CheckMyBlob

CheckMyBlob identifies and validates ligands in X-ray electron density maps to support structure-guided drug design.

Key Features:

• Machine learning-based ligand recognition: Employs a machine learning algorithm that generalizes ligand descriptions from extensive datasets of moieties in the Protein Data Bank (PDB) to identify ligands in electron density blobs.
• Detection of unmodeled fragments: Detects and interprets unmodeled fragments of X-ray electron density maps to locate candidate ligand sites.
• File format support: Processes PDB/mmCIF and MTZ file formats for input electron density and coordinate data.
• Candidate ranking: Generates a ranked list of the 10 most likely ligand candidates for each detected electron density blob.
• Coot plugin script generation: Produces plugin scripts for Coot to enable downstream analysis and model inspection.

Scientific Applications:

• Structure-guided drug design: Provides candidate ligand identifications to inform structure-based lead optimization and design decisions.
• Ligand model validation: Aids validation of proposed ligand models against experimental electron density to improve model accuracy.
• Exploration of unmodeled regions: Supports discovery and characterization of unmodeled ligands or fragments in crystallographic maps.
• Reduction of interpretation bias: Helps mitigate confirmation bias in manual electron density interpretation by offering data-driven candidate suggestions.

Methodology:

The method trains a machine learning model on known ligand structures and moieties from the PDB to learn features for recognizing ligands in electron density maps, predicts and ranks candidate ligands for detected density blobs, and outputs Coot plugin scripts for further analysis.