1Overview
How NUTRON handles nutritional data
NUTRON is a nutrition intelligence platform that aggregates, harmonises, and presents compositional data from multiple peer-reviewed, government-maintained food databases. Our goal is to surface the most accurate, scientifically grounded nutritional picture possible — and to present it to you transparently.
We do not invent nutritional values. Every figure displayed on NUTRON is derived from one or more of the source databases listed in this document, processed through our harmonisation pipeline, and scored using the FoodCompass 2.0 algorithm developed by researchers at Tufts University.
NUTRON is an informational platform. Nothing presented on this platform constitutes medical advice, dietary guidance, clinical nutrition recommendations, or a substitute for consultation with a qualified healthcare professional. Nutritional science is complex, individual metabolism varies significantly, and no scoring system captures the full picture of diet and health. Always consult a registered dietitian, nutritionist, or physician before making significant changes to your diet, particularly if you have a health condition, allergy, or are pregnant or breastfeeding.
2Data Sources
Five databases. One harmonised picture.
NUTRON draws on five primary food composition databases (including USDA FoodData Central), each maintained by government agencies or academic research institutions. These databases were selected for their scientific rigour, public availability, and complementary coverage of nutritional attributes — particularly micronutrients and phytochemicals that are underrepresented in single-source systems.
Danish Food Composition Database
~1,341 foods · 223 nutrients · Maintained by DTU Fødevareinstituttet
Denmark's national food composition database, maintained by the National Food Institute at the Technical University of Denmark (DTU). FRIDA serves as NUTRON's primary reference database for nutrient column naming conventions and structural harmonisation. It provides comprehensive coverage of vitamins, minerals, fatty acids, and amino acids for Scandinavian and European food items.
McCance & Widdowson's (CoF)
~6,981 foods · 284 nutrients · Public Health England / FSA
One of the world's most comprehensive food composition datasets, published by Public Health England and the Food Standards Agency. NUTRON's largest source by food count, it provides particularly strong coverage of fatty acid profiles, carotenoids, and vitamin subtypes. The UK dataset is standardised to align with FRIDA column definitions within NUTRON's pipeline.
Swiss Food Composition Database
~1,190 foods · 49 nutrients · Federal Food Safety and Veterinary Office
Published by the Swiss Federal Food Safety and Veterinary Office (FSVO), this database contributes European-specific food items and provides a cross-validation reference for nutrient values. Its standardised structure and high data quality make it a reliable secondary source for foods commonly consumed in continental Europe.
USDA FoodData Central
Supplementary nutrient reference · U.S. Department of Agriculture
USDA FoodData Central provides supplementary nutrient values used in NUTRON's harmonisation pipeline, improving coverage when certain nutrient attributes are sparse in European sources and supporting consistent nutrient mapping across databases.
Polyphenol Composition Database
~458 foods · Comprehensive phytochemical coverage · INRAE / Clermont-Fd
A specialised database focused exclusively on polyphenol and phytochemical content in foods, developed by INRAE (France). Phenol-Explorer is the primary source for NUTRON's flavonoid, anthocyanin, stilbene, and total polyphenol data — nutrients largely absent from conventional food composition databases but increasingly implicated in chronic disease prevention.
Data Currency & Updates
NUTRON's underlying database versions are detailed in our internal changelog. Source databases are updated periodically by their maintaining institutions; NUTRON integrates updates through its harmonisation pipeline on a best-effort basis. Users should be aware that database values may differ from those published after our last synchronisation date.
3Methodology
From raw data to meaningful insight
Raw food composition data across multiple national databases is inherently inconsistent — different units, naming conventions, food groupings, and measurement methodologies. NUTRON's harmonisation pipeline resolves these inconsistencies in a principled, auditable way.
Source Ingestion & Standardisation
Each database is ingested in its native format and mapped to a canonical nutrient schema derived from FRIDA's column conventions. Nutrient names, units of measurement, and food categorisation codes are normalised across all sources. Where mappings are scientifically ambiguous (e.g. isomers, sub-fractions), conservative rules are applied and documented in our internal mapping registry.
Unit Harmonisation
All nutrient values are converted to consistent SI-adjacent units prior to merging. For example, values originally reported in mg are converted to μg where standards require it (e.g. carotenoids, trace minerals). Conversion factors adhere to established analytical chemistry standards and are applied uniformly across all source databases.
Cross-Database Merging & Conflict Resolution
For nutrients present in multiple databases, NUTRON applies a reliability-weighted averaging approach. Each source is assigned a confidence weight reflecting the peer-review status, sample sizes, and analytical methods of the original database. FRIDA and UK CoF carry the highest baseline weights (0.95) due to their methodological rigour and breadth of documentation. Where values differ significantly between sources, the divergence is flagged internally for review.
Derived Nutrient Computation
Certain nutritional aggregates are computed from component nutrients rather than sourced directly. These "NUTRON-derived" values include total carotenoids (sum of α-carotene, β-carotene, lycopene, cryptoxanthin, and lutein in μg equivalents), total provitamin A carotenoids, and aggregated flavonoid classes from Phenol-Explorer data. Derived fields are clearly attributed in the data model.
FoodCompass 2.0 Scoring
The harmonised nutrient profile for each food item is passed through the FoodCompass 2.0 scoring algorithm. Each of the nine nutrient domains is evaluated independently against the reference criteria published in the original FoodCompass 2.0 methodology paper. The resulting domain scores are averaged to produce a single FoodCompass Score (FCS) on a 1–100 scale. Full details are provided in Section 4.
AI-Assisted Enrichment
For composite foods and novel food items where database coverage is incomplete, NUTRON may use large language model (LLM) inference to estimate categorical attributes such as NOVA processing classification and ingredient composition. These AI-enriched fields are distinguished from empirical database values and carry appropriately reduced confidence indicators. Full details are in Section 5.
4FoodCompass 2.0
A holistic, evidence-based scoring system
FoodCompass 2.0 is a nutrient profiling system developed by Dariush Mozaffarian, Venkat R. Bhupathiraju, and colleagues at the Friedman School of Nutrition Science and Policy, Tufts University. It was designed to overcome the limitations of single-nutrient scoring systems and provide a comprehensive, balanced assessment of food healthfulness across a wide range of food types.
NUTRON implements FoodCompass 2.0 as described in the original peer-reviewed publication (Mozaffarian et al., 2022, Nature Food). NUTRON is not affiliated with or endorsed by the FoodCompass research team or Tufts University. We implement the publicly published algorithm in good faith for informational purposes.
How the Score Works
FoodCompass evaluates each food across nine nutrient domains, scoring each domain on a 0–10 scale. The final FoodCompass Score (FCS) is the unweighted mean of domain scores, resulting in a score from 1 (least healthy) to 100 (most healthy). The nine domains are:
Not all nine domains can always be computed for every food item, as some attributes (notably D4 Ingredient Composition and D5 Additives) require information not available in standard food composition databases. Where domain data is unavailable, NUTRON applies the methodology's prescribed fallback rules, which may result in that domain receiving a neutral or estimated score. Scores computed from incomplete domain data are flagged accordingly in the interface.
Score Interpretation
Score Bands (Illustrative)
FoodCompass scores are continuous. Higher scores indicate a more favourable overall nutritional profile. The original authors suggest approximate interpretive categories, but these are guides rather than absolute thresholds. A score of 70+ generally indicates foods encouraged for regular consumption; scores below 31 generally correspond to foods consumed sparingly.
A score is not a verdict
FoodCompass scores reflect nutrient composition relative to the algorithm's reference criteria. They do not account for individual health status, dietary context, portion size, food interactions, cultural preferences, or the reality that dietary patterns — not individual foods — are the strongest determinants of long-term health outcomes.
5AI Use Disclosure
Where AI is used, and where it is not
Artificial intelligence plays a specific, bounded role in NUTRON. We believe in being precise about where AI inference is and is not involved, so you can calibrate your trust appropriately.
AI Disclosure
NUTRON uses large language model (LLM) inference — currently via Anthropic's Claude API — for a limited set of enrichment tasks involving composite or novel food items where structured database coverage is insufficient. AI-generated values are supplementary only. They do not replace or override empirical database values and are clearly indicated in the user interface where they appear.
Where AI is used
NOVA processing level classification for foods lacking explicit group assignments; ingredient disaggregation for complex composite dishes; food-to-database entity matching where name ambiguity exists; natural language queries and conversational nutrition guidance within the platform interface.
Where AI is not used
Nutrient values drawn directly from FRIDA, UK CoF, SWISS, USDA, or Phenol-Explorer are not generated, modified, or interpolated by AI. FoodCompass 2.0 scoring is performed deterministically using the published algorithm. No AI inference is involved in the core data pipeline for foods with full database coverage.
AI Limitations & Accuracy
AI-generated nutritional attributes carry inherent uncertainty. LLMs can produce plausible but incorrect outputs, particularly for niche food items, regional specialities, or novel ingredient combinations. NUTRON does not represent AI-inferred values as empirically verified data. Users should treat AI-enriched fields as provisional estimates and consult primary sources for clinical or professional use.
AI-generated content within NUTRON must not be used as the basis for clinical decision-making, dietary prescription, allergy management, or any other health-critical purpose. The inherent probabilistic nature of large language models makes them unsuitable for applications where accuracy is safety-critical.
6Known Limitations
What the data can and cannot tell you
Scientific integrity requires honesty about the boundaries of our data. The following limitations apply to NUTRON's nutritional data and should be considered when interpreting scores and values.
Database Coverage Gaps
No food composition database is complete. Polyphenol data exists for only ~458 foods in Phenol-Explorer; many common foods lack phytochemical data entirely. Micronutrient values may be absent, estimated, or averaged from limited analytical samples. Scores for foods with sparse coverage are inherently less reliable.
Biological Variability
Nutrient content in real foods varies substantially by variety, season, soil quality, storage conditions, preparation method, and ripeness. Published database values represent average analytical measurements and may not reflect the specific item you are consuming. Cooked versus raw preparations can alter nutrient availability significantly.
Bioavailability Is Not Captured
Nutrient amounts listed are compositional values, not measures of what the body actually absorbs and utilises. Bioavailability is highly context-dependent — affected by food matrix, co-consumed nutrients, cooking method, gut microbiome, and individual physiology. A food high in iron may provide negligible absorbable iron depending on these factors.
Scoring Algorithm Limitations
FoodCompass 2.0, like any nutrient profiling system, reflects the state of nutritional science at the time of its development. It embeds certain assumptions and scoring criteria that may be updated as evidence evolves. The algorithm was originally developed and validated using USDA data; NUTRON's application of it to European food databases involves some methodological extrapolation.
Dietary Context Is Everything
Individual food scores are poor predictors of health outcomes. A diet consisting exclusively of individually high-scoring foods could still be nutritionally inadequate or harmful in other ways. NUTRON scores should be interpreted as one input among many in a holistic dietary assessment, not as rankings of absolute health value.
Cultural & Culinary Context
Nutrition is inseparable from culture, pleasure, and social context. A food that scores modestly on FoodCompass may be central to a culturally balanced dietary pattern with proven health benefits. NUTRON intentionally surfaces flavour profile data alongside nutritional scores to acknowledge this reality — the tradeoffs are yours to make.
7Legal & Rights
Attribution, licensing & your rights
Third-Party Database Attribution
NUTRON uses data from publicly available food composition databases maintained by government and academic institutions. We acknowledge and respect the intellectual effort behind these resources and attribute them accordingly.
DTU Fødevareinstituttet — Technical University of Denmark
Data sourced from the FRIDA food composition database. © DTU Fødevareinstituttet. Used in accordance with the database's public access terms. NUTRON is not affiliated with or endorsed by DTU.
Public Health England & Food Standards Agency
Nutritional composition data from McCance and Widdowson's The Composition of Foods (7th Ed.) and associated supplementary volumes. © Crown copyright. Used under Open Government Licence v3.0. NUTRON is not affiliated with or endorsed by PHE or FSA.
Swiss Federal Food Safety and Veterinary Office (FSVO / BLV)
Data sourced from the Swiss Food Composition Database. © FSVO. Used in accordance with the database's public access terms. NUTRON is not affiliated with or endorsed by FSVO.
USDA FoodData Central — U.S. Department of Agriculture
Data sourced from USDA FoodData Central. Public domain U.S. government works. Used in accordance with USDA public access terms. NUTRON is not affiliated with or endorsed by USDA.
Phenol-Explorer — INRAE / Université de Clermont-Ferrand
Polyphenol composition data from Phenol-Explorer v3.6. © INRAE and Université de Clermont-Ferrand. Used in accordance with academic public access terms. NUTRON is not affiliated with or endorsed by INRAE.
FoodCompass Algorithm Attribution
The FoodCompass 2.0 scoring algorithm is the intellectual property of its authors (Mozaffarian et al.) and Tufts University. NUTRON implements the published algorithm as described in the primary research paper cited in Section 8. NUTRON is not affiliated with, endorsed by, or a licensee of Tufts University or the FoodCompass research programme. The name "FoodCompass" is used solely to accurately describe the scoring methodology applied.
NUTRON Platform Content
NUTRON's software, user interface, design system, harmonisation pipeline, derived data structures, and platform-generated content are proprietary to NUTRON. All rights reserved. Source database values remain the property of their respective maintaining institutions.
No Warranty
Data Accuracy Disclaimer
NUTRON provides nutritional data and scores on an "as is" basis. We make reasonable efforts to ensure accuracy but make no warranties, express or implied, regarding the completeness, accuracy, reliability, or fitness for purpose of any data presented. NUTRON shall not be liable for any damages arising from reliance on data presented within the platform, including but not limited to decisions made in relation to diet, health, or medical conditions.
Limitation of Liability
To the fullest extent permitted by applicable law, NUTRON and its operators disclaim all liability for any direct, indirect, incidental, special, consequential, or punitive damages resulting from your use of, or inability to use, this platform or the data it presents. This includes but is not limited to damages arising from dietary decisions, allergic reactions, health outcomes, or reliance on AI-generated content.
Governing Law
These terms and the NUTRON platform are governed by and construed in accordance with the laws of England and Wales. Any disputes arising from the use of this platform shall be subject to the exclusive jurisdiction of the courts of England and Wales.
Changes to This Page
NUTRON reserves the right to update this Data & Methodology page at any time. Material changes will be communicated through the platform. Continued use of NUTRON following any such changes constitutes acceptance of the updated terms. The date of last update is noted in the page header above.
8References
Primary literature & sources
The following references underpin the scientific methodology, data sources, and scoring system implemented in NUTRON.
- 1
Mozaffarian D, Malaeb D, Micha R, et al.
Food Compass is a nutrient profiling system using expanded characteristics for assessing healthfulness of foods. Nature Food. 2021;2:809–818. doi:10.1038/s43016-021-00381-y
↗ doi:10.1038/s43016-021-00381-y - 2
DTU Fødevareinstituttet
FRIDA — Food Data (version 4). National Food Institute, Technical University of Denmark. Accessed 2025. frida.fooddata.dk
↗ frida.fooddata.dk - 3
Finglas PM, Roe MA, Pinchen HM, et al.
McCance and Widdowson's The Composition of Foods, 7th Summary Edition. Royal Society of Chemistry, 2015. Supplementary volumes and updates maintained by Public Health England / UKHSA.
- 4
Federal Food Safety and Veterinary Office (FSVO)
Swiss Food Composition Database v6.4. Bern: FSVO; 2024. naehrwertdaten.ch
↗ naehrwertdaten.ch - 5
Rothwell JA, Pérez-Jiménez J, Neveu V, et al.
Phenol-Explorer 3.0: a major update of the Phenol-Explorer database to incorporate data on the effects of food processing on polyphenol content. Database (Oxford). 2013;2013:bat070. doi:10.1093/database/bat070
↗ doi:10.1093/database/bat070 - 6
Monteiro CA, Cannon G, Levy R, et al.
NOVA. The star shines bright. [Food classification. Public health]. World Nutrition. 2016;7(1–3):28–38. The reference classification for NOVA food processing levels used in FoodCompass Domain 6.
- 7
Hu FB
Dietary pattern analysis: a new direction in nutritional epidemiology. Current Opinion in Lipidology. 2002;13(1):3–9. Foundational context for dietary pattern vs. single food assessment.
- 8
Pérez-Jiménez J, Neveu V, Vos F, Scalbert A
Identification of the 100 richest dietary sources of polyphenols: an application of the Phenol-Explorer database. European Journal of Clinical Nutrition. 2010;64:S112–S120. doi:10.1038/ejcn.2010.221
↗ doi:10.1038/ejcn.2010.221