GCNCMI: A graph convolutional neural network approach for predicting circRNA-miRNA interactions
numpy==1.16.6
pandas==1.1.5
scikit-learn==0.24.2
tensorflow-gpu==1.15.0
Interactions between circular RNAs (circRNAs) and microRNAs (miRNAs) have been shown to alter gene expression and regulate disease genes. Most circRNA-miRNA interactions remain largely unknown because traditional experimental approaches are time-consuming and labor-intensive. Developing computational methods to explore the interactions between circRNAs and miRNAs at scale could help bridge this gap. In this paper, we propose a graph convolutional neural network-based method named GCNCMI to predict potential interactions between circRNAs and miRNAs. GCNCMI first mines the latent interactions of adjacent nodes in a graph convolutional neural network, and then recursively propagates the interaction information on the graph convolutional layers. Finally, it combines the embedded representations generated by each layer to make the final prediction. In five-fold cross-validation, GCNCMI achieved the highest AUC of 0.9312 and the highest AUPR of 0.9412. In addition, case studies on two miRNAs, hsa-miR-622 and hsa-miR-149-5p, showed that our model has a good effect on predicting circRNA-miRNA interactions.
Our dataset contains 821 miRNAs and 2115 circRNAs, which we divide into a 5-fold cross-validation dataset.
1.The GCNCMI.conf file in the config directory can modify the parameters of the model.
2.Run the main function in GCNCMI to start training.
3.If you want to reuse our code, you can replace your dataset in the dataprocess method, but you need to preprocess the dataset first.