Methodology RFC: terpene descriptor map refinement

[bpevreall]

Status: open for community input
Related: worker/migrations/0002_reference_data.sql, worker/src/lib/inference.ts

Context

Phenologue v0.1 ships an organoleptic descriptor → terpene weight mapping (terpene_descriptor_map table) for inferring chemotype profiles when a lab COA isn't available. The mappings are documented in methodology §7.2 and the migration file.

Three mappings were removed in migration 0003 because they were not supported by peer-reviewed evidence:

• skunk → caryophyllene (removed, refuted by Oswald et al. 2021, ACS Omega, doi:10.1021/acsomega.1c04196 — skunk smell is volatile sulfur compounds, not terpenes)
• cheese → caryophyllene (removed, no terpene-specific peer-reviewed support)
• diesel → caryophyllene (removed, no clean single-terpene mapping in literature)

The current map covers: citrus, pine, floral, fruit, sweet, earth, mint, hops, plus sharp herbal. See migration 0002 for the full list.

Goal of this RFC

Refine the descriptor → terpene mapping with peer-reviewed evidence as it accumulates. Specifically:

1. VSC-class descriptors — once methodology v0.2 supports a non-terpene volatile category, re-add skunk and diesel mapped to a vsc (volatile sulfur compound) bucket rather than to terpenes.
2. Cheese-family — investigate butyric-acid family volatiles for a mapping path; needs a peer-reviewed source for cannabis specifically.
3. Cultural / regional descriptors — gas, loud, haze are common in patient vocabulary but not in our controlled list. Should they be added? With what mapping?
4. Confidence weights — current weights (0.4–1.0) are eyeballed. Could they be calibrated against measured COA data once we have enough chemotype-tagged sessions?
5. Synergy mappings — many descriptors are multi-terpene (e.g. ""tropical"" = myrcene + esters + minor terpenes). Should the schema support multi-terpene weights per descriptor?

How to contribute

If you have a specific mapping correction backed by peer-reviewed evidence, comment with:

• Descriptor (existing or proposed)
• Terpene (or VSC class)
• Suggested weight
• Citation (DOI / PMID)

Folk evidence is welcome as a discussion starter, but only peer-reviewed citations land in the migration.