Cognate docking for evaluating algorithm of molecular docking
Before the onset of real molecular docking we tested the algorithm for its ability to detect the active site properly and to reproduce the experimentally solved pose of the ligand molecule. After redocking the co-crystallized or native ligand of the COX-2 enzyme, we carefully quantified the RMSD between the two ligand poses (one being experimentally derived pose and another being algorithm generated). The algorithm proved to detect active site very accurately and was quite capable of emulating the experimental (crystal) pose of the ligand . This conclusion has been derived as the RMSD value obtained was 0.4048 Å (Figure 1). RMSD values ranging from 0 to 2 Å are quite acceptable in small molecule docking studies [28, 34]. After assuring the precision and accuracy of docking algorithm we moved to actual molecular docking for ranking the ligand molecules.
Niflumic acid showed comparably more docking score against COX-2
Among the docked molecules, niflumic acid showed more negative value of docking score against COX-2 as compared to celecoxib. More negative docking score of niflumic acid indicates its stronger inclination towards this enzyme . While celecoxib showed a docking score of −9.31, niflumic acid displayed this score as −10 indicating its relatively stronger binding propensity (Figure 2). As a regular procedure it is common to supplement the findings of one experiment with other for extra surety thus in the subsequent experiment we tried to explore the binding free energy of aforementioned inhibitors with more reliable technique.
MMGBSA study further supported the molecular docking accuracy
Two MM-GBSA experiments were performed for estimating the binding free energy values of docked complexes. Niflumic acid like docking score showed more negative value (lowest score) of binding free energy than celecoxib. While a binding free energy value of −39.5573 kcal/mol was manifested by celecoxib-COX-2 complex, a more negative value was demonstrated by niflumic acid-COX-2 complex (−54.9838 kcal/mol). More negative values of binding free energy in case of niflumic acid-COX-2 indicate stronger interactions than celecoxib-COX-2 (Figure 3) . Now it is evident that niflumic acid showed not only more negative value of docking score but also more binding free energy value against COX-2 in comparison to celecoxib. Till here we proved through molecular docking and MMGBSA method that niflumic acid has relatively favourable affinity over celecoxib against COX-2. In the forthcoming experiments we are going to scrutinize the interaction of these inhibitors individually in docked state with COX-2.
Celecoxib showed more interactions over niflumic acid
Various types of interactions are known to exist between protein receptors and drug molecules. Among these interactions hydrophobic contacts, hydrogen bonds, pi-pi, pi-cation and salt bridges are prominent. Niflumic acid showed seven hydrophobic and two hydrogen bonding interactions with active site residues of COX-2. Among the residues showing hydrophobic contacts LEU 352, TYR 355, TYR 385, TRP 387, VAL 523 and ALA 527 were prominent. Among the hydrogen bonding displaying residues TYR 385 and SER 530 were noticeable. On the other hand celecoxib demonstrated ten hydrophobic and five hydrogen bonds with COX2. Apart from this, one halogen bond was seen with ARG 120 of this enzyme. PHE 518 showed dual contacts (hydrophobic and hydrogen bond) with celecoxib (Figure 4). Thus niflumic acid and celecoxib showed similar but surely not identical interaction profile in the bound state with this inflammation facilitating enzyme. The differences in interaction profile can be ascribed to structural differences between these inhibitors giving rise to distinct interactions within the COX-2 pocket .
After sequential binding proclivity and interaction studies we focussed on dynamic study which can answer the stability of niflumic acid/celecoxib in the binding pocket of COX-2. Thus, in the upcoming steps we employed molecular dynamics simulation approach to address this question.
Niflumic acid showed comparatively more stability in the COX-2 active site
It is well known that binding of ligands to proteins is dynamic thereby involving various dynamic and complex transitions. While a flexible ligand can alter its conformation on binding the target site in a best possible manner, the protein can change its dynamics and conformation for promoting ligand binding . Thus it is obvious that static approaches like molecular docking should be coupled to dynamic approaches like molecular dynamics simulations for gaining more insights about ligand-receptor stability [28, 32]. We subjected celecoxib-COX-2 complex and niflumic acid-COX-2 for 10 ns simulation using the Desmond software. Even molecular dynamics simulation supplemented the molecular docking and MMGBSA conclusions. Niflumic acid demonstrated relatively better stability in the binding pocket of COX-2 as compared to celecoxib. This crux has been taken as niflumic acid showed comparatively lesser fluctuations indicated by the lower RMSD values than celecoxib (Figure 5) . Further, the COX-2 showed lesser local fluctuations in docked state with niflumic acid than with celecoxib as indicated by the RMSF values (Figure 6). Residues of COX-2 crucial for interaction with celecoxib and niflumic acid were also explored by analysing the respective trajectory. Certain residues including SER 530 and TYR 385 were critical in interacting with niflumic acid whereas residues like LEU 352 and ARG 120 were crucial in case of celecoxib (Figure 7). Importantly, one interactions of niflumic acid sustained for 99% (TYR 385) of the simulation time while in case of celecoxib-COX-2 only one interaction remained for 91% (LEU 352) of the simulation duration which again favours stability of niflumic acid over celecoxib (Figure 8).