Fingerprint-based QSAR model generation to identify structural determinants of HCV NS5B inhibition

Official URL: https://dx.doi.org/10.29228/jrp.429

RNA-dependent RNA polymerase, non-structural protein 5B (NS5B), is an essential enzyme of HCV for viral transcription and genome replication. Its initial validation as a promising target for the treatment of chronic hepatitis and hepatocellular carcinoma has consequently prompted different research institutes and the pharmaceutical industry to find potential inhibitors for human therapies. Among those, anthranilic acid derivatives received increasing attention because of their promising drug-like properties. In order to design promising drug candidates, the structural determinants of NS5B inhibitors were determined by a robust fingerprint-based quantitative structure-activity relationship (QSAR) model which was depicted on atomic effect contribution maps to provide visual aids for medicinal chemists. In the present work, we used a combination of computational chemistry methods including ensemble docking, binding free energy calculations, and a fingerprint-based QSAR model. We built a robust in silico protocol to accelerate the structure-based design of HCV NS5B inhibitors. The QSAR model, kpls_linear_3, constructed by KPLS fitting with linear fingerprints produced the best predictive performance (a correlation coefficient for the training set R2 = 0.8900, and a correlation coefficient Q² = 0.9234 and RMSE = 0.3032 for the test compounds). The atomic effect contribution map that was generated based on this model showed a good agreement between the predictions and the experimental data. To the best of our knowledge, we illustrated for the first time the use of the atomic effect contribution map as a visual aid for assessing the structural determinants of NS5B inhibitors. The computational strategy represented herein can assist pharmaceutical chemists in the rapid identification of the important features to design novel inhibitors of other protein targets as well.