Microbe-Disease Association Prediction Using RGCN Through Microbe-Drug-Disease Network

Area of Science:

• Microbiology
• Computational Biology
• Bioinformatics

Background:

• Microbes significantly impact human health and disease.
• Identifying microbe-disease associations is vital for disease prevention strategies.

Purpose of the Study:

• To develop a predictive method, TNRGCN, for identifying microbe-disease associations.
• To leverage a Microbe-Drug-Disease network and Relation Graph Convolutional Network (RGCN) for enhanced prediction accuracy.

Main Methods:

• Constructed a Microbe-Drug-Disease tripartite network integrating data from HMDAD, Disbiome, MDAD, and CTD databases.
• Developed microbe, disease, and drug similarity networks using function, semantic, and Gaussian interaction profile kernel similarities.
• Applied Principal Component Analysis (PCA) for feature extraction and input into a two-layer RGCN model.

Main Results:

• TNRGCN outperformed existing methods in cross-validation tests for predicting microbe-disease associations.
• Case studies on Type 2 diabetes, Bipolar disorder, and Autism confirmed TNRGCN's effectiveness.

Conclusions:

• TNRGCN is an effective computational approach for predicting microbe-disease associations.
• The method holds promise for advancing disease prevention through a deeper understanding of microbial roles.