fix: normalise Title-Case metabolite (and CoA reaction) names

[edkerk]

Main improvements in this PR:

Tier 2 of #1037: normalise the Recon3D Title-Case metabolite names to the model's naming convention. Names only; model structure and balance are unchanged.

Three families, one commit each:

• acyl-CoA (105) - lowercase and Coenzyme A -> the model's -CoA, e.g. Trans,Cis-Dodeca-2,5-Dienoyl Coenzyme A -> trans,cis-dodeca-2,5-dienoyl-CoA. Stereo descriptors are preserved ((3S), (5Z), cis/trans); lowercase cis/trans matches 154 existing correct names, so no E/Z re-notation is imposed.
• acid / fibre-bile-acid complexes (34) - e.g. Lysophosphatidic Acid -> lysophosphatidic acid.
• acyl-[ACP] (3) - e.g. Cis-Tetradec-7-Enoyl-[Acyl-Carrier Protein] -> cis-tetradec-7-enoyl-[ACP], matching the existing (2E)-decenoyl-[ACP].

142 metabolite names changed; the reaction names that embed them are updated in sync (303 name lines total). The reaction names were edited surgically rather than by re-running the name generator, so existing reaction-name curation is untouched.

A fourth commit normalises the 138 Title-Case reaction names that use "Coenzyme A" as a generic term (e.g. Stearoyl Coenzyme A Desaturase -> stearoyl-CoA desaturase), preserving shorthand like N-C18:0CoA. After this, no "Coenzyme A" string remains anywhere in the model.

A fifth commit resolves #229: 10 bile-acid metabolites carried a semicolon-separated list of names as their name (a Recon3D/HMDB import artefact). Each is replaced with a single clean lowercase name (e.g. 12-Dehydrocholic acid; 12-Oxodeoxycholic acid; ... -> 12-dehydrocholic acid), fixing a typo (Chenodeoxychollyc -> chenodeoxycholic) and a wrong first name (MAM03704 is the trihydroxy isocholic acid, not the dihydroxy isochenodeoxycholic acid). Metabolite names only.

Scope note. The broader ~480 remaining Title-Case metabolite names are mostly standard glycan / glycolipid nomenclature (GalNAc4S-GlcA, IV2Fuc-Lc4Cer, Roman numerals) whose capitalisation is correct and must stay, so they are deliberately left alone.

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

[github-actions]

Model quality report

✅ Model QC checks fine, no changes compared to develop.

Model QC checks

Check	Result	Δ vs develop
Growth (biomass producible)	125	0	✅
Metabolites missing formula	7	0	✅
Metabolites missing charge	0	0	✅
Reaction bound / GPR issues	0	0	✅
Malformed cross-references	59	0	✅
Cross-refs inconsistent across compartments	59	0	✅
MEMOTE score (%)	pending		⏳

MACAW and mass/charge balance

MACAW: dead-end and duplicate tests

Mass and charge balance

(balance report unavailable)

Gene essentiality (Hart 2015)

cellLine	TP	TN	FP	FN	accuracy	sensitivity	specificity	F1	MCC
DLD1	38	2158	60	280	0.8659	0.1195	0.9729	0.1827	0.1588
GBM	34	2137	64	300	0.8564	0.1018	0.9709	0.1574	0.1276
HCT116	47	2181	55	308	0.8599	0.1324	0.9754	0.2057	0.1906
HELA	32	2241	70	250	0.8766	0.1135	0.9697	0.1667	0.1332
RPE1	15	2179	83	258	0.8655	0.05495	0.9633	0.08086	0.02935
all	7	2379	95	112	0.9202	0.05882	0.9616	0.06335	0.02199

Per-finding detail is committed to data/testResults/ (qc_metabolite_completeness.csv, qc_reaction_sanity.csv, qc_annotation_issues.csv, macaw_results.csv, balance_results.csv, gene-essential.csv); the full MEMOTE result is uploaded as a build artifact.

Results for commit 1569eee.

Full workflow run · this comment is edited on subsequent runs