RPMVCDA: Random Perturbation and Multi-View Graph Convolutional Networks for CircRNA-Disease Association Prediction.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Circular RNAs (circRNAs) play regulatory roles in various diseases.
• Accurate identification of circRNA-disease associations is crucial for understanding disease mechanisms.
• Existing computational models face challenges due to sparse circRNA-disease association networks.

Purpose of the Study:

• To develop an advanced computational model, RPMVCDA, for predicting circRNA-disease associations.
• To improve the accuracy and robustness of circRNA-disease association prediction.
• To address the sparsity issue in existing circRNA-disease association networks.

Main Methods:

• Proposed RPMVCDA model integrating random perturbation and multi-view graph convolutional networks (GCNs).
• Constructed multiple circRNA and disease similarity networks, utilizing multi-view GCNs for embedding representations.
• Incorporated a feature similarity association network for message passing and a random perturbation association network to explore potential associations.
• Employed a self-attention mechanism to generate high-quality features for association score calculation.

Main Results:

• RPMVCDA demonstrated superior performance compared to existing models in predicting circRNA-disease associations.
• Five-fold cross-validation and case studies on the CircR2Disease dataset validated the model's effectiveness.
• The proposed random perturbation approach effectively explored potential circRNA-disease associations.

Conclusions:

• RPMVCDA offers a promising alternative for predicting circRNA-disease associations.
• The model's performance highlights the potential of integrating multi-view GCNs and random perturbation techniques.
• Continued development of computational models is essential for advancing circRNA-disease association research.