Software suite

The amino acid sequence of HelD from B. subtilis was retrieved from UniProt database and the closest structural homologs were identified using BLAST search against the Protein Data Bank database. The ATPase domain of HelD (residue range 539–641) shares 29% sequence identity with the E. coli UvrD. The C-terminal domain (residue range 606–774) shares 39% sequence identity with Lactobacillus planetarium (PDB ID 3DMN). No sequence similarity was observed for the N-terminal domain. The homology model was built only for the ATPase and the C-terminal domain using Prime 3.1 module (Jacobson et al., 2004 (link)) in Schrödinger software suite (Schrödinger, LLC, New York, NY, United States). SSpro program in Prime was used to predict the secondary structure of HelD. Initially, four models were generated and the model with the least energy was selected for the loop refinement. For refinement of loops comprising of <5 and >5 amino acids, the number of output structures was set to 10 and 5, respectively. After loop refinement, the OPLS3 force field (Harder et al., 2016 (link)) was used to minimize the model which was further validated by PROCHECK (Laskowski et al., 1993 (link)) for the evaluation of Ramachandran plot (Ramachandran et al., 1963 (link)).

Covalent docking of PepE and TrxR1 was performed in the Schrödinger Software Suite (V2015-2, Schrödinger LLC, NY, USA). The structure of human TrxR1 protein (PDB ID: 2J3N, Chain A, B, C, D, E, and F) was obtained from the Protein Data Bank and further modified in the Protein Preparation Wizard module. The Sec498 residue in chain C was selected as the reactive residue involved in the Michael addition and also set as the centroid of the docking box (Box size: dock ligands with length ≤ 15 Å). The other dock parameters were set to default in the docking simulation.

Grid-based molecular docking was set up using the Schrödinger software suite to predict the binding orientation and interaction of the ligands. The processed compounds from CMNPD were screened using a virtual screening workflow which includes three stages: (1) high-throughput virtual screening (HTVS), (2) standard precision (SP) and extra-precision (XP) docking [35 (link),36 (link)]. The compounds retained from the HTVS stage were passed to the next stage, SP docking; the SP selected compounds were then docked using the more accurate and computationally intensive XP mode. At each stage, the top 10% of compounds were retained and proceeded to the next stage. Finally, the best compounds were selected based on XP GlideScore, a scoring scheme used to report the strength of the binding of a ligand to a protein. Protein–ligand interactions, including hydrogen bonds, hydrophobic interactions, π-π stacking and cation-π interactions, were assessed. The free energies for the binding of ligands with the proteins (ΔG_bind in kcal/mol) were calculated by using the MM-GBSA method. For each DPP-4-ligand complex, the MM-GBSA-based binding free energy was estimated using the equation: ΔG_bind = G_complex − G_protein − Gl_igand, where ΔG_bind is the binding free energy and G_complex, G_protein, and G_ligand are the free energies of complex, protein and ligand, respectively [37 (link)].

Pharmacophore hypotheses were developed using the E-Pharmacophore model in PHASE (Pharmacophore Alignment and Scoring Engine) [39 (link),40 (link)] within the Schrödinger Software Suite, Schrodinger LLC. The E-Pharmacophore model generates hypotheses from the co-crystal structure of protein ligand complexes based on complementarity between protein and ligand either automatically or manually. The automatic generation uses the GLIDE XP scoring terms to determine the most important features (interactions) contributing towards binding. In the manual options, the features can be picked or selected manually. Once they are generated, the pharmacophoric features and their positions in space are manually compared to generate a set of pharmacophoric criteria for their binding, activity, potency and selectivity. Six types of pharmacophoric features were generated in this study. These pharmacophoric features are represented either by colored spheres or by the colored ring in 3D. H-bond donors (D) are represented by light blue spheres, H-bond acceptors (A) by pink spheres, positives (P) charges by dark blue spheres, negatives (N) charges by dark red spheres, hydrophobic (H) groups by green spheres and aromatic (R) groups by orange rings. A combination of these features that correlates with the biological activity or potency is defined as pharmacophore hypotheses.