econvRBP

econvRBP predicts RNA-binding proteins from amino acid sequences using an ensemble convolutional neural network to identify RBPs for genome annotation and studies of post-transcriptional regulation.

Key Features:

• Ensemble Convolutional Neural Network: Employs an ensemble of convolutional neural networks to capture both local and global sequence-derived features for RBP prediction.
• Feature Encoding (One Hot and Conjoint Triad): Transforms amino acid sequences using One Hot encoding for local sequence information and Conjoint Triad encoding for broader structural context.
• High-Level Feature Extraction: Integrates local and global encoded features via CNNs to extract high-level protein patterns for classification.
• Robust Validation: Evaluated with 10-fold cross-validation on a dataset of 2,875 RBPs and 6,782 non-RBPs, achieving 99% accuracy.
• External Validation: Validated on RBPPred datasets with an accuracy of 0.87.

Scientific Applications:

• Genome Annotation: Predicts RBPs from sequence to support annotation of proteomes and identification of RNA-binding functions.
• Post-transcriptional Regulation Studies: Identifies candidate RBPs involved in RNA synthesis, transport, translation, and degradation for studies of post-transcriptional regulation.
• RBP Functional Characterization: Facilitates exploration of functional roles of RBPs through sequence-derived predictions.

Methodology:

Amino acid sequences are encoded with One Hot and Conjoint Triad encodings and processed by an ensemble of convolutional neural networks to extract high-level features; performance was assessed by 10-fold cross-validation and external validation on RBPPred datasets.