Deciphering common failures in molecular docking of ligand-protein complexes

Area of Science:

• Computational chemistry
• Structural biology
• Drug discovery

Background:

• Predicting ligand-protein complex crystal structures is crucial for drug discovery.
• Ligand-protein docking simulations often fail to accurately predict these structures.
• Failures are categorized as 'soft' (search algorithm limitations) or 'hard' (energy function inaccuracies).

Purpose of the Study:

• Investigate common failures in predicting ligand-protein complex crystal structures.
• Analyze thermodynamic and kinetic factors influencing binding energy landscapes.
• Develop an improved protocol to overcome docking prediction limitations.

Main Methods:

• Combined thermodynamic and kinetic analysis of binding energy landscapes for three ligand-protein systems.
• Classification of misdocked predictions into 'soft' and 'hard' failures.
• A hierarchical protocol involving structural similarity clustering and AMBER force field energy refinement.

Main Results:

• Neither identifying the lowest energy structure nor the most common binding mode guarantees correct crystal structure prediction.
• The proposed hierarchical approach, using a hierarchy of energy functions, successfully resolves some common docking failures.
• The protocol detected crystallographic binding modes missed by standard docking simulations.

Conclusions:

• Standard docking approaches are insufficient for accurately predicting ligand-protein complex crystal structures.
• A hierarchical protocol combining conformational clustering and energy refinement enhances prediction accuracy.
• This method offers a more reliable way to identify biologically relevant binding modes.