A Review About RNA–Protein-Binding Sites Prediction Based on Deep Learning

Jianrong Yan,Min Zhu

doi:10.1109/access.2020.3014996

Jianrong Yan, Min Zhu

Open Access

https://doi.org/10.1109/access.2020.3014996

Copy DOI

Abstract

RNA-binding proteins (RBPs) play crucial roles in gene regulation. The advent of high-throughput experimental methods, has generated a huge volume of experimentally verified binding sites of RNA-binding proteins and greatly advanced the genome-wide studies of RNA-protein interactions. Many computational approaches have been proposed, including deep learning models, which have achieved remarkable performance on the identification of RNA-protein binding affinities and sites. In this review, we discuss machine learning and deep learning approaches, mainly focusing on the prediction of RNA and proteins binding sites on RNAs by deep learning. Furthermore, we discuss the advantages and disadvantages of these approaches. The workflow of deep learning is also revealed. We recommend some promising future directions of deep learning models in the study of RBP-binding sites on RNAs, especially the embedding, generative adversarial net, and attention model. Extraction and visualization methods involving motif are illustrated. Finally, we summarize the previous studies, and then compare the performance on different dataset.

Highlights

RNA-binding proteins (RBPs) play important roles in various cellular processes, such as alternative splicing, RNA editing, and mRNA localization [1], [2]
We summarize the recent progress of RBP binding sites, focusing on deep learning methods
Word emdedding about Word2vec was applied to iDeepV [78] to represent k-mers in a lower dimensional space, attaching 1-D convolutional neural network (CNN) to predict the RBP binding sites

Summary

INTRODUCTION

RNA-binding proteins (RBPs) play important roles in various cellular processes, such as alternative splicing, RNA editing, and mRNA localization [1], [2]. Li et al [20] predicted protein-RNA binding residues using deep boosting-based approach via a total of 168 sequence features. We discuss the challenges and potential defects of the RBP binding sites prediction method based on deep learning. In multi-classification, the number of labels is the sum of all the specific RBP types plus the non-binding sites, with only one general model to make prediction unlike the specific model. A specific protein binds to a set of predesigned oligomers and measures binding by hybridizing with complementary probes on the microarray This dataset is used by many models, such as GraphProt [18], RCK [19], Deepbind [23], RNAContext [17], DLPRB [34], etc. We will introduce stacked deep neural network model for predicting RBP binding sites, such as CNN-LSTM, Embedding, GAN, attention model, etc

CONVENTIONAL DEEP LEARNING MODELS

APPLICATION OF DEEP LEARNING IN THE PREDICTION OF RBP BINDING SITES

EXTRACTING AND VISUALIZING BINDING MOTIFS

VALIDATION METRICS

Findings

DISCUSSION