Machine Learning to Predict CDK4 Inhibition.

Area of Science:

• Computational chemistry
• Drug discovery
• Machine learning

Background:

• Cyclin-dependent kinase 4 (CDK4) is a crucial target for anticancer drug development.
• Existing crystallographic structures enable computational docking simulations for CDK4 inhibition.
• Binding affinity data for CDK4 inhibitors facilitates the creation of predictive machine learning models.

Purpose of the Study:

• To describe an integrated workflow for constructing a neural network model to predict CDK4 inhibition.
• To utilize atomic coordinates and docking results for regression modeling of CDK4 inhibition.
• To provide accessible datasets and Jupyter Notebooks for reproducible research.

Main Methods:

• Employing Molegro Data Modeller (MDM) to build a regression model based on docking results.
• Utilizing protein-pose structures generated by Molegro Virtual Docker (MVD).
• Integrating experimental binding data from BindingDB for model training.

Main Results:

• A functional workflow integrating docking simulations and machine learning for CDK4 inhibition prediction was established.
• The workflow successfully builds regression models to calculate binding affinity based on atomic coordinates.
• Associated CDK4 datasets and Jupyter Notebooks are publicly available on GitHub.

Conclusions:

• The described workflow offers a robust method for predicting CDK4 inhibition using computational approaches.
• This integrated approach facilitates the identification and development of novel anticancer drugs targeting CDK4.
• The availability of code and data promotes further research and application in drug discovery.