Prediction of inhibitory constants of compounds against SARS-CoV 3CLpro enzyme with 2D-QSAR model

Abstract Developing broad-spectrum anti-coronavirus drugs is greatly important, since the novel SARS-CoV-2 has rapidly become a threat to the public health and economy worldwide. SARS-CoV 3-chymotrypsin-like protease (3CLpro), as highly conserved in betacoronavirus, is a viable target for anti-SARS drugs. A quantitative structure–activity relationship (QSAR) for inhibitory constants (pKi) of 89 compounds against SARS-CoV 3CLpro enzyme was developed by using support vector machine (SVM) and genetic algorithm. The optimal SVM model (C = 90.2339 and γ = 1.19826 × 10−5) based on six molecular descriptors has determination coefficients of 0.839 for the training set (65 compounds) and 0.747 for test set (24 compounds), and rms errors of 0.435 and 0.525, respectively. These results are accurate and acceptable compared with that in other models reported, although our SVM model deals with more samples in the dada set. The SVM model could be beneficial for search of novel 3CLpro enzyme inhibitors against SARS-CoV.