NutriHeal-Graph: Dataset for Selective Nutrition-Goal Suitability Verification

NutriHeal-Graph is a comprehensive nutrition-health knowledge graph dataset that integrates structured relationships between nutrients, foods, diseases, and therapeutic dietary interventions. This dataset is designed to support research in nutritional informatics, clinical dietetics, and biomedical knowledge graph applications. The knowledge graph encodes multi-relational data across four primary entity types: (1) nutrients and bioactive compounds, (2) food items and food groups, (3) diseases and health conditions, and (4) dietary patterns and therapeutic diets. Edges in the graph represent evidence-based relationships such as "contains", "deficiency_causes", "beneficial_for", "contraindicated_in", and "recommended_for", derived from peer-reviewed nutritional science literature and clinical dietary guidelines. Key features of NutriHeal-Graph include: Over 15,000 entities spanning nutrients, foods, diseases, and dietary regimens - More than 80,000 relational triples extracted from clinical nutrition databases and scientific literature - Integration of data from authoritative sources including USDA FoodData Central, NHANES, PubMed, and WHO dietary guidelines - RDF/OWL-compatible format supporting SPARQL queries and graph neural network (GNN) applications - Annotated with confidence scores and provenance metadata for each triple This dataset enables downstream tasks including disease-diet association prediction, nutrient deficiency inference, personalized dietary recommendation, and biomedical question answering. NutriHeal-Graph is intended for use by researchers in nutritional genomics, clinical informatics, and AI-driven healthcare applications.