CDK7 as a Target for Docking Screens.

Area of Science:

• Computational chemistry
• Structural biology
• Drug discovery

Background:

• Molegro Virtual Docker (MVD) facilitates molecular docking simulations.
• Machine learning (ML) models can enhance drug discovery by predicting binding affinity.
• Cyclin-dependent kinase 7 (CDK7) is a key target for anticancer drug development.

Purpose of the Study:

• To develop and validate an integrated computational workflow for molecular docking and ML-based binding affinity prediction.
• To assess the performance of a targeted regression model against a standard MVD scoring function for CDK7.

Main Methods:

• Utilized Molegro Virtual Docker (MVD) for docking simulations and Molegro Data Modeller (MDM) for ML modeling.
• Integrated MVD and MDM using Jupyter Notebooks to merge structural and binding data.
• Developed regression models using energy terms, scoring functions, and ligand descriptors as features.

Main Results:

• The integrated workflow successfully performed docking simulations and built regression models.
• A targeted regression model for CDK7 binding affinity prediction demonstrated superior performance compared to the MVD Rerank score.
• The workflow is adaptable for any protein target with available structural and binding data.

Conclusions:

• The developed workflow offers a powerful approach for enhancing drug discovery through accurate binding affinity prediction.
• ML-based regression models can outperform classical scoring functions in predicting drug-target interactions.
• The study provides a reproducible framework and accessible datasets for computational drug discovery research.