AlphaFold for Docking Screens.

Area of Science:

• Computational biology
• Artificial intelligence in drug discovery
• Structural bioinformatics

Background:

• AlphaFold provides AI-generated protein 3D structures, accessible via the AlphaFold Protein Structure Database (AlphaFoldDB).
• Experimental structural data for certain proteins, like cyclin-dependent kinase 19 (CDK19), is often unavailable.
• CDK19 is a significant target for developing novel anticancer therapeutics.

Purpose of the Study:

• To demonstrate the utility of AlphaFold models in virtual screening and drug discovery workflows.
• To develop a predictive regression model for CDK19 inhibition using computational methods.
• To integrate AI-driven structural predictions with molecular docking simulations for drug target analysis.

Main Methods:

• Utilized AlphaFold-generated protein structures for docking screens with Molegro Virtual Docker.
• Employed Jupyter Notebooks to integrate docking simulations and analyze protein-ligand complex atomic coordinates.
• Constructed a neural network regression model using Molegro Data Modeller based on docking results and binding affinity data for CDK19 inhibitors.

Main Results:

• Successfully built a neural network regression model to predict CDK19 inhibition.
• Demonstrated a workflow integrating AlphaFold structures, docking, and machine learning for drug target modeling.
• Provided open-access datasets and Jupyter Notebooks for reproducibility and further research.

Conclusions:

• AlphaFold models can be effectively used in docking screens for drug discovery targeting proteins lacking experimental structural data.
• The developed computational approach facilitates the prediction of inhibitor efficacy for targets like CDK19.
• This study highlights the potential of AI and computational tools in accelerating anticancer drug development.