Hands-On Docking with Molegro Virtual Docker.

Area of Science:

• Computational chemistry
• Drug discovery
• Machine learning in bioinformatics

Background:

• Protein-ligand docking is crucial for drug discovery.
• Predicting binding affinity and inhibition is computationally challenging.
• Cyclin-dependent kinase 2 (CDK2) is a key target in cancer research.

Purpose of the Study:

• To develop a workflow for building regression models to predict CDK2 inhibition.
• To integrate virtual docking simulations with machine learning for enhanced prediction.
• To create targeted scoring functions for improved accuracy.

Main Methods:

• Utilized Molegro Virtual Docker (MVD) for protein-ligand docking simulations.
• Employed differential evolution for search and MolDock/Plants scores for binding affinity.
• Constructed machine learning models using Scikit-Learn with docking-derived atomic coordinates.

Main Results:

• Developed regression models that accurately predict CDK2 inhibition.
• Achieved superior predictive performance compared to classical scoring functions.
• Demonstrated a practical approach to integrate docking and machine learning.

Conclusions:

• The proposed workflow effectively predicts CDK2 inhibition.
• Machine learning models trained on docking data offer a powerful alternative to standard scoring functions.
• The methodology provides a valuable tool for accelerating drug discovery efforts targeting CDK2.