Predicting potential microbe-disease associations based on dual branch graph convolutional network | JoVE Visualize

Area of Science:

Computational biology
Bioinformatics
Genomics and Microbiomics

Background:

Understanding microbe-disease associations is vital for disease prevention, diagnosis, drug development, and personalized medicine.
Traditional laboratory methods for confirming microbe-disease relationships are time-consuming and expensive.
There is a critical need for advanced computational frameworks to predict novel microbe-disease associations efficiently.

Purpose of the Study:

To develop and validate a novel computational approach, DBGCNMDA, for identifying microbe-disease associations.
To leverage dual graph convolutional networks (GCNs) for enhanced feature extraction and prediction accuracy.
To address data dimensionality issues and optimize network connectivity for robust association prediction.

Main Methods:

Calculated microbe and disease similarity matrices using integrated functional similarity and Gaussian association spectrum kernel (GAPK) similarity.
Employed a dual branch GCN module (GlobalGCN and LocalGCN) to extract semantic information from biological networks.
Optimized network connectivity with random walk with restart (RWR) and used the similarity matrix as the initial feature matrix.

Main Results:

The DBGCNMDA model achieved high accuracy in five-fold cross-validation (5-fold-CV), with an Area Under the ROC Curve (AUC) of 0.9559 and Area Under the Precision-Recall Curve (AUPR) of 0.9630.
Case studies using published experimental data confirmed a significant number of predicted microbe-disease associations.
The method effectively addresses low data dimensionality by extending disease nodes and incorporates homologous neighbor information.

Conclusions:

DBGCNMDA is a powerful and effective computational tool for predicting potential microbe-disease associations.
The dual branch GCN approach enhances the understanding of complex biological network relationships.
The findings support the utility of computational methods in accelerating the discovery of microbe-disease links for clinical applications.