DEPTH

DEPTH computes atomic and residue depth and solvent-accessible surface area (SASA) from protein structures to predict small-molecule ligand binding cavities and characterize residue burial for analyses of stability, hydrogen exchange, interaction hotspots, post-translational modification sites, and sequence variability.

Key Features:

• Depth Calculation: Measures the distance from any atom or residue to the closest bulk water to quantify burial within a protein structure.
• Solvent-Accessible Surface Area (SASA): Computes SASA values to identify residues with burial/exposure patterns relevant to ligand binding cavities.
• Cavity Prediction: Uses depth–SASA pair values to estimate per-residue likelihood of participation in small-molecule binding cavities, with prediction performance comparable to geometry-based methods LIGSITE, SURFNET, and Pocket-Finder (Matthew's correlation coefficient ≈ 0.4).
• Parameter Tuning: Provides adjustable parameters to detect cavities of varying sizes and shapes, including geometrically flat binding sites.
• Evolutionary Information Integration: Incorporates evolutionary information to enhance binding site prediction specificity and precision relative to methods such as MetaPocket 2.0.
• pKa Prediction: Predicts pKa values for GLU, ASP, LYS, and HIS with an average error of less than one pH unit over 60 predictions.

Scientific Applications:

• Protein Functional Annotation: Identifies potential ligand binding sites to support annotation of protein function.
• Drug Discovery: Guides site-directed mutagenesis and small-molecule docking by identifying likely binding cavities and residue environments.
• Structural Analysis: Enables analysis of residue burial and solvent exposure to study protein stability, hydrogen exchange, interaction hotspots, post-translational modification sites, and sequence variability.

Methodology:

Computational calibration over a training set of 900 high-resolution X-ray crystal structures of single-domain proteins bound to small molecules using depth and SASA calculations and depth–SASA pair values; incorporation of evolutionary information for binding site prediction; evaluation against geometry-based methods (LIGSITE, SURFNET, Pocket-Finder) and comparison of specificity/precision to MetaPocket 2.0; pKa predictions for GLU, ASP, LYS, HIS assessed with average error < 1 pH unit over 60 predictions.