GGAECDA: Predicting circRNA-disease associations using graph autoencoder based on graph representation learning.

Area of Science:

• Biochemistry and Molecular Biology
• Computational Biology and Bioinformatics
• Genomics and Genetics

Background:

• Circular RNAs (circRNAs) are crucial regulators in biological processes and potential disease biomarkers.
• Traditional methods for identifying circRNA-disease associations are resource-intensive and time-consuming.
• Accurate prediction of circRNA-disease associations is vital for understanding disease mechanisms and developing diagnostics.

Purpose of the Study:

• To develop an advanced deep learning model for predicting circRNA-disease associations.
• To overcome the limitations of traditional experimental methods in identifying novel circRNA-disease interactions.
• To provide a reliable computational tool for guiding research on circRNA functions in diseases.

Main Methods:

• A novel deep learning model, GGAECDA, was developed using a graph autoencoder (GAE) framework.
• The model integrates Graph Attention Network (GAT) for learning low-order neighbor information and Random Walk with Restart (RWR) for high-order neighbor information.
• Feature representations from GAT and RWR were combined and processed through co-trained GAEs to predict circRNA-disease associations.

Main Results:

• The GGAECDA model achieved an average AUC of 0.9359 in five-fold cross-validation.
• Case studies validated the model's capability to identify potential candidate circRNAs for human diseases.
• The model effectively mines low-dimensional representations from circRNA and disease similarity networks.

Conclusions:

• GGAECDA is a powerful and reliable computational tool for predicting circRNA-disease associations.
• The model offers a more efficient alternative to traditional experimental approaches.
• GGAECDA can guide future research into the roles of circRNAs in various pathological conditions.