QSAR Study of HCV Inhibitors

Hepatitis C is one of the most dangerous and rapidly spreading infections. Modern therapy is only effective in 50% cases and is associated with side effects. The development of direct-acting antiviral drugs can increase treatment efficacy and reduce side effects. However, laboratory drug searches can be expensive and time consuming, so pharmaceutical companies need proven methods to predict the activity of compounds in silico. In this work, we introduce the method to predict antiviral activity of chemical compounds against hepatitis C virus by constructing ensembles of models consisting of basic models based on various machine learning methods and types of molecular descriptors. The effectiveness of the obtained ensembles was demonstrated on an external validation sample. The developed methodology was used to predict the antiviral activity of known drugs and other compounds that have undergone clinical trials; the forecast results can be further used for virtual screening of prioritization of compounds before experimental determination of antiviral activity.