Antiinflammation Derived Suzuki-Coupled Fenbufens as COX-2 Inhibitors: Minilibrary Construction and Bioassay

A small fenbufen library comprising 18 compounds was prepared via Suzuki Miyara coupling. The five-step preparations deliver 9–17% biphenyl compounds in total yield. These fenbufen analogs exert insignificant activity against the IL-1 release as well as inhibiting cyclooxygenase 2 considerably. Both the para-amino and para-hydroxy mono substituents display the most substantial COX-2 inhibition, particularly the latter one showing a comparable activity as celecoxib. The most COX-2 selective and bioactive disubstituted compound encompasses one electron-withdrawing methyl and one electron-donating fluoro groups in one arene. COX-2 is selective but not COX-2 to bioactive compounds that contain both two electron-withdrawing groups; disubstituted analogs with both resonance-formable electron-donating dihydroxy groups display high COX-2 activity but inferior COX-2 selectivity. In silico simulation and modeling for three COX-2 active—p-fluoro, p-hydroxy and p-amino—fenbufens show a preferable docking to COX-2 than COX-1. The most stabilization by the p-hydroxy fenbufen with COX-2 predicted by theoretical simulation is consistent with its prominent COX-2 inhibition resulting from experiments.


Introduction
Nonsteroidal anti-inflammatory drugs (NSAIDs) are characterized as cyclooxygenase (COX) inhibitors. NSAIDs have recently received significant attention, predominantly due to its unidentified adverse effects on treatments of the COVID-19 pandemic. In the course of COVID-19 therapy, one of the NSAIDs, Ibuprofen, induces the overexpression of angiotensin converting enzyme (ACE-2) receptor which may enable the entrance of SARS coronaviruses into the host cells [1]. Thus, the World Health Organization discouraged the repurposing of COX inhibitor.
The rising concerns of the adverse effects have been challenged by the report by Ong et al. [2]. They performed a large randomized controlled trial for the infected patients to assess COX inhibitor NSAIDs and COXIBs as appropriate therapeutics or adjuvant drugs against COVID-19. The supporting drugs administered with the NSAID 'Etoricoxib' reduced the levels of Interleukin-6, thus requiring no noninvasive or invasive ventilation or transfer to the intensive care units. Interestingly, the supporting drugs administered with NSAIDs did not develop adverse effects typically found in the cyclooxygenase inhibition A facial method is mediated through Suzuki Miyara (SM) coupling of one benzene ring that is derivatized from commercial source or self-preparation. The corresponding fenbufen derivatives 5a-5o were generated as shown in Scheme 1. Starting from bromobenzene via Friedel-Crafts acylation to introduce side chain 2, followed by borylation 3 [19], deprotection and the final SM coupling with the bromoarene derivatives, 15 fenbufen analogs can be generated [20][21][22][23][24][25]. Followed by further deprotection of the acid moiety, the final products 6a-6o were obtained [21,[26][27][28][29]. The above reactions are all performable with a rational total yield of 2-19% in five steps. Hydrolysis of the boronic ester 3 is a reversible reaction. [30] Although both stoichiometric amount and catalytic amount of 1 N HCl(aq) have been reported, only catalytic amount is capable of delivering an optimized yield in our hand. The byproduct pinacol is easily removed and the crude product after extraction can be forwarded to the subsequent reaction.
SM coupling is a well-known carbon-carbon formation reaction. It can be affected by the catalyst, base and solvent. The combination of Pd(PPh3)4 or Pd(PPh3)2Cl2, Na2CO3 and A facial method is mediated through Suzuki Miyara (SM) coupling of one benzene ring that is derivatized from commercial source or self-preparation. The corresponding fenbufen derivatives 5a-5o were generated as shown in Scheme 1. Starting from bromobenzene via Friedel-Crafts acylation to introduce side chain 2, followed by borylation 3 [19], deprotection and the final SM coupling with the bromoarene derivatives, 15 fenbufen analogs can be generated [20][21][22][23][24][25]. Followed by further deprotection of the acid moiety, the final products 6a-6o were obtained [21,[26][27][28][29]. A facial method is mediated through Suzuki Miyara (SM) coupling of one benzene ring that is derivatized from commercial source or self-preparation. The corresponding fenbufen derivatives 5a-5o were generated as shown in Scheme 1. Starting from bromobenzene via Friedel-Crafts acylation to introduce side chain 2, followed by borylation 3 [19], deprotection and the final SM coupling with the bromoarene derivatives, 15 fenbufen analogs can be generated [20][21][22][23][24][25]. Followed by further deprotection of the acid moiety, the final products 6a-6o were obtained [21,[26][27][28][29]. The above reactions are all performable with a rational total yield of 2-19% in five steps. Hydrolysis of the boronic ester 3 is a reversible reaction. [30] Although both stoichiometric amount and catalytic amount of 1 N HCl(aq) have been reported, only catalytic amount is capable of delivering an optimized yield in our hand. The byproduct pinacol is easily removed and the crude product after extraction can be forwarded to the subsequent reaction.
SM coupling is a well-known carbon-carbon formation reaction. It can be affected by the catalyst, base and solvent. The combination of Pd(PPh3)4 or Pd(PPh3)2Cl2, Na2CO3 and The above reactions are all performable with a rational total yield of 2-19% in five steps. Hydrolysis of the boronic ester 3 is a reversible reaction [30]. Although both stoichiometric amount and catalytic amount of 1 N HCl (aq) have been reported, only catalytic amount is capable of delivering an optimized yield in our hand. The byproduct pinacol is easily removed and the crude product after extraction can be forwarded to the subsequent reaction.
SM coupling is a well-known carbon-carbon formation reaction. It can be affected by the catalyst, base and solvent. The combination of Pd(PPh 3 ) 4 or Pd(PPh 3 ) 2 Cl 2 , Na 2 CO 3 and dimethyloxyethane (DME) is mostly utilized [31]. When coupling the boronic acid 4 and the bromo counterpart according to the common reaction condition, only 6% yield of fenbufen analog 5d was obtained (Table 1, entry 1). When trying other solvent and catalyst (entry 2) [32], we could merely observe the formation of the desired product 5d at 4 h but it faded away at 18.5 h post the reaction along with the presence of a more concentrated polar unknown byproduct. Even after an attempt to shorten the reaction time, the yield remains unsatisfactory (entry 3 and 4). ous K2CO3 and toluene, or the solid of K3PO4·nH2O or K2CO3 in toluene. While trying the condition of K2CO3, toluene and PdCl2 [34], reaction did not take place until the addition of a small amount of EtOH under assistance of a gradual increase of temperature to 75 °C from 90 min to 100 min post the reaction and a prolonged stirring overnight (entry 5). The isolated product fraction from flash chromatography was purified with HPLC to afford two fractions which were identified as a methyl ester 5b and an ethyl ester 5bbyp at a ratio of 1:1. The transesterification was not observed at the beginning of the reaction but it was observable after a cook. Thus, to prevent S-M coupling from the thermodynamic predominant transesterification, it needs to be stopped at 70 °C at 30 min post reaction in spite of a small amount of starting material remaining (entry 6). When substituting MeOH for EtOH, although the reaction proceeded rather fast, the yield did not improve significantly (entry 7). The optimized condition was met without transesterification when carried out at rt for 3 h (entry 8). ous K2CO3 and toluene, or the solid of K3PO4·nH2O or K2CO3 in toluene. While trying the condition of K2CO3, toluene and PdCl2 [34], reaction did not take place until the addition of a small amount of EtOH under assistance of a gradual increase of temperature to 75 °C from 90 min to 100 min post the reaction and a prolonged stirring overnight (entry 5). The isolated product fraction from flash chromatography was purified with HPLC to afford two fractions which were identified as a methyl ester 5b and an ethyl ester 5bbyp at a ratio of 1:1. The transesterification was not observed at the beginning of the reaction but it was observable after a cook. Thus, to prevent S-M coupling from the thermodynamic predominant transesterification, it needs to be stopped at 70 °C at 30 min post reaction in spite of a small amount of starting material remaining (entry 6). When substituting MeOH for EtOH, although the reaction proceeded rather fast, the yield did not improve significantly (entry 7). The optimized condition was met without transesterification when carried out at rt for 3 h (entry 8). The SM coupling of the 1,2,4-tribromobenzene with organoboronic acid 4 generated three classes of products: tri-, di-and the major mono-coupled products. The fraction of mono coupled product mixtures from column chromatography could be further separated through HPLC to give three fractions at a yield ratio of 30:14:24 ( Figure 2). As expected, the most steric hinderance C-2 decreases the yield of 5h. Compared to Pd(PPh3)4, the less bulky catalyst PdCl2 assist the SM coupling to all three positions [35]. The three compounds are distinguishable using 1 H-NMR spectroscopy by comparing the deshielding effects arisen from the closeness to the two bromo groups [36].  Miyaura reported that basic condition may saponify the ester group, racemize chiral compounds and disable the condensation between aldehyde and alcohol [33]. Ester can be kept intact by employing a heterogeneous condition, such as the combination of aqueous K 2 CO 3 and toluene, or the solid of K 3 PO 4 ·nH 2 O or K 2 CO 3 in toluene. While trying the condition of K 2 CO 3 , toluene and PdCl 2 [34], reaction did not take place until the addition of a small amount of EtOH under assistance of a gradual increase of temperature to 75 • C from 90 min to 100 min post the reaction and a prolonged stirring overnight (entry 5). The isolated product fraction from flash chromatography was purified with HPLC to afford two fractions which were identified as a methyl ester 5b and an ethyl ester 5bbyp at a ratio of 1:1. The transesterification was not observed at the beginning of the reaction but it was observable after a cook. Thus, to prevent S-M coupling from the thermodynamic predominant transesterification, it needs to be stopped at 70 • C at 30 min post reaction in spite of a small amount of starting material remaining (entry 6). When substituting MeOH for EtOH, although the reaction proceeded rather fast, the yield did not improve significantly (entry 7). The optimized condition was met without transesterification when carried out at rt for 3 h (entry 8).
The SM coupling of the 1,2,4-tribromobenzene with organoboronic acid 4 generated three classes of products: tri-, di-and the major mono-coupled products. The fraction of mono coupled product mixtures from column chromatography could be further separated through HPLC to give three fractions at a yield ratio of 30:14:24 ( Figure 2). As expected, the most steric hinderance C-2 decreases the yield of 5h. Compared to Pd(PPh 3 ) 4 , the less bulky catalyst PdCl 2 assist the SM coupling to all three positions [35]. The three compounds are distinguishable using 1 H-NMR spectroscopy by comparing the deshielding effects arisen from the closeness to the two bromo groups [36]. The SM coupling of the 1,2,4-tribromobenzene with organoboronic acid 4 generated three classes of products: tri-, di-and the major mono-coupled products. The fraction of mono coupled product mixtures from column chromatography could be further separated through HPLC to give three fractions at a yield ratio of 30:14:24 ( Figure 2). As expected, the most steric hinderance C-2 decreases the yield of 5h. Compared to Pd(PPh3)4, the less bulky catalyst PdCl2 assist the SM coupling to all three positions [35]. The three compounds are distinguishable using 1 H-NMR spectroscopy by comparing the deshielding effects arisen from the closeness to the two bromo groups [36]. In the course of the coupling, a homocoupled byproduct was always observed from TLC [30]. The byproduct did not really disturb the purification except the compound 5o. The polar NH2 group forms a hydrogen bond with SiO2. NEt3 as a common co-eluent for chromatography will render the current mobile phases inadequate because of deterioration of the theoretical plates. Only when substituting EtOAc/CH2Cl2/NEt3 or CH3OH/CH2Cl2/NEt3 with acetone/n-hexane/NEt3 = 2:8:0.3, a rough purification was allowed. However, the solubility remains an unresolved issue even after adding CHCl3; the mixture will gradually precipitate in the column chromatography resulting in blockade. Nevertheless, the homocoupled byproduct could be removed in the next tosylation. The final deprotection of the acid group could be accomplished using CF3COOH at 120 °C whereas NaOH in CH3OH (aq) is a common condition [19,37].
In view of the contribution of sulfonamide and sulfonylurea to the pharmacophore for inhibiting NLRP3 [13,38], e.g., MCC950 and glyburide [39,40], we therefore introduced a tolylsulfonyl group to mimic the pharmacophore (Scheme 2). Followed by tosylation and acid deprotection, the tosylate 8 was assessed for its bioactivity. To prevent the deprotected product from forming a TFA-co-crystalized complex that may alter the bioassay, LiOH was employed instead.
The following compounds are also included in the bioassay ( Figure 3). Compound 9 [41] and 10 were each obtained from the ester congeners that had been reported before [19]. Furthermore, compound 11 was obtained from its ester precursor 12 that was prepared in a similar S-M coupling. In contrast to the present organoboron laying on the arene already installed with a carboxylic acid group, the corresponding boron building block for compound 12 is on the other benzene moiety. In the course of the coupling, a homocoupled byproduct was always observed from TLC [30]. The byproduct did not really disturb the purification except the compound 5o. The polar NH 2 group forms a hydrogen bond with SiO 2 . NEt 3 as a common coeluent for chromatography will render the current mobile phases inadequate because of deterioration of the theoretical plates. Only when substituting EtOAc/CH 2 Cl 2 /NEt 3 or CH 3 OH/CH 2 Cl 2 /NEt 3 with acetone/n-hexane/NEt 3 = 2:8:0.3, a rough purification was allowed. However, the solubility remains an unresolved issue even after adding CHCl 3 ; the mixture will gradually precipitate in the column chromatography resulting in blockade. Nevertheless, the homocoupled byproduct could be removed in the next tosylation. The final deprotection of the acid group could be accomplished using CF 3 COOH at 120 • C whereas NaOH in CH 3 OH (aq) is a common condition [19,37].
In view of the contribution of sulfonamide and sulfonylurea to the pharmacophore for inhibiting NLRP3 [13,38], e.g., MCC950 and glyburide [39,40], we therefore introduced a tolylsulfonyl group to mimic the pharmacophore (Scheme 2). Followed by tosylation and acid deprotection, the tosylate 8 was assessed for its bioactivity. To prevent the deprotected product from forming a TFA-co-crystalized complex that may alter the bioassay, LiOH was employed instead. The following compounds are also included in the bioassay ( Figure 3). Compound 9 [41] and 10 were each obtained from the ester congeners that had been reported before [19]. Furthermore, compound 11 was obtained from its ester precursor 12 that was prepared in a similar S-M coupling. In contrast to the present organoboron laying on the arene already installed with a carboxylic acid group, the corresponding boron building block for compound 12 is on the other benzene moiety.  As shown in Figure 4, all the fenbufen analogs do not inhibit the release of IL-1 to a satisfactory level. The two rigid biaryl rings without a heteroatom to link between them may hamper the activity as evidenced from the NLRP-3 potent flufenamic acid encompassing an azo between two arenes [13]. The COX inhibition was assessed using the commercial assay kit Cayman (No. 560131). The whole assay procedure is divided into two parts: COX inhibition and ELISA staining. The detection by an enzyme-linked immune assay (ELISA) is aimed at comparing 20 compounds tested to the standard COX-2 selective celecoxib and COX-1 selective resveratrol in a qualitative manner. To assess their bioactivities simultaneously, the procedure was modified to fit the requirements for a minimal volume of 20 μL by each multipipetting. Visible light absorbance of both the groups of void COX and fully active COX are well within the meaningful ranges guided by the assay kit. In some batches of experiments, the reaction tube embedded in a holder makes heat transfer inefficient. Thus, the reaction temperature may be lower than the optimal 37 °C rendering the reaction incomplete, thus voiding the result. Hence, the current data is grouped on the basis of a comparison of compounds using concentration of 22 μM  As shown in Figure 4, all the fenbufen analogs do not inhibit the release of IL-1 to a satisfactory level. The two rigid biaryl rings without a heteroatom to link between them may hamper the activity as evidenced from the NLRP-3 potent flufenamic acid encompassing an azo between two arenes [13]. The COX inhibition was assessed using the commercial assay kit Cayman (No. 560131). The whole assay procedure is divided into two parts: COX inhibition and ELISA staining. The detection by an enzyme-linked immune assay (ELISA) is aimed at comparing 20 compounds tested to the standard COX-2 selective celecoxib and COX-1 selective resveratrol in a qualitative manner. To assess their bioactivities simultaneously, the procedure was modified to fit the requirements for a minimal volume of 20 µL by each multipipetting. Visible light absorbance of both the groups of void COX and fully active COX are well within the meaningful ranges guided by the assay kit. In some batches of experiments, the reaction tube embedded in a holder makes heat transfer inefficient. Thus, the reaction temperature may be lower than the optimal 37 • C rendering the reaction incomplete, thus voiding the result. Hence, the current data is grouped on the basis of a comparison of compounds using concentration of 22 µM (Figures 5-7).
In general, all these fenbufen compounds are COX-2 selective (COX-2/COX-1 > 3). Some of them also show comparable COX-2 inhibition to that of celecoxib, such as the monosubstituted analogs para-fluoro 6a, p-hydroxy 6l and p-amino 6o analogs. The former two compounds show 8-fold inhibition of COX-2 compared to COX-1, whereas 6o shows 60-fold inhibition of COX-2 compared to COX-1. Concerning the disubstituted analogs, in spite of the high COX-2 selectivities by 6b, 6g, 6i and 6j, only 6b exhibits an acceptable COX-2 inhibition. Compound 6b contains one ortho-methyl as an electron-donating group and one para-fluoro as an electron-withdrawing group. The other three isosteres enclose both electron-withdrawing 4th periodic bromo groups. In addition, when encompassing two resonance-formable electron-donating groups, such as oand p-dihydroxy compound 6m, the COX-2 selectivity diminishes while COX-2 activity is reasonable.        In general, all these fenbufen compounds are COX-2 selective (COX-2/COX-1 > 3). Some of them also show comparable COX-2 inhibition to that of celecoxib, such as the monosubstituted analogs para-fluoro 6a, p-hydroxy 6l and p-amino 6o analogs. The former   In general, all these fenbufen compounds are COX-2 selective (COX-2/COX-1 > 3). Some of them also show comparable COX-2 inhibition to that of celecoxib, such as the monosubstituted analogs para-fluoro 6a, p-hydroxy 6l and p-amino 6o analogs. The former  Because of the substantial bioactivities of p-fluoro, p-hydroxy and p-amino fenbufen compounds 6a, 6l, 6o, a further docking study was performed using in silico simulation and modeling. Two enzyme-sequence templates were retrieved from PDB bank encompassing COX-1 (1EQG) complexation with ibuprofen, a nonselective NSAID [42] and COX-2 (1CX2) complexation with SC558 (bromocelecoxib) [43], a COX-2 specific inhibitor. Before the molecular docking (MD) simulation, the receptor moieties, i.e., the two COX enzymes were both administered in the CHARMm force field throughout the whole docking process.
The flexible receptor atom property is enabled by creating a sphere radiating from a center defined by the PDB crystal data with a radius of 4Ǻ (Section 4.4.1) [42,43]. The two sites contain mostly involved residues including Arg120 and Tyr355 for COX-1 and His90, Gln192, Arg513, Ser353, Tyr355 and Phe518 for COX-2. The in situ ligand minimization algorithm comprises a number of programmings, such as adopted basis Newton-Raphson (NR), steepest descent and conjugate gradient. NR is applied to a subspace of the coordinate vector spanned by the displacement coordinates of the last positions. Steepest descent and conjugate gradient are both used to improve a poor conformation through an iterative method via minimization steps as well as the current gradient to determine the next step. Energy minimization through these procedures will be scored in terms of the function of smart minimizer.
A further algorithm for generating conformations was enabled by adopting the option of FAST mode so that rational numbers of low-energy conformation were obtained at a reasonable of time cost. The entropy component for the ligand conformation was also optimized.
Because the solvent effect plays an important role in the binding calculation, an implicit solvent model was performed with respect to Coulomb repulsion and dielectric attraction using the mode of Poisson-Boltzmann with non-polar surface area (PBSA). PBSA is the most rigorous yet slowest solvent approximation method based on continuum electrostatics. A further scoring function, such as the salt concentration, was also addressed. For example, NaCl as the salt and concentration was set to be 0.145 M.
Through a preliminary flexible docking of COX-1 and COX-2, the fluoro analog 6a generated 36 and 30 docking poses, respectively; the hydroxy analog 6l generated 56 and 96 poses for each; the amino analog 6o provided 56 and 62 poses for each. According to the free energy derivation: ∆G binding = ∆G complex − ∆G ligand − G enzyme , the top five high-scoring poses of each group were included in the binding free energy calculation and an average of the five data of each set are grouped in the Table 2. An exception is the fifth data of 5-hydroxy fenbufen 5l docked to COX-2 showing an extraordinarily large value which is skipped. The current simulation was validated by redocking the benchmarking inhibitor ibuprofen to compare with the pose from the original crystallized COX-1 complex (1QEG) in the experimental (Section 4.4.4). The root-mean-square deviation (RMSD) value of 5.649 Å is larger than in the literature [44]. Similar findings were also observed in the case of redocking of celecoxib in comparison with the crystallized pose of bromo analog (SC558) complexed with COX-2 (1CX2). Whereas the electrondonating methyl group is virtually different from the electronwithdrawing bromo group, the RMSD value of 6.615 Å is comparable to the example of ibuprofen.
The values of free binding energy (∆G binding-CHARMm ) from the first calculation ranging from −60 to −250 kcal/mol is smaller than the reported values of around −10 kcal/mol [44,45]. Calibration by incorporating the solvent term provides the second set of data (∆G binding-MMPBSA ). Whereas the ∆G binding-MMPBSA of three compounds (−13~−35 kcal/mol) are comparable to the literature, the two most COX-2 inhibiting phydroxy fenbufen 6l and celecoxib scoring poor in ∆G binding-MMPBSA = 23 kcal/mol seems to deviate unusually. The results may be due partly to the erroneous input of the original crystallized data of the 1eqg (COX-1) and 1CX2 (COX-2) at resolutions of 2.6 and 3.0 Å, respectively. The preliminary Gibbs free binding energy data under CHARMm condition provided a more consistent trend and formed the basis for comparison. It is noted that the COX-2 benchmarking inhibitor celecoxib scores are very low (∆G binding-CHARMm = −60 kcal/mol) compared with fenbufen analogs (−156 to −248 kcal/mol). Fenbufen analogs may take advantage through the fitting feasibility of linear-like structural flexibility compared with the relatively rigid tri-cyclic structure of celecoxib.
The MD simulations of fluoro analog 6a with COX-1 and COX-2 show that both the long and narrow channels of the two active sites can accommodate fluoro analog 6a (Figures 8 and 9). Similar trends are also observed for the hydroxy and amino analogs 6l, 6o but are akin to bind more deeply in the active sites of COX-2 (Supplementary Materials Figures S1-S4). The free binding energy values of fluoro analog 6a to COX-2 are smaller but larger when docking to COX-1. This is also applicable to both the hydroxy and amino analogs 6l and 6o. Whereas van der Waals contacts between COX-1 and fluoro analog 6a constitute the major stabilization, unfavorable interactions are also emerged and even more than that of COX-2's docking. Hydroxy analog 6l attains stabilization with both dockings of COX-1 and COX-2 in all respects of attractive interaction, such as the van der Waals and electrostatic interactions. Similar interaction patterns are observed in the case of amino analog 6o. The RMSD values for hydroxy and amino analogs 6l, 6o coupled with both enzymes are within reasonable ranges (COX-1 and COX2; 0.91 Å and 1.22 Å; 1.59 Å and 1.25 Å) except fluoro analog 6a, which shows relatively resonant values when docking to COX-1 (2.9 Å) but preserves reasonable values when docking to COX-2 (1.95 Å). In brief, the MD simulation results imply that the three compounds prefer a binding toward COX-2 than COX-1.
As shown in Table 2, while COX-2 inhibition by the benchmarking inhibitor celecoxib is comparable to the three fenbufen compounds, the stabilization predicted from the MD simulation is significantly less than that of the three fenbufen analogs. The mismatch may be caused by the fitting ability in terms of the geometric restrictions as described above. Nevertheless, the comparison among the three fenbufen analogs shows a dependency of the COX-2 activity on the MD simulation. For example, p-hydroxy fenbufen 6l scores the highest stabilization and exerts the most COX-2 inhibition. On the other hand, a similar finding was also observed in the group of p-fluoro and p-amino fenbufen analogs 6a, 6o but with an inverted order, probably due to the very close scoring and the close COX-2 inhibiting efficacy. As shown in Figure 10, the superimposition of p-hydroxy fenbufen 6l and COX-2 inhibitor bromocelecoxib (SC558) arranged themselves in a similar spatial orientation with a similar binding mode prevailing. Whereas the OH group of compound 6l exerts a prominent H-bonding to the OH group of Tyr385; the bromo group of SC558 lacks the corresponding interaction. The van del Waals contact between the aromatic ring of 6l and the nonpolar residues of Val 349 and Ala 527 enhances the stabilization. SC558 exerts a similar interaction but with extra stabilization through the two benzene rings. In addition, the polar groups of Arg 120, Arg 513 and His 90 can engage in the dipole-dipole attractions. The polar sulfone group of SC558 exerts similar interactions. In spite of the structural difference, they both follow a similar binding pattern which may address their equivalent COX-2 inhibition if the difference between bromo and methyl group in SC558 and celecoxib, respectively, could be neglected.   As shown in Table 2, while COX-2 inhibition by the benchmarking inhibitor celecoxib is comparable to the three fenbufen compounds, the stabilization predicted from the MD simulation is significantly less than that of the three fenbufen analogs. The mismatch may be caused by the fitting ability in terms of the geometric restrictions as described above. Nevertheless, the comparison among the three fenbufen analogs shows a dependency of the COX-2 activity on the MD simulation. For example, p-hydroxy fenbufen 6l scores the highest stabilization and exerts the most COX-2 inhibition. On the other hand, a similar finding was also observed in the group of p-fluoro and p-amino fenbufen analogs 6a, 6o but with an inverted order, probably due to the very close scoring and the close COX-2 inhibiting efficacy. As shown in Figure 10, the superimposition of p-hydroxy fenbufen 6l and COX-2 inhibitor bromocelecoxib (SC558) arranged themselves in a similar spatial ori- structural difference, they both follow a similar binding pattern which may address their equivalent COX-2 inhibition if the difference between bromo and methyl group in SC558 and celecoxib, respectively, could be neglected. Figure 10. Overlaying of the most potent p-hydroxy fenbufen 6l with the original bromocelecoxib (SC58). 6l was specified in purple and bromocelecoxib was marked in blood red.

Conclusions
Through the Suzuki-Miyamura coupling reaction, a small library comprising 18 fenbufen compounds was generated. The optimized condition using PdCl2 and the cosolvents of toluene and EtOH can generally provide the coupled biphenyl products within 3 h. In spite of insignificant IL-1β inhibition activity, the NSAIDs derivatives show a typical COX inhibition. As anticipated, the NLRP-3 activity might be improved by introducing a linker such as an azo or a sulfide group. Among them, p-fluoro, p-hydroxyl and p-amino fenbufen analogs 6a, 6l, 6o are better COX-2 inhibitors. The p-hydroxy fenbufen 6l is even better than celecoxib at the concentration level of 22 μM at COX-2 inhibition. All these potential fenbufen analogs exert a better inhibition against COX-2 than COX-1. The COX-2 potent and highly COX-2 selective p-amino fenbufen 6o needs to be further assessed for their inflammatory efficacy in vivo. The p-hydroxy fenbufen 6l showing the most COX-2 inhibition also deserves further study. Sulfone analog 8 shows potential because of having minor NLRP3 activity, a remarkable COX-2 activity and a considerable COX-2 selectivity.
MD simulation using CHARMm force field along with the parameter settings and validation of the program using benchmarking COX-2 inhibitor bromocelecoxib (SC85) and COX-1 inhibitor ibuprofen generates two classes of binding free energy scoring expression with respect to the solvent effects. Whereas the data from incorporating solvent effects biases the results, the preliminary CHARMm-derived data exerts a consistent trend and forms the basis for discussion. The relatively higher stabilization gained by the current simulation (−70 to −190 kcal) than the reported data (~−10 kcal/mol) may be due to the erroneous input and structural features. The former arises from fair crystallographic resolution of the model systems 1QEG and 1CX2 (2.6 Å and 3.0 Å). The latter is related to the more feasibly structural flexibility exerted by the linear-like structural fenbufen than the tricyclic celecoxib analogs. The lowest free binding energy from the MD simulation by Figure 10. Overlaying of the most potent p-hydroxy fenbufen 6l with the original bromocelecoxib (SC58). 6l was specified in purple and bromocelecoxib was marked in blood red.

Conclusions
Through the Suzuki-Miyamura coupling reaction, a small library comprising 18 fenbufen compounds was generated. The optimized condition using PdCl 2 and the cosolvents of toluene and EtOH can generally provide the coupled biphenyl products within 3 h. In spite of insignificant IL-1β inhibition activity, the NSAIDs derivatives show a typical COX inhibition. As anticipated, the NLRP-3 activity might be improved by introducing a linker such as an azo or a sulfide group. Among them, p-fluoro, p-hydroxyl and p-amino fenbufen analogs 6a, 6l, 6o are better COX-2 inhibitors. The p-hydroxy fenbufen 6l is even better than celecoxib at the concentration level of 22 µM at COX-2 inhibition. All these potential fenbufen analogs exert a better inhibition against COX-2 than COX-1. The COX-2 potent and highly COX-2 selective p-amino fenbufen 6o needs to be further assessed for their inflammatory efficacy in vivo. The p-hydroxy fenbufen 6l showing the most COX-2 inhibition also deserves further study. Sulfone analog 8 shows potential because of having minor NLRP3 activity, a remarkable COX-2 activity and a considerable COX-2 selectivity.
MD simulation using CHARMm force field along with the parameter settings and validation of the program using benchmarking COX-2 inhibitor bromocelecoxib (SC85) and COX-1 inhibitor ibuprofen generates two classes of binding free energy scoring expression with respect to the solvent effects. Whereas the data from incorporating solvent effects biases the results, the preliminary CHARMm-derived data exerts a consistent trend and forms the basis for discussion. The relatively higher stabilization gained by the current simulation (−70 to −190 kcal) than the reported data (~−10 kcal/mol) may be due to the erroneous input and structural features. The former arises from fair crystallographic resolution of the model systems 1QEG and 1CX2 (2.6 Å and 3.0 Å). The latter is related to the more feasibly structural flexibility exerted by the linear-like structural fenbufen than the tricyclic celecoxib analogs. The lowest free binding energy from the MD simulation by p-hydroxy fenbufen 6l was consistent with its highest COX-2 inhibiting activity. In addition, p-fluoro and p-amino fenbufen analogs 6a, 6o exert comparable experimental COX-2 inhibitions which are consistent with their equivalent free binding energies as predicted by ∆G binding-CHARMm .
Both p-hydroxy fenbufen 6l and bromocelecoxib (SC558) follow the similar binding pattern to COX-2, irrespective of the difference between methyl group and bromo group.

General Information
Most reagents and solvents were purchased from Fluka (St. Louis, MO, USA), Sigma-Aldrich (St. Louis, MO, USA), Alfa (Binfield, Berkshire, UK), Acros (Geal, Belgium), Showa (Tokyo, Japan) or TCI (Tokyo, Japan). These compounds were performed in dried glassware under a purge of nitrogen at room temperature unless otherwise noted. CH 2 Cl 2 and toluene were dried over CaH 2 . CH 3 OH was dried over Mg and distilled prior to reactions. THF was treated with FeSO 4 ·5H 2 O and dried over KOH followed by filtration and distilled over Na. DMSO and DMF were distilled over CaH 2 under reduced pressure. NEt 3 and pyridine were distilled over CaH 2 . The related reagents and solvents were obtained in reagent grade. The eluents for flash chromatography (e.g., EtOAc, acetone, and n-hexane) were of industrial grade. They were distilled prior to use. CHCl 3 and CH 3 OH were of reagent grade and used without purification. NMR spectroscopy including 1 H-NMR (500 MHz) and 13 C-NMR (125 MHz, DEPT-135) was performed by using a Unity Inova 500 MHz instrument (Varian, USA). Deuterated-solvents including CDCl 3 , CD 3 OD, C 6 D 6 and DMSO-d 6 were purchased from Aldrich (St. Louis, MO, USA). Low-resolution mass spectrometry (LRMS) was carried out on an ESI-MS spectrometer using a Varian 901-MS Liquid Chromatography Tandem Mass Q-TOF Spectrometer at the Department of Chemistry of National Tsing-Hua University (NTHU). LRMS was also performed at the Department of Applied Chemistry of National Chiao-Tung University (NCTU). High-resolution mass spectrometry (HRMS) was carried out using a Varian HPLC (Prostar series ESI/APCI) system coupled with a Varian 901-MS (FT-ICR Mass) mass detector and a triple quadrupole setting. Thin layer chromatography (TLC) was performed with TLC silica gel 60 F 254 pre-coated plates (Machery-Nagel, Dueren, Germany) to monitor the starting materials and products upon visualization under UV light (254 nm). TLC plates were staining with either ninhydrin or ceric ammonium molybdate under heating. Celite 545 was obtained from Macherey-Nagel Inc. (Dueren, Germany). Strong acid cation exchange resin (H + ) was obtained from Amberlite IR-120. Column chromatography was performed using Silicycle 60 silica gel (60-200 mesh, Quebec City, QC, Canada) under a slight pressure. Melting points were measured with a MEL-TEMP instrument (Barnstead international, Dubuque, IA, USA) without correction.
Normal phase HPLC constitutes an Agilent isocratic 1100 pump that was connected by a UV-VIS detector (254 nm) and a column of ZORBAX SIL column (9.4 mm × 250 mm, 5 µm), a combination of EtOAc and n-hexane as the mobile phase at a flow rate of 3 mL/min. A Rheodyne injector with a loop of 0.5 mL was employed.

Chemical Preparation
para-Bromofenbufen methyl ester 2 p-hydroxy fenbufen 6l was consistent with its highest COX-2 inhibiting activity. In addition, p-fluoro and p-amino fenbufen analogs 6a, 6o exert comparable experimental COX-2 inhibitions which are consistent with their equivalent free binding energies as predicted by ΔGbinding-CHARMm.
Both p-hydroxy fenbufen 6l and bromocelecoxib (SC558) follow the similar binding pattern to COX-2, irrespective of the difference between methyl group and bromo group.

General Information
Most reagents and solvents were purchased from Fluka (St. Louis, MO, USA), Sigma-Aldrich (St. Louis, MO, USA), Alfa (Binfield, Berkshire, UK), Acros (Geal, Belgium), Showa (Tokyo, Japan) or TCI (Tokyo, Japan). These compounds were performed in dried glassware under a purge of nitrogen at room temperature unless otherwise noted. CH2Cl2 and toluene were dried over CaH2. CH3OH was dried over Mg and distilled prior to reactions. THF was treated with FeSO4⋅5H2O and dried over KOH followed by filtration and distilled over Na. DMSO and DMF were distilled over CaH2 under reduced pressure. NEt3 and pyridine were distilled over CaH2. The related reagents and solvents were obtained in reagent grade. The eluents for flash chromatography (e.g., EtOAc, acetone, and n-hexane) were of industrial grade. They were distilled prior to use. CHCl3 and CH3OH were of reagent grade and used without purification. NMR spectroscopy including 1 H-NMR (500 MHz) and 13 C-NMR (125 MHz, DEPT-135) was performed by using a Unity Inova 500 MHz instrument (Varian, USA). Deuterated-solvents including CDCl3, CD3OD, C6D6 and DMSO-d6 were purchased from Aldrich (St. Louis, MO, USA). Low-resolution mass spectrometry (LRMS) was carried out on an ESI-MS spectrometer using a Varian 901-MS Liquid Chromatography Tandem Mass Q-TOF Spectrometer at the Department of Chemistry of National Tsing-Hua University (NTHU). LRMS was also performed at the Department of Applied Chemistry of National Chiao-Tung University (NCTU). High-resolution mass spectrometry (HRMS) was carried out using a Varian HPLC (Prostar series ESI/APCI) system coupled with a Varian 901-MS (FT-ICR Mass) mass detector and a triple quadrupole setting. Thin layer chromatography (TLC) was performed with TLC silica gel 60 F254 precoated plates (Machery-Nagel, Dueren, Germany) to monitor the starting materials and products upon visualization under UV light (254 nm). TLC plates were staining with either ninhydrin or ceric ammonium molybdate under heating. Celite 545 was obtained from Macherey-Nagel Inc. (Dueren, Germany). Strong acid cation exchange resin (H + ) was obtained from Amberlite IR-120. Column chromatography was performed using Silicycle 60 silica gel (60-200 mesh, Quebec City, QC, Canada) under a slight pressure. Melting points were measured with a MEL-TEMP instrument (Barnstead international, Dubuque, IA, USA) without correction.
Normal phase HPLC constitutes an Agilent isocratic 1100 pump that was connected by a UV-VIS detector (254 nm) and a column of ZORBAX SIL column (9.4 mm × 250 mm, 5 μm), a combination of EtOAc and n-hexane as the mobile phase at a flow rate of 3 mL/min. A Rheodyne injector with a loop of 0.5 mL was employed.

Chemical Preparation para-Bromofenbufen methyl ester 2
To a three-neck round bottomed flask was charged a mixture of succinic anhydride (31 g, 0.31 mol, 1.2 eq). CH 2 Cl 2 (50 mL) was added and the mixture was stirred until dissolution. The mixture was cooled down by an ice bath followed by adding AlCl 3 (104 g, 0.78 mol, 3 eq). The bath was removed and bromobenzene (27 mL, 0.26 mol) was added. CH 2 Cl 2 (50 mL) was added and the sticky solid was agitated using a spatula. The mixture was poured into a mixture containing ice (120 g) and HCl (12 N, 65 mL) followed by Büchner filtration through suction. The white residue was further concentrated under reduced pressure. While attempting to recrystallize by using CH 3 OH for dissolution, a significant amount of methylated product was obtained but not to completeness. The mixture (35.5 g) after concentration under reduced pressure was submitted to an acid protection procedure as follows. The mixture was dried by distilling with toluene (10 mL) and CH 2 Cl 2 (10 mL). CH 3 OH (100 mL) and H 2 SO 4 (7.5 mL) was added. Stirring was allowed for 40 min followed by extraction using EtOAc (200 mL), Na 2 CO 3 (satd., 50 mL × 2). The organic layers were collected and dried over Na 2 SO 4 followed by gravitational filtration. The filtrate was concentrated under reduced pressure to give 49% yield of the white solid (35 g, 0.13 mol) over two steps (m.p. 48-50 • C (lit. [46] m.p. 51.5 • C). To a three-neck round bottomed flask was charged a mixture of succinic anhydride (31 g, 0.31 mol, 1.2 eq). CH2Cl2 (50 mL) was added and the mixture was stirred until dissolution. The mixture was cooled down by an ice bath followed by adding AlCl3 (104 g, 0.78 mol, 3 eq). The bath was removed and bromobenzene (27 mL, 0.26 mol) was added. CH2Cl2 (50 mL) was added and the sticky solid was agitated using a spatula. The mixture was poured into a mixture containing ice (120 g) and HCl (12 N, 65 mL) followed by Büchner filtration through suction. The white residue was further concentrated under reduced pressure. While attempting to recrystallize by using CH3OH for dissolution, a significant amount of methylated product was obtained but not to completeness. The mixture (35.5 g) after concentration under reduced pressure was submitted to an acid protection procedure as follows. The mixture was dried by distilling with toluene (10 mL) and CH2Cl2 (10 mL). CH3OH (100 mL) and H2SO4 (7.5 mL) was added. Stirring was allowed for 40 min followed by extraction using EtOAc (200 mL), Na2CO3 (satd., 50 mL × 2). The organic layers were collected and dried over Na2SO4 followed by gravitational filtration. The filtrate was concentrated under reduced pressure to give 49% yield of the white solid (35 g, 0.13 mol) over two steps (m.p. 48-50 °C (lit. [46] m.p. 51.5 °C). A two-necked round-bottomed flask (50 mL) encompassing KOAC (7.3 g, 73.8 mmol, 4.0 eq) was dried in an oven at 120 °C for over 96 h. A second flask (50 mL) charging diboronopinacol (9.4 g, 36.9 mmol, 2.0 eq) was dried by distillation with toluene at 50 °C under reduced pressure. Bromo compound 2 (5.0 g, 18.5 mmol, 1 eq) was added and codistilled with toluene. A spin-like flask (50 mL) containing a complex of dichloro-[1,1′bis(diphenylphosphino)ferrocene] palladium (II) [PdCl2(dppf)] (500 mg, 0.68 mmol, 10.0% wt) was dried twice with toluene by co-distillation at 55 °C under reduced pressure. A flask containing dried DMSO (over 4 Å MS for 48 h) was bubbled with N2 for 15 min. To the two-necked round bottomed flask containing KOAc, a mixture of bromo compound and diboronpinacol in DMSO (30 mL) and a solution of PdCl2(dppf) in DMSO (10 mL) were added sequentially under sufficient stirring. The mixture was moved to an oil bath that had been preheated to 90 °C and the stirring was continued through bubbling. The mixture turned from light orange to dark brown after 10 min. TLC (EtOAc/n-hexane 2:8) indicated the consumption of the starting material 2 (Rf = 0.52) and formation of the product 3 (Rf = 0.52) with intense blue after staining. The reaction was terminated at 110 min post reaction by partitioning between CH2Cl2 (40 mL) and HCl (aq. 0.2 N, 20 mL). The organic layer was dried over Mg2SO4 followed by filtration using a celite pad and concentration under reduced pressure. The black residue (25 g) contained a significant amount of DMSO that can be further reduced by a second extraction or purified through flash chromatography using the gradient mode of EtOAc/n-hexane 1:9 → 2:8 to give a pleasant Hinoki-essential oil-odor pale yellow viscous gum in a 78% yield (4.6 g).
A two-necked round-bottomed flask (50 mL) encompassing KOAC (7.3 g, 73.8 mmol, 4.0 eq) was dried in an oven at 120 • C for over 96 h. A second flask (50 mL) charging diboronopinacol (9.4 g, 36.9 mmol, 2.0 eq) was dried by distillation with toluene at 50 • C under reduced pressure. Bromo compound 2 (5.0 g, 18.5 mmol, 1 eq) was added and co-distilled with toluene. A spin-like flask (50 mL) containing a complex of dichloro-[1,1bis(diphenylphosphino)ferrocene] palladium (II) [PdCl 2 (dppf)] (500 mg, 0.68 mmol, 10.0% wt) was dried twice with toluene by co-distillation at 55 • C under reduced pressure. A flask containing dried DMSO (over 4 Å MS for 48 h) was bubbled with N 2 for 15 min. To the two-necked round bottomed flask containing KOAc, a mixture of bromo compound and diboronpinacol in DMSO (30 mL) and a solution of PdCl 2 (dppf) in DMSO (10 mL) were added sequentially under sufficient stirring. The mixture was moved to an oil bath that had been preheated to 90 • C and the stirring was continued through bubbling. The mixture turned from light orange to dark brown after 10 min. TLC (EtOAc/n-hexane 2:8) indicated the consumption of the starting material 2 (R f = 0.52) and formation of the product 3 (R f = 0.52) with intense blue after staining. The reaction was terminated at 110 min post reaction by partitioning between CH 2 Cl 2 (40 mL) and HCl (aq. 0.2 N, 20 mL). The organic layer was dried over Mg 2 SO 4 followed by filtration using a celite pad and concentration under reduced pressure. The black residue (25 g) contained a significant amount of DMSO that can be further reduced by a second extraction or purified through flash chromatography using the gradient mode of EtOAc/n-hexane 1:9 → 2:8 to give a pleasant Hinoki-essential oil-odor pale yellow viscous gum in a 78% yield (4.6 g).

COX-1 and COX-2 Inhibition Assay
Assay kit (No. 560131) from Cayman was used for screening bioactivity. The flowcharts have been appended as Supplementary Material. In brief, the volume of compound solution and the reagents used for this assay were increased from 10 to 20 µL in order that a multipipetting was performable. Ten microliters of Heme was diluted using 520 µL buffer solution. It was distributed to 7 tubes by 70 µL and the last tube by 40 µL. SnCl2 solution was prepared by dissolving 45 mg in HCl (0.9 mL) and distributing 120 µL to 8 tubes. The arachidonic acid solution was prepared by mixing a portion of 5 µL with 5 µL of KOH (aq) by vortex followed by dissolving with 1 mL H 2 O. It was distributed to 8 tubes in 125 µL amounts for each. A volume of 45 µL of COX enzyme dissolved in 415 µL buffer was ready for the subsequent procedure. In stage 2, all the reagents and compounds were added at volumes of 20 µL or 30 µL per multi pipetting, except the addition of COX enzymes. The last Elisa assay was performed by using the 2000-fold dilution and the addition stage was per multipipetting. The rest of development was the same as that described.

Algorithms and Scoring Functions
In Situ Ligand Minimization Algorithm Available options for establishing the algorithm encompasses adopted basis Newton-Raphson (NR), steepest descent and conjugate gradient. NR is applied to a subspace of the coordinate vector spanned by the displacement coordinates of the last positions. In each step of this iterative procedure, the coordinates are adjusted in the negative direction of the gradient. Steepest descent does not generally converge, but will rapidly improve a very poor conformation. Conjugate gradient is an iterative method which makes use of the previous history of minimization steps as well as the current gradient to determine the next step. It has better convergence characteristics but is subject to numerical overflows when starting with very poor conformations. Energy minimization through these procedures will be scored using the smart minimizer function.

Conformation Method
Algorithm for generating conformations was enabled by adopting the option of FAST mode. This could provide rational numbers of low-energy conformation in a reasonable time.
Entropy Minimization Similar to above in situ ligand minimization algorithm, the entropy component for the ligand conformation was also minimized using the tree approaches.

Implicit Solvent Model
Issues involved in this calculation regard the Coulomb repulsion and dielectric attraction. The option of Poisson-Boltzmann with non-polar surface area (PBSA) was Figure 11. The plots of three-dimensional COX-1 (1EQG) and COX-2 (1CX2) are shown in the left and right panels. Each of them contains a grey sphere redefined as an active site with a 4 Å radius. The residues marked with yellow were suggested to be involved in the binding.

Algorithms and Scoring Functions In Situ Ligand Minimization Algorithm
Available options for establishing the algorithm encompasses adopted basis Newton-Raphson (NR), steepest descent and conjugate gradient. NR is applied to a subspace of the coordinate vector spanned by the displacement coordinates of the last positions. In each step of this iterative procedure, the coordinates are adjusted in the negative direction of the gradient. Steepest descent does not generally converge, but will rapidly improve a very poor conformation. Conjugate gradient is an iterative method which makes use of the previous history of minimization steps as well as the current gradient to determine the next step. It has better convergence characteristics but is subject to numerical overflows when starting with very poor conformations. Energy minimization through these procedures will be scored using the smart minimizer function.

Conformation Method
Algorithm for generating conformations was enabled by adopting the option of FAST mode. This could provide rational numbers of low-energy conformation in a reasonable time.
Entropy Minimization Similar to above in situ ligand minimization algorithm, the entropy component for the ligand conformation was also minimized using the tree approaches.

Implicit Solvent Model
Issues involved in this calculation regard the Coulomb repulsion and dielectric attraction. The option of Poisson-Boltzmann with non-polar surface area (PBSA) was adopted. PBSA is the most rigorous yet slowest solvent approximation method based on continuum electrostatics. This may not be available if in situ ligand minimization is running.

Implicit Solvent Dielectric Constant
The model used for dielectric constant for bulk solvent used is PBSA.

Salt Concentration
The solvation condition was generated through common settings, such as using NaCl as the salt and concentration was set to 0.145 M.

Simulation Procedure
The docking simulation was performed using DS 2021 software integrated with the DS flexible docking protocol mode. The flexible algorithm allows the bound residue to fit in a reasonable manner. The pretreatment was the same as that described for the former docking simulation in Section 4.4.1. The solvation condition was generated through common settings, such as using NaCl as the salt and concentration in 0.145 M. The subsequent standard dynamics cascade protocol simulates the molecular dynamics through energy minimization and a number of stages, e.g., heating, equilibration and production. The procedures of steepest descent and conjugate gradient were used to converge as described above. The isothermal and isobaric ensemble conditions were set to perform the dynamics calculation. The candidate conformation was further analyzed in terms of trajectory based on the deviation from the initial atomic state which was described by two functions of root-mean-square deviation, RMSD, and root-mean-square fluctuation, RMSF.
The parameters were set to allow for maximum conformation numbers of ligand of 255, docking numbers of hotspot of 100 and energy threshold of 20. The probable conformations were submitted to DS Analyze Ligand Poses and Docking Pose in order to minimize binding free energy. A further validation using DS Calculate Binding Energies to compare the results from the two calculations can identify the most optimized conformation. All the free energy calculations will be enrolled in the following expression: ∆G binding = ∆G complex − ∆G ligand − G enzyme .
The free energy calculation generates two classes of data expression: the preliminary binding energy of the system in the CHARMm condition and, additionally, the free energy resulting from the presence of solvent effect.

Validation of Computational Program
The most optimal docking poses by the two benchmarking inhibitors, i.e., celecoxib and ibuprofen, were each superimposed with bromocelecoxib (SC58) and ibuprofen itself (Figures 12 and 13). The latter two conformations were derived from the original cocrystallized complexes 1EQG and 1CX2, respectively. The two orientations of ibuprofen from the original complex and from simulation showed an RMSD value of 5.649 Å, a value relatively larger than the reported 0.433 Å [44]. The subtle shift in placement may be related to broader defining of the sphere covering the active site. When comparing bromocelecoxib (SC558) with celecoxib, an RMSD value of 6.615 Å is almost comparable to that of the ibuprofen group.  Acknowledgments: Huai-En Yu was acknowledged for his assistance in experiment.