KGRACDA: A Model Based on Knowledge Graph from Recursion and Attention Aggregation for CircRNA-Disease Association...

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Circular RNAs (circRNAs) are increasingly recognized for their roles in human diseases.
• Accurate prediction of circRNA-disease associations (CDAs) is vital for understanding disease mechanisms.
• Existing computational methods, particularly deep learning, often overlook the extraction of local depth information in graph structures.

Purpose of the Study:

• To develop a novel computational model, KGRACDA, for predicting circRNA-disease associations.
• To integrate explicit structural features and implicit graph embedding information for enhanced prediction accuracy.
• To address the limitations of existing methods in capturing local depth information.

Main Methods:

• Construction of a large-scale, multi-source heterogeneous knowledge graph incorporating RNAs and diseases.
• Utilizing a recursive method to generate multi-hop subgraphs for mining local depth information.
• Employing an optimized graph attention mechanism with a gating mechanism and a multi-head attention mechanism to balance global and local graph features.

Main Results:

• The proposed KGRACDA model effectively captures both local and global depth features within the knowledge graph.
• KGRACDA demonstrates superior performance compared to existing methods in predicting circRNA-disease associations.
• An updated interactive web platform, HNRBase v2.0, is provided for data visualization and CDA prediction.

Conclusions:

• KGRACDA offers a powerful new approach for circRNA-disease association prediction by leveraging knowledge graphs and attention mechanisms.
• The model's ability to mine local and global depth information enhances prediction accuracy.
• The HNRBase v2.0 platform facilitates access to circRNA data and utilization of the KGRACDA model for research.