Abstract:
The hologram quantitative structure-activity relationship (HQSAR) was used to study 40 SARS-CoV 3CL protease inhibitors. The molecular fragment distinction parameter, fragment parameter and hologram length that affect the model quality were investigated. Using Partial least squares (PLS) to establish an optimal model with 30 compounds as the training set. In this model cross-validated
q2 was 0.604, and conventional
r2 was 0.904, standard deviation SEE was 0.125; The model was used to predict the test set consisting of 9 compounds, and
r2pred was 0.723, which indicated that the model had good prediction ability and fitting ability. The color code analysis provided the activity contribution of different groups in the model, and based on this result, a group of phenylisoerine inhibitors with good predictive activity was designed. It provides a reference for the design and optimization of new SARS-CoV 3CL protease inhibitors, and also provides a reference for the development of 3CL protease inhibitors for the treatment of novel coronavirus pneumonia (COVID-19).