LISE

LISE predicts small-molecule binding sites on proteins by leveraging geometric motifs from interaction networks of protein and ligand atoms to capture spatial and physicochemical properties of interacting surface atoms.

Key Features:

• Geometric motif extraction: Derives geometric motifs from interaction networks that connect protein and ligand atoms.
• Interaction-network representation: Represents protein-ligand contacts as interaction networks to encode atom-level connectivity.
• Spatial and physicochemical characterization: Focuses on spatial and physicochemical properties of interacting surface atoms.
• Network-motif scoring: Scores candidate binding sites by counting occurrences of network-derived geometric motifs.
• Methodological contrast: Employs an approach distinct from structural similarity, cavity identification, or binding energy estimation methods.
• Benchmark validation: Evaluated on two small benchmark test sets and on targets from community-based ligand-binding site prediction experiments.
• Large-scale evaluation: Validated on a dataset of over 2,000 protein–ligand complexes to assess accuracy and reliability.
• Challenging-class identification: Identifies protein classes that are difficult to predict, providing realistic performance benchmarks.

Scientific Applications:

• Binding-site prediction: Predicts small-molecule binding sites on protein structures at atom-level resolution.
• Proteome-scale benchmarking: Provides realistic expectations of binding-site prediction performance across proteome-scale datasets.
• Drug discovery support: Prioritizes putative ligand-binding sites for downstream drug-design and screening workflows.
• Characterization of difficult targets: Detects protein classes that are systematically challenging for binding-site prediction methods.

Methodology:

Derives geometric motifs from interaction networks connecting protein and ligand atoms; characterizes spatial and physicochemical properties of interacting surface atoms; scores candidate sites by counting network motifs; contrasted with methods based on structural similarity, cavity identification, and binding energy estimation.