Iyengar Flavor Index
A unified quantitative framework for flavor engineering across eight receptor systems — heat, sweetness, umami, bitterness, kokumi, cool-burn, TRPA1 pungency, and astringency.
The Formula
Every IFI score — across all eight subscales — is computed by the same formula:
IFI(c, m) = 1,000 × EC₅₀(ref) / EC₅₀(c)
c = compound or construct being evaluated | m = modality subscale | ref = reference compound
EC₅₀(ref)
The half-maximal effective concentration of the reference compound at the relevant receptor. Defined once per subscale; never changes.
EC₅₀(c)
The half-maximal effective concentration of the compound or construct being evaluated. Derived from receptor binding models or flux balance predictions.
Scale anchor
IFI = 1,000 is defined as equivalent potency to the reference compound. Values above 1,000 are stronger; values below are weaker. A value of 0 indicates no detectable receptor activity.
The Eight Receptor Axes
Each subscale targets a distinct receptor system. The eight axes are independent — a compound can score high on IFI-H and zero on IFI-S, or engineer all eight simultaneously.
TRPV1-mediated thermal nociception. The baseline for all capsaicinoid engineering. A value of 1,000 matches pure capsaicin's potency; higher values indicate a more potent TRPV1 agonist.
Sweetness is mediated by the TAS1R2/TAS1R3 heterodimer. IFI-S captures relative activation of this complex compared to a 34.2 mM sucrose solution (approximately 1% sucrose by mass).
Umami perception involves both the TAS1R1/TAS1R3 heterodimer and metabotropic glutamate receptor 1 (mGluR1). IFI-U normalizes against MSG (monosodium glutamate) at 820 µM — approximately the concentration in a standard light seasoning.
IFI-B is an inverse scale: it measures suppression of TAS2R38 relative to the baseline agonist phenylthiocarbamide (PTC) at 1.2 µM. An IFI-B of 1,000 represents complete suppression equivalent to baseline PTC sensitivity; higher values represent greater bitter reduction.
The calcium-sensing receptor (CaSR) mediates kokumi — the mouthful, lingering richness associated with γ-glutamyl peptides. IFI-K normalizes against γ-Glu-Val-Gly at 48 nM, the most potent known natural kokumi peptide.
Biphasic thermal sensation requiring co-activation of TRPM8 (cooling) and TRPV1 (heat). The reference is a menthol:capsaicin composite at 78 µM designed to represent the canonical cool-burn experience. The TRPM8/TRPV1 ratio within the composite is 2.5:1.
TRPA1 mediates nasal-clearing pungency distinct from capsaicin heat — the sharp, electrophilic character of mustard, wasabi, and horseradish. IFI-P normalizes against allyl isothiocyanate (the active compound in mustard oil) at 56 µM.
Astringency is not a receptor-mediated signal in the strict sense: it results from proanthocyanidin-salivary protein cross-linking that reduces lubrication. IFI-A normalizes against tannic acid at 1,100 µg/mL — the approximate astringency threshold of strong black tea.
Reference Compounds and EC₅₀ Values
These values are constants in the IFI formula — they never change between experiments.
| Subscale | Receptor(s) | Reference Compound | EC₅₀ |
|---|---|---|---|
| IFI-H | TRPV1 | Capsaicin | 3.16 nM |
| IFI-S | TAS1R2 + TAS1R3 | Sucrose | 34.2 mM |
| IFI-U | TAS1R1 + mGluR1 | MSG | 820 µM |
| IFI-B | TAS2R38 (suppression) | PTC | 1.2 µM |
| IFI-K | CaSR | γ-Glu-Val-Gly | 48 nM |
| IFI-C | TRPM8 + TRPV1 | Menthol + Capsaicin composite | 78 µM composite |
| IFI-P | TRPA1 | Allyl isothiocyanate (AITC) | 56 µM |
| IFI-A | Proanthocyanidin-salivary protein axis | Tannic acid | 1,100 µg/mL |
Computational Prediction Methods
IFI scores in the Scoville Splice catalog are design targets, not measured values. They are derived from two complementary computational methods:
Receptor Binding Models
Molecular docking simulations predict EC₅₀ by computing the binding affinity of candidate compounds to each receptor’s active site. Structures are derived from published crystal structures (TRPV1: PDB 3J5P; TAS1R2/3: PDB 7JYZ; TRPA1: PDB 3J9P). Docking uses AutoDock Vina with Gnina scoring function refinement. Predicted EC₅₀ values carry a typical uncertainty of ±0.5 log units (3× overestimate to 3× underestimate).
Pathway Flux Balance Analysis
For engineered constructs, flux balance analysis (FBA) predicts the final concentration of capsaicinoids or flavor compounds in plant tissue under the modeled metabolic state. The capsaicinoid network model covers 67 reactions and 58 metabolites. FBA predictions are converted to expected tissue concentration, then EC₅₀ is estimated from concentration-response curves.
All IFI scores are computational design targets. They represent expected potency based on receptor modeling and pathway flux predictions. None have been validated by in vitro binding assays or in vivo taste panel studies. Experimental confirmation is planned and will replace computational estimates as data becomes available.
Limitations
Computational, not measured
IFI scores are predictions. Computational binding models have real-world accuracy constraints. The scale is internally consistent — an IFI-H of 2,000 is predicted to be twice as potent as 1,000 — but absolute values should be treated as engineering estimates until confirmed by assay.
Receptor models for humans
IFI is calibrated to human receptor variants. In other organisms — particularly plants used as engineering hosts — receptor orthologs may not exist or may differ functionally. The IFI value describes expected human perception of the final product, not host organism biology.
No cross-modality interactions
The eight subscales are treated as independent. In reality, some interactions exist: sweetness suppresses bitterness, certain kokumi compounds enhance umami, and capsaicin partially desensitizes other TRPV-family channels. Cross-modality interaction modeling is not included in the current IFI version.
IFI-A is mechanism-different
Astringency (IFI-A) is the only subscale not mediated by a discrete receptor binding event. The proanthocyanidin-salivary protein cross-linking mechanism means IFI-A predictions are based on compound concentration thresholds rather than receptor EC₅₀ curves, and carry higher uncertainty than the other seven subscales.
IFI vs Brix vs Scoville
| Property | IFI | Scoville (SHU) | Brix (°Bx) |
|---|---|---|---|
| What it measures | Receptor activation (8 axes) | TRPV1 heat only | Total dissolved sugars |
| Method | Receptor binding model / FBA | HPLC capsaicinoid analysis | Refractometer |
| Scale type | Normalized (1,000 = reference) | Linear dilution ratio | % sucrose equivalent |
| Cross-modality | Yes — all 8 axes unified | No — heat only | No — sweetness only |
| Design use | Engineering target per axis | Potency benchmark (heat) | Ripeness/sweetness QC |
| Validated by assay | Not yet — computational | Yes — HPLC is standardized | Yes — refractometry |
Further Reading