FAIR²-FoodHealth: A Provenance-Rich Knowledge Graph Linking Empirical Food Composition, Nutrients, Health-Variable Definitions and Cardiometabolic Guidance (Dataset v0.4.0)

FAIR²-FoodHealth: Dataset Description FAIR²-FoodHealth (version 0.4.0) is a provenance-rich, machine-readable research dataset and knowledge graph designed to support semantic interoperability between food composition, nutrition, health-variable definitions, cardiometabolic guidance, and food-supply-chain information. The resource addresses a fundamental challenge in food and health data integration: heterogeneous sources frequently use different identifiers, terminology, units and measurement contexts. Integrating these sources through simple name matching can produce scientifically incorrect relationships. FAIR²-FoodHealth addresses this problem through explicit, qualified semantic mappings, documented provenance, and auditable acceptance and rejection decisions. Dataset contents The release contains: 24 interconnected CSV tables comprising 666 records. 52 empirical food records from USDA FoodData Central Foundation Foods and SR Legacy, including 398 nutrient composition values expressed per 100 g. An RDF knowledge graph comprising 15,028 triples, distributed in Turtle and JSON-LD formats. Semantic mappings incorporating FoodOn, ChEBI, NCIt, LOINC, UCUM and Mondo terminology. Population-level dietary guidance links covering hypertension, obesity and type 2 diabetes. A clearly identified synthetic demonstration of food-supply-chain, food-safety and food-fraud relationships. A comprehensive data dictionary, provenance metadata, JSON Schemas, SHACL validation shapes and SPARQL competency queries. Reproducibility scripts, validation tests, metadata and supporting documentation. Scientific contribution The principal contribution is a reproducible semantic integration resource that preserves source-native information while explicitly distinguishing empirical observations, inferred relationships, synthetic demonstrations and computationally drafted mappings. The accompanying evaluation demonstrates how qualified semantic mappings can prevent inappropriate connections between distinct nutritional concepts, including total versus added sugars and food energy content versus dietary energy intake. Reusability and applications The resource is intended to support research in food-data interoperability, semantic knowledge graphs, personalized nutrition informatics, cardiometabolic prevention, data provenance, FAIR data management and trustworthy AI. Potential applications include knowledge-graph-based retrieval, provenance-aware retrieval-augmented generation, federated data integration, food-composition comparisons and transparent data-quality assessment. Validation and limitations The dataset includes automated structural validation, referential-integrity checks, SHACL constraints, deterministic RDF generation, competency-query testing and reproducibility documentation. The empirical subset is limited and not representative of the broader food supply. Semantic mapping and matching decisions are computationally drafted and await independent expert review. The synthetic records do not represent real supply-chain observations. The resource contains no patient-level health observations and is not intended for clinical decision-making or individualized dietary recommendations. The FAIR² designation reflects the resource's intended alignment with FAIR and AI-ready data principles; formal FAIR² certification is not claimed. Data provenance and licensing Empirical food-composition values originate from USDA FoodData Central (CC0). Project-created data structures, documentation and curation resources are accompanied by source-level provenance and licensing information. Third-party terminology and referenced guidance retain their respective attribution and usage conditions. Version: 0.4.0Resource type: DatasetKeywords: FAIR data, knowledge graph, semantic interoperability, food composition, personalized nutrition, cardiometabolic health, provenance, RDF, AI-ready data, FoodData Central, dietary guidelines