Predicting Regioselectivity of AO, CYP, FMO, and UGT Metabolism Using Quantum Mechanical Simulations and Machine...

Area of Science:

• Drug metabolism and pharmacokinetics
• Computational chemistry
• Medicinal chemistry

Background:

• Drug candidate failure in late stages and withdrawal of approved drugs are often caused by unexpected metabolism.
• Early prediction of metabolism sites (SoM) is critical for efficient drug discovery and development.
• Existing methods may lack isoform-specificity or accuracy for diverse enzyme families.

Purpose of the Study:

• To develop and validate computational models for predicting isoform-specific drug metabolism sites.
• To cover human alcohol (AOs), flavin-containing monooxygenases (FMOs), and UDP-glucuronosyltransferases (UGTs), as well as general CYP metabolism in preclinical species.
• To enhance early-stage drug discovery by providing accurate SoM predictions.

Main Methods:

• Utilizing semi-empirical quantum mechanical simulations to estimate the reactivity of potential SoM.
• Validating simulation methods with experimental data and Density Functional Theory (DFT) calculations.
• Developing ligand-based models incorporating SoM reactivity, orientation, and steric effects within enzyme binding pockets.

Main Results:

• Achieved high predictive performance with kappa (κ) values up to 0.94.
• Demonstrated strong classification accuracy with Area Under the Curve (AUC) values up to 0.92.
• Successfully integrated quantum mechanical and ligand-based approaches for improved SoM prediction.

Conclusions:

• The developed models offer accurate and reliable prediction of drug metabolism sites for specific human enzymes and preclinical species.
• These computational tools can significantly de-risk drug development by identifying potential metabolic liabilities early.
• The study provides a valuable resource for medicinal chemists and pharmacologists to guide drug design and selection.