SGAEMDA: Predicting miRNA-Disease Associations Based on Stacked Graph Autoencoder.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• MicroRNA-disease association (MDA) prediction is crucial for disease management.
• Traditional experimental methods for MDA prediction are time-consuming and expensive.

Purpose of the Study:

• To develop an efficient computational model for predicting potential miRNA-disease associations.
• To overcome the limitations of traditional experimental approaches in MDA prediction.

Main Methods:

• Proposed SGAEMDA (Stacked Graph Autoencoder-Based Prediction of Potential miRNA-Disease Associations), a multi-layer collaborative unsupervised training model.
• Extracted similarity and association features from miRNA and disease data.
• Utilized stacked graph autoencoder for unsupervised learning of low-dimensional representations.
• Integrated learned representations with association features for final pair feature extraction.
• Employed a multilayer perceptron (MLP) for predicting unknown miRNA-disease associations.

Main Results:

• SGAEMDA achieved high performance with mean ROC AUC scores of 0.9585 (5-fold CV) and 0.9516 (10-fold CV).
• The model significantly outperformed existing baseline methods in MDA prediction accuracy.
• Case studies demonstrated SGAEMDA's ability to accurately identify candidate microRNAs for various neoplasms.

Conclusions:

• SGAEMDA provides an effective and accurate computational approach for predicting miRNA-disease associations.
• The model's unsupervised learning capability and feature integration enhance prediction performance.
• SGAEMDA holds promise for advancing disease prevention, diagnosis, and treatment strategies.