orthonym.data.coordination_retained#

Note

Internal API. Names and behaviour may change between releases.

Retained-name table for N-coordinated metal-tetrapyrrole macrocycles.

Milestone D1. These are Fe/Mg/Co/Ni tetrapyrrole coordination complexes

(heme, chlorophyll, cobalamin, siroheme, coenzyme F430). OPSIN 2.9.0 cannot parse any of their names (measured: heme b / chlorophyll a / cyanocobalamin / siroheme / F430 all -> opsin_parse None), so a round-trip oracle is impossible for this class. The 0-wrong guarantee here is therefore CURATION PROVENANCE + EXACT-InChIKey identity, exactly the contract already used by the amino-acid / sugar / metallocene retained tables:

  • Each key is the standard InChIKey of one exact ChEBI structure.

  • Each value is that structure’s ChEBI-accepted IUPAC/name annotation, taken VERBATIM from the ChEBI data (chebi_iupac_filtered.tsv / ChEBI id in the per-entry comment) – never guessed, hand-derived, or approximated.

  • The lookup fires ONLY on an exact InChIKey match, and ONLY on the descriptive-fallback path (namer._descriptive_fallback), i.e. after every real namer has already declined. A miss stays abstaining (still 0-wrong); a molecule the normal namer names never reaches this table.

Charge / protonation states are enumerated as SEPARATE keys (ChEBI stores one row per charge state; standard InChI does not renormalise formal charge), so each key is an exact, unambiguous structural match.

Scope (-D1 plan IN-list): heme b/c/o, siroheme, chlorophyll/bacteriochlorophyll family, the cobalamin group (cyano/adenosyl/methyl/aqua/hydroxo/nitrito/cob(I..III)), and coenzyme F430. -D1 additionally added the 13 cobyrinic-acid-type corrinoid / cobalt-precorrin rows (Co corrin/precorrin, no nucleotide loop) that each passed a strict per-row provenance check (verbatim ChEBI name, InChIKey re-derived == key, distinct id, opsin_parse == None); see internal notes SP5.1 addendum. Still OUT (left abstaining): chlorophyllide/protochlorophyllide precursors, Zn/Cu porphyrin dyes, and the general additive namer (D2). Rows whose ChEBI name is malformed (unsigned charge, empty parentheses, stray double-hyphen) or whose structure has no standard InChIKey (dative-bond SMILES) are also left out – they stay abstaining.