Integrating docking scores, interaction profiles and molecular descriptors to improve the accuracy of molecular docking: Toward the discovery of novel Akt1 inhibitors

Article abstract of DOI:10.1016/j.ejmech.2014.01.019

A set of forty-seven Akt1 inhibitors was used for the development of molecular docking based QSAR model by using nonlinear regression. The integration of docking scores, key interaction profiles and molecular descriptors remarkably improved the accuracy of the QSAR models, providing reasonable statistical parameters (R_train² = 0.948, R _test² = 0.907 and Q_cv² = 0.794). The established MD-SVR model based structural modification of new 4-amino-pyrimidine derivatives was further performed, and six compounds 56a,b and 60a-d with good prediction activities were synthesized and biologically evaluated. All of these compounds exhibited promising Akt1 inhibitory and antiproliferative activities, suggesting the reliability and good application value of the established MD-SVR model in the development of Akt1 inhibitors.

Full text of DOI:10.1016/j.ejmech.2014.01.019

European Journal of Medicinal Chemistry 75 (2014) 11e20
Contents lists available at ScienceDirect
European Journal of Medicinal Chemistry
journal homepage: http://www.elsevier.com/locate/ejmech
Original article
Integrating docking scores, interaction profiles and molecular
descriptors to improve the accuracy of molecular docking: Toward the
discovery of novel Akt1 inhibitors
Wenhu Zhan a, Daqiang Li a, Jinxin Che a, Liangren Zhang b, Bo Yang c, Yongzhou Hu a,
**
,
a,
*
Tao Liu a , Xiaowu Dong
a ZJU-ENS Joint Laboratory of Medicinal Chemistry, Zhejiang Province Key Laboratory of Anti-Cancer Drug Research, College of Pharmaceutical Sciences,
Zhejiang University, Hangzhou 310058, China
b State Key Laboratory of Natural and Biomimetic Drugs, Peking University, Beijing 100083, China
c Department of Pharmacology, School of Medicine, Zhejiang University, Hangzhou 310058, China
a r t i c l e i n f o
a b s t r a c t
Article history:
A set of forty-seven Akt1 inhibitors was used for the development of molecular docking based QSAR
model by using nonlinear regression. The integration of docking scores, key interaction profiles and
molecular descriptors remarkably improved the accuracy of the QSAR models, providing reasonable
statistical parameters (Rt2rain ¼ 0.948, Rt2est ¼ 0.907 and Qc2v ¼ 0.794). The established MD-SVR model
based structural modification of new 4-amino-pyrimidine derivatives was further performed, and six
compounds 56a,b and 60aed with good prediction activities were synthesized and biologically evalu-
ated. All of these compounds exhibited promising Akt1 inhibitory and antiproliferative activities, sug-
gesting the reliability and good application value of the established MD-SVR model in the development
of Akt1 inhibitors.
Received 4 September 2013
Received in revised form
8 January 2014
Accepted 13 January 2014
Available online 22 January 2014
Keywords:
Akt1 inhibitors
Molecular docking
Support vector regression (SVR)
Synthesis
Ó 2014 Elsevier Masson SAS. All rights reserved.
Antiproliferative activity
1. Introduction
et al., demonstrated that seven docking scores were integrated by
the support vector rank regression (SVRR) algorithm to improve the
Structure-based drug design (SBDD) is a rational strategy for
identification of bioactive hits and optimization of lead compounds.
Several tools are commonly used in SBDD, such as molecular
docking, NMR spectroscopy and X-ray crystallography. Molecular
docking is a computational technique that can rapidly propose a
binding mode between a ligand and its receptor. The reliability of
molecular docking in SBDD is significantly affected by the accuracy
of docking scores and the 3D-structure of receptor. Recently, there
is a growing interest in applying mathematical methods to improve
the accuracy of molecular docking, including multiple linear
regression (MLR), artificial neural network (ANN) and support
vector machine (SVM). Among them, SVM is a popular algorithm
developed from the machine learning community [1], showing
good predictive ability in QSAR studies [2,3]. For example, Chen
prediction for binding conformation [4]. In addition, the accuracy of
molecular docking can be improved by integrating additional var-
iables that indicate the fitness of a computed binding conformation
[5e7].
The serine/threonine kinase Akt (also known as protein kinase
B) has become a promising potential target because of its essential
regulatory role in cellular processes, including cancer progression
and insulin metabolism [8e12]. Akt regulates cell growth through
its effects on the TSC1/TSC2 complex and mTOR pathways, as well
as cell cycle and cell proliferation [13e15]. Inhibition of Akt activ-
ities can result in increasing apoptosis of cancer cells, and
decreasing tumor growth [11,16e18]. Currently, many companies
and academic laboratories have initiated a variety of approaches to
inhibit the Akt pathway [16,19e21], leading to the discovery of a
number of Akt inhibitors. However, there are still some issues (e.g.
selectivity, toxicity and efficacy) that need to be resolved. Thus, it is
still greatly of interest in the development of novel Akt inhibitors,
especially for reliable and efficient method to accelerate the pro-
cess. According to the presence of high resolution 3D structures of
* Corresponding author. Tel./fax: þ86 571 88981051.
** Corresponding author. Tel./fax: þ86 571 88208460.
E-mail addresses: lt601@zju.edu.cn (T. Liu), dongxw@zju.edu.cn (X. Dong).
0223-5234/$ e see front matter Ó 2014 Elsevier Masson SAS. All rights reserved.
http://dx.doi.org/10.1016/j.ejmech.2014.01.019

12
W. Zhan et al. / European Journal of Medicinal Chemistry 75 (2014) 11e20
Akt1 bound with its inhibitors, structure-based drug design would
be a promising strategy.
2.2. Molecular docking with LigandFit [29]
Although several attempts have been made recently in gener-
ating the predictive models for Akt1 inhibitors through ligand-
based methods [22,23], there is few structure-based predictive
QSAR models for identification and optimization of Akt1 inhibitors.
Herein, molecular docking based support vector regression (MD-
SVR) method was applied to establish a reliable structure-based
QSAR model. As revealed both in X-ray diffraction and molecular
docking of ligand-bound Akt1 complex, several crucial interactions
were commonly observed, such as amino acid residues Glu228,
Ala230, Glu234, Glu292 and Phe163. In addition, considering that
hydrophilic and hydrophobic interactions are two major intermo-
lecular forces, molecular descriptors (e.g. ClogP and tPSA) would be
important properties for binding affinity of the ligands with Akt1.
Thus, in addition to the generated docking scores by molecular
docking, key interactions profiles and molecular descriptors (Sup-
porting information, Table S1) were integrated, with aim to
enhance the accuracy and reliability of molecular docking (Scheme
1).
The robustness of the established MD-SVR model was validated
by the prediction for training and test set, as well as the internal
leave-one-out cross validation. Moreover, the MD-SVR model was
used in structure-based design of 4-aminopyrimidine derivatives as
novel Akt1 inhibitors. Six compounds with good prediction Akt1
inhibitory activities were synthesized and biological evaluated for
Akt1 inhibitory and antiproliferative activities against cancer cell
lines.
The docking pocket of Akt1 (PDB ID: 3OCB) was obtained using
the LigandFit site search utility in Discovery studio 2.5. For the
generation of ligands’ conformations, variable numbers of Monte
Carlo simulations were used. All the calculations were performed
under the PLP Force Field formalism [30]. A short rigid body
minimization was then performed, and 50 poses for each ligand
were saved for further analysis. Scoring was performed with a set of
scoring functions (i.e. Dock_score, LigScore2, PLP1, Jain and PMF)
implemented in LigandFit module. The combination of consensus
scoring method and the interaction mode analysis was applied to
select the preferable output conformation.
2.3. Support vector regression (SVR) method [1]
The basic idea of SVM regression is to map the data x into a
higher-dimensional feature space F via a nonlinear mapping v and
then to do linear regression in this space. Therefore, regression
approximation addresses the problem of estimating a function
based on a given data set G ¼ fðxi; diÞgni (xi is the input vector, di is
the desired value, and n is the total number of data patterns). SVM
regression approximates the function in the following form:
l
X
y ¼
w
ifðxiÞ þ b
(1)
i ¼ 1
where
f
ðxÞ is the high dimensional feature space, which is non-
linearly mapped from the input space x and wi and b are co-
efficients. They are estimated by minimizing the regularized risk
function R(C)
2. Computational method
2.1. Data set
N
X
1
N
1
2
For molecular docking based QSAR modeling, a set of forty-seven
molecules (Fig.1) were selected from the literature [18,24e27]. Akt1
inhibitory activity data (Table 1) of them span over 4 orders of
magnitude (from 2 to 4000 nM). The biological data of Akt1
inhibitory activities [IC50 (nM)] were converted to logarithmic scale
[ꢀLog IC50] and then used for subsequent QSAR analyses as the
response variable (dependent variable). For model training and
validation, the total data set (n ¼ 47) was randomly divided into
training set (n ¼ 36) and test set (n ¼ 11). The 3D-strucure of all
compounds were generated and optimized in Discovery Studio 2.0
software (Accelrys, Inc. San Diego, CA) using the CHARMm-like
force field [28].
RðCÞ ¼ C
Lxðdi; yiÞ þ kwk2;
(2)
i ¼ 1
where
ꢀ
jd ꢀ yj ꢀ
3
forjd ꢀ yj ꢀ
3 ꢁ 0
L 3 ðdi; yiÞ ¼
;
(3)
0
otherwise
P
N
The first term C N1 i ¼ 1 Lxðdi; yiÞ is the so-called empirical error
(risk), which is measured by the
3
2
1
second term 2kwk , on the other hand, is called the regularized
term. is called the tube size of SVM, and C is the regularization
3
Scheme 1. The evolution of the establishment of MD-SVR model and its application in development of novel 4-aminopyrimidine derivatives as Akt inhibitors.

W. Zhan et al. / European Journal of Medicinal Chemistry 75 (2014) 11e20
13
Fig. 1. The 2D structures of 47 Akt1 inhibitors used to generate the MD-SVR model.
constant determining the trade-off between the empirical error
2.4. Leave-one-out cross validation
and the regularized term. Introduction of slack variables ‘
x
’ leads Eq.
(2) to the following constrained function:
The leave-one-out cross validation was employed to find the
promising QSAR model. Given n samples available in a data set
and m candidate models, each model is trained with n ꢀ 1
samples and then is tested on the sample that was left out. This
process is repeated n times until every sample in the data set
have been used once as a cross-validation instance. Finally, cross
validation correlation coefficient (Q2loo), a measure of the model
generalization capability, for all candidate models can be ob-
tained as below:
n
ꢁ
ꢂ
ꢁ
ꢂ
X
1
2
Minimize R w; xi
;
x*
¼
kwk2 þ C
xi
þ
xi*
;
(4)
i ¼ 1
Thus, decision function of Eq. (1) changes to the following form:
l
ꢁ
ꢂ
ꢁ
ꢂ
X
f x; ai
;
ai*
¼
ai
ꢀ
a*i Kðx; xiÞ þ b;
(5)
i ¼ 1
where ai anda*i are the introduced Lagrange multipliers and K(x, xi)
is the kernel function. The value is equal to the inner product of two
n
P
2
b
i¼1ðyi ꢀ yiÞ
Q2 ¼ 1 ꢀ
P
i¼1ðyi ꢀ ymÞ2
n
vectors x and xi in the feature space
F(x) and F(xi), that is, K(x,
xi) ¼
F
(x)T
F
(xi). And the radial basis function (RBF) kernel
b
where yi is the desired output, yi is the predicted value by model
and ym is the mean value of dependent variable.
ꢃ
ꢃ
ꢃ
ꢃ
x ꢀ x 2Þ is commonly used.
ꢃ
Kðxi; xjÞ ¼ expðꢀ
g
ꢃ
i
j

14
W. Zhan et al. / European Journal of Medicinal Chemistry 75 (2014) 11e20
observed (Fig. 2) and their R2 values with respect to activity were
<0.417 (for overall set). Hence, these individual scoring functions
were not enough accurate to prediction the Akt1 inhibitory activ-
ities, prompting us to develop the consensus models that inte-
grating docking scores, the key interactions profiles and molecular
descriptors to improve the accuracy. The distance between the key
amino acid residue (e.g. Glu228, Ala230, Glu234, Glu292 and
Phe163) of Akt1 and docked compounds was recorded as key in-
teractions profiles. Further, molecular descriptors (e.g. ClogP and
tPSA) were also used to get better insight into the physicochemical
requirements for effective binding of ligands with Akt1.
Table 1
Compounds 1e47 in training set and test set for MD-SVR model, and their corre-
sponding experimental and estimated Akt1 inhibitory activities.
NO.
Experimental
IC50(nM)
Predictive
Res.
b
ꢀLog IC50
ꢀLog IC50
1
2
3
4
3
6
2
3
5
5
24
12
8
2
9
18
5
3
158
32
3
8.52
8.22
8.70
8.52
8.30
8.30
7.62
7.92
8.10
8.70
8.05
7.74
8.30
8.52
6.80
7.49
8.52
8.52
6.62
6.15
6.22
6.15
5.62
7.76
7.38
6.68
6.37
6.50
6.82
6.77
8.28
8.51
5.55
8.32
5.50
8.24
8.32
8.62
6.74
6.26
6.24
6.82
6.13
7.48
5.40
6.82
7.60
8.25
8.20
8.18
8.25
8.27
8.18
7.92
8.09
8.24
8.08
8.25
8.02
8.17
8.31
6.11
7.32
8.24
7.75
6.69
6.21
6.32
6.25
5.77
7.55
7.13
6.70
6.48
6.74
6.90
6.92
8.30
8.30
5.65
8.32
5.80
8.47
8.50
8.53
6.91
6.36
6.72
6.62
6.32
6.92
5.47
6.82
7.38
0.27
0.02
0.52
0.27
0.03
0.12
5
6a
7
ꢀ0.3
In MD-SVR modeling, the RBF kernel was used as kernel func-
8
ꢀ0.17
ꢀ0.14
0.62
9a
tion. Capacity parameter C,
3
of
3 -insensitive loss function and the
10
11
12
13
14
15a
16
17
18
19
20a
21
22
23
24
25
26a
27
28a
29
30
31a
32
33
34a
35
36
37a
38
39
40a
41
42
43
44a
45
46
47
corresponding parameters
g of RBF kernel need to be optimized.
ꢀ0.2
ꢀ0.28
0.13
0.21
0.69
0.17
0.28
0.77
ꢀ0.07
ꢀ0.06
ꢀ0.1
ꢀ0.1
ꢀ0.15
0.21
0.27
0.62
ꢀ0.2
ꢀ0.28
0.21
0.13
0.02
0.12
0.52
0.03
0.27
ꢀ0.3
ꢀ0.17
ꢀ0.14
0.69
0.17
0.28
0.77
0.21
ꢀ0.07
ꢀ0.06
ꢀ0.1
ꢀ0.1
The optimal parameters are found by grid search (GS) method.
Parameter lg2C, lg2 and lg2 was extensively investigated
g
3
between ꢀ10 and 10 with 1 as the increment. The result of this grid
search is an error-surface spanned by the model parameters. In
order to find the optimized combination of the parameters C,
, a process of leave-one-out (LOO) cross validation of training set
was performed as shown in Fig. 3. The best choices for C, and , are
3
, and
g
3
3
g
240
700
600
700
2400
17.5
42
2, 0.11 and 0.6, respectively, and the corresponding support vector
number is 18. The regression model developed was found to be
reliable statistically (R2train ¼ 0.948, Rt2est ¼ 0.907, Ro2verall ¼ 0.934,
R2cv ¼ 0.794) (Fig. 2F). Accordingly, the predicted Akt1 inhibitory
activities ꢀLog IC50 are listed in Table 1.
210
425
318
151
171
5.3
3.1
2790
4.8
3190
5.8
4.8
2.4
180
550
580
150
740
33
For example, both of compounds 17 and 45 were docked into
ATP-bound pocket of Akt1 using LigandFit program, and the bind-
ing modes of them were proposed as shown in Fig. 4. Two hydrogen
bonds are formed between 1H-pyrrolo[2,3-b]pyridine moiety of 17
(or aminopyrimidine moiety of 45) and Glu228 and Ala230 of Akt1.
The hydrophobic interaction was observed between 4- methox-
yphenyl ring of 17 (or biphenyl ring of 45) and Phe161 of Akt1. An
ionic interaction was observed between primary amino group of 17
and Glu292 of Akt1, while no ionic interaction was observed. For
quantitative analysis, the distances of the key interactions
mentioned above are recorded (Supporting information, Table S2).
As shown in Table 1, the superiority of docking scores (Dock_scores,
ligscores2, PMF, PLP1 and Jain) of 17 to that of 45 are not very
obvious, although 17 (ꢀLogIC50 ¼ 8.52) exhibited remarkably more
potent than 45 (-LogIC50 ¼ 5.40). Insight into the binding mode of
17 and 45 to Akt1, accordingly, 17 showed additional ionic inter-
action with Glu292 when comparing with that of 45, may due to
the much more bulky effect of biphenyl moiety of 45. Considering
the different profiles in key interaction analyses, together with five
docking scores and tow molecular descriptors, MD-SVR model can
accurately predict the potency of compounds 17 and 45, the devi-
ation for 17 and 45 are 0.28 and ꢀ0.06 respectively.
4000
150
25
a
The compounds were used as test set.
The ꢀLog IC50 was predicted by MD-SVR model.
b
3. Results and discussion
3.2. The prediction ability of MD-SVR model in design of novel Akt1
3.1. The development of MD-SVR model
inhibitors
The main objective of this work is to generate structure-based
quantitative models using a consensus of molecular docking
scores, key interactions profiles and molecular descriptors. Firstly,
the docking protocols have been validated by reproducing the
bound natural substrate conformation (PDB ID: 3OCB). The RMSD
value of proposed ligand from the reference coordinates is 0.24,
indicating the good accuracy and reliability of LigandFit program in
molecular docking of Akt1. Then, known Akt1 inhibitors 1e47 were
docked into ATP-bound pocket of Akt1 by LigandFit. A set of scoring
functions (i.e. Dock_score, LigScore2, PLP1, Jain and PMF) were
calculated (Supporting information, Table S2). Correlating docking
scores with Akt1 inhibitory activities was initially performed.
Feeble robustness of scoring functions toward prediction was
The aim of any QSAR modeling is that the developed model
should be strong enough to be capable of making accurate and
reliable predictions of biological activities of new compounds.
Hence, to examine predictive ability of the MD-SVR model, a set of
novel 4-aminopyrimidine derivatives 56a,b and 60aed were
designed and virtually evaluated by MD-SVR model. Firstly, com-
pounds 56a,b and 60aed were docked into ATP-bound pocket of
Akt1 using LigandFit program. Then, the five docking scores, key
interaction profiles and molecular descriptors were recorded
(Supporting information, Table S2). Based on MD-SVR model, the
predictive Akt1 inhibitory activities ꢀLog(IC50) of these compounds
were calculated and presented in Table 1. All of them showed good
prediction activities, especially for 60b and 60c, the ꢀLog(IC50
)

W. Zhan et al. / European Journal of Medicinal Chemistry 75 (2014) 11e20
15
Fig. 2. (A) Docking_score vs. experimental Akt inhibitory activities (ꢀLog IC50); (B) Ligscore2 vs. experimental Akt inhibitory activities (ꢀLog IC50); (C) ePLP vs. experimental Akt
inhibitory activities (ꢀLog IC50); (D) ePMF vs. experimental Akt inhibitory activities (ꢀLog IC50); (E) Jain vs. experimental Akt inhibitory activities (ꢀLog IC50); (F) Estimated vs.
experimental Akt inhibitory activities (ꢀLog IC50) of the MD-SVR model.
Fig. 3. Selection of g g
, C and 3 for training set data in MD-SVR model development. (A) C vs.Rc2v on LOO cross-validation. (B) vs.R2cv on LOO cross validation. (C) 3 vs.Rc2v on LOO cross
validation.
value of which are 7.88 and 7.93, respectively. As exemplified by
compound 60c, the most promising one, the interaction mode of
60c to Akt1 are similar to that of 17. As shown in Fig. 5, two
hydrogen bonds are formed between 4-aminopyrimidine moiety of
60c and Glu228 and Ala230 of Akt1. The hydrophobic interaction
was observed between 4-chlorophenyl ring of 60c and Phe161 of
Akt1. An additional ionic interaction was observed between pri-
mary amino group of 60c and Glu292 of Akt1.
Fig. 4. Interaction modes of 17 and 45 to Akt1 proposed by molecular docking. (A) Interaction mode between Akt1 and compound 17; (B) Interaction mode between Akt1 and
compound 45. Hydrogen bonds and ionic interactions were highlighted by dash green line and the hydrophobic interactions were highlighted by dash yellow line. (For inter-
pretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

16
W. Zhan et al. / European Journal of Medicinal Chemistry 75 (2014) 11e20
double alkylation of 3,4-dichlorophenylacetonitrile with 50 under
basic, phase-transfer conditions. Debenzylation of 51 under mild
conditionfurnished the intermediate 52 in excellent yield. Treatment
of 4,6-dichloropyrimidine 53 with either ethanolic methylamine (or
aqueous ammonia) afforded 4-amino-6-chloropyrimidines 54.
Treatment of compound 54 with 52 in the presence of K2CO3 in DMF
provides the corresponding 55. Then, target compound 56 was ob-
tained by reduction of the nitrile group of 55 in the presence of LiAlH4
and AlCl3 in THF.
On the other hand, The N-boc-piperazine and 4-amino-6-
chloropyrimidines 54b were condensed to give 57 in presence of
N,N-Diisopropylethylamine (DIPEA) in n-BuOH. Treatment of
compound 57 with NCS (or NBS) in CH2Cl2 gave 58. Deprotection of
N-Boc of 58 in acid condition was performed to give key in-
termediates 59. Finally, the preparation of the desired compound
60 was achieved by amide coupling, followed by N-Boc depro-
tection under acidic condition.
Fig. 5. Interaction mode between Akt1 and 60c proposed by molecular docking.
Hydrogen bonds and ionic interactions were highlighted by dash green line and the
hydrophobic interactions were highlighted by dash yellow line. (For interpretation of
the references to color in this figure legend, the reader is referred to the web version of
this article.)
3.4. Pharmacological assay
Akt1 inhibitory activities of 56a,b and 60aed were determined
using a competitive fluorescence polarization kinase activity assay.
Compound H-89 was used as positive control, the determined IC50
value of which is consistent with that of literature [31]. Table 2
presents the pharmacological assay results of 56a,b and 60aed.
All of them showed moderate to good Akt1 inhibitory activities.
Especially, compounds 60b and 60c showed the best Akt1 inhibi-
tory activities (IC50 of 60b and 60c were 17 nM and 7.7 nM).
Comparing with Akt1 inhibitory activities of 56a,b, the methylation
on 4-amino group of pyrimidine ring didn’t affect the Akt1
3.3. The synthesis of 4-aminopyrimidine derivatives 56a,b and
60aed
The synthesis of 4-aminopyrimidine derivatives 56a,b and 60ae
d is outlined in Scheme 2. On one hand, treatment of 48 with benzyl
bromide in the presence of K2CO3 in acetonitrile and subsequent
chlorinated with SOCl2 gave the N-protected bis(2-chloroethyl)
amine 50 as hydrochloride salt. Compound 51 was obtained by
Scheme 2. The synthetic route for compounds 56a,b and 60aed. Reaction conditions: a) Benzyl chloride, K2CO3, CH3CN; b) SOCl2, DCM; c) 2-(3,4-dichlorophenyl)acetonitrile,
tetrabutylammonium hydrogen sulfate, 50% NaOH (aq.); d) 2-chloropropanoyl chloride, DCE; e) MeOH, reflux; f) appropriate amine, i-PrOH; g) 52, DMF; h) LiAlH4, AlCl3, THF; i) N-
Boc-piperazine, DIPEA, n-BuOH; j) NBS or NCS, THF; k) TFA, DCM; l) appropriate N-Boc phenylalanine, EDCl, HOBt, DIPEA, DCM.

W. Zhan et al. / European Journal of Medicinal Chemistry 75 (2014) 11e20
17
Table 2
5. Experimental
The predictive and experimental Akt1 inhibitory and antiproliferative activities of
compounds 56a,b and 60aed.
5.1. Chemistry
Compd. Experimental
Predictive Antiproliferative activities (mM)
Melting points were obtained on a B-540 Büchi melting-point
apparatus and are uncorrected. 1H NMR spectra were recorded on
a 500 MHz, 13C NMR were recorded on a 125 or 100 MHz spec-
trometer (chemical shifts are given in ppm (d) relative to TMS as
internal standard, coupling constants (J) are in hertz (Hz), and
signals are designated as follows: s, singlet; d, doublet; t, triplet; m,
multiplet; br s, broad singlet, etc.). Mass spectral data were ob-
tained on an Esquire-LC-00075 spectrometer. High resolution mass
spectra were measured on an Agilent 1290 HPLC-6224 Time of
Fight Mass Spectrometer.
IC50
(
m
M) ꢀLog(IC50
)
ꢀLog(IC50
)
OVCAR-8
HCT116
56a
56b
60a
60b
60c
60d
H-89
1.83
4.32
1.83
5.70
5.36
5.74
7.77
8.11
6.36
/
6.85
7.18
7.59
7.88
7.93
7.85
/
N.T.a
N.T.a
N.T.a
26.07
22.67
N.T.a
10.50
16.70
19.84
28.69
4.36
5.15
6.19
0.017
0.0077
0.432
2.2
15.70
a
N.T. means no test.
5.1.1. N,N-Bis(2-hydroxyethyl)-N-benzylamine 49
inhibitory activities significantly. Insight into the observed sub-
stituent’s effects on 60aed revealed that the chlorine phenyl is
essential for Akt1 inhibitory activities. Compounds 60b,c exhibited
much potent (60- to 100-fold) than that of 60a,d. The substitution
(Cl or Br) on 5-pyrinmidine of 60 has a little effect on the Akt1
inhibitory activities, and bromine is slightly better. Moreover,
correlating predictive and experimental activities revealed that
moderate linear relation coefficient was observed (R2 ¼ 0.575),
indicating that the trend of them are the same. Except for 60a,
compounds 60bec with preferable predictive activities also
showed better experimental activities than compounds 56a,b.
Especially for compounds 60b and 60c, the prediction results are
very accurate that the derivation for them are ꢀ0.11 and 0.18,
respectively.
To a stirred solution of diethanolamine 48 (21 g, 200 mmol) and
anhydrous K2CO3 (55.2 g, 400 mmol) in dry acetonitrile (500 mL)
was added dropwise benzyl bromide (34.2 g, 200 mmol). The
mixture was stirred at 80 ꢂC for 10 h, after cooling, the potassium
salts were filtered off and the reaction mixture was concentrated.
The residue was extracted with ethyl acetate and washed with
water. The organic layer was further washed well with brine, dried
over anhydrous Na2SO4, and then concentrated under vacuum to
give the title compound (36.2 g, 93%) as a colorless oil, which was
characterized by LC-MS (m/z: 196 [M þ 1]þ) and then used directly
in the next step without further purification.
5.1.2. N,N-Bis(2-chloroethyl)benzylamine hydrochloride 50
To a solution of 49 (19.5 g, 0.1 mol) in dry CH2Cl2 (200 mL) at 0 ꢂC
was added SOCl2 (60 mL) in one portion. The mixture was stirred at
25 ꢂC for 2 h, then at 50 ꢂC for 1 h. After cooling, ethyl ether
(400 mL) was added, and then the precipitate was filtrated, rinsed
with ethyl ether and dried to afford the title compound (25 g, 94%)
as a white solid, mp: 148e150 ꢂC. 1H NMR (500 MHz, DMSO-d6)
The antiproliferative activities of compounds 56a,b and 60ae
d against cancer cell lines (OVCAR-8 and HCT116) was investi-
gated, since that Akt1 is an anti-apoptotic protein kinase and the
blockade of its activity leads to cell death. As expected, com-
pounds 60b and 60c showed good anti-proliferative activities
against HCT116 (60b: IC50 ¼ 4.36
more potent than that against OVCAR-8 (60b: IC50 ¼ 26.07
60c: IC50 ¼ 22.67 M) cell lines. In antiproliferative activities
against HCT116, three compounds 60bed with preferable Akt1
inhibitory activities (IC50 < 0.5 M) exhibited more potent
m
M; 60c: IC50 ¼ 5.15
m
M) and
m
M;
d
11.94 (s, 1H), 7.64 (s, 2H), 7.37 (d, J ¼ 15.4 Hz, 3H), 4.42 (s, 2H), 4.02
(s, 4H), 3.39 (s, 4H). ESI-MS (m/z): 232 [M þ 1]þ.
m
m
5.1.3. 1-Benzyl-4-(3,4-dichlorophenyl)piperidine-4-carbonitrile 51
To a well stirred suspension of 50 (5.37 g, 20 mmol), 2-(3,4-
dichlorophenyl)acetonitrile (4.1 g, 22 mmol) and tetrabuty-
lammonium hydrogen sulfate (0.68 g, 2 mmol) in toluene (15 mL)
was added dropwise 10 g of 50% (w/w) aqueous sodium hydroxide
solution. The reaction mixture was heated at 85 ꢂC for 4 h, then
poured into ice-water (50 mL) and extracted with ethyl ether
(25 mL ꢃ 3). The combined organic layers were dried over anhy-
drous Na2SO4, concentrated under vacuum. The residue was puri-
fied on silica gel to afford the title compound (4.3 g, 63%) as a yellow
antiproliferative activities (3- to 4-fold) than that of 56a,b and
60a.
4. Conclusion
In this study, we described an improved molecular docking
based QSAR model for Akt1 inhibitors. The MD-SVR model was
established using support vector regression (SVR) method. A
combination of docking scores, key interaction profiles and
molecular descriptors of the Akt1 inhibitors were considered in
QSAR modeling. The MD-SVR model was extensively validated
and showed good statistical parameters. Then, the application of
MD-SVR model in optimization of 4-aminopyrimidine de-
rivatives as novel Akt1 inhibitors was performed, and all of them
exhibited moderate to good Akt1 inhibitory activities. It was
demonstrated that the reliability and accuracy of the established
MD-SVR model is promising, since that moderate linear relation
coefficient was observed by correlating predictive and experi-
mental activities. To the best of our knowledge, the integration
of additional variables into molecular docking model for Akt1
inhibitors and its subsequent application in structure-based
drug design was reported for the first time. Moreover, com-
pound 60c, the most potent one (IC50 ¼ 7.7 nM), would be a
good lead to do further structure-activity research to find high
potent inhibitors.
oil. 1H NMR (500 MHz, CDCl3)
d
7.59 (d, J ¼ 2.3 Hz, 1H), 7.48e7.46
(m, 1H), 7.37e7.32 (m, 5H), 7.31e7.26 (m, 1H), 3.60 (s, 2H), 3.01 (d,
J ¼ 12.5 Hz, 2H), 2.56e2.42 (m, 2H), 2.09e2.05 (m, 4H). ESI-MS (m/
z): 345 [M þ 1]þ.
5.1.4. 4-(3,4-Dichlorophenyl)piperidine-4-carbonitrile
hydrochloride salt 52
To a solution of 51 (3.1 g, 9 mmol) in 1,2-dichloroethane (50 mL)
was added 2-chloroethyl chloroformate (1.5 g, 11 mmol), and the
mixture was heated to 100 ꢂC for 2 h. The mixture was cooled to
25 ꢂC and concentrated under reduced pressure. The crude residue
was dissolved in MeOH (50 mL) and heated to reflux for 3 h. The
mixture was cooled to room temperature, and concentrated under
reduced pressure to yield the title compound (2.35 g, 90%), as a
white solid, mp: ＞250 ꢂC. 1H NMR (500 MHz, DMSO-d6)
1H), 9.38 (s, 1H), 7.80e7.76 (m, 2H), 7.55 (dd, J ¼ 8.5, 2.4 Hz, 1H),
d 9.58 (s,

18
W. Zhan et al. / European Journal of Medicinal Chemistry 75 (2014) 11e20
3.48 (d, J ¼ 13.5 Hz, 2H), 3.12e3.00 (m, 2H), 2.47e2.41 (m, 4H). ESI-
MS (m/z): 255 [M þ 1]þ.
5.1.7.1. 6-(4-(Aminomethyl)-4-(3,4-dichlorophenyl)piperidin-1-yl)-
N-methylpyrimidin-4-amine 56a. Reagent: 55a (90 mg, 0.25 mmol).
The product was obtained as a white solid (61 mg, 67%), mp: 128e
130 ꢂC. 1H NMR (500 MHz, DMSO-d6)
d 8.02 (s, 1H), 7.82 (d,
5.1.5. General procedure for preparation of compound 54
To a suspension of 4,6-dichloropyrimidine 53 (3.0 g, 20 mmol) in
isopropanol (40 mL) was added appropriate amine at such a rate
that the internal temperature did not rise above 40 ꢂC. After
completion of the addition, the reaction mixture was stirred for 1 h
at 25 ꢂC. Water (30 mL) was added, and the resulting suspension
was cooled in an ice bath to 0 ꢂC. The precipitated product was
filtered off and washed with cold isopropanol/water (2:1, 50 mL)
and water. The collected material was dried in vacuo to afford the
title compounds.
J ¼ 2.2 Hz, 1H), 7.71 (d, J ¼ 8.5 Hz, 1H), 7.57 (dd, J ¼ 8.5, 2.3 Hz, 1H),
6.67 (d, J ¼ 4.7 Hz, 1H), 5.66 (s, 1H), 4.49 (s, 2H), 3.30 (d, J ¼ 11.7 Hz,
1H), 3.04 (t, J ¼ 12.3 Hz, 2H), 2.74 (d, J ¼ 4.6 Hz, 3H), 2.23 (d,
J ¼ 12.8 Hz, 2H), 2.00 (td, J ¼ 13.1, 3.8 Hz, 2H). 13C NMR (125 MHz,
DMSO-d6) d 164.5, 164.4, 162.2, 157.7, 141.2, 132.3, 131.6, 131.5, 128.6,
126.9,121.6, 42.7, 41.8, 34.8, 28.1. ESI-MS (m/z): 366 [M þ 1]þ. HRMS
(ESI): m/z calcd for (C17H21Cl2N5þH)þ: 366.1252; found: 366.1255.
5.1.7.2. 6-(4-(Aminomethyl)-4-(3,4-dichlorophenyl)piperidin-1-yl)
pyrimidin-4-amine 56b. Reagent: 55a (87 mg, 0.25 mmol) The
product was obtained as a white solid (55 mg, 62%), mp: 146e
5.1.5.1. 6-Chloro-N-methylpyrimidin-4-amine 54a. Reagent: 33%
methylamine in ethanol (20 mL, 165 mmol). The product was ob-
tained as a white solid (2.6 g, 91%), mp: 137e139 ꢂC. 1H NMR
148 ꢂC. 1H NMR (500 MHz, CDCl3)
d
8.18 (s, 1H), 7.47 (d, J ¼ 8.5 Hz,
1H), 7.43 (d, J ¼ 2.2 Hz,1H), 7.20 (dd, J ¼ 8.5, 2.3 Hz,1H), 5.58 (s, 1H),
4.52 (s, 2H), 3.88 (dt, J ¼ 10.7, 5.4 Hz, 2H), 3.22 (ddd, J ¼ 13.2, 10.0,
3.0 Hz, 2H), 2.82 (s, 2H), 2.19 (d, J ¼ 14.2 Hz, 2H), 1.83 (ddd, J ¼ 13.9,
10.0, 3.9 Hz, 2H). ESI-MS (m/z): 352 [Mþ1]þ. HRMS (ESI): m/z calcd
for (C16H19Cl2N5þH)þ: 352.1096; found: 352.1200.
(500 MHz, DMSO-d6)
d
8.23 (s, 1H), 7.67 (s, 1H), 6.49 (d, J ¼ 0.8 Hz,
1H), 2.77 (d, J ¼ 0.8 Hz, 3H). ESI-MS (m/z): 144 [M þ 1]þ.
5.1.5.2. 6-Chloropyrimidin-4-amine 54b. Reagent: ammonia solu-
tion (10 mL, 165 mmol). The product was obtained as a white solid
5.1.8. Tert-butyl 4-(6-aminopyrimidin-4-yl)piperazine-1-carboxylate
57
(2.6 g, 88%), mp: 205e207 ꢂC. 1H NMR (500 MHz, DMSO-d6)
(s, 1H), 7.22 (s, 2H), 6.43 (s, 1H). ESI-MS (m/z): 130 [M þ 1]þ.
d 8.19
A solution of 54b (1.3 g, 10 mmol), N-Boc-piperazine (2.0 g,
11 mmol), and N,N-diisopropylethylamine (3.9 g, 30 mmol) in n-
BuOH (50 mL) was heated to refluxing under a nitrogen atmo-
sphere for 10 h. The reaction mixture was concentrated under
vacuum, and the crude residue was purified by silica gel chroma-
tography to afford the title compound (2.15 g, 77%) as a white solid,
5.1.6. General procedure for preparation of compound 55
To a solution of 52(290 mg, 1 mmol) in DMF (15 mL) was added
54 and K2CO3 (414 mg, 3 mmol) and the reaction mixture was heat
to 100 ꢂC in a sealed tube for 5 h. Then, the reaction solution was
poured into water (30 mL) and extracted with ethyl acetate
(25 mL ꢃ 3). The combined organic layers were dried over Na2SO4,
concentrated under vacuum and purified on silica gel to afford the
title compounds.
mp: 187e189 ꢂC. 1H NMR (500 MHz, CDCl3)
d 8.18 (s, 1H), 5.56 (s,
1H), 4.66 (s, 2H), 3.54 (s, 4H), 3.50 (s, 4H), 1.48 (s, 9H). ESI-MS (m/z):
280 [M þ 1]þ.
5.1.9. General procedure for preparation of compound 58
A mixture of 57 (1.4 g, 5 mmol) and NCS or NBS in dry THF
(30 mL) was stirred at 50 ꢂC for 5 h. The reaction mixture was
concentrated and the crude residue was recrystallized from ethyl
acetate to give the title compound.
5.1.6.1. 4-(3,4-Dichlorophenyl)-1-(6-(methylamino)pyrimidin-4-yl)
piperidine-4-carbonitrile 55a. Reagent: 54a (172 mg, 1.2 mmol). The
product was obtained as a white solid (188 mg, 52%), mp: 181e
183 ꢂC. 1H NMR (500 MHz, CDCl3)
d
8.19 (s, 1H), 7.58 (d, J ¼ 2.3 Hz,
1H), 7.51 (d, J ¼ 8.5 Hz, 1H), 7.35 (dd, J ¼ 8.5, 2.3 Hz, 1H), 5.48 (s, 1H),
4.95 (s, 1H), 4.57 (d, J ¼ 14.0 Hz, 2H), 3.36e3.29 (m, 2H), 2.92 (d,
J ¼ 5.2 Hz, 3H), 2.19 (d, J ¼ 11.6 Hz, 2H), 2.00 (td, J ¼ 13.2, 4.1 Hz, 2H).
ESI-MS (m/z): 362 [M þ 1]þ.
5.1.9.1. Tert-butyl 4-(6-amino-5-chloropyrimidin-4-yl)piperazine-1-
carboxylate 58a. Reagent: NCS (0.8 g, 6 mmol). The product was
obtained as a white solid (1.3 g, 81%), mp: 149e151 ꢂC. 1H NMR
(500 MHz, CDCl3) d 8.10 (s, 1H), 5.28 (s, 2H), 3.53 (s, 8H), 1.47 (s, 9H).
ESI-MS (m/z): 314 [M þ 1]þ.
5.1.6.2. 1-(6-Aminopyrimidin-4-yl)-4-(3,4-dichlorophenyl)piperi-
dine-4-carbonitrile 55b. Reagent: 54a (155 mg, 1.2 mmol). The
product was obtained as a white solid (167 mg, 48%), mp: 195e
5.1.9.2. Tert-butyl 4-(6-amino-5-bromopyrimidin-4-yl)piperazine-1-
carboxylate 58b. Reagent: NBS (1.1 g, 6 mmol). The product was
obtained as white solid (1.55 g, 87%), mp: 172e174 ꢂC. 1H NMR
198 ꢂC. 1H NMR (500 MHz, CDCl3)
d
8.23 (s, J ¼ 0.7 Hz, 1H), 7.57 (d,
J ¼ 2.3 Hz, 1H), 7.50 (dd, J ¼ 11.0, 6.7 Hz, 1H), 7.34 (dd, J ¼ 8.5, 2.4 Hz,
1H), 5.67 (d, J ¼ 0.8 Hz, 1H), 4.70 (s, 2H), 4.53 (d, J ¼ 13.9 Hz, 2H),
3.35e3.28 (m, 2H), 2.21e2.15 (m, 2H), 1.99 (td, J ¼ 13.4, 4.2 Hz, 2H).
ESI-MS (m/z): 348 [M þ 1]þ.
(500 MHz, CDCl3) d 8.13 (s, 1H), 5.35 (s, 2H), 3.57e3.44 (m, 8H), 1.48
(s, 9H). ESI-MS (m/z): 358 [M þ 1]þ.
5.1.10. General procedure for preparation of compound 59
To a solution of 58 in CH2Cl2 (15 mL) was added trifluoroacetic
acid (3 mL) under ice-cooling. The reaction mixture was stirred at
25 ꢂC for 3 h and concentrated in vacuo. The residue was treated
with 1 N aqueous solution of NaOH and extracted with CH2Cl2. The
extract was dried over anhydrous Na2SO4 and concentrated in
vacuo to give the title compound.
5.1.7. General procedure for preparation of compound 56
A suspension of AlCl3 (66.5 mg, 0.5 mmol) in dry THF (10 mL)
was added to a suspension of LiAlH4 (20 mg, 0.53 mmol) in THF
(10 mL) at 0 ꢂC and this mixture was stirred for 10 min followed by
the dropwise addition of a solution of 55 in THF (5 mL). The reaction
mixture was allowed to warm to room temperature, stirred over-
night, cooled with an ice bath, and quenched with a saturated so-
lution of Na2CO3 in water until the foaming stopped. Subsequently,
the suspension was filtered, and the filtrate was dried over Na2SO4,
concentrated under vacuum and purified on silica gel to afford the
title compounds.
5.1.10.1. 5-Chloro-6-(piperazin-1-yl)pyrimidin-4-amine
59a.
Reagent: 58a (1.56 g, 5 mmol). The product was obtained as pale
yellow powder (1.0 g, 95%), mp: 167e169 ꢂC. 1H NMR (500 MHz,
DMSO-d6)
d 7.96 (s, 1H), 6.80 (s, 2H), 3.36e3.31 (m, 4H), 2.83e2.79
(m, 4H). ESI-MS (m/z): 214 [M þ 1]þ.

W. Zhan et al. / European Journal of Medicinal Chemistry 75 (2014) 11e20
19
5.1.10.2. 5-Bromo-6-(piperazin-1-yl)pyrimidin-4-amine
Reagent: 58b (1.79 g, 5 mmol). The product was obtained as a white
solid (1.25, 97%), mp: 163e165 ꢂC. 1H NMR (500 MHz, CDCl3)
8.13
59b.
butoxycarbonyl)amino)-3-phenylpropanoic acid (132 mg, 0.5 mmol)
and 59b (129 mg, 0.5 mmol). The product was obtained as a white
solid (119 mg, 59%), mp: 143e145 ꢂC. 1H NMR (500 MHz, CDCl3)
d
(s, 1H), 5.23 (s, 2H), 3.49e3.45 (m, 4H), 3.00e2.97 (m, 4H). ESI-MS
d 8.08 (s, 1H), 7.32e7.27 (m, 2H), 7.24e7.17 (m, 3H), 5.50 (s, 2H), 3.97
(m/z): 258 [M þ 1]þ.
(t, J ¼ 7.2 Hz,1H), 3.75 (m,1H), 3.60 (m,1H), 3.51e3.40 (m, 2H), 3.39e
3.31 (m, 1H), 3.32e3.24 (m, 1H), 3.23e3.15 (m, 1H), 2.98e2.90 (m,
5.1.11. General procedure for synthesis of compound 60
1H), 2.88e2.81 (m, 2H). 13C NMR (125 MHz, CDCl3)
d 173.4, 162.8,
A
mixture of N-Boc-phenylalanine, 1-hydroxybenzotriazole
161.5, 155.4, 137.4, 129.4, 128.7, 127.0, 89.0, 52.5, 48.2, 47.9, 45.0, 43.0,
41.8. ESI-MS (m/z): 405 [M þ 1]þ. HRMS (ESI): m/z calcd for
(C17H21BrN6O þ H)þ: 405.1038; found: 405.1040.
(91 mg, 0.6 mmol), 1-ethyl-3-(3-dimethyllaminopropyl)carbodii-
mide hydrochloride (115 mg, 0.6 mmol) and N,N-
diisopropylethylamine (129 mg, 1 mmol) in dry CH2Cl2 (10 mL) at
0
ꢂC was added compound 59. After stirring for 15 min, the cold
5.2. Pharmacological assay
bath was removed, and the reaction mixture was stirred at 25 ꢂC
overnight and concentrated in vacuo. The resulting residue was
added H2O (10 mL) and extracted with EtOAc, the combined
organic layers were dried over Na2SO4, concentrated to afford a
white solid. To a solution of the above solid in CH2Cl2 (7 mL) was
added trifluoroacetic acid (1 mL) under ice-cooling. The reaction
mixture was stirred at 25 ꢂC for 2 h and concentrated in vacuo. The
residue was treated with 1 N aqueous solution of NaOH and
extracted with CH2Cl2. The extract was dried over anhydrous
Na2SO4 and purified on silica gel to afford the title compound.
5.2.1. Akt1 assay
AKT1/PKB
a Kinase Assay Kit (catalog No. 32-021) and active
recombinant AKT1 kinase were commercially available from UP-
STATE (acquired by Merck Millipore, Billerica, MA). 4 ng of Akt1 in
8
m
L of 2.5 ꢃ kinase buffer [62.5 mM TriseHCl (pH 7.5), 25 mM
MgCl2, 12.5 mM -glycerophosphate, 0.25 mM Na3VO4, 5 mM
dithiothreitol (DTT)], was mixed with 2 L of dimethyl sulfoxide
(DMSO) vehicle or each of the compound (indicated concentra-
tions), incubated at room temperature for 5 min and 10 L of ATP/
substrate cocktail (20 mM ATP, 3 mM eNOS served as substrate) was
added. After incubation at room temperature for 30 min, add 20
b
m
m
5.1.11.1. (S)-2-Amino-1-(4-(6-amino-5-chloropyrimidin-4-yl)piper-
azin-1-yl)-3-phenylpropan-1-one 60a. Reagent: (S)-2-((tert-butox-
ycarbonyl)amino)-3-phenylpropanoic acid (132 mg, 0.5 mmol) and
59a (106.7 mg, 0.5 mmol). The product was obtained as a white
solid (95 mg, 53%), mp: 129e132 ꢂC. 1H NMR (500 MHz, CDCl3)
mL
of 50 mM ethylenediaminetetraacetic acid (EDTA) (pH 8.0) and
terminate the reaction. Then, PKB/Akt1 kinase activity was
analyzed according to the manufacturer’s instructions.
d
8.04 (s, 1H), 7.32e7.26 (m, 2H), 7.24e7.16 (m, 3H), 5.49 (s, 2H), 4.10
5.2.2. Cytotoxic activity assay. The cytotoxic activity of the tested
compounds in OVCAR-8 and HCT116 cells was measured using the
MTT (3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bro-
mide, a tetrazole) method. Cells were seeded in 96-well microtiter
plates (at a density of 4000 cells per well) for overnight attachment
(t, J ¼ 7.2 Hz, 1H), 3.74e3.67 (m, 1H), 3.64e3.55 (m, 1H), 3.54e3.43
(m, 2H), 3.42e3.34 (m, 2H), 3.34e3.27 (m, 1H), 2.99e2.89 (m, 2H),
2.89e2.81 (m, 1H). 13C NMR (125 MHz, CDCl3)
d
172.9, 160.5, 160.5,
154.4, 136.7, 129.4, 128.8, 127.2, 98.0, 52.2, 47.3, 47.1, 45.6, 45.1, 42.0.
ESI-MS (m/z): 361 [M
þ
1]þ. HRMS (ESI): m/z calcd for
and exposed to each of the test compound (1.0e100.0
The MTT solution (5.0 mg/mL in RPIM 1640 medium; Sigmae
Aldrich) was added (20.0 l/well), and plates were incubated for a
further 4 h at 37 ꢂC. The purple formazan crystals were dissolved in
100.0 L of DMSO. After 5 min, the plates were read on an auto-
mM) for 72 h.
(C17H21ClN6O þ H)þ: 361.1544; found: 361.1539.
m
5.1.11.2. (ꢄ)-2-Amino-1-(4-(6-amino-5-chloropyrimidin-4-yl)piper-
azin-1-yl)-3-(4-chlorophenyl) propan-1-one 60b. Reagent: 2-((tert-
m
butoxycarbonyl)amino)-3-(4-chlorophenyl)propanoic
acid
mated microplate spectrophotometer (Bio-Tek Instruments,
Winooski, VT) at 570 nm. Assays were performed in triplicate on
three independent experiments. The concentration of drug inhib-
iting 50% of cells (IC50) was calculated using the software of dose-
effect analysis.
(150 mg, 0.5 mmol) and 59a (106.7 mg, 0.5 mmol). The product was
obtained as a white solid (92.8 mg, 47%), mp: 151e153 ꢂC. 1H NMR
(500 MHz, CDCl3)
d 8.09 (s, 1H), 7.28e7.26 (m, 2H), 7.15e7.12 (m,
2H), 5.30 (s, 2H), 3.96e3.88 (m, 1H), 3.74e3.64 (m, 2H), 3.56e3.37
(m, 4H), 3.30e3.09 (m, 2H), 2.97e2.88 (m, 1H), 2.78 (dd, J ¼ 13.4,
7.2 Hz, 1H). 13C NMR (125 MHz, CDCl3)
d 173.2, 160.7, 160.4, 154.6,
Acknowledgment
136.0, 132.8, 130.7, 128.8, 98.3, 52.4, 47.7, 47.5, 45.1, 42.1, 41.9. ESI-
MS (m/z): 395 [M
þ
1]þ. HRMS (ESI): m/z calcd for
We thank Mrs. Jianyang Pan (Pharmaceutical Informatics Insti-
tute, Zhejiang University) for performing the NMR spectrometry.
This study was financially supported by National Natural Science
Foundation of China (No. 81001359), Medical Scientific Research
Foundation of Zhejiang Province (NO. 2010KYA061), Zhejiang Pro-
vincial Natural Science Foundation of China (LY13H300002) and
Program for Zhejiang Leading Team of S&T Innovation.
(C17H20Cl2N6O þ H)þ: 395.1154; found: 395.1158.
5.1.11.3. (ꢄ)-2-Amino-1-(4-(6-amino-5-bromopyrimidin-4-yl)piper-
azin-1-yl)-3-(4-chlorophenyl) propan-1-one 60c. Reagent: 2-((Tert-
butoxycarbonyl)amino)-3-(4-chlorophenyl)propanoic
acid
(150 mg, 0.5 mmol) and 59b (129 mg, 0.5 mmol). The product was
obtained as a white solid (138.5 mg, 63%), mp: 166e168 ꢂC. 1H NMR
(500 MHz, CDCl3)
d
8.13 (s, 1H), 7.28 (d, J ¼ 8.3 Hz, 2H), 7.15 (d,
Appendix A. Supplementary data
J ¼ 8.3 Hz, 2H), 5.27 (s, 2H), 3.93 (t, J ¼ 7.1 Hz, 1H), 3.72 (t, J ¼ 5.1 Hz,
2H), 3.56e3.44 (m, 2H), 3.43e3.35 (m, 2H), 3.31e3.26 (m, 1H),
3.15e3.07 (m, 1H), 2.94 (dd, J ¼ 13.4, 6.9 Hz, 1H), 2.79 (dd, J ¼ 13.4,
Supplementary data related to this article can be found at http://
dx.doi.org/10.1016/j.ejmech.2014.01.019.
7.2 Hz, 1H). 13C NMR (125 MHz, CDCl3)
d 173.3, 162.9, 161.5, 155.5,
136.0, 132.8, 130.7, 128.8, 89.3, 52.4, 48.3, 48.1, 45.1, 42.1, 41.9. ESI-
References
MS (m/z): 439 [M
þ
1]þ. HRMS (ESI): m/z calcd for
(C17H20BrClN6O þ H)þ: 439.0649; found: 439.0652.
[1] V. Vapnik, C. Cortes, Support vector networks, Mach. Learn. 20 (1995) 273e
297.
[2] J. Levatic, J. Curak, M. Kralj, T. Smuc, M. Osmak, F. Supek, Accurate models for
P-gp drug recognition induced from a cancer cell line cytotoxicity screen,
J. Med. Chem. 56 (2013) 5691e5708.
5.1.11.4. (ꢄ)-2-Amino-1-(4-(6-amino-5-bromopyrimidin-4-yl)piper-
azin-1-yl)-3-phenylpropan-1-one
60d. Reagent:
(S)-2-((tert-

20
W. Zhan et al. / European Journal of Medicinal Chemistry 75 (2014) 11e20
[3] R. Varela, W.P. Walters, B.B. Goldman, A.N. Jain, Iterative refinement of a
G.P. Vigers, B.J. Brandhuber, J. Liang, B.S. Safina, J. Li, B. Zhang, C. Chabot, S. Do,
L. Lee, J. Oeh, D. Sampath, B.B. Lee, K. Lin, B.M. Liederer, N.J. Skelton, Discovery
and preclinical pharmacology of a selective ATP-competitive Akt inhibitor
(GDC-0068) for the treatment of human tumors, J. Med. Chem. 55 (2012)
8110e8127.
binding pocket model: active computational steering of lead optimization,
J. Med. Chem. 55 (2012) 8926e8942.
[4] W. Wang, W. He, X. Zhou, X. Chen, Optimization of molecular docking scores
with support vector rank regression, Proteins 81 (2013) 1386e1398.
[5] P.J. Ballester, J.B. Mitchell, A machine learning approach to predicting proteine
ligand binding affinity with applications to molecular docking, Bioinformatics
26 (2010) 1169e1175.
[6] S. Li, L. Xi, C. Wang, J. Li, B. Lei, H. Liu, X. Yao, A novel method for proteine
ligand binding affinity prediction and the related descriptors exploration,
J. Comput. Chem. 30 (2009) 900e909.
[7] S. Das, M.P. Krein, C.M. Breneman, Binding affinity prediction with property-
encoded shape distribution signatures, J. Chem. Inf. Model. 50 (2010) 298e
308.
[8] S.P. Staal, Molecular cloning of the akt oncogene and its human homologues
AKT1 and AKT2: amplification of AKT1 in a primary human gastric adeno-
carcinoma, Proc. Natl. Acad. Sci. U. S. A. 84 (1987) 5034e5037.
[9] J.Q. Cheng, B. Ruggeri, W.M. Klein, G. Sonoda, D.A. Altomare, D.K. Watson,
J.R. Testa, Amplification of AKT2 in human pancreatic cancer cells and inhi-
bition of AKT2 expression and tumorigenicity by antisense RNA, Proc. Natl.
Acad. Sci. U. S. A. 93 (1996) 3636e3641.
[10] J.Q. Cheng, C.W. Lindsley, G.Z. Cheng, H. Yang, S.V. Nicosia, The Akt/PKB
pathway: molecular target for cancer drug discovery, Oncogene 24 (2005)
7482e7492.
[11] R.T. Joseph, B. Alfonso, AKT plays a central role in tumorigenesis, Proc. Natl.
Acad. Sci. U. S. A. 98 (2001) 10983e10985.
[12] L.C. Cantley, The phosphoinositide 3-kinase pathway, Science 296 (2002)
1655e1657.
[13] B.D. Manning, L.C. Cantley, AKT/PKB signaling: navigating downstream, Cell
129 (2007) 1261e1274.
[19] W.R. Chao, D. Yean, K. Amin, C. Green, L. Jong, Computer-aided rational drug
design: a novel agent (SR13668) designed to mimic the unique anticancer
mechanisms of dietary indole-3-carbinol to block Akt signaling, J. Med. Chem.
50 (2007) 3412e3415.
[20] J.G. Kettle, S. Brown, C. Crafter, B.R. Davies, P. Dudley, G. Fairley, P. Faulder,
S. Fillery, H. Greenwood, J. Hawkins, M. James, K. Johnson, C.D. Lane, M. Pass,
J.H. Pink, H. Plant, S.C. Cosulich, Diverse heterocyclic scaffolds as allosteric
inhibitors of AKT, J. Med. Chem. 55 (2012) 1261e1273.
[21] S.B. Kondapaka, S.S. Singh, G.P. Dasmahapatra, E.A. Sausville, K.K. Roy, Peri-
fosine, a novel alkylphospholipid, inhibits protein kinase B activation, Mol.
Cancer Ther. 2 (2003) 1093e1103.
[22] X. Dong, C. Jiang, H. Hu, J. Yan, J. Chen, Y. Hu, QSAR study of Akt/protein kinase
B (PKB) inhibitors using support vector machine, Eur. J. Med. Chem. 44 (2009)
4090e4097.
[23] X. Dong, X. Zhou, H. Jing, J. Chen, T. Liu, B. Yang, Q. He, Y. Hu, Pharmacophore
identification, virtual screening and biological evaluation of prenylated fla-
vonoids derivatives as PKB/Akt1 inhibitors, Eur. J. Med. Chem. 46 (2011)
5949e5958.
[24] M.A. Seefeld, M.B. Rouse, K.C. McNulty, L. Sun, J. Wang, D.S. Yamashita,
J.I. Luengo, S. Zhang, E.A. Minthorn, N.O. Concha, D.A. Heerding, Discovery of
5-pyrrolopyridinyl-2-thiophenecarboxamides as potent AKT kinase in-
hibitors, Bioorg. Med. Chem. Lett. 19 (2009) 2244e2248.
[25] K.D. Freeman-Cook, C. Autry, G. Borzillo, D. Gordon, E. Barbacci-Tobin,
V. Bernardo, D. Briere, T. Clark, M. Corbett, J. Jakubczak, S. Kakar, E. Knauth,
B. Lippa, M.J. Luzzio, M. Mansour, G. Martinelli, M. Marx, K. Nelson, J. Pandit,
F. Rajamohan, S. Robinson, C. Subramanyam, L. Wei, M. Wythes, J. Morris,
Design of selective, ATP-competitive inhibitors of Akt, J. Med. Chem. 53 (2010)
4615e4622.
[26] B. Lippa, G. Pan, M. Corbett, C. Li, G.S. Kauffman, J. Pandit, S. Robinson, L. Wei,
E. Kozina, E.S. Marr, G. Borzillo, E. Knauth, E.G. Barbacci-Tobin, P. Vincent,
M. Troutman, D. Baker, F. Rajamohan, S. Kakar, T. Clark, J. Morris, Synthesis
and structure based optimization of novel Akt inhibitors, Bioorg. Med. Chem.
Lett. 18 (2008) 3359e3363.
[14] L. Salmena, A. Carracedo, P.P. Pandolfi, Tenets of PTEN tumor suppression, Cell
133 (2008) 403e414.
[15] A. Carnero, C. Blanco-Aparicio, O. Renner, W. Link, J.F. Leal, The PTEN/PI3K/
AKT signalling pathway in cancer, therapeutic implications, Curr. Cancer Drug
Targets 8 (2008) 187e198.
[16] M. Addie, P. Ballard, D. Buttar, C. Crafter, G. Currie, B.R. Davies, J. Debreczeni,
H. Dry, P. Dudley, R. Greenwood, P.D. Johnson, J.G. Kettle, C. Lane, G. Lamont,
A. Leach, R.W. Luke, J. Morris, D. Ogilvie, K. Page, M. Pass, S. Pearson, L. Ruston,
Discovery of 4-amino-N-[(1S)-1-(4-chlorophenyl)-3-hydroxypropyl]-1-(7H-
pyrrolo[2,3-d]pyrimidin -4-yl)piperidine-4-carboxamide (AZD5363), an orally
bioavailable, potent inhibitor of Akt kinases, J. Med. Chem. 56 (2013) 2059e
2073.
[17] M.A. Ashwell, J.M. Lapierre, C. Brassard, K. Bresciano, C. Bull, S. Cornell-Ken-
non, S. Eathiraj, D.S. France, T. Hall, J. Hill, E. Kelleher, S. Khanapurkar, D. Kizer,
S. Koerner, J. Link, Y. Liu, S. Makhija, M. Moussa, N. Namdev, K. Nguyen,
R. Nicewonger, R. Palma, J. Szwaya, M. Tandon, U. Uppalapati, D. Vensel,
L.P. Volak, E. Volckova, N. Westlund, H. Wu, R.Y. Yang, T.C. Chan, Discovery
and optimization of a series of 3-(3-phenyl-3H-imidazo[4,5-b]pyridin-2-yl)
pyridin-2-amines: orally bioavailable, selective, and potent ATP-independent
Akt inhibitors, J. Med. Chem. 55 (2012) 5291e5310.
[27] S. Chang, Z. Zhang, X. Zhuang, J. Luo, X. Cao, H. Li, Z. Tu, X. Lu, X. Ren, K. Ding,
New thiazole carboxamides as potent inhibitors of Akt kinases, Bioorg Med.
Chem. Lett. 22 (2012) 1208e1212.
[28] B.R. Brooks, R.E. Brucolleri, B.D. Olafson, D.J. States, S. Swaminathan,
M. Karplus, CHARMM: a program for macromolecular energy, minimization,
and dynamics calculations, J. Comput. Chem. 4 (1983) 187e214.
[29] C.M. Venkatachalam, X. Jiang, T. Oldfield, M. Waldman, LigandFit: a novel
method for the shape-directed rapid docking of ligands to protein active sites,
J. Mol. Graphics Modell. 21 (2003) 289e307.
[30] M. Montes, M.A. Miteva, B.O. Villoutreix, Structure-based virtual ligand
screening with LigandFit: pose prediction and enrichment of compound col-
lections, Proteins 68 (2007) 712e725.
[18] J.F. Blake, R. Xu, J.R. Bencsik, D. Xiao, N.C. Kallan, S. Schlachter, I.S. Mitchell,
K.L. Spencer, A.L. Banka, E.M. Wallace, S.L. Gloor, M. Martinson, R.D. Woessner,
[31] C.C. Kumar, V. Madison, Akt crystal structure and Akt-specific inhibitors,
Oncogene 24 (2005) 7493e7501.