DeCban

DeCban predicts interactions between circular RNAs (circRNAs) and RNA binding proteins (RBPs) using deep learning methods to computationally identify circRNA–RBP interactions.

Key Features:

• Hybrid Double Embeddings: Employs hybrid double embeddings that incorporate pre-trained embedding vectors for RNA segments and their corresponding amino acids to capture comprehensive sequence information.
• Cross-Branch Attention Neural Network: Uses a cross-branch attention neural network for classification that integrates features across different scales and focuses on critical information within very long sequences.
• Benchmark Training: Trained and evaluated on benchmark datasets comprising 37 different sets.
• Improved Performance: Demonstrates enhanced prediction accuracy and computational efficiency compared to mainstream deep learning-based methods.

Scientific Applications:

• circRNA–RBP interaction prediction: Computational identification of interactions between circular RNAs (circRNAs) and RNA binding proteins (RBPs).
• Functional studies of circRNAs: Supports investigation of circRNA roles in gene regulation and disease mechanisms by providing predicted RBP partners.

Methodology:

Training deep learning models on 37 benchmark datasets using hybrid double embeddings (pre-trained RNA-segment and amino-acid vectors) combined with a cross-branch attention neural network for classification and handling very long sequences, with attention mechanisms to focus on essential sequence features.