Update ATP metabolism

[JonathanRob]

Main improvements in this PR:

A number of updates were made to ATP metabolism (primarily the electron transport chain) in HumanGEM to improve the biological accuracy. These updates were made to address Issues #98, #99, #100, #101, and #102.

The model updates were implemented using the new script curateATPmetabolism, and are summarized below:

Remove duplicate reaction
A duplicate proton transport reaction H+[i] => H+[m] was identified during curation (reactions HMR_7638 and Htmi). The Htmi reaction was associated to HMR_7638 in the humanGEMRxnAssoc.JSON annotation file, and subsequently removed from the model.

Update the stoichiometry of ATP synthase
Related to Issue #98. The stoichiometry for protons in ATP synthase HMR_691 were updated to 3 instead of 4. In addition, the reaction was re-balanced by removing one additional proton from the products:
Original: ADP[m] + Pi[m] + 4 H+[i] => ATP[m] + 4 H+[m] + H2O[m]
Revised: ADP[m] + Pi[m] + 3 H+[i] => ATP[m] + 2 H+[m] + H2O[m]

In addition, a permeability transition pore (PTP) reaction was added to move H+[c] -> H+[m] to account for the proton that is lost through the adenine nucleotide transporter (ATP [m] -> ATP[c]).

Prevent free transport of Pi into the mitochondria
Related to Issue #99. Phosphate (Pi) should not be freely transported into the mitochondria ([c] -> [m]), but should instead require expenditure of proton motive force via co-transport of a proton (H+[i] -> H+[m]). Several reactions enabling free Pi transport were made irreversible (only allowing [m] -> [c] direction) to prevent such activity:

Reaction	Equation
HMR_3971	fumarate[m] + Pi[c] <=> fumarate[c] + Pi[m]
HMR_4862	Pi[c] + succinate[m] <=> Pi[m] + succinate[c]
HMR_4865	malate[m] + Pi[c] <=> malate[c] + Pi[m]
HMR_4870	malonate[m] + Pi[c] <=> malonate[c] + Pi[m]
HMR_4940	GSH[c] + Pi[m] <=> GSH[m] + Pi[c]
HMR_6330	AKG[c] + Pi[m] <=> AKG[m] + Pi[c]
HMR_6331	oxalate[m] + Pi[c] <=> oxalate[c] + Pi[m]

ATP synthesis from propanoate fermentation
Related to Issue #100. In the presence of glucose and absence of oxygen the model synthesized propanoate instead of lactate with an ATP yield of ~5 ATP/glucose instead of 2. This was caused by the reversibility of the following reactions:
HMR_0153: AMP[c] + PPi[c] + propanoyl-CoA[c] <=> ATP[c] + CoA[c] + propanoate[c]
HMR_4459: ATP[c] + H+[c] + propanoate[c] <=> PPi[c] + propinol adenylate[c]

Both reactions were made irreversible, where HMR_0153 was turned around to proceed in the backward direction (consuming propanoate).

Flux through complex I under anaerobic conditions
Related to Issue #101. In the absence of oxygen and presence of glucose the model was able to carry flux through complex I and synthesizes 3-Methyl-Glutaconate and 2-Methyl-3-Hydroxy-Valerate. This resulted in ATP yield > 4 per glucose instead of 2. This was resolved by making RE1519X and HMR_3212 irreversible:
RE1519X: 4-cis-decenoyl-CoA[p] + FAD[p] <=> 2-trans-4-cis-decadienoyl-CoA[p] + FADH2[p]
HMR_3212: FAD[m] + propanoyl-CoA[m] <=> acrylyl-CoA[m] + FADH2[m]

ATP and carbon from Pi and O2
Related to Issue #102. The model was able to produce infinite ATP and CO2 from Pi and O2. This was resolved by inactivating the pool reaction HMR_0686.

Metabolic task list for model verification

The above fixes were verified with four new metabolic tasks. These metabolic tasks were added to the existing set of "leak check" tasks introduced with PR #69, initiating a combined document metabolicTasks_VerifyModel.xls that can be used to accumulate additional checks that should be performed on the model.

Updates to reaction change documentation functions

The functions docRxnChanges and writeRxnChanges that are used to document model changes were updated and now document changes to the grRules field.

Removal of the rxnRecon3DID model field

PR #107 updated the way in which external identifiers (for reactions and metabolites) are stored. Therefore, the rxnRecon3DID model field is outdated and no longer necessary to keep in the HumanGEM model structure, and is removed here.

I hereby confirm that I have:

• Tested my code on my own computer for running the model
• Selected devel as a target branch

[haowang-bioinfo]

@JonathanRob this PR involves additional rxn association between HMR2 and Recon3D, and probably should be implemented after #107?

[JonathanRob]

@Hao-Chalmers Do you mean this PR should be implemented after PR #120? If so, then I agree, and will update this PR to be consistent with the changes made in #120.

[haowang-bioinfo]

@JonathanRob yes, I mean after #120.

[haowang-bioinfo]

The modifications are fine, here are some minor comments:

1. It was confirmed that model's rxnRecon3DID field has identical information to that in humanGEMRxnAssoc.JSON file. Would be good to explicitely note this in the comments (i.e. line 33 of curateATPmetabolism.m)
2. In curateATPmetabolism.m, Htmi is intended to be removed and HMR_10025 added. How about only renaming the former rxn id to the latter and replacing H+[I] with H+[c]? That could save the code in lines 59-60 and 85-102.
3. This curation script aims to fix a bunch of issues that had been formatted into the metabolicTasks_VerifyModel.xls as metabolic tasks. It might be good to append a verification section to the script for showing the successful achieving of those tasks.

[JonathanRob]

@Hao-Chalmers Thanks for the suggestions. I have implemented some changes based on your ideas, which I liked.

1. It was confirmed that model's rxnRecon3DID field has identical information to that in humanGEMRxnAssoc.JSON file. Would be good to explicitely note this in the comments (i.e. line 33 of curateATPmetabolism.m)

This is now stated in the comment of the script.

2. In curateATPmetabolism.m, Htmi is intended to be removed and HMR_10025 added. How about only renaming the former rxn id to the latter and replacing H+[I] with H+[c]? That could save the code in line 59-60 and 85-102.

Good point. I realized that the original form of HMR_7638, which is currently identical to Htmi, actually involved proton transport from [c] to [m]. Therefore, instead of removing Htmi and creating a the new HMR_10025 reaction, I have now simplified the process to just revert the equation of HMR_7638 back to it's original form: H+[c] => H+[m]. Now, the script does not require the deletion or addition of any reactions.

3. This curation script aims to fix a bunch of issues that had been formatted into the metabolicTasks_VerifyModel.xls as metabolic tasks. It might be good to append a verification section to the script for showing the successful achieving of those tasks.

An optional section of code has been added to load and run the model verification tasks, to demonstrate improved model performance after the changes have been implemented.