DeepRKE

DeepRKE predicts RNA binding protein (RBP) binding sites by integrating primary RNA sequence and secondary structure information using a deep neural network to improve large-scale identification of RBP–RNA interactions.

Key Features:

• Integration of Sequence and Structure: Combines primary RNA sequences and secondary structure information to enhance prediction accuracy of RBP binding sites.
• Advanced Feature Extraction: Uses a word embedding algorithm to extract distributed k-mer representations from RNA sequences and secondary structures instead of one-hot encoding.
• Deep Learning Architecture: Employs convolutional neural networks (CNNs) to detect local sequence and structure patterns and bidirectional long short-term memory networks (BiLSTMs) to capture long-range dependencies.
• Performance Superiority: Demonstrates improved performance over existing methods on two large-scale benchmark datasets in reported experimental results.
• Generalized Representations: Learns generalized representations from DNA and protein sequences that contribute to predictive performance.

Scientific Applications:

• Post-transcriptional Gene Regulation: Predicts RBP binding sites to study mechanisms such as splicing, mRNA transport, translation modulation, and decay.
• Mapping RBP–RNA Interaction Networks: Aids elucidation of regulatory networks governing RNA metabolism by identifying putative RBP binding events.
• Disease-related RNA Processing: Supports investigation of diseases linked to dysregulated RNA processing by providing candidate RBP–RNA interactions for further study.

Methodology:

Integrates primary RNA sequence and secondary structure, extracts k-mer embeddings via a word embedding algorithm, processes inputs with CNN and BiLSTM layers, and is trained and evaluated on two large-scale benchmark datasets.