Rocs 3

C-terminal residues were sequentially removed from the active site cavity of the post-cleaved NS2 protease domain structure (PDB: 2HD0) using Maestro (Schrodinger) to produce the models NS2Δ_P1−P2, NS2Δ_P1−P5 and NS2Δ_P1−P10. Virtual screening was performed using eHITS (SymBioSys) to dock and score a library of >5 × 10⁵ commercially available lead-like molecules. Additionally, SPROUT (Gillet et al., 1994 (link)) was employed to construct models of molecules predicted to form favourable interactions. Top scoring molecules were expanded using ROCS 3.2.0.4 (OpenEye Scientific Software, Santa Fe, NM. http://www.eyesopen.com) (Hawkins et al., 2007 (link)) and filtered by favourable modelled binding pose, structural diversity and cLogP to yield a library of 200 compounds.

Compounds were selected by 2D and 3D ligand-based similarity searches. We calculated the Tanimoto coefficient between compound 1 and each of the 6,868,468 commercially available compounds on the basis of 2D fingerprints (MACCS public keys, ECFP4, FCFP4 and GpiDAPH3) and 3D shape metrics (ComboScore) with the software Pipeline Pilot 2017 (Dassault Systèmes, San Diego, CA), MOE 2015.1001 (Chemical Computing Group, Montreal, Quebec, Canada), and ROCS 3.2.0.4 (OpenEye Scientific Software, Santa Fe, NM.) [38 (link)]. Finally, 38 compounds showing high Tanimoto coefficient values from each metric were selected and purchased.