Identification of novel influenza polymerase PB2 inhibitors using virtual screening approach and molecular dynamics simulation analysis of active compounds

Hierarchical virtual screening combined with ADME prediction and cluster analysis methods were used to identify influenza virus PB2 inhibitors with high activity, good druggability properties, and diverse structures. The 200,000 molecules in the ChemDiv core library were narrowed down to a final set of 97 molecules, of which six compounds were found to rescue cells from both H1N1 and H3N2 virus-induced CPE with EC50 values ranging from 5.81 μM to 42.77 μM, and could bind to the PB2 CBD of H1N1, with Kd values of 0.11 μM to 6.4 μM. The six compounds have novel structures and low molecular weight and are, thus, suitable serve as lead compounds for development as PB2 inhibitors. A receptor-based pharmacophore model was successfully constructed using key amino acid residues for the binding of inhibitors to PB2, provided by the MD simulations. This pharmacophore model suggested that to improve the activity of our active compounds, we should mainly focus on optimizing their existing structures with the aim of increasing their adaptability to the binding site, rather than adding chemical fragments to increase their binding to adjacent sites. This pharmacophore construction method facilitates the creation of a reasonable pharmacophore model without the need to fully understand the structure-activity relationships, and our descriptions provide a useful reference for similar research.