Benchmarking Docking Protocols on Predicting Alternative Binding Modes

Area of Science:

• Computational Chemistry
• Structural Biology
• Drug Discovery

Background:

• Protein-ligand binding pose accuracy is critical for affinity prediction.
• High-throughput virtual screening prioritizes speed over pose accuracy.
• Docking protocols are used in lead optimization, but their accuracy with template structures is less understood.

Purpose of the Study:

• To assess the ability of templated docking protocols to predict alternative ligand binding modes.
• To evaluate pose prediction methods when a reference ligand structure is known.
• To focus on cases with multiple potential binding modes for new compounds.

Main Methods:

• Collected 60 complex structures from the Protein Data Bank (PDB) with multiple ligand binding modes.
• Assessed templated docking protocols using these structures as benchmarks.
• Compared success rates to standard docking benchmarks.

Main Results:

• Templated docking protocols showed modest success rates of 30-50% in identifying alternative binding modes.
• These success rates are significantly lower than those on standard diverse sets (70-90%).
• Ligand modifications breaking local symmetry pose challenges for prediction.

Conclusions:

• Current docking methods require further improvement for effective use in lead optimization.
• Enhanced pose prediction capabilities are needed to identify viable alternate binding modes.
• The study highlights limitations in applying standard docking benchmarks to lead optimization scenarios.