GCNCMI: A Graph Convolutional Neural Network Approach for Predicting circRNA-miRNA Interactions

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Circular RNAs (circRNAs) and microRNAs (miRNAs) interactions are vital in gene expression regulation and disease pathogenesis.
• Experimental identification of circRNA-miRNA interactions is challenging due to time and labor constraints.
• Computational methods are needed for large-scale prediction of these interactions.

Purpose of the Study:

• To develop an efficient computational approach for predicting circRNA-miRNA interactions.
• To leverage graph convolutional neural networks (GCNs) for interaction prediction.
• To provide a valuable tool for understanding gene regulation in diseases.

Main Methods:

• Proposed GCNCMI, a novel graph convolutional neural network-based model.
• GCNCMI mines potential interactions of adjacent nodes and propagates information through graph convolutional layers.
• The model integrates embedded representations from each layer for final prediction.

Main Results:

• GCNCMI achieved high performance in five-fold cross-validation, with an AUC of 0.9312 and AUPR of 0.9412.
• Case studies on hsa-miR-622 and hsa-miR-149-5p demonstrated the model's effectiveness.
• The method successfully predicts potential circRNA-miRNA interactions.

Conclusions:

• GCNCMI offers an effective computational strategy for predicting circRNA-miRNA interactions.
• The model can accelerate the discovery of disease-related gene regulatory mechanisms.
• The study provides open-source code and data for further research.