A computational model for potential microbe-disease association detection based on improved graph convolutional...

Area of Science:

• Microbiome research
• Computational biology
• Bioinformatics

Background:

• Human health and disease are intricately linked to the composition of the human microbiome.
• Accurate inference of microbe-disease associations is crucial for understanding health and disease pathogenesis.
• Existing methods for predicting microbe-disease associations require enhancement for improved accuracy and efficiency.

Purpose of the Study:

• To propose a novel computational model, GCANCAE, for inferring potential microbe-disease associations.
• To leverage graph attention networks and sparse autoencoders for robust association prediction.
• To integrate heterogeneous biological data, including known microbe-disease relationships and similarity networks.

Main Methods:

• Construction of a heterogeneous network incorporating microbe-disease relationships, disease similarity, and microbial similarity.
• Utilizing improved Graph Convolutional Networks (GCN) and Convolutional Sparse Autoencoders (CSAE) for feature extraction.
• Integration of extracted features to generate unique eigenmatrices for microbes and diseases to predict association likelihood.

Main Results:

• GCANCAE demonstrated superior prediction performance compared to state-of-the-art methods in 2-fold and 5-fold cross-validation.
• Experimental validation using established databases (HMDAD, Disbiome) confirmed the model's predictive power.
• Case studies on asthma, irritable bowel syndrome (IBS), and type 2 diabetes (T2D) highlighted GCANCAE's practical efficiency.

Conclusions:

• The GCANCAE model offers a powerful and accurate computational approach for identifying microbe-disease associations.
• This method advances the field of microbiome research by providing a robust tool for biological data analysis.
• GCANCAE's findings contribute to a deeper understanding of the role of the microbiome in various human diseases.