Novel QSAR model to predict Activity of natural Products against Covid-19.

Si, Yaru; Xu, Xinyue; Hu, Yingfeng; Si, Hongzong; Zhai, Honglin

Currently COVID-19 is spreading in a large scale while no efficient vaccine has been produced. How to found a high effective drug for COVID-19 is very necessarily. We established a satisfied quantitative structure-activity relationship model by gene expression programming to predict the IC50 value of natural compounds. 27 natural products were optimized by heuristic method in CODESSA program to build a liner model. Basis on it, only 2 descriptors were selected and utilized to build a nonlinear model in gene expression programming. The square of correlation coefficient and s2 of heuristic method were 0.80 and 0.10, respectively. In gene expression programming, the square of correlation coefficient and mean square error for training set were 0.91 and 0.04. The square of correlation coefficient and mean square error for test set are 0.86 and 0.1. This nonlinear model has stronger predictive ability to develop the targeted drugs of covid-19.