The CPH generation and Atom-based 3D-QSAR studies were processed using a phase module incorporated within the maestro software interface from Schrodinger, LLC, New York USA (release 2017) [6]. The studies related to docking were also additionally performed using Glide module (release 2017, Schrödinger, LLC, NY) installed on Linux based Computer. The molecular dynamic simulation study and Absorption, Distribution, Metabolism, Excretion and Toxicity (ADMET) properties predictions were done using DESMOND and QikProp modules respectively [7, 8].

A set of thirty-eight IPC derivatives having mycobacterium inhibitory activity was used for 3D-QSAR study as shown in Table 1. The compounds within the set were divided randomly into 70% of the training set and 30% of test set molecules. Twenty-seven molecules in the training test and eleven molecules in the test set were used for developing the Phase CPH.

Table 1. Structures of ligands used for study:

Compd.ID	R	MIC(µM)	Compd. ID	R	MIC (µM)
24		0.038	27f		10.58
25		0.62	27g		87
26a		0.81	28a		49.61
26b		1.53	28b		9.97
26c		0.76	29a		1.93
26d		0.64	29b		0.62
26e		1.39	30a		95.42
26f		0.18	30b		90.35
26g		0.044	30c		90
26h		0.041	30d		43.47
26i		90	30e		77
26j		42.55	30f		86
26k		5.31	30g		45.98
26l		47.31	30h		87
27a		6.21	30i		84.43
27b		1.47	31		0.66
27c		1.32	32a		95.04
27d		87	32b		95
27e		12.43	33		90.33

Overall, molecular docking comprises 5 main steps-1] protein preparation, 2] ligand preparation, 3] receptor grid generation, 4] ligand docking procedure and then 5] viewing the docking results. Prime MMGBSA generated properties are helpful in determining energies for ligand-receptor complex [9-12].

For predicting the ADMET properties, we utilized the QikProp module (Schrödinger, LLC, NY). The various physically significant descriptors and pharmaceutically relevant properties of organic molecules are nowadays predicted by QikProp. In addition to predicting molecular properties, QikProp provides ranges for comparing a particular molecule’s properties with those of 95% of known drugs. QikProp also flags 30 types of reactive functional groups that may cause false positives in high-throughput screening (HTS) assays [8].

Molecular dynamics (MD) simulations study was performed with the docking complex of Pantothenate synthase with molecule 26k by using Desmond 3.8 as implemented in Schrödinger package with 1.2 ns (nanoseconds) simulation time. The OPLS-2005 molecular mechanic's force field was used for performing the initial steps. The full system of 39341 atoms was simulated through the multistep MD protocols of Maestro version 11.1.011; release 2017. The trajectory was kept at 4.8 ps time and number of frames at 250. Ensemble class was kept at NPT with temperature 300k and pressure at 1.01325 bar. We used the option for relaxing the model before simulation. Finally, the job was allowed to run [7].

The actives and inactives are the two parts of our dataset molecules which were used for developing the CPH.

For CPH generation and Atom-based 3D-QSAR analysis, a dataset of 38 compounds was used. All these operations were carried out using PHASE drug design software (Schrodinger 2017 release, Inc.). The computations were performed on a Linux based computer having a processor of 2 GHz and memory 512 RAM. Macromodel with OPLS 2005 force field was used in minimizing the dataset structures. Previously minimized LigPrep structures were allowed to import in PHASE having protonation states at physiological pH 7.4 ± 0.0. Dataset was defined with criteria like - compounds with MIC (-log) >5.40 mol/l were considered as active, while those with MIC (-log) < 4.30 mol/l as inactive. Default pharmacophore features in PHASE are hydrogen bond acceptor (A), hydrogen bond donor (D), hydrophobic (H), negative (N), positive (P) and aromatic ring (R). Default definitions were used for generating the hypothesis. Finding common pharmacophore, the maximum number of sites was set to 5 and minimum to 3. Box size of pharmacophore was adjusted to 2 Å. Active and inactive molecules were scored for a given pharmacophore using default weights of scoring parameters. Top ranking hypotheses were subjected to 3D-QSAR analysis for which grid spacing was 1 Å and maximum PLS factors 3. 27 molecules to the training set and 11 molecules to the test set were assigned, based on the variation in structure and biological activity [13-18].

The 3DQSAR models from PHASE are either based on pharmacophore or on atoms [15]. Since all the ligands used are in the congeneric series, that is why we used 3D-QSAR models based on atoms. These were generated using the best 5 point HHPRR hypothesis. The hypothesis was generated using 27 molecule training set and a grid spacing of 1.0 Q. 3D-QSAR model with three PLS factors was generated and further, validated with predicting activities of molecules in the test set [15].

The data set including the inbound (cocrystallized ligand) was docked into the binding pocket of pantothenate synthetase (PDBID: 3IVX) target enzyme. For the reliability of the docking protocol, we evaluated it through the RMSD value. The dataset 38 molecules were allowed for docking protocol to dock into the active site of pantothenate synthetase (PDBID-3IVX), which indicates hydrogen bond interactions with residues Gly158, Met195, Pro38 and also shows further Pi-cation interactions with a residue like Hie47. The average RMSD value was found to be 0.083 Ǻ. The result of docking studies in form of docking score is shown in Table 2. From the results obtained, we found that molecule 26k Fig. (2) has the highest binding affinity towards Ps (Pantothenate synthetase) with XP Gscore of -7.370 kcal/mol while natural ligand has binding affinity compared to molecule 26k with XP Gscore of -9.746 kcal/mol. We also, observed that molecule 27g from the dataset has the lowest binding affinity towards Ps with XP Gscore of –0. 972 kcal/mol Fig. (3). From Prime MMGBSA, we found that the DG binding energy for most docked molecule 26k (-63.425749 Kcal/mol) is more negative than least dock, 27g (-62.070189 Kcal/mol).

Fig. (2). Ligand interaction diagram with the most docked molecule 26k .

Fig. (3). Ligand interaction diagram with least docked 27g with Ps.

QikProp predicted ADMET properties i.e parameters offered the information related with molecular weight, partition coefficient (QPlogPo/w), stars, % oral absorption values and Lipinski’s properties (Rule of Five) and apparent Caco-2 cell permeability (QPPCaco) in nm/sec. The partition coefficient (QPlogPo/w) was found within the range for all our dataset molecules. Overall, the percentage of human oral absorption for compounds was found to be very good. Table 3 indicates that all parameters obtained were found within the acceptable range for human use, thus indicating their potential as a drug-like molecule. The data obtained for most active and least active molecules are mentioned in Table 3.

The docking studies do not involve the flexible nature of the protein. For confirming the exact binding mode and stability, we have to study MD simulations with the Desmond program. The stability of Pantothenate synthetase (Ps) with molecule 26k complex was evaluated through 1.2 ns molecular dynamics simulations. The simulations provided exact binding interaction of the docking complex with system embedded with water molecules, temperature and pressure. The complex was originated in all proper binding poses with an acceptable RMSD value (< 3 Å). The RMSD of backbone atoms of the 3IVX models relative to initial structures indicates that stable molecular systems were obtained during MD (Fig. 4). The RMSD plots indicated that 3IVX-molecule 26k complex was more stable in the simulations study of 1.2ns. We also obtained the protein RMSF plot which shows the local changes that are useful for characterizing local changes in the protein chain (Fig. 5). According to MD simulation study, solvating water molecules were taken for checking the stability of the ligand-protein complex. MD simulations were closer to the physiological environmental conditions. The simulation study can reveal better binding conformations for the docked complex.

Fig. (4). The Root Mean Square Deviations (RMSD) of backbone atoms relative to the starting complexes during 1.2 ns MD.

Fig. (5). Protein RMSF plot (On this plot, peaks indicate areas of the protein that fluctuate the most during the simulation and Protein residues that interact with the ligand are marked with green-colored vertical bars.)