Homology modeling estimates the 3D structure of a target protein sequence by using its alignment to one or more protein templates of known structure . Though, Mirk plays a crucial role in various solid tumors and acts as a potential therapeutic drug target. Still, now there is no 3D structure of Mirk protein was deposited in PDB; hence a homology model was performed to determine the 3D coordination of Mirk protein. The sequence alignment of 4MQ1 (template) and (target) and the secondary structural elements are indicated in Figure 1. Template searching against Protein Data Bank (PDB) through BlastP was performed and 4MQ1 was selected based on identity (42%) with the lowest E-Value (3e-23). The three-dimensional structure provides valuable insight into its molecular function and also enables the analysis of its interactions with suitable substrates or inhibitors .
On the basis of the sequence alignment, the 3D structure of the target protein was predicted using the crystal structure of the template (4M) by comparative modeling using a modeler. It generates five models, based on the DOPE score model-4 was selected for further analysis (Figure 2). The root means square deviation of the model relative to the template (4MQ1) 0.111 A was calculated using Pymol. It indicates that the modeled protein show close homology to the template.
In order to evaluate whether a model satisfies standard steric and geometric criteria and select the best homologs model, we used several verification tools for model evaluation such as PROCHEK, WHAT_CHECK, ERRAT, VERIFI_3D, and PROVE provided by the SAVS server. PROCHEK was used to perform full geometric analysis as well as the stereo-chemical quality of a protein structure by analyzing residue-by-residue geometry and overall structure geometry [37, 38]. The stereochemistry assessment of the Mirk model showed more than 92.4% residues of psi and phi angles are in the favored region of the Ramachandran plot (Figure 3) and 6.4 %, 0.4% of residues belong to the generously allowed region and allowed respectively. Concerning the Ramachandran plot, it is observed that there are 0.4% residues in the disallowed region; it again indicates that the final model was more reliable and had good quality and was close to the template structure, and could be used for further studies. Figure 3B represent that the residue interaction energies; which shows the energies remain negative for almost all the residue indicating the acceptability of the model. Profile score above zero in the Verify 3D graph [39, 40] corresponds to the acceptable environment of the model. The ProSA  results in finding the potential error on the predicted model revealed the Z-score values of -4.56 against the template which indicates that protein structure is well correlated with the crystal structure of similar length and within the acceptable range.
Virtual screening and docking studies
The active site of the modeled Mirk was predicted by the sitemap program implemented in Schrodinger. The grid was generated considering the centroid of active site residues in the binding region for further virtual screening. Chembridge database was used to find potential inhibitors. The validated model of Mirk and ligands of Chembridge database were imported into Schrodinger LLC, 2018). Prior to docking ligands were prepared to expand protonation and tautomeric states of each ligand molecule at 7.0 pH with LigPrep module implemented in maestro [42, 43]. The high energy tautomer state or ionization state and the ligand did not follow Lipinski's rule, or the molecules with reactive functional groups are eliminated from the generated confirmation. Like ligand preparation, prior to docking the predicted Mirk model was also subjected to preprocess with Protein preparation Wizard in the maestro, where all the hydrogens were added and subsequently subjected into minimization with an OPLS-2005 force field and the impact of molecular mechanics engine [44, 45]. The minimization process restraining heavy atoms and allows free rotation of hydrogen setting with an RMSD value of 0.30Å . The active site was predicted with the Sitemap module and a grid box was generated around the active site of the model . Glide virtual screening, SP, (standard precision), and XP (extra precision) mode were applied to predict the potential lead molecules. The good scoring ligands were ranked based on glide score and binding energy with predicted Mirk modeled protein.
In silico predicted Physico-chemical parameters
The top-ranked compounds (table 1) were subjected to a QikProp module to examine the drug-like behavior of the compounds by evaluating their pharmacokinetic factors. All the important parameters were depicted in table 1. All the top-ranked molecules exhibited a molecular weight in the acceptable range. ADME properties of lead molecules play a crucial role in medicinal chemistry. The pharmacokinetic parameters will influence the drug-likeness of the lead molecules. Hence in the present study, in silico analysis of physicochemical properties of the top-ranked hits was predicted. The values of QPlogS, QPlogBB, and QPlogPo/w were noticed within the prescribed limit −6.5 to 0.5, −3.0 to 1.2, and −2.0 to 6.5 respectively. The value of predicted Caco-2 cell permeability of the top-ranked lead molecules was found to be very high except for the compound ID 7753623, this result suggested that the ChembrigeID: 7753623 may face some hurdles in the absorption process than other lead molecules. The value of permeability through MDCK cells of all the top-ranked compounds was found to be <500, except ChembrigeID: 7764195, the results depicted that the MDCK cells are considered to be a good mimic for the blood-brain barrier. Overall results indicated that all the top-ranked compounds are satisfied with rules with fewer violations. Additionally, prime MM-GBSA binding energy was calculated for the top two compounds based on their binding affinity towards the active site of Mirk protein. MM-GBSA score of Chembridge ID: 7768949 was -102.21 whereas the second topmost compound ID: 7771055 has shown a slightly lower MM-GBSA score of -96.04. With these results, it can be concluded that the binding energy of both the compound shows good binding interaction and binding energy with the active site of the Mirk protein.
Molecular docking studies revealed the apparent binding mode of lead molecules in the active site of the selected drug targets with maximum docking score and binding energy. The modeled Mirk was docked with Chembridge database ligand molecules to generate the binding mode and dynamic simulation was done to refine the best pose of top-ranked compounds with an allowed conformational change in the Mirk. The top-ranked compounds such as Chembridge ID; 7768949, 7771055, 7758866, and 7764195 were shows good docking scores ranging from -11.361 to -10.064. The docked poses of top-ranked compounds at the active site of modeled Mirk protein are shown in Figure 4. The detailed analysis of residues and hydrogen bond interactions of lead molecules were shown in the 2D diagram in Figure 5 and Table 2. The binding mode analysis shows that the binding energy range of lead molecules was ranged between -49.771Kcal/mol to -32.065 Kcal/mol. The amino acid residues involved in both bonding and non-bonding interaction with lead molecules with Asn86, Leu83. Lys17, Ile7 play a crucial role in ligand binding. Based on the modeled structure, we speculated that the identified lead molecules significantly positioned in the binding pocket of the modeled structure with good stability.
The molecular interaction diagram (Figure 5) shows the interaction of the best two compounds such as ChembridgeID: 7768949 and ChembridgeID: 7771055 with active site residues. Chembridge ID 7768949 formed 2 hydrogen bond interactions with Leu83 via –N- and –NH- atom (N…H, NH…Leu83) with 2.30Å distance and another hydrogen bond interaction with same residues via –C- and –OH atom (C…H, OH…Leu83), with a distance of 1.89Å also it forms hydrogen bond interaction with Ile7 via –C- and –OH atom (C…H, OH…Ile7) with 2.01Å distance. Compound ID 7771055 forms two hydrogen bond interactions with Leu83 via –N- and –NH- atom (N…H, NH…Leu83) and –C- and –OH atom (C…H, OH…Leu83) with the distance of 2.04Å and 2.34Å respectively, also form hydrogen bond interaction with Asn86 via –C- and –OH- atom (N…, O, HC…Asn86) with the distance of 1.97Å. Compound ID: 7758866 also have a similar mode of interaction like Chembridge ID 7768949, formed 2 hydrogen bond interaction with Leu83 via –N- and –NH- atom (N…H, NH…Leu83) and –C- and –OH atom (C…H, OH…Leu83) with a distance of 1.94Å and 2.30Å respectively, also form hydrogen bond interactions with Asn86 via –C- and –OH- atom (N…, O, HC…Asn86) with a distance of 2.10Å. compared to the first three compounds, compound ID: 7764195 shows the different mode of interaction in the active site of Mirk receptor, it forms hydrogen bond interaction with Leu via –N- and –NH- atom (N…H, NH…Leu83) with a distance of 2.10Å and with Lys17 via –N- and –NH- atom (N…H, NH…Lys17) with a distance of 2.60Å and form another hydrogen bond interaction with Ile7 via –C- and –OH atom (C…H, OH…Ile7) with 2.01Å distance This interaction studies revealed that hydrogen bond formations were responsible for he binding affinity of the compounds with active site residues. The molecular contact profiling suggested that virtually screened compound show the potential to inhibit the Mirk/Dyrk1B protein via action on the binding of crucial amino acid residues to distort the active binding pocket of Mirk/Dyrk1B.
A molecular dynamic simulation was performed using GROMACS to evaluate the structural stability of both modeled protein and complexes with top-ranked compounds. To confirm the structural stability and equilibrium state of the molecular simulation, root mean square deviation (RMSD) of backbone carbon atoms of both modeled protein and complexes with top-ranked compounds are calculated and plotted in figure 6. From the RMSD plot, it was observed that the modeled protein Mirk was significantly stable throughout the simulation time of 20 ns in the range of 0.14 to 0.17 nm. Whereas the top-ranked complexes such as Chembridge ID: 7768949 and Chembridge ID: 7771055 with modeled protein show higher RMSD fluctuation arise till 5 ns in the range of 0.25 and slightly down in the range of 0.15 to 020 ns and get stable remaining simulation time (figure 7). The RMSD of a compound such as Chembridge ID: 7758866 and Chembridge ID: 7764195 shows fluctuation similar to first two top-ranked compounds and arise up to 5 ns in the range of 0.25 and highly fluctuate up to in the range of 0.30 to 0.35 till 20 ns simulation period (Figure 7). The RMSF of modeled protein residues lies below 0.21 nm (figure 8).
The binding of top-ranked compounds such as (compound 3 and 4) with modeled protein shows little fluctuation and shows RMSF value up to 0.4 nm and 0.5 nm respectively, however, the active site residues do not show high fluctuation throughout the simulation time. After binding of compounds 1 and 2, the modeled protein does not show any high fluctuations throughout the simulation time and both the compounds retain stability in the whole simulation period (Figure 7). From the RMSD and RMSF results, it depicted that the top-ranked screened compounds effectively perturb to the binding pocket of the modeled protein. In addition to the stability analysis, the protein interaction with top-ranked compounds was supervised throughout the simulation period. The hydrogen bond interaction of crucial amino acids is present in the active site of the protein-making the protein-ligand complex stable.
The molecular simulation studies indicate that the both modeled and complex system shows significant stability, and results of hydrogen bond interaction of top-ranked complexes also support this statement (Figure 8). The amino acid residues Asn86, Lue83, and Ile7 exhibited strong hydrogen bond interaction with all the top-ranked compounds revealed by docking studies also existed in molecular dynamics simulation trajectory, with stable and strong interaction in the stipulated time. Compared to compounds 1 and 2, compounds 3 and 4 showed less stable interactions with similar amino acid residues during the entire simulation the dark color represents the number of interactions with respective of the amino acid high, and minimal backbone fluctuation has ensued in the system.