GraphCDA: a hybrid graph representation learning framework based on GCN and GAT for predicting disease-associated...

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Circular RNAs (circRNAs) are stable noncoding RNAs crucial in complex diseases.
• circRNAs serve as vital disease biomarkers and drug targets.
• Experimental identification of circRNA-disease associations is laborious and time-consuming.

Purpose of the Study:

• To develop an accurate computational method for predicting circRNA-disease associations.
• To facilitate the identification of disease-related circRNAs.

Main Methods:

• Constructed circRNA-circRNA and disease-disease similarity networks.
• Employed a hybrid graph embedding model combining Graph Convolutional Networks and Graph Attention Networks.
• Developed a prediction model using learned circRNA and disease representations.

Main Results:

• GraphCDA outperformed state-of-the-art methods in predicting circRNA-disease associations.
• The model demonstrated strong performance even with limited training data.
• Case studies confirmed GraphCDA's capability in identifying potential disease-related circRNAs.

Conclusions:

• GraphCDA is a reliable computational tool for exploring circRNA roles in complex diseases.
• The framework aids in discovering novel circRNA-disease associations.
• This approach accelerates research into the regulatory functions of circRNAs.