qsar based modeling on a series of α-hydroxy amides as a novel class of bradykinin b1 selective...
TRANSCRIPT
Journal of Saudi Chemical Society (2011) 15, 215–220
King Saud University
Journal of Saudi Chemical Society
www.ksu.edu.sawww.sciencedirect.com
ORIGINAL ARTICLE
QSAR based modeling on a series of a-hydroxy amides
as a novel class of bradykinin B1 selective antagonists
A.K. Srivastava *, Neerja Shukla, Avni Pandey, Akanchha Srivastava
QSAR Research Laboratory, Department of Chemistry, University of Allahabad, Allahabad 211 002, India
Received 15 April 2010; accepted 7 September 2010
Available online 19 September 2010
*
E
ne
m
13
El
Pe
do
KEYWORDS
QSAR;
Bradykinin B1 selective
antagonists;
Steric and topological
parameters
Corresponding author.
-mail addresses: draksrivasta
erjashuklaau2007@rediffmail
ail.com (A. Pandey), akanchh
19-6103 ª 2010 King Saud
sevier B.V. All rights reserve
er review under responsibilit
i:10.1016/j.jscs.2010.09.001
Production and h
va1@red
.com (N
a_au@red
Universit
d.
y of King
osting by E
Abstract G-protein coupled receptors like Bradykinin (BK) B1 represent a potential treatment
route for chronic pain and inflammation. Quantitative structure activity relationship has been per-
formed on a series of a-hydroxy amides as a novel class of bradykinin B1 selective antagonists, using
different physicochemical parameters along with appropriate indicator variables. It has been found
that physicochemical parameters such as connectivity indices 3v, 4v and 5v, molecular weight, molar
refractivity, density along with indicator variables are significantly correlated with activity. In this
paper best results were obtained by using multiple regression analysis. Different models were gen-
erated with high values of R2 and low values of PRESS/SSY ratio. The significant equations were
statistically tested by using leave one out (LOO) technique and cross validation methods.ª 2010 King Saud University. Elsevier B.V. All rights reserved.
1. Introduction
Bradykinin receptor B1 (B1) is a G-protein coupled receptorencoded by the BDKRB1 gene in humans. Its principal ligand
is bradykinin, a nine amino acid peptide generated in patho-
iffmail.com (A.K. Srivastava),
. Shukla), avnipandey@rediff
iffmail.com (A. Srivastava).
y. Production and hosting by
Saud University.
lsevier
physiologic conditions such as inflammation, trauma, burns,
shock, and allergy. The B1 receptor is one of the two Gprotein-coupled receptors that have been found to bind brady-kinin and to mediate responses to these pathophysiologicconditions. B1 protein is synthesized de novo following tissue
injury and receptor binding leading to an increase in the cyto-solic calcium ion concentration, ultimately resulting in chronicand acute inflammatory responses. The Bradykinin receptor is
not commonly expressed in healthy states but is automaticallyinduced upon injury. These receptors are expressed in centralNervous system of mice, rats and primates. [des-Arg]BK
(DABK) and [des-Arg] kallidin are natural agonists for brady-kinin B1 receptors, which are obtained by the metabolism ofbradykinin and kallidin.
2. Methodology
QSAR studies have been done on a series a-hydroxy amides
(Wood et al., 2008). The 2D structures of the molecules were
216 A.K. Srivastava et al.
drawn using Chem Sketch Software. Several physicochemical
parameters were used in the present study, which is discussedbelow:
(1) Density (D) – Density is a steric parameter and calcu-
lated by ACD Lab Chem Sketch Software. This param-eter is related with the bulk and size of the substituents.
D ¼MWðMolecular weightÞMvðMolecular volumeÞ
(2) Molecular weight (MW) – Molecular weight descriptor
has been used as a descriptor in systems such as trans-port studies where diffusion is the mode of operation.It is an important variable in QSAR studies pertainingto cross resistance of various drugs in multi-drug resis-
tant cell lines (Liu et al., 1997; Holder et al., 2006;Carton et al., 1983).
(3) Molar refractivity (MR) – Molar refractivity is gener-
ally considered to be a measure of overall bulk and isrelated to London dispersion forces as follows:
MR ¼ 4pNa=3
where N is Avogadro number and a is the polarizability
of the molecule. It gives no information about shape andis generally scaled 0.1 and has been extensively usedin QSAR (Hansch et al., 1973; Bauer et al., 1960;
Srivastava and Archana, 2003).
(4) Molecular connectivity (v) Randic, 1975; Randic, 1993 –Randic index 1v = 1v (G) of G was introduced by Ran-
dic in 1975 as the connectivity index.(5) Indicator variables (Senda and Fujita, 1991; Ismail
et al., 2002) – These are not QSAR parameters but areused to indicate the significance of any particular group
or species at a particular substitution site in a given ser-ies of drugs.
3. Results and discussion
QSAR was performed on a series of 21 compounds of a-hydroxyamides with the physicochemical parameters, which are listedin Table 1, where the biological activity (pIC50) is a measure
of inhibitory activity. All the QSARs reported here, were de-rived by us and were not reported in the original data set takenfrom the literature as reference. We have used Hansch analysis
(Hansch and Leo, 1995; Hansch et al., 1995) for developingthese models. The QSAR multiple regression analyses wereperformed with SPSS (7.5) version program. Their activity
data and the physicochemical parameters evaluated in the cor-relation are listed in Table 2. These parameters were found tobe useful earlier in QSAR based drug modeling (Srivastavaet al., 2008, 2009a,b,c, 2010; Srivastava and Archana,
2008a,b,c; Srivastava et al., 2009; Singh et al., 2008a,b).In order to study the role of different substituents at specific
positions indicator parameters such as I1 for H, I2 for –
CH2CH3 and I3 for –Cl at R1, R2 and R6 positions, respec-tively were introduced and are also listed in Table 1. Multiregression analysis of the data gave several models of which
the following equations were found to be significant.
pIC50 ¼ 1:510ð�0:817Þ5v� 0:672ð�1:314ÞI1� 1:122ð�2:065ÞI2 þ 0:834ð�1:560ÞI3 � 0:780 ð1Þ
n= 21, R = 0.942, R2 = 0.887, R2A ¼ 0:859, SE = 0.467,
F(4,16) = 31.344, Q= 2.017
pIC50 ¼ 0:766ð�0:576Þ4v� 0:567ð�1:716ÞI1� 1:357ð�2:615ÞI2 þ 0:910ð�1:986ÞI3 þ 1:679 ð2Þ
n= 21, R = 0.903, R2 = 0.816, R2A ¼ 0:770, SE = 0.595,
F(4,16) = 17.772, Q= 1.518
pIC50 ¼ 0:666ð�0:532Þ3v� 0:263ð�1:891ÞI1� 1:440ð�2:712ÞI2 þ 0:894ð�3:349ÞI3 þ 0:463 ð3Þ
n= 21, R = 0.895, R2 = 0.802, R2A ¼ 0:752, SE = 0.618,
F(4,16) = 16.182, Q= 1.448
pIC50 ¼ 0:015ð�0:012ÞMW� 0:540ð�1:863ÞI1� 1:199ð�2:824ÞI2 þ 0:579ð�2:241ÞI3 þ 1:363 ð4Þ
n= 21, R = 0.888, R2 = 0.788, R2A ¼ 0:735, SE = 0.639,
F(4,16) = 14.869, Q= 1.390
pIC50 ¼ 0:096ð�0:081ÞMR� 0:357ð�1:932ÞI1� 2:053ð�2:827ÞI2 þ 0:614ð�2:233ÞI3 � 1:604 ð5Þ
n= 21, R = 0.887, R2 = 0.787, R2A ¼ 0:734, SE = 0.640,
F(4,16) = 14.806, Q= 1.386
pIC50 ¼ 6:656ð�7:028ÞD� 1:169ð�1:961ÞI1� 0:634ðpm3:349ÞI2 þ 1:129ð�2:388ÞI3 � 1:545 ð6Þ
n= 21, R = 0.853, R2 = 0.727, R2A ¼ 0:659, SE = 0.726,
F(4,16) = 10.651, Q= 1.174In all the above models n is the number of data points, R is
correlation coefficient of determination, SE is the standard er-ror of estimate, R2
A represents adjusted R2 or explained vari-ance F is variance ratio (Diudea, 2000) between observed
and calculated activity. Q is the quality of fit (Pogliani, 1994;Pogliani, 1996) and data within parenthesis are for the 95%confidence intervals.
The positive coefficients of 3v, 4v and 5v, MW, MR and Din the above models indicate that more bulkier groups withthird, fourth and fifth orders of branching are preferable for
the activity.The positive sign of coefficient I3 makes it clear that Cl
group at R6 position is beneficial for the activity, while nega-tive coefficients of I1 and I2 explain that H and –CH2CH3 at
R1 and R2 positions respectively, should be strictly avoidedin future drug modeling.
Predicted and residual values for the best model Eq. (1) are
given in Table 3. In this equation the F ratio value is muchhigher than theoretical F value (F(4,16) = 3.01) indicating thestatistical significance of this equation. Predicted values are
the calculated activities of the equation and the residual valuesare the differences between the observed biological activitiesand the calculated activities, which are found to be low.
The plot of observed pIC50 versus predicted pIC50 based onthe Eq. (1) is shown in graph (Fig. 1) and the predicted R2 wasfound to be fairly large.
3.1. Cross validation
The cross validation analysis was performed using leave one
out (LOO) method (Carmer et al., 1988; Podloga andFerguson, 2000), in which one compound is removed from
Table 1 Biological activity, physicochemical data for Bradykinin B1 selective antagonists.
X
R6
R7
NH
R1
OH
R5
R2
R 4
R3
O
S.No. R1 R2 R3 R4 R5 R6 pIC50 R7 X D MR MW 3v 4v 5v I1 I2 I3
1. –CH2– –CH2– H F COOCH3 F 7.228 H H 1.321 81.94 335.3 9.591 6.911 5.147 0 0 0
2. H H H F COOCH3 F 6.287 H H 1.321 81.94 335.3 8.370 6.308 4.964 1 0 0
3. CH3 CH3 H F COOCH3 F 6.903 H H 1.266 91.17 363.3 9.091 6.503 5.052 0 0 0
4. CH2CH3 CH2CH3 H F COOCH3 F 5.962 H H 1.223 100.44 391.4 10.418 7.365 5.207 0 1 0
5. CF3 CF3 H F COOCH3 F 7.022 H H 1.474 92.09 471.3 11.715 8.962 5.401 0 0 0
6. CH3 H H F COOCH3 F 6.180 H H 1.290 86.53 349.3 8.773 6.416 5.011 0 0 0
7. H CH3 H F COOCH3 F 5.978 H H 1.290 86.53 349.3 8.773 6.416 5.011 1 0 0
8. CF3 H H F COOCH3 F 6.962 H H 1.404 86.99 403.3 9.855 7.186 5.213 0 0 0
9. H CF3 H F COOCH3 F 6.319 H H 1.404 86.99 403.3 9.855 7.186 5.213 1 0 0
10. i-Pr H H F COOCH3 F 7.145 H H 1.241 95.76 377.4 9.584 7.003 5.163 0 0 0
11. t-Bu H H F COOCH3 F 7.053 H H 1.222 100.40 391.4 9.855 7.186 5.213 0 0 0
12. i-Pr CH3 H F COOCH3 F 7.078 H H 1.222 100.40 391.4 10.371 7.011 5.183 0 0 0
13. CF3 CH3 H F COOCH3 F 7.610 H H 1.376 91.63 417.3 10.778 7.170 5.227 0 0 0
14. CH3 CF3 H F COOCH3 F 6.237 H H 1.376 91.63 417.3 10.778 7.170 5.227 0 0 0
15. CF3 CH3 H F COOCH3 F 7.610 H H 1.376 91.63 417.3 10.778 7.170 5.227 0 0 0
16. CF3 CH3 H F COOCH3 Cl 9.455 Cl N 1.449 104.15 483.2 11.404 8.155 5.662 0 0 1
17. CF3 CH3 CH3 F
ON
N
CH3Cl 9.102 Cl N 1.464 110.28 507.3 12.383 9.315 6.553 0 0 1
18. CF3 CH3 CH3 FO
N
N
CH3 F 9.180 Cl N 1.449 105.38 490.8 12.383 9.315 6.553 0 0 0
19. CF3 CH3 CH3 F NN
N
N
CH3 F 9.229 Cl N 1.560 107.85 490.8 12.383 9.315 6.533 0 0 0
20. CF3 CH3 CH3 ClN
N
N
N
CH3 F 9.366 Cl N 1.580 112.58 507.3 12.383 9.315 6.533 0 0 0
21. CF3 CH3 CH3 F
NO
N
CH3 F 9.180 Cl N 1.449 105.38 490.8 12.383 9.315 6.533 0 0 0
QSAR
based
modelin
gonaseries
ofa-hydroxyamides
asanovel
class
ofbradykinin
B1selectiv
eantagonists
217
Table 2 Correlation matrix between physicochemical parameters and indicator parameters.
pIC50 D MR MW 3v 4v 5v I1 I2 I3
pIC50 1.000
D 0.725 1.000
MR 0.790 0.475 1.000
MW 0.846 0.821 0.829 1.0003v 0.830 0.768 0.822 0.965 1.0004v 0.834 0.805 0.807 0.951 0.954 1.0005v 0.885 0.768 0.835 0.886 0.886 0.937 1.000
I1 �0.433 �0.121 �0.476 �0.396 �0.492 �0.382 �0.313 1.000
I2 �0.280 �0.312 0.114 �0.105 �0.026 �0.059 �0.117 �0.091 1.000
I3 0.481 0.269 0.406 0.435 0.331 0.323 0.326 �0.132 �0.073 1.000
Table 3 Comparison between observed and predicted activ-
ities and their residual values.
S. No. Observed Calculated Residual
1 7.228 6.994 0.234
2 6.287 6.046 0.241
3 6.903 6.850 0.053
4 5.962 5.962 0.000
5 7.022 7.377 �0.3556 6.180 6.788 �0.6087 5.978 6.117 �0.1398 6.962 7.093 �0.1319 6.319 6.422 �0.10310 7.145 7.018 0.127
11 7.053 7.093 �0.04012 7.078 7.048 0.030
13 7.610 7.114 0.496
14 6.237 7.114 �0.87715 7.610 7.114 0.496
16 9.455 8.606 0.849
17 9.102 9.951 �0.84918 9.180 9.117 0.063
19 9.229 9.087 0.142
20 9.366 9.087 0.279
21 9.180 9.087 0.093
218 A.K. Srivastava et al.
the data set and the activity is correlated using the rest of thedata set. The cross-validated R2 in each case was found to be
A plot showing comparison between using m
5
5.5
6
6.5
7
7.5
8
8.5
9
9.5
10
5 5.5 6 6.5 7Observ
Pred
icte
d pI
C50
Figure 1 A plot showing comparison between obse
very close to the value of R2 for the entire data set and hencethese models can be termed as statistically significant.
Cross validation provides the values of PRESS (predictedresidual sum of squares), SSY (sum of the squares of the re-
sponse) and R2cv (overall predicted ability) and PSE (predicted
squares error) from which we can test the predictive power ofthe proposed model. It is argued that PRESS, is a good esti-
mate of the real predictive error of the model and if it is smallerthan SSY the model predicts better than chance and can beconsidered statistically significant. Furthermore, the ratio
PRESS/SSY can be used to calculate approximate confidenceintervals of prediction of a new compound. To be a reasonableQSAR model PRESS/SSY should be smaller than 0.4. Also, ifPRESS value is transformed in a dimension less term by relat-
ing it to the initial sum of squares, we obtain R2cv, i.e. the com-
plement to the traces of unexplained variance over the totalvariance. The PRESS and R2
cv have good properties. However,
for practical purposes of end users the use of square root ofPRESS/N, which is called predictive square error (PSE), ismore directly related to the uncertainty of the predictions.
The PSE values also support our results. The calculatedcross-validated parameters confirm the validity of the models.All the requirements for an ideal model have been fulfilled by
model no. 1, that is why, we have considered it as the bestmodel.
R2A takes into account the adjustment of R2. R2
A is a mea-sure of the percentage explained variation in the dependent
variable that takes into account the relationship between the
observed and predicted pIC50 values odel 1
y = 0.8868x + 0.8467R2 = 0.8869
7.5 8 8.5 9 9.5 10ed pIC50
rved and predicted pIC50 values using model 1.
Table 4 Cross validated parameters and predictive error of coefficient of correlation (PE) for the proposed model.
S. No. Parameters used n PRESS SSY PRESS/SSY R2cv PSE R 1 � R2 PE 6PE
1. 5v + I1 + I2 + I3 21 3.492 27.366 0.128 0.872 0.089 0.942 0.113 0.016 0.096
2. 4v + I1 + I2 + I3 21 5.669 25.189 0.225 0.775 0.113 0.903 0.184 0.027 0.162
3. 3v + I1 + I2 + I3 21 6.116 24.742 0.247 0.753 0.118 0.895 0.198 0.029 0.174
4. MW+ I1 + I2 + I3 21 6.542 24.317 0.269 0.731 0.122 0.888 0.212 0.031 0.186
5. MR+ I1 + I2 + I3 21 6.563 24.295 0.270 0.730 0.122 0.887 0.213 0.031 0.186
6. D+ I1 + I2 + I3 21 8.425 22.434 0.376 0.624 0.138 0.853 0.273 0.040 0.240
QSAR based modeling on a series of a-hydroxy amides as a novel class of bradykinin B1 selective antagonists 219
number of cases and the number of independent variables inthe regression model, whereas R2 will always increase whenan independent variable is added. R2
A will decrease if the addedvariable does not reduce the unexplained variable enough to
offset the loss of decrease of freedom.
3.2. Predictive error of coefficient of correlation (PE)
The predictive error of coefficient of correlation (PE)(Chatterjee et al., 2000) is yet another parameter used to decide
the predictive power of the proposed models. We havecalculated PE value of all the proposed models and they arereported in Table 4. It is argued that if
(i) R< PE, then correlation is not significant;(ii) R> PE; several times (at least three times), then corre-
lation is indicated; and if
(iii) R> 6PE, then the correlation is definitely good.(iv) For all the models developed the condition R> 6PE is
satisfied and hence they can be said to have a good pre-
dictive power.
4. Conclusions
From the results and discussion made above, it may be con-
cluded that:
(1) Bulkier groups having more branchings should be usedin future drug designing.
(2) The group Cl at R6 position favours the inhibitoryactivity.
(3) Groups H and –CH2CH3 at R1 and R2 position respec-
tively, should be avoided in future drug modeling.
References
ACD – LAB Software for calculating the referred physiochemical
parameters. Chem Sketch www.acdlabs.com/acdlabs-rss-feed.xml.
Bauer, N., Fajars, K., Lewin, S.Z., 1960. In technique of organic
chemistry. In: Weissberger, A. (Ed.), Vol. 1, Part, and third ed.
Interscience, New York, pp. 1139.
Carmer III, R.D., Bunce, J.D., Patterson, D.E., Frank, I.E., 1988.
Quant. Struct. Act. Relat. 7, 18.
Carton, M., 1983. Steric effects in drug design. In: Charton, M.,
Motoc, I. (Eds.), Springer, Berlin, 57.
Chatterjee, S., Hadi, A.S., Price, B., 2000. Regression Analysis by
Example, third ed. Wiley VCH, New York.
Diudea, M.V., 2000. QSPR/QSAR studies for molecular descriptors,
Eds Nova Science. Huntingclon, New York.
Hansch, C., Leo, A., 1995. Exploring QSAR – Fundamentally and
Application in Chemistry and Biology. Amer. Chem. Soc., Wash-
ington, DC.
Hansch, C., Leo, A., Unger, S.H., Kim, K.H., Nikaitani, D., Lien, J.J.,
1973. Med. Chem. 161, 1207.
Hansch, C., Leo, A., Hochman, D., 1995. Exploring QSAR –
Hydrophobic, Electronic and Steric Constant. Amer. Chem. Soc.,
Washington, DC.
Holder, J., Lin, Ye., David, E.J., Cecil, C., Chappelow, C.C., 2006.
QSAR Comb. Sci. 25 (10), 905–911.
Ismail, Y., Llkay, O., Ozlem, Esin A.S., 2002. Acta Biochem Pol. 47
(2), 481.
Liu, S., Zhang, R., Liu, M., Hu, Z., 1997. J. Chem. Infcomput. Sci. 37,
1146–1151.
Podloga, B.L., Ferguson, D.M., 2000. Drug Des. Discov. 17, 4.
Pogliani, L., 1994. Structural property relationships of amine acids and
some Peptides. Amino Acids 6, 141.
Pogliani, L., 1996. Modeling with special descriptors derived from a
medium size set of connectivity indices. J. Phys. Chem. 100, 18065.
Randic, M., 1975. J. Am. Chem. Soc. 97, 6609–6615.
Randic, M., 1993. J. Am. Chem. Phys. Lett. 211, 478–483.
Senda, M., Fujita, T., 1991. Chem. Pharma. Bull. 39, 1736.
Singh, J., Dubey, V.K., Agarwal, V.K., Khadikar, P.V., 2008a. QSAR
study on octanol–water partitioning: dominating role of equalized
electronegativity. Oxid. Commun. 31 (1), 27–43.
Singh, J., Agarwal, V.K., Singh, S., Khadikar, P.V., 2008b. Use of
topological as well as quantum chemical parameters in modeling
antimalarial activity of 2,4-diamino-6-quinazoline sulphonamides.
Oxid. Commun. 31 (1), 17–26.
Srivastava, A.K., Archana, Jaiswal M., 2003. Oxid. Commun. 31(1),
44–51.
Srivastava, A.K., Archana, Jaiswal M., 2008a. Quantitative structure–
activity relationship studies of q-arylthiocinnamides as antagonists
of biochemical ICAM- 1/LFA-1 interaction in relation to antiin-
flammatory activity. Oxid. Commun. 31, 44–51.
Srivastava, A.K., Archana, Jaiswal M., 2008b. And QSAR studies on
indole substituted potent human histamine H4 antagonists: role of
physicochemical and statistical parameters. J. Saudi Chem. Soc. 12,
221–226.
Srivastava, A.K., Archana, Jaiswal M., 2008c. Exploring QSAR of
selective PDE4 inhibitors as 8-substituted analogues of 3-(3-
cyclopentyloxy-4-methoxy-benzyl)-8 isopropyl adenine. J. Saudi
Chem. Soc. 12, 227–232.
Srivastava, A.K., Srivastava, A., Archana, Jaiswal M., 2008. Role of
physico-chemical parameters in quantitative structure-activity
relationship based modeling of CYP26A1 inhibitory activity. J.
Indian Chem. Soc. 85, 721–727.
Srivastava, A.K., Archana, Jaiswal M., Srivastava, A., 2009. QSAR
modelling of selective CC chemokine receptor 3 (CCR3) antago-
nists using physicochemical parameters. Oxid. Commun. 32, 55–61.
Srivastava, A.K., Srivastava, A., Archana, Jaiswal M., Nath, A.,
2009a. QSAR studies on anti-apoptotic Bcl-2 protein inhibitors. J.
Saudi Chem. Soc. 13, 259–262.
Srivastava, A.K., Pandey, A., Nath, A., Chaurasia, Sh., 2009b. QSAR
based modeling of inhibitory activity of alkenyldiarylmethane
derivatives. J. Saudi Chem. Soc. 13, 263–267.
220 A.K. Srivastava et al.
Srivastava, A.K., Archana, Jaiswal M., Srivastava, A., 2009c.
QSAR modelling of selective CC chemokine receptor 3
(CCR3) antagonists using physicochemical parameters. Oxid.
Commun. 32, 55–61.
Srivastava, A.K., Pathak, V.K., Srivastava, A., Pandey, A., 2010.
Molecular modeling of 8-methoxy quinolone analogues by using
quantitative structure activity relationship. J. Saudi Chem. Soc. 14,
217–222.
Wood, M.R., Schirripa, K.M., Kim, J.J., Kuduk, S.D., Chang, R.K.,
Marco C.N. Di, DiPardo, R.M., Wan, B.-L., Murphy, K.L.,
Ransom, R.W., Chang, R.S.L., Holahan, M.A., Cook, J.J.,
Lemaire, W., Mosser, S.D., Bednar, R.A., Tang, C., Thomayant,
P., Wallace, A.A., Mei, Q., Yu, J., Bohn, D.L., Clayton, F.C.,
Adarayn, E.D., Sitko, G.R., Leonard, Y.M., Freidinger, R.M.,
Pettibone, D.J., Bock, M.G., 2008. Bioorg. Med. Chem. Lett. 18,
716–720.