Virtual screening of potential ATR kinase inhibitors based on machine learning and molecular docking

Screening potential active compounds from molecular libraries is a common method for drug discovery.However, with the continuous exploration of chemical space, there are already compound libraries with more than billions of molecules, so molecular docking is no longer enough to quickly screen specific target inhibitors from the ultra-large compound libraries.This study proposes a method for screening potential active compounds, which involves filtering and selecting compounds from a candidate compound library containing over 5.5 billion molecules through a series of steps, including calculating physical and chemical property similarities, constructing machine learning prediction models, and molecular docking.In the end, 51 compounds with potential ataxia telangiectasia-mutated and rad3-related (ATR) inhibitory activity were obtained.This method is effective for rapidly screening novel potential active compounds from large compound libraries.