Computational Method Using Attribute-Aware Message Passing and Graph Convolutional Network for Potential...

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• MicroRNA (miRNA) dysregulation is implicated in human diseases, particularly cancers.
• Identifying miRNA-disease associations is crucial for understanding pathogenesis and developing treatments.
• Computational prediction offers an efficient alternative to costly experimental methods.

Purpose of the Study:

• To develop a novel computational model, Attribute-Aware Message Passing Graph Convolutional Network (AAMPGCN), for predicting miRNA-disease associations.
• To integrate miRNA and disease attribute information into a graph convolutional network framework.
• To improve the accuracy and efficiency of miRNA-disease association prediction.

Main Methods:

• Developed the Attribute-Aware Message Passing Graph Convolutional Network (AAMPGCN) model.
• Constructed a miRNA-disease heterogeneous graph incorporating functional and semantic similarities.
• Partitioned the graph into attribute-homogeneous subgraphs to filter noise and preserve embedding discriminability.

Main Results:

• The AAMPGCN model achieved high performance with AUC of 94.06% and AUPR of 93.52% on the HMDD2.0 dataset.
• The model demonstrated superior performance compared to existing methods for miRNA-disease association prediction.
• Attribute integration effectively filtered cross-attribute noise and preserved node embedding discriminability.

Conclusions:

• AAMPGCN provides a novel and effective computational approach for predicting miRNA-disease associations.
• The model offers valuable insights into disease molecular mechanisms.
• It supports the identification of potential clinical therapeutic targets for various diseases.