DeepDRBP-2L

DeepDRBP-2L predicts DNA-binding proteins (DBPs), RNA-binding proteins (RBPs), and dual DNA/RNA-binding proteins (DRBPs) from protein sequences to enable accurate identification for studies of gene expression regulation and disease associations.

Key Features:

• Dual-Level Architecture: Integrates Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks to capture local sequence patterns and long-range dependencies.
• Comprehensive Identification: Distinguishes DBPs, RBPs, and DRBPs to reduce misclassification between DNA- and RNA-binding proteins.
• Enhanced Precision: Demonstrates improved predictive performance through cross-validation and independent testing.
• Validation on Tomato Genome: Applied to the tomato genome as a practical validation of prediction capability.
• Addresses Sequence-Based Limitations: Designed to overcome limitations of existing sequence-based prediction methods.

Scientific Applications:

• Gene Expression Studies: Enables identification of binding proteins involved in transcriptional and post-transcriptional regulation.
• Disease Association Research: Supports investigation of DBP/RBP/DRBP roles in disease mechanisms and potential therapeutic targets.
• Genomic Annotation: Assists in annotation of protein function in genome-scale studies, exemplified by application to the tomato genome.

Methodology:

Uses CNNs for feature extraction from protein sequences, capturing spatial hierarchies, and employs LSTMs to model temporal (long-range) dependencies within a two-level architecture, with performance assessed by cross-validation and independent testing.