Optimized Compound Screening Library

For Cn, Hbp and Cdk2, we performed virtual screening using the ICCL. In this library we removed all the compounds with molecular weight over 750 g mol⁻¹ or consisting of less than 10 atoms. The molecules chelating exotic atoms (Au, Cu, Hg, I, Sn, …) or ions (Fe²⁺, Mg²⁺, …) were also filtered because they were not correctly handled by the docking programs. Finally 856 of the 15,163 compounds were removed and the virtual screening was performed on 14,307 unique compounds. The preparation of the chemical library for docking required different steps to generate accurate 3D molecular structures. When chirality was not specified in the chemical library, all the possible enantiomers were generated. In addition, the protonation states of the compounds were adjusted according to the pH of the medium surrounding the target. In our case, the physiological pH at 7.4 ± 1 was retained. All the protonation states with a probability of existence over 10 % at the given pH were generated as well as all the likely tautomers. The preparation of these compounds was achieved using LigPrep 2.8.0 (Schrödinger). After this preparation, the library amounted to 24,186 structures. Since some docking programs, such as GOLD, do not alter bond lengths and angles, thereby the ligands energy was minimized using the OPLS2005 force field to ensure proper bond distances and angles. We generated one conformer per molecule, the exploration of the ligand conformational space being managed by each of the four docking programs.
The DUD-E database includes the multi-mol2 files of active compounds and decoys for each target, which were prepared by the DUD-E team, with their enantiomers, protomers and tautomers [18 (link)]. We used them without any modification.