DUckCov (https://doi.org/10.1002/cmdc.201900078), a computational platform designed for the virtual screening of covalent binders, was applied against the SARS-CoV-2 Mpro structure in PDB 6LU7. An acrylamide warhead was preferred over the chloroacetamide to reduce the reactivity of the electrophile. Fragment x1382 was modified by incorporating an acrylamide instead of the original warhead, and docked with CovDock-LO (Schrödinger, https://doi.org/10.1021/ci500118s) against Cys145 to obtain reference coordinates for warhead atoms. A library of ≈450k acrylamides was compiled from the Enamine REAL database. An Interaction Fingerprint (IFP) analysis on XChem hits highlighted four recurring ph4 features: an hydrophobic group in P2; an H-bond donor to E166-O; an H-bond acceptor to E166-N; an H-bond acceptor to H163-NE2. The hydrophobic feature in P2 was found to be the most frequent and was set as mandatory together with at least one of the three H-bond features for the constrained docking with rDock. Tethered and constrained HTVS by rDock was performed by allowing rotational and translational degree of freedom to a certain extent for warhead atoms in the reactive site. Dynamic Undocking (DUck, https://doi.org/10.1038/nchem.2660) was performed focusing on the H-bond established by best scoring rDock hits with any of the three ph4 residues. CovDock-LO was performed on the best scoring hits to obtain covalent docking poses.
All the compounds are from Enamine REAL acrylamide library