A joint model for predicting circRNA-RBP binding sites based on deep learning

The interaction between circRNAs and RBPs is related to many diseases especially cancers. Understanding the mechanism of the interaction is critical and predicting the binding sites of them is helpful for this. Existing methods for predicting circRNA-RBP binding sites have some limitations. They cannot fully learn and utilize features of circRNA sequences and the performance of them needs to be further improved.

To overcome these limitations, Guo Ya-Jing et al. published their new research on 15 Oct 2023 in Frontiers of Computer Science co-published by Higher Education Press and Springer Nature.

The authors proposed a novel deep learning-based model named circ2CBA which considering the context information between sequence nucleotides of circRNAs and the important position weight information of features. circ2CBA fully learns the features of circRNA sequences using various deep learning methods. The results of comparative and ablation experiments show that circ2CBA can achieve a good performance.

circ2CBA uses only sequence information of circRNAs to predict the binding sites between circRNAs and RBPs. The circRNA sequences are obtained from the CircInteractome database. Eight RBPs are selected to construct the data set.

One-hot method is used to encode circRNA sequences which are fed into subsequent models as input information. circ2CBA first utilizes two layers of CNN to extract the local features of circRNAs and to get a larger perception domain. To capture the context dependent information between sequence nucleotides, circ2CBA then introduces a BiLSTM network. Next, an attention layer is used to assign different weights to feature matrix before feeding it to the two-layer fully connected layer. Finally, the prediction result is obtained through a softmax function. Comparison experiments with other latest methods and ablation experiments are performed on the same dataset. Besides, A motif analysis is made to explore the reason for the remarkable performance improvement of circ2CBA on some sub-datasets. The experimental results show that circ2CBA is an effective method to predict the binding sites of circRNA-RBP.

DOI: 10.1007/s11704-022-2151-0