TSNAPred

TSNAPred predicts type-specific nucleic acid-binding residues in proteins to distinguish residues binding A-DNA, B-DNA, ssDNA, mRNA, tRNA, and rRNA.

Key Features:

• Type-Specific Prediction: Identifies residues that bind A-DNA, B-DNA, ssDNA, mRNA, tRNA, and rRNA.
• Ensemble Approach: Combines LightGBM and a capsule network trained on sequence-derived features to generate predictions.
• Sliding Window Technique: Applies a sliding-window to capture long-distance dependencies between residues along protein sequences.
• Weighted Ensemble Strategy: Integrates outputs from component models using a weighted combination to improve prediction accuracy.
• Sequence-Derived Feature Extraction: Derives features from protein sequences to serve as input for model training.

Scientific Applications:

• Protein function and interaction networks: Distinguishes nucleic-acid-binding residues to inform studies of protein function and interaction networks.
• DNA vs RNA binding discrimination: Enables residue-level differentiation of DNA-binding and RNA-binding sites for mechanistic analyses.
• Experimental design and therapeutics: Provides residue-level targets to aid experimental planning and development of interventions targeting protein–nucleic-acid interactions.

Methodology:

Feature extraction from protein sequences; training LightGBM and capsule networks on these features; applying a sliding-window technique to capture long-distance residue dependencies; integrating model outputs via a weighted ensemble strategy.