OBJECTIVES: A significant part of global public health problems can be traced back to environmental and lifestyle factors. The pivotal role of diet in these factors has been well supported by nutrition-related research; however, our current understanding of the way food molecules affect health is largely limited to a few hundred components tracked by food composition databases. Yet, these components represent only a tiny fraction of the total composition of the food supply, hindering research from discovering mechanistic effects and roles of food molecules. To solve this problem, we designed a Big Data and AI strategy for the creation of a high-resolution collection of food composition (Foodome).

METHODS: In this European pilot conducted in Hungary, data from 28 existing food composition, biochemical, and health effect databases were connected and curated. An AI-assisted pipeline (FoodMine) was designed to complement the database curation, reviewing and extracting data from scientific literature for the eight most relevant Hungarian food commodities. The resulting database was designed to show the interplay between foods, their compounds, and potential downstream effects of their consumption through a network science approach.

RESULTS: A unified Hungarian Foodome Knowledge Graph (HKG) was created. The HKG contains ontologies of 10 different entity types (FoodItem, CommonVariety, ChemicalCompound, ChemicalName, Concentration, Flavor, FlavorStrength, Disease, HealthEffect, HealthEffectStrength) and 20 relationships (e.g., chemical-disease associations, food item-health effect associations). The HKG alone contains 30k+ entities and 100k+ relationships, revealing connections between food compounds and potential mechanisms of action, both marker and therapeutic, and disease associations as well.

CONCLUSIONS: The complete biochemical composition of selected European foods is available through the HKG, enabling enterprises, research, and start-up communities to utilise it. As this pilot showed, a European Foodome could be developed, which could change the direction of food and nutrition research as well as highlight future development avenues in the health industry.