CoMFA, CoMSIA, kNN MFA and docking studies of 1,2,4-oxadiazole derivatives as potent caspase-3 activators

2.1 Data set

A data set of thirty-two 3-aryl-5-aryl-1,2,4-oxadiazole derivatives with caspase-3 enzyme activity that induced cytotoxicity on the T47D cell line was used in the present study (Table 1) (Zhang et al., 2005a,b; Kemnitzer et al., 2009). Test and training set contained a diverse set of compounds with low, moderate and high biological activity.

Table 1 1,2,4-Oxadiazole analogs and their experimental caspase-3 activator activity.

Close

Export to PPT

S. No.	Ar1	Ar2	Activity T47D (EC50 nM)	pIC50 (T47D)
1d			0.0012	2.9172
4a			0.0018	2.7520
4b			0.0028	2.5496
4c			0.0009	3.0506
4d			0.0042	2.3799
4e			0.0037	2.4365
4g			0.0009	3.0456
4h			0.0014	2.8508
4i			0.0009	3.0315
4j			0.0009	3.0406
4k			0.0014	2.8416
4l			0.0016	2.7986
4m			0.0009	3.0362
4n			0.0008	3.1249
4o			0.0033	2.4789
4v			0.0020	2.7055
10a			0.0011	2.9666
10b			0.0017	2.7595
10c			0.0174	1.7595
10d			0.0020	2.6968
10e			0.0005	3.2840
10f			0.0016	2.7932
10g			0.0006	3.2147
10h			0.0011	2.9547
11a			0.0017	2.7696
11b			0.0024	2.6198
11c			0.0031	2.5086
11d			0.0045	2.3468
11e			0.0011	2.9580
11f			0.0009	3.0268
11g			0.0022	2.6575
11h			0.0029	2.5376

2.2 Molecular structure and alignment

The 3D QSAR molecular modeling and statistical analysis were performed using the molecular modeling package SYBYL Version 6.7 (Tripos Inc., 2001) and V-Life Molecular Design Suit (MDS) software Version 3.5 using the CoMFA, CoMSIA and [(SW) kNN MFA] methods respectively. The 3D structures of all thirty-two oxadiazole compounds were built on the workspace of molecular modeling softwares SYBYL and V-Life MDS. Energy minimization was carried out using the standard tripos force field with distance dependent-dielectric function, energy gradient of 0.001 kcal/mol Å and an iteration limit of 10,000. Partial charges for all the molecules were assigned using the Gasteiger–Huckel method and they were submitted for a conformational search protocol. Bioactive conformations and molecular alignment are two imperative parameters to build more consistent 3D QSAR models. Structure alignment is considered as one of the most sensitive steps in 3D QSAR done with a template based alignment. The most active compound 10e was used as a template and each molecule was superimposed on the template via the field fit alignment (Figs. 1 and 2) (Jain et al., 2010; Vaidya et al., 2011).

Close

Figure 1

Field fit alignment of 1,2,4-oxadiazoles derivatives using SYBYL 6.7.

Export to PPT

Close

Figure 2

Field fit alignment of 1,2,4-oxadiazoles derivatives using V-Life MDS 3.5.

Export to PPT

2.3 Generation of CoMFA, CoMSIA and [(SW) kNN MFA] fields

After alignment molecules were placed in a common rectangular grid that extended 4 Å units beyond the aligned molecules in all directions with a grid step size of 2 Å; a probe sp³ hybridized carbon atom with +1 charge and Van Der Waals radius of 1.52 Å was employed.

In CoMFA, steric and electrostatic field descriptors were calculated with a default cut-off of 30 kcal/mol whereas for CoMSIA all five force field properties (i.e. steric, electrostatic, hydrophobic, hydrogen-bond donor and hydrogen-bond acceptor) were determined at a 30-kcal/mol energy cut-off, which meant that energy fields greater than 30 kcal/mol are curtailed to that value, and thus can avoid infinite energy values inside the molecule.

Similar to CoMFA in the [(SW) kNN MFA] method steric and electrostatic fields were generated with a default energy of 30 kcal/mol.

2.4 Partial Least-Square (PLS) regression analysis and validation of 3D-QSAR model

Partial Least-Square (PLS) regression analysis was used to construct a linear correlation between energy fields (independent variables) with anticancer activity values (dependent variable) and cross-validation was done using the leave-one-out (LOO) method with a 2.0 kcal/mol column filter for CoMFA, CoMSIA and [(SW) kNN MFA] analyses.

The final models were validated on the basis of obtained optimum number of components (N) yielding the highest cross-validated r² (q²) value with a minimum standard of error.

2.5 Molecular docking

Docking studies were employed to locate the appropriate binding orientations and conformations of these oxadiazole derivatives interacting with caspase-3 using the docking program GOLD. Gold is a fast, flexible docking method that uses an incremental construction algorithm to place ligands into an active site. By default, the docking program produces 10 docked structures for each 1,2,4-oxadiazole derivative. The conformation with the lowest docking energy in the most populated cluster is selected as the possible ‘active’ conformation against the 1RE1 active site. In the present study, 32 compounds were successfully docked into the 1RE1 site.

The X-ray crystal structure of caspase-3 taken from the Protein Data Bank (pdb: 1RE1) was used to dock. At the beginning of docking, all the water molecules were removed, hydrogen atoms added and AMBER7FF99 charges to the protein were applied.

It is critical to search for the binding pocket of the prepared protein in docking studies. In Surflex-docking, the binding pockets can be defined either from a cocrystallized ligand or from a list of residues known to be part of the interacting site (or predicted de novo). The active site was defined within 5 Å surrounding to the co-crystallized ligand and the specific residues and constraints information were obtained from crystallographic data as well as an earlier study. The docking poses were ranked by Gold docking and the top ten poses were selected. The ligands were then docked inside a cubic GRID box centered at the midpoint between the Cys205 and Gly238. Ten docking runs were performed for each compound in the dataset. In most cases the chosen pose was the top-ranked solution.

3.1 CoMFA results

3D-QSAR models were built by template based alignment method in SYBYL Molecular design Software (Fig. 1).

For the selected CoMFA model, the cross-validated r² (q²) value was 0.632 with four principal components. The non-cross-validated r² value was 0.817 with a standard error of estimate 0.152 and a covariance ratio (F) of 44.177 (significant at 99% level) (Table 2).

In CoMFA model, the steric parameter contributes 52.4%, while the electrostatic parameter accounts for 47.6%. Contributions of steric and electrostatic fields are shown in Fig. 3. The correlation between experimental and predicted activity for both training and test sets of compounds is shown in Table 3 and represented graphically in Fig. 4, respectively. These results authenticate the good prediction ability of the generated 3D QSAR model. The model summary dialog box, showed the relative positions of the local fields around aligned molecules that were important for activity variation in the model (Fig. 3). Greater values of “Bio-Activity Measurement” are correlated with more bulky near green, less bulky near yellow, more electro-negative near blue and less electro-negative or electro-positive charge near red. A large green counter near the 5th -position of the substituted furan ring (Ar₁ substitution) of 1,2,4-oxadiazole ring, indicates that the bulky substitution favors the activity. Thus the presence of a bulky substitution like bromine at 3rd position substitution (compound 10a) encourages the activity. Naturally, presence of chlorine atom at 3rd position in compound 10b is more bulky than unsubstituted in compound 10c thus compound 10b is more active than compound 10c. Activity may further be enhanced by replacing the chlorine atom with bromine (i.e., compound 10b and 10a, respectively) due to more bulkiness of bromine as compared to chlorine. Less bulky group substitution demonstrated with a yellow contour has no significant contribution to activity. Similarly, a blue contour near the 3rd position of the substituted oxadiazole ring (Ar₂), indicates that substitution with electronegative groups like CN, F, Cl increases the activity. Thus the presence of a trifluoromethyl group in compounds (1d, 4i, 4k and 10g) rendered them more active then compounds containing chlorine atoms (4a, 4h, 4l and 10f), respectively. Electropositive substitution demonstrated with a red contour has no contribution to activity.

Close

Figure 3

CoMFA contour plots for steric and electrostatic regions. Green counters indicate the bulky group region, whereas the yellow counter indicates the region where bulky groups are not required. Blue counters indicate the region needed for electronegative contribution, whereas red counters indicate the region where electropositive contribution is required.

Export to PPT

Close

Figure 4

Correlation between the experimental and predicted activities of the developed CoMFA model. (♦ Represent training data set, ● Represent test data set).

Export to PPT

3.2 CoMSIA results

For the selected CoMSIA model, the cross-validated r² (q²) value of the training set was 0.577 with the optimum number of components (N) six. The regression coefficient (r²) value was 0.822 with standard error of estimation (SEE) 0.150 and covariance ratio (F) of 38.623 (significant at 99% level) (Table 2). The correlation between experimental activity (EA) and predicted activity (PA) is shown in Fig. 5 and reported in Table 3. The CoMSIA model was externally validated with the test set of compounds. These results demonstrated that the obtained CoMSIA model has good self-consistency (r² > 0.8) and good prediction ability (q² > 0.5). The data clearly illustrate the steric, electrostatic, hydrogen-bond donor, hydrogen-bond acceptor and hydrophobic field interaction in the CoMSIA model. In the CoMSIA model, the contributions of the steric, electrostatic, hydrophobic, hydrogen-bond acceptor and hydrogen bond donor fields were 21.1%, 28.9%, 33.2%, 12.3%, and 4.5%, respectively (Table 2) and are represented graphically in Figs. 6 and 7. Similar to CoMFA the presence of a green contour near the fifth position substituted oxadiazole derivatives (Ar₁) indicated that the bulky groups at this position increased activity. The yellow contour above the third position of the substituted phenyl system (Ar₂) indicates that bulky substituents at this position decrease anticancer activity and small group substitution or unsubstitution adds to the biological activity. The blue contour near the third position of the substituted phenyl ring (Ar₂) indicates that the biological activity will increase by an electronegative group at the aforementioned positions (i.e., presence of CF₃ in compounds 1d, 4i, 4k and 10g) (Fig. 6). In the hydrophobic contour plots (Fig. 7) the white contour near 3rd and 5th positions of the substituted (Ar₂ and Ar₁, respectively) oxadiazole ring system suggests that hydrophobic groups like long chain alkyls at this position will add to activity, while a diminutive cyan color indicates no significant contribution of hydrophobic disfavor in biological activity. The purple counter near the 5th position of the substituted (Ar₂) system suggests that the hydrogen-bond donor favored activity and orange contours for hydrogen-bond donor not favored are absent showing no contribution in biological activity. Similarly contours for both hydrogen bond acceptors that were favored and disfavored (magenta contour and green contour respectively) showed no significant contribution in anticancer activity.

Close

Figure 5

Correlation between the experimental and predicted activities of the developed CoMSIA model. (♦ Represent training data set, ● Represent test data set).

Export to PPT

Close

Figure 6

CoMSIA contour plots for Steric and Electrostatic regions. Green counters indicate the bulky group region, whereas yellow counters indicate the region where less bulky groups are required. Blue counters indicate the region needed electronegative contribution.

Export to PPT

Close

Figure 7

CoMSIA contour plots for hydrophobic, hydrogen bond donor and hydrogen bond acceptor fields region. The white contour near suggest that hydrophobic groups will increase activity and cyan contour decrease the activity. The purple and orange contour for both hydrogen-bond donor favor and not favor respectively in biological activity are absent. Similarly contour for both hydrogen bond acceptor favor and disfavor are showed by magenta contour and green contour respectively.

Export to PPT

3.3 [(SW) kNN MFA] results

In the [(SW) kNN MFA] method the results of uni-column statistics are summarized in Table 4, which show that the test is interpolative. The final model was selected on the basis of statistical parameters of the models. Finally, the model with good internal and external predictive abilities was selected and is shown in Table 2.

For the selected [(SW) kNN MFA] model, the cross-validated r² (q²) value of the training set was 0.6278 with three principal components. The non-cross-validated r² value was 0.8413 with a standard error of estimate of 0.1852 and a covariance ratio (F) of 25.434 (significant at 99% level). The predictive ability of the model was also confirmed by external pred_r² having the value 0.5078 (Table 2).

In kNN MFA model, the steric parameter contributes 34%, while the electrostatic parameter accounts for 66%. Contributions of steric and electrostatic fields are shown in the contribution chart (Fig. 8). The correlations between experimental and predicted activity for both training and test set of compounds are shown in Table 3 and represented graphically in Fig. 9 respectively, authenticate the good prediction ability of the generated 3D QSAR model. The model summary dialog box, showed the relative positions of both steric (green counter) and electrostatic regions (blue counter) around aligned molecules that were important for activity variation in the model (Fig. 10). Like CoMFA and CoMSIA, the green counter near the 5th position of the substituted furan ring of 1,2,4-oxadiazole derivatives indicates that biological activity can be improved by introducing a bulky group. Presence of a blue counter near the 3rd and 5th positions of the substituted furan and phenyl rings respectively of 1,2,4-oxadiazole derivatives favors the presence of an electronegative group at aforementioned positions.

Close

Figure 8

Contribution of steric and electrostatic fields generated in 3D QSAR by [(SW) kNN MFA] model.

Export to PPT

Close

Figure 9

Correlation between the experimental and predicted activities of the developed 3D QSAR by [(SW) kNN MFA] model. (♦ Represent training data set, ● Represent test data set).

Export to PPT

Close

Figure 10

kNN MFA contour plots for steric and electrostatic regions. Green counters (S) indicate the steric region, whereas blue counters (E) indicate region where electrostatic contribution are required.

Export to PPT

Thus on the basis of 3D QSAR including CoMFA, CoMSIA and [(SW) kNN MFA] analyses the presence of bulky groups such as ethoxy and phenoxy at the 5th position of thesubstituted furan ring of 1,2,4-oxadiazole derivatives and substitution of electronegative groups including di-fluoro and trimethy fluoro at 3-substituted phenyl ring of 1,2,4-oxadiazole derivatives are conducive to caspase-3 activating activity.

3.4 Docking results

It is well-reported that the anticancer mechanism of 3-Aryl-5-aryl-1,2,4-oxadiazoles is due to caspase-3 activation. Therefore, a docking study could offer understanding the protein–activator interactions and the structural features of the active site of the protein.

All 1,2,4-oxadiazoles derivatives were docked into the binding site of caspase-3 and the energy scores of the activators are also shown in Table 3, where precise correlations could be found between docking scores and pIC₅₀ values.

A complete overview of Gold docking is presented in Figs. 11 and 12. Correlation between the docking score and experimental activity is shown graphically in Fig. 13. It vividly represents the interaction model of the most potent activator 10e with caspase-3. Docking studies showed that activator 10e is suitably situated at the binding site and there are various interactions between it and the binding region of the enzyme. The oxadiazole ring binds to the Caspase hinge region through three key hydrogen bond interactions: (1) between the NH of Cys205 and the O of the oxadiazole ring (2) between the NH of Gly238 and the O of the oxadiazole ring (3) between the NH of Cys205 and the N of the oxadiazole ring. Similarly the 5th position of the substituted furan ring also forms hydrogen bonding between the NH of Gly238 and the O of the furan ring. The hydrogen bonding distances observed were 2.319 Å (O_ _ _ H–NH– Cys205), 2.532 Å (O_ _ _ H–NH– Gly238), 2.558 Å (N_ _ _ H–NH– Cys205) and 1.557 Å between the O of the furan ring and NH of Gly238 (O_ _ _ H–NH– Gly238).

Close

Figure 11

Overlay of docked highest potent oxadiazole compound (10e) at the active site of 1RE1 produced using the Gold program.

Export to PPT

Close

Figure 12

Overlay of docked least potent oxadiazole compound (10c) at the active site of 1RE1 produced using the Gold program.

Export to PPT

Close

Figure 13

Correlation between the experimental activity and dock score in Gold docking. (♦ Represent training data set, ● Represent test data set).

Export to PPT

Fig. 12 shows the docking mode of the least active oxadiazole derivative compound 10c at the docking pocket. Similar to compound 10e, compound 10c was also docked at the same binding pockets having Cys205 and Gly238 amino acid residues. Results show that the O atom of the oxadiazole ring forms a conservative hydrogen bond with Cys205 residue (O_ _ _ H-NH-Cys205) having 2.414 Å bond distance. O atom of the furan ring substituted at the 5th position of oxadiazole (Ar₁) also forms hydrogen bonding with NH of Gly238 (O_ _ _ H-NH-Gly238) with 1.594 Å bond length. The docking results reported in Table 3, reveal that hydrogen bonding may be responsible for activity, which may be further increased on adding high electronegative substitutions.

Results of CoMFA, CoMSIA and [(SW) kNN MFA] methods, clearly show that the presence of an additional electronegative substitution at the 3rd position (Ar₂ substitution) of 1,2,4-oxadiazole enhances biological activity. The presence of electronegative substitution at the 3rd position is responsible for biological activity and related data were further confirmed by docking results. Docking results clearly reveal that the presence of an additional electronegative group at 3rd position substitution (10e) forms more hydrogen bonds with their surrounding amino acid residues (3 hydrogen bonds with Cys205 and Gly238) and thus possesses improved biological activity in comparison to compounds that possess less number of electronegative substitutions (10c) at the aformentioned positions. These data justify the results of 3D QSAR and docking results and also confirm the utility of this hybrid technique.