Identification of Natural Compounds with Anti-SARS-CoV-2 Activity using Machine Learning, Molecular Docking and Molecular Dynamics Simulation Studies

ABSTRACT

The coronavirus pandemic of 2019 (COVID-19) has adversely affected public health and the socioeconomic situation worldwide. Although there is no therapeutic drug to treat COVID, several treatment options are being considered to alleviate symptoms. Hence, researches on prophylactic treatment for COVID are being encouraged. Searching natural products is a rational strategy since it has served as a valuable source of lead compounds in drug discovery. In this study, three machine learning approaches, including Support Vector Machine (SVM), Random Forest (RF) and Gradient Boosting Machine (GBM), have been used to develop the classification model. The molecular docking was performed on AutoDock vina. Further, molecular dynamics (MD) simulation of the potential inhibitors was conducted using the AmberTools package. The accuracy for SVM, RF and GBM was found to be 60.45 %, 63.43 % and 64.93 %, respectively. Further, the model has demonstrated specificity range of 41.67 % to 50.00 % and sensitivity range of 74.32 % to 79.73 %. Application of the model on the NuBBE database, a repository of natural compounds, led us to identify 322 unique natural compounds, likely possessing anti-SARS-CoV-2activity. Further, molecular docking study has yielded three flavonoids and one lignoid compounds with comparable binding affinities to the standard compound. In addition, MD showed that these compounds form stable complexes with different magnitude of binding energy. The in silico investigations suggest that these four compounds likely demonstrate their anti-SARS-CoV-2activity by inhibiting the main protease enzyme. Our developed and validated in silico high-throughput investigations may assist in identifying and developing antiviral drug-like compounds from natural sources.