TargetS

TargetS predicts protein-ligand binding sites directly from primary amino acid sequences without relying on structural templates, enabling binding-site identification when 3D structures or homologous templates are unavailable.

Key Features:

• Template-Free Prediction: Performs binding-site prediction independently of structural templates using sequence information.
• Ligand-Specific Strategy: Applies a ligand-specific approach to predict binding residues along the protein sequence.
• Recursive Spatial Clustering Algorithm: Uses a recursive spatial clustering algorithm to delineate distinct binding pockets or sites from predicted binding residues.
• Feature Construction: Constructs discriminative features from protein evolutionary information, predicted protein secondary structures, and ligand-specific binding propensities of residues.
• Improved AdaBoost Classifier Ensemble: Employs an enhanced AdaBoost classifier ensemble with random undersampling to address class imbalance between binding and non-binding residues.

Scientific Applications:

• Protein Function Analysis: Identifies ligand-binding residues to support molecular-level protein function inference.
• Drug Design: Pinpoints potential ligand-binding regions to inform the design of molecules that interact with specific proteins.

Methodology:

Predicts binding residues from primary sequences using a ligand-specific strategy; constructs features from protein evolutionary information, predicted protein secondary structures, and ligand-specific residue propensities; classifies residues with an improved AdaBoost ensemble using random undersampling; refines residue-level predictions into binding pockets/sites via recursive spatial clustering.