Discovery of Potential Flavonoid Inhibitors Against COVID-19 3CL Proteinase Based on Virtual Screening Strategy
Zhongren Xu, Lixiang Yang, Xinghao Zhang, Qiling Zhang, Zhibin Yang, Yuanhao Liu, Shuang Wei, Wukun Liu
Abstract
The outbreak of 2019 novel coronavirus (COVID-19) has caused serious threat to public health. Discovery of new anti-COVID-19 drugs is urgently needed. Fortunately, the crystal structure of COVID-19 3CL proteinase was recently resolved. The proteinase has been identified as a promising target for drug discovery in this crisis. Here, a dataset including 2030 natural compounds was screened and refined based on the machine learning and molecular docking. The performance of six machine learning (ML) methods of predicting active coronavirus inhibitors had achieved satisfactory accuracy, especially, the AUC (Area Under ROC Curve) scores with fivefold cross-validation of Logistic Regression (LR) reached up to 0.976. Comprehensive ML prediction and molecular docking results accounted for the compound Rutin, which was approved by NMPA (National Medical Products Administration), exhibited the best AUC and the most promising binding affinity compared to other compounds. Therefore, Rutin might be a promising agent in anti-COVID-19 drugs development.