GATCDA: Predicting circRNA-Disease Associations Based on Graph Attention Network.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Circular RNAs (circRNAs) are non-coding RNA molecules implicated in disease development.
• Experimental prediction of circRNA-disease interactions is time-consuming.
• Computational approaches offer a more efficient alternative for predicting these associations.

Purpose of the Study:

• To develop and validate a novel computational method, GATCDA, for predicting circRNA-disease associations.
• To leverage graph attention networks (GAT) for enhanced prediction accuracy.
• To explore the utility of disease symptom similarity, network similarity, and information entropy in predicting these interactions.

Main Methods:

• Developed GATCDA, a computational method utilizing Graph Attention Networks (GAT).
• Incorporated circRNA-miRNA and disease-mRNA interactions to construct features, considering the circRNA-miRNA-mRNA axis.
• Employed disease symptom similarity, network similarity, and information entropy similarity for both circRNAs and diseases.

Main Results:

• GATCDA achieved a high Area Under the Curve (AUC) value of 0.9011 based on five-fold cross-validation.
• Case studies demonstrated GATCDA's ability to identify previously unknown circRNA-disease associations.
• The method effectively utilizes attention mechanisms to learn node representations in biological networks.

Conclusions:

• GATCDA is a robust and effective computational tool for predicting circRNA-disease associations.
• The developed method offers a valuable approach for advancing research into the roles of circRNAs in disease.
• GATCDA provides a promising avenue for exploring complex biological interactions relevant to human health.