The Metabolome of Food Knowledge Database: Development of a Nutrition Database to Support Precision Nutrition

OBJECTIVES: To develop a precision nutrition knowledge database, with the aim to provide individualized and actionable dietary recommendations to help prevent disease. However presently, dietary phytochemicals are poorly represented in current metabolomic databases. To address this gap, we are building a cloud-based knowledge database (KDB) named “The Metabolome of Food” (MetaboFood®) which focuses on phytochemical compositional, metabolite and pathway data. METHODS: MetaboFood® features P-MetDB®, a database in static tabular form, of nutritionally relevant phytochemicals and their metabolites derived from systematic literature reviews of 17 commonly consumed phytochemical-rich foods, matched to InChI key, physical and chemical properties (mass, formula) and database identifiers (i.e., PubChem ID, KEGG ID, SMILES etc.). To build MetaboFood®, information about metabolic pathways and diseases associated with these foods have been extracted from various pathway databases using APIs that these databases provide. Information can be searched in MetaboFood® and results are explored in a highly visual and interactive way, in the form of self-organizing maps, node-link diagrams, Sankey diagrams and other visual analytics techniques. RESULTS: MetaboFood® captures data on foods, their phytochemical compositions, human and microbial metabolites, and pathway and diseases linkages. Information in MetaboFood® facilitates both hypothesis generation and hypothesis testing relative to food and pathway analysis. Initial use of this database identifies significant interactions between polyphenol rich foods and numerous metabolic networks. CONCLUSIONS: MetaboFood® builds on traditional food composition databases by integrating biochemical and disease pathway data with diet metabolites. A key to moving forward is building data richness, enabling greater connections between diet and health. FUNDING SOURCES: Research reported in this abstract was supported by a NIEHS Human Health Exposure Analysis Resource (HHEAR) program grant under award number 1U2CES030857-01 and a NIH Nutrition for Precision Health (NPH) Metabolomics and Clinical Assay Center (MCAC) grant under the award number 1U24CA268153-01. CDK was also supported by the USDA National Institute of Food and Agriculture Hatch award (Kay-Colin; 1,011,757).