RNA-binding protein recognition based on multi-view deep feature and multi-label learning.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• RNA-binding proteins (RBPs) regulate biological processes and their aberrant expression is linked to diseases.
• Existing methods for predicting RNA-RBP interactions often lack accuracy due to insufficient feature learning and failure to account for RBP binding similarities.
• Multi-label learning and Long Short-Term Memory networks have been explored but show limited success.

Purpose of the Study:

• To develop a novel computational method, iDeepMV, for accurate prediction of RNA-RBP interactions.
• To leverage multi-view deep learning and RNA-RBP Binding Network (RRBN) concepts for improved multi-label learning.
• To enhance the identification of RBPs that bind to specific RNAs, aiding in understanding disease mechanisms.

Main Methods:

• Proposed the RNA-RBP Binding Network (RRBN) to support multi-label learning for RNA-RBP interaction prediction.
• Developed iDeepMV, integrating multi-view deep learning with a multi-label learning framework.
• Extracted features from amino acid sequences and dipeptide components, applied deep neural networks for feature learning, and utilized multi-label classifiers with a voting mechanism.

Main Results:

• RRBN information significantly improves the prediction of unknown RNA-RBP interactions.
• iDeepMV demonstrates significantly better prediction performance compared to state-of-the-art methods.
• The method effectively combines multi-view deep feature learning with RNA-RBP interaction data.

Conclusions:

• iDeepMV offers a powerful and accurate approach for predicting RNA-RBP interactions.
• The integration of multi-view deep learning and RRBN concepts advances the field of RNA-RBP binding prediction.
• iDeepMV provides a valuable tool for researchers studying RNA-protein interactions and associated diseases.