Predicting circRNA-RBP Binding Sites Using a Hybrid Deep Neural Network.

Area of Science:

• Bioinformatics
• Molecular Biology
• Genomics

Background:

• Circular RNAs (circRNAs) are key regulators in biological processes and diseases, interacting with RNA-binding proteins (RBPs).
• Predicting circRNA-RBP interactions is crucial for understanding gene regulation, but existing computational methods often rely on limited feature extraction.
• Traditional experimental methods for identifying these interactions are time-consuming and expensive.

Purpose of the Study:

• To develop a novel computational model, circ-FHN, for predicting circRNA-RBP interactions solely based on circRNA sequences.
• To enhance the accuracy and efficiency of circRNA-RBP interaction prediction by addressing the limitations of single-feature extraction methods.

Main Methods:

• The circ-FHN model utilizes a hybrid deep learning architecture combining a Convolutional Neural Network (CNN) for high-level feature learning and a Bidirectional Gated Recurrent Unit (BiGRU) for capturing sequential dependencies.
• Sequence features are extracted using four distinct coding methods that incorporate the physicochemical properties of circRNA sequences.

Main Results:

• circ-FHN demonstrated superior performance compared to existing computational methods across 16 diverse datasets.
• Ablation experiments and motif analysis validated the effectiveness of the proposed feature extraction and hybrid deep learning approach.
• The model achieved exceptional accuracy in predicting circRNA-RBP interactions.

Conclusions:

• The circ-FHN model offers a powerful and accurate computational tool for predicting circRNA-RBP interactions using only sequence information.
• This approach provides a cost-effective and efficient alternative to experimental methods for studying circRNA function and disease relevance.
• The developed model and its source code are publicly available for further research and application.