A unified knowledge graph linking foodomics to chemical-disease networks and flavor profiles
Fangzhou Li, Ilias Tagkopoulos, Keer Ni, Michael Gunning, Jason Youn, Kaichi Xie, Trevor Chan, Pranav Gupta, Arielle Yoo
npj Science of Food
Problems Identified (2)
Diet-molecular-effect map gap: Modern nutrition science lacks a comprehensive, machine-readable map connecting diet with molecular composition and biological effects.
Diet-molecular-effect map gap: Modern nutrition science lacks a comprehensive, machine-readable map connecting diet with molecular composition and biological effects.
Proposed Solutions (5)
FoodAtlas knowledge graph: FoodAtlas is a large-scale knowledge graph linking foods, chemicals, diseases, and flavor descriptors using provenance-tracked edges.
Transformer food-chemical mining: A transformer-based text-mining pipeline extracts quantitative food–chemical associations from literature sentences.
Integrated foodomics data fusion: The system integrates food–chemical associations with chemical–disease assertions, chemical-bioactivity records, flavor annotations, and taxonomic relationships.
FoodAtlas utility models: Models built on FoodAtlas include a bioactivity predictor and a substitution engine for evaluating practical utility.
FoodAtlas knowledge graph: FoodAtlas is a large-scale knowledge graph linking foods, chemicals, diseases, and flavor descriptors using provenance-tracked edges.
Results (3)
Provenance-tracked KG scale:
Text-mining extraction performance:
Dietary disease-risk modules:
Research Domain
Nutrition science / foodomics knowledge graphs