Efficient conformational sampling and weak scoring in docking programs? Strategy of the wisdom of crowds

Area of Science:

• Computational chemistry
• Drug discovery
• Structural biology

Background:

• Accurate protein-ligand interactions are crucial for structure-based drug design.
• Docking programs are essential for predicting ligand poses when experimental data is unavailable.
• Evaluating the performance of popular docking software is vital for successful drug optimization.

Purpose of the Study:

• To assess the performance of four leading docking programs: Gold, Glide, Surflex, and FlexX.
• To investigate the accuracy of ligand pose prediction and ranking in protein-ligand complexes.
• To explore methods for improving docking accuracy, including ensemble approaches.

Main Methods:

• Utilized 100 crystal structures of protein-ligand complexes from the Directory of Useful Decoys-Enhanced (DUD-E) database.
• Evaluated the semi-rigid docking capabilities of Gold, Glide, Surflex, and FlexX.
• Applied the United Subset Consensus (USC) method to combine results from multiple docking programs.

Main Results:

• Surflex achieved the highest success rate in predicting correct ligand poses (up to 84 complexes).
• Glide's scoring function (Glidescore) showed the best performance in ranking correct poses (up to 75 within top 4).
• The United Subset Consensus (USC) method improved results, yielding a correct pose in the top 4 for 87 complexes.
• No significant correlation was found between program performance and target/ligand properties, except for ligand rotatable bonds.
• Combining programs via USC was more effective than individual program rescoring or specific program combinations (e.g., Surflex-Glidescore).

Conclusions:

• Current scoring functions require improvement for accurate pose detection.
• Combining results from multiple docking programs, particularly using the USC method, enhances prediction accuracy.
• Ensemble docking strategies are recommended to increase the success rate in drug design and optimization.