CRPGCN: predicting circRNA-disease associations using graph convolutional network based on heterogeneous network.

Area of Science:

• Genomics and Bioinformatics
• Computational Biology
• Disease Biomarker Discovery

Background:

• Circular RNAs (circRNAs) show potential as biomarkers for disease diagnosis and treatment.
• The functional roles and disease associations of most circRNAs remain largely unknown.
• Current experimental methods for predicting circRNA-disease associations are often costly and time-consuming.

Purpose of the Study:

• To develop an efficient computational method for predicting circRNA-disease associations.
• To overcome limitations of existing methods that use insufficient circRNA attribute information.
• To provide a more accurate and cost-effective approach for identifying potential circRNA-disease links.

Main Methods:

• A novel algorithm, CRPGCN, integrating Graph Convolutional Network (GCN), Random Walk with Restart (RWR), and Principal Component Analysis (PCA) was developed.
• RWR was employed to enhance similarity associations between neighboring nodes.
• PCA was utilized for dimensionality reduction and feature extraction, improving the proximity of related circRNAs and diseases.

Main Results:

• The CRPGCN algorithm successfully learned features between circRNAs and diseases using heterogeneous adjacency and feature matrices.
• Cross-validation studies demonstrated the algorithm's predictive power.
• The method achieved high Area Under the ROC Curve (AUC) values: 0.9490 (2-fold), 0.9720 (5-fold), and 0.9722 (10-fold).

Conclusions:

• The proposed CRPGCN method is effective in predicting associations between circRNAs and diseases.
• CRPGCN offers a valuable computational tool for advancing research in circRNA-mediated diseases.
• This approach contributes to a more efficient discovery of disease biomarkers and therapeutic targets.