Comprehensive and Automated Linear Interaction Energy Based Binding-Affinity Prediction for Multifarious Cytochrome...

Area of Science:

• Computational chemistry
• Drug discovery
• Biochemistry

Background:

• Estrogen-dependent breast cancer treatment relies on aromatase inhibitors.
• Developing potent, selective, and safer inhibitors is crucial.
• In silico screening accelerates the identification of lead compounds.

Purpose of the Study:

• To create robust binding affinity prediction models for Cytochrome P450 aromatase (CYP19A1).
• To utilize an automated Linear Interaction Energy (LIE) workflow combined with machine learning.
• To enhance the discovery of novel aromatase inhibitors.

Main Methods:

• Applied an automated LIE workflow to 132 diverse aromatase inhibitors.
• Integrated machine learning for compound clustering and model optimization.
• Used Molecular Dynamics (MD) trajectories for protein-ligand interaction profiles.

Main Results:

• Developed 3 robust predictive models covering 86% of the dataset.
• Achieved high correlation between calculated and observed ligand-binding free energies (RMSE < 2.5 kJ mol⁻¹).
• Demonstrated good cross-validation statistics, indicating model reliability.

Conclusions:

• The LIE-based workflow with machine learning effectively predicts aromatase inhibitor binding affinity.
• This computational approach aids in identifying promising drug candidates for breast cancer therapy.
• The models provide a reliable tool for in silico screening in drug discovery.