A lightweight single-view contrastive learning hypergraph neural network for food-microbe-disease association...

Area of Science:

• Computational biology
• Nutritional science
• Microbiome research

Background:

• Understanding food-gut microbiota-disease links is crucial for personalized nutrition.
• Existing computational methods for microbiota-disease associations are limited, especially for higher-order food-microbiota-disease interactions.

Purpose of the Study:

• To develop a novel computational model for predicting ternary food-microbe-disease (FMD) associations.
• To address the limitations in predicting food-microbiota relationships and higher-order interactions.

Main Methods:

• Constructed a comprehensive Food-Microbe-Disease (FMD) database.
• Proposed a lightweight single-view contrastive learning hypergraph neural network (LSCHNN).
• Formulated FMD interactions as a hypergraph, incorporating biological features and employing contrastive learning for enhanced feature extraction and generalization on sparse data.

Main Results:

• The LSCHNN model achieved superior precision in predicting ternary FMD associations compared to state-of-the-art methods.
• LSCHNN successfully identified a greater number of potential FMD associations.
• Case studies validated the model's effectiveness in identifying microbe-specific FMD associations.

Conclusions:

• Introduced LSCHNN, the first model integrating hypergraph neural networks with lightweight single-view contrastive learning for FMD association prediction.
• LSCHNN offers a groundbreaking framework for advancing precision nutrition and personalized dietary interventions.