Data-driven discovery of chemical signatures for developing new inhibitors against human influenza viruses.

This study presents cheminformatics analysis of the antiviral chemical space targeting human influenza A and B viruses. By curating 407,366 small molecules from ChEMBL and PubChem, we evaluated physicochemical properties, structural motifs, and activity trends across phenotypic and target-based assays. We found that 90.6% of evaluated molecules met Lipinski's Rule of Five, with active compounds exhibiting higher topological polar surface area and hydrogen bond donor groups. Target-specific analyses revealed distinct profiles for neuraminidase (NA) and hemagglutinin (HA) inhibitors, including larger molecular weights and increased rotatable bonds. Structural characterization identified cyclohexene, dihydropyran, and pyrimidine rings as prevalent in highly active molecules, while phthalimide motifs correlated with inactivity. Clustering of phenotypic assay data highlighted four promising and unique antiviral candidates, with unexplored chemical space. We also identified five multi-target scaffolds, including the curcumin-like scaffold, demonstrating dual inhibitory potential against two viral proteins. Molecular docking experiments on molecules within one of these multi-target scaffolds indicated their potential as initial hit candidates. Combined RMSD, PDF and DCCM analyses across molecular dynamics simulations elucidated the binding behaviour of five curcumin-like candidates. Two ligands remained as stable as the reference antivirals, one showed target-specific loss of affinity, and two dissociated rapidly, indicating that the stable pair should be prioritised for subsequent in vitro validation. Overall, the findings of this study can aid computer-aided drug design efforts, contributing to the development of novel antiviral agents against human influenza viruses.