Prediction of COMT Inhibitors Using Machine Learning and Molecular Dynamics Methods.

Area of Science:

• Biochemistry
• Neuroscience
• Computational Chemistry

Background:

• Catechol O-methyltransferase (COMT) regulates neurotransmitter levels, and its dysfunction is linked to neurological conditions like Parkinson's disease.
• COMT inhibitors block enzyme activity by binding to the active site, offering a therapeutic strategy for managing neurotransmitter imbalances.

Purpose of the Study:

• To design novel COMT inhibitors using machine learning and computational methods.
• To predict the inhibitory activity and binding stability of newly designed molecules.

Main Methods:

• Development of classification models for COMT inhibitors.
• Application of regression techniques (Random Forest, AdaBoost, gradient boosting, SVM) for activity prediction.
• Molecular dynamics (MD) simulations and QM/MM calculations to assess binding stability and methyl transfer energy barriers.

Main Results:

• Machine learning models achieved R² > 70% for predicting inhibitor activity.
• MD simulations confirmed the stability of known and designed inhibitors within the COMT binding pocket.
• Free energy calculations provided insights into the methyl transfer mechanism.

Conclusions:

• Machine learning effectively aids in the design of novel COMT inhibitors.
• Computational methods are valuable for predicting inhibitor efficacy and stability.
• This study provides a foundation for developing new therapeutics for COMT-related neurological disorders.