Is Structure-Based Drug Design Ready for Selectivity Optimization?

Area of Science:

• Computational chemistry
• Drug discovery
• Molecular modeling

Background:

• Alchemical free-energy calculations (AFECs) are established for optimizing small-molecule potency.
• Their application in predicting compound selectivity across targets remains less explored.
• Binding site similarity may enhance AFEC accuracy for selectivity prediction due to error cancellation.

Purpose of the Study:

• To evaluate the accuracy of AFECs for predicting selectivity of kinase inhibitors.
• To investigate selectivity prediction between closely related kinases (CDK2/CDK9) and distantly related ones (CDK2/ERK2).
• To analyze the impact of systematic error correlation on selectivity prediction accuracy.

Main Methods:

• Utilized alchemical free-energy calculations for kinase inhibitor selectivity prediction.
• Employed a Bayesian analysis approach to differentiate systematic and statistical errors.
• Quantified the correlation of systematic errors across different kinase pairs.

Main Results:

• High systematic error correlation between CDK2 and CDK9 suggests AFECs can significantly aid selectivity optimization.
• For CDK2/ERK2, error correlation indicates potential for fortuitous cancellation even in less related kinases.
• Longer simulations are beneficial for balancing statistical and systematic errors in selectivity predictions.

Conclusions:

• AFECs show promise for guiding compound selectivity in drug discovery, especially between similar targets.
• Understanding systematic error correlations is crucial for maximizing AFEC utility in selectivity prediction.
• Simulation length optimization is key to leveraging AFECs for accurate selectivity predictions.