MIM #114 follow-ups: media→DEFINED_MEDIUM reclassification, Carnitine dedup decision, peptone prose

[realmarcin]

The final MIM #114 follow-ups. No ontology mappings changed (only ingredient_type + curation_history prose); reconcile in-sync, invariants A/B1/B2/B3 pass.

Media-vs-ingredient modeling

The schema already has DEFINED_MEDIUM ("complete medium formulation/recipe, e.g. LB broth") — so no schema change is needed; the commercial-media records just weren't all using it. Reclassified 53 media (broth/agar/medium-named) from UNDEFINED_MIXTURE → DEFINED_MEDIUM (that type is for ingredient-level mixtures like yeast extract, not whole media). 41 media were already correctly DEFINED_MEDIUM; the 10 media-named STOCK_SOLUTION records are correctly typed (they're stock components/bases for a medium — "Macro Component 1 for J Medium", "UW concentrated base", etc.). Precedent exists for mapping specific media to NCIT growth-medium terms (Middlebrook_7H10_Agar → NCIT:C85509) — a further optional enhancement, not done here.

Carnitine Hydrochloride dedup — decision: NOT a duplicate

Both Carnitine Hydrochloride and Carnitine (Dl) Hydrochloride are 0-occurrence and map to CHEBI:17126, but they're distinct name surface-forms (stereo-unspecified, from CultureBotHT, vs racemic "DL"), both correctly grounding to the stereo-unspecified parent. MIM keeps one record per surface-form, so kept both; documented with a DEDUP_REVIEWED curation_history entry (per-ingredient + aggregate).

Cosmetic

Fixed the 3 soy-peptone curation_history prose that said "EXACT_MATCH" (their mapping_quality fields were already correctly CLOSE_MATCH).

🤖 Generated with Claude Code

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[realmarcin]

Stray 4th prose edit, out of scope and now inconsistent. The PR scope is the 3 soy-peptone records (Phytone/Soya_Pepton/Soya_Peptone), whose mapping_quality is CLOSE_MATCH — so flipping their prose EXACT_MATCH→CLOSE_MATCH correctly aligns prose with the field. This 4th change is a different ingredient: Tryptone peptone (identifier MICRO:0000182, UNMAPPED_0558), whose mapping_quality is EXACT_MATCH (unchanged here, line ~140203). Flipping only the prose to CLOSE_MATCH therefore makes the aggregate self-contradictory (prose says CLOSE_MATCH, field says EXACT_MATCH), and it diverges from the per-ingredient file data/ingredients/mapped/Tryptone_Peptone.yaml, which still reads EXACT_MATCH in both its mapping_quality and its curation_history prose. The SSSOM reconcile/invariant gates derive the predicate from mapping_quality (not this prose), so they pass and do not catch the drift. Recommend reverting this line to EXACT_MATCH to restore aggregate↔per-ingredient consistency (tryptone→tryptone is a genuine exact match; changing it to CLOSE_MATCH would be a mapping change requiring the field + predicate + per-ingredient file to move too, which is out of scope for this cosmetic-prose PR).

[realmarcin]

Review (data-curation): mostly clean, one confirmed issue flagged inline.

Confirmed issue (1) — stray 4th prose edit in `data/curated/mapped_ingredients.yaml`: the cosmetic EXACT_MATCH→CLOSE_MATCH fix was also applied to `Tryptone peptone` (MICRO:0000182, UNMAPPED_0558), whose `mapping_quality` is `EXACT_MATCH`. That leaves the aggregate self-contradictory (prose CLOSE_MATCH vs field EXACT_MATCH) and out of sync with the untouched per-ingredient `Tryptone_Peptone.yaml` (still EXACT_MATCH in both field and prose). The SSSOM gates key off `mapping_quality`, not this prose, so they pass and dont catch it. Suggest reverting that one line to EXACT_MATCH. Details inline.

Verified clean:

• 53 media reclassifications: all names are genuine media (broth/agar/medium/base). Each record adds one RECLASSIFY_INGREDIENT_TYPE entry and flips UNDEFINED_MIXTURE→DEFINED_MEDIUM; nothing else changes (still UNMAPPED, mapping untouched, notes intact). No single-ingredient / stock-component / ingredient-mixture was wrongly swept — the "…Agar Base" records are medium bases (defined formulations), not stock solutions.
• Carnitine DEDUP_REVIEWED: present and byte-identical (modulo list indentation) in both `Carnitine_Hydrochloride.yaml` and the aggregate; no identifier/predicate/mapping change.
• 3 soy-peptones (Phytone / Soya_Pepton / Soya_Peptone): prose now correctly matches `mapping_quality: CLOSE_MATCH` in both per-ingredient and aggregate; no residual EXACT_MATCH; SSSOM predicate (closeMatch) unchanged.
• No ontology-mapping changes anywhere — only `ingredient_type` + curation_history prose moved. No stray files (58 files, all data YAMLs). All spot-checked YAML parses.