Voronoi Binding Site Models for Ligand Conformation and Energy Prediction

Gordon M. Crippen from the University of Michigan presents a novel computational method for modeling protein binding sites. The approach uses supplied ligand structures and observed binding free energies to predict binding conformations and energies for new compounds. The method was illustrated by modeling the binding of five polychlorinated biphenyls to thyroxine binding prealbumin and correctly predicted energies for six additional biphenyls.

Use Cases

• Predicting binding free energies for new compounds based on ligand structures and observed energies mentioned in the description
• Modeling the three-dimensional conformation and positioning of ligands within a binding site as described
• Developing compact representations of allowed ligand conformation space for computational efficiency
• Validating proposed site geometries against experimental binding data

Strengths

• Method demonstrated by fitting observed binding energies for five polychlorinated biphenyls without outliers
• Successfully predicted binding energies for six additional biphenyls in the illustrative example
• Algorithm requires only ligand structures, binding free energies, and a proposed site geometry with no further assumptions

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download
• Column-level documentation is absent; field semantics must be inferred after download
• Row count is unknown, which may limit suitability assessment

Provenance

Source

Gordon M. Crippen, University of Michigan

Collection Method

Novel computational algorithm for modeling binding sites, as described in the associated paper.