MultiRBP
Open Access
- 1 August 2021
- conference paper
- Published by Association for Computing Machinery (ACM)
- p. 1-9
Abstract
Protein-RNA binding plays vital roles in post-transcriptional gene regulation. High-throughput in vitro binding measurements were generated for more than 200 RNA-binding proteins, enabling the development of computational methods to predict binding to any RNA transcript of interest. In recent years, deep learning-based methods have been developed to predict RNA binding in vitro achieving state-of-the-art results. However, all methods train a single model per protein, under-utilizing the similarities in binding preferences shared by multiple RNA-binding proteins. In this work, we developed MultiRBP, a deep learning-based method to predict RNA binding of hundreds of proteins to a given RNA sequence. The innovation of MultiRBP is in its multi-task nature, i.e., predicting binding for hundreds of proteins at the same time. We trained MultiRBP on the RNAcompete dataset, the most comprehensive dataset of in vitro binding measurements. Our method outperformed extant methods in both in vitro and in vivo RNA-binding prediction. Our method achieved an average Pearson correlation of 0.692±0.17 for in vitro binding prediction, and a median AUROC of 0.668±0.09 for in vivo binding prediction. Moreover, by visualizing the learned binding preferences, MultiRBP provided more interpretable visualization than a single-task model. The code is publicly available at github.com/OrensteinLab/MultiRBP.Keywords
This publication has 20 references indexed in Scilit:
- Deep neural networks for interpreting RNA-binding protein target preferencesGenome Research, 2020
- RDense: A Protein-RNA Binding Prediction Model Based on Bidirectional Recurrent Neural Network and Densely Connected Convolutional NetworksIEEE Access, 2019
- Pyramid-Structured Depth MAP Super-Resolution Based on Deep Dense-Residual NetworkIEEE Signal Processing Letters, 2019
- A deep neural network approach for learning intrinsic protein-RNA binding preferencesBioinformatics, 2018
- Basset: learning the regulatory code of the accessible genome with deep convolutional neural networksGenome Research, 2016
- Predicting the sequence specificities of DNA- and RNA-binding proteins by deep learningNature Biotechnology, 2015
- Protein–RNA interactions: new genomic technologies and perspectivesNature Reviews Genetics, 2012
- RNAcontext: A New Method for Learning the Sequence and Structure Binding Preferences of RNA-Binding ProteinsPLOS Computational Biology, 2010
- The use of the area under the ROC curve in the evaluation of machine learning algorithmsPattern Recognition, 1997
- Multitask LearningMachine Learning, 1997