DeepBSRPred

DeepBSRPred predicts protein binding site residues from protein sequence and AlphaFold2-predicted structures to identify interaction interface residues in protein–protein interactions.

Key Features:

• Deep Neural Network-Based Approach: Employs a deep neural network to predict binding site residues by integrating sequence and structural signals.
• Feature Integration: Utilizes position-specific scoring matrix (PSSM), solvent accessible surface area, conservation score, amino acid properties, and residue depth derived from sequences and AlphaFold2-predicted structures.
• Performance Metrics: Achieved an average F1 score of 0.73 on a dataset of 1236 proteins.
• Benchmarking: Outperformed existing methods on four benchmark datasets.

Scientific Applications:

• Protein–Protein Interaction Analysis: Identifies interface residues to support studies of PPIs and their roles in cellular processes.
• Functional and Affinity Inference: Provides residue-level information to assist inference of binding affinities and functional roles of protein complexes.
• Drug Discovery: Enables identification of specific binding sites that can be targeted in therapeutic development.

Methodology:

DeepBSRPred trains deep neural networks on datasets comprising protein sequences and AlphaFold2-predicted structures using integrated features including PSSM, solvent accessible surface area, conservation score, amino acid properties, and residue depth.