MVGCNMDA: Multi-view Graph Augmentation Convolutional Network for Uncovering Disease-Related Microbes

Area of Science:

• Computational biology and bioinformatics
• Microbiome research
• Disease association studies

Background:

• Understanding microbe-disease interrelationships is vital for clinical decision-making and treatment planning.
• Current prediction methods relying on experiments and domain knowledge are inefficient and time-consuming.
• Automated algorithms are needed to uncover microbe-disease links, but data noise and biological complexity pose challenges.

Purpose of the Study:

• To develop an automated computational model for predicting potential microbe-disease associations.
• To address challenges of data noise and improve the efficiency of microbe-disease association prediction.
• To introduce a Multi-View Graph Augmentation Convolutional Network (MVGCNMDA) for this purpose.

Main Methods:

• Employed data augmentation techniques (edge perturbation, node dropping) to mitigate data noise.
• Utilized Gaussian interaction profile kernel similarity and cosine similarity for multi-view feature extraction.
• Integrated a multi-attention block and Convolutional Neural Network (CNN) combiner within a Graph Convolutional Network (GCN) framework, followed by matrix completion for prediction.

Main Results:

• Achieved high performance metrics: AUPR (0.9440), AUC (0.9428), F1 score (0.9383), and RECALL (0.8858) on multiple datasets.
• Demonstrated superior accuracy in predicting microbe-disease associations compared to state-of-the-art methods via cross-validation.
• Ablation studies confirmed the effectiveness of the proposed graph data augmentation, and case studies validated predictions.

Conclusions:

• The MVGCNMDA model effectively predicts potential microbe-disease associations with high accuracy and efficiency.
• The developed model offers a robust computational approach to advance microbiome and disease research.
• This work provides a valuable tool for identifying novel microbial roles in human diseases.