The interactions between proteins and RNAs play essential roles in many important biological processes. While the detection of protein-RNA interaction by biological experiments is a laborious and time-consuming task, computational prediction tools are highly in need. The prediction performance of the computational tools rely on two factors, namely feature representation of RNA sequences and classification models. In the existing methods, statistical features or one-hot vectors are adopted, and most of the classifiers are traditional machine learning models, while the distributed representation and flexible deep learning architectures have not been exploited. Therefore, in this study, we represent RNA sequences by continuous distributed features, and propose a hybrid deep learning architecture, which combines both CNN and RNN. The experiments are conducted on 31 benchmark datasets, corresponding to 31 RNA-binding-proteins. The results show that the new method achieves obvious advantages against the existing methods on most of the datasets.

, ,
doi.org/10.1145/3240876.3240912, hdl.handle.net/1765/111781
10th International Conference on Internet Multimedia Computing and Service, ICIMCS 2018
Department of Medical Informatics

Zhang, K. (Kaiming), Xiao, Y. (Yiqun), Pan, X., & Yang, Y. (Yang). (2018). Prediction of RNA-protein interactions with distributed feature representations and a hybrid deep model. In ACM International Conference Proceeding Series. doi:10.1145/3240876.3240912