TLDR: This research introduces “Flavonoid Fusion,” a knowledge graph designed to map the complex relationships between food, flavonoids, and health, particularly in the context of cancer. By integrating data from USDA databases and scientific literature, the study creates a machine-readable format to understand how dietary flavonoids impact disease management. It uses the KNARM methodology to build an ontology, aiming to facilitate new hypotheses and support the “food as medicine” concept, with future plans to incorporate AI for automated development.
The idea of ‘food as medicine’ is gaining significant attention in health circles, with many studies exploring how what we eat impacts our well-being. However, there’s been a notable gap in representing these intricate relationships between food and health in a standardized, machine-readable format that can be effectively used by computers and AI systems.
To bridge this gap, a recent study titled Flavonoid Fusion: Creating a Knowledge Graph to Unveil the Interplay Between Food and Health introduces an innovative approach: a knowledge graph. This graph is designed to link food and health by combining information from various sources, specifically focusing on the flavonoid content of foods found in the USDA’s extensive databases and connections to cancer identified in scientific literature.
Understanding Flavonoids and Their Health Benefits
Flavonoids are a diverse group of compounds found abundantly in plants. They are a subclass of polyphenols and are crucial for numerous health benefits. For instance, many flavonoids are known to reduce oxidative stress and inflammation in the human body, which are key factors in conditions like cardiovascular diseases. Specific examples include Anthocyanins, flavones, flavonols, and isoflavones, which have shown positive effects on asthma. Flavonoids like Luteolin have demonstrated anti-cancer properties, while Quercetin, found in common foods like apples, onions, and broccoli, exhibits anti-inflammatory activities.
Given the complexity and variety of flavonoids and their food sources, understanding their intricate relationships can be challenging. This is where knowledge graphs offer a powerful solution. They allow for the organization and analysis of complex data, helping researchers identify patterns, dependencies, and correlations that would be difficult to spot in isolated datasets. These graphs are also flexible, meaning they can be updated and refined as new research findings emerge, ensuring the information remains current and accurate.
The Methodology Behind Flavonoid Fusion
The researchers employed the KNARM methodology (Knowledge Acquisition and Representation Methodology) to develop this knowledge graph. The process involved several key steps:
- Dataset Collection: The primary data came from the United States Department of Agriculture (USDA) flavonoid datasets, which provide extensive information on flavonoid molecules and their food sources. This was complemented by literature exploring relationships between cancers and flavonoid components of food.
- Data Cleaning and Analysis: Initial steps involved converting and cleaning unstructured data, addressing issues like inconsistent spellings. A ‘Sub-language analysis’ was performed to identify key concepts and relationships relevant to the domain.
- Expert Input: The researchers, acting as domain experts due to their deep understanding of the data, conducted ‘Unstructured Interviews’ to filter relevant information and integrate cancer-food relationships from web literature.
- Standardization: Through ‘Sub-language Recycling,’ existing terms and definitions from established ontologies like Chemical Entities of Biological Interest (ChEBI), Compositional Dietary Nutrition Ontology (CDNO), and DOID (for disease names) were reused. This ensured consistency and reduced redundancy, for example, mapping ‘apple’ and ‘apples’ to the same set of nutrients.
- Knowledge Modeling: ‘Metadata Creation and Knowledge Modelling’ involved cross-referencing with ChEBI and NCBI taxonomies, establishing that flavonoids are secondary metabolites in food products. Crucially, a ‘has associated disease’ linkage was added to connect flavonoids with specific cancer subtypes.
- Database Creation and Validation: The Neo4J graph database application was used to store the data, establishing relationships like ‘Parent of,’ ‘Has An ID,’ and ‘has associated disease.’ The database was validated by querying it to ensure accuracy, for instance, checking the flavonoids present in milk.
- Semi-Automated Ontology Building: The dataset was expanded with information on breast and colon cancer associations. A command-line tool called ROBOT was used to semi-automate the ontology building process from organized CSV files, creating a three-layer structure for vocabulary, axioms, and merging flavonoid and food data.
A New Paradigm for Food and Health
The ‘Flavonoid Fusion’ knowledge graph creates a novel modeling paradigm that connects drugs-to-food-to-disease through ‘chemical composition components.’ This connection highlights the foundational belief that food possesses medicinal properties and opens new avenues for generating hypotheses in the ‘food as medicine’ approach. The research aims to facilitate more effective evaluation and decision-making in nutrition, health, and food science, potentially leading to new dietary guidelines and therapeutic targets.
Also Read:
- M-Reason: A Multi-Agent System for Auditable Biomedical Evidence Synthesis
- Intelligent IoT for Agriculture: A Semantic Layered System
Future Directions
The researchers acknowledge that ontology development is a human-intensive process. Future work involves expanding the scope of the knowledge graph and reducing dependency on domain experts by integrating Large Language Models (LLMs) with OLIVE (Ontology Learning with Integrated Vector Embeddings). This will help expedite the ontology development timeline and enable users to generate knowledge graphs more efficiently.
