feat: add reaction names from KEGG

[mihai-sysbio]

Main improvements in this PR:

This PR resolves #181 by taking 2000 reaction names from KEGG (the first supplied). The work was done fully scripted on Google Colab and also moved to the /code folder.

I hereby confirm that I have:

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

[JonathanRob]

it seems that only 2 of these reactions had names before

I compared the rxnNames fields for the model on develop and on this branch, and found more than 2 reactions whose names had been overwritten, where one case was a deletion of the name.

rxn	rxnName (original)	rxnName (new)
MAR04606	Formation of homocarnosine	L-histidine:4-aminobutanoate ligase (ADP-forming)
MAR04607	Hydrolysis of homocarnosine	alpha-Aminobutyryl histidine hydrolase
MAR05387	Acireductone Synthase	5-(methylthio)-2,3-dioxopentyl-phosphate phosphohydrolase (isomerizing)
MAR05389	Acireductone Dixoygenase 1	1,2-dihydroxy-5-(methylthio)pent-1-en-3-one:oxygen oxidoreductase (formate-forming)
MAR01940	Formation of 21-hydroxypregnenolone	pregnenolone,NADPH-hemoprotein reductase:oxygen oxidoreductase (21-hydroxylating)
MAR01941	Formation of 11-deoxycorticosterone	21-Hydroxypregnenolone:NAD+ 3-oxidoreductase
MAR02041	17-beta-estradiol 17-dehydrogenase	Estradiol-17beta:NADP+ 17-oxidoreductase
MAR01317	NAD Dependent 11-Hydroxythromboxane B2 Dehydrogenase
MAR09798	Formation of Alanylalanine	D-alanine:D-alanine ligase (ADP-forming)
MAR01044	Ferrochelatase, Mitochondrial	protoheme ferro-lyase (protoporphyrin-forming)

It would be good to double check these to ensure they make sense, as well as spot-check several other reactions with new name assignments to confirm they're OK.

[mihai-sysbio]

Interesting you found more conflicts @JonathanRob, I'm not sure how I missed them.

This conflict between curated data and database-fetched data is a recurring problem. My stance here is to not keep track of manual curations done on top of database-fetched data, so I'm going to propose a different approach.

Instead of overwriting reaction names, I can edit the script to only fill in reaction names where none existed, and push a new commit. With this approach we avoid the need to double-check the conflicts; only the spot-checking would need to be done.

[haowang-bioinfo]

Instead of overwriting reaction names, I can edit the script to only fill in reaction names where none existed, and push a new commit. With this approach we avoid the need to double-check the conflicts; only the spot-checking would need to be done.

this seems a decent solution, would be good to upload the code as well

[mihai-sysbio]

The force-pushed commit c5ec9fe overwrites the previous work. Now, only empty reaction names are filled in. The Python script is also attached. Given the changes, I'm re-requesting reviews.

[haowang-bioinfo]

@mihai-sysbio thanks for providing the script, whose output curated-Human-GEM.yml has exactly the same content as Human-GEM.yml after running.

However, the code still reports Modified reactions: 385. Not sure the reason but there should be no more modifications. Could you please check this out and maybe add more comments to the code for easier understanding and reviewing.

[mihai-sysbio]

The modified reaction increment is counting the attempts where a reaction with no name ("") is replaced with the name for the associated KEGG id from reactions.tsv. 385 indicates there are still quite a few cases with reactions that don't have a name even in KEGG. I guess I could change the output so as to more clearly indicate this, but as you note @haowang-bioinfo this wouldn't have any practical implications, since:

output curated-Human-GEM.yml has exactly the same content as Human-GEM.yml after running

My approach has been that of a one-time curation, so I am in favor of not over-engineering the script. If we are aiming for turning this into a scripted, recurring curation, I would prefer to have a different issue, since the parsing of the yml is a bit clunky anyway.

[haowang-bioinfo]

385 indicates there are still quite a few cases with reactions that don't have a name even in KEGG. I guess I could change the output so as to more clearly indicate this

That would be good. Agree to not over-engineer

[mihai-sysbio]

That would be good. Agree to not over-engineering

@haowang-bioinfo I've updated the script that now outputs (on the second run):

Reaction names adopted from KEGG: 0
Blank names also blank in KEGG: 385

The latest commit bd111e8 is force-pushed because it's now rebased on develop, making for an easy merge without needing to resolve conflicts.

[haowang-bioinfo]

I've updated the script that now outputs (on the second run)

thanks