Validation and extraction of molecular-geometry information from small-molecule databases

Area of Science:

• Crystallography
• Structural Biology
• Computational Chemistry

Background:

• Macromolecular modeling and structure refinement rely on accurate small-molecule ligand descriptions.
• The Crystallography Open Database (COD) is a valuable resource for molecular geometry information.

Purpose of the Study:

• To enhance the reliability of molecular geometry data extracted from the COD.
• To generate improved ligand descriptions for macromolecular software.

Main Methods:

• Implemented strict validation criteria for COD entries, including resolution, structure solution method, valence consistency, bond length deviations, and atomic collisions.
• Utilized high-order moment-based statistical techniques for validating atom types and bond classes.
• Iteratively refined atom typing and bond/angle classes through a four-step process.

Main Results:

• Developed a robust procedure for selecting high-quality crystal structures from the COD.
• Generated fine-grained atom typing, bond, and angle classes based on validated data.
• Demonstrated the potential to apply this procedure to other small-molecule databases.

Conclusions:

• Strict validation of the COD significantly increases the reliability of extracted molecular geometry data.
• The refined data and derived classes enhance the accuracy of ligand descriptions for macromolecular modeling.
• The methodology is adaptable for improving data quality from diverse small-molecule structure sources.