ORIGINAL RESEARCH

Role of antioxidants on docetaxel-induced in vitro lipidperoxidation using malondialdehyde as model marker:an experimental and in silico approach

Supratim Ray • Selim Mondal • Sarbani Dey Ray •

Partha Pratim Roy

Received: 1 December 2013 / Accepted: 18 April 2014

� Springer Science+Business Media New York 2014

Abstract The present in vitro study was designed to

explore antiperoxidative potential of 28 structurally diverse

classes of antioxidants on docetaxel-induced lipid peroxi-

dation (LPO). Both experimental and in silico approaches

were adopted to explore the potential of antioxidants. Goat

liver tissue homogenate was used as source of lipid. Esti-

mation of malondialdehyde in liver tissue homogenate was

used as model marker for docetaxel-induced LPO. The

computational part of the present study was confined to

QSAR analysis of 28 structurally diverse classes of anti-

oxidants having LPO-inhibition potency induced by doce-

taxel for better understanding of structural features

necessary for their LPO-inhibition properties. The study

was performed with freely available online 2D descriptor

on PaDEL-Descriptors (open source). Stepwise regression

analysis was used as chemometric tool. The study showed

the LPO induction capacity of docetaxel. Butylated

hydroxyl toluene demonstrated highest potential

(-21.3 %) and hesperidin the lowest potential (14.36 %) to

suppress the docetaxel-induced LPO. The computational

study indicates the importance of topological distances

among atoms with in a molecule, specific branching pattern

relative to molecular size presents in a molecule required

for the LPO-inhibition activity. The developed model was

validated both internally and externally by using several

parameters.

Keywords Antioxidant � Lipid peroxidation � Docetaxel �Malondialdehyde � QSAR

Introduction

The polyunsaturated fatty acids of membrane phospholip-

ids are particularly susceptible to peroxidation and undergo

significant modifications, including the rearrangement or

loss of double bonds and in some cases, the reductive

degradation of lipid acyl side chains (Leibowitz and

Johnson, 1971; Gardner, 1975). Reactive oxygen free

radicals are responsible for damage of tissues through lipid

peroxidation (LPO) (Guio et al., 1996). Free radicals are

constantly formed in the human body, but the protection of

cellular structures from damage by free radicals can be

accomplished through enzymatic and non-enzymatic

defense mechanisms (Durak et al., 1994). LPO leads to

generation of peroxides and hydroperoxides that can

decompose to yield a wide range of cytotoxic end-products

most of which are aldehydes as exemplified by malondi-

aldehyde (MDA), 4-hydroxy-2-nonenal (4-HNE), etc.

(Esterbauer et al., 1998). Free radicals are constantly being

generated in the body through various mechanisms and

also being removed by endogenous antioxidant defense

Electronic supplementary material The online version of thisarticle (doi:10.1007/s00044-014-1019-8) contains supplementarymaterial, which is available to authorized users.

S. Ray (&)

Department of Pharmaceutical Sciences, Assam University,

Silchar 788011, India

e-mail: supratimray_in@yahoo.co.in

S. Mondal

Dr B C Roy College of Pharmacy & A.H.S, Durgapur 713206,

India

S. D. Ray

Department of Pharmaceutical Technology, Jadavpur University,

Kolkata 700032, India

P. P. Roy

Department of Pharmacy, Guru Ghasidas Vishwavidyalaya,

Bilaspur 495009, India

123

Med Chem Res

DOI 10.1007/s00044-014-1019-8

MEDICINALCHEMISTRYRESEARCH

mechanism that acts by scavenging free radicals, decom-

posing peroxides, and/or binding with pro-oxidant metal

ion. Free radical-mediated oxidative stress results usually

from deficient natural antioxidant defense. In case of

reduced or impaired defense mechanism and excess gen-

eration of free radicals that are not counter balanced by

endogenous antioxidant defense exogenously administered

antioxidants have been proven useful to overcome oxida-

tive damage (Halliwell, 1991).

Docetaxel is a semi-synthetic derivative of paclitaxel

which is obtained from the rare pacific yew tree Taxus

brevifolia (Clarke and Rivory, 1999). Docetaxel is cyto-

toxic to all dividing cells in the body due to its specific

action on cell cycle (Rang et al., 2003). It produces several

toxic side effects due to damage of normal cell-like hair

follicles, bone marrow, and other germ cells. It is used

mainly for the treatment of breast, ovarian, and non-small

cell lung cancer. Docetaxel has the capability of inducing

lipid oxidization and membrane damage in human hepa-

toma cells (Yang et al., 2009). LPO induction capacity of

drugs may be related to their toxic potential as exemplified

by adriamycin-induced cardiotoxicity, which occurs

through free radical-mediated process (Luo et al., 1997).

So the evaluation of antioxidants as suppressor of drug-

induced LPO provides a scope of further investigation for

their co-administration with drugs to reduce drug-induced

toxicities that are possibly mediated by free radical

mechanism.

Considering the above findings the present work has

been divided into two parts. At first attempts have been

made to find out the LPO induction capacity of docetaxel

on goat liver tissue and explore the beneficial role of 28

antioxidants on docetaxel–lipid interaction. MDA is used

as laboratory tool. Second, the computational portion of the

present study is confined to QSAR analysis of those

structurally diverse classes of antioxidants having LPO-

inhibition potency to promote our understanding of the

structural features necessary for their activity.

Results and discussion

Experimental work

The percent changes in MDA content of different samples

at 5 h of incubation were calculated with respect to the

control of the corresponding time of incubation and was

considered as indicator of the extent of LPO. The results of

the studies on docetaxel-induced LPO and its inhibition

with alpha-tocopherol/apigenin/ascorbic acid/BHA/BHT/

caffeic acid/chrysin/curcumin/dextrose/fisetin/flavanone/

flavone/fustin/galangin/hesperidin/kaempferol/larycytrin/

morin/myricetin/n-propyl gallate/naringenin/naringin/

quercetin/robinetin/rutin/taxifolin/uric acid/vitexin are

shown in Table 1.

From Table 1, it was evident that tissue homogenates

treated with docetaxel showed an increase in MDA

(5.59–40.51 %) content in samples with respect to control to

a significant extent. The observations suggest that docetaxel

could significantly induce the LPO process. MDA is a highly

reactive three-carbon dialdehyde produced as a byproduct of

polyunsaturated fatty acid peroxidation and arachidonic acid

metabolism (Yahya et al., 1996). But the MDA contents

were significantly reduced (14.36 to -21.3 %) in compari-

son to docetaxel-treated group when tissue homogenates

were treated with docetaxel in combination with above-

mentioned antioxidant (Table 1). Again the tissue homoge-

nates were treated only with the abovementioned antioxidant

then the MDA level was reduced (-2.77 to -39.00 %) in

comparison to the control and the docetaxel-treated group

(Table 1). This decrease may be due to the free radical

scavenging property of the antioxidant.

From Table 1, it is observed that there are significant

differences among various groups (F1) such as docetaxel-

treated, docetaxel and antioxidant-treated, and only anti-

oxidant-treated group. But within a particular group, dif-

ferences (F2) are insignificant which shows that there are

no statistical differences in animals in a particular group. If

F test is significant and more than two treatments are

included in the experiment it may not be obvious imme-

diately which treatments are different. To solve the prob-

lem, multiple comparison analysis using least significant

different procedure is proposed (Snedecor and Cochran,

1967; Bolton, 2001) on the percent changes data of various

groups such as docetaxel-treated (D), docetaxel and anti-

oxidant (DA), and only antioxidant-treated (A) with respect

to control group of corresponding time. It was observed

that the level of MDA in docetaxel-treated group, doce-

taxel and chrysin/fisetin/flavanone/galangin/hesperidin/

kaempferol/larycytrin/n-propyl gallate/naringin/quercetin/

rutin/uric acid/vitexin-treated groups, and only chrysin/

fisetin/flavanone/galangin/hesperidin/kaempferol/larycytrin/

n-propyl gallate/naringin/quercetin/rutin/uric acid/vitexin-

treated groups are statistically significantly different from

each other. But the MDA content in docetaxel-treated

group is only statistically significantly different from doce-

taxel and alpha-tocopherol/apigenin/ascorbic acid/BHA/

BHT/caffeic acid/curcumin/dextrose/flavone/fustin/morin/

myricetin/naringenin/robinetin and taxifolin-treated groups.

But there is no statistically significantly difference among

the docetaxel and antioxidant-treated group and only

antioxidant-treated groups.

Med Chem Res

123

Ta

ble

1E

ffec

tso

fan

tio

xid

ant

on

Do

ceta

xel

-in

du

ced

lip

idp

ero

xid

atio

n:

chan

ges

inM

DA

pro

file

Sl

no

.

Nam

eo

fth

e

anti

ox

idan

t

%ch

ang

esin

MD

Aco

nte

nt

(wit

hre

spec

tto

corr

esp

on

din

gco

ntr

ol)

du

eto

trea

tmen

tw

ith

dru

gan

d

or

anti

ox

idan

t

An

aly

sis

of

var

ian

cean

dm

ult

iple

com

par

iso

n

Sam

ple

s(A

v±

SE

)(n

=3

)

DD

AA

1A

lph

a-

toco

ph

ero

l

8.7

8(±

2.7

1)

-1

0.5

(±1

.13

)-

6.0

5(±

1.1

9)

F1

=3

8.4

[df

=(2

,4)]

,F

2=

1.7

9[d

f=

(2,4

)],

po

ole

dv

aria

nce

(S2)*

=7

.96

,cr

itic

ald

iffe

ren

ce(p

=0

.05

)#

LS

D=

5.3

1,

ran

ked

mea

ns*

*(D

)(D

A,

A)

2A

pig

enin

9.1

4(±

0.8

2)

-4

.6(±

1.1

2)

-5

.42

(±1

.99

)F

1=

40

.86

[df

=(2

,4)]

,F

2=

1.5

9[d

f=

(2,4

)],

po

ole

dv

aria

nce

(S2)*

=4

.91

,cr

itic

ald

iffe

ren

ce(p

=0

.05

)#

LS

D=

4.1

7,

ran

ked

mea

ns*

*(D

)(D

A,

A)

3A

sco

rbic

acid

24

.33

(±4

.31

)-

17

.25

(±4

.5)

-2

3.9

8(±

2.6

)F

1=

8.8

46

7[d

f=

(2,4

)],

F2

=3

.03

33

[df

=(2

,4)]

,p

oo

led

var

ian

ce(S

2)*

=2

32

.16

7,

crit

ical

dif

fere

nce

(p=

0.0

5)#

LS

D=

28

.68

8,

ran

ked

mea

ns*

*(D

)(D

A,

A)

4B

HA

22

.05

(±6

.04

)-

9.2

(±5

.18

)-

17

.62

(±3

.36

)F

1=

51

.39

[df

=(2

,4)]

,F

2=

6.7

7[d

f=

(2,4

)],p

oo

led

var

ian

ce(S

2)*

=2

5.5

1,cr

itic

ald

iffe

ren

ce(p

=0

.05

)#

LS

D=

9.5

0,

ran

ked

mea

ns*

*(D

)(D

A,

A)

5B

HT

23

.25

(±4

.62

)-

21

.3(±

2.6

4)

-3

9.0

0(±

3.8

2)

F1

=1

4.1

3[d

f=

(2,4

)],

F2

=1

.74

5[d

f=

(2,4

)],

po

ole

dv

aria

nce

(S2)*

=2

18

.48

,cr

itic

ald

iffe

ren

ce

(p=

0.0

5)#

LS

D=

27

.83

0,

ran

ked

mea

ns*

*(D

)(D

A,

A)

6C

affe

icac

id2

6.0

1(±

5.9

)-

0.4

9(±

3.9

2)

-1

0.9

9(±

2.0

4)

F1

=2

6.6

2[d

f=

(2,4

)],F

2=

1.9

8[d

f=

(2,4

)],p

oo

led

var

ian

ce(S

2)*

=4

0.9

8,cr

itic

ald

iffe

ren

ce(p

=0

.05

)#

LS

D=

12

.04

,ra

nk

edm

ean

s**

(D)

(DA

,A

)

7C

hry

sin

5.5

9(±

0.3

4)

-1

2.5

(±0

.65

8)

-5

.19

(±0

.22

8)

F1

=6

89

.3[d

f=

(2,4

)],

F2

=2

.98

[df

=(2

,4)]

,p

oo

led

var

ian

ce(S

2)*

=0

.36

,cr

itic

ald

iffe

ren

ce(p

=0

.05

)#

LS

D=

0.1

76

,ra

nk

edm

ean

s**

(D)

(DA

)(A

)

8C

urc

um

in6

.84

(±1

.38

)-

3.1

(±0

.72

)-

3.7

4(±

0.8

78

)F

1=

47

.35

[df

=(2

,4)]

,F

2=

2.2

9[d

f=

(2,4

)],

po

ole

dv

aria

nce

(S2)*

=2

.23

,cr

itic

ald

iffe

ren

ce(p

=0

.05

)#

LS

D=

2.8

1,

ran

ked

mea

ns*

*(D

)(D

A,

A)

9D

extr

ose

15

.51

(±1

.37

)-

17

.48

(±5

.6)

-3

4.9

2(±

6.6

9)

F1

=2

2.7

3[d

f=

(2,4

)],

F2

=0

.70

5[d

f=

(2,4

)],

po

ole

dv

aria

nce

(S2)*

=8

6.5

7,

crit

ical

dif

fere

nce

(p=

0.0

5)#

LS

D=

17

.52

,ra

nk

edm

ean

s**

(D)

(DA

,A

)

10

Fis

etin

7.2

5(±

1.1

5)

-2

.53

(±0

.31

4)

-4

.03

(±0

.38

)F

1=

15

9.4

2[d

f=

(2,4

)],

F2

=4

.71

[df

=(2

,4)]

,p

oo

led

var

ian

ce(S

2)*

=0

.70

5,

crit

ical

dif

fere

nce

(p=

0.0

5)#

LS

D=

1.5

9,

ran

ked

mea

ns*

*(D

)(D

A)

(A)

11

Fla

van

on

e8

.09

(±1

.34

)4

.2(±

1.1

4)

-3

.56

(±0

.63

)F

1=

29

.55

[df

=(2

,4)]

,F

2=

0.9

4[d

f=

(2,4

)],

po

ole

dv

aria

nce

(S2)*

=3

.57

,cr

itic

ald

iffe

ren

ce(p

=0

.05

)#

LS

D=

3.5

,ra

nk

edm

ean

s**

(D)

(DA

)(A

)

12

Fla

vo

ne

34

.96

(±6

.32

)-

0.6

33

(±5

.29

)-

15

.25

(±5

.02

)F

1=

21

.77

[df

=(2

,4)]

,F

2=

1.0

38

[df

=(2

,4)]

,p

oo

led

var

ian

ce(S

2)*

=9

1.8

6,

crit

ical

dif

fere

nce

(p=

0.0

5)#

LS

D=

18

.04

,ra

nk

edm

ean

s**

(D)

(DA

,A

)

13

Fu

stin

10

.78

(±1

.62

)-

3.2

6(±

0.5

7)

-4

.86

(±0

.72

)F

1=

49

.95

[df

=(2

,4)]

,F

2=

0.3

4[d

f=

(2,4

)],

po

ole

dv

aria

nce

(S2)*

=4

.45

,cr

itic

ald

iffe

ren

ce(p

=0

.05

)#

LS

D=

3.9

7,

ran

ked

mea

ns*

*(D

)(D

A,

A)

14

Gal

ang

in7

.68

(±0

.35

)-

1.4

4(±

0.1

03

)-

2.9

9(±

0.4

8)

F1

=1

93

.68

[df

=(2

,4)]

,F

2=

0.1

39

[df

=(2

,4)]

,p

oo

led

var

ian

ce(S

2)*

=0

.51

5,

crit

ical

dif

fere

nce

(p=

0.0

5)#

LS

D=

1.3

6,

ran

ked

mea

ns*

*(D

)(D

A)

(A)

15

Hes

per

idin

40

.51

(±4

.23

)1

4.3

6(±

5.1

)-

13

.32

(±2

.4)

F1

=1

8.3

8[d

f=

(2,4

)],

F2

=3

.94

[df

=(2

,4)]

,p

oo

led

var

ian

ce(S

2)*

=1

18

.26

,cr

itic

ald

iffe

ren

ce

(p=

0.0

5)#

LS

D=

20

.47

,ra

nk

edm

ean

s**

(D)

(DA

)(A

)

16

Kae

mp

fero

l1

9.6

1(±

1.4

7)

6.5

2(±

0.6

3)

-2

.77

(±0

.32

)F

1=

19

2.9

8[d

f=

(2,4

)],

F2

=2

.07

9[d

f=

(2,4

)],

po

ole

dv

aria

nce

(S2)*

=1

.96

4,

crit

ical

dif

fere

nce

(p=

0.0

5)#

LS

D=

2.6

4,

ran

ked

mea

ns*

*(D

)(D

A)

(A)

17

Lar

ycy

trin

7.4

3(±

0.5

2)

-3

.83

(±0

.58

)-

5.9

3(±

0.4

3)

F1

=1

67

.02

[df

=(2

,4)]

,F

2=

0.5

6[d

f=

(2,4

)],p

oo

led

var

ian

ce(S

2)*

=0

.92

,cr

itic

ald

iffe

ren

ce(p

=0

.05

)#

LS

D=

1.8

1,

ran

ked

mea

ns*

*(D

)(D

A)

(A)

Med Chem Res

123

Ta

ble

1co

nti

nu

ed

Sl

no

.

Nam

eo

fth

e

anti

ox

idan

t

%ch

ang

esin

MD

Aco

nte

nt

(wit

hre

spec

tto

corr

esp

on

din

gco

ntr

ol)

du

eto

trea

tmen

tw

ith

dru

gan

d

or

anti

ox

idan

t

An

aly

sis

of

var

ian

cean

dm

ult

iple

com

par

iso

n

Sam

ple

s(A

v±

SE

)(n

=3

)

DD

AA

18

Mo

rin

35

.34

(±5

.17

)0

.34

(±3

.76

)-

10

.04

(±1

.47

)F

1=

32

.51

[df

=(2

,4)]

,F

2=

1.8

9[d

f=

(2,4

)],p

oo

led

var

ian

ce(S

2)*

=5

2.1

9,cr

itic

ald

iffe

ren

ce(p

=0

.05

)#

LS

D=

13

.60

,ra

nk

edm

ean

s**

(D)

(DA

,A

)

19

My

rice

tin

7.4

8(±

0.7

17

)-

4.3

5(±

1.7

2)

-6

.81

(±1

.78

)F

1=

31

.09

[df

=(2

,4)]

,F

2=

1.5

3[d

f=

(2,4

)],

po

ole

dv

aria

nce

(S2)*

=5

.63

,cr

itic

ald

iffe

ren

ce(p

=0

.05

)#

LS

D=

4.4

7,

ran

ked

mea

ns*

*(D

)(D

A,

A)

20

n-P

rop

yl

gal

late

15

.51

(±1

.2)

2.3

6(±

0.4

1)

-3

.91

(±0

.55

)F

1=

13

9.5

4[d

f=

(2,4

)],

F2

=0

.72

3[d

f=

(2,4

)],

po

ole

dv

aria

nce

(S2)*

=2

.11

,cr

itic

ald

iffe

ren

ce

(p=

0.0

5)#

LS

D=

2.7

3,

ran

ked

mea

ns*

*(D

)(D

A)

(A)

21

Nar

ing

enin

8.9

8(±

0.5

74

)-

3.1

1(±

0.8

)-

3.6

4(±

1.4

8)

F1

=5

0.7

6[d

f=

(2,4

)],

F2

=1

.15

[df

=(2

,4)]

,P

oo

led

var

ian

ce(S

2)*

=3

.01

,C

riti

cal

dif

fere

nce

(p=

0.0

5)#

LS

D=

3.2

7,

Ran

ked

mea

ns*

*(D

)(D

A,

A)

22

Nar

ing

in2

3.3

9(±

4.2

4)

10

.1(±

2.1

9)

-1

8.8

5(±

2.9

1)

F1

=9

6.7

8[d

f=

(2,4

)],F

2=

4.4

8[d

f=

(2,4

)],p

oo

led

var

ian

ce(S

2)*

=1

4.4

7,cr

itic

ald

iffe

ren

ce(p

=0

.05

)#

LS

D=

7.1

6,

ran

ked

mea

ns*

*(D

)(D

A)

(A)

23

Qu

erce

tin

19

.66

(±2

.42

)-

20

.06

(±2

.63

)-

9.7

7(±

0.9

8)

F1

=1

06

.88

[df

=(2

,4)]

,F

2=

1.4

6[d

f=

(2,4

)],

po

ole

dv

aria

nce

(S2)*

=1

1.9

3,

crit

ical

dif

fere

nce

(p=

0.0

5)#

LS

D=

6.5

0,

ran

ked

mea

ns*

*(D

)(D

A)

(A)

24

Ro

bin

etin

7.0

7(±

1.4

6)

-3

.26

(±0

.96

)-

5.7

5(±

1.7

3)

F1

=1

6.6

6[d

f=

(2,4

)],

F2

=0

.18

[df

=(2

,4)]

,p

oo

led

var

ian

ce(S

2)*

=8

.32

,cr

itic

ald

iffe

ren

ce(p

=0

.05

)#

LS

D=

5.4

3,

ran

ked

mea

ns*

*(D

)(D

A,

A)

25

Ru

tin

27

.34

(±6

.3)

5.6

5(±

1.1

2)

-1

1.7

2(±

1.3

2)

F1

=3

1.4

3[d

f=

(2,4

)],F

2=

1.5

1[d

f=

(2,4

)],p

oo

led

var

ian

ce(S

2)*

=3

6.5

5,cr

itic

ald

iffe

ren

ce(p

=0

.05

)#

LS

D=

11

.38

,ra

nk

edm

ean

s**

(D)

(DA

)(A

)

26

Tax

ifo

lin

11

.76

(±3

.15

)-

5.0

1(±

0.8

2)

-6

.71

(±2

.25

)F

1=

13

.4[d

f=

(2,4

)],F

2=

0.0

17

[df

=(2

,4)]

,p

oo

led

var

ian

ce(S

2)*

=2

3.3

2,cr

itic

ald

iffe

ren

ce(p

=0

.05

)#

LS

D=

9.0

9,

ran

ked

mea

ns*

*(D

)(D

A,

A)

27

Uri

cac

id3

5.7

1(±

4.6

)1

0.7

5(±

2.2

)-

11

.61

(±0

.59

)F

1=

17

.39

[df

=(2

,4)]

,F

2=

0.4

6[d

f=

(2,4

)],p

oo

led

var

ian

ce(S

2)*

=9

6.5

5,cr

itic

ald

iffe

ren

ce(p

=0

.05

)#

LS

D=

18

.51

,ra

nk

edm

ean

s**

(D)

(DA

)(A

)

28

Vit

exin

8.0

4(±

0.3

4)

3.3

8(±

0.3

44

)-

3.1

5(±

0.6

2)

F1

=1

11

.24

[df

=(2

,4)]

,F

2=

0.1

6[d

f=

(2,4

)],p

oo

led

var

ian

ce(S

2)*

=0

.85

,cr

itic

ald

iffe

ren

ce(p

=0

.05

)#

LS

D=

1.7

4,

ran

ked

mea

ns*

*(D

)(D

A)

(A)

Av

erag

eso

fth

ree

sets

;S

E=

stan

dar

der

ror

(n=

3);

theo

reti

cal

val

ues

of

F:

p=

0.1

lev

elF

1=

4.3

2[d

f=

(2,4

)],

F2

=4

.32

[df

=(2

,4)]

;p

=0

.05

lev

elF

1=

6.9

4[d

f=

(2,4

)],

F2

=6

.94

[df

=(2

,4)]

,F

1an

dF

2co

rres

po

nd

ing

tov

aria

nce

rati

ob

etw

een

gro

up

san

dw

ith

ing

rou

ps,

resp

ecti

vel

y;

D,

DA

,an

dA

ind

icat

ed

oce

tax

el-t

reat

ed,

do

ceta

xel

and

anti

ox

idan

t-tr

eate

d,

and

on

ly

anti

ox

idan

t-tr

eate

d,

resp

ecti

vel

y

*E

rro

rm

ean

squ

are

**

Tw

om

ean

sn

ot

incl

ud

edw

ith

insa

me

par

enth

esis

are

stat

isti

call

ysi

gn

ifica

ntl

yd

iffe

ren

tat

p=

0.0

5le

vel

#C

riti

cal

dif

fere

nce

acco

rdin

gto

leas

tsi

gn

ifica

nt

pro

ced

ure

Med Chem Res

123

Computational work

Membership of compounds in different clusters generated

using k-means clustering technique is shown in Table 2.

The PCA score plot of first three principal components of

the standardized descriptor matrix suggesting that test set

compounds lie in near vicinity of some training set mole-

cules (Fig. 1). In the developed model, difference between

R2 and Q2 values is not very high (less than 0.3) (Eriksson

et al., 2003).

%DA ¼ 6:514�470ETA EtaP B RC�38:2ATSc2

R2 ¼ 0:6416; R2a ¼ 0:602; Q2

int ¼ 0:571;

ntraining ¼ 21; r2mðLOOÞ ¼ 0:525;

Dr2mðLOOÞ ¼ 0:075; Q2

ext ¼ 0:576;

ntest ¼ 7; r2mðtestÞ ¼ 0:526;

D r2mðtestÞ ¼ 0:026;

r2mðoverallÞ ¼ 0:524;Dr2

mðoverallÞ ¼ 0:032:

The relative order of importance of the descriptors is as fol-

lows: ATSc2[ETA_EtaP_B_RC. The equation could explain

and predict, respectively, 60.2 and 57.1 % of variance. When

this equation was applied for prediction of test set compounds,

the predictive R2 value for the test set was found to be 0.576.

The autocorrelation of a topological structure of lag 2

(ATSc2) has negative coefficient toward activity. The

values are calculated considering weight equal to charges.

The 2D-autocorrelation descriptors in general explain how

the values of certain functions, at intervals equal to the lag

d, are correlated. In the case of the descriptors used, lag is

the topological distance, and the atomic properties (weight

or charge) are the functions correlated. The descriptors can

be obtained by summing up the products of certain prop-

erties of the two atoms located at a given topological dis-

tance or spatial lag. It can be expressed as follows:

ATSd ¼XA

i¼1

Xa

j¼1

dij:ðwi � wjÞd;

where w is any atomic property, A is the atom number, d is

the considered topological distance (i.e., the lag in auto-

correlation terms), dij is the Kronecker delta (dij = 1 if

dij = d, zero otherwise). Compounds like BHT, Alpha-

tocopherol possesses comparatively lower value showing

better LPO-inhibition potency. In all these molecules, less

number of functional groups present where the topological

distance between carbon and functional group is two. But

compounds like hesperidin, rutin, and naringin contain

more functional groups and the distance between carbon

and functional group is two showing comparatively poor

activity.

Eta_EtaP_B_RC is a measure of branching index rela-

tive to molecular size and has negative coefficient toward

activity. It reflects a measure of overall branchedness

present in a molecule. It is represented as gB

0= (gB/NV)

where gB = gNlocal - gR

local ? 0.086NR. The NR term rep-

resents a correction factor for cyclicity. Compound like

chrysin possesses less branching pattern in the molecular

structure, showing better activity. But galangin, morin

rutin, having higher branching pattern in the molecule

showing lesser activity.

Table 2 K-means clustering of compounds using standardized descriptors

Cluster

no.

No. of compounds in

cluster

Compounds (Sl nos.) in each clusters

1 3 8 17 28

2 4 1 15 22 25

3 21 2 3 4 5 6 7 9 10 11 12 13 14 16 18 19 20 21 23 24 26 27

Fig. 1 The PCA score plot of first three principal components of the

standardized descriptor matrix

Med Chem Res

123

Overview and conclusions

The present in vitro study was designed to explore anti-

peroxidative potential of 28 structurally diverse classes of

antioxidants on docetaxel-induced LPO. Both experimental

and theoretical approaches were adopted to explore the

potential of antioxidants. Goat liver tissue homogenate was

used as source of lipid. Estimation of MDA in liver tissue

homogenate was used as model marker for docetaxel-

induced LPO. The study showed the LPO induction

capacity of docetaxel. It was also observed that all 28

antioxidants had the ability to suppress the LPO. But

among them, BHT showed highest potential (-21.3 %)

and hesperidin showing lowest potential (14.36 %) to

suppress the docetaxel-induced LPO.

The theoretical study was designed to explore LPO-

inhibition potency of 28 structurally diverse classes of

antioxidants on docetaxel-induced LPO. The whole dataset

(n = 28) was divided into a training set (21 compounds)

and a test set (7 compounds) based on k-means clustering

of the standardized descriptor matrix and models were

developed from the training set. Stepwise regression ana-

lysis was used as chemometric tool. The predictive ability

of the models was judged from the prediction of the

activity of the test set compounds. The study indicates the

importance of topological distances among atoms with in a

molecule, specific branching pattern relative to molecular

size presents in a molecule required for the LPO-inhibition

activity. Figure 2 shows a scatter plot of observed versus

calculated/predicted values of the training and test set

compounds, respectively, of the developed model. The

developed model also passes the criteria of rm2 for test,

training, and overall set. The intercorrelation among the

parameters used in equation is shown in Table 3 and

utmost care was exercised to avoid collinearities among the

variables. From the total study, it is observed that BHT

having less number of functional group present possesses

highest activity.

Materials and methods

Experimental work

The pure sample of docetaxel used in the present study was

provided by Fresenius Kabi, Kalyani, India. Thiobarbituric

acid (TBA) and trichloroacetic acid (TCA) were purchased

from Ranbaxy Fine Chemicals Ltd., New Delhi; butylated

hydroxyl toluene (BHT) and butylated hydroxyl anisole

(BHA), alpha-tocopherol, and ascorbic acid were from

Merck, Mumbai; morin, rutin, dextrose, and uric acid were

from CDH Pvt. Ltd., New Delhi; naringin, flavone, flava-

none, hesperidin, quercetin, curcumin, and caffeic acid

were from Himedia Bioscience, Mumbai; apigenin, chry-

sin, kaempferol, fisetin, galangin, naringenin, taxifolin, and

vitexin were from Sigma-Aldrich, St. Louis, MO; myrice-

tin and n-propyl gallate were from SRL, Mumbai; larycy-

trin and fustin were from Triveni Aromatics and Perfumery

Pvt. Ltd., Vapi; robinetin was from Clearsynth Labs

(P) Ltd., Mumbai. 1,1,3,3-Tetraethoxypropane (TEP) was

from Sigma chemicals Co., St. Louis, MO, USA. All other

reagents were of analytical grade.

The study was performed on goat (Capra capra) liver.

MDA content of the tissue sample was used as marker of

LPO. The goat liver was selected because of its easy

availability and close similarity to the human liver in its

lipid profile (Hilditch and Williams, 1964).

Preparation of tissue homogenate

Goat liver was collected from Durgapur Municipal Cor-

poration (DMC) approved outlet. Goat liver perfused with

normal saline through hepatic portal vein was harvested

and its lobes were briefly dried between filter papers to

remove excess blood and thin cut with a heavy-duty blade.

Table 3 Intercorrelation among descriptors used in the model from

stepwise analysis

Eta_EtaP_B_RC ATSc2

Eta_EtaP_B_RC 1.000 0.002

ATSc2 0.002 1.000

Fig. 2 Scatter plot of observed versus calculated/predicted values of

the training and test set compounds, respectively, of the developed

model

Med Chem Res

123

The small pieces were then transferred in a sterile vessel

containing phosphate buffer (pH 7.4) solution. After

draining the buffer solution as completely, the liver was

immediately grinded to make a tissue homogenate (1 g/ml)

using freshly prepared phosphate buffer (pH 7.4) solution

(Pandey et al., 1994). The homogenate was divided into

four equal parts, which were then treated differently as

mentioned below.

Incubation of tissue homogenate with docetaxel and/

or antioxidant

The tissue homogenate was divided into four equal parts. The

first portion was kept as control (C), while the second portion

was treated with docetaxel (D) at a concentration of

0.143 lM/g wet liver tissue homogenate. The third portion

was treated both with docetaxel at a concentration of

0.143 lM/g wet liver tissue homogenate, and antioxidant

(alpha-tocopherol/apigenin/ascorbic acid/BHA/BHT/caffeic

acid/chrysin/curcumin/dextrose/fisetin/flavanone/flavone/

fustin/galangin/hesperidin/kaempferol/larycytrin/morin/myr-

icetin/n-propyl gallate/naringenin/naringin/quercetin/robine-

tin/rutin/taxifolin/uric acid/vitexin) at a concentration of

0.189 lM/g wet liver tissue homogenates (DA). The fourth

one was treated only with abovementioned antioxidant alone

at a concentration of 0.189 lM/g wet liver tissue homogenate

(A). After treatment with docetaxel and/or antioxidant, the

different portions of liver homogenate were shaken for 5 h at

ambient temperature and MDA content of different propor-

tions was estimated.

Estimation of malondialdehyde level from tissue

homogenate

The estimation was repeated in three animal sets. In each

set, three replicate samples of 2.5 ml of incubation mixture

were mixed with 2.5 ml of 10 % (w/v) TCA and centri-

fuged at room temperature at 3,000 rpm for 30 min to

precipitate protein. Then 2.5 ml of the supernatant was

treated with 5 ml of 0.002 (M) TBA solutions and then

volume was made up to 10 ml with distilled water. The

mixture was heated on a boiling water bath for 30 min.

Then tubes were cooled to a room temperature and the

absorbance was measured at 530 nm against a TBA blank

(prepared from 5 ml of TBA solution and 5 ml of distilled

water) (Ohkawa et al., 1979). The concentrations of MDA

were determined from standard curve. The standard cali-

bration curve was drawn based on the following procedure.

Different aliquots from standard TEP solution were taken

in graduated stoppered (10.00 ml) test tube and volume of

each solution was made up to 5.00 ml. To each solution,

5.00 ml of TBA reagent was added and the mixture was

heated in a steam bath for half an hour when a pink color

developed. The solutions were cooled to room temperature

and their absorbances were noted at 530 nm using TBA

reagent as blank. By plotting absorbances against concen-

trations, a straight line passing through the origin was

obtained. Beer’s law was obeyed over the entire concen-

tration range. The best fit equation is A = 0.00705m -

0.00107, where m is the amount of MDA in nM and A is

the absorbances at 530 nm. The statistical significance of

the equation is checked by, R = 0.9993, SEE = 0.0041

and F = 6073.95 (df = 1, 8).

Statistical analysis

The results were expressed as mean of percent changes of

various groups with respect to corresponding control along

with standard errors. Interpretation of the result was sup-

ported by analysis of variance (ANOVA) and multiple

comparison analysis (Snedecor and Cochran, 1967; Bolton,

2001).

Computational work

Data set

The percent changes in MDA content of 28 docetaxel–

antioxidant-treated groups were used as response variable

(% DA) for subsequent QSAR analyses (Table 1).

Descriptors

The structures of 28 compounds (Fig. 3) were sketched

using Chem Draw Ultra version 6.0 (CS ChemOffice is

software of Cambridge Soft Corporation, USA) and saved

in mol. format which is one of the suitable input formats

for PaDEL-Descriptors. The energies of structural config-

uration were minimized by AM-1 method using Chem 3D

Ultra version 6.0 and used as input structure for descriptor

calculations. Only 2D descriptors available on freely

available PaDEL-Descriptors were considered for the

present study (Yap, 2011). Initially, 1,660 descriptors were

calculated using PaDEL-Descriptors software version 2.12.

Then we deleted the descriptors with high intercorrelation

(0.95), as well as zero and constant value descriptors.

Finally, pruned 232 descriptors were chosen for QSAR

analysis of selected data set. The categorical lists of the

descriptors are listed in Table 4. The values of the

descriptors present in the developed model are listed in

Supplementary Materials.

Model development

To begin the model development process, the whole data

set (n = 28) was divided into training (n = 21, 75 % of the

Med Chem Res

123

Fig. 3 Structural features of antioxidants

Med Chem Res

123

Table 4 Categorical list of descriptors used in QSAR analysis

Category of descriptors Name of descriptors

ALOGP AlogP, AlogP2, AMR

Atom count nH, nC

Autocorrelation (charge) ATSc1, ATSc2, ATSc3, ATSc4, ATSc5

Autocorrelation (mass) ATSm5

Autocorrelation

(polarizability)

ATSp1, ATSp5

BCUT BCUTw-1l, BCUTw-1h, BCUTc-1l, BCUTc-1h, BCUTp-1l, BCUTp-1h

Bond count nBondS2, nBondS3

BPol bpol

Carbon types C2SP2, C3SP2, C1SP3, C2SP3

Chi chain SCH-6, SCH-7, VCH-6, VCH-7

Chi cluster SC-3, SC-5, VC-3, VC-5

Chi path cluster SPC-4, SPC-5, VPC-4

Chi path SP-4, SP-7, VP-2, VP-6

Crippen LogP Crippen LogP

Eccentric Connectivity

Index

ECCEN

Electrotopological state

atom type

nHBint3, nHBint4, nHBint5, nHBint6, nHBint7, nHBint8, nHBint9, nHBint10, nHCsats, nHCsatu,nssCH2, ndsCH,

nsssCH, ndssC, nsOH, ndO, nssO, SHBd, SwHBa, SHBint3, SHBint4, SHBint5, SHBint6, SHBint7, SHBint8,

SHBint9, SHBint10, SHsOH, SHCsats, SHCsatu, SHother, SsCH3, SssCH2, SdsCH, SsssCH, SdssC, minHBd,

minHBa, minwHBa, minHBint3, minHBint4, minHBint5, minHBint6, minHBint7, minHBint8, minHBint9,

minHBint10, minHsOH, minHdsCH, minHCsats, minHCsatu, minssCH2, mindsCH, minsssCH, minsOH,

minssO, maxHBd, maxwHBa, maxHBint3, maxHBint4, maxHBint5, maxHBint6, maxHBint7, maxHBint8,

maxHBint9, maxHBint10, maxHsOH, maxHdsCH, maxHCsats, maxsCH3, maxdsCH, maxdssC, maxdO, suml,

hmax, gmax, hmin, gmin, Lipoaffinity Index, MAXDN2, MAXDP2

Extended topochemical

atom type

ETA_dAlpha_B, ETA_Epsilon_3, ETA_Epsilon_4, ETA_Epsilon_5, ETA_dEpsilon_A, ETA_dEpsilon_B,

ETA_dEpsilon_C, ETA_dEpsilon_D, ETA_Psi_1, ETA_Shape_P, ETA_Shape_Y, ETA_Beta, ETA_BetaP,

ETA_BetaP_s, ETA_Beta_ns, ETA_BetaP_ns, ETA_dBeta, ETA_dBetaP, ETA_Eta, ETA_EtaP, ETA_Eta_F,

ETA_EtaP_F, ETA_EtaP_L, ETA_EtaF_L, ETA_EtaP_F_L, ETA_Eta_B, ETA_EtaP_B, ETA_Eta_B_RC,

ETA_EtaP_B_RC

FMF FMF

Fragment complexity fragC

PaDEL H bond acceptor

count

nHBAcc3,

PaDEL H bond donor

count

nHBDon_Lipinski

Hybridization ratio HybRatio

Kappa Shape Indices Kier1, Kier2, Kier3

Largest chain nAtomLC

Largest Pi system nAtomP

Longest aliphatic chain

descriptor

nAtomLAC

Mannhold log P descriptor MLogP

MDE descriptor MDEC-11, MDEC-12, MDEC-13, MDEC-22, MDEC-23, MDEC-33, MDEO-11, MDEO-12, MDEO-22

MLFER descriptor MLFER_A, MLFER_B, MLFER_S, MLFER_E, MLFER_L

Petitjean number descriptor Petitjean number

Ring count descriptor nRing, nFRing, nTRing, nT6Ring

Rotatable bonds count

descriptor

nRotB

Rule of five descriptor LipinskiFailures

VAdjMa descriptor vAdjMat

Weighted path descriptor WTPT-2, WTPT-3

Med Chem Res

123

total number of compounds) and test (n = 7, 25 % of the

total number of compounds) sets by k-means clustering

technique applied on standardized descriptor matrix. The

QSAR model was developed using the training set com-

pounds (optimized by Q2), and then the developed models

were validated (externally) using the test set compounds.

Chemometric tools

Stepwise regression is used as chemometric tool. In step-

wise regression, a multiple term linear equation was built

step-by-step (Darlington, 1990). The basic procedures

involve (1) identifying an initial model, (2) iteratively

‘‘stepping,’’ i.e., repeatedly altering the model of the pre-

vious step by adding or removing a predictor variable in

accordance with the ‘‘stepping criteria,’’ (F = 4.0 for

inclusion; F = 3.9 for exclusion) in our case, and (3) ter-

minating the search when stepping is no longer possible

given the stepping criteria, or when a specified maximum

number steps has been reached. Specifically, at each step

all variables are reviewed and evaluated to determine

which one will contribute most to the equation. That var-

iable will then be included in the model, and the process

started again. A limitation of the stepwise regression search

approach is that 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 fitted subsequently to study

whether some other subsets of X variables might be better.

Software used for model development

MINITAB version 14 software (MINITAB version 14 is

statistical software of Minitab Inc, USA) was used for

stepwise regression method. K-means clustering, stan-

dardization of the variables was performed in SPSS version

9.0 software (SPSS version 9.0 software is statistical soft-

ware of IBM Corporation). STAISTICA version 7 software

(STATISTICA version 7 is statistical software of Stat Soft

Inc.) was used for the determination of the leave-one-out

(LOO) values of the training set compounds.

Model validation

The statistical qualities of various equations were judged

by calculating several metrics namely determination coef-

ficient (R2) as a measure of the total variance of the

response explained by the regression models (fitting),

explained variance (Ra2), and variance ratio (F) at specified

degrees of freedom (df) (Snedecor and Cochran, 1967).

Both internal and external validations are performed to

assess to reliability and the predictive potential of the

developed models. To determine the predictive quality of

the models, models are required to be further validated

using different validation techniques: (a) internal validation

or cross-validation using the training set compounds, and

(b) external validation using the test set compounds.

Internal validation

The generated model was validated internally by the LOO

procedure (Qint2 ) (Wold and Eriksson, 1995). It can be

expressed as follows:

Q2int ¼ 1�

PðYobs � YcalÞ2PðYobs � �YtrainingÞ2

; ðiÞ

where Yobs and Ycal indicate observed and calculated

activity of training set compounds. �Ytraining indicates mean

of activity of training set respectively.

External validation

The developed models were judged by external validation

parameters like Qext2 (Hawkins, 2004). It is defined as

follows:

Q2ext ¼ 1�

PðYobsðtestÞ � YcalðtestÞÞ2PðYobsðtestÞ � �YtrainingÞ2

; ðiiÞ

Table 4 continued

Category of descriptors Name of descriptors

Wiener numbers descriptor WPATH

XLogP descriptor XLogP

Pubchem fingerprint PubchemFP2, PubchemFP12, PubchemFP20, PubchemFP21, PubchemFP181, PubchemFP184, PubchemFP185,

PubchemFP339, PubchemFP346, PubchemFP366, PubchemFP374, PubchemFP380, PubchemFP432,

PubchemFP516, PubchemFP535, PubchemFP537, PubchemFP542, PubchemFP553, PubchemFP571,

PubchemFP579, PubchemFP589, PubchemFP604, PubchemFP614, PubchemFP620, PubchemFP639,

PubchemFP642, PubchemFP661, PubchemFP662, PubchemFP663, PubchemFP667, PubchemFP672,

PubchemFP681, PubchemFP692, PubchemFP697, PubchemFP698, PubchemFP701, PubchemFP714,

PubchemFP798, PubchemFP803, PubchemFP819, PubchemFP824

Med Chem Res

123

where Yobs(test) and Ycal(test) indicate observed and calcu-

lated activity of test set compounds. �Ytraining indicates mean

of activity of training set.

Further test on external validation

As external validation is the optimum tool for establishing

the predictive QSAR models, so beside the above param-

eters two more external validation parameters were also

employed to check the predictive ability of the developed

models.

The parameters r2m and Drm

2 are utilized to indicate better

both the internal and external predictive capacities of a

model and to ascertain the proximity in the values of the

predicted and observed response data (Ojha et al., 2011;

Roy and Roy, 2008). They are calculated as follows:

r2m ¼ ðr2

m þ r02mÞ=2; ð1Þ

Dr2m ¼ jðr2

m � r02mÞj; ð2Þ

where r2m ¼ r2 � ð1�

ffiffiffiffiffiffiffiffiffiffiffiffiffiffir2 � r2

0

pÞ and r02m ¼ r2 � ð1�ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

r2 � r020p

Þ:Squared correlation coefficient values between the

observed and predicted values of the test set compounds

(LOO predicted values for training set compounds) with

intercept (r2) and without intercept (r02) were calculated for

determination of rm2 Change of the axes gives the value of

r02m and the r02m metric is calculated based on the value of r020 :The r2

m and Drm2 matrices are applied for internal validation

of training set compounds (r2mðLOOÞ as well as Drm(LOO)

2 ,

external validation of test set compounds (r2mðtestÞ as well as

Drm(test)2 ) and overall validation for all compounds

(r2mðoverallÞ, Drm(overall)

2 ). QSAR models bearing acceptable

values for all the traditional parameters can be finally

assessed based on the rm2 metrics. Those with r2

m value

above the threshold of 0.5 and with a Drm2 value less than

0.2 are considered to be predictive and reliable ones.

Conflict of interest The authors declare no conflict of interest.

References

Bolton S (2001) Statistics. In: Gennaro AR (ed) Remington: the

science and practice of pharmacy, vol 1, 20th edn. Lippincott

Williams & Wilkins, Philadelphia, pp 124–158

Clarke SJ, Rivory LP (1999) Clinical pharmacokinetics of docetaxel. Clin

Pharmacokinet 36:99–114. doi:10.2165/00003088-199936020-00002

CS ChemOffice is software of Cambridge Soft Corporation, USA.

www.cambridgesoft.com

Darlington RB (1990) Regression and linear models. McGraw Hill,

New York

Durak I, Perk H, Kavutcu M, Combolat O, Akyal O, Beduk Y (1994)

Adenosine deaminase, 50 nucleotidase, xanthine oxidase, superoxide

dismutase and catalase activities in cancerous and non cancerous

human bladder tissues. Free Radic Biol Med 16:825–831

Eriksson L, Jaworska J, Worth AP, Cronin MT, McDowell RM,

Gramatica P (2003) Methods for reliability and uncertainty

assessment and for applicability evaluations of classification-

and regression-based QSARs. Environ Health Perspect

111:1361–1375. doi:10.1289/ehp.5758

Esterbauer H, Zollner H, Schauer RJ (1998) Hydroalkenals: cytotoxic

products of lipid peroxidation. Atlas Sci Biochem 1:311–319

Gardner HW (1975) Decomposition of linoleic acid hydroperoxides.

Enzymic reactions compared with nonenzymic. J Agric Food

Chem 23:129–136. doi:10.1021/jf60198a012

Guio Q, Zhao B, Li M, Shen S, Xin W (1996) Studies on protective

mechanisms of four components of green tea polyphenols

against lipid peroxidation in synaptosomes. Biochim Biophys

Acta 1304:210–222. doi:10.1016/S0005-2760(96)00122-1

Halliwell B (1991) Drug antioxidant effects—a basis for drug selection.

Drugs 42:569–605. doi:10.2165/00003495-199142040-00003

Hawkins DM (2004) The problem of overfitting. J Chem Inf Comput

Sci 44:1–12. doi:10.1021/ci0342472

Hilditch TP, Williams PN (1964) The chemical constituents of fats.

Chapman & Hall, London

Leibowitz ME, Johnson MC (1971) Relation of lipid peroxidation to

loss of cations trapped in liposomes. J Lipid Res 12:662–670

Luo X, Evrovsky Y, Cole D, Trines J, Benson LN, Lehotay DC

(1997) Doxorubicin-induced acute changes in cytotoxic alde-

hydes, antioxidant status and cardiac function in the rat.

Biochem Biophys Acta 1360:45–52

MINITAB version 14 is statistical software of Minitab Inc, USA.

http://www.minitab.com

Ohkawa H, Ohishi N, Yagi K (1979) Assay of lipid peroxides in

animal tissues by thiobarbituric reaction. Anal Biochem

95:351–353. doi:10.1016/0003-2697(79)90738-3

Ojha PK, Mitra I, Das RN, Roy K (2011) Further exploring rm2 metrics

for validation of QSPR models dataset. Chemom Intell Lab Syst

107:194–205

Pandey S, Sharma M, Chaturvedi P, Tripathi YB (1994) Protective

effect of Rubia cordifolia on lipid peroxide formation in isolated

rat liver homogenate. Indian J Exp Biol 32:180–183

Rang HP, Dale MM, Ritter JM, Moore PK (2003) Pharmacology, 5th

edn. Churchill Livingstone, London, pp 694–698

Roy PP, Roy K (2008) On some aspects of variable selection for

partial least squares regression models. QSAR Comb Sci

27:302–313. doi:10.1002/qsar.200710043

Snedecor GW, Cochran WG (1967) Statistical methods. Oxford &

IBH Publishing Co Pvt Ltd., New Delhi

SPSS version 9.0 software is statistical software of IBM Corporation.

http://www-01.ibm.com/software/analytics/spss

STATISTICA version 7 is statistical software of Stat Soft Inc. www.

statsoft.com

Wold S, Eriksson L (1995) Validation tools. In: van de Waterbeemd

H (ed) Chemometric methods in molecular design (methods and

principles in medicinal chemistry). Weinheim-VCH, New York,

pp 312–317

Yahya MD, Pinnsa JL, Meinke GC, Lung CC (1996) Antibodies

against malondialdehyde (MDA) in MRL/lpr/lpr mice: evidence

for an autoimmune mechanism involving lipid peroxidation.

J Autoimmun 9:3–9

Yang Z, Fong DW, Yin L, Wong Y, Huang W (2009) Liposomes

modulate docetaxel-induced lipid oxidization and membrane

damage in human hepatoma cells. J Liposome Res 19:122–130.

doi:10.1080/08982100802632649

Yap CW (2011) PaDEL-Descriptor: an open source software to

calculate molecular descriptors and fingerprints. J Comput Chem

32:1466–1474. doi:10.1002/jcc.21707

Med Chem Res

123