Supervised machine learning and molecular docking modeling to identify potential Anti-Parkinson's agents.

Parkinson's disease is a neurodegenerative condition that affects the brain's neurons, and causes malfunction of nerve cells and their death. A neurotransmitter called dopamine interacts with the part of the brain in charge of coordination and movement. In general, the brain produces less dopamine as Parkinson's disease worsens; therefore, it becomes harder to control the movements. In this study, a dataset collected from CHEMBL library was applied to build four machine learning models using three different descriptors functions to determine the best models with the best features and suggest the best adenosine inhibitors. Molecular docking of adenosine A2A (PDB ID: 3UZA) receptor was applied to identify the potential inhibitors. The machine learning and molecular docking results indicate that XGBoost model with RDkit features is an excellent model for this dataset to explore new Anti-Parkinson's agents.