MVGNCDA: Identifying Potential circRNA-Disease Associations Based on Multi-view Graph Convolutional Networks and...

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Circular RNAs (circRNAs) are increasingly recognized for their roles in disease development.
• Traditional methods for identifying circRNA-disease associations are time-consuming and expensive.
• Existing computational methods struggle with data sparsity and inefficient similarity representation.

Purpose of the Study:

• To develop an innovative computational framework, MVGNCDA, for predicting circRNA-disease associations.
• To leverage multisource data for more effective detection of unknown associations.
• To overcome limitations of existing computational approaches in circRNA-disease association prediction.

Main Methods:

• Calculating disease semantic similarity, circRNA functional similarity, Gaussian interaction profile (GIP) kernel, and cosine similarity.
• Employing multi-view graph convolutional networks (GCNs) to extract local node embeddings.
• Constructing a heterogeneous network with integrated similarity and verified associations to learn global node embeddings.
• Utilizing a bilinear decoder for evaluating circRNA-disease associations based on fused local and global embeddings.

Main Results:

• MVGNCDA demonstrated superior performance compared to existing methods across five public datasets.
• Fivefold cross-validation confirmed the effectiveness of the proposed framework.
• A case study validated MVGNCDA's capability in efficiently identifying unknown circRNA-disease associations.

Conclusions:

• MVGNCDA offers an effective computational solution for predicting circRNA-disease associations.
• The framework successfully integrates multisource data and advanced network embedding techniques.
• MVGNCDA advances the field by enabling efficient discovery of novel circRNA-disease links.