Essential Structural Profile of Novel Adenosine Derivatives as Antiplatelet Aggregation Inhibitors Based on 3D-QSAR Analysis Using CoMFA, CoMSIA, and SOMFA

Article abstract of DOI:10.1134/S1068162020030103

Abstact: —In this study, comparative molecular field analysis (CoMFA), comparative molecular similarity indices analysis (CoMSIA), and the self-organizing molecular field analysis (SOMFA) were performed on a series of novel adenosine derivatives. Significant correlation coefficients (CoMFA, q² = 0.560, r² = 0.940, F value = 71.850, and SEE = 0.097; CoMSIA, q² = 0.528, r² = 0.943, F value = 29.29 and SEE = 0.108; SOMFA, r² = 0.615, r²_cr= 0.577, F value = 60.797, and SEE = 0.226) were obtained, and the generated models were validated using test sets. By analyzing the corresponding contour maps in detail, new adenosine derivatives with potential efficacy were designed for synthesis in the future.

Full text of DOI:10.1134/S1068162020030103

ISSN 1068-1620, Russian Journal of Bioorganic Chemistry, 2020, Vol. 46, No. 3, pp. 448–457. © Pleiades Publishing, Ltd., 2020.
LETTER TO THE EDITOR
Essential Structural Profile of Novel Adenosine Derivatives
as Antiplatelet Aggregation Inhibitors Based on 3D-QSAR Analysis
Using CoMFA, CoMSIA, and SOMFA
Shunlai Lia, XueFeng Baoa, Chenghu Lua, Chaorui Rena, Guocheng Liua, and Hongguang Dua, 1
aCollege of Science, Beijing University of Chemical Technology, Chaoyang District, Beijing, 100029 China
Received June 11, 2019; revised August 23, 2019; accepted December 29, 2019
Abstact—In this study, comparative molecular field analysis (CoMFA), comparative molecular similarity
indices analysis (CoMSIA), and the self-organizing molecular field analysis (SOMFA) were performed on a
series of novel adenosine derivatives. Significant correlation coefficients (CoMFA, q2 = 0.560, r2 = 0.940, F
value = 71.850, and SEE = 0.097; CoMSIA, q2 = 0.528, r2 = 0.943, F value = 29.29 and SEE = 0.108;
2
SOMFA, r2 = 0.615, = 0.577, F value = 60.797, and SEE = 0.226) were obtained, and the generated mod-
rcv
els were validated using test sets. By analyzing the corresponding contour maps in detail, new adenosine
derivatives with potential efficacy were designed for synthesis in the future.
Keywords: adenosine derivatives, design, antiplatelet aggregation activity, 3D-QSAR, CoMFA
DOI: 10.1134/S1068162020030103
produce activity without the metabolism, better con-
trolling the bleeding time [10]. However, for the mod-
erately high bleeding, it is worth to further optimize its
chemical structure and improve its activity. According
to the pharmacodynamic action, it is necessary and
urgent to further develop antiplatelet drugs with better
curative effect and lower side-effects for the treatment
of stroke, coronary heart disease, and thrombosis.
In order to dig out helpful information in designing
more highly active and safe drugs, the analysis of
quantitative structure activity relationship (QSAR)
[11–15], which predicts biological activity based on
chemical structures, could provide important basic
information. Three-dimensional quantitative struc-
ture activity relationship (3D-QSAR)analysis meth-
ods, such as comparative molecular field analysis
(CoMFA) [16], comparative molecular similarity
indices analysis (CoMSIA) [17], and the self-organiz-
ing molecular field analysis (SOMFA) [18] are well-
known methods in analysis of compounds.
The leading cause of stroke [1], coronary heart dis-
ease (CHD) [2, 3], is a huge threat to public health
worldwide, and antiplatelet drugs (antithrombotic
drugs) produce remarkable treatment effects in this
field [4]. Some antiplatelet drugs have been discovered
and applied in clinical treatment, such as aspirin,
clopidogrel, ticlopidine, platelet fibrinogen receptor
antagonists, and ticagrelor [5]. However, some causes,
such as the resistance of aspirin and clopidogrel, big-
ger side effect of bleeding, and high cost of fibrinogen
receptor antagonists, hindered the curative effects of
antiplatelet drugs [6, 7]. Under the previous guidance
and experience, some novel drugs need to be further
studied to make up the inadequacy of current drugs.
As approved for patients with acute coronary syn-
drome (ACS) and myocardial infarction (MI) in 2012,
ticagrelor (the structure shown in Fig. 1) [8] hinders
ADP binding to P2Y12 receptors in a direct and revers-
ible manner, resulting in faster work, migration pro-
files and genetic polymorphisms induce low pharma-
cokinetic changes. Although it is effective in reducing
the mortality of cardiovascular patients, a non-fatal
higher bleeding rate marked a significant adverse side
effect of ticagrelor [8]. It is noteworthy that cangrelor
(as shown in Fig. 1) [9], which was developed in 1999,
has a significant effect on antiplatelet aggregation and
a higher antiplatelet activity than clopidogrel. Mean-
while, it can rapidly inhibit platelet aggregation and
A series of N6-alkyl(aryl)-2-alkyl(aryl)thio-ade-
nosines have been synthesized and their antiplatelet
activity has been evaluated by our research group [19].
The structures and activities described by IC50 (μM)
are shown in Table 1, and the synthetic route is shown
in Scheme 1. In this paper, 3D-QSAR analysis on
these adenosine compounds was carried out by
CoMFA, CoMSIA, and SOMFA methods. According
to the final 3D-QSAR analysis, some structural help-
ful information in enhancing the platelet inhibitory
activity was obtained and twenty novel adenosine
1 Corresponding author: phone/fax: +86-10-64439218; e-mail:
dhg@mail.buct.edu.cn.
448

ESSENTIAL STRUCTURAL PROFILE
449
NH2
N
F
N
HN
N
F
N
N
N
N
N
S
O
P
O
P
O
P
N
O
HO
O
O
O
S
O
H
H
O
O
O
H
H
OH OH
OH OH
Ticagrelor
Cangrelor
Fig. 1. Structures of ticagrelor and cangrelor.
derivatives with potential antiplatelet efficacy were training set contained 40 compounds (80%) and the
designed for synthesis in future.
test set contained 5 compounds (20%). The 50%
inhibiting concentration IC50 values were translated
into the PAC values using the equation PAC = IC50 ×
50/300. Lower PAC value indicated greater inhibitory
activity. Then PAC values were translated into
The data set consisted of 45 adenosine derivatives,
which were taken from our research paper [19], was
subjected to 3D-QSAR analysis using CoMFA, CoM-
SIA, and SOMFA. The forty five compounds with
experimental data were divided randomly into training
set and test set in a ratio of about 8 : 1, in which the log(PAC) = log((1/PAC)/(300 × 10–6)).
Cl
N
O
N
O
N
N
N
N
N
N
N
N
NH
NH2
NH
NH2
a
b
NH2
AcO
HO
AcO
O
O
O
OAc OAc
Cl
OAc OAc
NR1R2
OH OH
N
N
N
N
N
N
c
d
N
SR
N
SR
AcO
HO
O
O
OAc OAc
R = alkyl, aryl; R1 = alkyl, aryl; R2 = H or R1
OH OH
Scheme 1. Synthesis of N6-alkyl(aryl)-2-alkyl(aryl)thioadenosines. Reagents and conditions: (a) Ac2O, DMAP, Et3N,
CH3CN, r.t.; (b) POCl3, Et4NCl, N,N-dimethylaniline, CH3CN, reflux; (c) isoamyl nitrite, MeCN, RSSR, 60°C; (d)
1) HNR1R2, Et3N, EtOH, reflux; 2) Na, reflux.
All 3D chemical structures of the 45 compounds The third superposition, which was based on the best
conformation in the receptor (alignment C), is dis-
were drawn and energy-minimized in SYBYLx1.3 [20]
using the MMFF94s force field with Delre charges
[21]. The alignment of molecules is a very consider-
able factor for 3D-QSAR studies. In our study, three
different kinds of alignment were selected to define
overlap. The first superposition of molecules, which
was based on purine ring as common structure (align-
ment A), is displayed in Fig. 2. The second superposi-
tion, which was based on using adenosine ring as com-
played in Fig. 4.
The CoMFA method was performed on molecular
alignment to generate the 3D-QSAR model. Adjusting
column filtering from 0.5 to 5 kcal/mol would improve
efficiency and decrease the noise. Cutoff values of two
CoMFA descriptors were changed at some point from
10 to 50 kcal/mol to cut the large steric domination
and minimize electrostatic energies. The remaining
mon structure (alignment B), is displayed in Fig. 3. parameters are the system defaults.
RUSSIAN JOURNAL OF BIOORGANIC CHEMISTRY Vol. 46 No. 3 2020

450
SHUNLAI LI et al.
Table 1. The structures of adenosine derivatives and their experimental and predictive activities
NR1R2
N
N
HO
N
N
SR
O
OH OH
IC50,
Predicted log(PAC)
PAC
(%, 300 μM) c
R1
R2
Entry Compd.
R
log(PAC)
μMa
CoMFA CoMSIA SoMFA2
Ncb
1
2
3
4
5a1
5a5
5a3
5a4
Et
Et
n-C6H13
c-C6H11
n-C6H13
n-C6H13
n-C6H13
c-C6H11
c-C6H11
c-C6H11
n-Bu
H
H
H
H
H
H
H
H
3.767
4.284
3.666
3.703
4.292
4.122
4.029
4.382
3.636
3.588
4.502
3.807
4.056
4.043
3.996
3.636
3.578
3.678
3.574
3.684
3.654
3.967
3.609
4.745
4.839
4.585
4.530
3.767
3.767
4.116
4.352
3.875
4.333
3.660
3.748
4.251
3.608
3.730
3.720
4.182
4.135
4.189
3.642
3.562
4.499
3.934
4.011
4.984
4.004
3.614
3.620
3.614
3.569
3.704
3.817
3.724
3.673
4.681
4.725
4.565
4.588
3.832
3.755
4.096
4.151
4.075
4.385
3.668
3.794
4.295
3.208
3.723
3.738
4.146
4.125
4.23
3.697
3.991
3.746
3.708
3.759
4.071
3.990
3.996
3.698
3.490
4.101
3.878
3.939
3.914
3.786
3.742
3.604
3.608
3.632
3.651
3.703
3.758
3.582
4.289
4.139
4.235
4.444
4.182
3.773
4.189
4.325
4.293
4.321
4.029
57.0
17.3
72.0
66.0
17.0
25.2
31.2
14.8
77.0
86.0
10.5
52.0
29.3
30.2
33.7
77.0
88.0
70.0
89.0
69.0
74.0
36.0
82.0
6.0
104
Nc
Nc
7
i-Pr
n-Bu
n-Pr
n-Pr
i-Pr
n-Bu
Et
5* 5a2
102 11
151
187 15
6
7
8
9
5a6
5a7
5a8
5a9
9
83
4
n-Bu Nc
n-Bu Nc
3.635
3.582
4.501
3.921
4.027
3.997
3.989
3.569
3.671
3.615
3.606
3.698
3.814
3.712
3.681
4.736
4.789
4.580
4.547
3.779
3.750
4.078
4.163
4.117
4.329
3.686
10 5a10
11 5a11
12 5b1
13 5b2
14 5b3
15* 5b4
16 5b5
17 5b6
18 5b7
19 5b8
20 5b9
21 5b10
22 5b11
23 5b12
24 5b13
25* 5b14
26 5b15
27 5b16
28 5b17
29 5b18
30 5b19
31 5b20
32 5b21
33 5b22
34 5b23
n-Pr
n-Bu
Me
n-Bu
CH3OCH2CH2
CH2Ph
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
63
Nc
5
Et
CH2Ph
176 16
181 14
202 20
Nc
n-Pr
n-Bu
Me
CH2Ph
CH2Ph
p-MePhCH2
p-MePhCH2
p-MePhCH2
p-MePhCH2
p-MeOPhCH2
p-MeOPhCH2
p-MeOPhCH2
p-MeOPhCH2
PhCH2CH2
PhCH2CH2
PhCH2CH2
PhCH2CH2
PhCH2CH2
Et
Nc
n-Pr
n-Bu
Me
Nc
Nc
Nc
Et
Nc
n-Pr
n-Bu
Me
216 12
Nc
36
29
5
3
3
6
Et
4.8
n-Pr
n-Bu
Ph
52
8.7
59
Nc
Nc
9.8
57.0
57.0
25.5
14.8
44.5
15.5
73.0
CH2Ph PhCH2CH2
Me
p-MeOPhCH2CH2
153 13
89 10
267 18
Et
p-MeOPhCH2CH2
p-MeOPhCH2CH2
p-MeOPhCH2CH2
n-Pr
n-Bu
93
7
CH2Ph p-MeOPhCH2CH2
Nc
RUSSIAN JOURNAL OF BIOORGANIC CHEMISTRY
Vol. 46
No. 3
2020

ESSENTIAL STRUCTURAL PROFILE
NR1R2
451
Table 1. (Contd.)
N
N
HO
N
N
SR
O
OH OH
IC50,
Predicted log(PAC)
CoMFA CoMSIA SoMFA2
PAC
(%, 300 μM) c
R1
R2
Entry Compd.
R
log(PAC)
μMa
35* 5b24
36 5b25
37 5b26
38 5b27
39 5b28
40 5b29
41 5b30
42 5b31
43 5b32
44 5b33
45* 5b34
Me
Et
m-MeOPhCH2CH2
m-MeOPhCH2CH2
m-MeOPhCH2CH2
m-MeOPhCH2CH2
CH(CH3)Ph
H
H
H
H
H
H
H
H
H
H
H
Nc
3.672
4.721
4.462
3.594
4.007
3.609
3.767
3.654
3.807
3.569
3.684
4.599
4.649
4.551
3.551
3.999
3.617
3.725
3.755
3.787
3.626
3.401
4.532
4.643
4.530
3.568
4.001
3.622
3.680
3.744
3.812
3.563
4.129
4.147
4.361
4.353
3.985
3.956
3.788
3.964
3.907
4.114
3.874
3.991
71.0
6.3
38
69
Nc
4
5
n-Pr
n-Bu
Et
11.5
85.0
32.8
82.0
57.0
74.0
52.0
90.0
69.0
197 12
Nc
n-Pr
Me
CH(CH3)Ph
p-CH2CH2C6H4F
p-CH2CH2C6H4F
p-CH2CH2C6H4F
p-CH2C6H4F
Nc
Et
Nc
n-Pr
Et
Nc
Nc
n-Pr
p-CH2C6H4F
Nc
a IC50 values are expressed as mean SEM. (n = 3) and calculated when platelet aggregation was below 50% of control (10 μM ADP as
agonist). b Nc, not calculated, because maximal inhibition of aggregation was lower than 50% at final concentration of 300 μM (10 μM
ADP as agonist). c PAC, platelet aggregation of control; the inhibition of aggregation at final concentration of 300 μM (10 μM ADP as
agonist). * Represent test set compounds, others are training set.
Except for steric (S) and electrostatic (E) fields, descriptors (as independent variables) to the activity
there are also three fields—hydrophobic property (H),
hydrogen bond donor (D), and hydrogen bond accep-
tor (A)—used to build models in the CoMSIA
method. Adjusting column filtering from 0.5 to
6 kcal/mol would improve efficiency and decrease the
noise. And the remaining parameters are the system
defaults. According to permutation and combination
of five descriptors, 45 generated models were prepared
for further 3D-QSAR analysis.
values (log(PAC)) of training set (as dependent vari-
ables) so as to build the 3D-QSAR models. To validate
the predictive ability of the built model by PLS, the
cross-validation analysis was performed using the
leave one out (LOO) method. Then the cross-valida-
tion correlation coefficient (q2) indicates predicative
power and robustness of built models. Some PLS
models with cross-validation analysis were accom-
plished in different components number to derive the
ONC (optimum number of components) with the
lowest standard error of estimate (SEE). According to
In this study, the partial least square (PLS) method
[22] was used to relate the CoMFA and CoMSIA
Fig. 2. Superposition of molecules using alignment A.
Superposition of all compounds.
Fig. 3. Superposition of molecules using alignment B.
Superposition of all compounds.
RUSSIAN JOURNAL OF BIOORGANIC CHEMISTRY Vol. 46 No. 3 2020

452
SHUNLAI LI et al.
40.6%, respectively. The predicted log(PAC) values of
best models calculated by CoMFA are listed in Table 1.
The correlation between experimental and predicted
log(PAC) values by CoMFA model is shown in
Fig. 5a. The high F and r2 with lower SEE value indi-
cates that the model has appropriate reliability and
predictive ability.
Compound 25 displayed in Fig. 6 to aid visualiza-
tion. For the best CoMFA model, the steric and elec-
trostatic contour maps are shown in Figs. 7a, 7b. In
Fig. 7a, the green contour parts (80% contributions)
indicating the bulky groups would increase the activi-
ties and yellow contour parts (20% contributions)
showing the bulky groups were unfavorable groups. In
Fig. 7b, blue and red contour parts (80% and 20%
contributions) represented positively and negatively
charged groups that were favorable for activities,
respectively.
Fig. 4. Superposition of molecules using alignment C.
Superposition of all compounds.
the optimum number of components, PLS analysis wa
then followed with non-cross-validation as final mod-
elling tool and a series of statistical parameter were
obtained, such as the squared correlation coefficient
(r2), standard error of estimate (SEE), and F values.
Finally, the predictive abilities of built models were
validated using the test set.
According to the steric contour maps in Fig. 7a,
great green contours appears at the end of region A,
which indicates that bulky group in this position would
increase activity. For example, incompounds 22 and
14 listed in Table 1 terminal structure of substitute R1
changes from –H to –OMe; log(PAC) values (4.043 >
3.967) increase with increase in the size of the substi-
tutes (–H < –OMe). At the same time, yellow con-
tours surrounding the middle part of region A indicate
that the bulky group in this position would decrease
activity, and then bulky group in region A can not be
The three different alignment CoMFA models
were obtained based on training set consisting of 40
compounds to analyze the relationship between chem-
ical structures and antiplatelet aggregation activity;
the statistical parameters associated with CoMFA are
summarized in Table 2.
The specific parameters of the best CoMFA model
(alignment A) are as follows: the cross-validated q2 of
0.560 with seven components, non-cross-validated r2 connected to C-6 position of purine ring directly, sug-
of 0.940, F = 71.850, and SEE of 0.090. The contribu- gesting connection through gracile link groups. Com-
tions of the steric and electrostatic fields were 59.4 and pound 13 and 39 listed in Table 1 have substitutes
Table 2. Statistical results of CoMFA and CoMSIA models
COMFA
COMSIA
alignment A
alignment B
alignment C
alignment A
alignment B
alignment C
q2a
ONCb
r2c
SEEd
Fe
Sf
Ef
Hf
Df
Af
0.560
7
0.058
1
0.048
2
0.528
4
0.120
2
0.157
4
0.940
0.097
71.850
59.4
40.6
–
0.506
0.260
38.976
49.3
50.7
–
0.375
0.293
11.081
58.2
0.943
0.108
29.290
6.1
0.556
0.250
23.141
–
0.676
0.216
18.294
15.3
39.0
33.0
10.0
2.7
41.8
16.2
39.7
38.0
–
26.7
73.3
–
–
–
–
–
–
0.092
(0.712)
0.097
(0.766)
2
g
–
–
–
–
rpred
b
c 2
aq2, cross-validated correlation coefficient. ONC, optimum number of componentsffrom PLS analysis. r , non-cross-validated cor-
relation coefficient. dSEE, standard error of estimate. F, the value of F statistic. Field contributions: steric (S), electrostatic (E),
e
hydrophobic (H), hydrogen bond donor (D), and gydrogen bond acceptor (A) fields. gPredictive r2 of all test set compounds (CoMFA:
compound 35 classified as outlier; CoMSIA: compound 25 classified as outlier).
RUSSIAN JOURNAL OF BIOORGANIC CHEMISTRY
Vol. 46
No. 3
2020

ESSENTIAL STRUCTURAL PROFILE
453
(a)
(b)
Training
Test
Training
Test
4.8
4.6
4.4
4.2
4.0
3.8
3.6
3.4
3.2
4.8
4.6
4.4
4.2
4.0
3.8
3.6
3.4
3.6 3.8 4.0 4.2 4.4 4.6 4.8
Experiment log(PAC)
3.6 3.8 4.0 4.2 4.4 4.6 4.8 5.0
Experiment log(PAC)
(c)
Training
4.8
4.6
4.4
4.2
4.0
3.8
3.6
Test
3.4
3.6 3.8 4.0 4.2 4.4 4.6 4.8
Experiment log(PAC)
Fig. 5. Correlation between experimental and predicted activities of the best CoMFA and CoMSIA models. (a) CoMFA;
(b) CoMSIA; and (c) SOMFA.
varying from –CH2Ph to –CH(CH3)Ph; the increase This red contour indicates that the electron-with-
of size of substitutes R1 (–CH2Ph < –CH(CH3)Ph)
however resulted in the decrease of log(PAC) values
(4.056 > 4.007). There is a small yellow polyhedron
near the ethylthio group in region B, indicating that
small groups would increase the activity.
drawing group could decrease the activity (log(PAC)).
In compounds 24 and 30, when structures changed
from –H to –OMe (–H < –OMe, –OMe, an elec-
tron-donating group), the activity decreased respec-
tively (4.745 > 4.166), due to the decreasing of elec-
tron-withdrawing potency in substituent at position 6.
In Fig. 7b, there are still some small blue contours near
the sugar ring in region C, which indicates that more
hydroxyl groups with positive charge would increase
the activity.
In Fig. 7b, the appearance of the two large blue
contours near the region A, one blue contour close to
the amino group and the other around the terminal of
phenyl rings, indicates that groups with positive charge
would increase the activity. For example, structures of
substituents in compounds 13 and 2 changed from
To constrcut CoMSIA models, there are also three
‒CH2Ph to –c-C6H11 (–CH2Ph < –c-C6H11); the different alignments applied in training set for the
combinations of different descriptors. Five CoMSIA
descriptors dependency would reduce the model sig-
nificance and predictive ability [23]. Therefore, all 31
activity increased respectively (4.056 < 4.284), due to
the increasing of electron-donating potency in substit-
uent at position 6. It also can be seen from Fig. 7b that
there is a red region appearing below the benzene ring. possible combinations of descriptors were calculated
RUSSIAN JOURNAL OF BIOORGANIC CHEMISTRY Vol. 46 No. 3 2020

454
SHUNLAI LI et al.
with their respective q2 value and optimum number of
Region A
components shown in Fig. 8. Electrostatic, steric,
hydrogen bond donor, and hydrophobic fields present
the highest q2 value (0.528) and were thus selected to
create the final CoMSIA model.
HN
7
6
N
5
4
N
1
Region C
8
N
N
3
9
The statistical parameters obtained by the best
CoMSIA model for different alignments are listed in
Table 2. The model of alignment A provided the high
q2 and r2 value of 0.528 and 0.943 with optimized com-
ponents of 14. The derived values for the best model
(CoMSIA) are fulfilling the threshold criteria. The
contributions of hydrogen bond donor, hydrophobic,
steric, and electrostatic fields were 38.0, 39.7, 6.1, and
2
HO
5'
O
1'
2'
OH OH
4'
3'
Region B
Fig. 6. Structure of template compound (the most active
molecule 25), the three depicted regions A, B, and C, and
the number of each substituent position in the purine ring.
(a) CoMFA steric
(b) CoMFA electrostatic
(c) CoMSIA steric
(d) CoMSIA electrostatic
(e) CoMSIA hydrophobic
(f) CoMSIA H-bond donor
(g) SOMFA electrostatic
(h) SOMFA steric
Fig. 7. Contour maps of CoMFA and CoMSIA: (a) CoMFA steric; (b) CoMFA electrostatic; (c) CoMSIA steric; (d) CoMSIA
electrostatic; (e) CoMSIA hydrophobic; (f) CoMSIA H-bond donor electrostatic; (g) SOMFA electrostatic; (h) SOMFA steric
fields based on compound 25.
RUSSIAN JOURNAL OF BIOORGANIC CHEMISTRY
Vol. 46
No. 3
2020

ESSENTIAL STRUCTURAL PROFILE
455
and 15 not having p-methoxy group. This phenome-
q2
non agreed with the fact of the CoMFA steric field.
0.5
Hydrogen bond donor contour maps of CoMSIA
models are shown in Fig. 7f. Cyan and purple contours
represent regions where hydrogen bond donor groups
would be favorable and unfavorable for activity,
respectively. As displayed in Fig. 7f, there is a large
cyan contour near the C-2 position of purine ring,
suggesting that the hydrogen bond donor substituents
in region A would increase the activity. The findings
also support our prior research conclusion: the com-
pounds substituted with alkylamino groups possess
higher antiplatelet aggregation inhibitory activities than
the compounds substituted with alkoxyl groups [24].
0.4
0.3
0.2
0.1
0
For SOMFA models, we added five different
charges (DELRE, PULLMAN, GASTEIGER,
GAST_HUCK, MMFF94) to three different align-
ments to establish fifteen models. In SOMFA2 soft-
ware, we calculated the electrostatic and steric fields of
fifteen models with two resolutions of grid (0.5, 1.0 Å)
inside a three-dimensional (coordinate origin: –20,
‒20, –20; volume: 40 × 40 × 40 Å). The nonlinear
equation (log(PAC) = c1log(PAC)ESP + (1 – c1) ×
log(PAC)shap) was used to measure the effects of elec-
trostatic and steric fields. The c1 of all thirty models
was calculated. At the same time, we got the corre-
Emembles
Fig. 8. Graph of the 31 possible CoMSIA descriptors com-
binations (S = steric, E = electrostatic, H = hydrophobic,
D/A = H-bond donor/acceptor) with their respective q2
values, optimal numbers of components are reported
above bars.
2
2
sponding values of SEE, F value, r2,
and
listed
rcv,
rpred
in Table 3.
16.2%, respectively. In this model, the hydrophobic
field was found to have higher contributions to the
activity. The predicted log(PAC) values calculated by
this CoMSIA model are also given in Table 1. The
log(PAC) values of compounds of the test set were also
predicted by this model. The correlation plot of pre-
dicted and experimental log(PAC) values were shown
in Fig. 5b.
The contour maps of CoMSIA fields (steric, elec-
trostatic, hydrophobic, and hydrogen bond donor)
with compound 25 are shown in Fig. 7. As shown in
Fig. 7d, the contour maps of CoMSIA fields (steric,
electrostatic) were basically similar to those of the
CoMFA method; for example, small groups with pos-
itive charge (such as alkyl amino group) near C-6 posi-
tion of purine ring and bulky electron withdrawing group
(such as phenyl rings with –F, –Cl, –Br, and –NO2) far
away from C-6 position of purine ring in region A
would increase the activity. Therefore, only contour
maps of hydrophobic and hydrogen bond donor fields
are discussed in detail.
For Table 3, it is obvious that the twelfth model is
2
the best model with highest
value of 0.577 which
rcv
indicates that the model can provide high predictabil-
ity; and the predicted activities (log(PAC)) are listed in
Table 1. An optimal coefficient c1 = 0.154 indicted that
the steric contribution has high importance. The liner
relation graph of experiment and predict log(PAC)
value are shown in Fig. 5c, including the training and
test set, with liner correlation coefficient 0.615 and
0.003 (0.781) (predicted r2 of all test or compound 25
as outliner).
For SOMFA models, the contour maps of electro-
static and steric fields of compound 25 are also shown
in Fig. 7. In Fig. 7g, the red grid dots at the C-5' posi-
tion indicate that a positive charge is favorable in this
region and the blue indicates that a negative charge is
favorable in this region. In Fig. 7h, the red grid dots at
the C-2, C-6, and C-5' positions indicate that steric
bulk enhances the activity and the blue ones decrese
the activity in these regions.
In hydrophobic field contour maps, yellow and
Based on the above analysis of CoMFA, CoMSIA,
gray contours showed the regions where hydrophobic and SOMFA method, we found many effects of each
groups were advantageous and disadvantageous for group as a substitute in the region A, B, and C. The
activity, respectively. As shown in Fig. 7e, the appear- results show that the inhibitory activities could be
ance of yellow contours at C-6 position of purine ring enhanced by introduction of medium-sized groups at
in region A indicates that terminal hydrophobic sub- the C-2 position of the purine ring; long groups with
stituent would increase the activity. For example, the strong electronegativity at the C-6 position of the
inhibitory activity of compound 14 containing purine ring and more hydrogen bonding donor groups
p-methoxy group was better than that of compounds 22 at the C-5' position of the sugar ring. Taking align-
RUSSIAN JOURNAL OF BIOORGANIC CHEMISTRY Vol. 46 No. 3 2020

456
SHUNLAI LI et al.
Table 3. Statistical results of SOMFA models with different alignments, charges, and resolution of grid
a
2
2
sb
Fc
r2d
f
e
Model no. Alignment
Charge
Resolution of grid, Å
c1
rpred
rcv
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
A
A
A
A
A
A
A
A
A
A
B
B
B
B
B
B
B
B
B
B
C
C
C
C
C
C
C
C
C
C
DELRE
DELRE
0.5
1.0
0.5
1.0
0.5
1.0
0.5
1.0
0.5
1.0
0.5
1.0
0.5
1.0
0.5
1.0
0.5
1.0
0.5
1.0
0.5
1.0
0.5
1.0
0.5
1.0
0.5
1.0
0.5
1.0
0.830 0.295 29.252 0.348 0.291 0.056 (0.003)
1.005 0.284 24.857 0.395 0.343 0.039 (0.000)
0.825 0.296 19.680 0.341 0.275 0.049 (0.034)
0.918 0.289 22.665 0.374 0.308 0.035 (0.033)
0.707 0.297 19.229 0.336 0.278 0.050 (0.013)
0.782 0.293 20.906 0.355 0.299 0.041 (0.009)
0.681 0.298 18.932 0.333 0.268 0.047 (0.035)
0.792 0.295 20.104 0.356 0.283 0.042 (0.027)
0.833 0.304 16.792 0.306 0.240 0.096 (0.006)
1.173 0.298 19.107 0.335 0.273 0.063 (0.013)
0.167 0.230 57.534 0.602 0.564 0.001 (0.679)
0.154 0.226 60.797 0.615 0.577 0.003 (0.781)
0.114 0.234 54.864 0.591 0.551 0.001 (0.658)
0.105 0.229 58.322 0.605 0.566 0.015 (0.773)
0.110 0.233 55.037 0.592 0.551 0.001 (0.673)
0.091 0.229 58.192 0.605 0.565 0.004 (0.777)
0.027 0.230 53.693 0.586 0.545 0.000 (0.670)
0.010 0.231 57.160 0.601 0.561 0.008 (0.782)
0.147 0.233 55.424 0.593 0.554 0.000 (0.668)
0.143 0.168 59.057 0.608 0.569 0.010 (0.778)
0.008 0.292 21.234 0.358 0.292 0.002 (0.353)
0.008 0.294 20.500 0.350 0.283 0.005 (0.321)
0.043 0.294 20.520 0.351 0.284 0.002 (0.353)
0.043 0.294 20.645 0.352 0.285 0.005 (0.321)
0.008 0.292 21.234 0.358 0.292 0.002 (0.353)
0.006 0.294 20.531 0.351 0.284 0.005 (0.321)
0.008 0.292 21.234 0.358 0.292 0.002 (0.353)
–0.140 0.294 20.520 0.351 0.284 0.004 (0.315)
0.246 0.291 21.882 0.365 0.301 0.015 (0.453)
0.202 0.293 20.898 0.355 0.289 0.012 (0.421)
PULLMAN
PULLMAN
GASTEIGER
GASTEIGER
GAST_HUCK
GAST_HUCK
MMFF94
MMFF94
DELRE
DELRE
PULLMAN
PULLMAN
GASTEIGER
GASTEIGER
GAST_HUCK
GAST_HUCK
MMFF94
MMFF94
DELRE
DELRE
PULLMAN
PULLMAN
GASTEIGER
GASTEIGER
GAST_HUCK
GAST_HUCK
MMFF94
MMFF94
aMixing coefficient of the SOMFA model. bStandard error of the estimate. cF-test value. dNon-cross-validated correlation coefficient.
eCross-validated correlation coefficient. fPredictive r2 of all test set compounds (compound 25 classified as outlier).
ACKNOWLEDGMENTS
ments into account, we found that the alignment had
a profound influence on the result. The model which
was built by alignment A showed the highest values of
r2 and q2 because it reflected the fact that purine ring
was the true basic backbone of adenosine derivatives.
The most potent compound 25 was regarded as the
reference molecule, and the new compounds could be
designed by replacing favorable groups at each posi-
tion of compound 25. These findings can be applied to
design new adenosine derivatives with bulky group at
position 6 in region A with higher electron withdraw-
ing ability, such as F, Cl, Br, and –NO2, and the stron-
ger hydrogen bond donor groups at position C-5', such
as –COOH, –CH2OH, etc.
The authors gratefully acknowledge professor Lei Ming.
We also thank the Chemical Grid Project at Beijing Univer-
sity of Chemical Technology (BUCT) for providing a part of
the computational resources. By analyzing the corresponding
contour maps detailly, new adenosine derivatives with poten-
tial efficacy were designed for synthesis in the future.
FUNDING
This work was supported by the National Natural Sci-
ence Foundation of China (no. 21272022).
RUSSIAN JOURNAL OF BIOORGANIC CHEMISTRY
Vol. 46
No. 3
2020

ESSENTIAL STRUCTURAL PROFILE
457
COMPLIANCE WITH ETHICAL STANDARDS
Clegg, J.A., Smith, J.A., and Tomlinson, W., J. Med.
Chem., 1999, vol. 42, pp. 213–220.
The work has no studies involving humans or animals as
subjects of the study.
10. Ferreiro, J.L., Ueno, M., and Angiolillo, D.J., Expert
Rev. Cardiovasc. Ther., 2009, vol. 7, pp. 1195–1201.
11. Langlois, M., Brémont, B., Rousselle, D., and Gaudy, F.
Eur. J. Pharmacol.: Mol. Pharmacol., 1993, vol. 244,
pp. 77–87.
12. Jiang, H.L., Chen, K.X., Wang, H.W., Tian, Y., Chen, J.Z.,
and Ji, R.Y., Acta Pharmacol. Sin., 1994, vol. 15,
pp. 481–487.
Conflict of Interests
Authors declare they have no conflicts of interest.
REFERENCES
1. Flügel, K.A., Fortschr. Med., 1980, vol. 98, pp. 773–
13. Sperandio, G.M., Sant’Anna, C.M., Barreiro, E., J.
Bioorgan. Med. Chem., 2004, vol. 12, pp. 3159–3166.
778.
14. Caballero, J., J. Mol. Graph. Model., 2010, vol. 29,
2. Wood, D.A., Lancet, 2001, vol. 357, pp. 995–1001.
pp. 363–371.
3. Gould, W.R. and Leadley, R.J., Curr. Pharm. Design,
15. Zhang, J., Wang, X.J., Dong, Z.K., Wang, S.Q.,
Xu, W.R., Fu, J.W., Cheng, X.C., and Wang, R.L., Lett.
Drug Des. Discov., 2016, vol. 13, pp. 376–386.
2003, vol. 9, pp. 2337–2347.
4. Lip, G.Y., Felmeden, D.C., and Dwivedi, G., Co-
chrane Db. Syst. Rev., 2011, vol. 8, CD003186.
16. Cramer, R.D., Patterson, D.E., and Bunce, J.D., J.
5. Springthorpe, B., Bailey, A., Barton, P., Birkinshaw, T.N.,
Bonnert, R.V., Brown, R.C., Chapman, D., Dixon, J.,
Guile, S.D., Humphries, R.G., Hunt, S.F., Ince, F.,
Ingall, A.H., Kirk, I.P., Leeson, P.D., Leff, P., Lewis, R.J.,
Martin, B.P., McGinnity, D.F., Mortimore, M.P.,
Paine, S.W., Pairaudeau, G., Patel, A., Rigby, A.J., Ri-
ley, R.J., Teobald, B.J., Tomlinson, W., Webborn, P.J.,
and Willis, P.A., Bioorg. Med. Chem. Lett., 2007,
vol. 17, pp. 6013–6018.
Am. Chem. Soc., 1988, vol. 110, pp. 5959–5967.
17. Klebe, G., Abraham, U., and Mietzner, T., J. Med.
Chem., 1994, vol. 37, pp. 4130–4146.
18. Li, Z., Zhou, M., Wu, F., Li, R., and Ding, Z., Eur. J.
Med. Chem., 2011, vol. 46, pp. 58–64.
19. Liu, G., Xu, J., Chen, N., Zhang, S., Ding, Z., and Du, H.,
Eur. J. Med. Chem., 2012, vol. 53, pp. 114–123.
20. Zhao, L.J., Zhang, L.R., and Lei, M., Sci. China
Chem., 2013, vol. 56, pp. 1550–1563.
6. Fosburgh, D., Mayo Clin. Proc., 2006, vol. 81, pp. 989–
989.
21. Gasteiger, J. and Marsili, M., Tetrahedron, 1980,
vol. 36, pp. 3219–3228.
7. Savi, P. and Herbert, J.M., Haematologica, 2000,
vol. 85, pp. 73–77.
22. Wold, S., Ruhe, A., Wold, H., and Dunn III, W.J.,
Siam J. Sci. Stat. Comput., 1984, vol. 5, pp. 735–743.
8. Husted, S. and van Giezen, J.J., Cardiovasc. Ther.,
2009, vol. 27, pp. 259–274.
23. Bringmann, G. and Rummey, C., J. Chem. Inf. Comput.
Sci., 2003, vol. 43, pp. 304–316.
9. Ingall, A.H., Dixon, J., Bailey, A., Coombs, M.E.,
Cox, D., McInally, J.I., Hunt, S.F., Kindon, N.D., 24. Du, H.G., Yu, M.W., Deng, C.E., Lu, C.H., and Li, S.L.,
Teobald, B.J., Willis, P.A., Humphries, R.G., Leff, P.,
J. Heterocycl. Chem., 2016, vol. 54, pp. 436–449.
RUSSIAN JOURNAL OF BIOORGANIC CHEMISTRY Vol. 46 No. 3 2020