QSAR studies and pharmacophore identification for arylsubstituted cycloalkenecarboxylic acid methyl esters with affinity for the human dopamine transporter

Article abstract of DOI:10.1016/j.bmc.2007.05.015

Data from a series of 29 monoamine transport inhibitors were used to generate 2D and 3D QSAR models for their binding affinity to the human dopamine transporter (hDAT). Among the inhibitors were many non-nitrogen containing compounds. The 2D QSAR analysis resulted in the equation -log K_i = 4.00 - 3.93E_LUMO - 0.67E_HOMO - 3.24σ_p, which predicted the importance of electron withdrawing groups in the aromatic moiety. However, the model failed to predict the observed poor binding of nitro-substituted compounds. In contrast, a derived 3D QSAR model was capable of predicting these more correctly.

Full text of DOI:10.1016/j.bmc.2007.05.015

Bioorganic & Medicinal Chemistry 15 (2007) 5262–5274
QSAR studies and pharmacophore identification for arylsubstituted
cycloalkenecarboxylic acid methyl esters with affinity for the
human dopamine transporter
Helena S. Christensen,a Søren V. Boye,a Jacob Thinggaard,a Steffen Sinning,b
Ove Wiborg,b Birgit Schiøtta,* and Mikael Bolsa,*
aDepartment of Chemistry, University of Aarhus, Langelandsgade 140, DK-8000 Aarhus C, Denmark
bDepartment of Biological Psychiatry, Psychiatric, University Hospital, Skovagervej 2, DK-8240 Risskov, Denmark
Received 18 September 2006; revised 25 February 2007; accepted 4 May 2007
Available online 10 May 2007
Abstract—Data from a series of 29 monoamine transport inhibitors were used to generate 2D and 3D QSAR models for their bind-
ing affinity to the human dopamine transporter (hDAT). Among the inhibitors were many non-nitrogen containing compounds. The
2D QSAR analysis resulted in the equation ꢀlogKi = 4.00 ꢀ 3.93ELUMO ꢀ 0.67EHOMO ꢀ 3.24rp, which predicted the importance of
electron withdrawing groups in the aromatic moiety. However, the model failed to predict the observed poor binding of nitro-substi-
tuted compounds. In contrast, a derived 3D QSAR model was capable of predicting these more correctly.
Ó 2007 Elsevier Ltd. All rights reserved.
1. Introduction
been accepted as the current view of the ‘dopamine
hypothesis’ that it is the increased amount of DA and
its subsequent action at one or more of the D1–D5 DA
receptors, and not a directly mediated message by the
binding of cocaine itself, that at large is responsible
for the behaviours associated with cocaine addiction.2,3
As a consequence of this assumption, the DAT has be-
come the major target in several studies in the search
of a treatment of cocaine addiction4 even though co-
caine’s affinities for the three major monoamine trans-
porters are roughly similar.2,5 Recent studies also
imply that the biochemistry may not be this simple, since
cocaine seems to interact with the metabotropic gluta-
mate receptor (mGluR) as well.6 However, most studies
conducted in this area are focused on the discovery of
selective DAT inhibitors that are either cocaine antago-
nists or partial agonists. The latter operate as competing
antagonists in the presence of a full antagonist (cocaine),
and can thus be utilized for potential treatment of
cocaine abuse provided the addictive properties of the
substitute are smaller.
Cocaine (1) is one of the most addictive substances
known, and currently no efficient treatment is available.1
It exerts its stimulating and reinforcing effect by inhibit-
ing the reuptake of the three major monoamines, dopa-
mine (DA), serotonin (SER) and norepinephrine (NE),
into the presynaptic neuron by blocking their respective
transporters [dopamine transporter (DAT), serotonin
transporter (SERT) and norepinephrine transporter
(NET)]. This blockade results in an elevated synaptic
content of monoamines and consequently an enhanced
response to the incoming signal. It has for some time
Abbreviations: hDAT, human dopamine transporter; hSERT, human
serotonin transporter; hNET, human norepinephrine transporter; DA,
dopamine; SER, serotonin; NE, norepinephrine; QSAR, quantitative
structure–activity relationship; 2D, 2 dimensional; 3D, 3 dimensional;
CoMFA, Comparative Molecular Field Analysis; NBS, N-bromosuc-
cinimide; TFA, trifluoroacetic acid; Tf2O, triflic anhydride; TS, train-
ing set; TES, test set; G/PLS, Genetic Partial Least Squares; PLS,
Partial Least squares; GFA, Genetic Function Approximation; PRE-
SS, predictive residual sum of squares; RMS, root mean square; DCM,
dichloromethane.
Over the past two decades, there have been an increasing
number of studies in the literature of the chemical and
pharmacological profiles of several classes of DAT
inhibitors, Figure 1, including tropanes (mainly of the
WIN based-analogues (2)),7,8 benztropine-based
Keywords: Monoamine transporters; QSAR; Inhibitors.
*
Corresponding authors. Tel.: +45 8942 3963 (M.B.); e-mail
addresses: birgit@chem.au.dk; mb@chem.au.dk
0968-0896/$ - see front matter Ó 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmc.2007.05.015

H. S. Christensen et al. / Bioorg. Med. Chem. 15 (2007) 5262–5274
5263
R
O
N
O
N
7
O
1
R
O
2
O
N
5
6
4
O
X
O
3
O
Y
1: (-)-Cocaine
2: WIN Based analogues
O
N
N
X
R
X
3: Benztropine Based
analogues
O
Y
Cl
4: Piperidine Based analogues
5: Pyrrolidine Based analogues
Figure 1. Structures of cocaine and selected analogues.
analogues (3),5,9 as well as piperidine-based analogues
(4).10,11 Lately, a new type of pyrrolidine-based ana-
logues (5)12,13 has gained interest in the continuing
search for potent DAT inhibitors. Many of these studies
have expanded from including only synthesis, structure–
activity relationships (SAR) and pharmacology to fur-
thermore include, for example, a Comparative Molecu-
lar Field Analysis (CoMFA)14 of the compounds.15–18
CoMFA-studies are conducted by using activity data
to gain valuable knowledge of structural features
required for the set of compounds studied to achieve
optimal binding to the DAT, that is, generation of a
pharmacophore model for optimal binding in the ligand
binding-site of the DAT. The studies resulted in predic-
tive models for the behaviour of benztropine analogues
3 and piperidine analogues 4 that among others came to
the conclusion that small halogens (–Cl, –F) in meta–
para positions on the phenyl-moiety were well tolerated.
To the best of our knowledge, no one has yet conducted
any CoMFA or other 3D QSAR studies on aryl-substi-
tuted cycloalkenecarboxylic acid methyl esters of the
type that can be seen in Figure 4. These compounds
are interesting to study because they, despite loosing
the important nitrogen atom in most compounds, pos-
sess considerable binding affinity. This raises the ques-
tion whether the compounds still bind in the same
mode as cocaine and phenyltropanes. Therefore, we
set out to fill this gap by exploiting first 2D QSAR (Ceri-
us219) and subsequently the 3D methodology imple-
mented in Catalyst20 in order to gain more knowledge
of this specific group of DAT inhibitors with respect
to extracting the most important structural and elec-
tronic properties required to achieve optimal binding
to hDAT.
O
O
N
O
N
N
O
O
O
6a (WIN35065-2)
Binding:
rDAT: Ki (nM) = 33 17
6b
Binding:
rDAT: Ki (nM) = 60400 4800
6c
Binding:
rDAT: Ki (nM) = 18900 1700
O
O
N
O
O
N
6d
6e
Binding:
rDAT: Ki (nM) = 96500 42
Binding:
rDAT: Ki (nM) = 5620 390
Figure 2. Ring-contracted analogues of the phenyl tropane WIN35065-2.

5264
H. S. Christensen et al. / Bioorg. Med. Chem. 15 (2007) 5262–5274
Structural modification in several, different positions of
the phenyl tropane has previously been addressed (2,
Fig. 1). However, simplification of the phenyl tropane
skeleton has not received much attention.
The present work was part of a line of research con-
ducted in our group intending to find new monoamine
transport inhibitors preferably with selectivity between
inhibition of DAT binding and uptake.25 One medicinal
chemistry strategy we wished to follow was simplifica-
tion of the tropane skeleton and here the bicyclic com-
pounds from the Trudell work became important as
simpler analogues. We therefore decided to carry out
an extended study of these compounds with many aro-
matic substituents. Since it has been reported that phe-
nyltropanes (Fig. 3) having a 2,3-double bond (8a–c)11
are equipotent with the saturated 2b,3b-substituted par-
ents (7a–c) we chose to investigate the unsaturated
derivatives thereby avoiding the problematic reduction
of the double bond. The potency towards the SERT
however was decreased dramatically, but since we were
One such work has been carried out by the Trudell
group.21 They synthesized the ring-contracted analogues
(6b–e, Fig. 2) of the phenyl tropane WIN35065-2 (6a)
and investigated their binding to the DAT. While the syn-
thesis of the 2:2:1-azabicyclic system was short and ele-
gant, mainly following the transformation 25 ! 26 !
11 (Scheme 1), reduction of the double bond and separa-
tion of the diastereoisomers 6b–e was a considerable chal-
lenge and the sequence was only carried out for the phenyl
derivative. The isomers 6b–e displayed markedly reduced
binding affinity to the DAT compared to 6a (Fig. 2).
O
X
O
X
O
O
ii
i
O
H
Br
O
Br
iii
22
23
25
24
H
N
O
O
O
X
X
vi
v
O
O
O
26
Br
Ar
9x-10x, 14-15, 19
Ar
11x, 16
v
H
O
N
O
Br
21
Scheme 1. Reagents and conditions: (i) AgNO3, NBS, acetone, rt, 2 h; (ii) 95°C, 24 h; (iii) H2, 10% Pd/C, cyclohexane, rt, 12 h; (iv) Ar–B(OH)2,
Na2CO3, Pd(OAc)2, PPh3, Benzene–EtOH, 75°C, 20 h; (v) TFA, DCM, rt, 20 h. X = –CH2–, –O–, –N(Boc)–.
O
N
O
O
O
O
Cl
Cl
O
O
Cl
Cl
Cl
Cl
7a (O-401 (R))
7b (O-914 (R/S))
7c (O-1414 (R/S))
myDAT: IC50 (nM) = 1,1
mySERT: IC50 = 2,5 0,1
SERT/DAT = 2,3
myDAT: IC50 (nM) = 3,4 0,5
mySERT: IC50 = 6,5 2,9
SERT/DAT = 1,9
myDAT: IC50 (nM) = 9,6 1,8
mySERT: IC50 = 33,4 0,6
SERT/DAT = 3,5
O
O
O
N
O
O
O
O
Cl
Cl
Cl
Cl
Cl
Cl
8a (O-1109 (R))
8b (O-1014 (R/S))
8c (O-1231 (R/S))
myDAT: IC50 (nM) = 1,2 0,2
mySERT: IC50 = 867 75
SERT/DAT = 722,5
myDAT: IC50 (nM) = 12,8 1,4
mySERT: IC50 = 1960 840
SERT/DAT = 153,1
myDAT: IC50 (nM) = 7,1 1,7
mySERT: IC50 = 5160 580
SERT/DAT = 726,8
Figure 3. An example of a phenyl tropane and structural modifications hereof.

H. S. Christensen et al. / Bioorg. Med. Chem. 15 (2007) 5262–5274
5265
mainly concerned with DAT affinity this was no
problem.
reach the conclusion that the derived 2D QSAR equa-
tion is not as powerful as the 3D QSAR derived phar-
macophore model with respect to predicting binding
affinity of similar new next generation compounds.
In the study we also addressed the structural simplifica-
tion offered by deleting the nitrogen atom in the bridge-
head position. In a very interesting recent study Meltzer
and collaborators have shown that the nitrogen atom in
the phenyl tropane (7a) is non-essential for binding. Oxa
and carba analogues 7b and 7c were only 3- and 9-fold
less potent than 7a for binding to the DAT.22–24 The
synthetic scheme was therefore varied to include oxa
and carba analogues of the 2:2:1-bicyclic compounds
(Fig. 4).
2. Methods
2.1. Synthesis
All compounds with a bicyclic core-structure were
prepared as outlined in Scheme 1. Compound 23 was
obtained by bromination of 22 utilizing N-bromosuccin-
imide (NBS) and AgNO3, 23 was subsequently mixed
with the cyclopentadiene 24 affording the Diels–Alder
coupling product 25. Upon hydrogenation of 25 with
H2 and Pd/C, 26 was produced, which was then
subjected to a Suzuki coupling with the appropriate
aromatic boronic acid to yield the desired coupling
products. Compounds containing the N-Boc-group
was finally subdued to a deprotection step with trifluo-
roacetic acid (TFA) resulting in the products outlined
in Figure 4.
The compounds 9x–11x, 14–16, 19–21 are derived from a
truncated tropane skeleton which arises when removing
C-421 (numbering from (ꢀ)-Cocaine (1)), while com-
pounds 12x and 17 arise from removing the bridge-head
atom from these compounds, and finally compounds 13x
and 18 arise from removing C-5 and C-6 (numbering
from 9) from compounds 9x. We herein describe the gen-
eration of 2D QSAR equations utilizing Cerius219 and
furthermore we identify a pharmacophore by the meth-
ods implemented in Catalyst 4.1020 both of which can
relate molecular properties to the experimentally mea-
sured hDAT binding affinities of the compounds. We
Compounds with a monocyclic core 12–13 and 17–18
were synthesized according to the reaction path outlined
O
O
O
O
O
O
NH
O
6
1
O
2
O
O
R
O
4
5
3
R
R
R
R
12a-c, 12e
9a-e, 9g-i, 9o-p
10a-d, 10f
11a-e, 11j, 11r
13c-f, 13k-n
g
m
R=3-NH2
a
R=2-Cl
R=H
n R=4-COCH3
o R=3-F, 4-CH3
p R=2,4,5-(CH3)3
b R=4-Ph
c R=4-Cl
d R=3,4-Cl2
h
R=2,3-Cl2
i R=2,4-Cl2
j R=3-NO2
k R=4-tBu
l R=4-OCF3
q
e
R= 3,4-(NO2)2
R=4-NO2
R=3-Cl
r
f
R=4-OCH3
O
O
O
O
O
O
NH
O
O
O
14
15
16
17
O
O
O
O
O
O
NH
O
O
O
N
Br
Br
18
19
21
20
Figure 4. Chemical structures of the data set molecules. The notation letter after the compound number is used to refer to all the variants included in
this figure for compounds 9x–13x.

5266
H. S. Christensen et al. / Bioorg. Med. Chem. 15 (2007) 5262–5274
O
O
O
O
O
O
n(H2C)
n(H2C)
n(H2C)
i
ii
O
OTf
27
Ar
12x-13x, 17-18
Scheme 2. Reagents and conditions: (i) NaH, TfO2, Et2O, rt, 1 h; (ii) Ar–B(OH)2, Na2CO3, Pd(OAc)2, PPh3, Benzene–EtOH, 75 °C, 20 h. n = 1,2.
Table 1. Actual and estimated pKi (ꢀlog(Ki)) values using the 2D
in Scheme 2. Starting either from commercially available
2-oxo-cyclopentanecarboxylic acid methyl ester (n = 1)
QSAR model
or 2-oxocyclohexanecarboxylic acid methyl ester
(n = 2) and reacting it with NaH and Triflic anhydride
Compound Actual ꢀlog(Ki) Estimated ꢀlog(Ki) Residual
2D-QSAR TS DAT binding
(Tf2O) gave rise to the triflate-derivative 27. Subsequent
Suzuki-coupling with the appropriate boronic acids
yielded the compounds 12–13 and 17–18.
9a
9b
3.19
3.22
4.60
5.17
3.43
4.83
4.64
3.35
4.61
4.57
5.96
3.20
4.07
3.93
4.74
4.46
3.85
3.52
2.58
3.06
3.48
5.37
5.76
4.93
5.38
3.13
3.98
4.16
5.54
3.35
4.76
4.08
3.92
4.28
4.68
5.97
3.03
3.87
4.08
4.15
4.68
3.82
3.66
2.92
2.89
3.92
5.35
5.73
4.83
5.14
0.06
ꢀ0.76
0.44
9c
9d
ꢀ0.37
0.08
0.07
9g
9h
2.2. Biology
9i
0.56
ꢀ0.57
0.33
All compounds were tested for inhibition of the hDAT.
This was carried out by determining displacement of a
radioligand, according to the procedure described in
Section 4.
11a
11b
11c
11d
12a
12c
12e
10c
13e
13f
13k
13l
13m
13n
14
ꢀ0.11
ꢀ0.01
0.17
2.3. Molecular modelling
0.20
ꢀ0.15
0.59
QSAR modelling concerns the extraction of a quantita-
tive relationship between the experimentally observed
biological activity of a set of compounds and a number
of descriptors related to physical and chemical proper-
ties of the compounds.25 The 29 molecules in the 2D
QSAR data set were divided into a training set of 25
compounds (2D-TS) and a test set with four molecules
(2D-TES)—see Table 1. A satisfactory 2D QSAR rela-
tion was generated by using the Genetic Partial Least
Squares methodology (G/PLS) as implemented in the
Cerius2 suite of programs.19 The G/PLS methodology
combines a Genetic Function Approximation (GFA)26
for selecting the appropriate descriptors for the
QSAR-model, and a Partial Least Squares (PLS)27
regression as the fitting technique to weigh the relative
contribution of each descriptor in the final model. The
GFA method provides a population of QSAR-models
rather than a single equation for correlation between
biological activity and physicochemical descriptors.
These models are created by evolving random initial
models using a genetic algorithm. The next generation
of models is computed by cross-over combinations
between the best scoring models. PLS has proved superior
in handling data with strongly correlated and numerous
independent variables.28 PLS, in particular, is able to
identify linear combinations of the original physico-
chemical descriptors that are better for correlating the
data thereby providing statistically more robust solu-
tions.28 Backtracking the final equation to the original
variables is easy within PLS and provides a better way
of chemical interpretation of the result. In this study,
all descriptors within the Cerius2 QSAR+ module were
included in the derivation of the QSAR equation, and
the G/PLS methodology thus is appropriate in deriving
robust QSAR relations consisting of a minimum num-
ber of descriptors.19 Once derived, the equation was
tested against the test set, 2D-TES. Furthermore it was
ꢀ0.22
0.03
ꢀ0.14
0.34
0.17
ꢀ0.44
0.02
0.03
16
17
0.10
0.24
18
2D QSAR TES DAT binding
9e
3.98
4.47
3.58
5.71
4.41
4.92
3.30
5.29
ꢀ0.43
ꢀ0.45
0.28
11e
12b
13d
0.42
checked by synthesizing and testing the biological activ-
ity of a set of analogues that were predicted to exhibit
better pharmacological activity towards DAT.
The Catalyst 4.10 software package can be applied for
automated ligand-based pharmacophore generation.
Using the algorithm HypoGen29 provided the quantita-
tive 3D QSAR models for describing the binding affinity
to hDAT. In this method an activity based alignment
among a collection of conformations for each com-
pound in the training set is obtained. The 3D pharmaco-
phore models, called ‘hypotheses’, were derived from a
data set containing 28 compounds of which 19 were used
as the training set (3D-TS) and the remaining 9 served
the test set (3D-TES)—see Table 2 for the details. The
molecules were carefully chosen with respect to discrep-
ancy in biological activity among a library consisting
of a total of 42 compounds in the structural distinct
sub-libraries (9–21).

H. S. Christensen et al. / Bioorg. Med. Chem. 15 (2007) 5262–5274
Table 2. Actual and estimated pKi values using the 3D QSAR model
5267
7.0
6.0
5.0
4.0
3.0
2.0
Compound
Actual Ki
Estimated Ki
Error score
3D QSAR TS DAT binding
9a
9h
2.44
4.82
5.12
3.35
5.96
4.47
3.20
4.67
5.71
4.46
3.85
3.06
5.37
5.76
4.93
4.48
2.44
2.39
2.92
4.70
5.51
2.64
5.49
4.39
3.00
4.68
5.92
4.48
4.68
3.08
5.07
5.11
5.00
4.80
2.57
2.64
ꢀ3.1
1.3
10d
11a
11d
11e
12a
12e
13d
13e
13f
13m
14
ꢀ2.4
5.2
K
2.9
1.2
1.7
ꢀ1.0
ꢀ1.6
ꢀ1.0
ꢀ6.7
ꢀ1.1
2.0
2.0
3.0
4.0
5.0
6.0
7.0
16
17
4.3
K
Estimated -log
i
ꢀ1.1
ꢀ2.0
ꢀ1.4
ꢀ1.8
19
20
Figure 5. Plot of actual ꢀlogKi versus estimated ꢀlogKi of training set
(Ç) and test set (j) using 2D QSAR. R2 = 0.8596.
21
3D QSAR TES DAT binding
9b
9d
3.22
5.17
4.46
3.24
4.60
4.57
4.07
4.74
5.38
3.43
5.28
3.48
4.22
3.39
4.18
4.41
4.55
4.81
ꢀ1.6
ꢀ1.3
9.5
Table 1 for actual numbers). The three descriptors that
G/PLS found most valuable for assessing the binding
to hDAT are all indicative of the importance of elec-
tronic effects of groups attached to the aromatic system.
Better binding is predicted by a low ELUMO energy (po-
sitive LUMO energies are computed), a large negative
HOMO energy (negative HOMO energies are found)
and a negative rp. Electron-withdrawing groups make
the two former terms positive, while the Hammett con-
10c
10f
11b
11c
12c
13c
18
ꢀ9.7
17.0
2.4
ꢀ2.2
1.5
3.3
3D QSAR prediction set DAT binding
9o
9p
4.25
4.49
4.74
4.47
ꢀ3.2
stant term is negative and thus
contribution.
a moderating
1.0
As the model contains only three descriptors, it is rela-
tively straightforward to suggest new potentially good li-
gands. By investigating different substituents at the
phenyl-moiety, it became obvious that nitro-groups
were predicted to have a very beneficial effect on the
Ki value. Compound 11r (with a para-nitro group) was
predicted to have a Ki value of 1.5 nM, whilst the 3,4-
disubstituted compound, 11q, was predicted to have a
Ki value of 0.3 pM. The 3,4-disubstituted compound,
11q, was not synthesized though several unsuccessful
attempts at synthesizing the equivalent boronic acid were
performed. We therefore synthesized the mono-substi-
tuted nitro-compounds 11j and 11r, and subsequently
tested them for biological activity giving the results
shown in Table 3. Instead of having affinities in the
low nanomolar range, these compounds give affinities
in the high micromolar range, 232 lM for 11j and
68 lM for 11r. Clearly, introduction of the nitro-groups
had an effect on binding affinity that was not described
well by the derived 2D QSAR equation. Therefore, the
3. Results and discussion
The best 2D QSAR model (highest q2) for hDAT bind-
ing calculated using G/PLS method found in Cerius2 re-
sulted in Eq. (1).
ꢀ log Ki ¼ 4:00 ꢀ 3:93ELUMO ꢀ 0:67EHOMO
ꢀ 3:24rp
ð1Þ
ELUMO and EHOMO are the energies of the LUMO and
HOMO computed by CNDO/2, respectively, and rp is
the Hammett constant of a substituent in the para-posi-
tion. Having r2 = 0.87 and q2 = 0.83, it is a case of a
good model bordering to a perfect model. The predictive
residual sum of squares (PRESS) is low (3.59) and com-
paring residual-values for the ligands in 2D-TES results
in high residual-values, yet not higher than residual-val-
ues contained in the 2D-TS. Figure 5 depicts the plot of
the actual activities versus the estimated activities (see
Table 3. Prediction of Ki valuesa (M) for the binding of nitro-substituted compounds to DAT with the 2D and 3D QSAR relations
Compound
Actual Ki
2D QSAR estimated Ki
Errorscore
3D QSAR estimated Ki
Errorscore
11j
11q
11r
2.32 Æ 10ꢀ4
n.a.
3.02 Æ 10ꢀ11
5.89 Æ 10ꢀ13
1.51 Æ 10ꢀ9
ꢀ77,00,000
n.a.
4.1 Æ 10ꢀ4
1.8 Æ 10ꢀ3
3.8 Æ 10ꢀ3
1.8
n.a.
56
6.79 Æ 10ꢀ5
ꢀ45,000
a All values specified in M. n.a., not available. Errorscore = ꢀAct.Ki/Est.Ki if Est. <Act.; = Act.Ki/Est.Ki if Est. >Act.

5268
H. S. Christensen et al. / Bioorg. Med. Chem. 15 (2007) 5262–5274
7.0
6.0
5.0
4.0
3.0
2.0
investigation was extended to include a 3D QSAR study
applying the Catalyst software to see whether this tool
could provide a better model. On the basis of manually
selected compounds, features and settings, this program
can design 3D models that not only suggest generalized
three dimensional orientations of chemical features, but
also estimate the biological activities of the compounds
in the training and test sets.
K
Using Catalyst’s HypoGen module, and selecting from a
set of 52 structurally diverse ligands that vary in activity
values over about 3 orders of magnitude, a 3D QSAR
model was obtained. The quality of the generated phar-
macophore hypothesis was evaluated from the cost
functions calculated by HypoGen during hypothesis
generation.29 The structures of the molecules used are
found in Figure 5 and all biological activities are shown
in Table 2. The result of the hypothesis generation can
be seen in Table 4 where the 10 generated hypotheses
are summarized.
2.0
3.0
4.0
5.0
6.0
7.0
K
Estimated -log
i
Figure 6. Plot of actual ꢀlogKi versus estimated ꢀlogKi of training set
(d), test set (j) and prediction set (m) using 3D QSAR. R2 = 0.8907.
Figure 6 depicts the plot of the actual activities versus
the estimated activities obtained with Hypo1 (the esti-
mated activities are tabulated in Table 2). The top scor-
ing hypothesis, Hypo1, with one hydrogen bond
acceptor, two hydrophobes, one aliphatic hydrophobe
and one aromatic hydrophobe features displays a fixed
cost of 72.245, total cost of 78.123 and null cost of
117.69 with a configuration cost of 7.209 bits, all of
which are acceptable values indicating a good fit. The
difference between the fixed and the null cost has to be
20–80 bits, and in this case the difference was 45.45,
which is reasonable. The difference in cost between total
and null of 39.568 indicates a 70–80% statistical signifi-
cance bordering the 80–90%, the consequence of this is
that there is approximately 80% chance of obtaining a
predictive hypothesis. This hypothesis also shows the
lowest RMS (0.787) and the highest correlation coeffi-
cient (0.944) and thus Hypo1 was retained for further
analysis. The test set was estimated with errorscores
ranging from ꢀ9.7 (10f) to 17 (11b), but most error-
scores are in the acceptable range of ꢀ4 to 4 (Table 2).
Table 5. Prediction set and Ki valuesa for DAT predicted by top
ranking Hypo1
Compound
Actual Ki
Estimated Ki
Errorscore
9o
9p
5.60 Æ 10ꢀ5
3.22 Æ 10ꢀ5
1.8 Æ 10ꢀ5
3.4 Æ 10ꢀ5
ꢀ3.2
1.0
a All values specified in M. Errorscore: ꢀAct./Est.
As can be seen from Table 5, Hypo1 demonstrated a
high degree of predictability for these two ligands with
errorscores in the low range (1.0 and ꢀ3.2). Figure 7 de-
picts 9p fitted in Hypo1 using the flexible fit option.
With this piece of evidence at hand, we turned to the
more difficult problem of predicting the activity of the
two nitro-substituted compounds (Table 3). While,
Hypo1 could not predict the Ki values of nitro-substi-
tuted compounds perfectly it nevertheless predicted
them considerably better than the 2D QSAR model
was able to do. This can also intuitively be rationalized
when visualizing the graphic model (Fig. 7), since it pos-
sesses two hydrophobic moieties in the position where
To validate Hypo1’s ability of prediction, we prepared
the two compounds, 9o and 9p, the activities of which
had been predicted by Hypo1 to be in the lM range.
Table 4. Summary of the HypoGen output for the 3D-TS data set
Hypoa
Total cost
Cost differenceb
RMS
Correlation (r)
Featuresc
Hypo1
Hypo2
Hypo3
Hypo4
Hypo5
Hypo6
Hypo7
Hypo8
Hypo9
Hypo10
78.123
78.400
78.527
78.824
79.372
79.731
81.955
84.055
84.094
84.722
39.568
39.290
39.163
38.866
38.318
37.959
35.735
33.636
33.596
32.969
0.787
0.801
0.805
0.831
0.864
0.883
1.010
1.094
1.111
1.129
0.944
0.942
0.941
0.937
0.932
0.929
0.905
0.888
0.885
0.881
HBA, HY, HYAl, HYAr, HY
HBA, HY, HYAl, HY, HY
HBA, HY, HYAr, HY, HY
HBA, HY, HY, HY, HY
HBA, HY, HYAl, HY, RAr
HBA, HY, HY, HY, RAr
HBA, HY, HYAl, HY, RAr
HBA, HY, HY, HY, RAr
HBA, HY, HYAl, HYAl, RAr
HBA, HY, HY, HY, RAr
a Numbers for the hypotheses are consistent with the numeration as obtained by the hypothesis generation.
b Difference between the null hypothesis and the total cost of each returned hypothesis.
c HBA: hydrogen bond acceptor; HY: Hydrophobic; HYAl: aliphatic hydrophobe; HYAr: aromatic hydrophobe; RAr: ring aromatic.

H. S. Christensen et al. / Bioorg. Med. Chem. 15 (2007) 5262–5274
5269
3.1. Conclusions and perspectives
2D and 3D QSAR models were developed to describe
the affinity for hDAT of the library of previously pre-
pared compounds. Both models show that the aromatic
moiety is very important. The 2D QSAR model is the
relatively simple relationship pKi = 4.00 ꢀ 3.93ELUMO
ꢀ 0.67EHOMO ꢀ3.24rp, which purely depends on elec-
tronic effects and which predicts that electron-withdraw-
ing groups are favourable. In contrast, the 3D model
predicts a number of hydrophobic areas around the
aromatic moiety and it is therefore better in predicting
the affinities of the nitro-substituted compounds than
the 2D model. The models complement each other and
appear to be useful predictive devices for further
compound design.
Figure 7. The top scoring Hypo1 is mapped to compound 9p. HBA:
hydrogen bond acceptor, HY: hydrophobe, HYAl: aliphatic hydro-
phobe, and HYAr: aromatic hydrophobe.
4. Experimental
4.1. General
the nitro-groups are. The nitro-group is polar and there-
fore, based on the 3D pharmacophore model, these
compounds are predicted and actually overestimated
to a poor affinity. In any case it is obvious that the nitro
derivatives are difficult to predict with models that are
based on training sets not containing nitro-substituted
compounds.
NMR spectra were recorded on a Varian Gemini 2000
Spectrometer (400 MHz concerning 1H and 100 MHz
concerning 13C) with tetramethylsilane (TMS) as the
internal standard and CDCl3 as the solvent and refer-
ence point (d 7.26 ppm and d 77.16 ppm, respectively).
Mass spectra were recorded at a Micromass LC-TOF
spectrometer. Thin layer chromatography (TLC) was
carried out on silica gel (Merck Kieselgel 60 F254) and
for visualization of the spots either UV light (254 nm)
or Ce-Mol solution (Ce(SO4)2 (10 g) and (NH4)2MoO4
(15 g) dissolved in 10% H2SO4) was applied. Flash col-
umn chromatography was carried out on Merck silica
gel (230–400 mesh). Reactions were conducted under
an atmosphere of nitrogen.
It is interesting to compare the two new models with
existing QSAR models of related compounds. Compar-
ison with the models made for phenyltropanes15,30 is
particularly relevant since the compounds in this study
are truncated versions of phenyltropanes and are likely
to have a similar binding at the ester and aromatic re-
gion. Carroll et al. have reported a 2D QSAR model
for para substituted phenyl tropanes that has the equa-
tion pIC50 = 0.46p + 0.20MR ꢀ 0.01MR2 ꢀ 1.51 mean-
ing that lipophilicity and size are the important factors
for the substituent in the aromat.30 This is in contrast
to our 2D model which exclusively correlates the activity
to electronic effects, which reflects that the binding of the
aromatic group, in the two classes of compounds, is not
identical. It is observed that, even though both classes of
compounds show similar trends (i.e., chloro-substitution
increases potency, etc.), the influence of substitution is
much larger for the small non-nitrogen containing com-
pounds. It is possible that these smaller compounds,
which do not interact at the nitrogen binding site, can
enter deeper into an aromatic binding pocket thereby
obtaining stronger and possibly different interactions.
4.2. Preparation of Suzuki coupling derivatives. General
synthetic procedure
1 Equiv methyl 3-bromo-bicyclo[2.2.1]hept-2-ene-2-car-
boxylate (26) was added to a 5 mL vial with PPh3 (0.1
equiv), Na2CO3 (2 equiv), Pd(OAc)2 (0.05 equiv) and
dissolved in ethanol–benzene (1/3, 4 mL). Finally, the
boronic acid (1.2 equiv) was added, and the resulting
mixture flushed with nitrogen. While stirring, the mix-
ture was heated to 75 °C for 20 h.
4.2.1. Methyl 3-phenyl-bicyclo[2.2.1]hept-2-ene-2-carbox-
ylate (9a). Purification by flash chromatography (diethyl
ether–pentane; 1/30) afforded 9a as colourless oil in 99%
yield (226 mg, 0.99 mmol). 1H NMR (CDCl3) d 7.53 (m,
2H), 7.33 (m, 3H), 3.67 (s, 3H), 3.39 (br s, 1H), 3.28 (br
s, 1H), 1.89 (m, 2H), 1.66 (ddd, 1H, J1 = 2.0 Hz,
J2 = 3.8 Hz, J3 = 8.5 Hz), 1.39 (m, 2H), 1.25 (d, 1H,
J = 8.5 Hz); 13C NMR (CDCl3) d 166.0, 158.0, 135.3,
132.2, 128.4, 128.3, 127.7, 51.1, 50.2, 46.7, 45.4, 26.1,
25.8; HR-MS (ES) calcd for C15H16O2 + Na+: m/z
251.1048. Found m/z 251.1044.
3D QSAR models, in the form of CoMFA, have also
been reported on phenyl tropanes15,31 and on structur-
ally related benztropines.17 These models indicate that
small to medium size steric groups are favourable on
the aromatic moiety while electronic effects are of less
importance. The Hypo1 model points to reach a very
similar conclusion since it points to hydrophilic groups
attached to the aromatic group as being important for
the binding. Thus 3D QSAR models of phenyl tropanes
and the non-nitrogen containing analogues are qualita-
tively consistent.
4.2.2. Methyl 3-(4-biphenyl)-bicyclo[2.2.1]hept-2-ene-2-
carboxylate (9b). Purification by flash chromatography
(diethyl ether–pentane; 1/30) afforded 9b as colourless

5270
H. S. Christensen et al. / Bioorg. Med. Chem. 15 (2007) 5262–5274
1
oil in 94% yield (69 mg, 0.23 mmol). H NMR (CDCl3)
d 7.67 (d, 2H, J = 7.8 Hz), 7.63 (d, 2H, J = 7.8 Hz), 7.60
(d, 2H, J = 7.8 Hz), 7.46 (t, 2H, J = 7.6 Hz), 7.36 (t, 1H,
J = 7.4 Hz), 3.72 (s, 3H), 3.44 (br s, 1H), 3.34 (br s, 1H),
1.93 (m, 2H), 1.70 (m, 1H), 1.43 (dd, 2H, J1 = 2.0 Hz,
J2 = 7.3 Hz), 1.29 (d, 1H, J = 8.6 Hz); 13C NMR
(CDCl3) d 166.0, 157.7, 141.0, 140.7, 134.0, 132.5,
128.9, 128.7, 127.4, 127.0, 126.4, 51.1, 50.1, 46.5, 45.5,
26.2, 25.7; HR-MS (ES) calcd for C19H18O2 + Na+:
m/z 327.1361. Found m/z 327.1358.
NMR (CDCl3) d 7.40 (d, 1H, J = 8.1 Hz), 7.15 (t, 1H,
J = 7.8 Hz), 6.95 (d, 1H, J = 7.6 Hz), 3.58 (s, 3H), 3.40
(br s, 1H), 3.19 (br s, 1H), 1.85 (m, 3H), 1.35 (m, 3H);
13C NMR (CDCl3) d 165.0, 155.6, 137.7, 136.2, 133.1,
131.0, 129.6, 127.5, 126.8, 51.2, 50.4, 47.4, 44.1, 25.7,
24.6; HR-MS (ES) calcd for C15H14Cl2O2 + Na+: m/z
319.0269. Found m/z 319.0259.
4.2.8. Methyl 3-(2,4-dichlorophenyl)-bicyclo[2.2.1]hept-2-
ene-2-carboxylate (9i). Purification by flash chromato-
graphy (diethyl ether–pentane; 1/30) afforded 9i as col-
1
4.2.3. Methyl 3-(4-chlorophenyl)-bicyclo[2.2.1]hept-2-ene-
2-carboxylate (9c). Purification by flash chromatography
(diethyl ether–pentane; 1/30) afforded 9c as colourless oil
ourless oil in 65% yield (50 mg, 0.17 mmol). H NMR
(CDCl3) d 7.41 (d, 1H, J = 2.2 Hz), 7.20 (dd, 1H,
J1 = 2.2 Hz, J2 = 8.3 Hz), 7.02 (d, 1H, J = 8.3 Hz), 3.59
(s, 3H), 3.40 (br s, 1H), 3.18 (br s, 1H), 1.85 (m, 3H),
1.32 (m, 3H); 13C NMR (CDCl3) d 165.1, 154.8, 136.6,
134.0, 133.5, 130.3, 129.3, 126.6, 51.2, 50.4, 47.4, 44.2,
25.6, 24.6; HR-MS (ES) calcd for C15H14Cl2O2 + Na+:
m/z 319.0269. Found m/z 319.0262.
1
in 70% yield (40 mg, 0.15 mmol). H NMR (CDCl3) d
7.48 (d, 2H, J = 8.8 Hz), 7.30 (d, 2H, J = 8.8 Hz), 3.68
(s, 3H), 3.38 (br s, 1H), 3.23 (br s, 1H), 1.89 (m, 2H),
1.65 (d, 1H, J = 8.8 Hz), 1.36 (m, 2H), 1.25 (d, 1H,
J = 8.8 Hz); 13C NMR (CDCl3) d 165.8, 156.8, 134.1,
133.6, 133.1, 129.8, 127.9, 51.2, 50.1, 46.7, 45.4, 26.1,
25.7; HR-MS (ES) calcd for C15H15ClO2 + Na+: m/z
285.0658. Found m/z 285.0667.
4.2.9. Methyl 3-(3-fluoro-4-methyl-phenyl)-bicyclo[2.2.1]-
hept-2-ene-2-carboxylate (9o). Prepared by employing
the Suzuki coupling procedure described above afford-
ing a colourless oil in 95% yield after purification by
4.2.4. Methyl 3-(3,4-dichlorophenyl)-bicyclo[2.2.1]hept-2-
ene-2-carboxylate (9d). Purification by flash chromato-
graphy (diethyl ether–pentane; 1/30) afforded 9d as
colourless oil in 93% yield (88 mg, 0.30 mmol). 1H
NMR (CDCl3) d 7.64 (s, 1H), 7.39 (s, 2H), 3.69 (s,
3H), 3.40 (br s, 1H), 3.22 (br s, 1H), 1.91 (m, 2H),
1.64 (ddd, 1H, J1 = 1.9 Hz, J2 = 4.0 Hz, J3 = 8.5 Hz),
1.36 (m, 2H), 1.26 (d, 1H, J = 8.5 Hz); 13C NMR
(CDCl3) d 165.5, 155.3, 135.2, 134.3, 132.1, 131.9,
130.2, 129.7, 127.9, 51.3, 50.0, 46.7, 45.5, 26.0, 25.7;
HR-MS (ES) calcd for C15H14Cl2O2 + Na+: m/z
319.0269. Found m/z 319.0268.
1
flash chromatography (diethyl ether–pentane; 1/25). H
NMR: d 7.25–7.23 (m, 1H), 7.22 (s, 1H), 7.13 (t, 1H),
3.68 (s, 3H), 3.37–3.36 (m, 1H), 3.24–3.23 (m, 1H),
2.27 (d, 3H), 1.92–1.86 (m, 2H), 1.64 (dt, 1H), 1.35 (tt,
2H), 1.24 (dt, 1H); 13C NMR: d 166.0, 159.6, 156.8,
134.7, 133.0, 130.7, 125.1, 124.1, 115.1, 51.3, 50.2,
46.7, 45.7, 26.3, 25.8, 14.6; HR-MS (ES) calcd for
C16H17FO2 + Na+ : 283.1110. Found: 283.1118.
4.2.10. Methyl 3-(2,4,5-trimethyl-phenyl)-bicyclo[2.2.1]-
hept-2-ene-2-carboxylate (9p). Prepared by employing
the Suzuki coupling procedure described above afford-
ing a colourless oil in 58% yield after purification by
4.2.5. Methyl 3-(3-chlorophenyl)-bicyclo[2.2.1]hept-2-ene-
2-carboxylate (9e). Purification by flash chromatography
(diethyl ether–pentane; 1/30) afforded 9e as colourless oil
1
flash chromatography (diethyl ether–pentane; 1/25). H
NMR: d 6.98 (s, 1H), 6.76 (s, 1H), 3.58 (s, 3H), 3.38–
3.37 (m, 1H), 3.08–3.06 (m, 1H), 2.22 (s, 3H), 2.20 (s,
3H), 2.16 (s, 3H), 1.88–1.79 (m, 2H), 1.72 (dt, 1H),
1.41–1.32 (m, 2H), 1.25 (dt, 1H); 13C NMR: d 165.9,
159.5, 136.1, 134.4, 133.7, 133.3, 132.9, 131.5, 128.7,
51.2, 51.1, 47.2, 44.4, 26.0, 24.9, 19.6, 19.4, 19.3; HR-
MS (ES) calcd for C18H22O2 + Na+: 293.1517. Found:
293.1517.
1
in 87% yield (50 mg, 0.19 mmol). H NMR (CDCl3) d
7.50 (br s, 1H), 7.40 (m, 1H), 7.27 (m, 2H), 3.68 (s,
3H), 3.39 (br s, 1H), 3.23 (br s, 1H), 1.90 (m, 2H),
1.65 (d, 1H, J = 8.8 Hz), 1.36 (m, 2H), 1.25 (d, 1H,
J = 8.8 Hz); 13C NMR (CDCl3) d 165.6, 156.8, 137.1,
133.8, 133.7, 129.0, 128.3, 128.2, 126.6, 51.2, 50.1,
46.8, 45.5, 26.0, 25.7; HR-MS (ES) calcd for
C15H15ClO2 + Na+: m/z 285.0658. Found m/z 285.0649.
4.2.11. Methyl 3-phenyl-7-oxabicyclo[2.2.1]hept-2-ene-2-
carboxylate (10a). Purification by flash chromatography
(diethyl ether–pentane; 1/30) afforded 10a as colourless
crystals (mp 89.5–90.0 °C) in 93% yield (107 mg,
4.2.6. Methyl 3-(2-chlorophenyl)-bicyclo[2.2.1]hept-2-ene-
2-carboxylate (9g). Purification by flash chromatography
(diethyl ether–pentane; 1/30) afforded 9g as colourless
1
oil in 50% yield (32 mg, 0.12 mmol). H NMR (CDCl3)
0.46 mmol). 1H NMR (CDCl3)
d 7.67 (d, 2H,
d 7.39 (m, 1H), 7.22 (m, 2H), 7.08 (m, 1H), 3.58 (s, 3H),
3.40 (br s, 1H), 3.21 (br s, 1H), 1.83 (m, 3H), 1.35 (m,
3H); 13C NMR (CDCl3) d 165.3, 155.9, 135.9, 135.5,
132.7, 129.5, 129.4, 128.9, 126.2, 51.1, 50.5, 47.3, 44.2,
25.7, 24.6; HR-MS (ES) calcd for C15H15ClO2 + Na+:
m/z 285.0658. Found m/z 285.0659.
J = 6.7 Hz), 7.38 (m, 3H), 5.35 (d, 1H, J = 3.0 Hz),
5.31 (d, 1H, J = 3.8 Hz), 3.74 (s, 3H), 2.03 (m, 2H),
1.53 (m, 2H); 13C NMR (CDCl3) d 164.2, 155.8, 131.8,
131.3, 129.4, 128.6, 128.2, 83.1, 81.0, 51.5, 25.5, 25.0;
HR-MS (ES) calcd for C14H14O3 + Na+: m/z 253.0841.
Found m/z 253.0838.
4.2.7. Methyl 3-(2,3-dichlorophenyl)-bicyclo[2.2.1]hept-2-
ene-2-carboxylate (9h). Purification by flash chromatog-
raphy (diethyl ether–pentane; 1/30) afforded 9h as
colourless oil in 59% yield (49 mg, 0.16 mmol). 1H
4.2.12. Methyl 3-(4-chlorophenyl)-7-oxabicyclo[2.2.1]hept-
2-ene-2-carboxylate (10c). The reaction mixture was
filtered and the solvents were evaporated under reduced
pressure. Flash chromatography (pentane/EtOAc 10:1)

H. S. Christensen et al. / Bioorg. Med. Chem. 15 (2007) 5262–5274
5271
afforded 10c in 79% yield (42 mg, 0.16 mmol). 1H NMR
(CDCl3) d 7.63 (d, 2H, J = 8.6 Hz), 7.35 (d, 2H,
J = 8.6 Hz), 5.34 (d, 1H, J = 3.4 Hz), 5.27 (d, 1H,
J = 3.7 Hz), 3.75 (s, 3H), 2.02 (m, 2H), 1.51 (m, 2H);
13C NMR (CDCl3) d 163.8, 154.4, 135.3, 132.2, 129.5,
129.8, 128.3, 82.7, 80.8, 51.4, 25.2, 24.8; HR-MS (ES)
calcd for C18H16O3 + Na+: m/z 287.0451. Found m/z
287.0450.
under reduced pressure and flash chromatography in
DCM/MeOH 10:1 afforded 11b as a yellow/brown oil
in 99% yield. 1H NMR (CDCl3) d 7.72 (d, 2H,
J = 8.4 Hz), 7.59 (d, 4H, J = 8.5 Hz), 7.42 (t, 2H,
J = 7.4 Hz), 7.34 (t, 1H, J = 7.4 Hz), 4.55 (d, 1H,
J = 1.6 Hz), 4.52 (d, 1H, J = 2.3 Hz), 3.73 (s, 3H), 2.06
(m, 2H), 1.44 (m, 2H); 13C NMR (CDCl3) d 165.2,
158.3, 141.6, 140.2, 134.7, 131.5, 128.8, 128.7, 126.9,
126.6, 127.5, 66.4, 63.5, 51.2, 25.4, 25.0; HR-MS (ES)
calcd for C20H19NO2 + H+: m/z 306.1494. Found m/z
306.1494.
4.2.13. Methyl 3-(3,4-dichlorophenyl)-7-oxabicyclo-
[2.2.1]hept-2-ene-2-carboxylate (10d). The reaction mix-
ture was filtered through Celite and purified by flash
chromatography (pentane/EtOAc 10:1) to yield the
product, 10d, as brown/yellowish oil (34 mg, 11 mmol,
4.2.18. Methyl 3-(4-chlorophenyl)-7-tert-butoxycarbonyl-
7-azabicyclo[2.2.1]hept-2-ene-2-carboxylate (Boc-11c).
The reaction mixture was filtered through Celite and
purified by flash chromatography (pentane/EtOAc
10:1) to yield the product, Boc-11c, as colourless oil
1
56%). H NMR (CDCl3) d 7.80 (d, 1H, J = 2.1 Hz),
7.53 (d, 1H, J = 2,1 Hz), 7.46 (s, 1H), 5.34 (dd, 1H,
J1 = 1.1 Hz, J2 = 4.3 Hz), 5.26 (dd, 1H, J1 = 1.0 Hz,
J2 = 4.6 Hz), 3.76 (s, 3H), 2.04 (m, 2H), 1.54 (m, 2H);
13C NMR (CDCl3) d 162.8, 152.3, 143.2, 139.8, 132.5,
132.2, 129.4, 129.2, 127.0, 81.6, 80.1, 50.7, 24.3, 24.0.
1
(65 mg, 0.18 mmol, 89%). H NMR (CDCl3) d 7.56 (d,
2H, J = 8.1 Hz), 7.35 (d, 2H, J = 8.1 Hz), 5.06 (s, 1H),
5.00 (s, 1H), 3.74 (s, 3H), 2.09 (m, 2H), 1.49 (m, 2H),
1.45 (s, 9H); 13C NMR (CDCl3) d = 163.5, 154.5,
154.4, 134.9, 129.9, 129.6, 127.9, 80.1, 65.0, 62.4, 51.1,
4.2.14. Methyl 3-(4-methoxy-phenyl)-7-oxobicyclo[2.2.1]hept-
2-ene-2-carboxylate (10f). Prepared by employing the
Suzuki coupling procedure described above affording a
yellow oil in 53% yield after purification by flash chro-
matography (EtOAc–pentane; 1/5). 1H NMR: d 7.67
(d, 2H); 6.89 (d, 2H) 5.30 (m, 2H); 3.81 (s, 3H); 3.73
(s, 3H); 1.89 (m, 2H); 1.46 (m, 2H).
27.7,
25.2,
25.1;
HR-MS
(ES)
calcd
for
C19H22NO4Cl + Na+: m/z 386.1135. Found m/z
386.1137.
4.2.19. Methyl 3-(4-chlorophenyl)-7-azabicyclo[2.2.1]hept-
2-ene-2-carboxylate (11c). Boc-11c (65 mg, 0.18 mmol)
was dissolved in 10 mL DCM and reacted with 2 mL
TFA (26 mmol) for 20 h which resulted in removal
of the Boc-group. The solvent and TFA were evapo-
rated under reduced pressure and flash chromatogra-
4.2.15. Methyl 3-phenyl-7-azabicyclo[2.2.1]hept-2-ene-2-
carboxylate (11a). Crude Boc-protected 11a, a yellowish
oil (320 mg, 0.97 mmol), was dissolved in 10 mL DCM
and reacted with 2 mL TFA (26 mmol) for 20 h which
resulted in removal of the Boc-group. Purification by
flash chromatography (1/15, MeOH–DCM) yielded
11a as a yellowish oil in 95% yield (211 mg, 0.92 mmol).
1H NMR (CDCl3) d 7.59 (m, 2H), 7.36 (m, 3H), 4.77
(br s, 1H), 4.76 (br s, 1H), 3.71 (s, 3H), 2.20 (m, 2H),
1.51 (m, 2H); 13C NMR (CDCl3) d 163.7, 155.0, 130.9,
129.8, 128.7, 128.2, 65.7, 62.9, 51.7, 24.1, 23.4; HR-MS
(ES) calcd for C14H15NO2 + H+: m/z 230.1181. Found
m/z 230.1016.
phy in DCM/MeOH 10:1 afforded 11c as
a
1
colourless oil in 99% yield. H NMR (CDCl3) d 7.59
(d, 2H, J = 8.5 Hz), 7.33 (d, 2H, J = 8.5 Hz), 4.55 (d,
1H, J = 2.6 Hz), 4.45 (d, 1H, J = 2.9 Hz), 3.73 (s,
3H), 2.04 (m, 2H), 1.43 (m, 2H); 13C NMR (CDCl3)
d = 164.9, 157.3, 135.2, 134.8, 131.0, 129.6, 128.2,
66.3, 63.5, 51.3, 25.3, 24.9; HR-MS (ES) calcd for
C14H14NO2Cl + H+: m/z 264.0791. Found m/z
264.0785.
4.2.20. Methyl 3-(3,4-dichlorophenyl)-7-N-tert-butyl-7-
azabicyclo[2.2.1]hept-2-ene-2-carboxylate (Boc-11d). The
reaction mixture was filtered through Celite and puri-
fied by flash chromatography (pentane/EtOAc 10:1) to
yield the product, Boc-11d, as brown/yellowish oil in
4.2.16. Methyl 3-(4-biphenyl)-7-tert-butoxycarbonyl-7-
azabicyclo[2.2.1]hept-2-ene-2-carboxylate (Boc-11b). The
reaction mixture was filtered through Celite and purified
by flash chromatography (pentane/EtOAc 10:1) to yield
the product, Boc-11b, as a colourless solid (mp 92–
1
97% (77 mg, 0.19 mmol). H NMR (CDCl3) d 7.91 (s,
1
95 °C) (71 mg, 0.18 mmol, 87%). H NMR (CDCl3) d
1H), 7.64 (s, 1H), 7.45 (s, 1H), 5.25 (s, 1H), 5.13 (s,
1H), 3.94 (s, 3H), 2.30 (m, 2H), 1.63 (m, 2H), 1.61
(s, 9H).
7.72 (d, 2H, J = 8.5 Hz), 7.62 (m, 4H), 7.45 (t, 2H,
J = 7.2 Hz), 7.37 (t, 1H, J = 7.2 Hz), 5.09 (s, 1H), 5.06
(s, 1H), 3.76 (s, 3H), 2.11 (m, 2H), 1.53 (m, 2H), 1.44
(s, 9H); 13C NMR (CDCl3) d 164.6, 155.2, 142.4, 140.6,
131.1, 129.9, 129.8, 129.5, 129.1, 127.3, 127.1, 127.9,
80.7, 65.8, 63.2, 51.8, 28.5, 28.3, 28.2; HR-MS (ES) calcd
for C25H27NO4 + Na+: m/z 428.1840. Found m/z
428.1831.
4.2.21. Methyl 3-(3,4-dichlorophenyl)-7-azabicyclo[2.2.1]-
hept-2-ene-2-carboxylate (11d). Boc-11d (77 mg, 0.19 mmol)
was dissolved in 10 mL DCM and reacted with 2 mL TFA
(26 mmol) for 20 h, which resulted in removal of the Boc-
group. The solvent and TFA were evaporated under re-
duced pressure and flash chromatography in DCM/MeOH
15:2 afforded 11d as a colourless oil in 97% yield. 1H NMR
(CDCl3) d 7.75 (s, 1H), 7.50 (d, 1H, J = 8.4 Hz), 7.43 (d,
1H, J = 8.4 Hz), 4.56 (m, 1H), 4.44 (m, 1H), 3.74 (s, 3H),
2.03 (m, 2H), 1.41 (m, 2H); 13C NMR (CDCl3) d 164.6,
155.8, 136.2, 132.6, 132.4, 132.0, 129.8, 129.7, 127.4, 66.0,
4.2.17. Methyl 3-(4-biphenyl)-7-azabicyclo[2.2.1]hept-2-
ene-2-carboxylate (11b). Boc-11b (71 mg, 0.18 mmol)
was dissolved in 10 mL DCM and reacted with 2 mL
TFA (26 mmol) for 20 h which resulted in removal of
the Boc-group. The solvent and TFA were evaporated

5272
H. S. Christensen et al. / Bioorg. Med. Chem. 15 (2007) 5262–5274
63.4, 51.3, 25.1, 24.8; HR-MS (ES) calcd for
4.2.27. Methyl 2-(4-acetylphenyl)cyclopent-1-enecarb-
oxylate (12n). Purification by flash chromatography
(diethyl ether–pentane; 1/1) afforded 9n as colourless
C14H13NO2Cl2 + H+: m/z 298.0402. Found m/z 298.0396.
1
4.2.22. Methyl 3-(3-chlorophenyl)-7-N-tert-butoxycarbonyl-
7-azabicyclo[2.2.1]hept-2-ene-2-carboxylate (Boc-11e). The
reaction mixture was filtered through Celite and purified
by flash chromatography (pentane/EtOAc; 10:1) to yield
the product, Boc-11e, as brown/yellowish oil in 91%
oil in 79% yield (50 mg, 0.20 mmol). H NMR (CDCl3)
d 7.92 (d, 2H, J = 8.5 Hz); 7.39 (d, 2H, J = 8.5 Hz); 3.62
(s, 3H); 2.85 (m, 4H); 2.59 (s, 3H); 2.01 (p, 2H,
J = 7.7 Hz). 13C NMR (CDCl3) d 197.6; 166.1; 152.5 ;
142.0; 136.2; 130.5; 127.9, 127.8; 51.3; 40.0; 35.1; 26.6;
22.0. HR-MS (ES) calcd for C15H16O3 + Na+:
267.0997. Found: 267.1000.
1
(66 mg, 0.18 mmol). H NMR (CDCl3) d 7.59 (s, 1H),
7.48 (d, 1H, J = 6.9 Hz), 7.32 (m, 2H), 5.06 (s, 1H),
5.00 (s, 1H), 3.74 (s, 3H), 2.09 (m, 2H), 1.48 (m, 2H),
1.43 (s, 9H); 13C NMR (CDCl3) d 163.9, 155.0, 154.9,
134.1, 133.8, 129.4, 129.3, 128.6, 126.8, 80.6, 65.6,
62.9, 51.6, 28.2, 25.5, 25.1; HR-MS (ES) calcd for
C19H22NO4Cl + Na+: m/z 386.1135. Found m/z
386.1136.
4.2.28. Methyl 3-(2-naphthyl)-bicyclo[2.2.1]hept-2-ene-2-
carboxylate (14). Purification by flash chromatography
(diethyl ether–pentane; 1/30) afforded 14 as colourless
1
oil in 63% yield (42 mg, 0.15 mmol). H NMR (CDCl3)
d 8.00 (s, 1H), 7.83 (m, 2H), 7.79 (d, 1H, J = 8.4 Hz),
7.63 (d, 1H, J = 8.4 Hz), 7.47 (m, 2H), 3.68 (s, 3H),
3.44 (br s, 1H), 3.40 (br s, 1H), 1.95 (m, 2H), 1.74 (d,
1H, J = 8.8 Hz), 1.45 (m, 2H), 1.30 (d, 1H,
J = 8.8 Hz); 13C NMR (CDCl3) d 166.0, 158.0, 133.1,
4.2.23. Methyl 3-(3-chlorophenyl)-7-azabicyclo[2.2.1]hept-
2-ene-2-carboxylate (11e). Boc-11e (66 mg, 0.18 mmol)
was dissolved in 10 mL DCM and reacted with 2 mL
TFA (26 mmol) for 20 h, which resulted in removal of
the Boc-group. The solvent and TFA were evaporated un-
der reduced pressure and flash chromatography in DCM/
MeOH 15:2 afforded 11e as a colourless oil in 99% yield.
1H NMR (CDCl3) d 7.62 ( s, 1H), 7.52 (m, 1H), 7.31 (m,
2H), 4.58 (d, 1H, J = 3.4 Hz), 4.49 (d, 1H, J = 3.8 Hz),
3.74 (s, 3H), 2.07 (m, 2H), 1.44 (m, 2H); HR-MS (ES)
calcd for C14H14NO2Cl + H+: m/z 264.0791. Found m/z
264.0793.
132.9 2, 132.8, 128.3, 127.5 2, 127.1, 126.5, 126.3,
*
*
126.1, 51.1, 50.3, 46.7, 45.5, 26.2, 25.8; HR-MS (ES)
calcd for C19H18O2 + Na+: m/z 301.1204. Found m/z
301.1205.
4.2.29. Methyl 3-(2-naphthyl)-7-oxabicyclo[2.2.1]hept-2-
ene-2-carboxylate (15). The reaction mixture was filtered
through Celite and purified by flash chromatography
(pentane–EtOAc 10/1) to yield the product, 15, as a yel-
low solid (38 mg, 0.14 mmol, 68%). 1H NMR (CDCl3) d
8.13 (s, 1H), 7.83 (m, 4H), 7.50 (m, 2H), 5.44 (d, 2H,
J = 4.6 Hz), 5.41 (d, 2H, J = 4.2 Hz), 3.76 (s, 3H), 2.06
(m, 2H), 1.59 (m, 2H). 13C NMR (CDCl3) d 164.3,
155.8, 133.6, 132.9, 132.1, 28.8, 128.6, 128.4, 127.8,
127.6, 127.1, 126.4, 126.1, 83.2, 81.2, 51.5, 25.6, 25.2;
HR-MS (ES) calcd for C18H16O3 + Na+: m/z 303.0997.
Found m/z 303.0999.
4.2.24. Methyl 3-(3-nitro-phenyl)-7-azabicyclo[2.2.1]hept-
2-ene-2-carboxylate (11j). Prepared by employing the
Suzuki coupling procedure described above affording a
colourless oil in 97% yield after purification by flash
chromatography (MeOH–DCM; 1/12). 1H NMR: d
8.46 (s, 1H); 8.20 (d, 1H); 7.94 (d, 1H); 7.56 (t, 1H);
4.79 (s, 1H); 4.78 (s, 1H); 3.75 (s, 3H); 2.23 (m, 2H);
1.57 (m, 2H). 13C NMR: d 162.9, 152.7, 147.7, 134.2,
129.0, 123.8, 123.2, 133.9, 132.7, 65.3, 62.6, 51.7, 23.8,
23.5. HR-MS (ES) calcd for C14H14N2O4 + H+:
275.1032. Found: 275.1027.
4.2.30. Methyl 3-(2-naphthyl)-7-tert-butoxycarbonyl-7-
azabicyclo[2.2.1]hept-2-ene-2-dicarboxylate (Boc-16). The
reaction mixture was filtered through Celite and purified
by flash chromatography (pentane–EtOAc, 10/1) to yield
the product, Boc-16, as brown/yellowish oil in 90%
4.2.25. Methyl 3-(4-nitro-phenyl)-7-azabicyclo[2.2.1]hept-
2-ene-2-carboxylate (11r). Prepared by employing the
Suzuki coupling procedure described above affording a
colourless oil in 97% yield after purification by flash
chromatography (MeOH–DCM; 1/12) 1H NMR: d.
8.20 (d, 2H) 7.77 (d, 2H); 4.60 (d, 1H); 4.48 (d, 1H);
3.73 (s, 3H); 2.07 (m, 2H); 1.46 (m, 2H). 13C NMR: d
164.6, 156.3, 147.5, 139.6, 129.9, 129.1, 128.3, 123.2,
66.6, 63.8, 51.7, 25.2. HR-MS (ES) calcd for
C14H14N2O4 + H+: 275.1032. Found: 275.1040.
1
(68 mg, 0.18 mmol). H NMR (CDCl3) d 8.09 (s, 1H),
7.87 (m, 2H), 7.83 (d, 1H, J = 8.4 Hz), 7.69 (d, 1H,
J = 8.4 Hz), 7.50 (m, 2H), 5.12 (s, 2H), 3.75 (s, 3H),
2.13 (m, 2H), 1.57 (m, 2H), 1.45 (s, 9H); 13C NMR
(CDCl3) d 164.1, 154.9, 142.2, 133.3, 132.6, 128.2,
128.3, 127.4 2, 126.8, 126.2, 126.0, 80.3, 62.8, 58.7,
*
51.3, 28.0, 27.9 2; HR-MS (ES) calcd for C H NO + -
*
23 25
4
Na+: m/z 402.1681. Found m/z 402.1678.
4.3. Binding assays
4.2.26. Methyl 2-(3-aminophenyl)cyclopent-1-enecarb-
oxylate (12m). Purification by flash chromatography
(diethyl ether–pentane; 1/1) afforded 9m as yellowish
1. Cell cultures.8 Cell lines stably expressing hDAT were
established by transfecting COS-1 cells with hDAT
inserted in the pIRES vector (BD Biosciences Clon-
tech) also carrying a Blasticidin resistance gene. Cells
were cultured in DMEM (BioWhitaker) supple-
mented with 10% FCS (Gibco Life Technologies),
1% penicillin/streptomycin (BioWhitaker) and 10 lg/
mL of Blasticidin (Cayla) selection of transfected
cells. After 14 days of selection, blasticidin was
1
oil in 86% yield (30 mg, 0.14 mmol). H NMR (CDCl3)
d 7.11 (t, 1H, J = 7.8 Hz); 6.70 (d, 1H, J = 7.6 Hz); 6.65
(s, 1H); 6.62 (d, 1H, J = 7.9 Hz); 3.63 (s, 3H); 2.81 (t,
4H, J = 7.7 Hz + H5a+b); 1.97 (p, 2H, J = 7.7 Hz). 13C
NMR (CDCl3) d 166.8; 153.5 ; 145.9; 138.0; 128.7;
118.1; 114.8; 114.3; 51.1; 40.1; 35.1; 21.9. HR-MS (ES)
calcd for C13H15NO2 + H+: 218.1181. Found: 218.1179.

H. S. Christensen et al. / Bioorg. Med. Chem. 15 (2007) 5262–5274
5273
adjusted to 2 lg/mL in the culture medium and the
cells were subcultured under this selection regime
and grown at 37 °C, 5% CO2 and 95% humidity.
2. Binding assay.8 Membrane preparations for the bind-
ing assay were produced by scraping the stably trans-
fected cells from cell culture dishes (Nunc), pelleting
the cells in ice-cold PBSCM by centrifugation and
homogenizing the cells in ice-cold harvest buffer I
(150 mM NaCl, 50 mM Tris, 20 mM EDTA) using
an Ultra-Turrax (Janke and Kunkel AG) for 60 s.
The membrane was pelleted by centrifugation at
12000g for 10 min at 4 °C and washed in ice-cold
Harvest buffer I. The membranes were pelleted again
and, finally, resuspended in PBSCM using Ultratur-
rax briefly. Membrane preparations were aliquoted
into 2 mL portions and stored at ꢀ80 °C until use.
The concentration of total protein in the membrane
preparation was determined with the MicroBCA kit
(Pierce). A concentration of 5 lg/well of membrane
preparation was used with the chosen concentration
of drug of interest in combination with 0.1–0.25 nM
125I-RTI-55. Membrane and ligands were incubated
for 1 h at 20 °C using a Filtermate cell harvester
(Packard), membranes were captured on GF/B
96-well filterplates (Packard) presoaked with 0.5%
polyethyleneimine (Merck) and washed thrice with
ice-cold water. The filter in each well was dissolved
in 40 lL Microscint 20 and scintillation counts were
determined with a Packard Topcounter. Precise
concentration of radioligand was quantified by liquid
scintillation counting on a Packard Tri-Carb.
tions will be included in the parent population and sub-
stitute for the worst of the parents. This will leave the
best 100 equations for the next evolution. The scoring
of the equations was validated by calculating the predic-
tive residual sum of squares (PRESS) and through the
cross-validation test implemented in Cerius2.19 The
highest q2 and the lowest difference between q2 and r2
were used to select the best equation. Furthermore, the
ability of the best equation to predict binding of com-
pounds in the test set was assessed.
5.3. 3D QSAR
Minimized structures of the compounds were imported
into Catalyst. Compounds not containing conforma-
tional information were drawn and minimized in Cata-
lyst and subsequently subjected to a conformational
search using ‘Best’ conformational generation, as rec-
ommended when utilizing the HypoGen module. The
training set was chosen from the libraries with respect
to activities, so that the most active and the least active
along with some intermediate actives from each library
were chosen for the HypoGen generation. The biologi-
cal data were entered manually. All parameter values
were left to default, and the features were chosen to be
hydrogen-bond acceptor, hydrophobe, hydrophobe ali-
phatic, hydrophobic aromatic and ring aromatic. Each
hypothesis was set to contain a minimum of 5 and a
maximum of 10 features. The generated models were
evaluated in terms of cost analysis, and the capability
of predicting activities was evaluated with a test set
and a prediction set.
3. Data analysis. Counts from the Packard Topcounter
were fitted to a sigmoidal dose–response curve using
the built-in nonlinear regression tool in the Graphpad
Prism 3 software. From at least three independent
experiments, the resulting IC50 values were trans-
formed into Ki values using the equation described
by Cheng and Prusoff.32
Acknowledgments
We are indebted to Steffen Sinning and Ove Wiborg for
performing the biological tests. We also thank the Lund-
beck Foundation, the Carlsberg Foundation and FNU
for financial support.
5. Molecular modelling
5.1. General
Supplementary data
Calculations were performed on a Silicon Graphics Oc-
tane2 R12000 workstation running under the IRIX
6.5.12 operating system. QSAR relations were computed
using either Cerius2 or Catalyst software packages.19,20
Initial drawing and minimization of the compounds
were done in Cerius2 or Catalyst.
Supporting information contains carbon-13 NMR spec-
tra of compounds 9o, 9p, 10f, 11j, 11r, 12m and 12n (8
pages), and the results of catscramble analysis of hypo
1. Supplementary data associated with this article can
be found, in the online version, at doi:10.1016/
j.bmc.2007.05.015.
5.2. 2D QSAR
References and notes
The biological activities were entered manually. The 2D
QSAR equations were generated using the G/PLS meth-
od in Cerius2. A population of 100 randomly con-
structed equations was initially built by use of the
default settings and the built-in descriptors. These equa-
tions were then evolved for 50,000 generations (default
5000). Evolving the equations means that, for each gen-
eration, two better scoring equations are selected as par-
ents. Parts of each parent equation are then combined
into a child equation. Only better scoring child equa-
1. Carroll, F. I.; Howell, L. L.; Kuhar, M. J. Med. Chem.
1999, 42, 2721–2736.
2. Singh, S. Chem. Rev. 2000, 100, 925–1024.
3. Kuhar, M. J.; Ritz, M. C.; Boja, J. W. Trends Neurosci.
1991, 14, 299–302.
4. Carroll, F. I. J. Med. Chem. 2003, 46, 1775–1794.
5. Pedersen, H.; Sinning, S.; Bulow, A.; Wiborg, O.; Falborg,
¨
L.; Bols, M. Org. Biomol. Chem. 2004, 2, 2861–2869.
6. Iso, Y.; Grajkowska, E.; Wroblewski, J. T.; Davis, J.;
Goeders, N. E.; Johnson, K. M.; Sanker, S.; Roth, B. L.;

5274
H. S. Christensen et al. / Bioorg. Med. Chem. 15 (2007) 5262–5274
Tueckmantel, W.; Kozikowski, A. P. J. Med. Chem. 2006,
49, 1080–1100.
7. Carroll, F. I.; Lewin, A. H.; Boja, J. W.; Kuhar, M. J.
21. Zhang, C.; Izenwasser, S.; Katz, J. L.; Terry, P. D.;
Trudell, M. L. J. Med. Chem. 1998, 41, 2430–2435.
22. Meltzer, P. C.; Liang, A. Y.; Blundell, P.; Gonzalez, M.
D.; Chen, Z.; George, C.; Madras, B. K. J. Med. Chem.
1997, 40, 2661–2673.
23. Meltzer, P. C.; Blundell, P.; Young, Y. F.; Chen, Z.;
George, C.; Gonzalez, M. D.; Madras, B. K. J. Med.
Chem. 2000, 43, 2982–2991.
J. Med. Chem. 1992, 35, 969–981.
8. Bulow, A.; Sinning, S.; Wiborg, O.; Bols, M. J. Comb.
¨
Chem. 2004, 6, 509–519.
9. Newman, A. H.; Allen, A. C.; Izenwasser, S.; Katz, J. L.
J. Med. Chem. 1994, 37, 2258–2261.
10. Kozikowski, A. P.; Araldi, G. L.; Boja, J. W.; Meil, W.
M.; Johnson, K. M.; Flippen-Anderson, J. L.; George, C.;
Saiah, E. J. Med. Chem. 1998, 41, 1962–1969.
11. Tamiz, A. P.; Zhang, J.; Flippen-Anderson, J. L.; Zhang,
M.; Johnson, K. M.; Deschaux, O.; Tella, S.; Kozikowski,
A. P. J. Med. Chem. 2000, 43, 1215–1222.
12. Enyedy, I. J.; Zaman, W. A.; Sakamuri, S.; Kozikowski,
A. P.; Johnson, K. M.; Wang, S. Bioorg. Med. Chem. Lett.
2001, 11, 1113–1118.
24. Goulet, M.; Miller, G. M.; Bendor, J.; Liu, S.; Meltzer, P.
C.; Madras, B. K. Synapse 2001, 129–140.
25. Petersen, M. D.; Boye, S. V.; Nielsen, E. H.; Sinning, S.;
Wiborg, O.; Bols, M. Bioorg. Med. Chem. 2007, 15, 4159–
4174.
26. Fan, Y.; Shi, L. M.; Kohn, K. W.; Pommier, Y.;
Weinstein, J. N. J. Med. Chem. 2001, 44, 3254–3263.
27. Rogers, D.; Hopfinger, A. J. J. Chem. Inf. Comput. Sci.
1994, 34, 854–866.
13. Meltzer, P. C.; Butler, D.; Deschamps, J. R.; Madras, B.
K. J. Med. Chem. 2006, 49, 1420–1432.
28. Glen, W. G.; Dunn, W. J., III; Scott, D. R. Tetrahedron
Comput. Meth. 1989, 2, 349–376.
14. Crammer, R. D.; Patterson, D. E.; Bunce, J. D. J. Am.
Chem. Soc. 1988, 110, 5959–5967.
15. Carroll, F. I.; Gao, Y. G.; Rahman, M. A.; Abraham, P.;
Parham, K.; Lewin, A. H.; Boja, J. W.; Kuhar, M. J.
J. Med. Chem. 1991, 34, 2719–2725.
29. Wold, S. In Chemometric Methods in Molecular Design;
Van der Waterbeemd, H., Ed.; VCH Weinheim, 1995; pp
195–218.
30. Li, H.; Sutter, J.; Hoffmann, R. HypoGen: an automated
system for generating 3D predictive pharmacophore
models. In: Pharmacophore Perception, Development, and
Use in Drug Design; International University: La Jolla,
CA, 2000, pp 171–189.
31. Carroll, F. I.; Mascarella, S. W.; Kuzemko, M. A.; Gao,
Y.; Abraham, P.; Lewin, A. H.; Boja, J. W.; Kuhar, M. J.
J. Med. Chem. 1994, 37, 2865–2873.
32. Cheng, Y.; Prusoff, W. H. Biochem. Pharmacol. 1973, 22,
3099–3108.
16. Newman, A. H.; Izenwasser, S.; Robarge, M. J.; Kline, R.
H. J. Med. Chem. 1999, 42, 3502–3509.
17. Kulkarni, S. S.; Grundt, P.; Kopajtic, T.; Katz, J. L.;
Newman, A. H. J. Med. Chem. 2004, 47, 3388–3398.
18. Yuan, H.; Kozikowski, A. P.; Petukhov, P. A. J. Med.
Chem. 2004, 47, 6137–6143.
19. Cerius2 version 4.9, Accelrys: San Diego, USA, 2004.
20. Catalyst version 4.10, Accelrys: San Diego, USA, 2005.