Structure-based virtual screening (SBVS) framework

The structure-based virtual screening (SBVS) framework applied in this study was previously described ^{28, 46} and included four main steps: (1) target and library pre-processing, (2) docking, (3) scoring and (4) post-processing of top-scoring hits. In brief, a library of commercially available compounds (the fragment-based library including 4,352 chemicals) was downloaded from Specs and processed by the Lig Prep tool within Maestro v10.1 ⁴⁷; AutoDock could alternatively be used for this function ⁴⁸. The target model (Smp_138030) was pre-processed using the Protein Preparation Wizard within Maestro (otherwise AutoDock) by assigning bond orders, adding hydrogens and performing a restrained energy minimisation of the added hydrogens using the OPLS_2005 force field. Docking simulations were performed on the substrate binding pocket of the target to evaluate the binding affinity of each Specs compound to a 12 Å docking grid (inner-box 10 Å and outer-box 22 Å) previously prepared using, as a centroid, the substrate peptide.

Initially, the in silico molecular docking was performed using the Glide docking software within Maestro (Schrödinger Release 2017 ⁴⁷) using the standard precision function (SP; all 4,532 compounds). AutoDock could be alternatively used as docking program. The results were subsequently refined using the more accurate extra precision (XP) function. The resulting conformations (or poses) of the compounds were ranked according to the Glide XP scoring function with the top 500 distinct compounds identified and retained for a similar docking (both SP and XP) to the corresponding human homologue H. sapiens MLL3 (PDB ID: 5F6K), previously prepared as described above for the schistosome protein. As a result, each compound was associated with a pair of docking scores (the SP and XP scoring function) for both the schistosome and the corresponding human homologue. The compounds with a more favourable docking score (i.e. the lower energy value represented by more negative values for both XP and SP scores) for the parasite protein compared to the human template were selected. A subset of compounds (defined here as first set of compounds) was carefully chosen to encompass maximal chemical diversity and purchased for biological screens. Refinement of the first selection criterion led to the identification of chemicals (defined here as second set of compounds) having a more favourable docking score for the parasite protein compared to the human template for at least one of the scoring function (either SP or XP score).

In a final stage of the study, we used the central scaffold of compound 7 (6-(piperazin-1-yl)-1,3,5-triazine-2,4-diamine, simplified molecular-input line-entry) as query structure to search for any remaining structural analogues of compound 7 present in the Specs fragment-based library. This structure-based search resulted in 17 compounds, of which nine chemicals were already included in the second set of compounds. The remaining eight small molecules were purchased and screened to fully explore the chemical space around the central scaffold of compound 7. These chemicals were not previously identified because they did not have a more favourable docking score (neither SP nor XP score) for the parasite protein compared to the human template. Overall, this approach exhaustedly investigated the whole Specs database for any closely structural related analogues of compound 7.