Inferring disease-associated circRNAs by multi-source aggregation based on heterogeneous graph neural network.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Circular RNAs (circRNAs) are increasingly recognized for their roles in disease pathogenesis.
• Their unique structure makes them promising biomarkers for diagnosis.
• Computational methods offer a cost-effective alternative to traditional experiments for identifying circRNA-disease associations.

Purpose of the Study:

• To develop an effective computational model for predicting circRNA-disease associations.
• To address the limitations of existing methods that overlook data heterogeneity.
• To leverage multi-source pathogenesis data for robust association inference.

Main Methods:

• Proposed a novel model, CDHGNN, utilizing edge-weighted graph attention and heterogeneous graph neural networks.
• Constructed integrated networks including circRNA, microRNA, disease, and heterogeneous networks from multi-source data.
• Employed an edge-weighted graph attention network for node feature representation and heterogeneous neural networks for association prediction.

Main Results:

• CDHGNN demonstrated superior accuracy in predicting circRNA-disease associations compared to state-of-the-art algorithms.
• Both edge-weighted graph attention and heterogeneous graph networks significantly improved model performance.
• Case studies validated CDHGNN's ability to identify specific molecular associations and regulatory relationships.

Conclusions:

• CDHGNN provides an effective and accurate computational approach for circRNA-disease association prediction.
• The model's ability to integrate heterogeneous data enhances the inference of complex biomolecular relationships.
• This work contributes a valuable tool for advancing our understanding of circRNA functions in pathogenesis.