Predicting Microbe-Disease Association by Learning Graph Representations and Rule-Based Inference on the...

Area of Science:

• Microbiology
• Computational Biology
• Bioinformatics

Background:

• Clinical observations increasingly highlight the significant impact of microbes on human diseases.
• Understanding microbe-disease relationships is crucial for effective disease prevention and therapeutic strategies.

Purpose of the Study:

• To develop a predictive model, LGRSH (Learning Graph Representations and a modified Scoring mechanism on the Heterogeneous network), for discovering potential microbe-disease associations.
• To integrate known microbe-disease associations with network analysis for enhanced prediction accuracy.

Main Methods:

• Constructed microbe and disease similarity networks using Gaussian interaction profile kernel similarity.
• Developed a heterogeneous network integrating similarity networks and known microbe-disease associations.
• Employed Node2vec for learning node representations within the heterogeneous network.
• Applied a modified rule-based inference method to calculate microbe-disease relevance based on learned representations.

Main Results:

• The LGRSH model demonstrated superior performance compared to existing methods (LRLSHMDA, KATZHMDA, BiRWHMDA).
• Case studies on asthma, Chronic Obstructive Pulmonary Disease, and Inflammatory Bowel Disease showed high validation rates for top-predicted microbe-disease associations (8/10, 8/10, and 10/10, respectively).

Conclusions:

• LGRSH effectively predicts potential microbe-disease associations.
• The model holds promise for advancing our understanding of the microbiome's role in human health and disease.