2.1 Data set

A set comprising of 63 compounds bearing cinnamic acyl 1,3,4-thiadiazole amides and sulfonamides was taken from the available literatures (Yang et al., 2012; Luo et al., 2011) and was used in the present study. The selected compounds for the data set shared the same assay procedure (Hamel, 2003; Hamel and Lin, 1984) with significant variations in their structures and potency profiles. Inhibitory potencies of the compounds included in data set, had IC₅₀ values varying from 3.16 to 505.76 μM which were converted into molar values. These were then converted into pIC₅₀ values using the formula given below. $${\text{pIC}}_{50}=-{\log }_{10}[{\text{IC}}_{50}]$$

The 3D structures of ligands were generated using the builder panel in Maestro and subsequently optimized using LigPrep module (v3.1, Schrödinger 2016-1). Partial atomic charges were ascribed and possible ionization states were generated at a pH of 7.0. The OPLS_2005 force field was used for optimization for production of low energy conformer of the ligand (Shivakumar et al., 2010). The energy minimization was performed for each ligand till it reached a root mean square deviation (RMSD) cutoff of 0.01 Å. The resulting structures were then taken for performing modeling studies.

2.2 Pharmacophore 3D-QSAR modeling

Phase (v4.6, Schrödinger 2016-1) was used for generation of pharmacophore and 3D-QSAR models for antitubulin agents (Dixon et al., 2006). The prepared ligands were imported for developing pharmacophore model panel of Phase with their respective biological activity values. The ligands were assigned as actives with a threshold of pIC₅₀ > 5.1 and inactives with a threshold of pIC₅₀ < 3.45. Remaining compounds were considered as moderately active. Phase (v4.0) has been proven to be an important tool for flexible ligand superposition (Sastry et al., 2011; Miller et al., 1999). In the current study, we used phase (v4.0) shape screening for flexible alignment of the selected antitubulin compounds. The most active compound 25 was taken as the template and pharmacophore type volume scoring was kept as default. The default settings were used with a maximum of 10 conformations per rotatable bonds and with this a maximum of 100 conformers were generated. Conformations of these compounds were varied and at most one alignment for every ligand was retained. For assigning training and test set for 63 compounds, random selection was made using the Automated Random Selection option present in Phase (v4.0) software. Test set comprised of 18 compounds and remaining 45 compounds were included in training set. Pharmacophore sites for these compounds included default set of chemical features of Phase: two hydrogen bond acceptors (A), aromatic ring (R) and a hydrophobe (H). Pharmacophore matching tolerance was set as 1 Å. A total of 6 variants were generated by keeping 5 and 4 as the maximum and minimum number of sites respectively. It was specified that at least 5 of the actives must match. Finally, it resulted in 10 possible combinations of features which could give rise to common pharmacophores. The resulting hypotheses were scored and ranked in accordance with their vector, volume, site scores, survival scores and survival actives (Dixon et al., 2006) shown in Table 3.

In the generated hypotheses, four sites were found to be common for all selected compounds. 3D-QSAR model was generated using Partial Least Square (PLS) regression statistics and keeping the grid spacing of 1 Å. The number of PLS factors included in model development is 6 because an incremental increase in the statistical significance and predictivity was observed till 6 PLS factor (Table 4).

The distance and angles between different sites of the model AAHR.11 are shown in Fig. 1a and 1b. This is also presented in Tables 1 and 2. Alignment of actives and inactives with the generated common pharmacophore is shown in Fig. 2a and b. Using the model, AAHR.11, fitness score of all the ligands was evaluated (Table 5). Contour plot analysis was also done to identify the distinct imperative pharmacophoric requirements at spatial sites of structure.

Close

Figure 1 Intersite (a) distances and (b) angles between the pharmacophoric points of model AAHR.11. All distances are in Å unit. Pink spheres with arrow, hydrogen bond acceptor (A); orange open circle, aromatic ring (R); green sphere, hydrophobe.

Export to PPT

Close

Figure 2 (a) Mapping of the active compounds onto the pharmacophore. (b) Mapping of inactive compounds onto the pharmacophore.

Export to PPT

Table 5 Calculated pIC₅₀ for compounds in predicted set.

Ligand	Chemical structure	Experimental activity (pIC50)	Predicted activity (pIC50)	Residual activity	Fitness
1t		4.75	4.89	−0.14	2.37
2		4.74	4.81	−0.07	2.66
3		4.83	4.77	0.06	2.73
4t		4.85	4.88	−0.03	2.73
5t		5.02	4.50	0.52	1.86
6		4.71	4.78	−0.07	2.40
7		4.70	4.75	−0.05	2.69
8		4.71	4.66	0.05	2.75
9		4.87	4.83	0.04	2.76
10		4.81	5.02	−0.21	2.95
11t		4.73	4.80	−0.07	2.40
12		4.66	4.71	−0.05	2.69
13		4.63	4.66	−0.03	2.75
14t		4.90	4.79	0.11	2.74
15		5.00	5.04	−0.04	2.95
16		4.82	4.88	−0.06	2.40
17t		4.82	4.86	−0.04	2.69
18t		4.87	4.76	0.11	2.75
19		4.90	4.92	−0.02	2.74
20		5.12	5.11	0.01	2.95
21		5.04	5.12	−0.08	2.45
22		5.06	5.08	−0.02	2.72
23		5.10	5.04	0.06	2.76
24		5.13	5.15	−0.02	2.77
25		5.50	5.15	0.35	3.00
26		3.90	3.86	0.04	1.64
27t		4.83	4.79	0.04	1.64
28		4.82	4.54	0.28	1.64
29		3.73	4.36	−0.63	1.64
30		4.83	4.78	0.05	1.86
31		3.90	4.19	−0.29	1.84
32		5.12	5.11	0.01	1.84
33		4.87	4.81	0.06	1.85
34t		4.41	4.59	−0.18	1.85
35		4.32	4.30	0.02	1.92
36		4.04	3.73	0.31	1.91
37		4.58	4.65	−0.07	1.91
38t		4.58	4.33	0.25	1.91
39t		4.56	4.19	0.37	1.91
40		3.69	3.92	−0.23	1.94
41		3.44	3.45	−0.01	1.93
42		5.00	4.44	0.56	1.92
43		3.56	4.12	−0.56	1.93
44t		3.92	3.98	−0.06	1.93
45		3.29	3.66	−0.37	1.89
46		3.35	3.25	0.10	1.87
47t		3.87	4.04	−0.17	1.87
48t		3.82	3.79	0.03	1.88
49		3.87	3.68	0.19	1.88
50		4.10	4.21	−0.11	1.88
51		3.96	3.73	0.23	1.86
52		4.50	4.56	−0.06	1.86
53t		4.08	3.98	0.10	1.87
54t		3.82	3.70	0.12	1.82
55		3.35	3.52	−0.17	1.67
56		4.74	4.56	0.18	1.65
57		3.43	3.49	−0.06	2.12
58		3.40	3.39	0.01	0.72
59		3.65	3.83	−0.18	1.94
60		3.53	4.02	−0.49	1.04
61t		3.70	3.34	0.36	1.57
62		3.67	3.53	0.14	1.87
63t		3.48	3.57	−0.09	0.72

^t: test.

2.3 Model validation

The QSAR model AAHR.11 with 6 component PLS factor was characterized as the best model. Pharmacophoric model was validated by its accuracy in estimation of the activity of training set ligands. Scatter plots for experimental and predicted activities of ligands elicited significant linear correlation and moderate difference between the experimental and predicted values shown in Fig. 3a and b. The efficacy of model AAHR.11 was also examined with external validation. Biological test used for compounds of both the “internal” training and test sets was the same as used for compounds of the “external test set”. In order to clearly distinguish between the test set and external test set, it is mandatory to mention that compounds of training and test share structural similarity while compounds of external test set are from a different publication. This external validation was done to check further reliability of the model (Table 1: Supplementary Material and Fig. 4). Graphs of actual value vs. predicted value and residual value vs. predicted value were plotted.

Close

Figure 3 (a): Scatter plot of the observed versus phase-predicted activity for (a) training set and (b) test set compounds with best fit line y = 0.84x + 0.66 (R² = 0.83).

Export to PPT

Close

Figure 4 (a) Plot of actual value vs. predicted value of external training set y = 0.84x + 0.66 (R² = 0.83). (b) Plot of residual value vs predicted value.

Export to PPT

The predictive power of the produced 3D-QSAR model was also assessed by using Enalos Model Acceptability Criteria KNIME node (Melagraki and Afantitis, 2013), which includes all proposed tests by Tropsha (Zhang et al., 2006). Criteria followed in developing activity/property predictors, especially for continuous QSAR, are as follows: (i) correlation coefficient R between the predicted and observed activities; (ii) coefficients of determination (predicted versus observed activities R²₀ and observed versus predicted activities R′₀² for regressions through the origin); and (iii) slopes k and k′ of regression lines through the origin. The criteria for the acceptability of the QSAR model are represented in left column of Fig. 5.

Close

Figure 5 Screenshot of results obtained by (a) Enalos Model Acceptability Criteria KNIME node; (b) Enalos Domain-Leverage node.

Export to PPT

Further validation of the model was done by enrichment studies using the decoy test. 1000 decoy test set compounds (Kirchmair et al., 2008) retrieved from the Schrödinger data base were taken to evaluate predictive power of the model. Enrichment factor (EF) and Robust initial enhancement (RIE) were calculated to ensure the reliability of the model and for the accurate ranking of compounds (Sheridan et al., 2001).

2.4 Domain of applicability

In order to screen a QSAR model for new compounds, its domain of application (APD) (Tropsha et al., 2003; Shen et al., 2004) must be defined and predictions for only those compounds that fall into this domain may be considered reliable. Extent of Extrapolation (Tropsha et al., 2003) is one of the simple approaches based on the calculation of the leverage (h_i) for each chemical, where the QSPR model is used to predict its activity. $$h_i=x_i{\left(X^{\text{T}}X\right)}^{-1}{x}_i^{\text{T}}$$

In the above equation, x_i is the row vector containing the k model parameters of the query compound and X is the n × k matrix containing the k model parameters for each one of the n training compounds. A leverage value greater than 3k/n is considered large. It means that the predicted response is the result of a substantial extrapolation of the model and may not be reliable.

2.5 Y-Randomization test

Y-Randomization technique ensures the robustness of a QSPR model (Tropsha et al., 2003; Shen et al., 2004). The dependent variable vector is randomly shuffled and a new QSPR model is generated. The procedure is repeated several times and the new QSPR models are expected to have low R² and Q² values. In case of opposite results, an acceptable QSPR model cannot be obtained.

2.6 Virtual screening

PubChem database serves as an important source of compounds which could be prioritized for screening for a wide range of targets. We decided to use PubChem database within our in silico framework for virtual screening compounds which could act as tubulin inhibitors. For this purpose, we developed a KNIME workflow to assess the compounds that will be prioritized for virtual screening. Our first aim was to mine all compounds having structural similarity with our set of compounds within PubChem database. To achieve that, we have used Substructure Matcher node included in Indigo cheminformatics toolkit. All these compounds were compared to the most active compound of our original data set in terms of similarity given by Indigo Fingerprints. It is known that similarity searching refers to the calculation of the similarity coefficients for a given molecule and each compound included within a database. Among the different proposed methodologies, similarity searching using fingerprints is one of the most widely applied methods for ligand-based screening. Fingerprints are representations of the structural features of a molecule given as a series of binary digits (bits) that account for the presence or absence of particular substructures in the molecule. For this study, Fingerprint Similarity node using Tanimoto metric for similarity computation was used to afford the most similar compounds that could be prioritized for screening.

All the retrieved compounds were virtually screened with the ultimate goal of prioritizing the most promising new synthetic targets. The proposed 3D-QSAR model was used to identify the most active compounds that are proposed for further investigation. For this, find matches option of Phase module was used.

2.7 Molecular docking

The catalytic domain of tubulin enzyme in complex with colchicine (PDB Code: 1SA0) was obtained from protein data bank and prepared using the protein preparation wizard (Sastry et al., 2013) available in Schrödinger suite 2016-1. Crystallographic water molecules i.e. without 3H bonds were deleted and hydrogen bonds corresponding to pH 7 were added considering the appropriate ionization states for both acidic and basic amino acid residues. OPLS_2005 force field was used for energy minimization of the crystal structure (Shivakumar et al., 2010). Active site was defined with a radius of 14 Å around the ligand present in crystal structure. Grid box was generated at a centroid of active site. For docking, low energy conformations of all the compounds were docked into the catalytic domain of tubulin protein (PDB Code: 1SA0) using Grid based Ligand Docking with Energetics (Glide v7.0, Schrödinger 2016-1) (Friesner et al., 2006) in extra precision mode without applying any constraint. The best docked structure was identified using Glide score function, Glide energy and Glide Emodel energy (Table 8: Supplementary Material). The lowest energy docked compound 25 was selected for further studies. A map of hydrophobic and hydrophilic fields for inhibitor 25 was also generated (Fig. 8).

2.8 Lipinski’s rule for drug likeliness and in silico ADME prediction

We further predicted the drug-like behavior of the compounds through the analysis of pharmacokinetic profile of the compounds by using Qikprop module (v4.3, Schrödinger 2016-1). The compounds prepared by LigPrep module (v3.1, Schrödinger 2016-1) were utilized for the calculation of pharmacokinetic parameters by QikProp v4.3. Physically significant descriptors and pharmaceutically relevant properties of all the test compounds such as molecular weight, log P, H-bond donors, and H-bond acceptors were analyzed in accordance with Lipinski’s rule of five (Lipinski et al., 2001).

3.1 Pharmacophore and 3D-QSAR models

Phase (v4.0) (Dixon et al., 2006) module of Schrödinger 2016-1 was used for the development of pharmacophore model. Atom based 3D-QSAR revealed the effect of substituents on activity. Based on the sites, a maximum of four features were allowed for developing the hypotheses. For this, 6 common hypotheses were generated in all for 63 compounds (Table 3). The best fitted model AAHR.11 (R² = 0.8925, Q² = 0.8204 and F = 44.3) consists of two hydrogen bond acceptor, one hydrophobic and aromatic ring features (Fig. 1a and b) with highest survival score (3.640). This is also apparent from the comparison made by the deduction of survival inactive from survival active. It deducts inactive features from the hypothesis and was decisively highest for the hypothesis AAHR.11. Among the pharmacophore features, intersite distances and angles between site points were found to be the principal attribute and the point of difference between active and inactive is due to the interstitial site distances as evident by the pharmacophore hypothesis AAHR.11 alignment over active (pIC₅₀ > 5.1) and inactive compounds (pIC₅₀ < 3.45) in Fig. 2a and b respectively.

3.2 Model validation

The predictability and validity of the common pharmacophore model, AAHR.11 (test set), based on active compounds were judged by cross validation coefficient (Q² = 0.8204) (Table 4). Regression coefficient (R²) of the training set was 0.89, which exhibited relevance of the model. Stability of the generated model ranges from 0.7355 to 0.9890 on the maximum scale of 1. The F value was found to be 44.3. Additionally, P value of 3.959e-014 and Pearson r of 0.9124 indicated greater degree of confidence on the model. Standard deviation (SD) value of 0.2134 and root mean square error (RMSE) of 0.1962 indicate the stability of the generated model for estimation of unknown compounds in the test. In order to evaluate the efficacy of the generated model, it was further validated using an external test set (Li et al., 2014). The calculated pIC₅₀ values of the compounds included in predicted set are given in Table 5. Predicted values for external test set are given in Table 1 (Supplementary Material). A plot of experimental vs. predicted pIC₅₀ of external test set is shown in Fig. 4a and a plot of residual vs. predicted value is shown in Fig. 4b. These two plots are vital for the predictive ability of QSAR. Scatter residual plots were used to identify the outliers from the QSAR model (Golbraikh and Tropsha, 2002a,b). Fig. 4b elicits that there is no outlier in the study. The presence or absence of outliers can be easily determined from the scatter plot by comparing the number of compounds taken in the test set and number of dots present in the plot. In this case, both the numbers were same, thereby rendering the model stable and reliable. It was able to support the experimental pIC₅₀ values for compounds in the external test set. Predictive correlation coefficient R² value of 0.83 was observed for the external test set for the developed QSAR model. R² value of more than 0.5 between the predicted and experimental values renders the model to be good and was able to predict the inhibitory activity of compounds not included in the model development process (Golbraikh and Tropsha, 2002a,b).

Furthermore, the Enalos Model Acceptability Criteria KNIME node has also been applied to the data. The model passed Tropsha’s recommended tests for predictive ability and the results are depicted in Fig. 5a. These results suggest that this alignment can effectively take into consideration the ligand-receptor interactions, and the QSAR model is thus reliable and could be used in the design of new tubulin inhibitors within this structural motif of molecules.

In order to validate the discriminatory ability, the model was screened against 1000 decoy molecules retrieved from the Schrödinger database. The model was able to find 100% of active compounds in the hit list. RIE was calculated for the generated models to estimate the contribution of the active compounds ranking in the enrichment. For AAHR.11, RIE value of 8.24 indicated the superiority of the pharmacophore model ranking over random distribution. AUC of the ROC curve is also considered as a reliable factor to assess the performance of the developed model. AAHR.11 showed a good value of 0.95 AUC and 0.99 ROC (Fig. 1 and Tables 2–5 of Supplementary Data).

3.3 Domain of applicability

It is important that the limitations of the model are also described via the APD. This gives an important indication as the user can freely and creatively design novel molecules but will be warned for the reliability of the estimation when the structural characteristics cannot be tolerated by the model. After model validation, the APD of our model was also defined to ascertain that a given estimation can be considered reliable. Enalos Domains – Leverage (Extent of Extrapolation) node that executes the aforementioned procedure is included in our workflow and was used to assess APD of the proposed model. In this method, the APD limit was defined as 0.167. This limit was defined on the basis of equation provided in Methods section. In the APD, all the compounds fell within the range i.e. none of the compounds was in the group of outliers (Fig. 5b). All these compounds were in the domain of the applicability, and hence can be considered reliable.

3.4 Y-Randomization test

The model was further validated by applying Y-randomization. Ten random shuffles of the Y vector were performed and the low R² and Q² values were obtained. This showed that the good results in the original model were not due to a chance correlation or structural dependency of the training set. The R² and Q² values were in the range of 0.05 to 0.345 and 0.00 to 0.11 respectively. It should be noted that for each random permutation of the Y vector, the complete training procedure was followed for developing the new QSPR model, including the selection of the most appropriate descriptors.

3.5 3D-QSAR contour map analysis

Contour plot analysis was performed to identify the effect of spatial arrangement of structural features such as electrostatic, ionic, hydrophobic, H-bond donor and H-bond acceptor regions on antitubulin activity. Their individual positive contribution is shown in blue cubes and negative contribution is indicated by red cubes. Fig. 6a–f shows the comparison of most significant favorable and unfavorable interactions, which arise on the application of QSAR model to the most active and least active ligands. Results of our study were found to be in concordance with those reported by the researchers who had synthesized and biologically evaluated these compounds (Yang et al., 2012; Luo et al., 2011).

Close

Figure 6 QSAR model visualized in the context of favorable and unfavorable hydrogen bond donor effects in (a) compound 25 (5.50) and (b) compound 45 (3.29). QSAR model visualized in the context of favorable and unfavorable hydrophobic interactions in (c) compound 25 (5.50) and (d) compound 45 (3.29). QSAR model visualized in the context of favorable and unfavorable electron withdrawing groups in (e) compound 25 (5.50) and (f) compound 45 (3.29).

Export to PPT

The hydrogen-bond donor nature for the most active compound 25 (Fig. 6a) and the least active compound 45 (Fig. 6b) when compared showed their most favorable region with blue color and unfavorable regions with red color. Hydrogen-bond donor mapping revealed that favorable regions lied near the nitrogen atoms of thiadiazole, indicating their importance for activity as compared to the least active compound 45 bearing sulfone group. Therefore, the presence of thiadiazole ring with hydrogen donor group in the scaffold backbone is vital for the tubulin inhibitory activity. Almost all the thiadiazole containing compounds were found to have better activity profile in comparison with the ones with sulfone moiety.

Another significant component that impacts the antitubulin activity is the hydrophobic character. Fig. 6c and d when compared for their hydrophobic nature for the most active compound 25 and least active compound 45 reveals that the blue region, an indicative of hydrophobic character, around the thiadiazole ring is essential for a compound to be an antitubulin agent. In Fig. 6d, the presence of red cubes at para position of phenyl ring directly attached to the sulfone group indicates that hydrophobic groups are unfavourable at this position. This assumption is supported by the low activity of methyl substituted compounds when compared to their unsubstituted derivatives. This is evident while comparing the compounds 31 with 30, 36 with 35, 41 with 40, 51 with 50, 55 with 54 and 60 with 59.

In contour plot of compound 25 (Fig. 6e), the presence of red cubes at para position of phenyl ring attached directly to the thiadiazole moiety indicates that the presence of electron withdrawing groups is undesirable at this position. This is supported by less activity of compound 6 (pIC₅₀ 4.71) having Cl group at para position. A comparison of the para-substituents on the same ring depicted that an electron-donating group (16, 21) (pIC₅₀ 4.82, 5.04) have slightly improved antitubulin activity with the potency order OCH₃ > CH₃. On the contrary, the presence of blue cubes at para position of cinnamoyl moiety indicated the preference of electron withdrawing groups at this position. This is explained by the significant antitubulin activity of compounds with para halogen substitution (30, 35, 40) (pIC₅₀ 4.83, 4.32, 3.69) in the order of F > Cl > Br. The results demonstrated that an electron-withdrawing group may have slightly improved antitubulin activity. From the values themselves, it is observed that the compounds substituted with electron donating groups (compound 45: pIC₅₀ 3.29) had relatively lower pIC₅₀ values in comparison with the compounds with electron withdrawing substituents and to the unsubstituted compound 26 (pIC₅₀ 3.90).

3.6 Virtual screening

As described in the Materials and Methods section, the produced QSAR model can be used as a useful tool for screening existing databases or virtual libraries in an effort to prioritize potent compounds for experimental evaluation. In this context, a virtual screening of structurally similar compounds included in PubChem database was carried out using a KNIME workflow.

First, 53,691 compounds were retrieved from the PubChem database. Lipinski’s rule of five was used to scrutinize these compounds. A total of 47,615 compounds were obtained after this. These compounds were compared to the active compound within our original database, compound 25 using fingerprints based on the Tanimoto similarity metric as described in Materials and Methods section. The 2542 compounds with a Tanimoto similarity metric of over 0.80 were selected to assess the potency of these compounds. Our approach that is based on the scaffold and fingerprint similarity search within PubChem database resulted in a narrower chemical space that increases the chance of success.

The pIC₅₀ values of the 13 most potent compounds (Vs1–Vs13) predicted with the established pharmacophore model are listed in Table 6 of Supplementary Material. All the virtual screening compounds were within the domain of the applicability of the model, and therefore their activity estimations can be accepted with confidence. The predicted compounds were searched for their activity using the PubChem database and their biological data have been reported in Table 6 (Supplementary Material).

On the whole, obtained hits can serve as new chemical starting points for further structural optimization of tubulin inhibitors.

3.7 Molecular docking

The scores of docking studies are shown in Table 7 (Supplementary Material). From 63 listed total-scores, pIC₅₀ of most compounds were in accordance with the Glide score. Docking study revealed that interactions were dominated by the hydrophobicity and aromaticity due to the presence of cinnamoyl moiety. The interactions were dominated in the region of Tyr 202, Ala 316, Ala 250 and Val 239 amino acid residues content due to pronouncing existence of active site in the region (Fig. 7a and b). In general, phenyl ring directly attached to thiadiazole ring exhibited hydrophobic interactions with Lys 254, Leu 242, Ala 250, Leu 248, Ala 354, Val 318, Lys 352, Ile 378 and Leu 255, the key residues necessary for binding of inhibitors (Fig. 7c). This is also evident by analyzing the generated map of hydrophobic and hydrophilic fields for inhibitor 25 (Fig. 8), where phenyl ring attached to thiadiazole and cinnamoyl moiety is buried in the hydrophobic pocket (orange color) while the ethylenic part is located in the hydrophilic pocket (cyan color). This is in correlation with the 3D-QSAR result where hydrophobic R11 pharmacophoric feature (cinnamoyl moiety) and thiadiazole ring of 25 are found to be preferable for the activity (Fig. 7c). Our Glide XP-docking result also revealed solvent exposure in the region of thiadiazole and cinnamoyl moiety plays an important role in stabilization of inhibitor at the active site. The compound also showed polar interactions with amino acid residues such as Asn 350, Thr 314, Asn 258, Asn 349, Thr 240 and Thr 239.

Close

Figure 7 (a) Binding mode of compound 25 in the catalytic pocket of 1SA0, (b) 2D-ligand interaction diagram of 25 in the catalytic pocket of 1SA0, (c) Hydrophobic interactions of 25 in the catalytic pocket of 1SA0, (d) Overlay of docked pose (magenta) of colchicine with its crystal structure conformation (pink).

Export to PPT

Close

Figure 8 Map of hydrophobic and hydrophilic fields for inhibitor 25 into the catalytic protein of tubulin (1SA0).

Export to PPT

Binding mode and 2D ligand interaction diagram of least active compound of the series (45) are shown in Fig. 2 of Supplementary Material. This compound showed polar interactions with Thr 314, Asn 350, Asn 349, Asn 258 and Thr 353. aa residues Met 259, Ile 378, Val 315, Leu 248, Leu 255 and some others were responsible for formation of hydrophobic bonds with the compound. The only positively charged residue responsible for interactions was Lys 352. On the whole, extent of interactions for least active compound (45) was much less in comparison with that exhibited by the most active compound (25).

In addition, the accuracy of the docking procedure was determined by examining how closely the lowest energy poses (binding conformation) predicted by the object scoring function, Glide score (GScore), which resembles an experimental binding mode as determined by X-ray crystallography. The root mean square deviation (RMSD) between the predicted conformation and the observed X-ray crystallographic conformation of colchicines (PDB Code: 1SA0) was found to be 0.853 Å (Fig. 7d), a value that suggests the reliability of Glide XP docking mode in reproducing the experimentally observed binding mode of tubulin polymerization inhibitors and the parameter set for Glide XP docking is reasonable to reproduce the X-ray structure. Further, similar orientation was observed between the superposition of conformation 25 best XP-docking pose and 3D-QSAR pose (RMSD: 0.870 Å) (Fig. 9).

Close

Figure 9 Superimposition of conformations of inhibitor 25: best docking pose and pose of the AAHR.11 model (RMSD: 0.870 Å).

Export to PPT

3.8 Lipinski’s rule for drug likeliness and in silico ADME prediction

Different pharmacokinetic parameters of the compounds taken for the study were subjected to ADME predictions by Qikprop v4.3. The compounds were assessed for their basic parameters of Lipinski’s rule of 5 and other pharmacokinetic parameters. Table 8 (Supplementary Material) shows the results obtained from Qikprop with their permissible range. In general, an orally active compound should not have more than 2 violations of the Lipinski rule. The active test compounds in present study were not found violating the rule more than the maximum permissible limits and thus proving their drug likeness properties.

The optimum values of the descriptors, polar surface area and rotatable bonds also have great influence on oral bioavailability of the drug molecules. The important parameters with their permissible ranges are delineated in Table 8 (Supplementary Material). The optimum value of rotatable bonds (0–15) and polar surface area (7–200 Å) holds a great importance on the oral bioavailability of the drug molecules. The active test acyl 1,3,4-thiadizole amide and sulfonamides derivatives demonstrated results of the descriptors, which were found to be within the prescribed range, thus owing to good bioavailability. Intestinal absorption or permeation is also one of the important factors to be studied in concern with the absorption of the drug molecule, which was further confirmed by predicted Caco-2 cell permeability (QPPCaco), used as model for gut-blood barrier. Caco-2 cell permeability prediction of the test compounds indicates excellent results, predicting good intestinal absorption. Further, the test results for QPlogkhsa descriptor of Qikprop indicating the predicted values of human serum albumin binding indicated that test molecules were found to fall within the permissible range (−1.5 to 1.5). Also, the Qikprop descriptor for blood/brain partition coefficient QPlogBB showed reliable prediction for all the test compounds and reference drugs. The cell permeability of the blood brain barrier that mimics MDCK cells (QPPMDCK) also displayed reliable results as they were well within the prescribed range. The aqueous solubility parameter (QPlog S) of the test entities was assessed and the compounds were also found to be in the permissible range (<0.5).