Discovery of new TLR7 agonists by a combination of statistical learning-based QSAR, virtual screening, and molecular dynamics.

Search for new antiviral medications has surged in the past two years due to the COVID-19 crisis. Toll-like receptor 7 (TLR7) is among one of the most important TLR proteins of innate immunity that is responsible for broad antiviral response and immune system control. TLR7 agonists, as both vaccine adjuvants and immune response modulators, are among the top drug candidates for not only our contemporary viral pandemic but also other diseases. The agonists of TLR7 have been utilized as vaccine adjuvants and antiviral agents. In this study, we hybridized a statistical learning-based QSAR model with molecular docking and molecular dynamics simulation to extract new antiviral drugs by drug repurposing of the DrugBank database. First, we manually curated a dataset consisting of TLR7 agonists. The molecular descriptors of these compounds were extracted, and feature engineering was done to restrict the number of features to 45. We applied a statistically inspired modification of the partial least squares (SIMPLS) method to build our QSAR model. In the next stage, the DrugBank database was virtually screened structurally using molecular docking, and the top compounds for the guanosine binding site of TLR were identified. The result of molecular docking was again screened by the ligand-based approach of QSAR to eliminate compounds that do not display strong EC