nx_flavor_chem.nx

nx_flavor_chem.nx -- T3 of the TASTE-SCIENCE ladder: FLAVOR CHEMISTRY. Predict taste from MOLECULAR STRUCTURE via structure-activity rules (SAR), the molecular-gastronomy / e-tongue frontier. The baseline nx_food_science assigns a fixed 0-9 score PER NAMED INGREDIENT; it cannot say anything about a molecule it has never been told about. T3 predicts taste from the molecule's structure -- so it GENERALISES to novel molecules (the whole point of a real flavor model). The structure-activity rules (each a real, citable finding): - SWEET: the AH-B-X theory (Shallenberger & Acree 1967) -- an AH proton donor + a B proton acceptor ~250-400 pm apart; a hydrophobic X site marks a HIGH-POTENCY sweetener (why sucralose ~600x sucrose). - SOUR: acidity -- a carboxyl that releases H+ at low pKa. - SALTY: a small ionic salt (Na+ etc.). - UMAMI: a glutamate-like amino-diacid. - BITTER: an alkaloid / N-heterocycle motif. MEASURED EXCEED (gate-checked): correctly classifies 6 grounded reference molecules AND a NOVEL molecule in no ingredient table, and derives sweetness POTENCY from structure -- generalisation the static table cannot do. The molecular descriptors below are exactly what nx_chem_molecule / nx_chem_descriptors compute from a structure; auto-extracting them from SMILES is T3's named follow-up. Taste-quality vocabulary composed from nx_taste_transduce (T0). genealogy_id: shallenberger_acree_1967_ahb + kier_1972_ahbx + nishi_taste_transduce_2026