Fix: the Marvin 26.1 protonation formula column was dropping R groups

[freiburgermsu]

Follow-up to #295, which is already merged. Review of it found compounds losing their R groups — confirmed, and the cause was that PR's formula column, not Marvin and not the structures.

What was wrong

#295 wrote Marvin's own getFormula() into the formula column and deferred Print_Structure_Formula_Charge.py to a follow-up step. Marvin omits wildcard atoms from a formula; this repository renders them as R, in one line of that script:

formula = re.sub(r'\*', 'R', formula)

SMILE rows whose formula contains an R group therefore fell from 23.4's 8,704 to zero. Stearoyl-ACPs went from C32H60N3O9PR2S to C32H60N3O9PS, and Update_Compound_Structures_Formulas_Charge.py carried that into 6,052 compound records — cpd00049 "carboxylic acid" became CHO2, a generic compound quietly ceasing to be generic.

A counting note, because it changes the headline. Count R groups with R(?![a-z]), not "R" in formula — the elements Ru, Rb, Rh, Re and Rn match the naive test and inflate every count here by 8. The first report of this regression (mine, and the review that caught it) said 8,712 → 8. The true figures are 8,704 → 0: #295 did not leave eight R groups standing, it left none, and the apparent survivors were ruthenium and rubidium compounds.

The structures were never affected

8,727 SMILE structures carry a * in both bundles, identically, and this fix changes 0 structures. Nothing was ever stripped from a molecule.

That is also why it shipped: coverage, compound sets, InChI agreement, InChIKey consistency and stereochemical fidelity all inspect structures, and the structures were always right. A column-only defect was invisible to every check in the suite. A formula-column comparison against 23.4 is now part of the report.

The fix

formula and charge now come from Print_Structure_Formula_Charge.parse_structure — this repository's own function, imported rather than reimplemented, computed per row from that row's structure string. Deferring it was the error: the cascade consumes this column directly, so the bundle has to be correct as written rather than correct after a second script runs.

Two further cases had to be reconciled to make that hold everywhere:

• The substitution only fires on the RDKit path. RDKit renders a dummy atom as *; OpenBabel and Marvin both omit them. So ISOCITHASE-P's *OP(=O)(=O)=O — an invalid valence only OpenBabel will read — still lost its R. The convention is now enforced as an invariant: a SMILE structure carrying * gets an R in its formula. This bundle violates it 0 times; 23.4 violates it 23 times.
• Counting wildcards has its own trap. A molblock R atom — how RDKit writes every dummy atom, and therefore how every structure arrives through the import bridge — reads back from Marvin as symbol R#, not R. A wildcard set without R# counts zero on a molecule that plainly has them.

Result

	this PR	23.4	as merged in #295
SMILE rows with an R group	8,727	8,704	0
ChEBI / KEGG / Rhea	2,071 / 957 / 30	2,071 / 957 / 30	—
MetaCyc	5,669	5,646	—
formula-column rows matching 23.4	76,479	—	68,466
invariant violations (SMILE)	0	23	—

Reading the diff

It looks total — ~141,900 insertions — because the bundle was regenerated and generated_on moved on every row. The content change is narrow:

	rows
content identical (only generated_on differs)	132,528
formula changed	9,185
charge changed	84
structure changed	0
rows added or removed	0

Re-verified against the committed bundles: coverage is 141,715 rows and 53,127 compounds, both exactly 23.4's; 0 non-standard InChI; 0 InChIKeys disagreeing with the InChI beside them; one generated_on, tool_version and ph across all four files; and re-running a source rewrites it byte for byte.

---

Generated by Claude Code