QSAR modeling of antiradical and antioxidant activities of flavonoids using electrotopological state...
-
Upload
supratim-ray -
Category
Documents
-
view
213 -
download
1
Transcript of QSAR modeling of antiradical and antioxidant activities of flavonoids using electrotopological state...
DOI: 10.2478/s11532-007-0047-3Research article
CEJC 5(4) 2007 1094–1113
QSAR modeling of antiradical and antioxidantactivities of flavonoids using electrotopological state
(E-State) atom parameters
Supratim Ray, Chandana Sengupta and Kunal Roy∗
Drug Theoretics & Cheminformatics Lab,Division of Medicinal & Pharmaceutical Chemistry,
Department of Pharmaceutical Technology,Jadavpur University, Kolkata 700 032, India
Received 10 July 2007; accepted 13 August 2007
Abstract: In the present paper QSAR modeling using electrotopological state atom (E-state)parameters has been attempted to determine the antiradical and the antioxidant activities of flavonoidsin two model systems reported by Burda et al. (2001). The antiradical property of a methanolic solutionof 1, 1-diphenyl-2-picrylhydrazyl (DPPH) and the antioxidant activity of flavonoids in a β-carotene-linoleic acid were the two model systems studied. Different statistical tools used in this communicationare stepwise regression analysis, multiple linear regressions with factor analysis as the preprocessingstep for variable selection (FA-MLR) and partial least squares analysis (PLS). In both the activities thebest equation is obtained from stepwise regression analysis, considering, both equation statistics andpredictive ability (antiradical activity: R2 = 0.927, Q2 = 0.871 and antioxidant activity: R2 = 0.901,Q2 = 0.841).c© Versita Warsaw and Springer-Verlag Berlin Heidelberg. All rights reserved.
Keywords: Flavonoids, antioxidant, QSAR
1 Introduction
Flavonoids are a group of naturally occurring polyphenolic compounds ubiquitously found
in fruits and vegetables [1–3]. Common family members of flavonoids include flavones,
flavonols, flavanones, isoflavones, biflavanones, catechins and anthocyanidins. Flavonoids
are a chemical class of benzo-γ-pyrone derivatives. The structural differences in each
flavonoid family result from the variation in the number and substitution patterns of
∗ E-mail: kunalroy [email protected]
S. Ray et al. / Central European Journal of Chemistry 5(4) 2007 1094–1113 1095
the hydroxyl groups and the extent of glycosylation of these groups [4]. Structural di-
versity of flavonoids allows them to exhibit antineoplastic, antihepatitis, antibacterial,
anti-inflammatory, antimutagenic, antiallergic, antithrombic, antiviral and vasodilatory
activities [5–7]. The potent antioxidant activity of flavonoids, their ability to scavenge hy-
droxyl radicals, superoxide anions and lipid peroxy radicals could be their most important
functions that underlie many of the above processes in the body [8].
Oxidative damage is implicated in many disease conditions as is shown in epidemiolog-
ical, clinical and laboratory research on flavonoids and other antioxidants. Quercetin, one
of the potent antioxidant biflavonoid compounds, has been shown to alleviate oxidative
injury by modulation of gene expression leading to suppression of production of reactive
oxygen and nitrogen species conferring an antiapoptotic activity [9]. It has been reported
recently that fisetin, chrysin, morin, 3-hydroxyflavone have antioxidant and antihemolytic
effects [10]. Galangin, its derivatives viz. quercetin, morin, myricetin, oligomeric proan-
thocyanidins and flavanones have been shown to inhibit the oxidation of lipids hence can
be utilized in prevention and treatment protocols for inflammatory conditions like cancer,
asthma, liver disease, cardiovascular disease and muscular degeneration [11, 12]. Intake
of antioxidant compounds present in food is an important factor in protecting health.
Flavonoids, which occur both in edible plants and in foodstuffs derived from plants as
in fruits, vegetables, red wine and tea, form substantial constituents of the human diet.
Therefore, flavonoids given as additives in foodstuffs have the potential to be applied to
the prevention and treatment of human diseases [13].
Flavonoids and its derivatives account for more than 4000 in number, further, the
antioxidant properties of each are very different. It is a complex task to select the
most effective of antioxidants from this large pool of flavonoids [14]. These character-
istics make flavonoids popular subjects for Quantitative structure-activity relationship
(QSAR) studies. Khlebnikov et al. [15] reported a QSAR model to predict antioxidant
activity of flavonoids in chemical, enzymatic and cellular systems using physicochemical
and structural descriptors. A QSAR study was performed on flavonoids derivatives as
p561ck tyrosinekinase inhibitor using hydration energy and logP as a predictor parame-
ter [16]. Badhan et al. [17] computed in silico modeling of the interaction of flavonoids
with human-p-glycoprotein nucleotide-binding domain. Karawajczyk et al. [18] reported
a QSAR model by neural network modeling to investigate the properties of flavonoids
influencing the binding to bilitransloase. A density functional study of flavonoids com-
pounds with anti-HIV activity was done using the multiple linear regression method [19].
QSAR for flavonoids-mediated inhibition of breast cancer resistance protein was done us-
ing three structural descriptors [20]. A study involving multiple linear regression (MLR)
and partial least squares (PLS) regression applied on flavonoid compounds with anti-HIV
activity have been reported [21]. Structure-radical scavenging activity relationships of
flavonoids were performed using simple indicator variables by MLR method [22]. Farkas
et al. [14] reported quantitative structure-antioxidant activity relationships of flavonoids
using constitutional, topological and connectivity indices by PLS method. In the present
paper, we have modeled the antiradical and antioxidant activities of flavonoids data set
1096 S. Ray et al. / Central European Journal of Chemistry 5(4) 2007 1094–1113
reported by Burda et al. [23] using electrotopological state atom (E-state) parameters by
stepwise regression, factor analysis followed by multiple linear regression (FA-MLR) and
PLS.
2 Computational
2.1 Electrotopological state atom (E-state) index
Structural specificity of a drug molecule is exhibited at an atomic or fragmental level
rather than at the molecular level. In the drug-receptor interaction phenomenon, a por-
tion of the molecule (pharmacophore) may play a more important role than the other
segments. Though basic information for constitution of topological indices are derived
from the atomic level such as the count of atoms, bonds, paths of bonds, etc., most of
the indices are applied to the entire molecule after summation of all components over the
whole molecule. Thus QSAR studies at the atomic or fragmental level are justified in the
present context [24].
The electrotopological state (E-state) atom index developed by Hall and Kier [25] is an
atom level descriptor encoding both the electronic character and topological environment
of each skeletal atom in a molecule. The E-state of a skeletal atom is formulated as an
intrinsic value Ii plus a perturbation term ΔIi, arising from the electronic interaction
within the molecular topological environment of each atom in the molecule.
The intrinsic value has been defined as the ratio of a measure of electronic state (Kier-
Hall valence state electronegativity) to the local connectivity. The number of the most
reactive valence electrons that are involved in chemical reactions and bond formations are
considered in the expression of I to encode the electronic feature. To reflect differences
in electronegativity among the atoms, principal quantum number is employed in the
expression of I. The topological attribute is accounted for by using adjacency count of
atom. The intrinsic value of an atom i is defined as
I = [(2/N)2δv + 1]/δ (1)
In Eq. (1), N stands for principal quantum number, δv and δ indicate the count of va-
lence electrons and sigma electrons, respectively, associated with the atom i in the hydro-
gen suppressed graph. The intrinsic electrotopological state calculated according to Eq.
(1) produces different values for an atom in different degrees of substitution (branching).
The values are also different for different atoms having differences in electronegativity.
The intrinsic values increase with increase in electronegativity or electron-richness and
decrease with increase in branching (substitution).
The perturbation factor for the intrinsic state of atom i is defined as
ΔIi =∑j �=i
Ii − Ij
r2ij
(2)
S. Ray et al. / Central European Journal of Chemistry 5(4) 2007 1094–1113 1097
In Eq. (2) rij stands for the graph separation factor, i.e., count of skeletal atoms in
the shortest path connecting the atoms i and j (including both atoms).
Summation of intrinsic state of an atom and influence of the field is called electro-
topological state of the atom.
Si = Ii +∑j �=i
ΔIij (3)
It is a representation of molecular structure information as it varies with changes in
structural features including branching, cyclicity, homologation, heteroatom variation,
and changes in relative positions of different groups. The electrotopological state consid-
ers both bonded and non-bonded interactions. The bonded component depends simply
on differences in electronegativity among the adjacent atoms. The non-bonded inter-
actions may be through inductive effect across the skeleton and is a function of graph
separation factor and electronegativity differences. Thus, electrotopological state repre-
sents electronic distribution information modified by both local and global topology. The
information encoded in the E-state value for an atom is the electronic accessibility at that
atom.
The E-state index has been projected as a useful tool in the context of QSAR studies
and reported to have the power to identify atoms or fragments in the molecules that are
important for the biological activity [26–28]. We have in our earlier studies, also used
E-state parameters to explore QSAR of ligands acting on pharmacologically relevant
targets of contemporary interest [29–37]. In continuation with these efforts, the present
communication will show the utility of E-state parameters in QSAR studies by exploring
QSAR of antiradical properties of flavonoids as studied in a methanolic solution of 1,
1-diphenyl-2-picrylhydrazyl (DPPH) [38] and antioxidant activity of flavonoids in a β-
carotene-linoleic acid model system [39] as reported by Burda et al. [23].
2.2 Methods
The equations were developed using 1) stepwise regression, 2) factor analysis followed by
multiple linear regression (FA-MLR) and 3) partial least squares analysis.
2.2.1 Stepwise Regression
In stepwise regression [40], a multiple term linear equation was built step-by-step. The
basic procedure involved (1) identifying an initial model, (2) iterative “stepping”, i.e.,
repeatedly altering the model of the previous step by adding or removing a predictor
variable in accordance with the “stepping criteria”, (F = 4 for inclusion; F = 3.9 for
exclusion in our case) and (3) terminating the search when stepping was no longer possible
as per the stepping criteria, or when a specified maximum number of steps was reached.
At each step all variables were reviewed and evaluated to determine which one would
contribute most to the equation. That variable was then included in the model, and the
process started again. A limitation of the stepwise regression search approach is that
1098 S. Ray et al. / Central European Journal of Chemistry 5(4) 2007 1094–1113
it presumes that there is a single “best” subset of X variables and seeks to identify it.
There is often no unique “best” subset, and all possible regression models with a similar
number of X variables as in the stepwise regression solution should be retro-fitted to check
whether some other subsets of X variables might be better.
2.2.2 FA-MLR
In the case of FA-MLR [41, 42], a classical approach to the MLR technique was used as
the final statistical tool for developing classical QSAR relations and FA was used in the
data-preprocessing step to identify the important descriptors contributing to the response
variable and avoid collinearities among them. In a typical FA procedure, the data matrix
is first standardized, and subsequently, the reduced correlation matrix is constructed. An
eigenvalue problem is then solved and the factor pattern can be obtained from the cor-
responding eigenvectors. The principal objectives of FA are to display multidimensional
data in a space of lower dimensionality with minimum loss of information (explaining
>95% of the variance of the data matrix) and to extract the basic features behind the
data with the ultimate goal of interpretation and/or prediction. FA was performed on the
data set(s) containing biological activity and all descriptor variables, which were to be
considered. The factors were extracted by principal component method and then rotated
by VARIMAX rotation (a kind of rotation which is used in principal component analysis
so that the axes are rotated to a position in which the sum of the variances of the loadings
is the maximum possible) to obtain Thurston’s simple structure. The simple structure is
characterized by the property that as many variables as possible fall on the coordinate
axes when presented in common factor space, so that the largest possible number of factor
loadings becomes zero; this is done to obtain numerically comprehensive picture of the
relatedness of the variables. Only variables with non-zero loadings in such factors where
biological activity also has non-zero loading were considered important in explaining the
variance of the activity. Further, variables with non-zero loadings in different factors were
combined in a multivariate equation.
2.2.3 PLS
PLS is a generalization of regression, which can handle data with strongly correlated
and/or noisy or numerous X variables [43, 44]. It gives a reduced solution, which is
statistically more robust than MLR. The linear PLS model finds “new variables” (latent
variables or X scores) which are linear combinations of the original variables. To avoid
overfitting, a strict test for the significance of each consecutive PLS component is nec-
essary and then stopping when the components are nonsignificant. Application of PLS
thus allows the construction of larger QSAR equations while still avoiding overfitting and
eliminating most variables. PLS is normally used in combination with crossvalidation to
obtain the optimum number of components. This ensures that the QSAR equations are
selected based on their ability to predict the data rather than to fit the data. In case of
PLS analysis on the present data set, based on the standardized regression coefficients,
the variables with smaller coefficients were removed from the PLS regression until there
S. Ray et al. / Central European Journal of Chemistry 5(4) 2007 1094–1113 1099
was no further improvement in Q2 value irrespective of the components.
2.3 Data treatment and software
The antiradical activities of flavonoid [23] compounds were converted to logarithmic scale
(pC) (Table 1) and then used for subsequent QSAR analyses as the response variables.
The antioxidant activities of flavonoid [23] compounds (Table 3) were used as % antioxi-
dant activity, AAFL as reported, for subsequent QSAR analyses as the response variable.
For antiradical activity, 29 compounds (set 1) and for antioxidant activity 24 compounds
(set 2) were considered in the present study. All the compounds (sets 1 and 2) contain 17
common atoms (excluding hydrogens). The atoms were numbered keeping serial numbers
of the common atoms same in all the compounds (as shown in Fig. 1). The electrotopo-
logical states of the 17 common atoms for all of the compounds (set 1 and set 2) were
determined using a GW-BASIC program ELECTRO1 developed in our group [45]. The
program uses only the connection table in a specific format along with values for intrin-
sic state of different atoms as input. To the output file obtained, the biological activity
data was introduced to make it ready for subsequent regression analysis. The stepwise
regression and FA and MLR were performed using the statistical software SPSS [46]. PLS
was performed using statistical software MINITAB [47]. The statistical qualities of the
MLR equations [48] were judged by the following parameters: variance (R2a), correlation
coefficient (R), standard error of estimate (s) and variance ratio (F ) at specified degree
of freedom (df). If not stated otherwise, all accepted MLR equations have regression co-
efficients and F ratio significant at 95 and 99% levels, respectively,. The generated QSAR
equations were validated by leave-one-out or LOO method [49, 50] using MINITAB soft-
ware [47] and the calculated parameters are predicted residual sum of squares (PRESS),
standard deviation based on PRESS (SPRESS), standard deviation of error of prediction
(SDEP) [50] and cross validation R2 (Q2). Q2 is calculated according to the following
formula
Q2 = 1 −∑
(Yobs − Ycal)2∑
(Yobs − Y )2(4)
In Eq. (4), Y means average activity value of the entire data set while Yobs and Ycal
represent observed and LOO estimated activity values. Standard deviation of error of
prediction (SDEP) [51] is calculated according to the formula
SDEP =
√PRESS
n(5)
In Eq. (5) PRESS is the predicted residual sum of squares using (leave-one-out)
statistics and n is the number of components.
1100 S. Ray et al. / Central European Journal of Chemistry 5(4) 2007 1094–1113
Fig. 1 General structure of flavonoids: the common atoms have been numbered 1-17.
3 Results and Discussion
3.1 QSAR of antiradical activity of flavonoids
The results obtained from different statistical methods are described below and the in-
terpretations of the equations are also depicted.
3.1.1 Stepwise regression
Using stepping criteria based on F value (F = 4.0 for inclusion; F = 3.9 for exclusion),
different equations were derived after successive addition of E-state parameters.
pC = −0.750(±0.079)S3 + 1.287(±0.073)
n = 29, R2 = 0.770, R2a = 0.762, Q2 = 0.773, R = 0.878
s = 0.329, PRESS = 4.804, F = 90.40, (df1.27), SDEP = 0.407, SPRESS = 0.422
(6)
Using single best parameter (S3), Eq. (6) could explain and predict 76.2% and 73.3%
respectively of the antiradical scavenging potency.
pC = −0.628(±0.086)S3 − 0.732(±0.281)S10 + 1.343(±0.069)
n = 29, R2 = 0.818, R2a = 0.804, Q2 = 0.776, R = 0.904
s = 0.355, PRESS = 4.212, F = 58.24, (df2.26), SDEP = 0.381, SPRESS = 0.402
(7)
On inclusion of S10 along with S3, it is observed that there is significant increase
in the statistical quality of the equation (explained and predicted variance of 80.4% and
76.6% respectively of antiradical scavenging potency). The best trivariate and tetravariate
equations are given below.
pC = −0.895(±0.111)S3 − 0.800(±0.243)S10 + 0.699(±0.219)S2 + 1.428(±0.065)
n = 29, R2 = 0.870, R2a = 0.855, Q2 = 0.805, R = 0.933
s = 0.305, PRESS = 3.513, F = 55.96, (df3.25), SDEP = 0.348, SPRESS = 0.375
(8)
S. Ray et al. / Central European Journal of Chemistry 5(4) 2007 1094–1113 1101
Table
1C
hem
ical
stru
cture
san
dob
serv
edan
dca
lcula
ted
free
radic
alsc
aven
ging
pot
ency
ofth
ese
lect
edflav
onoi
ds.
Com
p.
Com
pound
R1
R2
R3
R4
R5
R6
R7
R8
C2
pC
No.
Nam
e=
C3
Obsa
Calb
A1
Mori
nO
HO
HO
HH
OH
HO
HH
+1.9
84
1.9
15
A2
Taxifolin
OH
OH
OH
HH
OH
OH
H−
1.9
76
1.9
72
A3
Kaem
pfe
rol
OH
OH
OH
HH
HO
HH
+1.9
70
1.9
08
A4
Fust
inO
HH
OH
HH
OH
OH
H−
1.9
63
2.0
60
A5
Gala
ngin
OH
OH
OH
HH
HH
H+
1.9
62
1.8
80
A6
Ruti
nO
-GR
OH
OH
HH
OH
OH
H+
1.9
58
1.8
52
A7
Quer
ceti
nO
HO
HO
HH
HO
HO
HH
+1.9
53
1.9
32
A8
Lute
olin-7
-O-g
luH
OH
O-g
luH
HO
HO
HH
+1.9
42
1.7
33
A9
Quer
ceti
n-3
-O-g
lu-7
-O-r
ha
O-g
luO
HO
-rha
HH
OH
OH
H+
1.9
38
1.8
71
A10
Lari
cytr
inO
HO
HO
HH
HO
HO
HO
CH
3+
1.9
27
1.9
44
A11
Laricy
trin
-3’-O
-glu
OH
OH
OH
HH
O-g
luO
HO
CH
3+
1.9
23
2.0
24
A12
Robin
etin
OH
HO
HH
HO
HO
HO
H+
1.9
15
2.0
50
A13
Fiset
inO
HH
OH
HH
OH
OH
H+
1.8
97
2.0
25
A14
Myrice
tin
OH
OH
OH
HH
OH
OH
OH
+1.8
62
1.9
56
A15
Kaem
pfe
rol
O-r
ha
OH
O-r
ha
HH
HO
HH
+1.8
48
1.8
31
3,7
-di-rh
aA
16
3-h
ydro
xy
OH
HH
HH
HH
H+
1.8
19
1.4
95
flav
one
A17
Apig
enin
-7-O
-glu
HO
HO
-glu
HH
HO
HH
+1.5
41
1.7
09
A18
Hes
per
etin
HO
HO
HH
HO
HO
CH
3H
−1.4
77
0.8
59
A19
Vitex
in3,5
,7,3
’,H
OH
OH
O-g
luH
HO
HH
+1.3
22
1.0
56
A20
4’,5’-hex
am
ethox
yO
CH
3O
CH
3O
CH
3H
HO
CH
3O
CH
3O
CH
3+
1.1
00
1.3
72
flav
one
A21
Nari
ngen
inH
OH
OH
HH
HO
HH
−0.7
99
0.8
68
A22
Nari
ngin
HO
HO
-neo
hes
pH
HH
OH
H−
0.6
72
1.0
16
A23
7-h
ydro
xy
HH
OH
HH
HH
H+
0.4
47
0.3
16
Fla
vone
A24
Fla
vanone
HH
HH
HH
HH
−0.4
14
0.4
71
A25
Fla
vone
HH
HH
HH
HH
+0.1
76
−0.1
51
A26
Chry
sin
HO
HO
HH
HH
HH
+0.0
41
0.2
09
A27
Apig
enin
HO
HO
HH
HH
OH
H+
−0.1
54
0.2
27
A28
8-m
ethox
yH
HH
OC
H3
HH
HH
+−0
.154
0.1
41
flav
one
A29
5-h
ydro
xy
HO
HH
HH
HH
H+
−0.2
21
−0.2
54
flav
one
Glu
=glu
cose
,R
ha
=rh
am
nose
,N
eohes
p=
neo
hes
per
idin
,G
R=
β-D
-Glu
-α-L
-Rha;T
hes
esu
gars
are
connec
ted
via
thei
rglu
cosi
de
OH
gro
up
toth
eflav
onoid
saTaken
from
Ref
.[2
3];
bFro
mE
q.
(10).
1102 S. Ray et al. / Central European Journal of Chemistry 5(4) 2007 1094–1113
pC = −1.393(±0.197)S3 − 2.537(±0.633)S10 + 1.346(±0.294)S2 + 2.009(±0.690)S4
+ 2.725(±0.449)
n = 29, R2 = 0.904, R2a = 0.888, Q2 = 0.837, R = 0.951
s = 0.268, PRESS = 2.932, F = 56.63, (df4.24), SDEP = 0.318, SPRESS = 0.349(9)
From Eqs. (8) and (9) it is observed that successive addition of S2 and S4 increases
the explained variance (85.5% and 88.8% respectively) and predicted variance (80.5%
and 83.7% respectively) of antiradical scavenging potency of flavonoids. The parameter
S2 implies the impact of the phenyl ring at position 2. Presence of electron withdrawing
groups on the phenyl ring increases the value of S2 activity in comparison to a phenyl ring,
which is unsubstituted or substituted with other groups. Again S4 implies the importance
of the keto group at position 4. Van Acker et al. [52] also reported the presence of 4-keto
functionality as a structural requirement for free radical scavenging activity of flavonoids.
The best pentavariate equation is given below.
pC = −1.378(±0.176)S3 − 3.432(±0.656)S10 + 1.359(±0.262)S2
+ 2.231(±0.621)S4 + 0.236(±0.888)S6 + 2.884(±0.405)
n = 29, R2 = 0.927, R2a = 0.911, Q2 = 0.871, R = 0.963
s = 0.238, PRESS = 2.328, F = 58.48, (df5.23), SDEP = 0.283, SPRESS = 0.318
(10)
The standard errors of the respective E-state indices are mentioned within parentheses.
From the above equations (6–10) it was observed that there is significant gradual increase
in the values of R2 (0.770 to 0.927), R2a (0.762 to 0.911) and Q2 (0.733 to 0.871) after
gradual addition of parameters. The final Eq. (10) could explain and predict 91.1%
and 87.1% respectively of the variance of antiradical scavenging potency. The positive
coefficient of S2, S4 and S6 indicate that activity increases with increase in E-state value
of atoms 2, 4 and 6 respectively while the negative coefficient of S3 and S10 indicate that
activity decreases with increase in E-state value of atoms 3 and 10 respectively. The
values of E-state parameters of the compounds present in the final equations and the
intercorrelation among the E-state parameters for antiradical activity data modeling are
shown in Table 1A and 2A respectively in the supplementary materials section. Table 2A
indicates that some of the E-state parameters used in the stepwise regression equations
are highly intercorrelated. However, as the leave-one-out prediction statistics of these
equations are very good (for example Eq.(10), Q2 = 0.871), we have accepted these
equations with a note of the intercorrelation among descriptors.
3.1.2 FA-MLR
From the factor analysis on the data matrix consisting of free radical scavenging potential
data (pC) of flavonoids and E-state indices, it was observed that 7 factors could explain
the data matrix to the extent of 95.8%. The free radical scavenging potency is highly
loaded with factor 3 (loaded in S3, S4) and moderately loaded in factor 2 (loaded in
S. Ray et al. / Central European Journal of Chemistry 5(4) 2007 1094–1113 1103
S14, S16 and S17) and factor 4 (loaded in S2). Based on the results of FA (Table 2), the
following equation with two variables was derived:
pC = 0.635(±0.256)S2 − 1.003(±0.125)S3 + 1.360(±0.073)
n = 29, R2 = 0.814, R2a = 0.800, Q2 = 0.761, R = 0.902
s = 0.358, PRESS = 4.297, F = 56.89, (df2.26), SDEP = 0.384, SPRESS = 0.406
(11)
Table 2 Factor loading of the variables (antiradical inhibition potency (pC) and predictor
variables) after VARIMAX rotation.
F1 F2 F3 F4 F5 F6 F7 Communalities
pC −0.258 −0.354 −0.768 −0.340 −0.019 −0.075 −0.166 0.931S1 0.202 0.256 0.262 0.025 0.879 0.041 0.195 0.988S2 0.234 0.226 0.224 0.889 −0.006 0.014 0.108 0.958S3 0.197 0.373 0.650 0.577 0.056 −0.122 0.163 0.977S4 0.590 0.449 0.574 0.073 0.000 0.141 0.263 0.974S5 0.827 0.298 0.350 0.115 −0.094 0.181 0.198 0.990S6 0.954 0.074 0.053 0.061 −0.126 −0.001 0.076 0.945S7 0.896 0.217 0.171 0.206 0.083 0.220 0.132 0.994S8 0.765 0.311 0.101 0.319 0.090 0.247 0.073 0.868S9 0.337 0.086 0.027 −0.008 −0.019 0.927 0.057 0.985S10 0.773 0.360 0.297 0.150 −0.106 0.326 0.201 0.995S11 −0.344 −0.193 −0.226 −0.023 0.875 −0.069 −0.091 0.987S12 0.426 0.650 0.456 0.153 −0.003 0.139 0.375 0.995S13 0.320 0.359 0.281 0.222 0.137 0.081 0.775 0.985S14 0.175 0.751 0.267 0.401 −0.023 0.036 0.165 0.856S15 0.458 0.638 0.151 0.426 −0.091 0.161 0.263 0.924S16 0.171 0.931 0.146 0.062 0.070 0.025 0.032 0.929S17 0.417 0.718 0.351 0.205 0.065 0.157 0.308 0.978% Variance 0.282 0.216 0.127 0.104 0.090 0.069 0.068 0.958
The standard errors of the respective E-state indices are mentioned within parenthesis.
Eq. (11) could explain and predict 80.0% and 76.1% respectively of the variance in the
antiradical scavenging potency. The negative coefficient of S3 indicates that activity
decreases with an increase in E-state value of atom 3. The parameter S3 indicates the
importance of hydroxyl group at position 3 for antiradical activity. The parameter S2
implies the impact of phenyl group at position 2 and its substituents on the biological
activity. Presence of hydroxyl group on the phenyl ring increases the activity.
3.1.3 PLS
The number of optimum components optimized by crossvalidation was 2 to obtain the
final equation. Based on the standardized regression coefficients, the following variables
were selected for the final equation:
pC = 0.137S6 − 0.328S5 − 0.782S4 − 0.521S3 + 0.845
n = 29, R2 = 0.826, R2a = 0.813, Q2 = 0.783, R = 0.908
PRESS = 3.90, F = 61.81, (df2.26), SDEP = 0.367, SPRESS = 0.387
(12)
1104 S. Ray et al. / Central European Journal of Chemistry 5(4) 2007 1094–1113
Eq. (12) could explain and predict 81.3% and 78.3% respectively of the variance of
the antiradical activity. The positive coefficient of the variable S6 indicates that the anti-
radical activity increases with increase in the E-state value of atom 6 while the negative
coefficients of S5, S4, and S3 indicate that the activity decreases with increase in E-state
values of atoms 5, 4 and 3. However, S4 showed positive coefficients in Eqs. (9) and (10).
The positive coefficients of S4 in eqs. (9) and (10) may be due to its high intercorrelation
with S10 (r = 0.897).
3.2 QSAR of antioxidant activity of flavonoids
The results obtained from different statistical methods are described below and the in-
terpretations of the equations are also depicted.
3.2.1 Stepwise regression
Using stepping criteria based on F value (F = 4.0 for inclusion; F = 3.9 for exclusion),
different equations were derived after successive addition of E-state indices.
AAFL = −104.95(±23.35)S1 + 608, 81(±131.45)
n = 24, R2 = 0.479, R2a = 0.455, Q2 = 0.381, R = 0.692
s = 25.70, PRESS = 17258.3, F = 20.19, (df1.22), SDEP = 26.81, SPRESS = 28.01(13)
On using single best parameter (S1), Eq. (13) could explain and predict 45.5% and
38.1% respectively of the antioxidant activity. The parameter S1 implies the importance of
oxygen atom at position 1. The parameter S1 has negative contribution to the antioxidant
activity.
AAFL = −16.59(±5.10)S3 − 85.27(±20.41)S1 + 496.73(±115.05)
n = 24, R2 = 0.653, R2a = 0.620, Q2 = 0.564, R = 0.808
s = 21.46, PRESS = 12147.0, F = 19.76, (df2.21), SDEP = 22.49, SPRESS = 24.05(14)
AAFL = −14.71(±4.66)S6 − 24.31(±4.93)S3 − 79.59(±17.19)S1 + 463.21(±96.91)
n = 24, R2 = 0.768, R2a = 0.734, Q2 = 0.691, R = 0.877
s = 17.97, PRESS = 8605.45, F = 22.11, (df3.20), SDEP = 18.93, SPRESS = 20.74(15)
From Eqs. 14 and 15 it is observed that inclusion of S6 and S3 along with S1 increase
the explained variance (62.0 and 73.4% respectively) and predicted variance (56.4 and
69.1% respectively) of the antioxidant activity of flavonoids. The parameter S6 implies
the effect of hydroxyl substituent at position 6. Presence of hydroxyl group increases the
activity.
S. Ray et al. / Central European Journal of Chemistry 5(4) 2007 1094–1113 1105
The best tetravariate equation is given below.
AAFL = 16.86(±3.35)S14 + 13.40(±3.14)S6 − 38.95(±4.40)S3
− 61.05(±12.12)S1 + 348.43(±68.99)
n = 24, R2 = 0.901, R2a = 0.888, Q2 = 0.841, R = 0.949
s = 12.07, PRESS = 4436.55, F = 43.07, (df4.19), SDEP = 13.59, SPRESS = 15.28(16)
The standard errors of the respective E-state parameters are mentioned within parenthe-
ses. From the above equations (13–16) it was observed that there is significant gradual
increase in the values of R2 (0.479 to 0.901), R2a (0.455 to 0.888) and Q2 (0.381 to 0.841)
after gradual addition of E-state indices. The final Eq. (16) could explain and predict
88.8% and 84.1% respectively of the variance of antioxidant activity. The positive co-
efficient of S14 and S6 indicate that activity increases with increase in E-state values of
atoms 14 and 6 respectively while the negative coefficient of S3 and S1 indicate that ac-
tivity decreases with increase in E-state value of atoms 3 and 1 respectively. The values
of E-state parameters of the compounds present in the final equations and the intercorre-
lation among the E-state parameters for antiradical activity data modeling are shown in
Table 3A and 4A respectively in the supplementary materials section. Table 4A indicates
that some of the E-state parameters used in the stepwise regression equations are highly
intercorrelated. However, as the leave-one-out prediction statistics of these equations are
very good (for example, Eq. (16), Q2 = 0.841), we have accepted these equations with a
note of the intercorrelation among descriptors.
3.2.2 FA-MLR
From the factor analysis on the data matrix consisting of antioxidant activity data (AAFL)
of flavonoids and E-state indices, it was observed that 6 factors could explain the data
matrix to the extent of 96.2%. The antioxidant activity is highly loaded with factor 4
(loaded in S3) and moderately loaded with factor 3 (loaded in S1 and S11). Based on the
results of FA (Table 4), the following equation was derived with three variables:
AAFL = 21.25(±9.08)S9 − 16.546(±4.63)S3 − 86.91(±18.55)S1 + 481.15(±104.68)
n = 24, R2 = 0.728, R2a = 0.687, Q2 = 0.582, R = 0.853
s = 19.49, PRESS = 11655.6, F = 17.80, (df3.20), SDEP = 22.04, SPRESS = 24.14(17)
The standard errors of the respective E-state parameters are mentioned within paren-
thesis. Eq. (17) could explain and predict 68.7% and 58.2% respectively of the variance
of the antioxidant activity. The negative coefficient of S3 and S1 indicates that activity
decreases with increase in E-state values of atoms 3 and 1 respectively. The positive
coefficient of S9 indicates that the antioxidant activity increases with increase in E-state
value of atom 9.
1106 S. Ray et al. / Central European Journal of Chemistry 5(4) 2007 1094–1113
Table
3C
hem
ical
stru
cture
san
dob
serv
edan
dca
lcula
ted
antiox
idan
tac
tivity
ofth
ese
lect
edflav
onoi
ds.
Com
p.
Com
pound
R1
R2
R3
R4
R5
R6
R7
R8
AA
FL
No.
Nam
eO
bsa
Calb
B1
Kaem
pfe
rol
OH
OH
OH
HH
HO
HH
65.3
060.5
8B
2G
ala
ngin
OH
OH
OH
HH
HH
H64.9
061.5
6B
3Q
uer
cetin
OH
OH
OH
HH
OH
OH
H63.6
036.7
6B
4R
obin
etin
OH
HO
HH
HO
HO
HO
H61.7
054.7
9B
5Fis
etin
OH
HO
HH
HO
HO
HH
61.6
051.7
3B
6K
aem
pfe
ride
OH
OH
OH
HH
HO
CH
3H
60.0
059.2
2B
73-h
ydro
xy
OH
HH
HH
HH
H59.4
072.0
5flav
one
B8
Lari
cytr
inO
HO
HO
HH
HO
HO
HO
CH
328.5
034.0
0B
9Lari
cytr
in-3
’-O
HO
HO
HH
HO
-glu
OH
OC
H3
26.2
030.5
0O
-glu
B10
Myri
cetin
OH
OH
OH
HH
OH
OH
OH
18.4
039.7
7B
11
3,5
,7,3
’,4’,5’-
OC
H3
OC
H3
OC
H3
HH
OC
H3
OC
H3
OC
H3
2.6
0−1
1.6
9hex
am
ethox
yflav
one
B12
3,5
,7,3
’,4’-
OC
H3
OC
H3
OC
H3
HH
OC
H3
OC
H3
OH
1.1
0−5
.92
pen
tam
ethox
yflav
one
B13
Lari
cytr
in-3
,3’-O
-di-glu
O-g
luO
HO
HH
HO
-glu
OH
OC
H3
1.1
03.1
7B
14
Quer
cetin-3
-O-g
lu-7
-O-r
ha
O-g
luO
HO
-rha
HH
OH
OH
H−6
.20
−2.2
9B
15
Lari
cytr
in-3
,7,3
’-O
-tri
-glu
O-g
luO
HO
-glu
HH
O-g
luO
HO
CH
3−6
.20
−8.4
5B
16
Rutin
O-G
RO
HO
HH
HO
HO
HH
−10.2
6.5
0B
17
Mori
nO
HO
HO
HH
OH
HO
HH
63.5
064.4
4B
18
Fla
vone
HH
HH
HH
HH
−1.5
0−4
.88
B19
5-h
ydro
xy-fl
avone
HO
HH
HH
HH
H−4
.00
−19.7
4B
20
7-h
ydro
xy-fl
avone
HH
OH
HH
HH
H0.0
0−0
.815
B21
Chry
sin
HO
HO
HH
HH
HH
−20.8
−15.9
5B
22
8-m
ethox
y-fl
avone
HH
OC
H3
HH
HH
H−2
9.2
−10.4
8B
23
Vitex
inH
OH
OH
glu
HH
OH
H−2
9.6
−27.8
2B
24
Lute
onin
-7-O
-glu
HO
HO
-glu
HH
OH
OH
H−2
5.3
−22.2
1
Glu
=glu
cose
,R
ha
=rh
am
nose
,G
R=
β-D
-Glu
-α-L
-Rha;th
ese
sugars
are
connec
ted
via
thei
rglu
cosi
de
OH
gro
up
toth
eflav
onoid
s.aTaken
from
Ref
.[2
3];
bFro
mE
q.
(16)
S. Ray et al. / Central European Journal of Chemistry 5(4) 2007 1094–1113 1107
Table 4 Factor loading of the variables (antioxidant activity (AAFL) and predictor vari-
ables) after VARIMAX rotation.
F1 F2 F3 F4 F5 F6 Communalities
AAFL −0.008 −0.070 −0.550 −0.798 0.158 −0.069 0.979S1 0.170 0.024 0.938 0.236 0.028 0.117 0.989S2 0.557 0.664 −0.062 0.391 0.128 0.253 0.974S3 0.414 0.586 0.044 0.645 −0.106 0.173 0.985S4 0.700 0.524 −0.069 0.394 0.126 0.212 0.989S5 0.852 0.387 −0.111 0.181 0.178 0.188 0.938S6 0.944 0.158 −0.112 −0.029 −0.018 0.095 0.991S7 0.903 0.286 0.110 0.088 0.236 0.136 0.843S8 0.764 0.417 0.150 0.070 0.239 0.024 0.987S9 336 0.050 −0.024 −0.125 0.924 0.047 0.994S10 0.787 0.461 −0.120 0.116 0.316 0.186 0.990S11 −0.349 −0.388 0.841 −0.030 −0.044 −0.087 0.995S12 0.492 0.757 −0.099 0.244 0.107 0.314 0.976S13 0.361 0.464 0.136 0.167 0.071 0.760 0.910S14 0.273 0.891 −0.121 0.025 −0.032 0.160 0.935S15 0.419 0.809 −0.047 0.114 0.167 0.248 0.900S16 0.141 0.927 −0.115 0.036 −0.012 −0.069 0.983S17 0.415 0.839 0.016 0.172 0.132 0.245 0.975% Variance 0.315 0.314 0.112 0.090 0.068 0.060 0.962
3.2.3 PLS
The number of optimum components was 4 to obtain the final equation (optimized by
crossvalidation). Based on the standardized regression coefficients, the following variables
were selected for the final equation:
AAFL = 13.93S14 − 11.25S11 + 32.17S10 + 27.55S1
− 21.18S4 − 32.66S3 − 20.36S2 − 46.18S1 + 361.05
n = 24, R2 = 0.898, R2a = 0.876, Q2 = 0.834, R = 0.947
PRESS = 4615.1, F = 41.86, (df4.19), SDEP = 13.86, SPRESS = 15.58
(18)
Eq. (18) could explain and predict 87.6% and 83.4% respectively of the variance of the
antioxidant activity. The positive coefficient of the variable S14, S10 and S7 indicate that
the antioxidant activity increases with increase in the E-state value of atom 14, 10 and
7 respectively while the negative coefficients of S11, S4, S3, S2 and S1 indicate that the
activity decreases with increase in E-state values of atoms 11, 4, 3, 2 and 1 respectively.
The values of E-state parameters of the compounds present in the final equations and the
intercorrelation among the E-state parameters for antiradical activity data modeling are
shown in Table 3A and 4A respectively in the supplementary materials section.
1108 S. Ray et al. / Central European Journal of Chemistry 5(4) 2007 1094–1113
4 Overview
For antiradical activity of flavonoids in set 1, the Q2 value (0.871) is highest in Eq. (10)
(Table 5). Figure 2 also shows that the observed and calculated values according to
Eq. (10) show good linear relation. The equations obtained from all three statistical
techniques, indicate that S3 has negative coefficient for antiradical activity. S3 indicates
the importance of hydroxyl group at position 3 for antiradical activity. In the previous
QSAR report [22] on the same data set the reported R value is 0.938 (n = 29), while in
our best model the R value is 0.963 (n = 29). In case of antioxidant activity of flavonoids
(set 2), the Q2 value (0.841) is highest in Eq. (16) (Table 5). Figure 3 also shows that
the observed and calculated values according to Eq. (16) show good linear relation. The
equations obtained from all three statistical techniques, indicates that S3 and S1 have
negative coefficient for antioxidant activity. S1 implies the effect of oxygen function at
position 1 and S3 indicates the importance of hydroxyl group at position 3 needed for
antioxidant activity. In the previous QSAR report on antiradical activity [14] on the
same data set with n = 24 the squared correlation coefficient (r2) between observed and
calculated values and sum of squared residual between observed and calculated values are
0.285 and 1042.37 respectively, while in our model (Eq. 10) with same data set (n = 24)
the r2 and sum of squared residual values (n = 24) between observed and predicted value
are 0.888 and 128.55 respectively. For both the models (antiradical and antioxidant), the
intercorrelation among the parameters used in equations are shown in Table 2A and 4A,
respectively, in the supplementary materials section and utmost care was exercised to
avoid collinearities among the variables. Flavonoids are polyphenolic compounds classified
as benzo-γ-pyrone derivatives with a basic structure that consists of 15 carbon atoms
arranged in three rings (Fig. 1). There is diversity in the structures among flavonoid
compounds, the structural differences in each flavonoid family result from the variation in
the number/substitution pattern of the hydroxyl groups and the extent of glycosylation of
these groups. These structural variations are responsible for various activities of flavonoid
compounds. On the other hand, E-state indices is a representation of molecular structure
information as it varies with changes in structural features including branching, cyclicity,
homologation, heteroatom variation, and changes in relative positions of different groups.
Thus, electrotopological state represents electronic distribution information modified by
both local and global topology. The information encoded in the E-state value for an atom
is the electronic accessibility at that atom. The E-state index has been projected as a
useful tool in the context of QSAR studies and reported to have power to identify atoms
or fragments in the molecules, which are important for the biological activity.
S. Ray et al. / Central European Journal of Chemistry 5(4) 2007 1094–1113 1109
Table 5 Statistical qualities of the equations in two activity models.
Antiradical activity (pC) Antioxidant activity (AAFL)
Types of Equation R2 Q2 F Types of Equation R2 Q2 Fstatistical statisticalAnalysis Number Analysis NumberStepwise 6 0.770 0.733 90.40 Stepwise 13 0.479 0.381 20.19
7 0.818 0.766 58.24 14 0.653 0.564 19.768 0.870 0.805 55.96 15 0.768 0.691 22.119 0.904 0.837 56.63 16 0.901 0.841 43.0710 0.927 0.871 58.48
FA-MLR 11 0.814 0.761 56.89 FA-MLR 17 0.728 0.582 17.80PLS 12 0.826 0.783 61.81 PLS 18 0.898 0.834 41.89
Fig. 2 Correlation between observed (pC) and calculated antiradical activity values ac-
cording to Eq. (10)
5 Conclusions
For both, antiradical and antioxidant activity models, the final equations (10, 11, 12
and 16, 7, 18) obtained from three techniques are of acceptable statistical quality and
predictive potential considering the leave-one-out prediction statistics. The models also
show the utility of E-state parameters in QSAR study for a better understanding about
the contribution of atoms or fragments in the molecules towards imparting biological
activity.
1110 S. Ray et al. / Central European Journal of Chemistry 5(4) 2007 1094–1113
Fig. 3 Correlation between observed (AAFL) and calculated antioxidant activity values
according to Eq. (16)
References
[1] G.D. Carlo, N. Mascolo, A.A. Izzo and F. Capasso: “Flavonoid: old and new aspects
of a class of natural therapeutic drugs”, Life Sci., Vol. 65, (1999), pp. 337–353.
[2] P.C.H. Haollman and I.C.W. Arts: “Flavonols, flavones and flavanols-nature, occur-
rence and dietary burden”, J. Sci. Food Agric., Vol. 80, (2000), pp. 1081–1093.
[3] S.A. Aherne and N.M. O’Brien: “Dietary Flavonols: Chemistry, Food Content, and
Metabolism”, Nutrition, Vol. 18, (2002), pp. 75–81.
[4] J.B. Harborne: The Flavonoids, Advances in Research Since 1986, Chapman and
Hall, London, 1994.
[5] O.P.Agarwal: “The anti-inflammatory action of nepitrin, a flavonoid”, Agents Ac-
tions, Vol. 12, (1982), pp. 298–302.
[6] M.K. Church: “Cromoglycate-like anti-allergic drugs: a review”, Drugs Today, Vol.
14, (1978), pp. 281–341.
[7] B F Rasulev, N D Abdullaev, Vladimir N Syrov and J Leszczynski: “A quanti-
tative structure-activity relationship (QSAR) study of the antioxidant activity of
flavonoids”, QSAR & Comb. Sci, Vol. 24, (2005), pp. 1056–1065.
[8] T. Ozawa: “Genetic and functional changes in mitochondria associated with aging”,
Physiol. Rev., Vol. 77, (1997), pp. 425–464.
S. Ray et al. / Central European Journal of Chemistry 5(4) 2007 1094–1113 1111
[9] S. Luangaram, U. Kukongviriyapan, P. Pakdeechote, V. Kukongviriyapan and P.
Pannangpetch: “Protective effects of quercetin against phenylhydrazine-induced vas-
cular dysfunction and oxidative stress in rats”, Food Chem. Toxicol., Vol. 45, (2007),
pp. 448–455.
[10] S. Chaudhuri, A. Banerjee, K. Basu, B. Sengupta and P.K. Sengupta: “Interaction
of flavonoids with red blood cell membrane lipids and proteins: Antioxidant and
antihemolytic effects”, Int. J. Biol. Macromol., Vol. 41, (2007), pp. 42–48.
[11] A.L. Miller: “Antioxidant flavonoids: structure, function and clinical usage”, Alt.
Med. Rev., Vol. 1, (1996), pp. 103–111.
[12] N.C. Cook and S. Samman: “Flavonoids—chemistry, metabolism, cardioprotective
effects, and dietary sources”, J. Nutr. Biochem., Vol. 7, (1996), pp. 66–76.
[13] E. Middleton: “Biological properties of plant flavonoids: An overview”, Int. J. Phar-
macog., Vol. 34, (1996), pp. 344–348.
[14] O. Farkas, J. Jakus and K. Heberger: “Quantitative structure-activity relationships
of flavonoid compounds”, Molecules, Vol. 9, (2004), pp. 1079–1088.
[15] A.I. Khlebnikov, I.A. Schepetkin, N.G. Domina, L.N. Kirpotina and M.T. Quinn:
“Improved quantitative structure-activity relationship models to predict antioxidant
activity of flavonoids in chemical, enzymatic and cellular systems”, Bioorg. Med.
Chem., Vol. 15, (2007), pp. 1749–1770.
[16] A. Thakur, S. Vishwakarma and M. Thakur: “QSAR study of flavonoid derivatives
as p561ck tyrosinkinase inhibitors”, Bioorg. Med. Chem., Vol. 12, (2004), pp. 1209–
1214.
[17] R. Badhan and J. Penny: “In silico modeling of the interaction of flavonoids with
human P-glycoprotein nucleotide-binding domain”, Eur. J. Med. Chem., Vol. 41,
(2006), pp. 285–295.
[18] A. Karawajczyk, V. Drgan, N. Medic, G. Oboh, S. Passamonti and M. Novic: “Prop-
erties of flavonoids influencing the binding of bilitranslocase investigated by neural
network modeling”, Biochem. Pharmacol., Vol. 73, (2007), pp. 308–320.
[19] J. Lameira, C.N. Alves, V. Moliner and E. Silla: “A density functional study of
flavonoid compounds with anti-HIV activity”, Eur. J. Med. Chem., Vol. 41, (2006),
pp. 616–623.
[20] S. Zhang, X. Yang, R.A. Coburn and M.E. Morris: “Structure activity relationships
and quantitative structure activity relationships for the flavonoid-mediated inhibition
of breast cancer resistance protein”, Biochem. Pharmacol., Vol. 70, (2005), pp. 627–
639.
[21] C.N. Alves, J.C. Pinheiro, A.J. Camargo, M.M.C. Ferreira, R.A.F. Romero and
A.B.F. da Silva: “A multiple linear regression and partial least squares study of
flavonoid compound with anti-HIV activity”, J. Mol. Struct., Vol. 541, (2001), pp.
81–88.
[22] D. Amic, D.D. Amic, D. Beslo and N. Trinajstic: “Structure-radical scavenging
activity relationships of flavonoids”, Croat. Chem. Acta, Vol. 76, (2003), pp. 55–61.
1112 S. Ray et al. / Central European Journal of Chemistry 5(4) 2007 1094–1113
[23] S. Burda and W. Oleszek: “Antioxidant and antiradical activities of flavonoids”, J.
Agric. Food Chem., Vol. 49, (2001), pp. 2774–2779.
[24] L.H. Hall, B. Mohney and L.B. Kier: “The Electrotopological State: An Atom Index
for QSAR”, Quant. Struct-Act. Rel., Vol. 10, (1991), pp. 43–51.
[25] L.B. Kier and L.H. Hall: “An Electrotopological State Index for Atoms in Molecules”,
Pharm. Res., Vol. 7, (1990), pp. 801–807.
[26] L.H. Hall, B. Mohney and L.B. Kier: “Comparison of Electrotopological State In-
dexes with Molecular Orbital Parameters: Inhibition of MAO by Hydrazides”, Quant.
Struct-Act. Rel., Vol. 12, (1993), pp. 44–48.
[27] L.B. Kier and L.H. Hall: Advances in Drug Research, Vol. 22, Academic Press, New
York, 1992.
[28] L.B. Kier and L.H. Hall: Molecular Structure Description: The Electrotopological
State, Academic Press, San Diego, 1999.
[29] K. Roy, A.U. De and C. Sengupta: “QSAR with electrotopological state atom in-
dex: Antiadrenergic activity of N,N-dimethyl-2-bromo-2-phenylethylamines”, Indian
J. Chem., Vol. 38B, (1999), pp. 942–947.
[30] A. Saha, K. Roy, K. De and C. Sengupta: “QSAR with electrotopological state atom
index. Part 3: Receptor binding affinity of estrogens and non-steroidal estrogen
analogs”, Indian J. Chem. Soc., Vol. 78 (2001), pp. 92–96.
[31] K. Roy, D.K. Pal and C. Sengupta: “QSAR of antineoplastics V: Exploration of
Receptor Interaction Sites of Antitumor N-(7-Indolyl)benzenesulfonamides Targeting
G1 Phase Using Electrotopological State Atom Index”, Drug Des. Discov., Vol. 17,
(2001), pp. 207–218.
[32] K. Roy, A.U. De and C. Sengupta: “QSAR of Human Factor Xa Inhibitor N2-
aroylanthranilamides Using Principal Component Factor Analysis”, Drug Des. Dis-
cov., Vol. 18, (2002), pp. 23–31.
[33] A. Saha, K. Roy, K. De and C. Sengupta: “QSAR with Electrotopological State Atom
Index : Part IV. - Receptor Binding Affinity of Progestagens”, Indian J. Chem., Vol.
41B, (2002), pp. 1268–1275.
[34] K. Roy, S. Chakraborty and A. Saha: “Exploring Selectivity Requirements for COX-2
vs. COX-1 binding of 3,4-diaryloxazolones using E-state index”, Bioorg. Med. Chem.
Lett., Vol. 13, (2003), pp. 3753–3757.
[35] K. Roy and J.T. Leonard: “QSAR Modeling of HIV-1 Reverse Transcriptase In-
hibitor 2-Amino-6-arylsulfonylbenzonitriles and Congeners Using Molecular Connec-
tivity and E-State Parameters”, Bioorg. Med. Chem., Vol. 12, (2004), pp. 745–754.
[36] S. Chakraborty, C. Sengupta and K. Roy: “Exploring QSAR with E-state index:
Selectivity Requirements for COX-2 versus COX-1 Binding of Terphenyl Methyl
Sulfones and Sulfonamides”, Bioorg. Med. Chem. Lett., Vol. 14, (2004), pp. 4665–
4670.
[37] C. Sengupta, J.T. Leonard and K. Roy: “Exploring QSAR of melatonin receptor
ligand benzofuran derivatives using E-state index”, Bioorg. Med. Chem. Lett., Vol.
14, (2004), pp. 3435–3439.
S. Ray et al. / Central European Journal of Chemistry 5(4) 2007 1094–1113 1113
[38] S.K. Lee, Z.H. Mbwambo, H.S. Chung, L. Luyengi, E.J.C. Games, R.G. Mehta,
A.D. Kinghorn and J.M. Pezzuto: “Evaluation of the antioxidant potential of natural
products”, Combin. Chem. High Throughput Screen., Vol. 1, (1998), pp. 35–46.
[39] D.E. Pratt: Phenolic Compounds in Food and Their Effect on Health II: Antioxidants
and Cancer Prevention, American Chemical Society, Washington DC, 1992.
[40] R.B. Darlington: Regression and linear models, McGraw Hill, New York, 1990.
[41] R. Franke: Theoretical Drug Design Methods, Elsevier, Amsterdam, 1984.
[42] R. Franke and A. Gruska: “Principal component and factor analysis”, in H. Van
de Waterbeemd (Ed): Chemometric Methods in molecular Design (Methods and
Principles in Medicinal Chemistry, Vol. 2, R. Manhold, P. Krogsgaard-Larsen and
H. Timmerman, Eds), VCH, Weinheim, 1995, pp. 113–126.
[43] S. Wold: “PLS for multivariate linear modeling” in H. Van de Waterbeemd (Ed):
Chemometric Methods in Molecular Design (Methods and Principles in Medicinal
Chemistry), VCH, Weinheim, 1995, pp. 195–218.
[44] Y. Fan, L.M. Shi, K.W. Kohn, Y. Pommier and J.N. Weinstein: ”Quantitative
structure-antitumor activity relationships of camptothecin analogs: Cluster analysis
and genetic algorithm-based studies”, J. Med. Chem., Vol. 44, (2001), pp. 3254-3263.
[45] The GW-BASIC program ELECTRO1 was developed by Kunal Roy and standard-
ized on known data sets.
[46] SPSS is statistical software of SPSS Inc., IL, USA.
[47] MINITAB is statistical software of Minitab Inc, USA, http://www.minitab.com.
[48] G.W. Snedecor and W.G. Cochran: Statistical Methods, Oxford & IBH publishing
Co. Pvt. Ltd, NewDelhi, 1967.
[49] S. Word and L. Eriksson: “Statistical validation of QSAR results”, in H. Van de Wa-
terbeemd (Ed): Chemometric Methods in Molecular Design (Methods and Principles
in Medicinal Chemistry), VCH, Weinheim, 1995, pp. 312–318.
[50] A.K. Debnath: “Quantitative structure-activity relationship (QSAR): A versatile
tool in drug design”, in A.K. Ghose and V.N. Viswanadhan (Eds): Combinatorial
Library Design and Evaluation: Principles, Software, Tools and Application in Drug
Discovery, Marcel Dekker Inc, NewYork, 2001, pp. 73–129.
[51] J.D. Walker, J. Jaworska, M.H. Comber, T.W. Schultz and J.C. Dearden: “Guide-
lines for developing and using quantitative structure-activity relationships”, Environ
Toxicol. Chem., Vol. 22, (2003), pp. 1653–1665.
[52] S.A.B.E. Van Acker, M.J. De Groot, D.J. Van Den Berg, M.N.J.L. Tromp, G.D.O.
Den Kelder, W.J.F. Van Der Vijgh and A. Bust: “A quantum chemical explanation
of the antioxidant activity of flavonoids”, Chem. Res. Toxicol., Vol. 9, (1996), pp.
1305–1312.