Food Graph Sources Research

Status

Research completed 2026-04-10. SIGHI covers histamine well. Other categories need additional sources.

The Problem

The app’s 7-category system requires per-food severity data (low/medium/high) across all categories. No single source covers all 7. The SIGHI data already in the vault covers histamine-pathway mechanisms comprehensively but has gaps in salicylate, oxalate, FODMAP, lectin, glutamate, and capsaicin coverage.

Source Assessment

✅ Histamine — SIGHI (Already Have)

The SIGHI Food Compatibility List (2024-08-29 edition) is the gold standard for histamine-related food classification. Already extracted into the vault as structured markdown files.

Coverage: Histamine content, histamine liberators, DAO blockers, biogenic amines. Excellent.

Gap: SIGHI’s mechanism codes (H, H!, A, L, B) map to histamine sub-mechanisms, not to the other 6 categories. See Issue 6 SIGHI Category Mapping Gap.

🔶 FODMAP — Monash University

Monash maintains the largest laboratory-tested FODMAP database in the world (~450 foods, 900 serving sizes). They use a traffic light system (green/amber/red) with per-FODMAP-subtype breakdown (fructans, GOS, lactose, fructose, polyols).

Access problem: No public API or downloadable dataset. Data is locked inside the Monash FODMAP app ($8.99 AUD, iOS/Android). Monash explicitly states that food scanning by ingredients is insufficient because FODMAP levels depend on processing methods, not just ingredient lists. Foods must be laboratory-tested.

Workaround options:

  1. Manual extraction from the app for the top 50-100 foods relevant to the suspect food pool — labor-intensive but doable for MVP
  2. Use published FODMAP food lists from peer-reviewed papers (Monash publishes some in journals) as a lower-fidelity seed
  3. Community-sourced FODMAP lists exist but Monash warns these are often inaccurate

Recommendation: Purchase the app, manually seed the top 100 FODMAP-relevant foods for MVP. Treat as approximate (Monash’s traffic light → our low/medium/high mapping). Note: Monash’s FODMAP subcategories (fructans, GOS, lactose, fructose, sorbitol, mannitol) are more granular than our single “fodmap” category — consider whether the schema needs a subcategory field on food_category_memberships.

🔶 Salicylates, Amines, Glutamates — RPAH Elimination Diet Handbook

The Royal Prince Alfred Hospital (RPAH) handbook is the gold standard for food chemical classification — salicylates, amines, and glutamates. Also known as the FAILSAFE diet. Developed by Anne Swain, Velencia Soutter, and Robert Loblay at the RPAH Allergy Unit.

Coverage: Comprehensive food chemical charts categorizing foods as Low, Moderate, High, or Very High with per-chemical-type labels (S = salicylate, A = amine, G = glutamate).

Access: The handbook is a published book (ISBN 9780980616408, ~$30 AUD). No public API or downloadable dataset. The food charts are copyrighted.

Workaround options:

  1. Purchase the handbook and manually extract data for the top 100 foods
  2. Community-maintained lists exist (Cooking for Oscar, idealnutrition.com.au) but the RPAH handbook author warns these are often inaccurate
  3. Published research papers by Swain et al. contain some food chemical composition data

Recommendation: Purchase the handbook. Manually seed salicylate, amine, and glutamate severity for the top 100 foods. This single source covers 3 of our 7 categories.

Important note: The RPAH “amines” category overlaps significantly with SIGHI’s histamine/biogenic amines data. Need to reconcile: RPAH amines include tyramine, phenylethylamine, and other biogenic amines beyond histamine. For the app’s purposes, the SIGHI histamine data + RPAH amine data together provide comprehensive biogenic amine coverage.

✅ Oxalates — Harvard Oxalate Table

Harvard T.H. Chan School of Public Health publishes the gold-standard oxalate food composition table. Updated 2023/2024 with hundreds of foods and mg/serving measurements.

Access: Freely downloadable as an Excel spreadsheet (.xlsx) from: https://hsph.harvard.edu/wp-content/uploads/2024/07/OXALATE-TABLE-1.xlsx

Also available: oxalate.org maintains a community database cross-referencing Harvard 2008, NUTTAB 2010, USDA 1984, and other sources for 750+ foods.

Quality note: Oxalate measurement is notoriously variable — different growing conditions, soil, and measurement techniques produce different numbers for the same food. Harvard is the most trusted single source, but the data should be treated as approximate.

Mapping: Harvard uses mg/serving with categories from “little to none” through “very high.” Map to our severity enum:

  • Little to none / Low → no membership (or low if >2mg/serving)
  • Medium → medium (1)
  • High / Very High → high (2)

Recommendation: Download the Harvard .xlsx, write an import script, and seed directly. This is the easiest category to populate — the data is free, structured, and machine-readable.

🔴 Lectins — No Single Authoritative Source

Lectin data is fragmented. No equivalent of SIGHI or RPAH exists for lectins.

Available sources:

  • Steven Gundry’s “The Plant Paradox” contains a lectin food list but is commercially motivated and not peer-reviewed
  • USDA databases include some lectin content data for raw legumes
  • Published research on specific lectins (PHA in kidney beans, WGA in wheat, SBA in soybeans) exists but is per-food, not a comprehensive database

Recommendation: Hand-curate the lectin category for MVP based on established food groups known to contain lectins: legumes (raw/undercooked), nightshades (tomatoes, peppers, eggplant, potatoes), grains (wheat, corn, rice bran), and dairy (A1 casein). Use a conservative severity scale — mark only the strongest evidence foods as “high.” Note the evidence tier caveat: lectin sensitivity evidence outside of acute toxicity is weaker than for other categories.

🔴 Capsaicin — Derives from Scoville Scale

Capsaicin content roughly tracks Scoville Heat Units (SHU), which are well-documented.

Mapping:

  • Bell peppers (0 SHU) → low or no membership
  • Mild peppers (100-2,500 SHU: poblano, Anaheim) → low
  • Medium peppers (2,500-30,000 SHU: jalapeño, serrano) → medium
  • Hot peppers (30,000+ SHU: cayenne, habanero, ghost) → high

Recommendation: Seed from any standard Scoville reference. This is a small category — maybe 20-30 foods. Manual entry is trivial.

Seeding Strategy for MVP

Phase 1: Core 100 foods (covers POC needs)

Manually seed the top 100 foods most likely to appear in the suspect food pool and daily meals. Sources per category:

CategorySourceFormatCostEffort
HistamineSIGHI (in vault)Structured markdownFree (done)Import script
FODMAPMonash app + published papersManual extraction$9~4 hours
SalicylateRPAH handbookManual extraction~$30~4 hours
GlutamateRPAH handbookManual extraction(same book)~2 hours
OxalateHarvard .xlsxSpreadsheetFreeImport script
LectinHand-curated from literatureManualFree~2 hours
CapsaicinScoville referencesManualFree~30 min

Total estimated effort: ~15 hours of data entry + import script writing.

Phase 2: Expand to 300+ foods

After MVP validation, expand the food graph using the full SIGHI dataset (already extracted), full Harvard oxalate spreadsheet, and broader RPAH/Monash coverage.

Phase 3: Community contribution + verification

Allow users to flag foods with suspected category memberships. Require 3+ independent reports before adding to the graph. Cross-reference with published data where possible.

Schema Considerations

Two issues surfaced during this research:

1. FODMAP subcategories: Monash tracks 6 FODMAP subtypes (fructans, GOS, lactose, fructose, sorbitol, mannitol). Our schema has a single “fodmap” category. Consider adding a subcategory string to food_category_memberships — e.g. { food: "garlic", category: "fodmap", subcategory: "fructans", severity: "high" }. This preserves granularity without adding 6 top-level categories.

2. Mechanism field for histamine: See Issue 6 SIGHI Category Mapping Gap. Adding a mechanism enum (direct, liberator, blocker) to food_category_memberships enables the quality scorer to distinguish between histamine-containing foods and DAO-blocking foods without proliferating categories.

References