Pharmacophore modeling of substituted 1,2,4-trioxanes for quantitative prediction of their antimalarial activity

Article abstract of DOI:10.1021/ci100180e

A pharmacophore model has been developed for determining the essential structural requirements for antimalarial activity from the eight series of substituted 1,2,4-trioxanes. The best pharmacophore model possessing two aliphatic hydrophobic, one aromatic

Full text of DOI:10.1021/ci100180e

1510
J. Chem. Inf. Model. 2010, 50, 1510–1520
Pharmacophore Modeling of Substituted 1,2,4-Trioxanes for Quantitative Prediction of
their Antimalarial Activity†
Amit K. Gupta,‡ S. Chakroborty,‡ Kumkum Srivastava,§ Sunil K. Puri,§ and Anil K. Saxena*,‡
Medicinal and Process Chemistry Division and Parasitology Division, Central Drug Research Institute,
CSIR, Lucknow, 226001, India
Received May 8, 2010
A pharmacophore model has been developed for determining the essential structural requirements for
antimalarial activity from the eight series of substituted 1,2,4-trioxanes. The best pharmacophore model
possessing two aliphatic hydrophobic, one aromatic hydrophobic, one hydrogen-bond (H-bond) acceptor,
and one H-bond acceptor (lipid) feature for antimalarial activity showed an excellent correlation coefficient
for the training (r2
) 0.85) and a fair correlation coefficient for the test set (r2 ) 0.51) molecules.
training
test
The model predicts well to other known substituted 1,2,4-trioxanes including those which either are drugs
or are undergoing clinical trials. In order to further validate this model, five substituted 1,2,4-trioxanes were
synthesized from the generated focused library and screened for antimalarial activity. The observed activity
of these molecules was consistent with the pharmacophore model, suggesting that the model may be useful
in the design of potent antimalarial agents.
1. INTRODUCTION
eliminate artemisinin-resistant malaria in less than five years
in Combodia.11
In spite of worldwide efforts to combat malaria, ap-
proximately 247 million malaria cases were reported in 2006,
causing nearly a million deaths, with the majority in children
under five years.1 According published reports, 109 countries
were found endemic for malaria in 2008,1 and even in 2009,
the disease was responsible for a child’s death every 30 s.2,3
The major problem in the chemotherapy of malaria is the
development of resistance of the Plasmodium falciparum
malaria parasites to many of the standard quinoline antima-
larial drugs, like chloroquine.4
The exact mode of action of artemisinin and its analogues
is still unknown. However, one of the widely known
mechanisms is that, during the erythrocytic stage of the
parasite life cycle, iron(II) liberated during hemoglobin
degradation12 reduces the peroxide bond in the 1,2,4-trioxane
to form a sequentially highly unstable oxygen (O)-centered
radical, which then rearranges to a more stable carbon (C)-
centered free radical, high-valent iron oxo species, and
reactive epoxides. These reactive intermediates alone, or in
combination, kill the parasite, possibly by disrupting the vital
biochemical processes or by alkylating or oxidizing the
critical biomolecules.13 Others have argued that endoperoxide
cleavage is catalyzed by a cytoplasmic iron(II) source. The
resulting reactive species then specifically inhibits an ad-
enosine 5′-triphosphate-dependent (ATP-dependent) Ca2+
pump called PfATP6 located on the endoplasmic reticulum
and results in killing the parasite.14
The major drawback of all semisynthetic trioxanes is that
their synthesis needs artemisinin as a starting material, and
currently the production of artemisinin from natural source
remains relatively low. To address the supply issue, a number
of groups have attempted to produce totally synthetic 1,2,4-
trioxane analogues, some of which have demonstrated
remarkable antimalarial activity,15-19 while Vennerstrom and
co-workers have reported the first 1,2,4-trioxolane endop-
The discovery of artemisinin, an active principle of the
Chinese traditional drug Qinghaosu extracted from the plant
Artemisia annua, has opened a new era in the malarial
chemotherapy. Artemisinin and its more potent analogues
viz. artemether, arteether (Figure 1), and artesunic acid
represent the endoperoxide class of compounds which are
highly active against both chloroquine-sensitive and -resistant
strains of P. falciparum.5 These fast-acting drugs are suitable
for the treatment of cerebral malaria caused by multidrug-
resistant P. falciparum.6-8
Although the first incidence of stable artemisinin resistance
was reported in 2006 by Cravo et al.9 and thereafter in
December 2008, both were in the western part of Combo-
dia,10 and there is no evidence of artemisinin resistance
occurring anywhere else in the world. The World Health
Organization (WHO)-recommended artemisinin combination
therapy (ACT) is the best option available to date for the
chemotherapy of malaria, and according to a study carried
out by Maude et al., immediate implementation of ACT may
* Corresponding author. E-mail: anilsak@gmail.com. Telephone:
+9152226124112/ex 4386. Fax: +9152226123405.
† CDRI communication no. 7888.
‡ Medicinal and Process Chemistry Division.
§ Parasitology Division.
Figure 1. (1) Artemether, (2) arteether, and (3) artemisinin.
10.1021/ci100180e  2010 American Chemical Society
Published on Web 07/27/2010

PHARMACOPHORE DESIGN FOR ANTIMALARIAL ACTIVITY
J. Chem. Inf. Model., Vol. 50, No. 8, 2010 1511
Table 1. Structures of the Molecules Used in Training and Test Setsa
a C. N. ) compound name, and compounds marked with asterisk (*) are training set molecules, while the remaining are test set molecules.
eroxide, OZ-277, which has recently entered into clinical
trials.20 However, according to Uhlemann and co-worker,
although the mode of action of both types of antimalarials
(artemisinins and OZ-277) are similar in many aspects, as
both concentrate in intraerythrocytic parasites 200-300 fold
compared with extracellular concentrations, there are also
remarkable differences in their behavior particularly in the
potency to inhibit PfATP6 and the variability of OZ-277 in
the subcellular localization pattern.21
Few comparative molecular field analysis (CoMFA) stud-
ies on the artemisinin derivatives have been reported,22,23
Figure 2. Common structures of molecules used in pharmacophore
and a pharmacophore model was developed by Grigorov et
al. using bi- and tricyclic 1,2,4-trioxanes as antimalarial
agents.24 But to date, no pharmacophore modeling study has
been reported on the artemisinin family of substituted 1,2,4-
trioxanes.
Considering the importance of artemisinin and its ana-
logues as a potent class of antimalarial drugs effective against
the multidrug-resistant P. falciparam strains and unavail-
ability of the exact target for this class of molecule,25 it
appeared of interest to develop a pharmacophore model for
determining the essential structural requirements for this class
of molecules for their antimalarial activity. Such a model
may be useful in designing of this class of molecules for
antimalarial activity. The development and validation of such
a model is presented in this paper.
study.
homogeneity of the biological assays is one of the important
aspects in quantitative structure-activity relationship (QSAR)
studies, therefore, the data set was collected from the same
research group following the same biological testing protocol
and were rationally divided into a training set of 45
compounds (marked with an asterisk in Table 1) and a
complementary test set of 43 compounds according to
published guidelines.33 The structures of test and training
set molecules have been given in Figure 2 and Table 1.
The HypoGen algorithm of CATALYST resulted in the
generation of 10 alternative pharmacophores describing
the antimalarial activity of the training set compounds. The
details of the pharmacophore generation procedure have been
described in the Experimental Section of this manuscript.
The quality of the generated pharmacophore hypotheses were
evaluated by considering the cost functions calculated by
the HypoGen module during hypothesis generation. The
composition (in terms of chemical features that are necessary
for activity), ranking score, and statistical parameters as-
sociated with the hypotheses are reported in Table 2.
2. RESULTS AND DISCUSSION
2.1. Pharmacophore Description. The QSAR studies
were performed using eight series of substituted 1,2,4-
trioxanes comprising 88 artemisinin analogues (activity
ranges from 1.4 to 2000 nM) reported in literature26-31 by
adopting the HypoGen algorithm32 of CATALYST. The

1512 J. Chem. Inf. Model., Vol. 50, No. 8, 2010
GUPTA ET AL.
Table 2. Composition (Features), Cost (Bits), and Statistical Parameters (Rmsd and Correlation) Associated with the 10 Best Hypotheses
(Pharmacophore Models)
serial no.
hypothesis
featuresa
total cost
∆costb
rmsd
correlation (R)
1
2
3
4
5
6
7
8
9
10
Hypo 1
Hypo 2
Hypo 3
Hypo 4
Hypo 5
Hypo 6
Hypo 7
Hypo 8
Hypo 9
Hypo 10
HBA, HBAL, 2HYAl, HYAr
HBA, HBAL, HYAl, HYAr
HBA, HBAL, 2HYAl, HYAr
HBA, HBAL, HYAl, HYAr
HBA, HBAL, 2HYAl, HYAr
HBA, HBAL, 2HYAl, HYAr
HBA, HBAL, 2HYAl, HYAr
HBA, HBAL, HYAl, HYAr
HBA, HBAL, HYAl, HYAr
HBA, HBAL, 2HYAl, HYAr
174.7
181.8
183.1
184.6
185.2
185.3
185.5
185.6
186.0
186.1
28.81
21.7
20.4
18.9
18.3
18.2
18
17.9
17.5
17.4
0.60
0.80
0.85
0.88
0.86
0.83
0.84
0.86
0.89
0.93
0.92
0.86
0.83
0.82
0.82
0.84
0.84
0.82
0.81
0.80
a HBA: hydrogen-bond acceptor, HBAL: hydrogen-bond acceptor lipid, HYAl: hydrophobic aliphatic, and HYAr: hydrophobic aromatic.
b ∆cost ) (null cost - total cost).
of uncertainty 3, indicating the high predictive ability of the
pharmacophore (Figure 4).
Previously a four feature hypothesis was reported by
Grigorov et al. using bi- and tricyclic 1,2,4,-trioxanes, viz.,
one HBD, one HBA, and two HY features for the antima-
larial activity.24 Comparison of the above hypothesis with
our model depicted well-differentiated hydrophobic features
(two HYAl and one HYAr) with an additional HBAL feature.
However, the HBD feature was absent in our model, which
may be due to the absence of such functional groups in the
training set molecules. Although there are two CoMFA and
comparative molecular similarity indices analysis (CoMSIA)
studies have been reported earlier using semisynthetic
derivatives of artemisinin,22,23 their findings vary in terms
of steric and electrostatic contours distribution around the
molecule. These variations may be due to the limitations of
CoMFA and CoMSIA studies, as they are highly specific to
the molecular alignment. However, our study corroborates
with a study carried out by Jung et al. in terms of HYAl
feature location in a 3D space, where they described the
favorable steric bulk requirement, and corroborates with both
studies in terms of HBA feature location, where the favorable
partial negative charge requirement near the peroxide group
of trioxane moiety has been described for antimalarial activity
of artemisinin analogues.
Figure 3. (A) Three dimensional arrangement of pharmacophoric
features generated from the training set of 45 substituted 1,2,4-
trioxanes; HBA and HBAL as green, HYAl-1 and HYAl-2 as dark
blue; and HYAr as light blue. (B) Interfeature distance between
different features in the pharmacophore model.
The fixed cost of the 10 top-scored hypotheses was 166.6
bits, well separated from the null hypothesis cost of 203.5
bits. The top-ranked pharmacophore model (Hypo-1) had the
best predictive power and statistical significance described
by the high correlation coefficient (r2training ) 0.85), low root-
mean-square deviation (rmsd, 0.60), weight (1.13) and error
(159.403) costs and a cost difference of 28.81, satisfying
the acceptable range recommended in the cost analysis of
the CATALYST procedure.33 The configuration cost was
14.15, indicating that all generated models have been
thoroughly analyzed. The cost difference between total and
fixed costs for the best hypothesis was only 8.0 bits,
indicating the high probability of the true correlation of the
data. It is due to the fact that the lower the cost difference
between the total and fixed costs, the higher the probability
is for the true correlation of the data. Thus Hypo-1 was
retained for further analysis as the best pharmacophore model
for antimalarial activity with five features, viz., one H-bond
acceptor (HBA), one H-bond acceptor lipid (HBAL), two
aliphatic hydrophobic (HYAl), and one aromatic hydrophobic
(HYAr) function at specific geometric locations in three-
dimensional (3D) space (Figure 3A) and is also statistically
the most relevant model. The interfeature distance between
all these features has been shown in Figure 3B.
Pharmacophoric representation of the most potent com-
pound 40 of the data set showed the presence of 2-furyl group
at C-10 position provided HYAr function to the molecule,
whereas the trioxanes ring provided HBA and HBAL
function (Figure 5A). The three chain carbon linked with
the C-3 and C-12a positions of the trioxane ring and the
methyl group at C-6 position of the molecule 40 provided
the HYA1-1and HYAl-2 functions, respectively, to the
molecule. The same study carried with the least active
compound of the data set 78 (IC50 ) 2000 nM) depicted the
absence of HYAl-2 function mapping due to absence of an
alkyl group at the C-6 position of the molecule, while at the
same time, the molecule failed to access the HBAL feature
in order to fit other features of the pharmacophore (Figure
5B).
The pharmacophore model mapped very well to the
training set molecules. The observed IC50 along with
predicted IC50 values for the antimalarial activity of the
training set compounds are given in Table 3.
A plot between the pIC50 values of observed versus
estimated activity demonstrated a good correlation coefficient
(r2training ) 0.85) for training set molecules within the range
It has been observed from the mapping of the training set
molecules to the pharmacophore that, among all the active
trioxanes, the peroxide group of trioxane ring provided a
HBA feature, while the oxygen atom at position 4 of 1,2,4-
trioxane motif provided a HBAL feature. The importance
of peroxide group in trioxane moiety is well-known since it

PHARMACOPHORE DESIGN FOR ANTIMALARIAL ACTIVITY
J. Chem. Inf. Model., Vol. 50, No. 8, 2010 1513
Table 3. Observed and Predicted Activity Values of the Compounds in the Training Set
obs. activity
(IC50) nM
pred. activity
(IC50) nM
obs. activity
(pIC50) nM
pred. activity
(pIC50) nM
obs. activity
scaleb
pred. activity
scale
serial no.
comp. name
errora
1
2
3
4
5
6
7
8
1
2
3
5
8
110
76
68
44
600
78
51
79
44
23
97.17
101.46
123.26
64.12
102.89
73.32
36.67
65.19
76.33
35.17
57.99
66.82
85.74
194.77
406.55
549.03
4.7
4.53
5.33
3.63
4.13
3.49
5.94
7.79
6.84
-2.04
-1.88
-1.83
-1.64
-2.78
-1.89
-1.71
-1.9
-1.64
-1.36
-1.81
-1.48
-1.92
-2.51
-2.2
-2.98
-0.62
-0.82
-0.89
-0.15
-0.71
-0.66
-0.96
-0.92
-0.89
-0.66
-0.65
-1.2
-0.91
-1.75
-1.52
-1.95
-1.63
-2.04
-1.36
-0.89
-3.11
-2.08
-3.3
-1.99
-2.01
-2.09
-1.81
-2.01
-1.87
-1.56
-1.81
-1.88
-1.55
-1.76
-1.82
-1.93
-2.29
-2.61
-2.74
-0.67
-0.66
-0.73
-0.56
-0.62
-0.54
-0.77
-0.89
-0.84
-0.69
-0.81
-0.86
-0.82
-2.04
-1.5
-1.13
1.33
1.81
++
++
++
++
++
++
1.45
+++
+
++
-5.83
-1.06
-1.39
-1.21
1.73
1.53
-1.12
2.22
++
9
++
++
12
13
14
17
18
19
24
25
26
30
36
37
38
40
44
45
48
52
54
57
58
60
61
64
65
67
69
70
71
73
74
76
78
79
80
82
83
85
86
++
+++
++
++
9
+++
+++
++
++
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
+++
++
65
30
84
+++
++
++
1.02
-1.64
2.54
++
320
160
960
4.2
6.6
7.8
1.4
5.1
4.6
9.1
8.4
7.8
4.6
4.5
16
+
+
+
+
-1.74
1.12
-1.46
-1.47
2.59
-1.23
-1.32
-1.53
-1.07
-1.14
1.05
+
+
+++
+++
+++
+++
+++
+++
+++
+++
+++
+++
+++
+++
+++
++
+++
+++
+++
+++
+++
+++
+++
+++
+++
+++
+++
+++
+++
++
4.85
6.49
7.27
6.55
1.45
-2.19
-1.23
1.94
-1.03
1.17
1.53
-2.26
1.39
8.1
56
33
89
43
108.97
31.82
104.13
66.16
48.62
32.14
51.47
214.34
72.56
1299.85
364.19
368.25
9.92
+++
++
++
++
+++
++
++
-2.02
-1.82
-1.69
-1.51
-1.71
-2.33
-1.86
-3.11
-2.56
-2.57
-1
110
23
+++
+++
++
+++
+++
+
7.7
1300
120
2000
230
600
25
31
11
13
6.68
-6.06
-1.65
-1.53
1.58
-1.63
-2.52
1.53
+
++
++
+
+
-2.36
-2.78
-1.4
-1.49
-1.04
-1.11
+
+
+
+
+++
++
+++
++
47.45
10.03
7.5
-1.68
-1
-0.88
-1.1
-1.73
+++
+++
+++
+++
a Values in the error column represent the ratio of predicted activity to observed activity, or its negative inverse if the ratio is less than 1.
b Activity scale: highly active (<50 nM, +++), moderately active (51-200 nM, ++), and less active (>200 nM, +).
Figure 5. (A) Mapping of the most active compound of the data
Figure 4. Correlation displaying the observed versus estimated
pIC50 (nM) values in the 0 to -3.5 range for the training set of 45
(shown in red color) and test set of 43 molecules (shown in black
color).
set 40 onto the pharmacophore. (B) Pharmacophore mapping of
the least potent analogue of the data set 78 showing two missed
features of the pharmacophore, i.e., HBAL and HYAl-2 (shown as
nonhighlighted features).

1514 J. Chem. Inf. Model., Vol. 50, No. 8, 2010
GUPTA ET AL.
set containing 43 substituted 1,2,4-trioxanes. The observed
IC50 along with predicted IC50 values of the test set
compounds are given in Table 4. The overall correlation
coefficient value of (r2test ) 0.51) between the observed and
estimated activity of the 43 test set molecule was indicative
of its good predictive quality (Figure 4). Among the test set
molecules, the molecule 27 (obs. act 8.2 and pred. act ∼422)
has shown maximum deviation in our QSAR analysis. This
compound had also shown high activity as compared to its
other closely related structural analogues in the reported
paper.26 Its removal improved the r2 value from 0.51 to 0.61.
This molecule follows the same pharmacophore mapping
trend as others of its series analogues, where it has also
missed the HBAL and HYAl-2 feature. The poor prediction
(poor mapping) of this molecule seemed to be due to the
inability of these molecules to achieve an energetically
favorable conformation as well as the interfeature distances
(geometrical distance) between the important features found
in the best pharmacophore.
In the most potent analogue of the test set 49, the trioxane
ring provided HBA and HBAL features, while the three chain
carbon linked with the C-3 and C-12a positions and the
methyl group at C-6 of the 49 provided the HYA1-1and
HYAl-2 functions, respectively, to the molecule. The missed
HYAr feature was compensated by the high-fitted score
values of other essential features, viz., one HBAL, one HBA,
and two HYAl, for ranking this molecule as active. The least
potent member of the test set (75) failed to map the HYAr
and HYAl-1 features of the hypothesis completely and hence
was ranked as least active.
Figure 6. The cost difference of between the Hypo-1(blue color)
and the scrambled runs (other colors).
is responsible for the generation of free radicals. A study
led by Wei et al. has shown that artemisinin treatment of
membranes (especially in the presence of heme) can cause
lipid damage.34 The oxygen atom at position 4 in the 1,2,4-
trioxane moiety seemed to be important in this context, as
described by the lipid kind of HBAL feature of the
pharmacophore as it may cause the phospholipid instability
and the membrane damage in the parasite. In view of above,
it may be concluded that both of these features are crucial
for antimalarial activity of substituted 1,2,4-trioxane mol-
ecules. Further, on the basis of the mapping experiment of
the training set molecules, it may be concluded that the
addition of a bulky aliphatic group at the region joining C-3
to C-12a and C-6 positions, while addition of aromatic groups
at C-10 position of the substituted 1,2,4-trioxanes, may lead
to a significant increase in antimalarial activity.
2.2. Evaluation of the Hypogen Model. 2.2.1. Score
Hypothesis. The training set was submitted to a score
hypothesis process. Within this procedure, the activity of
every single training set molecule was estimated by the
hypothesis. The actual and estimated antimalarial activity
of the 45 training set compounds based on Hypo-1 are listed
in Table 3, where 42 molecules out of the 45 in the training
set have errors less than 3, and only three molecules (8, 73,
and 74) have errors less than 7, confirming that our
hypothesis is a reliable model for describing the structure-
activity relationship (SAR) in the training set.
2.4. External Test Set Validation. 2.4.1. Validation
through Known Synthetic Substituted 1,2,4-Trioxanes. As the
main purpose of the quantitative model is to identify the
active structures through virtual screening, hence to check
the predictive ability of the pharmacophore model, an
external data set comprising first-, second-, and third-
generation synthetic substituted 1,2,4-trioxanes36 was sub-
mitted to pharmacophore mapping. External test set mol-
ecules, along with their observed and predicted activity, have
been given in Table 5, where the predicted values well
described the observed antimalarial activity of these
compounds.
2.4.2. Validation of the Generated Hypothesis for Its
Predictability against Clinically ActiVe Trioxanes. The
validity of this pharmacophore model was further examined
on highly active trioxane molecules which either are drugs
or are undergoing clinical trials, viz., artemisinin,26 arte-
mether,37 arteether,26 artesunate,21 artelinate, and artemi-
sone.38 The observed and predicted activity of these mol-
ecules are listed in Table 6. Interestingly these molecules
mapped well to the features of the best hypotheses, and the
model also ranked them as active. Since the model discrimi-
nated well between active and inactive, it may be considered
for virtual screening to prioritize the molecules for the
synthesis and evaluation as antimalarial agents.
The poor predicted values for two compounds of the
external test set and for artemisone of the clinically active
compounds may be due to the following reasons. The
biological activity of these molecules has came from different
research groups/laboratories and has not been determined
under similar conditions than those of training set molecules,
hence the chances of error are expected in their activity
2.2.2. Cat Scramble Validation (Fisher Test). To evaluate
the statistical relevance of the model, the Fisher’s random-
ization test has been applied. This test involves thorough
randomization of the training set to validate and derive the
significance of the generated best model. As a consequence,
the pharmacophore model corresponding to the Hypo-1 was
evaluated for the statistical significance using a randomization
trial procedure derived from the Fisher method.35 These
randomized spreadsheets should yield hypotheses with lesser
statistical significance than the original model to suggest that
the original hypothesis represents a true correlation. The
number of such random trials depends on what level of
statistical significance is to be achieved. For a 95% confi-
dence level, 19 spreadsheets are created, while for 98% and
99% confidence levels, 49 and 99 spreadsheets are created,
respectively. Our model was found to be 99% significant in
the F-randomization test which substantiates the significance
of the model (Figure 6).
2.3. Test Set Validation. The robustness of the generated
model was further assessed by its predictability on the test

PHARMACOPHORE DESIGN FOR ANTIMALARIAL ACTIVITY
J. Chem. Inf. Model., Vol. 50, No. 8, 2010 1515
Table 4. Observed and Predicted Activity Values of the Compounds in the Test Set
obs. activity
(IC50) nM
pred. activity
(IC50) nM
obs. activity
(pIC50) nM
pred. activity
(pIC50) nM
obs. activity
scaleb
pred. activity
scale
serial no.
comp. name
errora
1
2
3
4
5
6
7
8
4
6
7
170
49
55
15
39
20
42
99
34
38.99
116.01
158.42
39.23
68.76
29.87
16.38
15.43
46.12
71.72
28.97
422.73
319.77
301.95
425.1
115.84
8.87
232.44
138.55
5.55
4.76
3.93
3.06
0.93
2.87
5.67
6.48
7.48
7.48
8.02
8.79
5.17
9.48
7.27
26.8
246
20.95
642.62
301.95
8.52
4.3
12.36
73.92
-2.23
-1.69
-1.74
-1.18
-1.59
-1.30
-1.62
-2
-1.53
-1.59
-1.72
-0.92
-2.38
-2.53
-2.85
-2.2
-2.04
-1.66
-1.58
-0.95
-0.72
-0.93
-1
-1.59
-2.06
-2.2
-4.36
2.37
2.88
2.61
1.76
++
+++
++
+++
++
++
10
11
15
16
20
21
22
23
27
28
29
31
32
33
34
35
39
41
42
43
46
47
49
50
51
53
55
56
59
62
63
66
68
72
75
77
81
84
87
88
-1.59
-1.84
-1.48
-1.21
-1.19
-1.66
-1.86
-1.46
-2.63
-2.5
-2.48
-2.63
-2.06
-0.95
-2.37
-2.14
-0.74
-0.68
-0.59
-0.49
0.03
-0.46
-0.75
-0.81
-0.87
-0.87
-0.9
-0.94
-0.71
-0.98
-0.86
-1.43
-2.39
-1.32
-2.81
-2.48
-0.93
-0.63
-1.09
-1.87
+++
+++
+++
+++
++
+++
++
1.49
+++
+++
+++
+++
++
-2.56
-6.42
1.36
9
+++
+++
++
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
39
53
1.84
-1.83
50.93
1.33
-1.13
-1.67
-1.38
-12.41
5.05
+++
+
8.3
240
340
710
160
110
46
38
9
5.2
8.6
10
16
9.4
4
11
8.3
14
37
4.3
28
11
10
+++
+
+
+
+
+
+
++
++
++
+++
+
+++
+++
+++
+++
+++
+++
+++
+++
+++
+++
+++
+++
++
3.64
++
-1.62
-1.09
-2.19
-3.27
-17.26
-3.27
1.41
+++
+++
+++
+++
+++
+++
+++
+++
+++
+++
+++
+++
+++
+++
+++
+++
+
-1.2
-0.97
-0.6
-1.04
-0.92
-1.15
-1.57
-0.63
-1.45
-1.04
-1
-1.77
-1.48
-1.4
-3.23
-2.49
-1.15
-0.84
-1.28
-2.38
-1.7
-1.11
-1.87
-4.62
2.04
+++
++
+++
+++
++
-5.41
-1.16
-1.37
-2.2
8.2
-1.2
-2.64
-1.02
-1.64
-1.6
-1.54
-3.25
59
30
25
1700
310
14
6.9
19
+++
+++
+
+++
+
+
+++
+++
+++
++
+
+++
+++
+++
+
240
a Values in the error column represent the ratio of predicted to observed activity, or its negative inverse if the ratio is less than 1. b Activity
scale: highly active (<50 nM, +++), moderately active (51-200 nM, ++), and less active (>200 nM, +).
predictions. The other reason for the outlier behavior of these
compounds may be due to their inability to achieve an
energetically favorable conformation as well as the inter-
feature distances (geometrical distance) between the impor-
tant features found in the best pharmacophore.
in a user-friendly interface. The “Protocores” (Figure 7)
consistent with 1,2,4-trioxane ring were considered for the
library generation using the default parameter setting imple-
mented in the CombiGlide.
A fragment set was prepared in the “Reagent Preparation
Wizard” of the maestro panel of Schro¨dinger software, where
among 41 different substitution reactions with reactants, the
“R-I” type of reactants have been chosen for reagent
preparation. Based on the pharmacophoric features require-
ment various hydrophobic substituent’s (R), viz., cyclohex-
ane, cyclopentane, ethane, propane, butane, benzene, dim-
ethylbenzene, trimethylbenzene, 2-methoxybenzene, o-tolyl-
methanol, o-tolylethanol, naphthalene, phenanthrene, indole,
thiophene, coumaron, and benzisoxazole were drawn and
used for the reagent preparation. The focused virtual library
generation was carried out using “Combinatorial Library
Enumeration” wizard of the maestro panel where the virtually
prepared reagents were selectively added to attachment points
“R1” and “R2” (Figure 7) which resulted into 600 new
2.5. Focused Virtual Library Screening. In silico screen-
ing is not restricted to commercial databases of existing
compounds but can be performed with focused virtual
libraries containing unknown compounds for new lead
discovery, synthesis, and further lead optimization. The high
extent of structural diversity, the high degree of drug likeness,
and eventually the synthetic accessibility are among the
criteria that may be useful for virtual library generation.
Several methods to screen the focused virtual library with
desired properties have been recently discussed in a review.39
Among these methods, the fragment-based method for the
generation of a focused virtual library has been used in our
studies, using the software CombiGlide40 which allows the
user to a run fragment-based generation of a virtual library

1516 J. Chem. Inf. Model., Vol. 50, No. 8, 2010
GUPTA ET AL.
Table 5. Structures of the Known Synthetic Trioxanes along with Their Observed and Predicted Activity Based on Hypo-1a
a Obs. A. ) observed activity, and Pred. A. ) predicted activity.
Table 6. Observed and Predicted Activity of Clinically Active
Trioxanes Based on Hypo-1
Table 7. Observed and Predicted Activity of Five Synthesized
Trioxanes along with Standard Drugs Based on Hypo-1
observed
IC50 (nM)
predicted
IC50 (nM)
serial
no.
observed activity predicted activity
a
serial no.
name
compound
5a
6a
6b
7a
IC50 (nM)
IC50 (nM)
LogP
1
2
3
4
5
6
artemisinin
artemether
arteether
artesunate
artelinate
artemisone
11
6.8
8.2
7.4
6.7
0.7
1
2
3
5
6
7
8
9
35 ( 2.375
76 ( 5.88
83.66
18
4.516
4.444
3.523
4.796
3.852
2.422
2.845
5.875
1.2
6.1
1.6
-
85.40 ( 4.93
55.70 ( 4.42
47.70 ( 3.85
2.20 ( 0.13
2.84 ( 0.22
16.3 ( 1.10
14.30
20.68
14.61
8.20
7.40
-
7b
0.88
33.6
artemether
arteether
chloroquine
a IC50: concentration corresponding to 50% growth inhibition of
chloroquine sensitive strain 3D7 of P. falciparum (data are
expressed as means ( SD from three different experiments in
duplicate).
trioxanes. The whole process of the generation of a focused
virtual library was carried out on a Red Hat Enterprise Linux
(RHEL) 5 based Operating system having an Intel Pentium
dual core 2.8 GHz processor. The generated virtual library
was submitted to the build database protocol implemented
in the Discovery Studio 2.0 (DS 2.0), which provided a
multiconformer database. The 3D quantitative pharmacoph-
ore Hypo-1 was used to estimate the biological activities of
the above database in DS 2.0, followed by a drug-likeness
screen to ensure high drug-likeness of the compounds. On
the basis of their predicted IC50 values, the top 50 scoring
molecules were considered as active new hits. In order to
further evaluate the predictability of the pharmacophore
model, five substituted 1,2,4-trioxanes among the top 50
scoring hits were synthesized and evaluated for their anti-
malarial activity (Table 7).
2.6. Synthesis. Allylic alcohols (3a-c) were prepared
from the corresponding ketones (1a-c) according to the
reported procedure41 and were photo-oxygenated to give
ꢀ-hydroxy hydroperoxides (4a-c). The photo-oxygenated
products of allylic alcohols (ꢀ-hydroxy hydroperoxides) were
not isolated and were condensed in situ, where 4a on
condensation with 2-adamantanone gave trioxane 5a, while
4b and 4c in a similar reaction with norcamphor and acetone
gave trioxanes 6a-b and 7a-b, respectively, in the presence
of a catalytic amount of HCl to furnish these trioxanes with
49-58% overall yields.
Figure 7. Protocores (a,b and c) used in the focused virtual library
generation. R1 and R2 denote the attachment points where prepared
reagents have been selectively attached.

PHARMACOPHORE DESIGN FOR ANTIMALARIAL ACTIVITY
J. Chem. Inf. Model., Vol. 50, No. 8, 2010 1517
2.7. Biological Evaluation. All the five synthesized
compounds along with the standard drug artemether, ar-
teether, and chloroquine were evaluated for their in vitro
antimalarial activity against chloroquine (CQ) sensitive 3D7
strain of P. falciparum BY SYBER green I-based fluores-
cence (MSF) assay42 (Table 7).
As shown in Table 7, the reported pharmacophore has been
successfully applied in predicting the in vitro antimalarial
activity of 1,2,4-trioxanes, where the error factor between
estimated and experimental activity of the three molecules
(viz., 5a, 7a, and 7b) out of five were ∼3 (the uncertainty
ratio of the CATALYST), while two molecules, viz., 6a and
6b have the error factor of 4.2 and 6, respectively.
pharmacophore model may be useful in the design of
substituted trioxanes as potent antimalarial agents.
4. EXPERIMENTAL SECTION
4.1. Pharmacophore Modeling. Pharmacophore genera-
tion was performed on a Windows-based operating system
having an Intel Pentium dual core 2.8 GHz. processor using
the HypoGen algorithm32 of CATALYST implemented in
the Discovery Studio 2.0 software package (DS 2.0).43 All
compounds used in the QSAR study were built using ISIS
Draw 2.5 and imported to DS 2.0 Windows, where the
stereochemistry in the molecules was specified. These
compounds were optimized using the CHARMm force field,
and a maximum of 255 conformations were generated using
the “best” conformational search, using the “Poling” algo-
rithm44 within an energy threshold of 20.0 kcal/mol with
respect to the global minimum. A minimum number of 16
compounds is recommended for the training set for assured
statistical power of the generated hypothesis with HypoGen
algorithm.31 The important aspect of the training set com-
pounds selection is that each active compound should provide
new information to the HypoGen to help it uncover as much
critical information as possible for predicting biological
activity. The training set molecules associated with their
conformations were submitted to the pharmacophore genera-
tion procedure (HypoGen).
The pharmacophore mapping experiment of the five
synthesized compounds revealed that aliphatic substituents
at position 3′ of trioxane ring, viz., admantyl, cycloheptane,
and dimethyl groups, provided the HYAl functionality, while
2,5-dimethyl phenyl, phenyl, and 2-methoxy phenyl groups
provided the HYAr functionality to the molecule. The
trioxane ring itself provided the HBA and HBAL function-
alities to the molecule. The active molecules (5a and 7b)
among the five are simple trioxanes which contained the
required pharmacophoric features.
To check the absorption, distribution, metabolism, and
excretion (ADME)/toxicity of the synthesized compounds,
the five synthesized compounds were submitted to the
Mobyle server based program for ADME/toxicity filtration,
where all the synthesized compounds passed these filtration
criteria necessary for drug-likeness (http://mobyle.rpbs.univ-
paris-diderot.fr/cgi-bin/portal.py?form)admetox).
The log P value in the water/octanol system has been
calculated for all five synthesized trioxanes in the QuickProp
protocol of Schro¨dinger software (Table 7), where it has been
found that their Log P value is comparable with the most
active molecule of the training set (molecule 40, Log P )
3.57) as well as some clinically active molecules. (viz.,
artelinate, Log P ) 3.79; and arteether, Log P ) 2.85).
Further, all the five synthesized compounds were soluble in
ethanol, dimethyl sulfoxide (DMSO), and dichloromethane
(DCM).
HypoGen algorithm allows identification of hypotheses
that are common to the “active” molecules in the training
set but at the same time not present in the “inactives”.32 This
algorithm allows a maximum of five features in the phar-
macophore generation out of an inbuilt collection of 11,
which are HBA, HBAL, HBD, HY, HYAl, HYAr, positively
(PC) and negatively (NC) charged, positively (PI) and
negatively (NI) ionizable, and ring aromatic (RA) features.
Among these features available for selection, four features,
viz., HBA, HBAL, HYAl, and HYAr present in the training
set were selected for pharmacophore generation. Pharma-
cophore generation was carried out by using the default
parameters and the setting implemented in the HypoGen
generation procedure of the CATALYST except for the
interfeature distance, where a default value (2.97 Å) was
reduced to 2 Å due to small molecular size of the active
compounds used in the training set. The choice and number
of features used in the hypothesis construction were HBA
(min. 1 - max. 5), HBAL (min. 1 - max. 5), HYAr (min.
1 - max. 5), and HYAl (min. 1 - max. 5). A default activity
uncertainty value of 3 has been used in the pharmacophore
generation. The specifications regarding the pharmacophore
generation have been well documented by Kristam et al. to
perform the reproducibility of the pharmacophore.45
In summary, the reported pharmacophore was found useful
in the prediction of antimalarial activity of substituted 1,2,4-
trioxanes with significantly high reliability, which provided
a cost-effective and significant solution for identifying novel
hits.
3. CONCLUSION
In the above study a quantitative pharmacophore model
has been developed using a training set of 45 substituted
1,2,4-trioxanes. The best hypothesis consisted five features,
viz., two HYAl, one HYAr, one HBA, and one HBAL with
an excellent correlation (r2training ) 0.85) for the training set
and fair (r2test ) 0.51) for test set of 43 substituted trioxanes.
The model also well explained the observed antimalarial
activity of known synthetic trioxanes as well as clinically
effective antimalarial drugs. In order to further validate this
model, five substituted 1,2,4-trioxanes were synthesized from
the generated focused library and screened for antimalarial
activity, where the observed activity of these molecules was
consistent with the pharmacophore model. Accordingly, the
An uncertainty ∆ in the CATALYST paradigm indicates
an activity value lying somewhere in the interval from
“activity divided by ∆” to “activity multiplied by ∆”. The
resultant best 10 generated hypotheses were assessed on
the basis of the cost relative to the null hypothesis and the
correlation coefficients. The total cost value is the summation
of weight, error, and a configuration cost value. The weight
component is a value that increases in a Gaussian form
because of the deviation of function weights from the ideal
value of 2. Increase of the root-mean-square (rms) divergence
between estimated and measured activity for the training set

1518 J. Chem. Inf. Model., Vol. 50, No. 8, 2010
GUPTA ET AL.
molecules results in a higher error cost value. The config-
uration cost is a fixed cost that quantifies the entropy of the
hypothesis space. In the standard HypoGen mode, the
configuration cost should not exceed a maximum value of
17 (corresponds to a number of 217 pharmacophore models)
because high values may lead to chance correlation of the
generated hypothesis, since CATALYST cannot consider
more than 217 models in the optimization phase, and so the
rest are left out of the process. The HypoGen module
performs two additional theoretical cost calculations (rep-
resented in bit units) to help the user in assessing the
statistical significance of the generated hypothesis. The fixed
cost is the lowest possible cost representing a hypothetical,
simplest model that fits all data perfectly. It is calculated by
adding the minimum achievable error, the weight cost, and
the constant configuration cost. The null cost represents the
maximum cost of a pharmacophore with no features and
estimates activity to be the average of the training set
molecule’s activity data. Its absolute value is equal to the
maximum occurring error cost. The statistical significance
of the hypothesis increases when the total cost value is close
to the fixed cost value and when the difference between null
and total costs is high. The estimation of the activity of each
training set compound is based on the regression analysis
between the parameters computed by using the relationship
of the geometric fit value versus the negative logarithm of
activity, and so the higher the geometric fit value is, the
greater the activity prediction is of the compound. The fit
function does not depend only on the mapping of the feature
but also contains a distance term measuring the distance
between the feature on the molecule and the centroid of the
hypothesis feature, and both these terms are used in the
calculation of geometric fitness. The method has been
documented to perform better than a structure-based phar-
macophore generation.46,47
for 10 h. The crude product ꢀ-hydroxy hydroperoxide
obtained was directly subjected to condensation with 2-ada-
mantanone (2.6 mL) in CH3CN (80 mL) and CH2Cl2 and
1-2 drops of conc. HCl were added, and the reaction mixture
was stirred for 2.5 h at room temperature. The reaction
mixture was concentrated to yield 3.81 g of the crude
product. It was chromatographed on silica gel (150 g) using
EtOAc:hexane, (0.5:99.5) as an eluent to furnish 2.37 g (58%
yield, based on allylic alcohol) pure trioxane 5a as a white
solid (mp ) 58 °C).
Trioxanes 6a-b were prepared from allylic alcohols 3b,
while trioxanes 7a-b were prepared from allylic alcohols
4b in a similar way by replacing 2-adamantanone with
norcamphor (560 mg, for trioxane 6a and 7a) and acetone
(5 mL, for trioxane 6b and 7b), respectively.
4.4. Trioxane 5a. Yield 58%, white solid, mp ) 57-59
°C, FT-IR (KBr, cm-1) 3460.7, 2361.5, 1597.9, 1458.2,
1
1245.6, 1109.3, 1000.7, 917, 757; H NMR (2918.2, 300
MHz, CDCl3), δ 1.57-2.09 (m, 13H), 2.96 (bs, 1H), 3.81(dd,
1H, J ) 11.8 and 3.3 Hz), 3.85 (s, 3H), 3.92 (dd, 1H, J )
11.8 and 10.3 Hz), 5.2 (dd, 1H, J ) 10.3 and 2.3 Hz), 5.33
(s, 1H), 5.43 (s, 1H), 6.89-6.95 (m, 2H), 7.13-7.16 (m,
1H), 7.29-7.35(m, 1H); ESMS (m/z): 343.1 [M + H]+;
HRMS for fragment ion [M + 2H]2+ C21H28O4: 344.1979
(obs.); 344.1988 (calcd.).
4.5. 6′-(1-(2,4-Dimethylphenyl)vinyl)spiro[bicyclo[2.2.1]-
heptane-2,3′-[1,2,4]trioxane] (6a). Yield 55%, yellow liquid
FT-IR (KBr, cm-1) 3015.3, 2966.9, 2874.0, 2362.1, 1450.8,
1328.6, 1217.5, 1132.3, 1105.8, 1055.1, 968.6, 824.2, 760.1,
669.2; 1H NMR (300 MHz, CDCl3), δ 1.27-1.80 (m, 10H),
2.28 (s, 3H), 2.33 (s, 3H), 3.60-3.88 (m, 2H), 4.89-5.01(m,
1H), 5.15 (s, 1H), 5.44 (s, 1H), 6.94-7.35 (m, 3H); ESMS
(m/z): 300.4 [M]+; HRMS for fragment ion [M + H]1+
C19H25O3: 301.1785(obs.); 301.1804 (calcd.).
4.6. 6-(1-(2,4-Dimethylphenyl)vinyl)-3,3-dimethyl-1,2,4-
trioxane (6b). Yield 49%, Yellow liquid; FT-IR (KBr, cm-1)
3428.8, 3018.8, 2927.4, 2361.2, 1612.3, 1449.9, 1376.5,
4.2. Chemistry. Melting points were determined in open
capillaries on a COMPLAB melting point apparatus and are
uncorrected. IR spectra were recorded on a Perkin-Elmer RXI
FT-IR spectrophotometer. 1H NMR was recorded on a Bruker
1
1215.0, 1162.0, 1049.9, 927.6, 760.7, 669.8; H NMR(200
MHz, CDCl3), δ 1.36 (s, 3H), 1.63 (s, 3H), 2.26 (s, 3H),
2.31(s, 3H), 3.687 (dd, 1H, J ) 11.84 and 3 Hz), 3.87 (dd,
1H, J ) 11.87 and 10.2 Hz), 4.93(dd, J ) 10.24 and 2.82
Hz), 5.17 (s, 1H), 5.47(s, 1H), 6.96-7.26 (m, 3H); ESMS
(m/z): 247.2 [M - H]-, HRMS for fragment ion C15H19O3
[M - H]- 247.1307(obs.); 247.1334 (calcd.).
1
Supercon Magnet DPX-200 (operating at 200 MHz for H)
spectrometer using CDCl3 as the solvent. Tetramethylsilane
1
(0.00 ppm) served as an internal standard in H NMR.
Multiplicities are represented by s (singlet), d (doublet), dd
(doublet of doublet), t (triplet), q (quartet), bs (broad singlet),
and m (multiplet). High-resolution electron impact mass
spectra (HR-EIMS) were obtained on a JEOL JMS-600H
instrument. For all trioxanes, satisfactory high-resolution
mass spectra were obtained, confirming purity. Reactions
were monitored on silica gel thin-layer chromatography
(TLC) plates. Column chromatography was performed over
silica gel (60-120 mesh) procured from Qualigens (India)
using freshly distilled solvents. All the chemicals and
reagents were obtained from Aldrich (USA), Lancaster
(England), or Spectrochem (India) and were used without
purification.
4.3. Trioxane 5a. 3-(2-Methoxyphenyl)but-2-en-1-ol (1
g) was dissolved in acetonitrile (60 mL) in a double-jacketed
round-bottomed flask assembled with a cryostat (instrument
used for maintaining low temperature between -10 and 0
°C) and methylene blue (10 mg) was added. The reaction
mixture was irradiated with a tungsten halogen filament lamp
(500 W), and oxygen was bubbled into the reaction mixture
4.7. (Z)-6′-(1-Phenylpent-1-enyl)spiro[bicyclo[2.2.1]heptane-
2,3′-[1,2,4]trioxane] (7a). Yield 56%, Yellow liquid, FT-IR
(KBr, cm-1) 3020.4, 2963.2, 2361.7, 1680.9, 1647.7, 669.5,
1
1328, 1216, 1107.3, 762.0; H NMR (300 MHz, CDCl3), δ
0.97(t, 3H, J ) 7.26 Hz), 1.27-1.79 (m, 12H), 2.34-2.39
(m, 2H), 3.52-3.87 (m, 2H), 5.41-5.51 (m, 1H), 5.84 (t,
1H, J ) 7.53 Hz), 7.47-7.99 (m,5H); ESMS (m/z): 315.2
[M + H]+; HRMS for fragment ion [M - H]- C20H25O3:
313.1811(obs.); 313.1804 (calcd.).
4.8. (Z)-3,3-Dimethyl-6-(1-phenylpent-1-enyl)-1,2,4-triox-
ane (7b). Yield 52%, Yellow liquid, FT-IR (KBr, cm-1)
3376.0, 3019.6, 2962.2, 2873.5, 1732.1, 1374.9, 1215.4,
1
1161.4, 1048.7, 903.1, 759.5, 668.4; H NMR (300 MHz,
CDCl3), δ 1(t, 3H, J ) 7.32 Hz), 1.39 (s, 3H), 1.49-1.57
(m, 2H), 1.61 (s, 3H), 2.33-2.41 (m, 2H), 3.53(dd, 1H, J )
12 and 3 Hz), 3.93 (dd, 1H, J ) 11.94 and 11.1 Hz), 5.42
(dd, 1H, J ) 11 and 3 Hz), 5.82 (t, 1H, 7.59 Hz), 7.28-7.35

PHARMACOPHORE DESIGN FOR ANTIMALARIAL ACTIVITY
J. Chem. Inf. Model., Vol. 50, No. 8, 2010 1519
(m, 5H); ESMS (m/z): 261.1[M - H]-; HRMS for C16H22O3:
262.1548(obs.); 262.1569 (calcd.).
(7) Ploypradith, P. Development of Artemisinin and its Structurally
Simplified Trioxane Derivatives as Antimalarial Drugs. Acta Trop.
2004, 89, 329–342.
(8) White, N. J. Qinghaosu (Artemisinin): The Price of Success. Science
2008, 320, 330–334.
(9) Afonso, A.; Hunt, P.; Cheeseman, S.; Alves, A. C.; Cunha, C. V.;
Rosario, V. do; Carvo, P. Malaria Parasites can Develop Stable
Resistance to Artemisinin but Lack Mutations in Candidate Genes atp6
(Encoding the Sarcoplasmic and Endoplasmic Reticulam Ca2+
ATPase), tctp, mdr1 and cg10. Antimicrob. Agents Chemother. 2006,
50, 480–489.
(10) Noedl, H.; Se, Y.; Schaecher, K.; Smith, B. L.; Socheat, D.; Fukuda,
M. M. Evidence of Artemisinin-Resistant Malaria in Western Cam-
bodia. N. Engl. J. Med. 2008, 359, 2619–2620.
(11) Maude, R. J.; Pontavornpinyo, W.; Saralamba, S.; Aguas, R.; Yeung,
S.; Dondorp, A. M.; Day, N. P. J.; White, N. J.; White, L. J. The Last
Man Standing is the Most Resistant: Eliminating Artemisinin Resistant
Malaria in Cambodia. Malaria J. 2009, 8, 31.
(12) O’Neill, P. M.; Posner, G. H. A Medicinal Chemistry Perspective on
Artemisinin and Related Endoperoxides. J. Med. Chem. 2004, 47,
2945–2964.
(13) Posner, G. H.; O’Neill, P. M. Knowledge of the Proposed Chemical
Mechanism of Action and Cytochrome P450 Metabolism of Antima-
larial Trioxanes Like Artemisinin Allows Rational Design of New
Antimalarial Peroxides. Acc. Chem. Res. 2004, 37, 97–404.
(14) U.Eckstein-Ludwig, R. J.; Web, I. D. A.; Van Goethem, J. M.; East,
A. G.; Lee, M. K. P. G.; Bray, S. A.; Ward, S. K. Artemisinins Target
the SERCA of Plasmodium falciparum. Nature 2004, 424, 957–961.
(15) Posner, G. H.; Chang, W.; Hess, L.; Woodard, L.; Sinishtaj, S.; Usera,
A. R.; Maio, W.; Rosenthal, A. S.; Kalinda, A. S.; D’Angelo, J. G.;
Petersen, K. S.; Stohle, R.; Chollet, J.; Santo-Tomas, J.; Snyder, C.;
Rottmann, M.; Wittlin, S.; Brun, R.; Shapiro, T. A. Malaria-Infected
Mice are Cured by Oral Administration of New Artemisinin Deriva-
tives. J. Med. Chem. 2008, 51, 1035–1042.
Biological Testing. The compounds were dissolved in
DMSO at 5 ng/mL. For the assays, fresh dilutions of all
compounds in the screening medium were prepared, and 50
µL of highest starting concentration (500 ng/mL) was
dispensed in duplicate wells in row “B” of 96 well tissue
culture plates. The highest concentration for chloroquine was
25 ng/mL. Subsequently two-fold serial dilutions were
prepared up to row “H” (seven concentrations). Finally 50
µL of a 2.5% parasitized cell suspension containing 0.5%
parasitaemia was added to each well except four wells in
row “A” which received non infected cell suspension. These
wells containing noninfected erythrocytes in the absence of
drugs served as negative controls, while parasitized eryth-
rocytes in the presence of CQ served as positive controls.
After 72 h of incubation, 100 µL of lysis buffer [20 mM tris
(pH 7.5), 5 mM EDTA, 0.008% (wt/vol) saponin, and 0.08%
(v/v) Triton X-100] containing one times the concentration
of SYBER green I (Invitrogen) was added to each cell. The
plates were reincubated for 1 h at room temperature and
examined for the relative fluoroscence units (RFUs) per well
using the FLUO star, BMG lab technologies. The IC50 has
been determined using nonlinear regression analysis dose
response curves. It is worthwhile to note that the standard
deviation (SD) in the measurement of the IC50 value for five
synthesized compounds along with standard drugs arte-
mether, arteether, and chloroquine was an average of 7% of
the mean, which is comparable with the SD in the measure-
ment of IC50 value (8-10% of the mean) for the training
and test sets considered for QSAR study.
(16) O’Neill, P. M.; Mukhtar, A.; Ward, S. A.; Bickley, J. F.; Davies, J.;
Bachi, M. D.; Stocks, P. A. Application of Thiol-Olefin Cooxygenation
Methodology to a New Synthesis of the 1,2,4-Trioxane Pharmacopoeia.
Org. Lett. 2004, 18, 3035–3038.
(17) Cabaret, O. D.; Vical, F. B.; Loup, C.; Robert, A.; Gornitzka, H.;
Bonhoure, A.; Vial, H.; Magnaval, J. F.; Seguela, J. P.; Meunier, B.
Synthesis and Antimalarial Activity of Trioxaquines Derivatives.
Chem.sEur. J. 2004, 10, 1625–1636.
(18) Griesbeck, A. G.; El-Idreesy, T. T.; Ho¨inck, L. O.; Lex, J.; Brun, R.
Novel Spiroanellated 1,2,4-Trioxanes with High in Vitro Antimalarial
Activities. Bioorg. Med. Chem. Lett. 2005, 15, 595–597.
(19) Singh, C.; Verma, V. P.; Naikade, N. K.; Singh, A. S.; Hassam, M.;
Puri, S. K. Novel Bis- and Tris-1, 2, 4-trioxanes: Synthesis and
Antimalarial Activity Against Multidrug-Resistant Plasmodium yoelii
in Swiss Mice. J. Med. Chem. 2008, 51, 7581–7592.
(20) Vennerstrom, S. J. L.; Brun, A.B. R.; Charman, S. A.; Chlu, F. C. K.;
Chollet, J.; Dong, Y.; Dorn, A.; Hunziker, D.; Matile, H.; McIntosh,
K.; Padmanilayam, M.; Tomas, J. S.; Scheurer, C.; Scorneaux, B.;
Tang, Y.; Urwyler, H.; Wittlin, S.; Charman, W. N. Identification of
an Antimalarial Synthetic Trioxolane Drug Development Candidate.
Nature 2004, 430, 900–904.
(21) Uhlemann, A. C.; Wittlin, S.; Matile, H.; Bustamante, L. Y.; Krishna,
S. Mechanism of Antimalarial Action of the Synthetic Trioxolane,
RBX11160 (OZ277). Antimicrob. Agents Chemother. 2007, 51, 667–
672.
(22) Jung, M.; Kim, H. CoMFA of Artemisinin Derivatives: Effect of
Location and Size of Lattice. Bioorg. Med. Chem. Lett. 2001, 11, 2041–
2044.
ACKNOWLEDGMENT
The authors are thankful to Mrs. S. Rastogi, Mr. A. K.
Srivastava, and Mr. A. S. Kushwaha for the technical
assistance and to SAIF and CDRI, Lucknow, for providing
spectroscopic data. Two of authors (A.K.G. and S.C.)
acknowledge CSIR, New Delhi, for financial support.
Supporting Information Available: Schemes for the
syntheses of trioxane 5a, 6a-b and 7a-b, 1H NMR spectra of
trioxane 5a, 6a-b and 7a-b and 3D structures of all the
molecules used in the QSAR study in their best scoring
conformations in the SDF format. This material is available
free of charge via the Internet at http://pubs.acs.org.
(23) Avery, M. A.; Gaston, M. A.; Rodrigues, C. R.; Barreiro, E. J.; Cohen,
F. E.; Sabnis, Y. A.; Woolfrey, J. R. Structure-Activity Relationships
of the Antimalarial Agent Artemisinin: The Development of Predictive
in Vitro Potency Models Using CoMFA and HQSAR Methodologies.
J. Med. Chem. 2002, 45, 292–303.
(24) Grigorov, M.; Weber, J.; Tronchet, J. M. J.; Jefford, C. W.; Millhous,
W. K.; Maric, D. A QSAR Study of the Antimalarial Activity of Some
Synthetic 1, 2, 4-Trioxanes. J. Chem. Inf. Comput. Sci. 1997, 37, 124–
130.
(25) Egan, T. J. Artemisinin-Resistant Plasmodium falciparum: Can the
Genie be Put Back in the Bottle. Future Microbiol. 2009, 4, 637–
639.
(26) Posner, G. H.; Park, S. B.; Gonza’lez, L.; Wang, D.; Cumming, J. N.;
Klinedinst, D.; Shapiro, T. A.; Bachi, M. D. Evidence for the
Importance of High-Valent Fe)O and of a di Ketone in the Molecular
Mechanism of Action of Antimalarial Trioxane Analogs of Artemisi-
nin. J. Am. Chem. Soc. 1996, 118, 3537–3538.
REFERENCES AND NOTES
(1) Aregawi, M.; Williams, R.; Dye, C.; Cibulskis, R.; Otten, M. World
Malaria Report 2008; World Health Organization (WHO): Geneva,
Switzerland, 2008.
(2) Rosenthal, A. S.; Chen, X.; Liu, J. O.; West, D. C.; Hergenrother,
P. J.; Shapiro, T. A.; Posner, G. H. Malaria-Infected Mice are Cured
by a Single Oral Dose of New Dimeric Trioxane Sulfones Which are
also Selectively and Powerfully Cytotoxic to Cancer Cells. J. Med.
Chem. 2009, 52, 1198–1203.
(3) Snow, R. W.; Guerra, C. A.; Noor, A. M.; Myint, H. Y.; Hay, S. I.
The Global Distribution of Clinical Episodes of Plasmodium falci-
parum Malaria. Nature 2005, 434, 214–217.
(4) Antimalarial drug resistance. In Antimalarial Chemotherapy: Mech-
anisms of Action, Resistance, and New Directions in Drug DiscoVery;
Rosenthal, P. J., Ed.; Humana Press: Totowa, NJ, 2001; pp 65-83.
(5) Wright, C. W. Traditional Antimalarials and the Development of Novel
Antimalarial Drugs. J. Ethnopharmacol. 2005, 100, 67–71.
(6) Klayman, D. L. Qinghaosu (Artemisinin) An Antimalarial Drug from
China. Science 1985, 228, 1049–1055.
(27) Posner, G. H.; Cumming, J. N.; Woo, S. H.; Ploypradith, P.; Xie, S.;
Shapiro, T. A. Orally Active Antimalarial 3-Substituted Trioxanes:
New Synthetic Methodology and Biological Evaluation. J. Med. Chem.
1998, 41, 940–951.

1520 J. Chem. Inf. Model., Vol. 50, No. 8, 2010
GUPTA ET AL.
(28) Cumming, J. N.; Wang, D.; Park, S. B.; Shapiro, T. A.; Posner, G. H.
Design, Synthesis, Derivatization, and Structure-Activity Relationships
of Simplified, Tricyclic, 1,2,4-Trioxane Alcohol Analogues of the
Antimalarial Artemisinin. J. Med. Chem. 1998, 41, 952–964.
(29) Posner, G. H.; O’Dowd, H.; Caferro, T.; Cumming, J. N.; Ploypradith,
P.; Xie, S.; Shapiro, T. A. Antimalarial Sulfone Trioxanes. Tetrahedron
Lett. 1998, 39, 2273–2276.
(30) Posner, G. H.; Parker, M. H.; Northrop, J.; Elias, J. S.; Ploypradith,
P.; Xie, S.; Shapiro, T. A. Orally Active, Hydrolytically Stable,
Semisynthetic, Antimalarial Trioxanes in the Artemisinin Family.
J. Med. Chem. 1999, 42, 300–304.
(31) O’Dowd, H.; Ploypradith, P.; Xie, S.; Shapiro, T. A.; Posner, G. H.
Antimalarial Artemisinin Analogs. Synthesis via Chemoselective C-C
Bond Formation and Preliminary Biological Evaluation. Tetrahedron.
1999, 55, 3625–3636.
(32) Gund, P. In Pharmacophore Perception, DeVelopment, and Drug
Design; Gunner, O. F., Ed.; International University Line: La Jolla,
CA, 2000; pp 3-11.
(33) Kim, S. G.; Yoon, C.; Kim, S.; Cho, Y.; Kang, D. Building a Common
Feature Hypothesis for Thymidylate Synthase Inhibition. Bioorg. Med.
Chem. 2000, 8, 11–17.
(34) Wei, N.; Sadrzadeh, S. M. Enhancement of Hemin-Induced Membrane
Damage by Artemisinin. Biochem. Pharmacol. 1994, 48, 737–41.
(35) The Principle of Experimentation Illustrated by a PsychoPhysical
Experiment, 8th ed.; Fischer, R., Ed.; Hafner Publishing Co.: New
York, 1966.
(36) Schlitzer, M. Malaria Chemotherapeutics Part I: History of Antimalarial
Drug Development, Currently Used Therapeutics and Drugs in Clinical
Development. ChemMedChem. 2007, 2, 944–986.
(37) Snydera, C.; Cholleta, J.; Santo-Tomasa, J.; Scheurera, C.; Wittlin, S.
In vitro and in vivo Interaction of Synthetic Peroxide RBx11160
(OZ277) with Piperaquine in Plasmodium Models. Exp. Parasitol.
2007, 115, 296–300.
(38) Vivas, L.; Rattray, L.; Stewart, L. B.; Robinson, B. L.; Fugmann, B.;
Haynes, R. K.; Peters, W.; Croft, S. L. Antimalarial Efficacy and Drug
Interactions of the Novel Semi-Synthetic Endoperoxide Artemisone
in Vitro and in ViVo. J. Antimicrob. Chemother. 2007, 59, 658–665.
(39) Langer, T.; Krovat, E. M. Chemical Feature-Based Pharmacophores
and Virtual Library Screening for Discovery of New Leads. Curr.
Opin. Drug DiscoVery DeV. 2003, 6, 370–376.
(40) CombiGlide, version 2.5; Schro¨dinger, LLC: New York, 2009.
(41) Singh, C.; Kanchan, R.; Srivastava, D.; Puri, S. K. 8-(1-Naphthalen-
2-yl-vinyl)-6, 7, 10-Trioxaspiro (4,5)Decane, a New 1,2,4-Trioxane
Effective Against Rodent and Simian Malaria. Bioorg. Med. Chem.
Lett. 2006, 16, 584–586.
(42) Smilkstein, M.; Sriwilaijaroen, N.; Kelly, J. X.; Wilairat, P.; Riscoe,
M. Simple and Inexpensive Fluorescence-Based Technique for High-
Throughput Antimalarial Drug Screening. Antimicrob. Agents Chemoth-
er. 2004, 48, 1803.
(43) DiscoVery Studio, version 2.0; Accelrys Software Inc.: San Diego,
CA 2001.
(44) Smellie, A.; Teig, S. L.; Poling, T. P. Promoting Conformational
Variation. J. Comput. Chem. 1995, 16, 171–187.
(45) Kristam, R.; Gillet, V. J.; Lewis, R. A.; Thorner, D. Comparision of
Conformational Analysis Techniques to Generate Pharmacophore
Hypotheses Using Catalyst. J. Chem. Inf. Model. 2005, 45, 461–476.
(46) Greenidge, P. A.; Weiser, J. A Comparison of Methods for Pharma-
cophore Generation with the Catalyst Software and Their Use for 3D-
QSAR: Application to a Set of 4-Aminopyridine Thrombin Inhibitors.
Mini-ReV. Med. Chem. 2001, 1, 79–87.
(47) Bhattacharjee, A. K.; Hartell, M. G.; Nichols, D. A.; Hicks, R. P.;
Stanton, B. Structure-Activity Relationship Study of Antimalarial
Indolo [2,1-B]Quinazoline-6,12-Diones (Tryptanthrins).Three-Dimen-
sional Pharmacophore Modeling and Identification of New Antima-
larial Candidates. Eur. J. Med. Chem. 2004, 39, 59–67.
CI100180E