RNA Secondary Structure Representation Network for RNA-proteins Binding Prediction

RNA-binding proteins (RBPs) play a significant part in several biological processes the living cell, such as gene regulation and mRNA localization. Several deep learning methods, especially model based on convolutional neural network(CNN), have been used to predict binding sites. However, previous methods fail represent RNA secondary structure features. The traditional generally transform regular matrix that cannot reveal topological information of RNA. To effectively extract features RNA, we propose an representation network (RNASSR-Net) graph (GCN) convolution (CNN) for RBP prediction. RNASSR-Net constructs derived from learn properties Then, it obtains spatial importance each base with CNN guide structure. Finally, combines sequence Experimental results demonstrate proposed method outperforms few state-of-the-art benchmark datasets gets higher improvement small-size data. Besides, is also detect accurate motifs compared experimentally verified motifs, which reveals region location interpretation some guidance future.