction joins the closed-system flux set.

Expected effect

Growth on MAR13082 and MAR00021, the essential and verification tasks, and gene essentiality should not change; the model should make FAD from riboflavin instead of taking it up.

Related

#945, #1028, #1039, #1041, #1090, #1096, #1103, #1105, #1106, #1107 (where errors came in or were left half-done); #1112 (rules for non-enzymatic reactions); #1117 (MAR00222, MAR01572); #1118 (MAR02767, MAR06513, the vitamin E thiolase steps); #1119 (the ACER3 ceramides); #1133 (dead ends); #1153 (charge balance); #1177.