feat: curate metabolite annotations, add mitochondrial acetate transport and RNDR4

[edkerk]

Main improvements in this PR:

• Fixes:
  • Correct erroneous metabolite annotations on 22 metabolites, where distinct metabolites shared identifiers, or carried the identifier of an entirely different compound. All replacement identifiers were verified against the ChEBI, KEGG, BiGG and MetaNetX web services (closes ~6 incorrect annotations #378).
• Features:
  • Acetate transport across the mitochondrial membrane, r_4785 (closes feat: add Acetate transport to the mitochondria #388).
  • Ribonucleotide reductase acting on UDP, cytosolic r_4786 and nuclear r_4787 (closes feat: add RNDR4 reaction #380).

Curations are consolidated in code/modelCuration/v9_2_0.m. The new code/modelCuration/removeMiriam.m complements RAVEN's editMiriam, which can add, fill or replace annotations, but not delete them.

Notable annotation corrections (#378)

Metabolite	Was	Now
s_4313 ITP[m]	a complete set of succinate annotations (succ, SUC, MNXM25, cpd00036, CID 160419)	aligned with the already correct ITP[c] (s_0950), plus MetaCyc and ModelSEED identifiers for ITP
s_0558/s_0559/s_3780 D-galactose, s_3862 alpha-D-galactose	MNXM390, which is L-galactose	MNXM735266 and MNXM1364486
s_3717 protein	bigg:protein (= Torasemide-M3), C00492 and MNXM621 (= raffinose)	removed; the pseudometabolite keeps only its SBO term
s_3718 carbohydrate	C05402 (= melibiose), MNXM1434 (= polydextrose)	removed
s_4180 L-xylulose	ChEBI and MetaNetX identifiers of D-xylulose	CHEBI:17399, MNXM1371095
s_3768/s_3852 ribose 5-phosphate forms	shared C00117 and MNXM722712	MNXM1363910 and MNXM1363911; the unspecific C00117 is dropped from both
s_0331/s_3879 GlcNAc-PI	shared CHEBI:57265, which is the (1-) form	CHEBI:12194 for the neutral s_0331, MetaNetX identifiers split
s_3807/s_3837/s_4176 methionine sulfoxides	swapped BiGG identifiers, both MetaNetX identifiers resolving to the (S)-epimer	metsox_S__L for the (S)-epimer, metox for the unspecific one, MNXM1364200
s_1473/s_1474 and s_4267/s_4268	names of taurocholate and taurocholic acid were swapped relative to their formula, charge and ChEBI	renamed to match

Note that BiGG (tchola) and KEGG (C05122) do not distinguish protonation states, so those identifiers legitimately remain shared. Several of the pairs above are genuine duplicate metabolites, which remains tracked in #338. MetaNetX identifiers that are merely deprecated but still refer to the correct compound are left untouched, as their bulk update is tracked in #245.

Mitochondrial acetate transport (#388)

Mitochondrial acetate is produced by Ach1 (r_0111) and the mitochondrial aldehyde dehydrogenases (r_0174, r_0175) and consumed by Acs1 (r_0113), but cytosolic and mitochondrial acetate could only be connected via acetaldehyde diffusion.

r_4785 is irreversible in the export direction. Allowing import by default lets cytosolic acetate feed mitochondrial acetyl-CoA via Acs1, which shifts the anaerobic redox balance and degrades the predictions that release 9.1.0 was curated for: glycerol drops from 4.35 to 3.81 (experimental 4.5 +/- 0.4), the mean relative error of the main fermentation products rises from 0.065 to 0.103, and the anaerobic flux prediction R² drops from 0.9967 to 0.9950. Acetate import is reported in the context of acetate detoxification, which is a stress response rather than a feature of growth on defined medium, so a rxnNotes entry documents that the lower bound should be set to -1000 when explicitly simulating acetate stress.

Validation

Test	v9.1.0	this PR
Gene essentiality accuracy	0.9024	0.9024
Growth prediction R2 (growth)	0.9011	0.9019
Anaerobic flux prediction R2 (anaerobic_flux_predictions)	0.9967	0.9970
Glycerol / ethanol / CO2 / biomass at qGlc = 23	4.353 / 35.707 / 38.774 / 0.3792	identical
Mean relative error of main fermentation products	0.0645	0.0645
Ammonium exchange / ATPase ratio	1.072	1.072

The three new reactions are elementally and charge balanced, and no energy- or redox-generating cycle is introduced: with all uptake blocked, maximum ATP and NADH production remain 0. Blocking r_4785 reproduces every v9.1.0 metric exactly, confirming that the annotation corrections and both ribonucleotide reductases are flux-neutral.

model/yeast-GEM.mat and model/yeast-GEM.xlsx are not regenerated, as those are written at release time.

I hereby confirm that I have:

• Selected develop as a target branch (top left drop-down menu)