- Email: [email protected]
Pharmacophore based virtual screening, molecular docking and biological evaluation to identify novel PDE5 inhibitors with vasodilatory activity
Accepted Manuscript Pharmacophore Based Virtual Screening, Molecular Docking and Biological Evaluation to Identify Novel PDE5 Inhibitors with Vasodilatory Activity Anupama Mittal, Sarvesh Paliwal, Mukta Sharma, Aarti Singh, Swapnil Sharma, Divya Yadav PII: DOI: Reference:
S0960-894X(14)00491-0 http://dx.doi.org/10.1016/j.bmcl.2014.05.004 BMCL 21617
To appear in:
Bioorganic & Medicinal Chemistry Letters
Received Date: Revised Date: Accepted Date:
8 February 2014 30 April 2014 2 May 2014
Please cite this article as: Mittal, A., Paliwal, S., Sharma, M., Singh, A., Sharma, S., Yadav, D., Pharmacophore Based Virtual Screening, Molecular Docking and Biological Evaluation to Identify Novel PDE5 Inhibitors with Vasodilatory Activity, Bioorganic & Medicinal Chemistry Letters (2014), doi: http://dx.doi.org/10.1016/j.bmcl. 2014.05.004
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Pharmacophore Based Virtual Screening, Molecular Docking and Biological Evaluation to Identify Novel PDE5 Inhibitors with Vasodilatory Activity Anupama Mittal, Sarvesh Paliwal,* Mukta Sharma, Aarti Singh, Swapnil Sharma, Divya Yadav Department of Pharmacy, Banasthali University, Banasthali, Rajasthan, India
*Corresponding Author‘s mailing address and email ID: Prof. Sarvesh Paliwal Department of Pharmacy Banasthali University, P.O. Banasthali Rajasthan-304022 Phone no. +91-9352141469 Email I.D. [email protected] Email addresses of all authors: AM: [email protected] SP: [email protected] MS: [email protected] AS: [email protected] SS: [email protected] DY: [email protected]
ABSTRACT: Prompted by the role of PDE5 and its closely associated cAMP and cGMP in hypertension, we have attempted to discover novel PDE5 inhibitors through ligand based virtual screening. Rigorously validated model comprising of one HBA, one HY and one RA was used as a query to search the NCI database leading to retrieval of many compounds which were screened on the basis of estimated activity, fit value and Lipinski’s violation. Selected compounds were subjected to docking studies which resulted into visualization of potential interaction capabilities of NCI compounds in line to pharmacophoric features. Finally three compounds were subjected to in-vitro evaluation using the isolated rat aortic model. The results showed that all three compounds are potent and novel PDE5 inhibitors with vasodilatory activity range from 10-2 to 10-5M. Keywords: Phosphodiesterase-5, cyclic guanosine monophosphate, discovery studio, hydrogen bond acceptor, cyclic adenosine monophosphate A number of endogenous substances are implicated in vasorelaxation among which cyclic adenosine monophosphate (cAMP) and cyclic guanosine monophosphate (cGMP) are major ones.1 Phosphodiesterase enzyme (PDE) hydrolyze the cAMP and cGMP to inactive metabolite 5’AMP and 5’GMP respectively and play a very crucial role in regulating the concentration level of these secondary messengers.2 The capacity of PDEs to degrade cAMP and cGMP is tenfold greater as compared to the rate of formation of secondary messenger, hence modification in the activity of PDE enzyme has a great impact on physiological activities. In the human genome 21 genes encoding PDEs have been identified.3 These 21 genes are assembled into 11 PDE (PDE1-PDE11) families on the basis of amino acid sequences, structures, enzyme kinetics, mode of action and tissue distributions.4 Out of the 11 families of phosphodiesterases, PDE5 is a cGMP specific enzyme which contains two catalytic domains GAF-A and GAF-B. cGMP bind to the GAF-A domain which facilitates the phosphorylation of N-terminal with the help of protein kinase-A and protein kinase-G
thus stimulates the hydrolysis of cGMP.5 PDE5 is located in many organs such as lung, heart, cerebellum, kidney, and smooth muscle.6 The vasodilation of vascular smooth muscle in hypertensive patients can be enhanced by blocking the hydrolysis of cGMP via PDE-5 isoenzyme. FDA in 1998 approved three PDE-5 inhibitors namely sildenafil, vardenafil, and tadalafil for the treatment of erectile dysfunction.7 Later it was observed that PDE5 inhibitor also have the ability to reduce pulmonary arterial resistance and pressure which may be helpful for the treatment of pulmonary hypertension but it is noticeable that out of three PDE5 inhibitors sildenafil and tadalfil are approved for their therapeutic use in pulmonary hypertension.8 In view of potential role of PDE5 inhibitors in hypertension we envisaged that identification of novel PDE5 inhibitors with diverse structures, druggability and high potency can enrich this class of antihypertensive drugs. The utility of molecular modeling and virtual screening in lead identification and optimization are unquestionable.9-10 Prompted by the successful application of in-silico pharmacophore based virtual screening in lead identification, we have made an effort to implement in-silico protocols in association with wet lab experimentation to identify novel PDE5 inhibitors. As a starting point a data set of 48 PDE-5 inhibitors with wide range of activity (4nM to 26,000 nM) and structural diversity (Supplementary data Table S1) were used for development of pharmacophore model.11 The Hypogen module of Accelrys Discovery studio v2.0 (DS) was used to generate the pharmacophore models. Energy of all the compounds was minimized using CHARMm force field with an energy gradient of 0.001 kcal/mol/A°.12 Total 34 compounds were used as training set while the rest of 14 compounds were used as internal test set to validate the pharmacophore model. A maximum of 255 diverse conformers were generated for each molecule using best flexible conformation generation module which performs a rigorous energy minimization in both torsional and cartesia to attain the best
coverage of conformational space.13 Features such as hydrogen bond acceptor (HBA), hydrogen bond donor (HBD), hydrophobic (HY), positive ionizable (PI) and ring aromatic (RA) were used to generate various hypotheses keeping minimum and maximum features value between 0 and 5. Two compounds (66, 74) showed a greater error ratio and removal of these compounds improved the quality of model hence they were removed during the course of pharmacophore modeling. The quality and integrity of all generated hypotheses were evaluated on the basis of cost function analysis, RMS, and correlation coefficient (Table 1). The chosen set of hypothesis exhibited a fixed cost value of 118.63 bits and as recommended it was well separated from the null cost (178.27 bits). Out of 10 hypotheses hypo1 showed maximum difference of 45 bits between total cost and the null cost representing 75-90% chances of true correlation. For a significant model the value of configuration14 should be less than 17 and the hypo1 showed a value of 10.07 with RMS value of 0.95 and squared correlation coefficient value of 0.79 (Figure 1) confirming the significance of the model.
Table1: Results of the top 10 pharmacophore hypotheses generated by the Hypogen. Hypo
Total cost
Cost difference
RMSD
R2
Feature
1
133.23
45.03
0.95
0.79
1HBA,1HY,1RA
2
146.10
32.16
1.29
0.60
1HBA,1HY,1RA
3
160.48
17.79
1.61
0.40
1HBA,1HY,1RA
4
166.74
11.52
1.72
0.32
1HBA,1HY,1RA
5
166.82
11.45
1.73
0.31
1HBA,1HY,1RA
6
168.42
9.85
1.75
0.29
1HBA,1HY,1RA
7
169.10
9.16
1.77
0.29
1HBA,1HY,1RA
8
170.25
8.02
1.79
0.27
1HBA,1HY,1RA
9
170.98
7.28
1.80
0.25
1HBA,1HY,1RA
10
171.09
7.17
1.80
0.25
1HBA,1HY,1RA
Figure 1. Plot of predicted versus the corresponding actual activity (IC50) for the training and test compounds. Based on the above results Hypo 1 comprising of one HBD, one RA, and one HY features was identified as statistically significant hypothesis and was subjected to CatScramble validation which generates pharmacophore hypotheses by randomizing the activity data of the training set compounds by using the formula,15 significance = 100 (1-(1+x/y)), where x signifies the total number of hypotheses with a total cost value lower than the original hypothesis, and y symbolize the total number of HypoGen runs i.e. initial and random runs. During the validation process random spreadsheets were generated using the same training set compounds. A total of 19 random spreadsheets16 (random hypotheses) were generated to achieve a confidence level of 95%. The results of the randomized data were analyzed to evaluate the quality of the model and it was observed that the statistics of Hypo 1 were far better as compared to the 19 random hypotheses (Supplementary data Table S2).The results of CatScramble clearly shows that the generation of Hypo 1 was not by chance rather it has been an outcome of true correlation between structures of training set compounds and PDE 5 inhibitory activity. After initial round of validation, mapping of training set compounds was carried out and it was observed that most active compound (50~IC50 4nM) of training set
mapped well over the hypo1 with a fit value of 5.56, the HY feature (light blue color) mapped over benzene ring containing methyl group, the HBA (green color) mapped over the ketonic group of pyrimidine ring and ring aromatic (orange color) group mapped over the imidazole ring (Figure 2A). On the other hand least active compound (47~IC5026000 nM) missed one hydrophobic feature when mapped over Hypo 1 (Figure 2B) and also showed a poor fit value of 2.507.
Figure 2. Best pharmacophore model (Hypo1) of PDE-5 inhibitors generated by hypogen module. (A) Pharmacophore mapping aligned with the most active training set compound 50. (B) Pharmacophore mapping of the least active training set compound 47. The utility of any pharmacophore model lies in its ability to correctly predict the activity of compounds within the domain and outside the domain. The internal test set comprising of 14 compounds was mapped onto hypo-1 and a good correlation was observed as depicted in Figure 1. Roy et al. have recently developed a new parameter, modified r² (rm²), which considers the actual difference between the observed and predicted response data, thereby serving as a more stringent measure for assessment of model predictivity compared to the traditional validation parameters.17-18 In the present study, we performed rm2 metric test on training and test sets in which the value of "average rm2" was found to be 0.70 and 0.50 and the value of "delta rm2" was found to be 0.16 and 0.05 for training and test sets respectively. Ideally the values of “Average rm2” must be greater than 0.5 and values of “Delta rm2" must be lower than 0.2.
As a final step towards the assurance of applicability, predictivity and soundness of the model an external dataset of 12 structurally diverse compounds comprising of 1-arylnaphthalene lignan derivative,19 sildenafil, vardenafil and tadalfil were used as an external test set. The pharmacophore mapping pattern of external test set compounds were analyzed and it was observed that the entire test set compounds mapped very well over the hypo 1 with good fit values ranging within 3.727 to 5.47. Most importantly all the three marketed drugs and the most active lignan derivative exhibited a perfect three-feature mapping with good fit values as shown in Figure 3A,3B,3C and 3D. The r2 value of 0.68 (Fig. 3E) between predicted and actual activity for external test set compounds testified the universality of the hypo-1.
Figure 3: Pharmacophoric feature mapping of external and marketed drug. (A) Three pharmacophoric feature mapping of most active lignan derivative with a fit value of 5.47. (B) Three pharmacophoric feature mapping of sildenafil with a fit value of 5.55 (C) Three pharmacophoric feature mapping of tadalfil with a fit value of 5.86 (D) Three Pharmacophoric feature mapping of vardenafil with a fit value of 4.48. (E) Plot of experimental values versus the corresponding actual values (IC50) for the external test compounds.
Needless to mention that utility of validated pharmacophore model lies in virtual screening and since the developed pharmacophore model showed all the signs of its soundness and universality, it was used to screen National Cancer Institute database comprising of 260,071 structurally diverse compounds20 using best flexible database search option. The retrieved hits were screened on the basis of their fit and estimated value which led to retention of 25 out of 300 hits. Lipinski’s rule of five was applied to check their druggable properties which further reduced the list of hits to 6 with fit value ranging within 5.6 -5.9 as shown in Table 2.
Table 2: Hits retrieved from NCI database as potent PDE5 inhibitors Name of Hits
NSC 960
Estimated value IC50 (nM) 15.467
Fit value
Structure
O
5.621
O N
NSC 22
14.485
5.649
H2N
N
N O OH O
NSC 1015
13.314
5.686 N OH HN O O
NSC 412
11.444
5.751 HO
N N
HO
NSC 619
9.128
OCH3
5.85 H3CO
O
NSC 1012
7.323
5.945 N OH HN O OH
Since out of six lead compounds only three compounds were available with NCI for experimental validation, all the three compounds were subjected to docking to evaluate the binding interaction between different amino acids of target enzyme and lead compounds. A high resolution crystal structure of PDE-5 bound with an inhibitory molecule, sildenafil citrate (PDB code1UDT) was selected for molecular docking studies employing LibDock module available within DS. The crystal structure was checked for valency, missing hydrogen and any structural disorders like connectivity and bond orders. All the water molecules were removed from the protein hierarchy and only A chain was retained which was split into the protein and crystal ligand part.21 In the protein structure hydrogen atoms
were added and CHARMm forcefield was applied for energy minimization. Binding site present in the A chain was defined within the radius of 9 Å. All the three hits (NSC22, NSC412, NSC619) were docked into the crystal structure of PDE-5 enzyme and various poses were examined on the basis of LibDocker score (protein-ligand interaction energy) to evaluate the nature and type of interactions. It was observed that NSC22 docked (Figure 4A) very well with a LibDock score of 70.071 and the hydroxyl group of naphthyridinol moiety showed hydrogen bond interaction with Tyr612 whereas the
6th and 7th position of
naphthyridinol functionality showed Vander Waals interaction with Gln817 and Ala 767. Four amino acids namely Met760, Met758, Leu756 and Thr723 made the hydrophobic environment around methyl group. The second lead compound NSC 412 also showed prominent interaction (Figure 4B) with a good LibDock score of 74.443. The hydroxyl group present on each phenyl rings showed hydrogen bond interaction with Tyr612 and Thr723 along with Vander waals interaction between oxygen atom of hydroxyl group and Leu765. Another Vander Waals interaction was seen between Ala 767 and the phenol ring. Hydrophobic environment was constituted around methyl group by Met758, Leu754, Met760 and Leu756. Another lead compound NSC 619 (Figure 4c) with a prominent LibDock score of 79.956 showed hydrogen bond interaction between the ketonic group of penatanone moiety and His617. Whereas, methoxy group present at the 3rd position exhibited Vander waals interaction with Leu725. Methyl group of pentanone was surrounded by various hydrophobic amino acids including Met760, Leu756, Thr723 and Met758. Review of previous literature has also revealed the importance of Gln817 in PDE-5 and ligand interaction.22 In our docking study also Gln817 has appeared as an important amino acid but it has also been observed that in addition to Gln817 other amino acids such as Tyr 612, Thr723, Leu765, Ala767 and Leu725 also play an important role in molecular interaction.
Figure 4: Docked conformation of NCI compounds on the active site of PDE5. The green dotted lines represent hydrogen bond interaction and pink dotted lines represent the Vander Waals interaction (A) NSC 22 showed interactions with Tyr612, Gln817 and Ala 767. The hydrophobic environment around methyl group was characterized by four amino acids namely Met760, Met758, Leu756 and Thr723. (B) NSC 412 showed interactions with Tyr612, Thr723, Leu765 and Ala 767. Hydrophobic environment was constituted around methyl group by Met758, Leu754, Met760 and Leu756. (C) NSC 619 showed interactions with His617 and Leu725. Methyl group of pentanone was surrounded by various hydrophobic amino acids including Met760, Leu756, Thr723, Met758. Since, NSC 22, NSC 412 and NSC 619 showed good interactions during docking studies, before proceeding to experimental validation they were checked for novelty employing pairwise tanimoto23 similarity indices protocol available in DS. The similarity of NSC 22, NSC 412 and NSC 619 were compared with known PDE 5 inhibitors from binding DB. NSC 22, NSC 412 and NSC 619 showed low tanimto similarity indices of 0.175, 0.224 and 0.257 to all the structures of established PDE5 inhibitors. In view of the fact that all the three novel
lead compounds showed good estimated activity, fit value, LibDock score and no violation to Lipinski’s, they were subjected to experimental validation using an in-vitro vasodilation assay on rat aortic ring as reported in previous literature.24 The results of contraction and relaxation were recorded using student physiograph and it was observed that on precontracted rat aortic ring the reference drug Sildenafil, NSC619, NSC412 and NSC22 induced 50% relaxation at 10-6 M, 10-5 M, 10-4 M and 10-2 M respectively. The dose-response curves for sildenafil, NSC22, NSC 412 and NSC619 is shown in Fig. 5. In comparison to standard drug it was observed that out of three; NSC 619 appeared to be most potent followed by NSC 412 and NSC 22.
Figure 5: Vasorelaxation activity of sildenafil and hits (NSC22, NSC412, and NSC619) on different concentrations.
Most importantly the results of vasorelaxation activity of all the three compounds were found to be in full agreement to the pharmacophore based virtual screening results. The most active compound NSC 619 was estimated high in terms of fit and estimated value of 5.85 and 9.12 nM respectively. On the other hand, the second most active compound showed fit and estimated value of 5.75 and 11.44 nM and as expected the moderately active compound was also estimated at 14.48nM with fit value of 5.64.
In conclusion, through our pharmacophore based virtual screening workflow we have identified three novel and structurally diverse PDE5 inhibitors with vasorelaxation activity ranging from 10-2 and 10-5M. The result of our study clearly shows that if elucidated and validated properly the ligand based pharmacophore can be a powerful source for identification of novel leads from chemical compound databases. Acknowledgment Computational resources were provided by Banasthali University, and the authors thank Vice Chancellor for providing all the necessary facilities.
Supplementary data 1. Methodology of in-vitro vasodilation assay performed on lead compounds. 2. Table S1: Structures and biological activity of tetracyclic guanine as PDE-5 inhibitors 3. Table S2: Fisher’s Randomization test 95 % results of the pharmacophore hypothesis (Hypo-1) References 1. Echeverri, D.; Montes, F.R.; Cabrera, M.; Galan, A.; Prieto, A. Int. J. Vasc. Med. 2010, 2010, 1. 2. Halpin, D.M.G. Int. J. Chron. Obstruct. Pulmon. Dis. 2008, 3, 543. 3. Conti, M. Mol. Endocrinol. 2000, 14, 1317. 4. Mehats, C.; Andersen, C.B.; Filopanti, M.; Jin, S.L.; Conti, M.Trends. Endocrinol. Metab. 2002, 13, 29. 5. Rybalkina, I.G.; Tang, X.B.; Rybalkin, S.D. Mol. Pharmacol. 2010, 77, 670. 6. Kulkarni, S.K.; Patil, C.S. Methods Find Exp. Clin. Pharmacol. 2004, 26, 789. 7. Schwarz, E.R.; Kapur, V.; Rodriguez, J.; Rastogi, S.; Rosanio, S. Int. J. Impot. Res. 2007, 19, 139. 8. Singh, T.P. Expert. Rev. Respir. Med. 2010, 4, 13. 9. Valasani, K.R.; Vangavaragu, J.R.; Day, V.W.; Yan, S.S. J. Chem. Inf. Model. 2014, 54, 902. 10. Hou, T.; Xu, X. Curr. Pharm.Des. 2004, 10, 1011. 11. Ahn, S.H.; Bercovici, A.; Boykow, G.; Bronnenkant, A.; Chackalamannil, S.; Chow, J.; Cleven, R.; Cook, J.; Czarniecki, M.; Domalski, C.; Fawzi, A.; Green, M.; Gundes,
A.; Ho, G.; Laudicina, M.; Lindo, N.; Ma, K.; Manna, M.; McKittrick, B.; Mirzai, B.; Nechuta, T.; Neustadt, B.; Puchalski, C.; Pula, K.; Silverman, L.; Smith, E.; Stamford, A.; Tedesco, R.P.; Tsai, H.; Tulshian, D.; Vaccaro, H.; Watkins, R.W.; Weng, X.; Witkowski, J.T.; Xia, Y.; Zhang, H. J .Med. Chem. 1997, 40, 2196. 12. Suresh, N.; Vasanthi, N.S. J.P.B. 2010, 3, 020. 13. Chen, I.J.; Foloppe, N. J. Chem. Inf. Model. 2008, 48, 1773. 14. Arooj, M.; Thangapandian, S.; John, S.; Hwang, S.; Park, J.K; Lee, K.W. Int. J. Mol. Sci. 2011, 12, 9236. 15. Bharatham, N.; Bharatham, K.; Lee, K.W. J. Mol. Graph. Model. 2007, 25, 813. 16. Tsai, K.C.; Chen, S.Y.; Liang, P.H.; Lu, I.L.; Mahindroo, N.; Hsieh, H.P.; Chao, Y.S.; Liu, L.; Liu, D.; Lien, W.; Lin, T.H.; Wu, S.Y. J .Med. Chem. 2006, 49, 3485. 17. Roy, K.; Chakraborty, P.; Mitra, I.; Ojha, P.K.; Kar, S.; Das, R.N. J. Comput. Chem. 2013, 34, 1071. 18. Mitra, I.; Saha, A.; Roy, K. Sci. Pharm. 2013, 81, 57. 19. Ukita, T.; Nakamura, Y.; Kubo, A.; Yamamoto, Y.; Takahashi, M.; Kotera, J.; Ikeo, T. J. Med. Chem. 1999, 42, 1293. 20. Xiang, M.; Lei, K.; Fan, W.; Lin, Y.; He, G.; Yang, M.; Chen, L.; Mo, Y. Drug Des. Devel. Ther. 2013, 7, 789. 21. Shin, H.L.; Josephine, W.W.; Hsuan, L.L.; Jian, H.Z.; Kung, T.L.; Chih, K.C.; Hsin, Y. L.; Wei, B.T.; Yih, H. J. Biomed. Sci. 2011, 18, 1. 22. Chandrasekaran, M.; Sakkiah, S.; Lee, K.W. J. Taiwan Inst. Chem. E. 2011, 42, 709. 23. Shen,M.Y.;Yu,H.;Li,Y.;Li,P.;Pan,P.;Zhou,S.;Zhang,L.;Li,S.;Lee,S.M.Y.;Hou,T.Mol. Biosyst.2013,9,1511. 24. Liu, J.; Zhao, M.; Cui, G.; Zhang, X.; Wang, J.; Peng, S. J. Med. Chem. 2008, 51, 4715.
Table1: Results of the top 10 pharmacophore hypotheses generated by the Hypogen. Hypo
Total cost
Cost difference
RMSD
R2
Feature
1
133.23
45.03
0.95
0.79
1HBA,1HY,1RA
2
146.10
32.16
1.29
0.60
1HBA,1HY,1RA
3
160.48
17.79
1.61
0.40
1HBA,1HY,1RA
4
166.74
11.52
1.72
0.32
1HBA,1HY,1RA
5
166.82
11.45
1.73
0.31
1HBA,1HY,1RA
6
168.42
9.85
1.75
0.29
1HBA,1HY,1RA
7
169.10
9.16
1.77
0.29
1HBA,1HY,1RA
8
170.25
8.02
1.79
0.27
1HBA,1HY,1RA
9
170.98
7.28
1.80
0.25
1HBA,1HY,1RA
10
171.09
7.17
1.80
0.25
1HBA,1HY,1RA
Table 2: Hits retrieved from NCI database as potent PDE5 inhibitors
Name of Hits
Estimated value
Fit value
Structure
(nM) NSC 960
15.467
5.621
O O N
NSC 22
14.485
5.649
H2N
N
N O OH O
NSC 1015
13.314
5.686 N OH HN O O
NSC 412
11.444
5.751 HO
N N
HO
NSC 619
9.128
5.85
OCH3
H3CO O
NSC 1012
7.323
5.945 N OH HN O OH
Figure Captions
Figure 1. Plot of predicted versus the corresponding actual activity (IC50) for the training and test compounds. Figure 2. Best pharmacophore model (Hypo1) of PDE-5 inhibitors generated by hypogen module. (a) Pharmacophore mapping aligned with the most active training set compound 50. (b) Pharmacophore mapping of the least active training set compound 47. Figure 3: Pharmacophoric feature mapping of external and marketed drug. (a) Three pharmacophoric feature mapping of most active lignan derivative with a fit value of 5.47. (b) Three pharmacophoric feature mapping of sildenafil with a fit value of 5.55 (c) Three pharmacophoric feature mapping of tadalfil with a fit value of 5.86 (d) Three Pharmacophoric feature mapping of vardenafil with a fit value of 4.48. (e) Plot of experimental values versus the corresponding actual values (IC50) for the external test compounds.
Figure 4: Docked conformation of NCI compounds on the active site of PDE5. The green dotted lines represent hydrogen bond interaction and pink dotted lines represent the Vander Waals interaction (a) NSC 22 showed interactions with Tyr612, Gln817 and Ala 767. The hydrophobic environment around methyl group was characterized by four amino acids namely Met760, Met758, Leu756 and Thr723. (b) NSC 412 showed interactions with Tyr612, Thr723, Leu765 and Ala 767. Hydrophobic environment was constituted around methyl group by Met758, Leu754, Met760 and Leu756. (c) NSC 619 showed interactions with His617 and Leu725. Methyl group of pentanone was surrounded by various hydrophobic amino acids including Met760, Leu756, Thr723, Met758. Figure 5: Vasorelaxation activity of sildenafil and hits (NSC22, NSC412, NSC619) at different concentrations.
Figure 1 -5 Training s et
-4.5
Tes t s et
-3.5 -3 -2.5 -2 -1.5 -1 -5
-4
-3
-2
Actual activity
-1
-0.5 0 0
Predicted activity
-4
L inear (Training s et) L inear (Tes t s et) Training set r² = 0.7944 Test set r² = 0.616
Figure 2
Figure 3
Figure 4
Figure 5
S0960-894X(14)00491-0 http://dx.doi.org/10.1016/j.bmcl.2014.05.004 BMCL 21617
To appear in:
Bioorganic & Medicinal Chemistry Letters
Received Date: Revised Date: Accepted Date:
8 February 2014 30 April 2014 2 May 2014
Please cite this article as: Mittal, A., Paliwal, S., Sharma, M., Singh, A., Sharma, S., Yadav, D., Pharmacophore Based Virtual Screening, Molecular Docking and Biological Evaluation to Identify Novel PDE5 Inhibitors with Vasodilatory Activity, Bioorganic & Medicinal Chemistry Letters (2014), doi: http://dx.doi.org/10.1016/j.bmcl. 2014.05.004
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Pharmacophore Based Virtual Screening, Molecular Docking and Biological Evaluation to Identify Novel PDE5 Inhibitors with Vasodilatory Activity Anupama Mittal, Sarvesh Paliwal,* Mukta Sharma, Aarti Singh, Swapnil Sharma, Divya Yadav Department of Pharmacy, Banasthali University, Banasthali, Rajasthan, India
*Corresponding Author‘s mailing address and email ID: Prof. Sarvesh Paliwal Department of Pharmacy Banasthali University, P.O. Banasthali Rajasthan-304022 Phone no. +91-9352141469 Email I.D. [email protected] Email addresses of all authors: AM: [email protected] SP: [email protected] MS: [email protected] AS: [email protected] SS: [email protected] DY: [email protected]
ABSTRACT: Prompted by the role of PDE5 and its closely associated cAMP and cGMP in hypertension, we have attempted to discover novel PDE5 inhibitors through ligand based virtual screening. Rigorously validated model comprising of one HBA, one HY and one RA was used as a query to search the NCI database leading to retrieval of many compounds which were screened on the basis of estimated activity, fit value and Lipinski’s violation. Selected compounds were subjected to docking studies which resulted into visualization of potential interaction capabilities of NCI compounds in line to pharmacophoric features. Finally three compounds were subjected to in-vitro evaluation using the isolated rat aortic model. The results showed that all three compounds are potent and novel PDE5 inhibitors with vasodilatory activity range from 10-2 to 10-5M. Keywords: Phosphodiesterase-5, cyclic guanosine monophosphate, discovery studio, hydrogen bond acceptor, cyclic adenosine monophosphate A number of endogenous substances are implicated in vasorelaxation among which cyclic adenosine monophosphate (cAMP) and cyclic guanosine monophosphate (cGMP) are major ones.1 Phosphodiesterase enzyme (PDE) hydrolyze the cAMP and cGMP to inactive metabolite 5’AMP and 5’GMP respectively and play a very crucial role in regulating the concentration level of these secondary messengers.2 The capacity of PDEs to degrade cAMP and cGMP is tenfold greater as compared to the rate of formation of secondary messenger, hence modification in the activity of PDE enzyme has a great impact on physiological activities. In the human genome 21 genes encoding PDEs have been identified.3 These 21 genes are assembled into 11 PDE (PDE1-PDE11) families on the basis of amino acid sequences, structures, enzyme kinetics, mode of action and tissue distributions.4 Out of the 11 families of phosphodiesterases, PDE5 is a cGMP specific enzyme which contains two catalytic domains GAF-A and GAF-B. cGMP bind to the GAF-A domain which facilitates the phosphorylation of N-terminal with the help of protein kinase-A and protein kinase-G
thus stimulates the hydrolysis of cGMP.5 PDE5 is located in many organs such as lung, heart, cerebellum, kidney, and smooth muscle.6 The vasodilation of vascular smooth muscle in hypertensive patients can be enhanced by blocking the hydrolysis of cGMP via PDE-5 isoenzyme. FDA in 1998 approved three PDE-5 inhibitors namely sildenafil, vardenafil, and tadalafil for the treatment of erectile dysfunction.7 Later it was observed that PDE5 inhibitor also have the ability to reduce pulmonary arterial resistance and pressure which may be helpful for the treatment of pulmonary hypertension but it is noticeable that out of three PDE5 inhibitors sildenafil and tadalfil are approved for their therapeutic use in pulmonary hypertension.8 In view of potential role of PDE5 inhibitors in hypertension we envisaged that identification of novel PDE5 inhibitors with diverse structures, druggability and high potency can enrich this class of antihypertensive drugs. The utility of molecular modeling and virtual screening in lead identification and optimization are unquestionable.9-10 Prompted by the successful application of in-silico pharmacophore based virtual screening in lead identification, we have made an effort to implement in-silico protocols in association with wet lab experimentation to identify novel PDE5 inhibitors. As a starting point a data set of 48 PDE-5 inhibitors with wide range of activity (4nM to 26,000 nM) and structural diversity (Supplementary data Table S1) were used for development of pharmacophore model.11 The Hypogen module of Accelrys Discovery studio v2.0 (DS) was used to generate the pharmacophore models. Energy of all the compounds was minimized using CHARMm force field with an energy gradient of 0.001 kcal/mol/A°.12 Total 34 compounds were used as training set while the rest of 14 compounds were used as internal test set to validate the pharmacophore model. A maximum of 255 diverse conformers were generated for each molecule using best flexible conformation generation module which performs a rigorous energy minimization in both torsional and cartesia to attain the best
coverage of conformational space.13 Features such as hydrogen bond acceptor (HBA), hydrogen bond donor (HBD), hydrophobic (HY), positive ionizable (PI) and ring aromatic (RA) were used to generate various hypotheses keeping minimum and maximum features value between 0 and 5. Two compounds (66, 74) showed a greater error ratio and removal of these compounds improved the quality of model hence they were removed during the course of pharmacophore modeling. The quality and integrity of all generated hypotheses were evaluated on the basis of cost function analysis, RMS, and correlation coefficient (Table 1). The chosen set of hypothesis exhibited a fixed cost value of 118.63 bits and as recommended it was well separated from the null cost (178.27 bits). Out of 10 hypotheses hypo1 showed maximum difference of 45 bits between total cost and the null cost representing 75-90% chances of true correlation. For a significant model the value of configuration14 should be less than 17 and the hypo1 showed a value of 10.07 with RMS value of 0.95 and squared correlation coefficient value of 0.79 (Figure 1) confirming the significance of the model.
Table1: Results of the top 10 pharmacophore hypotheses generated by the Hypogen. Hypo
Total cost
Cost difference
RMSD
R2
Feature
1
133.23
45.03
0.95
0.79
1HBA,1HY,1RA
2
146.10
32.16
1.29
0.60
1HBA,1HY,1RA
3
160.48
17.79
1.61
0.40
1HBA,1HY,1RA
4
166.74
11.52
1.72
0.32
1HBA,1HY,1RA
5
166.82
11.45
1.73
0.31
1HBA,1HY,1RA
6
168.42
9.85
1.75
0.29
1HBA,1HY,1RA
7
169.10
9.16
1.77
0.29
1HBA,1HY,1RA
8
170.25
8.02
1.79
0.27
1HBA,1HY,1RA
9
170.98
7.28
1.80
0.25
1HBA,1HY,1RA
10
171.09
7.17
1.80
0.25
1HBA,1HY,1RA
Figure 1. Plot of predicted versus the corresponding actual activity (IC50) for the training and test compounds. Based on the above results Hypo 1 comprising of one HBD, one RA, and one HY features was identified as statistically significant hypothesis and was subjected to CatScramble validation which generates pharmacophore hypotheses by randomizing the activity data of the training set compounds by using the formula,15 significance = 100 (1-(1+x/y)), where x signifies the total number of hypotheses with a total cost value lower than the original hypothesis, and y symbolize the total number of HypoGen runs i.e. initial and random runs. During the validation process random spreadsheets were generated using the same training set compounds. A total of 19 random spreadsheets16 (random hypotheses) were generated to achieve a confidence level of 95%. The results of the randomized data were analyzed to evaluate the quality of the model and it was observed that the statistics of Hypo 1 were far better as compared to the 19 random hypotheses (Supplementary data Table S2).The results of CatScramble clearly shows that the generation of Hypo 1 was not by chance rather it has been an outcome of true correlation between structures of training set compounds and PDE 5 inhibitory activity. After initial round of validation, mapping of training set compounds was carried out and it was observed that most active compound (50~IC50 4nM) of training set
mapped well over the hypo1 with a fit value of 5.56, the HY feature (light blue color) mapped over benzene ring containing methyl group, the HBA (green color) mapped over the ketonic group of pyrimidine ring and ring aromatic (orange color) group mapped over the imidazole ring (Figure 2A). On the other hand least active compound (47~IC5026000 nM) missed one hydrophobic feature when mapped over Hypo 1 (Figure 2B) and also showed a poor fit value of 2.507.
Figure 2. Best pharmacophore model (Hypo1) of PDE-5 inhibitors generated by hypogen module. (A) Pharmacophore mapping aligned with the most active training set compound 50. (B) Pharmacophore mapping of the least active training set compound 47. The utility of any pharmacophore model lies in its ability to correctly predict the activity of compounds within the domain and outside the domain. The internal test set comprising of 14 compounds was mapped onto hypo-1 and a good correlation was observed as depicted in Figure 1. Roy et al. have recently developed a new parameter, modified r² (rm²), which considers the actual difference between the observed and predicted response data, thereby serving as a more stringent measure for assessment of model predictivity compared to the traditional validation parameters.17-18 In the present study, we performed rm2 metric test on training and test sets in which the value of "average rm2" was found to be 0.70 and 0.50 and the value of "delta rm2" was found to be 0.16 and 0.05 for training and test sets respectively. Ideally the values of “Average rm2” must be greater than 0.5 and values of “Delta rm2" must be lower than 0.2.
As a final step towards the assurance of applicability, predictivity and soundness of the model an external dataset of 12 structurally diverse compounds comprising of 1-arylnaphthalene lignan derivative,19 sildenafil, vardenafil and tadalfil were used as an external test set. The pharmacophore mapping pattern of external test set compounds were analyzed and it was observed that the entire test set compounds mapped very well over the hypo 1 with good fit values ranging within 3.727 to 5.47. Most importantly all the three marketed drugs and the most active lignan derivative exhibited a perfect three-feature mapping with good fit values as shown in Figure 3A,3B,3C and 3D. The r2 value of 0.68 (Fig. 3E) between predicted and actual activity for external test set compounds testified the universality of the hypo-1.
Figure 3: Pharmacophoric feature mapping of external and marketed drug. (A) Three pharmacophoric feature mapping of most active lignan derivative with a fit value of 5.47. (B) Three pharmacophoric feature mapping of sildenafil with a fit value of 5.55 (C) Three pharmacophoric feature mapping of tadalfil with a fit value of 5.86 (D) Three Pharmacophoric feature mapping of vardenafil with a fit value of 4.48. (E) Plot of experimental values versus the corresponding actual values (IC50) for the external test compounds.
Needless to mention that utility of validated pharmacophore model lies in virtual screening and since the developed pharmacophore model showed all the signs of its soundness and universality, it was used to screen National Cancer Institute database comprising of 260,071 structurally diverse compounds20 using best flexible database search option. The retrieved hits were screened on the basis of their fit and estimated value which led to retention of 25 out of 300 hits. Lipinski’s rule of five was applied to check their druggable properties which further reduced the list of hits to 6 with fit value ranging within 5.6 -5.9 as shown in Table 2.
Table 2: Hits retrieved from NCI database as potent PDE5 inhibitors Name of Hits
NSC 960
Estimated value IC50 (nM) 15.467
Fit value
Structure
O
5.621
O N
NSC 22
14.485
5.649
H2N
N
N O OH O
NSC 1015
13.314
5.686 N OH HN O O
NSC 412
11.444
5.751 HO
N N
HO
NSC 619
9.128
OCH3
5.85 H3CO
O
NSC 1012
7.323
5.945 N OH HN O OH
Since out of six lead compounds only three compounds were available with NCI for experimental validation, all the three compounds were subjected to docking to evaluate the binding interaction between different amino acids of target enzyme and lead compounds. A high resolution crystal structure of PDE-5 bound with an inhibitory molecule, sildenafil citrate (PDB code1UDT) was selected for molecular docking studies employing LibDock module available within DS. The crystal structure was checked for valency, missing hydrogen and any structural disorders like connectivity and bond orders. All the water molecules were removed from the protein hierarchy and only A chain was retained which was split into the protein and crystal ligand part.21 In the protein structure hydrogen atoms
were added and CHARMm forcefield was applied for energy minimization. Binding site present in the A chain was defined within the radius of 9 Å. All the three hits (NSC22, NSC412, NSC619) were docked into the crystal structure of PDE-5 enzyme and various poses were examined on the basis of LibDocker score (protein-ligand interaction energy) to evaluate the nature and type of interactions. It was observed that NSC22 docked (Figure 4A) very well with a LibDock score of 70.071 and the hydroxyl group of naphthyridinol moiety showed hydrogen bond interaction with Tyr612 whereas the
6th and 7th position of
naphthyridinol functionality showed Vander Waals interaction with Gln817 and Ala 767. Four amino acids namely Met760, Met758, Leu756 and Thr723 made the hydrophobic environment around methyl group. The second lead compound NSC 412 also showed prominent interaction (Figure 4B) with a good LibDock score of 74.443. The hydroxyl group present on each phenyl rings showed hydrogen bond interaction with Tyr612 and Thr723 along with Vander waals interaction between oxygen atom of hydroxyl group and Leu765. Another Vander Waals interaction was seen between Ala 767 and the phenol ring. Hydrophobic environment was constituted around methyl group by Met758, Leu754, Met760 and Leu756. Another lead compound NSC 619 (Figure 4c) with a prominent LibDock score of 79.956 showed hydrogen bond interaction between the ketonic group of penatanone moiety and His617. Whereas, methoxy group present at the 3rd position exhibited Vander waals interaction with Leu725. Methyl group of pentanone was surrounded by various hydrophobic amino acids including Met760, Leu756, Thr723 and Met758. Review of previous literature has also revealed the importance of Gln817 in PDE-5 and ligand interaction.22 In our docking study also Gln817 has appeared as an important amino acid but it has also been observed that in addition to Gln817 other amino acids such as Tyr 612, Thr723, Leu765, Ala767 and Leu725 also play an important role in molecular interaction.
Figure 4: Docked conformation of NCI compounds on the active site of PDE5. The green dotted lines represent hydrogen bond interaction and pink dotted lines represent the Vander Waals interaction (A) NSC 22 showed interactions with Tyr612, Gln817 and Ala 767. The hydrophobic environment around methyl group was characterized by four amino acids namely Met760, Met758, Leu756 and Thr723. (B) NSC 412 showed interactions with Tyr612, Thr723, Leu765 and Ala 767. Hydrophobic environment was constituted around methyl group by Met758, Leu754, Met760 and Leu756. (C) NSC 619 showed interactions with His617 and Leu725. Methyl group of pentanone was surrounded by various hydrophobic amino acids including Met760, Leu756, Thr723, Met758. Since, NSC 22, NSC 412 and NSC 619 showed good interactions during docking studies, before proceeding to experimental validation they were checked for novelty employing pairwise tanimoto23 similarity indices protocol available in DS. The similarity of NSC 22, NSC 412 and NSC 619 were compared with known PDE 5 inhibitors from binding DB. NSC 22, NSC 412 and NSC 619 showed low tanimto similarity indices of 0.175, 0.224 and 0.257 to all the structures of established PDE5 inhibitors. In view of the fact that all the three novel
lead compounds showed good estimated activity, fit value, LibDock score and no violation to Lipinski’s, they were subjected to experimental validation using an in-vitro vasodilation assay on rat aortic ring as reported in previous literature.24 The results of contraction and relaxation were recorded using student physiograph and it was observed that on precontracted rat aortic ring the reference drug Sildenafil, NSC619, NSC412 and NSC22 induced 50% relaxation at 10-6 M, 10-5 M, 10-4 M and 10-2 M respectively. The dose-response curves for sildenafil, NSC22, NSC 412 and NSC619 is shown in Fig. 5. In comparison to standard drug it was observed that out of three; NSC 619 appeared to be most potent followed by NSC 412 and NSC 22.
Figure 5: Vasorelaxation activity of sildenafil and hits (NSC22, NSC412, and NSC619) on different concentrations.
Most importantly the results of vasorelaxation activity of all the three compounds were found to be in full agreement to the pharmacophore based virtual screening results. The most active compound NSC 619 was estimated high in terms of fit and estimated value of 5.85 and 9.12 nM respectively. On the other hand, the second most active compound showed fit and estimated value of 5.75 and 11.44 nM and as expected the moderately active compound was also estimated at 14.48nM with fit value of 5.64.
In conclusion, through our pharmacophore based virtual screening workflow we have identified three novel and structurally diverse PDE5 inhibitors with vasorelaxation activity ranging from 10-2 and 10-5M. The result of our study clearly shows that if elucidated and validated properly the ligand based pharmacophore can be a powerful source for identification of novel leads from chemical compound databases. Acknowledgment Computational resources were provided by Banasthali University, and the authors thank Vice Chancellor for providing all the necessary facilities.
Supplementary data 1. Methodology of in-vitro vasodilation assay performed on lead compounds. 2. Table S1: Structures and biological activity of tetracyclic guanine as PDE-5 inhibitors 3. Table S2: Fisher’s Randomization test 95 % results of the pharmacophore hypothesis (Hypo-1) References 1. Echeverri, D.; Montes, F.R.; Cabrera, M.; Galan, A.; Prieto, A. Int. J. Vasc. Med. 2010, 2010, 1. 2. Halpin, D.M.G. Int. J. Chron. Obstruct. Pulmon. Dis. 2008, 3, 543. 3. Conti, M. Mol. Endocrinol. 2000, 14, 1317. 4. Mehats, C.; Andersen, C.B.; Filopanti, M.; Jin, S.L.; Conti, M.Trends. Endocrinol. Metab. 2002, 13, 29. 5. Rybalkina, I.G.; Tang, X.B.; Rybalkin, S.D. Mol. Pharmacol. 2010, 77, 670. 6. Kulkarni, S.K.; Patil, C.S. Methods Find Exp. Clin. Pharmacol. 2004, 26, 789. 7. Schwarz, E.R.; Kapur, V.; Rodriguez, J.; Rastogi, S.; Rosanio, S. Int. J. Impot. Res. 2007, 19, 139. 8. Singh, T.P. Expert. Rev. Respir. Med. 2010, 4, 13. 9. Valasani, K.R.; Vangavaragu, J.R.; Day, V.W.; Yan, S.S. J. Chem. Inf. Model. 2014, 54, 902. 10. Hou, T.; Xu, X. Curr. Pharm.Des. 2004, 10, 1011. 11. Ahn, S.H.; Bercovici, A.; Boykow, G.; Bronnenkant, A.; Chackalamannil, S.; Chow, J.; Cleven, R.; Cook, J.; Czarniecki, M.; Domalski, C.; Fawzi, A.; Green, M.; Gundes,
A.; Ho, G.; Laudicina, M.; Lindo, N.; Ma, K.; Manna, M.; McKittrick, B.; Mirzai, B.; Nechuta, T.; Neustadt, B.; Puchalski, C.; Pula, K.; Silverman, L.; Smith, E.; Stamford, A.; Tedesco, R.P.; Tsai, H.; Tulshian, D.; Vaccaro, H.; Watkins, R.W.; Weng, X.; Witkowski, J.T.; Xia, Y.; Zhang, H. J .Med. Chem. 1997, 40, 2196. 12. Suresh, N.; Vasanthi, N.S. J.P.B. 2010, 3, 020. 13. Chen, I.J.; Foloppe, N. J. Chem. Inf. Model. 2008, 48, 1773. 14. Arooj, M.; Thangapandian, S.; John, S.; Hwang, S.; Park, J.K; Lee, K.W. Int. J. Mol. Sci. 2011, 12, 9236. 15. Bharatham, N.; Bharatham, K.; Lee, K.W. J. Mol. Graph. Model. 2007, 25, 813. 16. Tsai, K.C.; Chen, S.Y.; Liang, P.H.; Lu, I.L.; Mahindroo, N.; Hsieh, H.P.; Chao, Y.S.; Liu, L.; Liu, D.; Lien, W.; Lin, T.H.; Wu, S.Y. J .Med. Chem. 2006, 49, 3485. 17. Roy, K.; Chakraborty, P.; Mitra, I.; Ojha, P.K.; Kar, S.; Das, R.N. J. Comput. Chem. 2013, 34, 1071. 18. Mitra, I.; Saha, A.; Roy, K. Sci. Pharm. 2013, 81, 57. 19. Ukita, T.; Nakamura, Y.; Kubo, A.; Yamamoto, Y.; Takahashi, M.; Kotera, J.; Ikeo, T. J. Med. Chem. 1999, 42, 1293. 20. Xiang, M.; Lei, K.; Fan, W.; Lin, Y.; He, G.; Yang, M.; Chen, L.; Mo, Y. Drug Des. Devel. Ther. 2013, 7, 789. 21. Shin, H.L.; Josephine, W.W.; Hsuan, L.L.; Jian, H.Z.; Kung, T.L.; Chih, K.C.; Hsin, Y. L.; Wei, B.T.; Yih, H. J. Biomed. Sci. 2011, 18, 1. 22. Chandrasekaran, M.; Sakkiah, S.; Lee, K.W. J. Taiwan Inst. Chem. E. 2011, 42, 709. 23. Shen,M.Y.;Yu,H.;Li,Y.;Li,P.;Pan,P.;Zhou,S.;Zhang,L.;Li,S.;Lee,S.M.Y.;Hou,T.Mol. Biosyst.2013,9,1511. 24. Liu, J.; Zhao, M.; Cui, G.; Zhang, X.; Wang, J.; Peng, S. J. Med. Chem. 2008, 51, 4715.
Table1: Results of the top 10 pharmacophore hypotheses generated by the Hypogen. Hypo
Total cost
Cost difference
RMSD
R2
Feature
1
133.23
45.03
0.95
0.79
1HBA,1HY,1RA
2
146.10
32.16
1.29
0.60
1HBA,1HY,1RA
3
160.48
17.79
1.61
0.40
1HBA,1HY,1RA
4
166.74
11.52
1.72
0.32
1HBA,1HY,1RA
5
166.82
11.45
1.73
0.31
1HBA,1HY,1RA
6
168.42
9.85
1.75
0.29
1HBA,1HY,1RA
7
169.10
9.16
1.77
0.29
1HBA,1HY,1RA
8
170.25
8.02
1.79
0.27
1HBA,1HY,1RA
9
170.98
7.28
1.80
0.25
1HBA,1HY,1RA
10
171.09
7.17
1.80
0.25
1HBA,1HY,1RA
Table 2: Hits retrieved from NCI database as potent PDE5 inhibitors
Name of Hits
Estimated value
Fit value
Structure
(nM) NSC 960
15.467
5.621
O O N
NSC 22
14.485
5.649
H2N
N
N O OH O
NSC 1015
13.314
5.686 N OH HN O O
NSC 412
11.444
5.751 HO
N N
HO
NSC 619
9.128
5.85
OCH3
H3CO O
NSC 1012
7.323
5.945 N OH HN O OH
Figure Captions
Figure 1. Plot of predicted versus the corresponding actual activity (IC50) for the training and test compounds. Figure 2. Best pharmacophore model (Hypo1) of PDE-5 inhibitors generated by hypogen module. (a) Pharmacophore mapping aligned with the most active training set compound 50. (b) Pharmacophore mapping of the least active training set compound 47. Figure 3: Pharmacophoric feature mapping of external and marketed drug. (a) Three pharmacophoric feature mapping of most active lignan derivative with a fit value of 5.47. (b) Three pharmacophoric feature mapping of sildenafil with a fit value of 5.55 (c) Three pharmacophoric feature mapping of tadalfil with a fit value of 5.86 (d) Three Pharmacophoric feature mapping of vardenafil with a fit value of 4.48. (e) Plot of experimental values versus the corresponding actual values (IC50) for the external test compounds.
Figure 4: Docked conformation of NCI compounds on the active site of PDE5. The green dotted lines represent hydrogen bond interaction and pink dotted lines represent the Vander Waals interaction (a) NSC 22 showed interactions with Tyr612, Gln817 and Ala 767. The hydrophobic environment around methyl group was characterized by four amino acids namely Met760, Met758, Leu756 and Thr723. (b) NSC 412 showed interactions with Tyr612, Thr723, Leu765 and Ala 767. Hydrophobic environment was constituted around methyl group by Met758, Leu754, Met760 and Leu756. (c) NSC 619 showed interactions with His617 and Leu725. Methyl group of pentanone was surrounded by various hydrophobic amino acids including Met760, Leu756, Thr723, Met758. Figure 5: Vasorelaxation activity of sildenafil and hits (NSC22, NSC412, NSC619) at different concentrations.
Figure 1 -5 Training s et
-4.5
Tes t s et
-3.5 -3 -2.5 -2 -1.5 -1 -5
-4
-3
-2
Actual activity
-1
-0.5 0 0
Predicted activity
-4
L inear (Training s et) L inear (Tes t s et) Training set r² = 0.7944 Test set r² = 0.616
Figure 2
Figure 3
Figure 4
Figure 5