Cross-docking of inhibitors into CDK2 structures. 1

Area of Science:

• Computational chemistry and molecular modeling
• Drug discovery and development

Background:

• Predicting protein/ligand binding affinity is a critical but challenging task in computational chemistry.
• Existing methods often fail to adequately account for protein flexibility, limiting their accuracy.
• The protein kinase Cyclin-Dependent Kinase 2 (CDK2) is a key target in various diseases.

Purpose of the Study:

• To evaluate the performance of docking software (Gold and Glide) in predicting binding poses for CDK2 inhibitors.
• To assess the ability of scoring functions to predict relative binding potency in a cross-docking scenario.
• To investigate the impact of protein flexibility on docking accuracy and affinity prediction.

Main Methods:

• A cross-docking study was performed using a dataset of 150 inhibitor complexes of the protein kinase CDK2.
• The Gold and Glide docking programs were used to predict ligand poses.
• Scoring functions were evaluated for their ability to predict relative binding potency based on docked poses.

Main Results:

• Gold and Glide demonstrated good performance in terms of docking accuracy, achieving a 50% success rate for poses within 2 Å RMSD of the experimental structure.
• Despite accurate pose prediction, scoring functions failed to reliably predict relative binding potency.
• This indicates a disconnect between predicting binding pose and predicting binding affinity.

Conclusions:

• Docking accuracy for protein-ligand complexes can be achieved with current software, but this does not guarantee accurate binding affinity prediction.
• Protein flexibility remains a significant challenge in developing reliable scoring functions for binding affinity prediction.
• Further development of scoring functions that incorporate protein flexibility is crucial for improving drug discovery efforts.