Discovery of new cholesteryl ester transfer protein inhibitors via ligand-based pharmacophore modeling and QSAR analysis followed by synthetic exploration

Article abstract of DOI:10.1016/j.ejmech.2009.12.070

Cholesteryl ester transfer protein (CETP) is involved in trafficking lipoprotein particles and neutral lipids between HDL and LDL and therefore is considered a valid target for treating dyslipidemic conditions and complications. Pharmacophore modeling and quantitative structure-activity relationship (QSAR) analysis were combined to explore the structural requirments for potent CETP inhibitors. Two pharmacophores emerged in the optimal QSAR equation (r² = 0.800, n = 96, F = 72.1, r²_LOO = 0.775, r²_PRESS against 22 external test inhibitors = 0.707) suggesting the existence of at least two distinct binding modes accessible to ligands within CETP binding pocket. The successful pharmacophores were complemented with strict shape constraints in an attempt to optimize their receiver-operating characteristic (ROC) curve profiles. The validity of our modeling approach was experimentally established by the identification of several CETP inhibitory leads retrieved via in silico screening of the National Cancer Institute (NCI) list of compounds and an in house built database of drugs and agrochemicals. Two hits illustrated low micromolar IC₅₀ values: NSC 40331 (IC₅₀ = 6.5 μM) and NSC 89508 (IC₅₀ = 1.9 μM). Active hits were then used to guide synthetic exploration of a new series of CETP inhibitors.

Full text of DOI:10.1016/j.ejmech.2009.12.070

European Journal of Medicinal Chemistry 45 (2010) 1598–1617
Contents lists available at ScienceDirect
European Journal of Medicinal Chemistry
journal homepage: http://www.elsevier.com/locate/ejmech
Original article
Discovery of new cholesteryl ester transfer protein inhibitors via ligand-based
pharmacophore modeling and QSAR analysis followed by synthetic exploration
,
Reema Abu Khalaf a, Ghassan Abu Sheikha a, Yasser Bustanji b, Mutasem O. Taha c
*
a Faculty of Pharmacy, Al-Zaytoonah Private University of Jordan, Amman, Jordan
b Department of Biopharmaceutics and Clinical Pharmacy, Faculty of Pharmacy, University of Jordan, Amman, Jordan
c Department of Pharmaceutical Sciences, Faculty of Pharmacy, University of Jordan, Amman, Jordan
a r t i c l e i n f o
a b s t r a c t
Article history:
Cholesteryl ester transfer protein (CETP) is involved in trafficking lipoprotein particles and neutral lipids
between HDL and LDL and therefore is considered a valid target for treating dyslipidemic conditions and
complications. Pharmacophore modeling and quantitative structure–activity relationship (QSAR)
analysis were combined to explore the structural requirments for potent CETP inhibitors. Two phar-
Received 20 July 2009
Received in revised form
16 December 2009
Accepted 23 December 2009
Available online 14 January 2010
macophores emerged in the optimal QSAR equation (r2 ¼ 0.800, n ¼ 96, F ¼ 72.1, r2
¼ 0.775, r2
LOO
PRESS
against 22 external test inhibitors ¼ 0.707) suggesting the existence of at least two distinct binding
modes accessible to ligands within CETP binding pocket. The successful pharmacophores were com-
plemented with strict shape constraints in an attempt to optimize their receiver-operating characteristic
(ROC) curve profiles.
Keywords:
CETP inhibitors
In silico screening
The validity of our modeling approach was experimentally established by the identification of several
CETP inhibitory leads retrieved via in silico screening of the National Cancer Institute (NCI) list of
compounds and an in house built database of drugs and agrochemicals. Two hits illustrated low
Pharmacophore modeling
Quantitative Structure–activity relationship
Shape constraints
Receiver-operating characteristic
micromolar IC50 values: NSC 40331 (IC50 ¼ 6.5
m
M) and NSC 89508 (IC50 ¼ 1.9
mM). Active hits were then
used to guide synthetic exploration of a new series of CETP inhibitors.
Ó 2010 Elsevier Masson SAS. All rights reserved.
1. Introduction
lipoproteins (LDL). CETP, as revealed by X-ray crystallography (PDB
code: 2OBD, resolution 2.2 Å), has a large highly hydrophobic
binding site capable of simultaneously binding up to four lipid
molecules [4]. In human plasma, CETP plays a potentially proa-
therogenic role by moving CE from HDL to very-low-density lipo-
protein (VLDL) and low-density lipoprotein (LDL) particles, thereby
lowering atheroprotective HDLc and raising proatherogenic VLDLc
and LDLc. Apparently, the risk of CAD is proportional to the plasma
levels of CETP [5]. In fact, It is quite common within the CAD pop-
ulation to have elevated CETP plasma protein levels that are 2- to 3-
fold higher than concentrations typically found in the plasma of
Atherosclerosis describes the principal progression in arterial
dysfunction and remodeling that restricts blood flow to vessels in
the peripheral vasculature and is ultimately manifested as coronary
artery disease (CAD) [1]. Several epidemiological studies have
demonstrated an inverse relationship between serum high-density
lipoprotein cholesterol (HDLc) levels and the incidence of ischemic
heart disease [2]. HDL mediates the reverse cholesterol transport
pathway which removes excess cholesterol from peripheral tissues
to the liver for biliary elimination [3].
CETP, a 476-residue glycoprotein, is involved in trafficking lipo-
proteinparticles and neutral lipids, including cholesteryl esters (CE),
phospholipids and triglycerides between HDL and low-density
normal subjects (1–3
mg/mL) [6].
Evidence exists that the consequences of CETP activity may
depend on the metabolic setting, particularly on triglyceride levels.
Accordingly, pharmacological CETP inhibition may reduce the risk
of CAD in humans, but only in those with high triglyceride levels [5].
The unavailability of satisfactory high resolution crystallo-
graphic structures for CETP combined with its prohibitively large
binding pocket confined most modeling-related discovery projects
to ligand-based approaches particularly quantitative structure–
activity relationship analysis (QSAR) [7–9].
Abbreviations: CETP, Cholesteryl ester transfer protein; QSAR, quantitative
structure–activity relationship; ROC, receiver-operating characteristic; NCI, national
cancer institute; CAD, coronary artery disease; HDLc, high-density lipoprotein
cholesterol; CE, cholesteryl esters; LDLc, low-density lipoprotein cholesterol; VLDLc,
very-low-density lipoprotein cholesterol; GFA, genetic function algorithm; MLR,
multiple linear regression; VS, virtual screening; DAC, drugs and agrochemicals.
* Corresponding author. Tel.: þ962 6 5355000x23305; fax: þ962 6 5339649.
E-mail address: mutasem@ju.edu.jo (M.O. Taha).
Despite the excellent predictive potential of 3D-QSAR method-
ologies (e.g., CoMFA and CoMSIA), they generally lack the ability to
0223-5234/$ – see front matter Ó 2010 Elsevier Masson SAS. All rights reserved.
doi:10.1016/j.ejmech.2009.12.070

R. Abu Khalaf et al. / European Journal of Medicinal Chemistry 45 (2010) 1598–1617
1599
act as effective search queries to mine virtual three-dimensional
(3D) databases for new hits [10,11].
The pharmacophoric space of CETP inhibitors was explored
employing three structurally diverse training subsets of inhibitors
(Table 1). The pharmacophoric space of each training subset was
explored via four HYPOGEN runs (Table B under Supplementary
material). We restricted HYPOGEN to explore pharmacophoric
models incorporating from zero to three features of any particular
selected feature type (i.e., HBA, HBD, hydrophobic and Ring Aromatic)
to limit the number of explored pharmacophoric models and there-
fore improve the quality of emerging binding hypotheses [12–17].
Only four and five-featured pharmacophores were explored. Three-
and two-featured pharmacophores are rather promiscuous as 3D
search queries and probably not adequate descriptions of ligand-
CETP binding as judged from the structural diversity of the training
compounds. CATALYST–HYPOGEN can generate pharmacophore
hypotheses of a maximum of five features [22,23].
The continued interest in the development of new CETP inhib-
itors combined with the lack of adequate CETP crystallographic
structures and adequate computer-aided drug discovery efforts in
this area, prompted us to explore the possibility of developing
ligand-based three-dimensional (3D) pharmacophore(s) integrated
within self-consistent QSAR model for CETP inhibitors. The phar-
macophore model(s) can be used as 3D search query(ies) to mine
3D libraries for new CETP inhibitors, while the QSAR model helps to
predict the biological activities of the captured compounds and
therefore prioritize them for in vitro evaluation. We previously
reported the use of this innovative approach towards the discovery
of new inhibitory leads against glycogen synthase kinase 3
b (GSK-
3
b) [12], dipeptidyl peptidase [13], hormone sensitive lipase (HSL)
[14], bacterial MurF [15], protein tyrosine phosphatase 1B (PTP 1B)
[16] and influenza neuraminidase [17].
We allowed a maximum of 4–5 features per pharmacophore in
runs 1 and 2; however, runs 3 and 4 were limited to five-featured
pharmacophores only. The same trend is followed with other
training subsets (i.e., sets II and III). Despite bias towards five-
featured hypotheses in this conduct, differences in the ranges of
allowed pharmacophoric features should force CATALYST to
explore different sections of the pharmacophoric space of CETP
inhibitors. This assumption is supported by significant differences
in the success profiles of resultant pharmacophore models, i.e., cost
criteria, particularly config. costs (Table C under Supplementary
material). Similarly, to explore the optimal pharmacophoric inter-
feature distances, we evaluated two inter-feature spacing values:
100 and 300 pm (as in Table B under Supplementary material).
Eventually, 10 optimal pharmacophoric hypotheses were
generated for each run, yielding 40 models for each set of inhibi-
tors, i.e., 120 models from the three training subsets. Table C, under
Supplementary material, shows the pharmacophoric features and
success criteria of best-performing representative pharmacophores
[26,38]. The resulting pharmacophore models shared comparable
features and acceptable statistical criteria. Emergence of several
comparable binding hypotheses suggests the existence of multiple
binding modes assumed by different CETP ligands within the
binding pocket.
We employed the HYPOGEN module from the CATALYST software
package to construct numerous plausible binding hypotheses for
CETP inhibitors [18]. Subsequently, genetic function algorithm (GFA)
and multiple linear regression (MLR) analysis were employed to
search for an optimal QSAR that combines high-quality binding
pharmacophores with other molecular descriptors and capable of
explaining bioactivity variation across a collection of diverse CETP
inhibitors. The optimal pharmacophores were further validated by
evaluating their ability to successfully classify a list of compounds as
actives or inactives by assessing their receiver-operating character-
istic (ROC) curves. Subsequently, the optimal pharmacophores were
complemented with tight shape constraints to enhance their ROC
profiles. Thereafter, the resulting shape-complemented pharmaco-
phores were used as 3D search queries to screen several available
virtual molecular databases for new CETP inhibitors. Active hits were
employed asguides to synthesize newseries of active CETPinhibitors.
CATALYST models drug–receptor interactions using information
derived from the ligand structures [18–26]. HYPOGEN identifies
a 3D array of a maximum of five chemical features common to
active training ligands that provides relative alignment for each
input molecule consistent with binding to a proposed common
receptor site. The conformational flexibility of training ligands is
modeled by creating multiple conformers that cover a specified
energy range for each input molecule [16,21–23,27–31].
2.2. QSAR modeling
The SHAPE module in CATALYST is a shape-based similarity
searching method. The Van der Waals surface of a molecule (in
a certain conformation) is calculated and represented as a set of
points of uniform average density on a grid. The surface points
enclose avolume on the grid. The geometric centerof the set of points
is computed along with the three principal component vectors
passing through the center. The maximum extents along each prin-
cipal axis and the total volume are calculated. These provide shape
indices that can be compared with the query and used in an initial
screening step to eliminate poor matches from further consideration
[32]. CATALYST pharmacophores, with or without shape constraints,
have been used as 3D queries for database searching and in 3D-QSAR
studies [21,23,27,32].
Despite the significance of pharmacophoric hypotheses in
understanding ligand-macromolecule affinity and as 3D search
queries, their predictive value as 3D-QSAR models is generally
limited by steric shielding and bioactivity-modulating auxiliary
groups (electron-donating or electron-withdrawing functional-
ities) [12–17,25]. This point combined with the fact that our phar-
macophore exploration of CETP inhibitors identified numerous
successful binding hypotheses (as in Table C under Supplementary
material) prompted us to employ classical QSAR analysis to search
for the best combination of orthogonal pharmacophores and other
structural descriptors (connectivity, topological, 2D, etc.) capable of
explaining bioactivity variation across the whole training list
Table 1
2. Results and discussion
Training subsets employed in exploring the pharmacophoric space of CETP inhibi-
tors, numbers correspond to compounds in Table A (under Supplementary material).
2.1. Pharmacophore modeling
Training set Most active subseta Intermediate subset
Least active subseta
I
1, 2, 4, 7, 21
23, 42, 43, 44, 53, 59, 92, 93, 96, 97
The literature was extensively surveyed to collect diverse CETP
inhibitors. A dataset of 118 N,N-disubstituted-3-amino-2-propanol
derivatives (1–118, Table A under Supplementary material) was used
for pharmacophore modeling and subsequent QSAR analysis [33–36].
The conformational space of each inhibitor was sampled utilizing the
poling algorithm implemented within CATALYST [21–23,37].
67, 76, 80, 88, 114
7, 29, 42, 53, 58, 76,
110, 111
II
III
1, 4, 102, 104,
105, 106
3, 8, 10, 11, 16,
17, 21, 23
86, 88, 89, 93, 96,
97, 112, 113
93, 94, 95, 96,
97, 99
36, 50, 54, 67, 68,
80, 83, 89
a
Potency categories as defined by Eqs. (2) and (3).

1600
R. Abu Khalaf et al. / European Journal of Medicinal Chemistry 45 (2010) 1598–1617
(1–118, Table A under Supplementary material). Eq. (1) shows our
best-performing QSAR model. Fig. 1 shows the corresponding
scatter plots of experimental versus estimated bioactivities for the
training and inhibitors.
depends on molecular conformation and orientation. AtypeH47 is
atom-type-based AlogP descriptor. SaaN is the electro-topological
sum descriptor (E-state descriptor). E-state descriptors encode
information about both the topological environment of the partic-
ular atom and the electronic interactions due to all other atoms in
the molecule. SaaN is the sum of E-State values of all nitrogen atoms
with two aromatic bonds found in the molecule [40,44]. Hypo4/8
and Hypo12/4 represent the fit values of the training compounds
against the 8th and 4th pharmacophores generated in the 4th and
12th modeling runs, respectively (Tables B and C under Supple-
mentary material) [40].
The excellent qualities of Eq. (1) are evident not only from its
excellent statistical criteria, i.e., r2, F, r2BS, r2LOO, r2PRESS, but also
from the orthogonality of its descriptor variables (see Table 2).
Several descriptors emerged in Eq. (1) in spline format, e.g.,
Hypo4/8. The spline terms employed herein are ‘‘truncated power
splines’’ and are denoted by bolded brackets ([ ]). For example,
[f(x) ꢀ a] equals zero if the value of (f(x) ꢀ a) is negative; otherwise,
it equals (f(x) ꢀ a) [40].
logð1=IC50Þ ¼ ꢀ1:642 þ 0:186 Hypo 12=4 þ 0:351½Hypo 4=8
ꢀ4:98ꢁ ꢀ 0:032½ShadowXZ ꢀ 70:306ꢁ
ꢀ0:184½AtypeH47 ꢀ 18:00ꢁ ꢀ 0:147 SaaN
r926 ¼ 0:800; F ¼ 72:1; n ¼ 96; rB2S ¼ 0:800;
rL2OO ¼ 0:775; rP2RESS ¼ 0:707
(1)
where, r2 is the correlation coefficient, F is Fisher statistical
96
parameter, n is the number of observations, r2 is the boot-
BS
strapping regression coefficient, r2
is the leave-one-out corre-
LOO
lation coefficient and r2
is the predictive r2 determined for 22
PRESS
randomly selected external test compounds [39–43]. Shadow-XZ is
the area of the molecular shadow in the XZ plane. It is a geometric
descriptor that characterizes the shape of the molecule and
4
3
A
2
1
0
-4
-3
-2
-1
0
1
2
3
4
-1
-2
-3
-4
Experimental Log(1/IC50)
4
B
3
2
1
0
-4
-3
-2
-1
0
1
2
3
4
-1
-2
-3
-4
Experimental Log(1/IC50)
Fig. 1. Experimental versus fitted (A, 96 compounds, rL2OO ¼ 0:775) and predicted (B, 22 compounds, r2
¼ 0.707) bioactivities calculated from the best QSAR model (Eq. (1)). The
PRESS
solid lines are the regression lines for the fitted and predicted bioactivities of training and test compounds, respectively, whereas the dotted lines indicate the 1.0 log point error
margins.

R. Abu Khalaf et al. / European Journal of Medicinal Chemistry 45 (2010) 1598–1617
1601
Table 2
Interestingly, the combination Hypo4/8 and Hypo12/4 fre-
quented in the highest-ranking QSAR equations suggesting they
represent two complementary binding modes accessible to ligands
within the binding pocket of CETP, i.e., one of them can optimally
explain the bioactivities of some training inhibitors, while other
inhibitors are more appropriately explained by the second phar-
macophore [28]. Fig. 2 shows the two pharmacophores and how they
Cross-correlation coefficients (r2) of different descriptors in QSAR Eq. (1).a
Hypo12/4
AtypeH47
Shadow-XZ
SaaN
0.00
AtypeH47
Shadow-XZ
SaaN
0.09
0.20
0.01
0.57
0.19
0.03
0.04
0.04
0.28
Hypo4/8
a
Correlation coefficients were determined against compounds 1–118 (Table A in
map to the most potent inhibitor 100 (IC50 ¼ 0.0012
mM), while Table
Supplementary material).
D, under Supplementary material, shows the X, Y, and Z coordinates,
weights and tolerances of their pharmacophoric features.
Fig. 2. The binding pharmacophore hypotheses emerging in the optimal QSAR model (Hydrogen bond acceptor as green vectored spheres, hydrophobic features as blue spheres,
ring aromatic as orange vectored spheres, Hydrogen bond donor as violet vectored spheres): (A) the most potent inhibitor 100 (Table A under Supplementary material,
IC50 ¼ 0.0012
mM), (B) Hypo4/8, (C) Hypo4/8 mapped against 100, (D) shape-complemented Hypo4/8 mapped against 100, (E) Hypo4/8 mapped against most potent 30 training
compounds, (F) Hypo12/4, (G) Hypo12/4 mapped against 100, and (H) Hypo12/4 mapped against most potent 30 training compounds. (For interpretation of the references to colour
in this figure legend, the reader is referred to the web version of this article.)

1602
R. Abu Khalaf et al. / European Journal of Medicinal Chemistry 45 (2010) 1598–1617
Intriguingly, Hypo4/8 emerged in Eq. (1) in spline format indi-
limitations [18]. We employed the most potent training compound
cating that this binding mode contributes to ligand/CETP affinity
only if it fits a particular ligand above the corresponding spline
threshold. Accordingly, the ability of certain ligand to fit Hypo4/8
will impact its actual affinity to CETP only if its fit value exceeds
4.98 (the spline intercept associated with this pharmacophore in
Eq. (1)). Since this spline cutoff resembles moderate ligand/phar-
macophore mapping, as the maximum fit value is 12.0, it appears
that ligand binding to CETP is sensitive to moderate misalignments
among the attracting moieties within the complex, such that
lowering the fit value below 4.98 nullifies any affinity gains from
mapping this pharmacophore.
Emergence Shadow descriptors (i.e., shadow-xz) in Eq. (1)
illustrates certain role played by the ligands’ topology in the
binding process. However, despite their predictive significance, the
information content of such descriptors is quite obscure [44].
Nevertheless, emergence of SaaN in combination with a negative
regression slope suggests that the presence of nitrogen-containing
heterocycles reduces anti-CETP activity, as seen in compounds 57,
66, 70, 71, 79, 81, and 87 (Table A under Supplementary material).
100 (IC50 ¼ 0.0012
mM) as shape template (see Section 4.4 for more
details).
Fig. 3 and Table 3 show the ROC result of the shape-decorated
version of Hypo4/8. Clearly, the performance of shape-com-
plemented Hypo4/8 improved significantly as reflected by its ROC-
AUC, which shifted from 71.95% to 94.20%.
2.4. In silico screening of databases
Despite its less-than-optimal ROC profile, we were prompted by
group of factors to employ Hypo4/8 as the primary 3D search query
against available 3D structural databases. These factors can be
summarized as follows: (i) Hypo4/8 frequented in high-ranking
QSAR models significantly more than Hypo12/4, (ii) Hypo4/8 was
associated with higher regression coefficient in the optimal QSAR
Eq. (1) compared to Hypo12/4 and (iii) the term [Hypo4/8 ꢀ 4.98],
in Eq. (1), is statisitcally more significantly connected to bioactivity
(i.e., Log(1/IC50)) compared to Hypo12/4 with coresponding
significance F-values of 134.0 and 42.0, respectively.
This conduct might be related to certain aromatic
p
-stacking
However, to improve the success rate of our virtual screening,
we decided to refine Hypo4/8 search hits by either: (A) subsequent
screening of the hit list by Hypo12/4 (select hits captured by both
Hypo4/8 and Hypo12/4), or (B) shape constraints implemented on
Hypo4/8 (Fig. 2D).
In silico screening was conducted against the national cancer
institute list of compounds (NCI, includes 238,819 compounds) [18]
as well as our in house built list of drugs and agrochemicals (DAC,
3002 compounds). NCI hits were subsequently filtered by Lipinski’s
[46] and Veber’s criteria [47]. However, DAC hits were left without
subsequent filtration. Table 4 shows the number captured hits by
both methods, i.e, sequential pharmacophore refinement or shape
constraints.
Surviving hits were fitted against Hypo4/8 and Hypo12/4 (fit
values determined by Eq. (5) in Section 4) and their fit values were
substituted in QSAR Eq. (1) to determine their predicted bioactiv-
ities. However, in order to minimize the impact of any possible
extrapolatory QSAR prediction errors on decisions regarding which
hits merit subsequent in vitro testing [10,41], we employed Log
(1/IC50) predictions merely to rank the corresponding hits and
prioritize subsequent in vitro testing [12–17]. Table 5 and Fig. 4
show the highest predicted hits, their QSAR-based predictions, as
well as their experimental in vitro bioactivities.
interactions involving the ligands and certain electron-deficient
center(s) in the binding pocket such that the switching electron-
rich aromatic rings with electron-deficient nitrogen heterocycles
within the ligand structures disrupts this interaction causing the
apparent reduction in activity.
On the other hand, emergence of AtypeH47, which encodes for
the hydrophobic contribution of hydrogen atoms [40], in associa-
tion with a negative slope, suggests that the binding pocket favors
hydrophilic ligands as it seems to occur at the hydrophilic outer
surface of CETP.
2.3. Receiver-operating characteristic (ROC) curve analysis and
shape constraints
To further validate the resulting models (both QSAR and phar-
macophores), we subjected our QSAR-selected pharmacophores to
receiver-operating curve (ROC) analysis (see Section 4.3 for more
details).
Table 3 and Fig. 3 show the ROC results of our QSAR-selected
pharmacophores. Hypo4/8 illustrated mediocre overall performance
with an AUC of 71.95%. On the other hand, Hypo12/4 exhibited
excellent performance with AUC value of 99.91%. This is not unex-
pected, as the presence of HBA and HBD features in Hypo12/4 points
to its hydrophilic nature and therefore its better selectivity. While on
the other hand, the hydrophobic nature of Hypo4/8 might explain its
observed inferior selectivity. Well-positioned hydrophilic groups
should promote selective ligand-receptor interactions and accord-
ingly promote pharmacophoric selectivity and ROC-AUC [45].
In order to enhance the ROC profile of Hypo4/8, we decided to
decorate it with shape constraints derived from the most potent
Out of the 52 highest-ranking hits acquired for experimental
validation, 25 were found to possess inhibitory activities against
CETP ranging from 5.7 to 82.5% at 10
(NSC 40331), 130 (Glypuride), 141 (NSC 186323) and 162 (NSC
89508) exhibited >30% CETP inhibition at 10 M prompting us to
evaluate their IC50 values. Fig. 5 shows how Hypo4/8 and Hypo12/4
fits 120 (IC50 ¼ 6.5 M), while Fig. 6 shows how shape-com-
plemented Hypo4/8 maps 162 (IC50 ¼ 1.9 M).
mM. Four hits, namely, 120
m
m
m
Interestingly, although the sequential search method achieved
higher percentage of active hits (23 actives from 43 captured)
compared to shape-constrained Hypo4/8 (2 actives from 9 hits), the
later method captured the most potent hit (162) with IC50 value of
training inhibitor 100 (IC50 ¼ 0.0012
mM). Shape constraints encode
for the degree of 3D spatial similarity between screened
compounds and the template ligand used to build the shape
1.9 mM. The capture of a majority of inactive hits by shape-com-
Table 3
plemented Hypo4/8 is probably due to factors related to favorable
hydration (163 and 164, Fig. 4) or excessive structural rigidity (165–
170, Fig. 4).
Performance of QSAR-selected pharmacophores and the shape-complemented
version of Hypo4/8 as 3D search queries.
Pharmacophore model
ROC–AUC ACC
SPC
TPR
FNR
Although Eq. (1) failed to identify inactive search hits, it pre-
dicted the bioactivities of active hits excellently, as evident in Table
5. In fact, it predicted the IC50 values of 120, 130, 141 and 162 to be
Hypo4/8
Hypo12/4
0.7195
0.9991
0.9522 0.9532 0.9333 0.0468
0.9522 0.9498 1.0000 0.0502
0.9522 0.9565 0.8667 0.0435
Shape-complemented Hypo4/8 0.9420
7.9, 794.3, 33.0 and 0.9
the experimental values 6.5, 238.6, 60.3 and 1.9
This conduct is probably due to the lack of structural diversity in
m
M, respectively, which agrees nicely with
ROC, receiver-operating characteristic; AUC, area under the curve; ACC, overall
accuracy; SPC, overall specificity; TPR, overall true positive rate; FNR, overall false
negative rate.
mM, respectively.

R. Abu Khalaf et al. / European Journal of Medicinal Chemistry 45 (2010) 1598–1617
RECEIVER OPERATING CHARACTERISTIC (ROC)
1603
RECEIVER OPERATING CHARACTERISTIC (ROC)
B
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
A
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
FALSE POSITIVE RATE
FALSE POSITIVE RATE
RECEIVER OPERATING CHARACTERISTIC (ROC)
C
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
FALSE POSITIVE RATE
Fig. 3. Receiver operating characteristic curves (ROCs) of (A) Hypo4/8, (B) shape-complemented Hypo4/8, (C) and Hypo12/4.
our training list of compounds thus restricting the extrapolatory
potential of resulting QSAR models to limited structural categories.
class of inhibitors that enclose at least three aromatic rings.
Moreover, we hypothesized that introducing two well-positioned
central HBA features within the scaffold should maximize ligand-
CETP binding by allowing the ligands to tightly map Hypo4/8 in
at least two poses. In this context, we prepared 11 new benzy-
lidene-amino methanones containing two opposing HBAs: imine
and ketone.
Schemes 3 and 4 show the new structures (187–190 and 193–
199) while Fig. 7 shows how Hypo4/8 maps compounds 189 and
190. The HBA of Hypo4/8 fits the imine of 189 and ketone of 190 as
anticipated.
2.5. Synthesis and scaffold exploration
Convergence of in silico modeling studies and subsequent in
vitro validation on CETP inhibitors based on three-aromatic ring
scaffold interrupted with hydrogen-bond acceptor feature, e.g.,
120 and 162 (see Figs. 5 and 6), prompted us to envisage a new
The synthesis commenced by preparing different substituted
aminobenzophenone intermediates (175–177 and 180–182), in
modest to reasonable yields, via Friedel–Crafts acylation of benzene
and substituted derivatives (172–174 and 179) with 3- and 4-ami-
nobenzoic acids (171 and 178) in the presence of polyphosphoric
acid (PPA) [48] as in Schemes 1 and 2.
The best yield was obtained upon reacting 4-aminobenzoic acid
with 1,3-dimethoxybenzene to yield 182 (89%). Unsurprisingly,
benzenes substituted with electron-donating groups afforded good
yields, while electron-deficient benzene rings and biphenyl
required higher reaction temperatures and gave products with low
yields.
Table 4
The number of captured hit compounds by search methods.
3D Databasea Post-screening Search method
filteringb
Sequential Hypo4/8
followed by Hypo12/4 Hypo4/8
Shape-complemented
NCI
Before
After
None
5205
516
121
1232
527
55
DACc
a
NCI: national cancer institute list of available compounds (238,819 structures),
DAC: the list of established drugs and agrochemicals (3002 structures).
b
Post-screening filtering employing Lipinski’s and Veber’s rules.
This list of compounds was in silico scanned without post-screening filtering.
c

1604
R. Abu Khalaf et al. / European Journal of Medicinal Chemistry 45 (2010) 1598–1617
Table 5
The captured hit molecules with their fit values, their corresponding QSAR estimates from Eq. (1) and their in vitro bioactivities.
Tested hitsa
Search methodb
Fit values againstc
Hypo4/8
In vitro Anti-CETP activity
Hypo12/4
QSAR-based estimates
% Inhibition at 10.0
m
M
IC50 (mM)
Log(1/IC50
)
IC50 (mM)
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
B
B
B
B
B
B
B
B
B
2.6
7.8
6.7
2.8
8.3
7.7
6.2
5.2
2.3
5.6
7.5
3.2
7.1
3.2
7.0
2.3
9.0
4.1
1.0
8.0
9.1
8.9
6.3
8.6
8.5
9.3
4.1
8.0
9.1
1.9
9.7
1.5
2.7
4.9
3.6
7.8
8.3
7.8
7.0
0.0
5.5
5.1
7.6
9.2
9.2
9.3
9.5
9.7
9.7
9.1
9.1
9.0
7.3
2.2
2.3
3.9
1.2
6.4
4.4
5.0
1.7
3.3
6.0
0.0
7.4
9.1
2.4
5.9
3.4
4.0
3.3
4.9
5.4
5.9
3.6
3.6
6.2
5.7
6.5
7.9
0.4
6.5
2.9
6.2
7.8
7.1
7.4
7.5
7.7
4.1
5.4
4.1
7.8
4.2
8.1
1.2
0.0
0.0
0.0
9.0
0.0
3.0
0.0
4.0
ꢀ1.9
ꢀ0.9
ꢀ0.7
ꢀ1.8
ꢀ2.7
ꢀ0.5
ꢀ1.8
ꢀ1.8
ꢀ2.1
ꢀ1.2
ꢀ1.3
ꢀ2.9
ꢀ0.8
ꢀ0.2
ꢀ0.9
ꢀ0.9
0.2
ꢀ1.0
ꢀ1.0
ꢀ0.9
ꢀ0.3
0.4
ꢀ1.0
ꢀ1.0
0.8
79.4
7.9
5.0
12.3
52.5
22.1
18.6
20.8
0.0
10.6
0.0
0.0
0.0
5.7
35.0
15.9
11.4
0.0
–
6.5 (0.99)d
–
–
–
–
–
–
–
–
63.1
501.2
3.2
63.1
63.1
125.9
15.8
20.0
794.3
6.3
1.6
7.9
7.9
0.6
10.0
10.0
7.9
2.0
0.4
10.0
10.0
0.2
0.4
2.5
1.0
7.9
3.2
0.4
3.2
6.3
2.5
2.5
0.3
0.2
4.0
1.3
–
238.6 (1.00)d
–
–
–
–
–
–
–
–
0.0
0.0
22.6
25.5
13.4
20.3
0.0
33.0
0.0
0.0
0.0
0.0
0.0
9.0
14.0
0.0
20.8
25.4
0.0
8.1
0.0
9.1
0.0
6.7
0.0
27.4
0.0
0.0
82.5
6.0
0.0
0.0
0.0
–
–
60.3 (0.99)d
–
–
–
–
–
–
–
–
–
–
–
–
–
–
–
–
–
–
–
–
1.9
–
–
–
–
–
–
–
–
0.4
ꢀ0.4
0.0
ꢀ0.9
ꢀ0.5
0.4
ꢀ0.5
ꢀ0.8
ꢀ0.4
-0.4
0.5
0.8
ꢀ0.6
ꢀ0.1
ꢀ1.0
ꢀ1.0
ꢀ2.0
0.0
10.0
10.0
100.0
1.0
0.9
15.9
2.6
2.0
0.1
1.4
3.2
0.1
ꢀ1.2
ꢀ0.4
ꢀ0.3
0.8
ꢀ0.2
ꢀ0.5
ꢀ0.6
ꢀ0.4
0.0
0.0
0.0
0.0
3.8
2.2
a
Compound numbers as in Fig. 4.
A: Sequential search by Hypo4/8 followed by Hypo12/4, B: search by shape-complemented Hypo4/8.
Best-fit values against each binding hypothesis calculated by Eq. (5).
b
c
d
This value represents the correlation coefficient of the corresponding dose-response line at three concentrations.
Subsequently, the aminobenzophenone intermediates were
The final products were tested against CETP at 10 mM concen-
used to prepare the final imines. Imines are formed typically by
reversible acid-catalyzed condensation of amines and aldehydes
with extrusion of water through either azeotropic distillation or by
employing chemical drying agents [49].
In the current work, the imine products were prepared from
reaction of benzaldehyde, trifluoro-m-tolualdehyde, trifluoro-p-
tolualdehyde, 3-methoxylbenzaldehyde, 4-methoxylbenzaldehyde
and 4-tert-butylbenzaldehyde with aminobenzophenone inter-
mediates (175–177 and 180–182) as illustrated in Schemes 3 and 4
[35]. The best yield was obtained when 176, dissolved in cyclo-
hexane, reacted with benzaldehyde to yield 187 (70%).
trations and exhibited anti-CETP activities ranging from 9.8 to
33.8% as in Table 6. The less-than-optimal bioactivities of the
prepared compounds might be related to their rigid scaffold. We
are currently in the process of preparing more flexible analogs of
better bioactivity profiles.
3. Conclusions
This work includes elaborate pharmacophore exploration of
CETP inhibitors utilizing CATALYST-HYPOGEN. QSAR analysis was
employed to select the best combination of molecular descriptors

R. Abu Khalaf et al. / European Journal of Medicinal Chemistry 45 (2010) 1598–1617
1605
O
HN
O
HN
Br
N
O
HO
H
.HCl
O
H
N
NH
O
N
N
119 (Bromocriptine)
120 (NSC 40331)
CF
H
3
Cl
O
N
H
N
N
O
O
Cl
N
HO
121 (NSC 216941)
122 (Amodiaquine)
F
O
O
H
N
N
OH
Br
N
N
OH
O
123 (Astemizole)
124 (Bromodiolone)
Cl
OH
H N
2
H
N
O
O
N
OCH
O
N
3
OH
F
H CO
3
OH
125 (Cisapride)
126 (Fexofenadine)
N
O
F
F
O
O
F C
3
Cl
N
HN
O
NH
Cl
O
N
O
H
F C
3
Cl
127 (Fluazuron)
128 (Fluvalinate)
Fig. 4. The chemical structures of the tested highest-ranking hits (as suggested by the best QSAR model and the associated pharmacophores).
and pharmacophore models capable of explaining bioactivity
variation across an informative list of training compounds. The
successful pharmacophores were complemented with strict shape
constraints to optimize their receiver-operating characteristic
(ROC) curve profiles. The best binding hypotheses were used as 3D
search queries to screen the NCI, drugs and agrochemicals data-
bases for new CETP inhibitors. From the highest-ranking hits, four
were found to possess promising inhibitory IC50 values against
CETP. Modeling results were then used to guide synthetic explo-
ration of a new series of aromatic imines as CETP inhibitors.
4. Experimental section
4.1. Molecular modeling
4.1.1. Software and hardware
The following software packages were utilized in the present
research.
ꢂ CATALYST (Version 4.11), Accelrys Inc. (www.accelrys.com),
USA.

1606
R. Abu Khalaf et al. / European Journal of Medicinal Chemistry 45 (2010) 1598–1617
O
S
HO
H
N
H
N
O
F
F
S
O
O
O
O
F C
3
H
N
H
H
H
Cl
OH
129 (Fluvestrant)
130 (Glypuride)
OH
HO
I
I
OH
O
I
I
HN
HO
H
N
O
I
HO
O
O
O
I
O
131 (Iodipamide)
132 (Lanatoprost)
Cl
NH
OH
O
N
H
Cl
Cl
O
S
NH
Cl
S
N
H
Cl
Cl
134 (NSC 58231)
133 (NSC 36385)
OH
O
NH
NH
HN
N
O
HN
S
O
Cl
135 (NSC 78365)
136 (NSC 120184)
N
O
.2HCl
NH
HO
HO
N
O
O
HO
Cl
N
137 (NSC 125527)
138 (NSC 130108)
Fig. 4. (continued).
ꢂ CERIUS2 (Version 4.10), Accelrys Inc. (www.accelrys.com), USA.
ꢂ CS ChemDraw Ultra 7.01, Cambridge Soft Corp. (http://www.
cambridgesoft.com), USA.
4.1.2. Dataset
The structures of 118 CETP inhibitors (Table A under Supple-
mentary material) were collected from published literature [33–
36]. The in vitro bioactivities of the collected inhibitors were
expressed as the concentration of the test compound that inhibited
the activity of CETP by 50% (IC50). The logarithm of the measured
1/IC50 values were used in pharmacophore modeling and QSAR
analysis, thus correlating the data linear to the free energy change.
Pharmacophore modeling and QSAR analysis were performed
using CATALYST (HYPOGEN module) and CERIUS2 software suites
from Accelrys Inc. (San Diego, California, www.accelrys.com)
installed on a Silicon Graphics Octane2 desktop workstation
equipped with a 600 MHz MIPS R14000 processor (1.0 GB RAM)
running the Irix 6.5 operating system.
In cases where IC50 is expressed as being higher than 100
93, 94, 95, 96, 97, 98 and 99), we assumed it equals 101 M. In cases
mM (e.g.,
m

R. Abu Khalaf et al. / European Journal of Medicinal Chemistry 45 (2010) 1598–1617
1607
N
Cl
O
NH
N
O
N
NH
O
Cl
O
O
139 (NSC 130795)
140 (NSC 160919)
Cl
NH
O
O
O
H
N
Cl
NH
O
N
H
Cl
Cl
Cl
141 (NSC 186323)
142 (NSC 204231)
Cl
NH
Cl
HN
O
O
N
HN
HN
Cl
N
O
Cl
O
O
143 (NSC 211295 )
144 (NSC 211303)
Cl
NH
Cl
NH
HN
NH
O
Cl
O
O
N
O
Cl
Cl
Cl
145 (NSC 211844)
146 (NSC 211848)
NH
O
OH
O
O
OH
OH
O
OH
NH
O
147 (NSC 215255)
148 (NSC 300513)
Fig. 4. (continued).
where IC50 is expressed as being higher than 50
m
M (e.g., 91, 116,
4.1.3. Conformational analysis
117 and 118), we assumed it equals 51 M, these assumptions are
m
The conformational space of each inhibitor (1–118, Table A
under Supplementary material) was explored adopting the ‘‘best
conformer generation’’ option within CATALYST which is based on
the generalized CHARMm force field implemented in the program.
necessary to allow statistical correlation and QSAR analysis. The
logarithmic transformation of IC50 values should minimize any
potential errors resulting from this assumption.

1608
R. Abu Khalaf et al. / European Journal of Medicinal Chemistry 45 (2010) 1598–1617
Cl
Cl
NH
NH
NH
O
O
O
Cl
Cl
Cl
O
O
Cl
149 (NSC 310137)
150 (NSC 319991)
N
O
O
HN
N
OH
H
O
N
O
Cl
151 (NSC 332413)
152 (NSC 366099)
O
O
OH
H
N
O
N
O
Cl
Cl
N
H
HN
HO
153 (NSC 366100)
154 (NSC 372280)
O
N
N
H
Cl
N
O
Cl
N
H
HN
HO
155 (NSC 372302)
156 (NSC 665363)
S
N
O
OH
O
O
HO
O
Cl
O
NH
Cl
157 (NSC 694168)
158 (Propafenone )
Fig. 4. (continued).
Conformational ensembles were generated with an energy
threshold of 20 kcal/mol from the local minimized structure and
a maximum limit of 250 conformers per molecule [18–20,21–
26,37].
4.1.4. Generation of pharmacophoric hypotheses
All 118 molecules with their associated conformational models
were regrouped into a spreadsheet. The biological data of the
inhibitors were reported with an ‘‘Uncertainty’’ value of 3, which

R. Abu Khalaf et al. / European Journal of Medicinal Chemistry 45 (2010) 1598–1617
1609
HO
HO
OH
OH
NH
N
O
OH
159 (Salmeterol)
160 (Terfenadine)
OH
HO
O
S
NH
F C
3
HO
O
O
Cl
S
O
N
O
161 (Travoprost)
162 (NSC 89508)
Cl
O
+
N
N
+
Cl
N
. Br-
2
N
O
N
163 (NSC152458)
164 (NSC 67226)
OH
O
N
N
O
S
H
N
N
H
Cl
N
H
O
S
S
O
Cl
165 (NSC 110302)
166 (NSC 327360)
O
S
N
O
O
HN
Br
N
N
N
O
N
Br
NH
O
167 (NSC 647954)
168 (NSC 35927)
O
O
H
N
O
N
H
N
S
S
N
H
O
H
O
O
169 (NSC 111347)
170 (NSC 144469)
Fig. 4. (continued).

1610
R. Abu Khalaf et al. / European Journal of Medicinal Chemistry 45 (2010) 1598–1617
Fig. 5. The mapping of Hypo4/8 and Hypo12/4 against hit 120 (Hydrogen bond acceptor as green vectored spheres, hydrophobic features as blue spheres, ring aromatic as orange
vectored spheres, Hydrogen bond donor as violet vectored spheres): (A) inhibitor 120, (B) Hypo4/8 mapped against 120 and (C) Hypo12/4 mapped against 120. (For interpretation of
the references to colour in this figure legend, the reader is referred to the web version of this article.)
means that the actual bioactivity of a particular inhibitor is assumed
to be situated somewhere in an interval ranging from one-third to
three-times the reported bioactivity value of that inhibitor [22–26].
Three training subsets (Table 1) were selected for pharmacophore
modeling. Each subset was utilized to conduct four modeling runs to
explore the pharmacophoric space of CETP inhibitors. Different
hypotheses were generated by altering the inter-feature spacing
and the number of allowed features in the resulting pharmaco-
phores (Table B under Supplementary material).
Pharmacophore modeling employing CATALYST proceeds
through three successive phases: the constructive phase, subtractive
phase and optimization phase [22–26]. During the constructive
phase, CATALYST generates common conformational alignments
among the most active training compounds. Only molecular align-
ments based on a maximum of five chemical features are considered.
The program identifies a particular compound as being within the
most active category if it satisfies Eq. (2) [22–26].
the activity of the training compounds under question. In the
subsequent subtractive phase, CATALYSTeliminates some hypotheses
that fit inactive training compounds. A particular training compound
is defined as being inactive if it satisfies Eq. (3) [22–26]:
logðActÞ ꢀ logðMActÞ > BS
(3)
where, ‘‘BS’’ is the bioactivity spread (equals 3.5 by default).
In the optimization phase, CATALYST applies fine perturbations
in the form of vectored feature rotation, adding new feature and/or
removing a feature, to selected hypotheses that survived the
subtractive phase to find new models of enhanced bioactivity-to-
mapping correlations. Eventually, CATALYST selects the highest-
ranking models (10 by default) and presents them as the optimal
pharmacophore hypotheses resulting from the particular automatic
modeling run [19].
4.1.5. Assessment of the generated hypotheses
When generating hypotheses, CATALYST attempts to minimize
a cost function consisting of three terms: Weight cost, Error cost
and Configuration cost [19,22–26]. Weight cost is a value that
increases as the feature weight in a model deviates from an ideal
ðMAct ꢃ UncMActÞ ꢀ ðAct=UncActÞ > 0:0
where ‘‘MAct’’ is the activity of the most active compound in the
training set, ‘‘Unc’’ is the uncertainty of the compounds and ‘‘Act’’ is
(2)

R. Abu Khalaf et al. / European Journal of Medicinal Chemistry 45 (2010) 1598–1617
1611
Z
COOH
Z
Y
1) PPA, CH2Cl2, 80-190°C
2) NH3, 0°C
+
O
Y
NH
H N
2
2
172 Y = H, Z = H
174 Y = H, Z = Phenyl
178
180 Y = H, Z = H
181 Y = H, Z = Phenyl
182 Y = OCH , Z = OCH
179 Y = OCH , Z = OCH
3
3
3
3
Scheme 2. Synthesis of aminobenzophenone intermediates 180–182.
equals to 1.6 Å by default). S
(disp/tol)2 is the summation of (disp/tol)2
values for all pharmacophoric features that successfully superimpose
corresponding chemical functionalities in the fitted compound
[22–24].
The third term, i.e., the configuration cost, penalizes the
complexity of the hypothesis, i.e. the configuration cost. This is
a fixed cost, which is equal to the entropy of the hypothesis space.
The more the numbers of features (a maximum of five) in
Fig. 6. The mapping of shape-complemented Hypo4/8 pharmacophore hypotheses
against hit 162 (Hydrogen bond acceptor as green vectored spheres, hydrophobic
features as blue spheres, ring aromatic as orange vectored spheres, Hydrogen bond
donor as violet vectored spheres): (A) inhibitor 162 and (B) Hypo4/8 fitted against 162.
(For interpretation of the references to colour in this figure legend, the reader is
referred to the web version of this article.)
value of 2. The deviation between the estimated activities of the
training set and their experimentally determined values adds to the
error cost [19,22–26]. The activity of any compound can be esti-
mated from a particular hypothesis through Eq. (4) [19].
logðEstimated activityÞ ¼ I þ Fit
(4)
where, I ¼ the intercept of the regression line obtained by plotting
the log of the biological activity of the training set compounds
against the Fit values of the training compounds. The Fit value for
any compound is obtained automatically employing Eq. (5) [22–24].
h
i
X
X
Fit ¼
mapped hypothesis featuresꢃW 1ꢀ
ðdisp=tolÞ2
(5)
where,
S mapped hypothesis features represents the number of
pharmacophore features that successfully superimpose (i.e., map or
overlap with) corresponding chemical moieties within the fitted
compound, W is the weight of the corresponding hypothesis feature
spheres. This value is fixed to 1.0 in CATALYST-generated models. disp
is the distance between the center of a particular pharmacophoric
sphere (feature centroid) and the center of the corresponding
superimposed chemical moiety of the fitted compound; tol is the
radius of the pharmacophoric feature sphere (known as Tolerance,
X
COOH
X
H N
2
O
+
1) PPA, CH2Cl2, 80-190°C
2) NH3, 0°C
NH
2
171
172 X = H
175 X = H
173 X = OCH
176 X = OCH
3
3
174 X = Phenyl
177 X = Phenyl
Fig. 7. The mapping of Hypo 4/8 against synthesized compounds 189 (CETP inhibition at
10
mM ¼ 33.8%) and 190 (CETP inhibition at 10
m
M ¼ 26.3%): (A) structure of 189, (B)
Scheme 1. Synthesis of aminobenzophenone intermediates 175–177.
structure of 190, (C) Hypo4/8 mapped against 189 and (D) Hypo4/8 mapped against 190.

1612
R. Abu Khalaf et al. / European Journal of Medicinal Chemistry 45 (2010) 1598–1617
X
O
Y
H
X
Cyclohexane, HCl,
80-90°C, 3-4 h
O
R
+
O
O
H
R
NH
Z
2
Cyclohexane,
184 R = 4-OCH
185 R = 3-CF
186 R = 4-CF
191 R = 3-OCH
180-182
3
+
N
HCl, 80-90°C, 3-4 h
3
H N
2
R
O
3
3
192 R = 4-tert-Bu
175-177
183 R = H
187 X = OCH , R = H
O
Y
3
184 R = 4-OCH
188 X = OCH , R = 4-OCH
3
3
3
185 R = 3-CF
189 X = Phenyl, R = 3-CF
3
3
186 R = 4-CF
190 X = H, R = 4-CF
3
3
N
Z
R
Scheme 3. Synthesis of imine compounds 187–190.
193 Y = H, Z = Phenyl, R = 3-CF
3
a generated hypothesis, the higher is the entropy with subsequent
increase in this cost. The overall cost (total cost) of a hypothesis is
calculated by summing over the three cost factors. However, error
cost is the main contributor to total cost.
194 Y = OCH , Z = OCH , R = 4-tert-Bu
3
3
195 Y = OCH , Z = OCH , R = 4-CF
3
3
3
196 Y = OCH , Z = OCH , R = 3-CF
3
3
3
197 Y = OCH , Z = OCH , R = 4-OCH
CATALYST also calculates the cost of the null hypothesis, which
presumes that there is no relationship in the data and that
experimental activities are normally distributed about their mean.
Accordingly, the greater the difference from the null hypothesis
cost (residual cost, Table C under Supplementary material), the
more likely that the hypothesis does not reflect a chance corre-
lation [19,20,22–26]. In a successful automatic modeling run,
CATALYST ranks the generated models according to their total
costs [22–24].
An additional approach to assess the quality of CATALYST–
HYPOGEN pharmacophores is to cross-validate them using the
Cat-Scramble algorithm implemented in CATALYST [26,38]. This
validation procedure is based on Fisher’s randomization test [38]. In
this validation test, we selected a 95% confidence level, which
instruct CATALYST to generate 19 random spreadsheets by the
Cat-Scramble command. Subsequently, CATALYST-HYPOGEN is
challenged to use these random spreadsheets to generate hypo-
theses using exactly the same features and parameters used in
generating the initial unscrambled hypotheses [43]. Success in
generating pharmacophores of comparable cost criteria to those
produced by the original unscrambled data reduces the confidence
in the training compounds and the unscrambled original pharma-
cophore models [26,38]. Based on Fisher randomization criteria, all
120 pharmacophores achieved 95% significance threshold, and
were therefore, submitted for subsequent processing (clustering
and QSAR analysis).
3
3
3
198 Y = OCH , Z = OCH , R = 3-OCH
3
3
3
199 Y = H, Z = H, R = 4-OCH
3
Scheme 4. Synthesis of imine compounds 193–199.
QSAR models. The test molecules were selected as follows: the 118
inhibitors were ranked according to their IC50 values, then every
fifth compound was selected for the test set starting from the high-
potency end (Table A under Supplementary material).
The chemical structures of the inhibitors were imported into
CERIUS2 as standard 3D single conformers (of the lowest energy
within the conformational ensemble generated by CATALYST)
representations in SD format. Subsequently, 100 molecular
descriptors were calculated for each compound employing the
C2.DESCRIPTOR module of CERIUS2. The calculated descriptors
included various simple and valence connectivity indices, electro-
topological state indices, single point quantum-mechanical
descriptors (via the AM1 model) and other molecular descriptors
(the detailed molecular descriptors are listed under Supplemen-
tary material) [40]. Furthermore, the training compounds were
fitted against the 26 representative CETP pharmacophore
hypotheses, generated by the CATALYST-HYPOGEN automatic runs
(shown in Tables A and B), and their fit values (produced by the
best-fit command within CATALYST via Eq. (5)) were added as
additional molecular descriptors.
Genetic function approximation (GFA) was employed to search
for the best possible QSAR regression equation capable of
4.1.6. Clustering of the generated pharmacophore hypotheses
The resulting models (120) were clustered into 26 groups
utilizing the hierarchical average linkage method available in
CATALYST. Subsequently, the highest-ranking representatives, as
judged based on their fit-to-bioactivity correlation values, were
selected to represent their corresponding clusters in subsequent
QSAR modeling. Table C, under Supplementary material, shows the
representative pharmacophores features, success criteria, the cor-
responding Cat.scramble confidence levels and differences from
corresponding null hypotheses.
Table 6
The synthesized compounds with their in vitro bioactivities.
Synthesized compounda
% Inhibition of CETP at 10.0 mM
187
188
189
190
193
194
195
196
197
198
199
12.6
14.8
33.8
26.3
9.8
24.9
15.4
14.9
19.5
23.7
19.9
4.2. QSAR modeling
A set of 96 compounds was employed as the training set for
QSAR modeling. The remaining 22 molecules (ca. 20% of the data-
set) were employed as an external test subset for validating the
a
Compound numbers as in Schemes 3 and 4.

R. Abu Khalaf et al. / European Journal of Medicinal Chemistry 45 (2010) 1598–1617
1613
correlating the variations in biological activities of the training
compounds with variations in the generated descriptors, i.e.,
multiple linear regression modeling (MLR) [40].
selected compounds (n) as a function of (1-Sp). Sp is defined as
specificity or true negative rate (Eq. (9)) [51,53].
Number of Selected Actives
Total Number of Actives
TP
TP þ FN
Our preliminary diagnostic trials suggested the following
optimal GFA parameters: Explore linear equations at mating and
mutation probabilities of 50%; population size ¼ 500; number of
genetic iterations ¼ 30000 and LOF smoothness parameter ¼ 1.0.
However, to determine the optimal number of explanatory terms
(QSAR descriptors), it was decided to scan and evaluate all possible
QSAR models resulting from 3 to 6 explanatory terms. All QSAR
models were validated employing leave-one-out cross-validation
Se ¼
Sp ¼
¼
(8)
(9)
Number of Discarded Inactives
Total Number of Inactives
TN
TN þ FP
¼
where, TP is the number of active compounds captured by the
virtual screening method (true positives), FN is the number of
active compounds discarded by the virtual screening method, TN is
the number of discarded decoys (presumably inactives), while FP is
the number of captured decoys (presumably inactives).
(r2LOO), bootstrapping (r2BS) [39,40] and predictive r2 (r2
)
PRESS
calculated from the test subsets. The predictive r2
is defined
PRESS
as:[39–42]
If all molecules scored by a virtual screening (VS) protocol
with sufficient discriminatory power are ranked according to
their score (i.e., fit values), starting with the best-scored molecule
and ending with the molecule that got the lowest score, most of
the actives will have a higher score than the decoys. Since some
of the actives will be scored lower than decoys, an overlap
between the distribution of active molecules and decoys will
occur, which will lead to the prediction of false positives and
false negatives [51,53]. An ROC curve is plotted by setting the
score of the active molecule as the first threshold. Afterwards, the
number of decoys within this cutoff is counted and the corre-
sponding Se and Sp pair is calculated. This calculation is repeated
for the active molecule with the second highest score and so
forth, until the scores of all actives are considered as selection
thresholds.
The ROC curve representing ideal distributions, where no
overlap between the scores of active molecules and decoys exists,
proceeds from the origin to the upper-left corner until all the
actives are retrieved and Se reaches the value of 1. In contrast to
that, the ROC curve for a set of actives and decoys with randomly
distributed scores tends towards the Se ¼ 1-Sp line asymptotically
with increasing number of actives and decoys [51]. The success of
a particular virtual screening workflow can be judged from the
following criteria (shown in Table 3):
rP2RESS ¼ SD ꢀ PRESS=SD
where SD is the sum of the squared deviations between the bio-
logical activities of the test set and the mean activity of the training
set molecules, PRESS is the squared deviations between predicted
and actual activity values for every molecule in the test set.
(6)
4.3. Receiver-operating characteristic (ROC) curve analysis
Selected pharmacophore models (i.e., Hypo4/8, Hypo12/4 and
shape-complemented Hypo4/8) were validated by assessing their
abilities to selectively capture diverse CETP active compounds from
a large testing list of actives and decoys.
The testing list was prepared as described by Verdonk et al.
[50,51]. Briefly, decoy compounds were selected based on three
basic one-dimensional (1D) properties that allow the assessment of
distance (D) between two molecules (e.g., i and j): (1) the number of
hydrogen-bond donors (NumHBD); (2) number of hydrogen-bond
acceptors (NumHBA) and (3) count of nonpolar atoms (NP, defined
as the summation of Cl, F, Br, I, S and C atoms in a particular
molecule). For each active compound in the test set, the distance to
the nearest other active compound is assessed by their Euclidean
Distance (Eq. (7)):
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
À
Á
À
Á
À
Á
2
Dði; jÞ ¼
NumHBDi ꢀ NumHBDj 2þ NumHBAi ꢀ NumHBAj 2þ NPi ꢀ NPj
(7)
(1) Area under the ROC curve (AUC) [53]. In an optimal ROC curve
an AUC value of 1 is obtained; however, random distributions
cause an AUC value of 0.5 [51,53].
(2) Overall accuracy (ACC): describes the percentage of correctly
classified molecules by the screening protocol (Eq. (10)). Testing
compounds are assigned a binary score value of zero (compound
not captured) or one (compound captured) [51,54,55].
The minimum distances are then averaged over all active
compounds (Dmin). Subsequently, for each active compound in the
test set, around 33 decoys were randomly chosen from the ZINC
database [52]. The decoys were selected in such a way that they did
not exceed Dmin distance from their corresponding active
compound.
To diversify active members in the list, we excluded any active
compound having zero distance (D(i,j)) from other active
compound(s) in the test set. Active testing compounds were defined
ꢀ
ꢁ
TP þ TN
A
N
A
N
ACC ¼
¼
ꢃ Se þ 1 ꢀ
ꢃ Sp
(10)
as those possessing CETP affinities ranging from 0.0012 to 6.72
mM.
N
The test set included 15 active compounds and 474 ZINC decoys.
The test set (489 compounds) was screened by each particular
pharmacophore employing the ‘‘Best flexible search’’ option
implemented in CATALYST, while the conformational spaces of the
compounds were generated employing the ‘‘Fast conformation
generation option’’ implemented in CATALYST. Compounds missing
one or more features were discarded from the hit list. In silico hits
were scored employing their fit values as calculated by Eq. (5).
The ROC curve analysis describes the sensitivity (Se or true
positive rate, Eq. (8)) for any possible change in the number of
where, N is the total number of compounds in the testing database,
A is the number of true actives in the testing database.
(3) Overall specificity (SPC): describes the percentage of discarded
inactives by the particular virtual screening workflow. Inactive
test compounds are assigned a binary score value of zero
(compound not captured) or one (compound captured)
[51,54,55].

1614
R. Abu Khalaf et al. / European Journal of Medicinal Chemistry 45 (2010) 1598–1617
(4) Overall true positive rate (TPR or overall sensitivity): describes
the fraction percentage of captured actives from the total
number of actives. Active test compounds are assigned a binary
score value of zero (compound not captured) or one
(compound captured).
Negative controls lacking rabbit serum were used as background.
All measurements were conducted in duplicates.
4.7. Synthetic procedures
(5) Overall false negative rate (FNR or overall percentage of dis-
carded actives): describes the fraction percentage of active
compounds discarded by the virtual screening method. Dis-
carded active test compounds are assigned a binary score value
of zero (compound not captured) or one (compound captured).
Melting points were measured using Gallenkampf melting point
apparatus and are uncorrected. 1H-NMR and 13C-NMR spectra were
collected on a Varian Oxford NMR300 spectrometer. The samples
were dissolved in CDCl3 at a concentration of 0.3–0.7 wt % and
placed in 5 mm NMR tubes. Mass spectrometry was performed
using LC Mass Bruker Apex-IV mass spectrometer utilizing an
electrospray interface.
4.4. Addition of shape constraints to pharmacophore models
Infrared spectra were recorded using Shimadzu IRAffinity-1
spectrophotometer. The samples were dissolved in CHCl3 and
analysed as thin solid films using NaCl plates. Analytical thin layer
chromatography (TLC) was carried out using pre-coated aluminum
plates and visualized by UV light (at 254 and/or 360 nm). Elemental
analysis was performed using EuroVector elemental analyzer.
Chemicals and solvents were purchased from corresponding
companies (Sigma-Aldrich, Riedel-de Haen, Fluka, BDH Laboratory
Supplies and Promega Corporation) and were used in the experi-
mentation without further purification.
Shape constraints were added using the CatShape module of
CATALYST [18,56]. To generate merged shape-pharmacophore
queries for Hypo4/8 and Hypo12/4, the most potent training
compound 100 (IC50 ¼ 0.0012
mM) was first fitted against the
pharmacophore models, thereafter, the best-fitted conformer of the
inhibitor was used to generate default shape constraints (70%–130%
tolerance similarity) that were subsequently merged with each
pharmacophore.
4.5. In silico screening of databases for new CETP inhibitors
4.7.1. General procedure for the synthesis of (3-amino-phenyl)-
(un)substituted phenyl-methanone (175–177)
Shape-complemented Hypo4/8 and combined Hypo4/8–
Hypo12/4 were employed as 3D search queries against the NCI and
DAC libraries using the ‘‘Best Flexible Database Search’’ option
implemented within CATALYST. NCI hits were filtered based on
Lipinski’s and Veber’s rules [46,47]. The remaining hits were fitted
against Hypo4/8 and Hypo12/4 using the ‘‘best-fit’’ approach
implemented within CATALYST. Subsequently, the fit values
together with the relevant molecular descriptors were substituted
in QSAR Eq. (1) to predict anti-CETP IC50 values. The highest-
ranking 52 hits were subsequently tested in vitro.
3-Aminobenzoic acid 171 (0.411 g, 3 mmol) was dissolved in
CH2Cl2 (5 mL), and polyphosphoric acid (10 g) was added. Then
(un)substituted benzene (6–30 mmol) was added. The mixture was
stirred carefully at 80- 180 ꢄC for 3 h and then poured on crushed
ice. The solution was carefully made alkaline with 25% ammonia
and then extracted with CH2Cl2 (3 ꢃ 20 mL). The combined extracts
were dried on anhydrous Na2SO4 and filtrated.
4.7.1.1. (3-Amino-phenyl)-phenyl-methanone (175). Evaporation of
the solvent gave 175 as a yellow powder (0.23 g, 39%) [48]: mp: 84–
86 ꢄC (Lit. mp: 109–109.5 ꢄC); 1H-NMR (CDCl3,
d ppm): 3.82 (bs, 2H,
4.6. CETP inhibition assay
NH2), 6.88 (d, J ¼ 7.2 Hz, 1H), 7.11 (m, 2H), 7.32 (d, J ¼ 7.2 Hz, 1H),
7.47 (d, J ¼ 6.8 Hz, 2H), 7.56 (d, J ¼ 6.8 Hz, 1H), 7.80 (d, J ¼ 6.8 Hz,
4.6.1. Quantification of CETP activity
2H); 13C-NMR (CDCl3,
d ppm): 116.13 (1C), 119.20 (1C), 120.86 (1C),
CETP inhibitory bioactivities were assayed by fluorescent-CE
transfer employing commercially available kit (BioVision, Linda Vista
Avenue, USA). The assay kit is based on donor molecule containing
fluorescent self-quenched neutral lipid that is transferred to an
acceptor molecule in the presence of CETP (from rabbit serum). CETP-
mediated transfer of the fluorescent neutral lipid to the acceptor
molecule results in increase in fluorescence. Inhibition of CETP will
prevent lipid transfer and therefore decrease fluorescence intensity.
The assay procedure can be described brieflyas follows. An aliquot
128.41 (2C), 129.30 (1C), 130.26 (2C), 132.52 (1C), 138.01 (1C),
138.90 (1C), 146.73 (1C), 197.22 (1C); IR (thin film, cmꢀ1): 3480,
3379, 3017, 1651, 1620, 1600, 1490, 1454.
4.7.1.2. (3-Amino-phenyl)-(4-methoxy-phenyl)-methanone (176). Eva-
poration of the solvent gave 176 as a yellow powder (0.35 g, 52%)
[48]: mp: 109–111 ꢄC (Lit. mp: 114–116 ꢄC); 1H-NMR (CDCl3,
d
ppm): 3.83 (bs, 2H, NH2), 3.86 (s, 3H, OCH3), 6.85 (dd, J ¼ 6.4,
1.8 Hz, 1H), 6.93 (dd, J ¼ 6.7, 2.1 Hz, 2H), 7.05 (dd, J ¼ 6.4, 1.8 Hz, 1H),
7.08 (t, J ¼ 1.8 Hz, 1H), 7.22 (t, J ¼ 6.4 Hz, 1H), 7.81 (dd, J ¼ 6.7, 2.1 Hz,
of 1.5
20 L of the master mix, provided in the assay kit (donor molecule,
acceptor molecule and assay buffer), was added, mixed well, and the
volume was completed to 203 L with the provided assay buffer.
After incubation at 37 ꢄC for 1 h, fluorescence intensity (Excitation
465 nm; Emission : 535 nm) was read in a FLX800TBI Microplate
mL of rabbit serum was added to 160 mL of testing sample. Then
m
2H); 13C-NMR (CDCl3,
d ppm): 55.72 (1C, OCH3), 113.69 (2C), 115.94
(1C), 118.69 (1C), 120.39 (1C), 129.20 (1C), 130.53 (1C), 132.77 (2C),
139.59 (1C), 146.75 (1C), 163.37 (1C), 196.05 (1C); IR (thin film,
cmꢀ1): 3464, 3364, 3017, 1651, 1613, 1597, 1503, 1475, 1451, 1250.
m
l:
l
Fluorimeter (BioTek Instruments, Winooski, USA).
4.7.1.3. (3-Amino-phenyl)-biphenyl-4-yl-methanone
(177). The
residue, after evaporation of the solvent, was purified by column
chromatography eluting with CHCl3/EtOAc (95:5) to give 177 pure
as a yellow powder (0.25 g, 30%): mp: 120–122 ꢄC; 1H-NMR (CDCl3,
4.6.2. Preparation of tested compounds
The tested compounds were initially dissolved in DMSO to yield
10 mM stock solutions and subsequently diluted to the required
concentrations using distilled deionized water. The final concen-
tration of DMSO was adjusted to 0.1%. The percentage of residual
activity of CETP was determined for each compound by comparing
the activity of CETP in the presence and absence of the tested
compound. Positive controls were tested to assess the degree of
CETP inhibition by 0.1% DMSO. CETP was not affected by DMSO.
d
ppm): 3.84 (bs, 2H, NH2), 6.90 (d, J ¼ 7.6 Hz, 1H), 7.16 (d, J ¼ 8.2 Hz,
2H), 7.26 (t, J ¼ 7.6 Hz, 1H), 7.42 (d, J ¼ 7.0 Hz, 1H), 7.48 (m, 2H), 7.67
(m, 4H), 7.90 (d, J ¼ 8.2 Hz, 2H); 13C-NMR (CDCl3,
d ppm): 114.81
(1C), 117.89 (1C), 119.46 (1C), 125.82 (2C), 126.24 (2C), 127.11 (1C),
127.91 (2C),128.05 (1C), 129.67 (2C), 135.37 (1C), 137.77 (1C),138.96
(1C), 144.03 (1C), 145.52 (1C), 195.55 (1C); IR (thin film, cmꢀ1):
3427, 3399, 3021, 1651, 1608, 1601, 1515, 1451, 1458.

R. Abu Khalaf et al. / European Journal of Medicinal Chemistry 45 (2010) 1598–1617
1615
4.7.2. General procedure for the synthesis of (4-amino-phenyl)-
(un)substituted phenyl-methanone (180–182)
4.7.3.2. {3-[(4-Methoxy-benzylidene)-amino]-phenyl}-(4-methoxy-
phenyl)-methanone (188). The residue was purified by crystalliza-
tion from petroleum ether to afford the title compound 188 as
a white powder (0.26 g, 25%): Rf ¼ 0.90 (CHCl3-MeOH, 97:3); mp:
4-Aminobenzoic acid 178 (0.411 g, 3 mmol) was dissolved in
CH2Cl2 (5 mL), and polyphosphoric acid (10 g) was added. Then
(un)substituted benzene (6–30 mmol) was added. The mixture was
stirred carefully at 80–190 ꢄC for 3 h and then poured on crushed
ice. The solution was carefully made alkaline with 25% ammonia
and then extracted with CH2Cl2 (3 ꢃ 20 mL). The combined extracts
were dried on anhydrous Na2SO4 and filtrated.
77–78 ꢄC; 1H-NMR (CDCl3,
d ppm): 3.87 (s, 3H, OCH3), 3.89 (s, 3H,
OCH3), 6.98 (m, 4H), 7.41 (dd, J ¼ 7.6, 1.5 Hz, 1H), 7.48 (t, J ¼ 7.6 Hz,
1H), 7.55 (dd, J ¼ 7.6, 1.5 Hz, 1H), 7.58 (t, J ¼ 1.5 Hz, 1H), 7.86 (m, 4H),
8.41 (s, 1H, N]CH); 13C-NMR (CDCl3,
d ppm): 55.69 (1C, OCH3),
55.74 (1C, OCH3), 113.84 (2C), 114.48 (2C), 121.85 (1C), 125.10 (1C),
127.07 (1C), 129.20 (1C), 129.23 (1C), 130.34 (1C), 130.92 (2C),
132.77 (2C),139.57 (1C),152.55 (1C),160.88 (1C),162.71 (1C),163.51
(1C), 195.60 (1C); IR (thin film, cmꢀ1): 1721, 1651, 1601, 1574, 1512,
1458, 1254, 1165; MS (ESI, positive mode): m/z [M þ H]þ 346.14377
(C22H20NO3 requires 346.13649); Anal. Calcd for C22H19NO3: C
76.50, H 5.54, N 4.06, found: C 76.33, H 5.45, N 3.68.
4.7.2.1. (4-Amino-phenyl)-phenyl-methanone (180) [48]. Evaporation
of the solvent gave 180 as a yellow powder (0.24 g, 41%) [48]: mp:
120–122 ꢄC (Lit. mp: 106–110 ꢄC); 1H-NMR (CDCl3,
d
ppm): 4.19 (bs,
2H, NH2), 6.66 (m, 2H), 7.46 (m, 2H), 7.53 (m, 2H), 7.70 (m, 3H); 13C-
NMR (CDCl3, ppm): 113.86 (2C), 127.57 (1C), 128.31 (2C), 129.72
d
(2C), 131.65 (1C), 133.18 (2C), 139.11 (1C), 151.25 (1C), 195.58 (1C); IR
(thin film, cmꢀ1): 3495, 3406, 3017, 1636, 1615, 1597, 1510, 1448.
4.7.3.3. Biphenyl-4-yl-{3-[(3-trifluoromethyl-benzylidene)-amino]-
phenyl}-methanone (189). The residue was purified by column
chromatography using CHCl3/EtOAc (99:1) as eluent, to afford the
title compound 189 as a light-brown powder (0.58 g, 45%): Rf ¼ 0.63
4.7.2.2. (4-Amino-phenyl)-biphenyl-4-yl-methanone (181). The re-
sidue, after evaporation of the solvent, was purified by column
chromatography eluting with CHCl3/MeOH (99:1) to give 181 pure
as a yellow powder (0.31 g, 38%): mp: 198–200 ꢄC; 1H-NMR (CDCl3,
(CHCl3-MeOH, 97:3); mp: 119–120 ꢄC; 1H-NMR (CDCl3,
d ppm): 7.32
(d, J ¼ 7.2 Hz, 1H), 7.46 (d, J ¼ 8.3 Hz, 2H), 7.51 (t, J ¼ 7.2 Hz, 1H), 7.62
(d, J ¼ 7.8 Hz, 1H), 7.71 (m, 2H), 7.80 (dd, J ¼ 7.8, 2.7 Hz, 4H), 7.86 (d,
J ¼ 8.1 Hz, 1H), 7.97 (d, J ¼ 8.1 Hz, 2H), 8.08 (d, J ¼ 8.3 Hz, 2H), 8.23
d
ppm): 4.15 (bs, 2H, NH2), 6.69 (dd, J ¼ 8.4, 1.8 Hz, 2H), 7.43 (m, 5H),
7.67 (m, 4H), 7.79 (m, 2H); 13C-NMR (CDCl3,
d
ppm): 112.64 (2C),
125.76 (2C),126.24 (2C),126.57 (1C),126.93 (1C),127.89 (2C),129.18
(2C), 131.87 (2C), 136.51 (1C), 139.21 (1C), 143.23 (1C), 149.86 (1C),
193.87 (1C); IR (thin film, cmꢀ1): 3433, 3341, 3021, 1652, 1613, 1597,
1539, 1448.
(s, 1H), 8.54 (s, 1H, N]CH); 13C-NMR (CDCl3,
d ppm): 119.29 (1C),
120.96 (1C), 123.25 (1C), 126.47 (2C), 127.77 (1C), 128.34 (2C),
129.09 (1C), 129.87 (2C), 130.11 (1C), 130.43 (2C), 131.78 (2C), 132.69
(2C), 135.21 (1C), 135.88 (1C), 137.56 (1C), 138.78 (1C), 144.16 (1C),
145.23 (1C), 155.52 (1C), 160.35 (1C), 195.46 (1C); IR (thin film,
cmꢀ1): 1721, 1651, 1601, 1582, 1485, 1454; MS (ESI, positive mode):
m/z [M þ H]þ 430.14133 (C27H19F3NO requires 430.13405); Anal.
Calcd for C27H18F3NO: C 75.52, H 4.22, N 3.26, found: C 75.47, H
4.24, N 3.23.
4.7.2.3. (4-Amino-phenyl)-(2,4-dimethoxy-phenyl)-methanone (182).
Evaporation of the solvent gave 182 as a yellow powder (0.69 g,
89%): mp: 164–166 ꢄC; 1H-NMR (CDCl3,
d ppm): 3.72 (s, 3H,
OCH3), 3.84 (s, 3H, OCH3), 4.14 (bs, 2H, NH2), 6.50 (s, 1H), 6.53
(dd, J ¼ 7.6, 2.4 Hz, 1H), 6.60 (dd, J ¼ 6.7, 1.9 Hz, 2H), 7.29 (dd,
J ¼ 7.6, 0.9 Hz, 1H), 7.65 (dd, J ¼ 6.7, 1.9 Hz, 2H); 13C-NMR (CDCl3,
4.7.3.4. Phenyl-{3-[(4-trifluoromethyl-benzylidene)-amino]-phenyl}-
methanone (190). The residue was purified by crystallization from
petroleum ether to afford the title compound 190 as a white
powder (0.11 g, 10%): Rf ¼ 0.67 (CHCl3-MeOH, 97:3); mp: 90–
d
ppm): 55.71 (1C, OCH3), 55.86 (1C, OCH3), 99.07 (1C), 104.48
(1C), 113.72 (2C), 122.71 (1C), 128.89 (1C), 131.37 (1C), 132.79
(2C), 151.23 (1C), 159.10 (1C), 162.67 (1C), 194.26 (1C); IR (thin
film, cmꢀ1): 3435, 3356, 3020, 1651, 1610, 1597, 1503, 1455, 1277,
1215.
92 ꢄC; 1H-NMR (CDCl3,
d
ppm): 7.49 (dd, J ¼ 2.0, 0.6 Hz, 1H), 7.52
(dd, J ¼ 7.6, 2.0 Hz, 2H), 7.61 (dd, J ¼ 7.8, 1.5 Hz, 1H), 7.65 (t,
J ¼ 7.6 Hz, 1H), 7.70 (dd, J ¼ 7.8, 1.5 Hz, 2 H), 7.74 (d, J ¼ 8.2 Hz,
2H), 7.84 (dd, J ¼ 7.8, 1.5 Hz, 2H), 8.03 (d, J ¼ 8.2 Hz, 2H), 8.55 (s,
4.7.3. Generalprocedureforthesynthesisofbenzylidene-aminophenyl-
methanone compounds (187–190, 193–199)
1H, N]CH); 13C-NMR (CDCl3,
d ppm): 121.99 (1C), 125.56 (1C),
One of the aminobenzophenone intermediates (175–177, 180–
182) (3 mmol) was dissolved in cyclohexane (20–25 mL). Then
(un)substituted benzaldehyde (7.5 mmol) was added. The mixture
was refluxed at 85–90 ꢄC for 3 h and then poured on crushed ice
and cooled. The resulting suspension was filtered, and washed with
cold aqueous ethanol (50%).
126.01 (1C), 126.06 (1C), 128.36 (1C), 128.62 (2C), 129.35 (2C),
129.45 (1C), 130.02 (2C), 130.32 (2C), 132.86 (1C), 137.63 (1C),
139.03 (1C), 151.71 (1C), 159.89 (1C), 159.92 (1C), 196.56 (1C); IR
(thin film, cmꢀ1): 1722, 1651, 1601, 1574, 1500, 1462; MS (ESI,
negative mode): m/z Mþ 353.10330 (C21H14F3NO requires
353.10275); Anal. Calcd for C21H14F3NO: C 71.38, H 3.99, N 3.96,
found: C 71.32, H 3.81, N 3.86.
4.7.3.1. [3-(Benzylidene-amino)-phenyl]-(4-methoxy-phenyl)-metha-
none (187). The residue was purified by column chromatography
using CHCl3/MeOH (99:1) as eluent, to afford the title compound
187 as yellow oil (0.66 g, 70%): Rf ¼ 0.50 (CHCl3-MeOH, 97:3); 1H-
4.7.3.5. Biphenyl-4-yl-{4-[(3-trifluoromethyl-benzylidene)-amino]-
phenyl}-methanone (193). The residue was purified by crystalli-
zation from petroleum ether to afford the title compound 193 as an
off-white powder (0.21 g, 16%): Rf ¼ 0.74 (CHCl3-MeOH, 97:3); mp:
NMR (CDCl3,
d
ppm): 3.87 (s, 3H, OCH3), 6.83 (dd, J ¼ 7.9, 2.3 Hz,
1H), 6.86 (dd, J ¼ 7.9, 2.3 Hz, 1H), 6.94 (dd, J ¼ 6.9, 2.1 Hz, 2H), 7.07
(t, J ¼ 7.9 Hz, 1H), 7.10 (t, J ¼ 2.3 Hz, 1H), 7.21 (t, J ¼ 8.4 Hz, 1H), 7.28
(dd, J ¼ 8.4, 2.4 Hz, 2H), 7.35 (dd, J ¼ 8.4, 2.4 Hz, 2H), 7.82 (dd,
165–166 ꢄC; 1H-NMR (CDCl3,
d
ppm): 7.30 (d, J ¼ 8.4 Hz, 2H), 7.43
(d, J ¼ 7.0 Hz, 1H), 7.50 (t, J ¼ 7.6 Hz, 2H), 7.66 (m, 3H), 7.73 (d,
J ¼ 8.4 Hz, 2H), 7.79 (d, J ¼ 7.6 Hz, 1H), 7.92 (m, 4H), 8.11 (d,
J ¼ 7.9 Hz, 1H), 8.23 (s, 1H), 8.54 (s, 1H, N]CH); 13C-NMR (CDCl3,
J ¼ 6.9, 2.1 Hz, 2H), 8.41 (s, 1H, N]CH); 13C-NMR (CDCl3,
d ppm):
55.72 (1C, OCH3), 111.71 (1C), 113.70 (2C), 115.71 (1C), 115.95 (1C),
118.69 (1C),119.89 (1C),120.46 (2C),120.93 (1C),129.31 (2C),130.53
(1C), 131.85 (1C), 132.77 (2C), 149.59 (1C), 161.72 (1C), 163.37 (1C),
196.05 (1C); IR (thin film, cmꢀ1): 1717, 1643, 1597, 1505, 1489, 1454,
1250; MS (ESI, negative mode): m/z [M þ Na]þ 338.11625
(C21H17NNaO2 requires 338.12593).
d ppm): 120.88 (2C), 125.93 (1C), 127.25 (2C), 127.55 (2C), 128.45
(2C), 128.52 (1C), 129.22 (2C), 129.68 (1C), 130.84 (2C), 131.82 (2C),
132.42 (1C), 132.43 (1C), 135.67 (1C), 136.67 (1C), 136.72 (1C),
140.22 (1C), 145.41 (1C), 155.35 (1C), 160.15 (1C), 195.74 (1C); IR
(thin film, cmꢀ1): 1720, 1643, 1601, 1580, 1435; MS (ESI, positive
mode): m/z [M þ H]þ 430.14133 (C27H19F3NO requires 430.13405);

1616
R. Abu Khalaf et al. / European Journal of Medicinal Chemistry 45 (2010) 1598–1617
Anal. Calcd for C27H18F3NO: C 75.52, H 4.22, N 3.26, found: C 75.42,
H 4.26, N 3.26.
OCH3), 55.77 (1C, OCH3), 55.82 (1C, OCH3), 99.05 (1C), 104.76 (1C),
114.51 (2C), 120.69 (2C), 122.06 (1C), 129.15 (1C), 131.06 (2C), 131.45
(2C), 132.17 (1C), 135.99 (1C), 156.36 (1C), 159.63 (1C), 160.93 (1C),
162.85 (1C), 163.39 (1C), 195.00 (1C); IR (thin film, cmꢀ1): 1721,
1647, 1601, 1574, 1512, 1308, 1254; MS (ESI, positive mode): m/z
[M þ H]þ 376.15433 (C23H22NO4 requires 376.14706); Anal. Calcd
for C23H21NO4: C 73.58, H 5.64, N 3.73, found: C 73.12, H 5.67, N
4.23.
4.7.3.6. {4-[(4-tert-Butyl-benzylidene)-amino]-phenyl}-(2,4-dimeth-
oxy-phenyl)-methanone (194). The residue was purified by crys-
tallization from diethyl ether to afford the title compound 194 as
a yellow powder (0.12 g, 10%):) Rf ¼ 0.65 (CHCl3-MeOH, 97:3); mp:
139–141 ꢄC; 1H-NMR (CDCl3,
d
ppm): 1.36 (s, 9H, 3 ꢃ CH3), 3.73 (s,
3H, OCH3), 3.88 (s, 3H, OCH3), 6.52 (d, J ¼ 2.2 Hz, 1H), 6.56 (dd,
J ¼ 8.4, 2.2 Hz, 1H), 7.18 (dd, J ¼ 6.7, 1.9 Hz, 2H), 7.41 (d, J ¼ 8.4 Hz,
1H), 7.51 (dd, J ¼ 6.7, 1.9 Hz, 2H), 7.84 (m, 4H), 8.43 (s, 1H, N]CH);
4.7.3.10. (2,4-Dimethoxy-phenyl)-{4-[(3-methoxy-benzylidene)-amino]-
phenyl}-methanone (198). The residue was purified by crystalliza-
tion from petroleum ether to afford the title compound 198 as
a yellow powder (0.29 g, 26%): Rf ¼ 0.83 (CHCl3-MeOH, 97:3); mp:
13C-NMR (CDCl3,
d ppm): 31.41 (3C, CH3), 35.34 (1C), 55.76 (1C,
OCH3), 55.82 (1C, OCH3), 99.05 (1C),104.77 (1C),120.64 (2C), 122.03
(1C), 126.08 (2C), 129.12 (2C), 131.43 (2C), 132.20 (1C), 133.49 (1C),
136.15 (1C),155.78 (1C),156.30 (1C), 159.64 (1C),161.56 (1C), 163.42
(1C), 195.00 (1C); IR (thin film, cmꢀ1): 1724, 1651, 1589, 1566, 1504,
1462, 1273, 1211; MS (ESI, positive mode): m/z [M þ H]þ 402.20637
(C26H28NO3 requires 402.19909); Anal. Calcd for C26H27NO3: C
77.78, H 6.78, N 3.49, found: C 77.73, H 6.79, N 3.17.
48–50 ꢄC; 1H-NMR (CDCl3,
d ppm): 3.73 (s, 3H, OCH3), 3.86 (s, 3H,
OCH3), 3.89 (s, 3H, OCH3), 6.52 (m, 4H), 7.19 (d, J ¼ 8.2 Hz, 1H), 7.29
(d, J ¼ 8.2 Hz, 1H), 7.53 (s, 1H), 7.66 (d, J ¼ 8.5 Hz, 2H), 7.83 (d,
J ¼ 8.5 Hz, 2H), 8.43 (s,1H, N]CH); 13C-NMR (CDCl3,
d ppm): 55.69
(1C, OCH3), 55.74 (1C, OCH3), 55.84 (1C, OCH3), 99.06 (1C), 104.63
(1C), 112.18 (1C), 113.73 (2C), 119.04 (1C), 120.65 (1C), 122.78 (2C),
131.40 (1C), 132.24 (1C), 132.79 (2C), 137.50 (1C), 151.12 (1C), 159.11
(1C), 160.27 (1C), 161.62 (1C), 162.67 (1C), 194.23 (1C); IR (thin film,
cmꢀ1): 1721, 1651, 1597, 1505, 1462, 1312, 1277; MS (ESI, positive
mode): m/z [M þ Na]þ 398.13628 (C23H21NNaO4 requires
398.14706), Anal. Calcd for C23H21NO4: C 73.58, H 5.64, N 3.73,
found: C 73.22, H 5.69, N 3.92.
4.7.3.7. (2,4-Dimethoxy-phenyl)-{4-[(4-trifluoromethyl-benzylidene)-
amino]-phenyl}-methanone (195). The residue was purified by
crystallization from petroleum ether to afford the title compound
195 as an off-white-yellowish powder (0.32 g, 26%): Rf ¼ 0.70
(CHCl3-MeOH, 97:3); mp: 102–104 ꢄC; 1H-NMR (CDCl3,
d ppm):
3.72 (s, 3H, OCH3), 3.88 (s, 3H, OCH3), 6.52 (d, J ¼ 2.1 Hz, 1H), 6.57
(dd, J ¼ 8.5, 2.3 Hz, 1H), 7.22 (dd, J ¼ 6.6, 1.8 Hz, 2H), 7.42 (d,
J ¼ 8.5 Hz, 1H), 7.75 (d, J ¼ 8.2 Hz, 2H), 7.85 (dd, J ¼ 6.6, 1.8 Hz, 2H),
8.04 (d, J ¼ 8.2 Hz, 2H), 8.51 (s, 1H, N]CH); 13C-NMR (CDCl3,
4.7.3.11. {4-[(4-Methoxy-benzylidene)-amino]-phenyl}-phenyl-meth-
anone (199). The residue was purified by crystallization from
diethyl ether to afford the title compound 199 as a yellowish-
orange powder (0.17 g, 18%): Rf ¼ 0.68 (CHCl3-MeOH, 97:3); mp:
d
ppm): 55.78 (1C, OCH3), 55.81 (1C, OCH3), 99.05 (1C), 104.85 (1C),
120.63 (2C),121.82 (1C),125.97 (1C),126.07 (2C),129.42 (2C),131.43
(2C), 132.32 (1C), 136.95 (1C), 139.10 (1C), 139.11 (1C), 155.16 (1C),
159.71 (1C), 159.95 (1C), 163.58 (1C), 194.89 (1C); IR (thin film,
cmꢀ1): 1721, 1651, 1601, 1574, 1501, 1462, 1308, 1277; MS (ESI,
positive mode): m/z [M þ H]þ 414.13115 (C23H19F3NO3 requires
414.12388); Anal. Calcd for C23H18F3NO3: C 66.86, H 4.39, N 3.39,
found: C 66.04, H 4.28, N 3.34.
114–116 ꢄC; 1H-NMR (CDCl3,
d ppm): 3.89 (s, 3H, OCH3), 7.00 (dd,
J ¼ 6.7, 2.1 Hz, 2H), 7.24 (dd, J ¼ 6.7, 2.1 Hz, 2H), 7.51 (dd, J ¼ 6.5,
1.5 Hz, 2H), 7.57 (dd, J ¼ 6.0, 1.8 Hz, 1H), 7.81 (dd, J ¼ 6.0, 1.8 Hz, 2H),
7.87 (m, 4H), 8.40 (s, 1H, N]CH); 13C-NMR (CDCl3,
d ppm): 55.72
(1C, OCH3), 114.54 (2C), 120.91 (2C), 128.49 (2C), 129.09 (1C), 130.13
(2C), 131.13 (2C), 131.87 (2C), 132.39 (1C), 134.63 (1C), 138.25 (1C),
156.54 (1C), 161.17 (1C), 162.94 (1C), 196.23 (1C); IR (thin film,
cmꢀ1): 1721, 1651, 1601, 1574, 1512, 1447, 1254; MS (ESI, positive
mode): m/z [M þ H]þ 316.13321 (C21H18NO2 requires 316.12593);
Anal. Calcd for C21H17NO2: C 79.98, H 5.43, N 4.44, found: C 79.39, H
5.45, N 4.47.
4.7.3.8. (2,4-Dimethoxy-phenyl)-{4-[(3-trifluoromethyl-benzylidene)-
amino]-phenyl}-methanone (196). The residue was purified by
crystallization from petroleum ether to afford the title compound
196 as an off-white powder (0.61 g, 49%): Rf ¼ 0.67 (CHCl3-MeOH,
97:3); mp: 79–80 ꢄC; 1H-NMR (CDCl3,
d ppm): 3.73 (s, 3H, OCH3),
Acknowledgments
3.88 (s, 3H, OCH3), 6.52 (d, J ¼ 2.3 Hz, 1H), 6.57 (dd, J ¼ 8.5, 2.3 Hz,
1H), 7.21 (dd, J ¼ 6.5, 1.9 Hz, 2H), 7.42 (d, J ¼ 8.5 Hz, 1H), 7.62 (t,
J ¼ 7.7 Hz, 1H), 7.76 (d, J ¼ 7.7 Hz, 1H), 7.84 (dd, J ¼ 6.5, 1.9 Hz, 2H),
8.09 (d, J ¼ 7.7 Hz, 1H), 8.21 (s, 1H), 8.51 (s, 1H, N]CH); 13C-NMR
This project was partially sponsored by the Faculty of Graduate
Studies (Ph.D. Thesis of Reema Abu Khalaf). The authors wish to
thank the Deanship of Scientific Research and Hamdi-Mango
Center for Scientific Research at the University of Jordan for their
generous funds. The authors are also indebted to national cancer
institute for freely providing hit molecules for evaluation.
(CDCl3,
d ppm): 55.78 (1C, OCH3), 55.80 (1C, OCH3), 99.04 (1C),
104.85 (1C), 120.62 (2C), 121.84 (1C), 125.80 (1C), 125.85 (1C),
128.38 (1C),129.64 (1C),131.44 (2C),131.48 (1C), 132.31 (1C),132.39
(1C), 136.78 (1C), 136.89 (1C), 155.17 (1C), 159.71 (1C), 159.86 (1C),
163.56 (1C), 194.90 (1C); IR (thin film, cmꢀ1): 1719, 1651, 1601, 1505,
1458, 1331, 1277; MS (ESI, positive mode): m/z [M þ H]þ 414.13115
(C23H19F3NO3 requires 414.12388); Anal. Calcd for C23H18F3NO3: C
66.86, H 4.39, N 3.39, found: C 66.93, H 4.34, N 3.23.
Appendix. Supplementary material
Supplementary data associated with this article can be found in
the online version, at doi:10.1016/j.ejmech.2009.12.070.
4.7.3.9. (2,4-Dimethoxy-phenyl)-{4-[(4-methoxy-benzylidene)-amino]-
phenyl}-methanone (197). The residue was purified by crystalliza-
tion from petroleum ether to afford the title compound 197 as an
off-white powder (0.12 g, 11%): Rf ¼ 0.61 (CHCl3-MeOH, 97:3); mp:
References
[1] G.K. Hansson, Inflammation, atherosclerosis, and coronary artery disease. N.
Engl. J. Med. 352 (2005) 1685–1695.
[2] B. Lamarche, J.P. Despres, S. Moorjani, B. Cantin, G.R. Dagenais, P.J. Lupien,
Triglycerides and HDL-cholesterol as risk factors for ischemic heart disease,
Results from the Quebec cardiovascular study. Atherosclerosis 119 (1996)
235–245.
127–128 ꢄC; 1H-NMR (CDCl3,
d ppm): 3.72 (s, 3H, OCH3), 3.88 (s, 3H,
OCH3), 3.89 (s, 3H, OCH3), 6.52 (d, J ¼ 2.3 Hz, 1H), 6.57 (dd, J ¼ 8.5,
2.3 Hz,1H), 6.99 (dd, J ¼ 6.8, 2.1 Hz, 2H), 7.18 (dd, J ¼ 6.7, 2.1 Hz, 2H),
7.41 (d, J ¼ 8.5 Hz, 1H), 7.82 (dd, J ¼ 6.8, 2.1 Hz, 2H), 7.86 (dd, J ¼ 6.7,
[3] A.R. Tall, N. Wang, P. Mucksavage, Is it time to modify the reverse cholesterol
transport model? J. Clin. Invest 108 (2001) 1273–1275.
2.1 Hz, 2H), 8.39 (s,1H, N]CH); 13C-NMR (CDCl3,
d ppm): 55.70 (1C,

R. Abu Khalaf et al. / European Journal of Medicinal Chemistry 45 (2010) 1598–1617
1617
[4] X. Qiu, A. Mistry, M.J. Ammirati, B.A. Chrunyk, R.W. Clark, Y. Cong, J.S. Culp,
D.E. Danley, T.B. Freeman, K.F. Geoghegan, M.C. Griffor, S.J. Hawrylik,
C.M. Hayward, P. Hensley, L.R. Hoth, G.A. Karam, M.E. Lira, D.B. Lloyd,
K.M. McGrath, K.J. Stutzman-Engwall, A.K. Subashi, T.A. Subashi,
J.F. Thompson, I.K. Wang, H. Zhao, A.P. Seddon, Crystal structure of cholesteryl
ester transfer protein reveals a long tunnel and four bound lipid molecules.
Nat. Struct. Mol. Biol. 14 (2007) 106–113.
[29] P.A. Keller, M. Bowman, K.H. Dang, J. Garner, S.P. Leach, R. Smith,
A.J. McCluskey, Pharmacophore development for corticotropin-releasing
hormone: new insights into inhibitor activity. J. Med. Chem. 42 (1999) 2351–
2357.
[30] R.G. Karki, V.M. Kulkarni, A feature based pharmacophore for Candida albicans
MyristoylCoA: protein N-myristoyltransferase inhibitors. Eur. J. Med. Chem. 36
(2001) 147–163.
[5] S.M. Boekholdt, J.A. Kuivenhoven, N.J. Wareham, R.J.G. Peters, J.W. Jukema,
R. Luben, S.A. Bingham, N.E. Day, J.J.P. Kastelein, K.T. Khaw, Plasma levels of
cholesteryl ester transfer protein and the risk of future coronary artery disease
in apparently healthy men and women. Circulation 110 (2004) 1418–1423.
[6] R. McPherson, C.J. Mann, A.R. Tall, M. Hogue, L. Martin, R.W. Milne, Y.L. Marcel,
Plasma concentration of cholesteryl ester transfer protein in hyper-
lipoproteinemia. Arterioscler. Thromb 11 (1991) 797–804.
[7] M.S. Castilho, S. Marcelo, R.V.C. Guido, A.D. Andricopulo, 2D Quantitative
structure–activity relationship studies on a series of cholesteryl ester transfer
protein inhibitors. Bioorg. Med. Chem. 15 (2007) 6242–6252.
[8] P. Hanumantharao, S.V. Sambasivarao, L.K. Soni, A.K. Gupta, S.G. Kaskhedikar,
QSAR analysis of analogs of bis[2-(acylamino) Ph] disulfides, 2-(acylamino)
benzenethiols and S-[2-(acylamino) Ph] alkanethioates as antihyperlipidemic
agents. Ind. J. Chem., Sect. B, Org. Chem. Includ. Med. Chem. 44B (2005)
1481–1486.
[9] M.A. Kelkar, D.V. Pednekar, S.R. Pimple, K.G. Akamanchi, 3D QSAR studies of
inhibitors of cholesterol ester transfer protein (CETP) by CoMFA, CoMSIA and
GFA methodologies. Med. Chem. Res. 13 (2004) 590–604.
[31] M.O. Taha, A.G. Al-Bakri, W.A. Zalloum, Discovery of potent inhibitors of
pseudomonal quorum sensing via pharmacophore modeling and In silico
screening. Bioorg. Med. Chem. Lett. 16 (2006) 5902–5906.
[32] K. Moffat, V.J. Gillet, M. Whittle, G. Bravi, A.R. Leach, A comparison of field-
based similarity searching methods, CatShape, FBSS, and ROCS. J. Chem. Inf.
Model 48 (2008) 719–729.
[33] R.C. Durley, M.L. Grapperhaus, M.A. Massa, D.A. Mischke, B.L. Parnas,
Y.M. Fobian, N.P. Rath, D.D. Honda, M. Zeng, D.T. Connolly, D.M. Heuvelman,
B.J. Witherbee, K.C. Glenn, E.S. Krul, M.E. Smith, J.A. Sikorski, Discovery of
chiral N, N-disubstituted trifluoro-3-amino-2-propanols as potent inhibitors
of cholesteryl ester transfer protein. J. Med. Chem. 43 (2000) 4575–4578.
[34] M.A. Massa, D.P. Spangler, R.C. Durley, B.S. Hickory, D.T. Connolly,
B.J. Witherbee, M.E. Smith, J.A. Sikorski, Novel heteroaryl replacements of
aromatic 3-tetrafluoroethoxy substituents in trifluoro-3-(tertiary amino)-2-
propanols as potent inhibitors of cholesteryl ester transfer protein. Bioorganic.
Med. Chem. Lett. 11 (2001) 1625–1628.
[35] R.C. Durley, M.L. Grapperhaus, B.S. Hickory, M.A. Massa, J.L. Wang,
D.P. Spangler, D.A. Mischke, B.L. Parnas, Y.M. Fobian, N.P. Rath, D.D. Honda,
M. Zeng, D.T. Connolly, D.M. Heuvelman, B.J. Witherbee, M.A. Melton,
K.C. Glenn, E.S. Krul, M.E. Smith, J.A. Sikorski, Chiral N,N-disubstituted tri-
fluoro-3-amino-2-propanols are potent inhibitors of cholesteryl ester transfer
protein. J. Med. Chem. 45 (2002) 3891–3904.
[10] M.T.D. Cronin, T.W. Schultz, Pitfalls in QSAR. J. Mol. Struct. (Theochem.) 622
(2003) 39–51.
[11] M. Akamatsu, Current state and perspectives of 3D-QSAR. Curr. Top. Med.
Chem. 12 (2002) 1381–1394.
[12] M.O. Taha, Y. Bustanji, M. Al-Ghussein, M. Mohammad, H. Zalloum, I.M. Al-
Masri, N. Atallah, Pharmacophore modeling, quantitative structure–activity
relationship analysis, and in-silico screening reveal potent glycogen synthase
kinase-3beta inhibitory activities for cimetidine, hydroxychloroquine, and
gemifloxacin. J. Med. Chem. 51 (2008) 2062–2077.
[13] I.M. Al-masri, K. Mohammad, M.O. Taha, Discovery of DPP IV inhibitors by
pharmacophore modeling and QSAR analysis followed by in silico screening.
Chem. Med. Chem. 3 (11) (2008) 1763–1779.
[14] M.O. Taha, L.A. Dahabiyeh, Y. Bustanji, H. Zalloum, S. Saleh, Combining ligand-
based pharmacophore modeling, QSAR analysis and in-silico screening for the
discovery of new potent hormone sensitive lipase inhibitors. J. Med. Chem. 51
(2008) 6478–6494.
[15] M.O. Taha, N. Atallah, A.G. Al-Bakri, C. Paradis-Bleau, H. Zalloum, K. Younis,
R.C. Levesque, Discovery of new MurF Inhibitors via pharmacophore modeling
and QSAR analysis followed by in silico screening. Bioorg. Med. Chem. 16
(2008) 1218–1235.
[36] E.J. Reinhard, J.L. Wang, R.C. Durley, Y.M. Fobian, M.L. Grapperhaus,
B.S. Hickory, M.A. Massa, M.B. Norton, M.A. Promo, M.B. Tollefson,
W.F. Vernier, D.T. Connolly, B.J. Witherbee, M.A. Melton, K.J. Regina,
M.E. Smith, J.A. Sikorski, Discovery of a simple picomolar inhibitor of cho-
lesteryl ester transfer protein. J. Med. Chem. 46 (2003) 2152–2168.
[37] R.P. Sheridan, S.K. Kearsley, Why do we need so many chemical similarity
search methods? Drug Discov. Today 7 (2002) 903–911.
[38] R. Fisher, in: The Principle of Experimentation Illustrated by a Psycho-Physical
ExpeHafner Publishing Co, eighth ed. Hafner Publishing, New York, 1966.
[39] A. Tropsha, P. Gramatica, V.K. Gombar, Quant. structure–activity relationship.
Comb. Sci. 22 (2003) 69–77.
[40] CERIUS2 QSAR Users’ Manual. Accelrys Inc., San Diego, CA, 2005.
[41] F.P. Maguna, M.B. Nunez, N.B. Okulik, E.A. Castro, Methodologies QSAR/QSPR/
QSTR: current state and perspectives. Int. J. Chem. Model 1 (2) (2009) 221–243.
[42] L.F. Ramsey, W.D. Schafer, in: The Statistical Sleuth, first ed. Wadesworth
Publishing Company, USA, 1997.
[16] M.O. Taha, Y. Bustanji, A.G. Al-Bakri, M. Yousef, W.A. Zalloum, I.M. Al-Masri,
N. Atallah, Discovery of new potent human protein tyrosine phosphatase
inhibitors via pharmacophore and QSAR analysis followed by in silico
screening. J. Mol. Graphics Model 25 (2007) 870–884.
[17] A.M. Abu Hammad, M.O. Taha, Pharmacophore modeling, quantitative struc-
ture–activity relationship analysis, and shape-complemented in silico
screening allow access to novel influenza neuraminidase inhibitors. J. Chem.
Inf. Model. 49 (2009) 978–996.
[43] E.M. Krovat, T. Langer, Non-peptide angiotensin ii receptor antagonists:
chemical feature based pharmacophore identification. J. Med. Chem. 46
(2003) 716–726.
[44] L.B. Kier, L.H. Hall, in: J. Devillers, A.T. Balaban (Eds.), Topological Indices and
Related Descriptors in QSAR and QSPR, Gordon and Breach, London, 1999, pp.
491–562.
[45] S.W. Homans, Water, water everywhere d except where it matters? Drug
Discov. Today 12 (2007) 534–539.
[18] Catalyst User Guide. Accelrys Software Inc., San Diego, CA, 2005.
[19] P.W. Sprague, R. Hoffmann, in: H. Van de Waterbeemd, B. Testa, G. Folkers
(Eds.), CATALYST Pharmacophore Models and Their Utility As Queries for
Searching 3D Databases, Current Tools for Medicinal Chemistry, VHCA, Basel,
1997, pp. 223–240.
[20] D. Barnum, J. Greene, A. Smellie, P.J. Sprague, IdentificationofCommonFunctional
configurations among molecules. Chem. Inf. Comput. Sci. 36 (1996) 563–571.
[21] A. Smellie, S. Teig, P. Towbin, Poling: promoting conformational variation. J.
Comput. Chem. 16 (1995) 171–187.
[22] H. Li, J. Sutter, R. Hoffmann, in: O.F. Guner (Ed.), Pharmacophore Perception,
Development, and Use in Drug Design, International University Line, Cal-
ifornia, 2000, pp. 173–189.
[23] J. Sutter, O.F. Guner, R. Hoffmann, H. Li, M. Waldman, Effect of Variable
Weights and Tolerances on Predictive Model Generation, Pharmacophore
Perception, Development, and Use in Drug Design. International University
Line, California, 2000, pp. 501–511.
[46] C.A. Lipinski, F. Lombardo, B.W. Dominy, P.J. Feeney, Experimental and
computational approaches to estimate solubility and permeability in drug
discovery and development settings. Adv. Drug Del. Rev. 46 (2001) 3–26.
[47] D.F. Veber, S.R. Johnson, H.Y. Cheng, B.R. Smith, K.W. Ward, K.D. Kopple,
Molecular properties that influence the oral bioavailability of drug candidates.
J. Med. Chem. 45 (2002) 2615–2623.
[48] I. Ivanov, S. Nikolova, S. Statkova-Abeghe, Efficient one-pot friedel–crafts
acylation of benzene and its derivatives with unprotected aminocarboxylic
acids in polyphosphoric acid. Synth. Commun. 36 (2006) 1405–1411.
[49] V. del Amo, D. Philp, Making imines without making water-exploiting
a recognition-mediated Aza-Wittig reaction. Org. Lett. 11 (2009) 301–304.
[50] M.L. Verdonk, L. Marcel, V. Berdini, M.J. Hartshorn, W.T.M. Mooij, C.W. Murray,
R.D. Taylor, P. Watson, Virtual screening using protein-ligand docking:
avoiding artificial enrichment. J. Chem. Inf. Comput. Sci. 44 (2004) 793–806.
[51] J. Kirchmair, P. Markt, S. Distinto, G. Wolber, T. Langer, Evaluation of the
performance of 3D virtual screening protocols: RMSD comparisons, enrich-
ment assessments, and decoy selectiondwhat can we learn from earlier
mistakes? J. Comput. Aided. Mol. Des. 22 (2008) 213–228.
[52] J.J. Irwin, B.K. Shoichet, ZINC – a free database of commercially available
compounds for virtual screening. J. Chem. Inf. Comput. Sci. 45 (2005) 177–182.
[53] N. Triballeau, F. Acher, I. Brabet, J.P. Pin, H.O. Bertrand, Virtual screening
workflow development guided by the ‘‘receiver-operating characteristic’’
curve approach. application to high-throughput docking on metabotropic
glutamate receptor subtype 4. J. Med. Chem. 48 (2005) 2534–2547.
[54] M. Jacobsson, P. Liden, E. Stjernschantz, H. Bostroem, U. Norinder, Improving
structure-based virtual screening by multivariate analysis of scoring data. J.
Med. Chem. 46 (2003) 5781–5789.
[24] Y. Kurogi, O.F. Guner, Pharmacophore modeling and three-dimensional database
searching for drug design using catalyst. Curr. Med. Chem. 8 (2001) 1035–1055.
[25] I.B. Bersuker, S. Bahçeci, J.E. Boggs, Pharmacophore Perception, Development,
}
and Use in Drug Design. in: O.F. Guner (Ed.). International University Line,
California, 2000, pp. 457–473.
[26] K. Poptodorov, T. Luu, R. Hoffmann, in: T. Langer, R.D. Hoffmann (Eds.),
Methods and Principles in Medicinal Chemistry, Pharmacophores and Phar-
macophores Searches, Wiley-VCH, Weinheim 2, 2006, pp. 17–47.
[27] J. Singh, C.E. Chuaqui, P.A. Boriack-Sjodin, W.C. Lee, T. Pontz, M.J. Corbley,
H.K. Cheung, R.M. Arduini, J.N. Mead, M.N. Newman, J.L. Papadatos, S. Bowes,
S. Josiah, L.E. Ling, Successful shape-based virtual screening: the discovery of
a potent inhibitor of the type I TGF
Chem. Lett. 13 (2003) 4355–4359.
b
receptor kinase (T
b
RI). Bioorg. Med.
[55] H. Gao, C. Williams, P. Labute, J. Bajorath, Binary quantitative structure–
activity relationship (QSAR) analysis of estrogen receptor ligands. J. Chem. Inf.
Comput. Sci. 39 (1999) 164–168.
[56] M. Hahn, Three-dimensional shape-based searching of conformationally
flexible compounds. J. Chem. Inf. Comput. Sci. 37 (1997) 80–86.
[28] M.O. Taha, A.M. Qandil, D.D. Zaki, M.A. AlDamen, Ligand-based assessment of
factor Xa binding site flexibility via elaborate pharmacophore exploration and
genetic algorithm-based QSAR modeling. Eur. J. Med. Chem. 40 (2005) 701–727.