Synthesis of novel 5-aryl-2-thio-1,3,4-oxadiazoles and the study of their structure-anti-mycobacterial activities

Article abstract of DOI:10.1016/j.bmc.2005.05.011

The preparation of novel 5-aryl-2-thio-1,3,4-oxadiazoles 4a-41 and the computer-aided study of their in vitro anti-tubercular activity against Mycobacterium tuberculosis H₃₇Rv (ATCC 27294) are reported. The average accuracy of the electronic-topological method and neural network methods applied to the activity prediction in leave-one-out cross validation is 80%.

Full text of DOI:10.1016/j.bmc.2005.05.011

Bioorganic & Medicinal Chemistry 13 (2005) 4842–4850
Synthesis of novel 5-aryl-2-thio-1,3,4-oxadiazoles and the study
of their structure–anti-mycobacterial activities
Fliur Macaev,a,* Ghenadie Rusu,a Serghei Pogrebnoi,a Alexandru Gudima,a
Eugenia Stingaci,a Ludmila Vlad,a Nathaly Shvets,b Fatma Kandemirli,c
d
Anatholy Dimogloa,b, and Robert Reynolds
*
aInstitute of Chemistry, Academy of Sciences of Moldova, Chisinau, MD-2028, Republic of Moldova
bGebze Institute of Technology, 41400, Turkey
cKocaeli University, 41400, Turkey
dSouthern Research Institute, Birmingham, AL 35255-5305, USA
Received 13 April 2005; revised 5 May 2005; accepted 6 May 2005
Abstract—The preparation of novel 5-aryl-2-thio-1,3,4-oxadiazoles 4a–41 and the computer-aided study of their in vitro anti-tuber-
cular activity against Mycobacterium tuberculosis H37Rv (ATCC 27294) are reported. The average accuracy of the electronic-topo-
logical method and neural network methods applied to the activity prediction in leave-one-out cross validation is 80%.
ꢀ 2005 Elsevier Ltd. All rights reserved.
1. Introduction
drugs.16,17 We also report our efforts to relate the depen-
dence of the anti-mycobacterial activity of new
compounds on the nature of substitution in the 5-aryl-
2-thio-1,3,4-oxadiazoles 4a–41.
Data from the World Organization of Health show a
significant rise in drug-resistant tuberculosis.1–3 Tuber-
culosis (TB) is one of the leading causes of death and
suffering worldwide among the infectious diseases. The
ever-increasing drug resistance, toxicity and side effects
of currently used anti-tuberculosis drugs and the ab-
sence of their bactericidal activity highlight the need
for new, safer and more effective anti-tuberculosis
drugs.4–8 The computer-aided prediction of biological
activity in relation to the chemical structure of a com-
pound is now a commonly used technique in drug dis-
covery.9–14 Modern drug discovery also relies on the
interface of chemical and biological diversity through
high throughput screening.15 Generation of functional
molecular diversity to probe biological activity space re-
quires robust molecular scaffolds that are low in molec-
ular weight and are easily modified to create a variety of
chemically diverse, biologically active potential
The present study uses both the electronic-topological
method (ETM) and neural network (NN) methods to
predict anti-tuberculosis activity. The better the descrip-
tion of a molecule in terms of structural parameters rep-
resenting its activity, the better the results of pattern
recognition and separation of the molecules by activity.
The ETM is a structure-based approach for studying
structure–activity relationships (SAR) of molecules.18
ETM is also capable of taking into account individual
properties of separate atoms and bonds that may be cru-
cial for revealing details of interactions between a bio-
logic receptor and an active molecule. A number of
studies have been reported that use the ETM to find
SAR models involving a representative list of activities
and thousands of compounds belonging to different
chemical classes.19–22 The ETM approach and software
have both undergone considerable development.23 The
new ETM-based system capable of unifying Web-based
SAR resources and using Internet communications in
the framework of a project, ETOSAR, makes the
ETM a valuable tool for SAR studies and provides rules
for the synthesis of new potentially active compounds.
Keywords: Anti-tubercular activity; 5-Aryl-2-thio-1,3,4-oxadiazoles;
Structure–activity relationships; Electronic-topological method; Neu-
ral network method.
*
Corresponding authors. Tel.: +90 262 653 84 97/1428; fax: +90 262
653 84 90; e-mail addresses: flmacaev@cc.acad.md; dimoglo@gyte.
edu.tr
0968-0896/$ - see front matter ꢀ 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmc.2005.05.011

F. Macaev et al. / Bioorg. Med. Chem. 13 (2005) 4842–4850
4843
For the analysis of data on the pharmacophores in this
study we have used one of the most well known NN—
the feed forward neural networks (FFNNs) trained with
the back propagation algorithm.24,25
example, reaction of thiol 4a–j with various x-bromo-
acetophenones was carried out at room temperature in
acetone in the presence of Et3N. The synthesis of new
1,3,4-oxadiazoles 25–39 as the carbonyl-protected ana-
logs of compound 7–23 by alkylation of oxadiazole
4a–j with the corresponding 2-phenyl-2-bromomethyl-
1,3-dioxolanes in DMF/K2CO3 at elevated temperature
is depicted in Scheme 1. Allylic bromination of com-
pound 7 by NBS following condensation of product
24 with thiourea gave the 2-aminothiazole 41. Under
Knoevenagel conditions, a condensation of 7 with p-
Br-benzaldehyde was preformed, and compound 40
was obtained as the main reaction product. All com-
pounds were obtained with high yields. Physical, analy-
tical and spectral data are presented in Tables 2 and 3.
2. Chemistry and biological activity testing
This chemistry utilizes different benzoic acids 1a–j as
starting materials. The synthesis of targets 1,3,4-
oxadiazoles 4a–41 was accomplished, following a
four-step synthetic route according to Scheme 1. The
synthesized compounds are given in Table 1.
Hydrazinolysis of esters 2a–j with 100% hydrazine
hydrate in refluxing EtOH gave the corresponding
hydrazides 3a–j. We have reported earlier that arylhydr-
azines react with tetramethylthiuram disulfide (TMTD)
in DMF to form 5-aryl-2-thio-1,3,4-oxadiazoles.26 Dur-
ing the course of this study, compounds 3a–j were cyc-
lized into thiols 4a–j with TMTD in DMF by the
reported procedure.
The synthesized compounds 4a–41 in the present study
were tested for the in vitro anti-mycobacterial activity
against Mycobacterium tuberculosis H37Rv using the
Alamar Blue assay method.15 Predicted and experimen-
tally-measured percent inhibition (PI—relative inhibi-
tion of bacterial growth at 6.25 lg/mL) values for the
synthesized compounds against M. tuberculosis H37Rv
are given in Table 1. Rifampicin was used as the control
standard in the assays for in vitro inhibition of M. tuber-
culosis H37Rv.
The synthesis of 2-R-thio-5-phenyl-1,3,4-oxadiazoles 5–
23 and 25–39 was carried out in acetone or DMF,
depending on the nature of the alkylating reagent. For
Scheme 1. The synthesis of compounds 2a–41.

4844
F. Macaev et al. / Bioorg. Med. Chem. 13 (2005) 4842–4850
Table 1. Synthesized compounds and anti-mycobacterial activity against M. tuberculosis H37Rv (in vitro)
Compound
R1
R2
R3
R4
R5
R6
Inhibition (%)
Prediction (%)
4f
CI
H
H
CI
OMe
H
H
—
—
—
—
—
—
H
—
—
—
—
—
—
H
20
3
19
4
4j
OMe
H
H
Me
5b
5c
5f
OH
Br
H
30
47
20
39
28
45
19
17
58
27
24
22
12
17
48
37
33
27
31
35
37
19
35
23
54
19
13
54
30
31
32
7
H
H
Me
Me
OMe
OMe
H
OMe
OMe
H
6
H
CH2Py
H
7
H
8
OH
Me
CI
H
H
H
H
H
H
9
H
H
H
H
H
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24a
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
H
H
H
H
H
H
H
H
H
F
H
OMe
H
OMe
H
H
H
F
H
H
H
NO2
Br
CI
H
H
OMe
OMe
OMe
OMe
OMe
OMe
OMe
OMe
OMe
OMe
H
OMe
OMe
OMe
OMe
OMe
OMe
OMe
OMe
OMe
OMe
H
H
Cl
H
H
H
H
H
OMe
H
20
36
34
33
30
30
H
H
Me
Phenyl
NO2
CI
OMe
31
H
H
H
H
H
H
H
H
H
H
H
H
2-Naphthyl
OMe
39
H
H
H
H
19
33
33
31
43
39
H
H
H
H
OH
OH
OH
H
H
H
H
2-Naphthyl
H
F
H
H
15
H
15
Br
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
13
82
55
0
16
53
37
4
H
H
H
OH
H
H
H
H
H
H
OEt
OMe
OMe
OH
OH
OMe
OMe
OMe
OMe
OMe
H
H
H
H
H
OMe
H
H
0
4
H
H
CI
H
15
0
14
3
H
H
OMe
OMe
H
H
H
OMe
F
3
11
4
H
OMe
OMe
OMe
OMe
OMe
H
6
H
H
Br
CI
H
7
4
H
H
28
5
37
5
H
OMe
H
H
H
14
30
55
11
18
58
H
H
H
H
H
H
H
H
a Note. X = Br.
In the first part of this study, we investigated depen-
dence of target activity on the nature of substituents in
the 1,3,4-oxadiazoles 4f,j; 5b,c,j and 6. Thiol 4j had the
lowest activity (PI 3%). Its methyl ether 5j and the 2,4-
dichloro analog 4f of 4j showed 20% higher activity.
Salicylic acid derivative 5b possessed even greater activ-
ity (percent inhibition of bacterial growth of 30%). It
was found that substituting a halogen (compound 5c)
to hydroxyl group (compound 5b) increased bioactivity
to 17%. The ether 6 of 4-sulfanylmethylpyridine showed
activity with a PI of 39%.
benzoic acid 7–11, 13. Salicylic acid derivative 29 with
a similar structure showed practically the same level of
activity and was the most active among the ortho-hy-
droxy compounds. Chloride 10 was less active than the
analogous non-halogenated product 7. It was found that
1,3-dioxolane derivatives are less active than the corre-
sponding keto-precursor compounds. Compounds 29
and 28 are notable exceptions. The latter showed the
highest target activity (82%) among all the 5-aryl-2-
thio-1,3,4-oxadiazoles derived from benzoic acids 1a–j.
In the case of the gallic acid derivatives, the presence of
a halogen (12, 14, 20) leads to a decrease in activity. The
maximum activity among 3,4,5-trimethoxy analogs was
observed for compound 17. It was noted that decreasing
the number of methoxy groups to one in the aromatic
ring of 5-aryl-1,3,4-oxadiazole 38 reduces its activity to
only 5% inhibition. The fluoro-substituted product 11
showed the highest activity among the derivatives of
3. Results and discussion
3.1. The search for pharmacophores (Ph) and
anti-pharmacophores (APh) by using ETM
The compounds under study (41 molecules) are shown
in Table 1. Molecules were classified as active (21 com-
pounds showing PI P 24), and inactive (20 compounds

F. Macaev et al. / Bioorg. Med. Chem. 13 (2005) 4842–4850
Table 2. Physical and analytical data of synthesized compounds 3a–41
4845
Compound
Mp (ꢁC) from EtOHa
Yield (%)
Molecular formula
Elemental analyses (%)
H, Calcd/Found
C, Calcd/Found
N, Calcd/Found
3a
3b
3c
3d
3ea
3f
113–117
147–150
150–152
Decomp.
117–118
Decomp.
264–266
136–140
129–130
Decomp.
219–222
121–214
210–211
Decomp.
Decomp.
177–178
234–236
Decomp.
175–176
Decomp.
73–75
94
82
92
92
87
88
91
95
94
80
90
68
87
91
86
84
72
81
79
85
87
91
92
76
93
78
89
91
92
89
79
92
80
89
79
79
87
88
84
92
88
90
71
68
73
81
65
88
69
71
67
68
82
81
76
77
81
76
88
C7H8N2O
61.75/61.52
55.26/55.19
39.10/39.02
63.98/63.91
49.28/49.22
41.00/40.96
55.26/55.22
57.82/57.77
59.99/59.92
53.09/53.02
53.92/53.88
49.47/49.52
37.37/37.48
56.23/56.27
45.18/45.22
38.89/38.94
49.47/49.52
51.91/51.96
54.04/54.12
49.24/49.29
51.91/51.96
39.87/39.67
51.05/51.09
56.81/56.70
64.85/64.88
61.53/6158
65.79/65.83
58.09/58.12
61.14/61.17
56.43/56.48
56.30/56.33
49.04/49.10
50.12/50.16
57.88/57.92
59.99/60.03
64.92/64.95
52.90/52.94
54.22/54.26
56.49/56.53
63.29/63.33
59.06/59.02
51.21/51.17
57.75/57.71
65.01/64.94
49.67/49.62
63.51/63.55
60.66/60.72
62.48/62.36
59.99/60.03
56.36/56.38
59.06/59.11
57.68/57.62
56.24/56.28
49.52/49.56
54.25/54.28
57.38/57.42
58.59/58.62
59.62/59.58
57.93/57.88
5.92/6.04
5.30/5.39
3.28/3.34
6.71/6.65
4.14/4.08
2.95/2.91
5.30/5.26
6.07/5.99
6.71/6.67
6.24/6.19
3.39/3.42
3.11/3.18
1.96/2.01
4.19/4.24
2.37/2.44
1.63/1.67
3.11/3.18
3.87/3.92
4.53/4.57
4.51/4.55
3.87/3.82
2.60/2.49
5.00/5.03
4.77/4.79
4.08/4.12
3.87/3.92
4.55/4.61
3.35/3.39
3.53/3.58
4.24/4.26
3.25/3.28
3.68/3.72
3.54/3.59
4.84/4.87
5.03/5.08
4.79/4.84
3.97/4.02
4.07/4.12
4.97/5.03
4.62/4.67
4.70/4.74
2.95/2.89
4.04/3.97
4.46/4.38
3.47/3.42
4.74/4.78
4.53/4.57
5.24/5.11
5.03/5.10
4.23/4.25
4.70/4.73
4.84/4.78
4.72/4.77
4.16/4.21
4.55/4.59
5.25/5.29
5.15/5.20
3.26/3.24
3.43/3.37
20.58/20.72
18.41/18.55
13.03/13.14
18.65/18.72
16.42/16.48
13.66/13.72
18.41/18.33
16.86/16.92
15.55/15.48
12.38/12.31
15.72/15.84
14.42/14.52
10.90/10.94
14.57/14.53
13.17/13.24
11.34/11.38
14.42/14.47
13.45/13.48
12.60/12.64
10.44/10.48
13.45/13.49
10.33/10.42
9.92/9.96
C7H8N2O2
C7H7BrN2O
C8H10N2O
C7H7ClN2O
C7H6Cl2N2O
C7H8N2O2
C8H10N2O2
C9H12N2O2
3g
3h
3i
3j
C10H14N2O4
C8H6N2OS
4a
4b
4c
4d
4e
4f
C8H6N2O2S
C8H5BrN2OS
C9H8N2OS
C8H5ClN2OS
C8H4Cl2N2OS
C8H6N2O2S
C9H8N2O2S
C10H10N2O2S
C11H12N2O4S
C9H8N2O2S
C9H7BrN2OS
C12H14N2O4S
C17H17N3O4S
C16H12N2O2S
4g
4h
4i
4j
5b
5c
5j
60–62
142–143
Decomp.
167–169
195–198
139–142
127–129
160–162
159–160
210–211
199
6
11.69/11.45
9.45/9.49
8.97/8.94
7
8
C
16H12N2O3S
C17H14N2O2S
16H11ClN2O2S
C16H11FN2O2S
19H17FN2O5S
C16H11N3O4S
19H17BrN2O5S
C19H16Cl2N2O5S
20H20N2O6S
C20H20N2O5S
25H22N2O5S
9
9.03/9.10
8.47/8.52
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28a
29
30
31
32
33
34
35
36
37
38
39
40
41
C
8.91/8.94
6.93/6.98
C
12.31/12.35
6.02/6.09
6.15/6.19
C
153–155
123–125
130–132
166–167
236–237
191–192
157–158
156–157
99–101
C
6.73/6.78
7.00/7.06
C
6.06/6.11
9.74/9.77
C19H17N3O7S
C19H17ClN2O5S
6.66/6.69
6.27/6.33
C
21H22N2O7S
C23H20N2O5S
C19H18N2O5S
C16H11BrN2O2S
6.42/6.48
7.25/7.32
127–129
133–135
Decomp.
134–136
129–130
147–149
Decomp.
98–99
7.47/7.41
7.48/7.54
C
18H15FN2O4S
C22H18N2O4S
18H15BrN2O4S
6.89/6.94
6.44/6.53
C
C18H16N2O3S
C18H16N2O4S
8.23/8.28
7.86/7.92
C20H20N2O4S
C20H20N2O5S
7.29/7.41
7.00/7.05
112–114
153–154
201–202
128–130
92–94
C19H17ClN2O4S
6.92/6.95
7.25/7.29
C
C
19H18N2O5S
20H20N2O6S
6.73/6.78
6.25/6.28
C21H21FN2O6S
21H21BrN2O6S
C21H21ClN2O6S
22H24N2O7S
C
5.50/5.52
6.03/6.08
102–103
119–121
148–149
177–178
227–229
C
6.08/6.12
6.51/6.55
C21H22N2O6S
C23H15BrN2O2S
6.05/6.12
15.90/15.96
C
17H12N4OS2
a Note. 3e from Et2O; 28 from benzene.
with PI 6 22). Conformational analysis and quantum
chemistry calculations were carried out by means of
the molecular mechanics method (MMP2) and the
semi-empirical quantum chemistry method (AM1),
respectively.27 Diagonal elements of the matrices called
electronic-topological matrices of contiguity (ETMC)

4846
F. Macaev et al. / Bioorg. Med. Chem. 13 (2005) 4842–4850
Table 3. 1H NMR spectral data of compounds 4a–39
Comp. d (ppm)
4a
4b
4c
4d
4e
4f
7.30–8.18 m (5H, arom), 12.4 s (1H, SH)
7.23–7.91 m (4H, arom), 9.68 s, 10.4 s (2H, SH, OH)
7.03–7.85 m (4H, arom), 10.9 s (1H, SH)
2.54 s (3H, Me), 7.49–7.84 m (4H, arom), 9.6 s (1H, SH)
7.26–7.96 m (4H, arom), 11.3 s (1H, SH)
7.38–7.98 m (3H, arom), 10.9 s (1H, SH)
4g
4h
4i
5.12 w s (1H, OH), 7.08 d, 7.18 d (4H, J = 8.46 Hz, arom), 11.4 s (1H, SH)
3.75 s (3H, Me), 7.01 d, 7.97 d (4H, J = 8.46 Hz, arom), 10.1 s (1H, SH)
1.35 t (3H, Me, J = 6.99 Hz), 4.11 q (2H, CH2, J = 6.88 Hz), 7.53 d, 7.79 d (4H, J = 8.86 Hz, arom), 14.5 s (1H, SH)
3.88 s (9H, 3 Me), 7.72–7.75 m (2H, arom), 13.1 s (1H, SH)
4j
5b
5e
5j
2.59 s (3H, Me), 6.89–7.82 m (4H, arom), 10.28 s (1H, OH)
2.53 s (3H, Me), 7.26–7.93 m (4H, arom)
2.68 s (3H, MeS), 3.76 s (3H, MeO), 3.90 s (6H, 2MeO), 7.18 s (2H, arom)
3.77 s (3H, Me), 3.81 s (6H, 2Me), 4.59 s (2H, CH2), 7.12–7.13 m (4H, arom), 7.48–8.61 m (4H, Py)
5.08 s (2H, CH2), 7.50–8.10 m (10H, arom)
6
7
8
5.13 s (2H, CH2), 6.92–8.12 m (9H, arom), 9.94 s (1H, OH)
2.50 s (3H, Me), 5.11 s (2H, CH2), 7.38–8.20 m (9H, arom)
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
5.18 s (2H, CH2), 7.43–8.12 m (9H, arom)
5.15 s (2H, CH2), 6.84–8.25 m (9H, arom)
3.81 s (3H, Me), 3.83 s (6H, 2Me), 5.10 s (2H, CH2), 7.16–8.15 m (6H, arom)
4.18 s (2H, CH2), 7.40–8.31 m (9H, arom)
3.78 s (3H, Me), 3.81 s (6H, 2Me), 5.19 s (2H, CH2), 7.19 s (2H, arom), 7.81 d, 8.01 d (4H, J = 8.84 Hz, arom)
3.77 s (3H, Me), 3.85 s (6H, 2Me), 4.78 s (2H, CH2), 7.31–7.92 m (6H, arom)
3.76 s (3H, Me), 3.81 s (9H, 3Me), 5.14 s (2H, CH2), 6.87–8.12 m (6H, arom)
2.46 s (3H, Me), 3.88 s (3H, MeO), 3.92 s (6H, 2MeO), 5.10 s (2H, CH2), 7.12 s (2H, arom), 7.31 d, 9.91 d (4H, J = 8.84 Hz, arom)
3.79 s (3H, Me), 3.82 s (6H, 2Me), 5.15 s (2H, CH2), 7.08–8.09 m (11H, arom)
3.71 s (3H, Me), 3.82 s (6H, 2Me), 5.08 s (2H, CH2), 7.11–8.43 m (6H, arom)
3.78 s (3H, Me), 3.83 s (6H, 2Me), 5.11 s (2H, CH2), 7.11 s (2H, arom), 7.52 d, 8.09 d (4H, J = 8.85 Hz, arom)
3.82 s (3H, Me), 3.90–3.91 m (12H, 4Me), 5.0 s (2H, CH2), 6.94–7.77 m (5H, arom)
3.72 s (3H, Me), 3.80 s (6H, 2Me), 5.29 s (2H, CH2), 7.18–8.16 m (9H, arom)
3.8 s (3H, Me), 3.90 s (6H, 2Me), 5.20 s (2H, CH2), 7.25–8.78 m (7H, arom)
7.50–8.10 m (11H, CHBr, arom)
3.58–4.29 m (6H, 3CH2), 6.90–8.11 m (8H, arom), 10.27 s (1H, OH)
3.72–4.29 m (6H, 3CH2), 6.88–8.03 m (11H, arom), 10.35 s (1H, OH)
3.74–4.18 m (6H, 3CH2), 6.96–7.80 m (8H, arom), 10.27 s (1H, OH)
3.70–4.15 m (6H, 3CH2), 7.00–7.88 m (10H, arom)
3.74–4.11 m (6H, 3CH2), 6.89–7.80 m (9H, arom), 10.26 s (1H, OH)
1.35 t (3H, Me, J = 7.1 Hz), 3.75–4.21 m (8H, 4CH2), 7.11 d, 7.90 d (4H, J = 8.84 Hz, arom), 7.26–7.55 m (5H, arom)
3.81 s, 3.82 s (6H, 2Me), 3.72–4.16 m (6H, 3CH2), 6.82–7.84 m (8H, arom)
3.35 s (3H, 2Me), 3.67–4.18 m (6H, 3CH2), 6.96–7.95 m (8H, arom)
3.73 s (3H, Me), 3.83 s (2H, CH2S), 3.66–4.17 m (4H, dioxolane), 6.56–7.88 m (8H, arom), 10.30 s (1H, OH)
3.70 s, 3.75 s (6H, 2Me), 3.94 s (2H, CH2S), 3.70–4.19 m (4H, dioxolane), 6.82–7.84 m (8H, arom), 10.34 s (1H, OH)
3.70 s (3H, Me), 3.89 s (6H, 2Me), 3.67–4.18 m (6H, 3CH2), 7.45–8.47 m (6H, arom)
3.65 s (3H, Me), 3.67 s (6H, 2Me), 3.54–4.16 m (6H, 3CH2), 7.40–8.21 m (6H, arom)
3.77 s (3H, Me), 3.82 s (6H, 2Me), 3.54–4.16 m (6H, 3CH2), 7.08–8.05 m (6H, arom)
3.82 s (3H, Me), 3.83–3.86 m (9H, 3Me), 3.54–4.16 m (6H, 3CH2), 7.26–8.0 m (6H, arom)
3.73 s (3H, Me), 3.84 s (5H, Me, CH2S), 3.50–4.22 m (4H, dioxolane), 6.83–7.95 m (8H, arom)
reflect one or more atomic properties (represented by a
separate value or by a vector of characteristics). Off-di-
agonal elements characterize bonds between pairs of
atoms, if they exist, or distances otherwise. Usually, only
the upper triangle of each matrix is used in calculations
due to the symmetry of bonds. Also, more than one
property can be taken for bonds. For the sake of simpli-
fication, the ETM calculations generally use only by one
property for atoms and bonds. In cases where there are
more than one property for atoms and bonds, the ETM
calculations can be repeated separately for each prop-
erty. In our case, effective charges on atoms are taken
as diagonal elements, and the values of WibergÕs index
represent off-diagonal elements corresponding to bonds;
if no bond, then off-diagonal elements are distances for
corresponding pairs of atoms.
The computational part of the ETM is a sequence of the
following steps:
• Conformational analysis,
• Quantum-chemistry calculations,
• ETMC formation,
• The search for structural features, responsible for a
compoundÕs activity/inactivity (the features are
referenced as pharmacophores/anti-pharmacophores,
correspondingly). To find pharmacophores, a tem-
plate active compound and the rest of the compound

F. Macaev et al. / Bioorg. Med. Chem. 13 (2005) 4842–4850
4847
set are compared as weighted graphs. To find anti-
pharmacophores, an inactive compound is used as a
template for the comparison. At the same time, the
molecular flexibility is taken into account by the com-
parison of atomic and bond weights.
phore (Ph) in the series under study is given by the fol-
lowing equation:
CAðPhÞ ¼ ðnA þ 1Þ=ðnA þ nIA þ 2Þ;
ð1Þ
where nA, nIA are numbers of active/inactive com-
pounds, respectively, that contain the Ph. For an anti-
pharmacophore (APh), (nIA+1) is to be used at place
of (nA+1) in Eq. 1. The remaining pharmacophores,
Ph2–Ph15, were found analogously, and the probabili-
ties of their realization in the class of active compounds
varied between 0.86 and 0.95.
The last two steps represent the essential part of the
ETM. The main advantages of the ETM are that its
molecular descriptions reflect a moleculeÕs electronic
and 3D conformational properties and do not depend
on the numbers and types of atoms.
Thus, for each template compound (active or inactive),
its ETMC was compared with the ETMCs of the rest
of the compounds in both classes of the series taken
for this study. The comparison resulted in a few com-
mon structural fragments for the two cases. The frag-
ments were found as submatrices of the ETMCs
corresponding to templates (they will be referenced as
electron-topological submatrices of contiguity, or
ETSCs). Consequently, all pharmacophores and anti-
pharmacophores found from the ETM calculations
form a system for the activity identification. For the ser-
ies studied, the system includes 15 pharmacophores and
10 anti-pharmacophores. From compound 11, the cho-
sen active template compound, an activity feature 1
(or the Ph1 pharmacophore) was found. It is given in
Figure 1 with the corresponding ETSC, which describes
electronic-topological characteristics of the fragment
(see Fig. 1).
To determine anti-pharmacophores, ETMCs of inactive
compounds were taken as templates. Fifteen anti-phar-
macophores, APh1–APh10, were found overall. The
ETSCs that correspond to APh1 are given in Figure 2
along with structures of the corresponding template
after which the anti-pharmacophore is found.
As seen from Figure 2, APh1 (based on the inactive tem-
plate 33) consists the nine atoms. APh1 is present in 10
inactive molecules and 1 active molecule, and the prob-
ability of its realization is 0.85.
When comparing the structures of the pharmacophores
and anti-pharmacophores, one should pay close atten-
tion to the differences in their spatial and electronic
characteristics. Thus, the pharmacophores and anti-
pharmacophores when considered as a whole play an
important role in activity predictions and the search for
new drugs. The set of activity/inactivity fragments found
as a result of this study forms a basis for the development
of a system for anti-tuberculosis activity prediction.
As seen from the Ph1 pharmacophore structure, it con-
sists of 7 atoms (C1, C7, C16–C19), which relate structur-
ally to the phenyl attached to the oxadiazole ring. The
submatrix given in Figure 1 is found after setting allow-
able limits for finding equivalent matrix elements. The
limits are d1 = 0.05 for its diagonal elements and
d2 = 0.15 for the off-diagonal ones. The pharmaco-
phore found from the ETM-calculations is realized in
nine active compounds, and the probability PA of its
realization in this class is about 0.91.
4. Neural network application
Artificial neural networks (ANNs) comprise a group of
methods that are increasingly being used in drug design
to study quantitative/qualitative SAR (QSAR). This ap-
proach is able to elucidate SARs and account for any
non-linear character of these relationships. Thus, this
method can be of significant interest in three-dimen-
sional (3D) QSAR studies.
One common criterion for structural methods (CA) that
evaluates the probability of occurrence of a pharmaco-
Figure 1. The Ph1 pharmacophore found relative to active molecule 11.

4848
F. Macaev et al. / Bioorg. Med. Chem. 13 (2005) 4842–4850
Figure 2. The AP1 anti-pharmacophore found relative to inactive molecule 33.
The architecture of the artificial supervised NN applied
to our study consists of three-layers, with five neurons in
one hidden layer. A single output node was used to code
activities of inhibitors. The bias neurons were presented
on the input and hidden layers. At least M = 200 inde-
pendent FFNNs were trained to analyze each set of vari-
ables. The values predicted for each analyzed case were
averaged over all M network predictions, and the means
were used to calculate statistical coefficients for targets.
Further details of the algorithm can be found
elsewhere.28,29
The LOO cross-validation procedure that supervises the
predictive performance of the ANN, has shown that
pruning algorithms31,32 can be used to optimize the
number of input parameters for the ANNs training
and to select the most significant ones. These algorithms
operate in a manner similar to step-wise multiple regres-
sion analysis and, at each step, exclude by one input
parameter that has been estimated as non-signifi-
cant.31–34 The pruning algorithms were used in the cur-
rent study to determine significant parameters from the
input data points of the analyzed molecules, as described
in the references.
The avoidance of overfitting/overtraining has been
shown to be an important factor for improving the pre-
dictive ability and correct selection of variables in the
FFNNs. The early stopping over ensemble (ESE) tech-
nique was used in the current study to achieve this
end. We used a subdivision of the initial training set into
two equal learning/validation subsets. The first set was
used to train the neural network, while the second was
used to monitor the training process measured by root
mean square error. An early stopping point determined
as a best fit of the network to the validation set was used
to stop the neural network learning. Thus, statistical
parameters calculated at the early stopping point were
used. The training was terminated by limiting the net-
work run to 10,000 epochs (total number of epochs) or
after 2000 epochs (local number of epochs) following
the last improvement of root-mean-square error in the
early stopping point. The root-mean-square error E
was computed as a criterion for network learning to
determine the stop points of the training procedure.
The quality of the model was tested by the leave-one-
out (LOO) cross-validation procedure carried out on
the training set with q2 value (Cramer et al.30) and is de-
fined as
As the second stage, we examined whether all 25 molec-
ular fragments (pharmacophores and anti-pharmaco-
phores) are relevant for the anti-tuberculosis activity
prediction. For these 25 fragments and 41 compounds
in the series studied, a table of weights was formed as
consisting of 41 rows of 25 elements being 1 (the
fragment is present in the corresponding molecular
structure) or 0 (the fragment is absent).
Application of pruning methods allowed for the selec-
tion of only five of the most appropriate parameters
responsible for the anti-tuberculosis inhibitory activity.
q2 ¼ ðSD ꢀ pressÞ=SD.
ð2Þ
In Eq. 2, SD represents the variance of a target value
relative to its mean and ÔpressÕ is the average squared
errors of predicted values obtained from the LOO
procedure.
Figure 3. Theoretical and experimental % inhibitions results.

F. Macaev et al. / Bioorg. Med. Chem. 13 (2005) 4842–4850
4849
The reduction of the initially selected seven parameters
to five did not decrease the cross-validated q2 coeffi-
cient (0.80 0.01). As illustrated in Figure 3, there
was a good concordance between the predicted and
experimental values of the anti-tuberculosis activities.
the 1,3,4-oxadiazoles 4a–j according to a known
procedure.26
6.1.2. General procedure for the preparation of the
thioethers 5–23. Et3N (0.303 g, 3 mmol) was added to
a suspension of (3 mmol) 1,3,4-oxadiazole in 15 mL of
acetone. To the resulting homogeneous solution, the
appropriate x-phenacyl bromide or MeI (3 mmol) was
added as one portion. The reaction mixture was stirred
for 1 h at 40 ꢁC. When the reaction was completed
(TLC control), the precipitate was filtered and crystal-
lized from the appropriate solvent.
This result confirms our hypothesis in that both
pharmacophores and anti-pharmacophores must be
considered as parameters when building a model for
anti-tuberculosis activity prediction.
5. Conclusions
6.1.3. Synthesis of 2-bromo-1-phenyl-2-(5-phenyl-1,3,4-
oxadiazol-2-ylsulfanyl)-1-ethanone 24. To a solution of 7
(2.96 g, 0.01 mol) in CCl4 (30 mL) and NBS (1.96 g,
0.011 mol) 200 mg of benzoyl peroxide were added.
The mixture was refluxed for 2 h (TLC control). The
reaction mixture was purified over a short plug of silica
(5 g). After evaporation of the solvent under reduced
pressure, the crude product 24 (yield 3.37 g, 90%) was
used without purification in the next step. IR (m/cmꢀ1):
the sample (24) in liquid paraffin showed absorption in
the region 650 (BrAC) and 1780 (C@O).
A series of 2,5-disubstituted-1,3,4-thiadiazoles were de-
signed and synthesized, and these compounds were
screened for anti-tuberculosis activity against M. tuber-
culosis H37Rv. A systematic SAR study was performed
through application of the ETM approach to the series
of compounds relative to their experimentally measured
anti-tuberculosis activity. Data obtained from confor-
mation and quantum-chemistry calculations were used
to form electronic-topological matrices. These matrices
were effectively used in the search for a system of phar-
macophores and anti-pharmacophores capable of effec-
tive separation of compounds from the examination
set into groups of active and inactive compounds. Low
activity molecules are badly responsive to the activity
prognostication, because they form a buffer zone con-
sisting of compounds that can include both pharmaco-
phores and anti-pharmacophores. The model
developed in this study is supposed to be applied to
the design, preparation and screening of new com-
pounds of similar structure in order to further test and
optimize the model with the eventual goal of preparing
new anti-tubercular agents.
6.1.4. General procedure for the preparation of the
thioethers 25–39. 5-Aryl-1,3,4-oxadiazole (3 mmol) in
5 mL DMF and 0.62 g (4.5 mmol) of potassium carbon-
ate were placed into a 3-necked round bottom flask,
equipped with thermometer, stirrer, and condenser.
The mixture was stirred for 5 min, and of 2-bromom-
ethyl-2-phenyl-1,3-dioxolane (3 mmol) was added. The
reaction temperature was increased to 130 ꢁC , and the
mixture was stirred for 3 h (TLC control). The cooled
reaction mixture was poured into 50 mL of water while
stirring. The resulting precipitate was filtered, washed
with water, dried, and crystallized from the specified
solvent (Table 2).
6. Experimental
6.1.5. Synthesis of 3-(4-bromophenyl)-1-phenyl-2-(5-phen-
yl-1,3,4-oxadiazol-2-ylsulfanyl)-2-propen-1-one 40. To a
solution of 7 (2.96 g, 0.01 mol) in toluene (50 mL) and
4-bromobenzaldehyde (1.85 g, 0.01 mol), 50 mg of pyr-
rolidine and 50 mg acetic acid were added. The mixture
was refluxed with Dean-Stark apparatus for 15 h. The
reaction mixture was washed with water (3 · 50 mL)
and dried (MgSO4). After removal of the toluene, the
solid was crystallized from EtOH. IR (m/cmꢀ1): 650
6.1. Chemical methods
All the solvents used were reagent quality, and all com-
mercial reagents were used without additional purifica-
tion. Removal of all solvents was carried out under
reduced pressure. Physical and analytical data of the
synthesized compounds 3a–41 are given in Table 2. Ana-
lytical TLC plates were Silufolꢂ UV-254 (Silpearl on
aluminium foil, Czecho-Slovakia). IR spectra were re-
corded on a Specord 75 IR instrument (suspension in
Vaseline oil). Melting points (uncorrected) were deter-
1
(BrAC) and 1675 (C@O). H NMR (d, ppm): 7.35 s
(1H, CH@), 7.40–8.0 m (14H, arom). MS, m/z (relative
intensity %): 462 [M]+ (25), 359(5), 285(35), 212(15),
178(35), 145(50), 105(90), 89(15), 77(100).
1
mined on a Boetius apparatus. H NMR spectra were
recorded for d6-DMSO solution on a Varian XL-400
spectrometer (399.95 MHz) and
a
Bruker AC-80
6.1.6. Synthesis of 4-phenyl-5-(5-phenyl-1,3,4-oxadiazol-
2-ylsulfanyl)-1,3-thiazol-2-amine 41. To a solution of 24
(1.87 g, 0.005 mol) in EtOH (20 mL), thiourea (0.76 g,
0.005 mol) in 10 mL of EtOH was added. The mixture
was refluxed for 1 h. After evaporation of EtOH under
reduced pressure, the residue was suspended in a 10%
solution of NaHCO3 and heated to 50 ꢁC. The crystal-
line product was filtered, washed with water, and crys-
tallized from EtOH to give the final 2-aminothiazole
41 (1.54 g, 88%) as a white powder. IR (m/cmꢀ1):
(80 MHz) apparatus and are given in Table 3. Mass
spectra were recorded on a Finnigan MAT.INCOS-50
instrument (ionizing voltage 70 eV).
6.1.1. Precursors of compounds 7–23, 25–39. The starting
x-phenacyl bromides and 2-bromomethyl-2-phenyl-1,3-
dioxolanes were obtained according to the published
methods.35–39 The arylhydrazines 3a–j have been
prepared from esters 2a–j and were used to prepare

4850
F. Macaev et al. / Bioorg. Med. Chem. 13 (2005) 4842–4850
1
15. Collins, L.; Franzblau, S. G. Antimicrob. Agents Chemo-
ther. 1997, 41, 1004–1009.
16. Dimoglo, A. S.; Sim, E.; Shvets, N. M.; Ahsen, V. Mini
Rev. Med. Chem. 2003, 3, 293.
3450 (NH2). H NMR (d, ppm): 5.97 w s (1H, CH@),
7.14–8.21 m (10H, arom). MS, m/z (relative intensity
%): 352 [M]+ (100), 207(30), 165(70), 145(70), 133(5),
121(20), 103(20), 89(20), 77(50), 60(5).
17. Dimoglo, A. S.; Shvets, N. M.; Tetko, I. V.; Livingstone,
D. J. Quant. Struct.-Act. Relat. 2001, 20, 31.
18. Dimoglo, A. S. Khim.- pharm. Zhur. 1985, 4, 438.
19. Bersuker, I. B.; Dimoglo, A. S. The electron-topological
approach to the QSAR problem. In Reviews in Computa-
tional Chemistry; Lipkowitz, K. B., Boyd, D. B., Eds.;
VCH: New York, 1991, Chapter 10.
20. Shvets, N. M.; Dimoglo, A. S. Nahrung 1998, 42, 364.
21. Dimoglo, A. S.; Beda, A. A.; Shvets, N. M.; Gorbachov,
M. Yu.; Kheifits, L. A.; Aulchenko, I. S. N. J. Chem. 1995,
19, 149.
6.1.7. In vitro evaluation of the anti-mycobacterial
activity. Primary screening was conducted at 6.25 lg/
mL against M. tuberculosis H37Rv (ATCC 27294) in
BACTEC 12B medium using a broth microdilution
assay, the Microplate Alamar Blue Assay (MABA)
according to the reported method.15
Acknowledgments
22. Dimoglo, A. S.; Vlad, P. F.; Shvets, N. M.; Coltsa, M. N.
N. J. Chem. 2001, 25, 283.
The authors gratefully acknowledge funding through a
grant from the Moldavian–U.S. Bilateral Grants Pro-
gram of the U.S. Civilian Research and Development
Foundation (Grant MC2-3007). The authors also
acknowledge the Tuberculosis Antimicrobial Acquisi-
tion and Coordinating Facility provided through the
U.S. National Institutes of Health (Contract N01-AI-
95364) for providing the anti-tubercular activity data.
23. Shvets, N. M.; Dimoglo, A. S. The electronic-topological
method (ETM): its further development and use in the
problems of SAR study. In Molecular Modeling and
Prediction of Bioactivity; Gundertofte, K., Jorgensen, F.
S., Eds.; Kluwer Academic/Plenum Publishers: New York,
1999, p 418.
24. Kohonen, T. Self-organisation Maps; Springer-Verlag:
Berlin, 1995.
25. Simon, V.; Gasteiger, J.; Zupan, J. A. J. Am. Chem. Soc.
1993, 115, 9148–9159.
26. Rusu, G. G.; Gutu, E. E.; Barba, N. A. Russ. J. Org.
Chem. 1995, 31, 1721.
References and notes
27. Frisch, M. J.; Trucks, G. W.; Schlegel, H. B. et al.
Gaussian 98, Revision A.9, Gaussian, Inc., Pittsburgh,
PA, 1998.
28. Kovalishyn, V. V.; Tetko, I. V.; Luik, A. I.; Kholodovych,
V. V.; Villa, A. E. P.; Livingstone, D. J. J. Chem. Inf.
Comput. Sci. 1998, 38, 651–659.
1. World Health Organization. In Anti-tuberculosis Drug
Resistance in the World; The WHO/IUATLD global
project on anti-tuberculosis drug resistance surveillance,
1997.
2. Bastian, I.; Colebunders, R. Drugs 1999, 58, 633–661.
3. Butler, D. Nature 2000, 406, 670–672.
4. Bodiang, C. K. Scott. Med. J. 2000, 45, 25.
5. Van Scoy, R. E.; Wilkowske, C. J. Mayo Clin. Proc. 1999,
74, 1038.
6. Long, R. Can. Med. Assoc. J. 2000, 163, 425–428.
7. Gazdic, A. Med. Arch. 1998, 52, 207.
8. Pozniak, A. Int. J. Tuberc. Lung Dis. 2000, 4, 993.
9. Poroikov, V. V.; Filimonov, D. A.; Borodina, Yu. V.;
Lagunin, A. A.; Kos, A. J. Chem. Inf. Comput. Sci. 2000,
40, 1349–1355.
10. Stepanchikova, A. V.; Lagunin, A. A.; Filimonov, D. A.;
Poroikov, V. V. Curr. Med. Chem. 2003, 10, 225.
11. Poroikov, V. V.; Filimonov, D. A. J. Comput. Aided Mol.
Des. 2003, 16, 819–824.
12. Geronikaki, A. A.; Dearden, J. C.; Filimonov, D.;
Galaeva, I.; Garibova, T. L.; Gloriozova, T.; Krajneva,
V.; Lagunin, A.; Macaev, F. Z.; Molodavkin, G.; Poroi-
kov, V. V.; Pogrebnoi, S. I.; Shepeli, F.; Voronina, T. A.;
Tsitlakidou, M.; Vlad, L. J. Med. Chem. 2004, 47, 2870–
2876.
29. Tetko, I. V.; Livingstone, D. J.; Luik, A. I. J. Chem. Inf.
Comput. Sci. 1995, 35, 826–833.
30. This coefficient was introduced as r2 in Cramer III, R. D.;
Patterson, D. E.; Bunce, J. D. J. Am. Chem. Soc. 1988,
110, 5959–5967. However, in order to avoid confusion
with the analogous conventional r2 = R R value the new
*
designation q2 was recommended in Cramer III, R. D.; De
Priest, S. A.; Patterson, D. E.; Hecht, P. The Developing
Practice of Comparative Field Analysis. In 3D QSAR in
Drug Design: Theory Methods and Applications; Kubinyi,
H., Ed.; ESCOM: The Netherlands, 1993, pp 443–486.
31. Tetko, I. V.; Villa, A. E. P.; Livingstone, D. J. J. Chem.
Inf. Comput. Sci. 1996, 36, 794–803.
32. Tetko, I. V. J. Chem. Inf. Comput. Sci. 2002, 42, 717–728.
33. Tetko, I. V.; Kovalishyn, V. V.; Livingstone, D. J. J. Med.
Chem. 2001, 44, 2411–2420.
34. Tetko, I. V.; Kovalishin, V. V. http://146.107.217.178/lab/
asnn/index.html.
35. Ma, Y.; Liu, H.; Chen, L.; Cui, X.; Zhu, J.; Deng, J. Org.
Lett. 2003, 5, 2103.
13. Geronikaki, A.; Babaev, E.; Dearden, J.; Dehaen, W.;
Filimonov, D.; Galaeva, I.; Krajneva, V.; Lagunin, A.;
Macaev, F.; Molodavkin, G.; Poroikov, V.; Pogrebnoi,
S.; Saloutin, V.; Stepanchikova, A.; Stingaci, E.; Tkach,
N.; Vlad, L.; Voronina, T. Bioorg. Med. Chem. 2004, 12,
6559–6568.
36. Kim, Y. H.; Cheong, C. S.; Lee, S. H.; Jun, S. J.; Kim, K.
S.; Cho, H. S. Tetrahedron Asymmetry 2002, 13, 2501–
2508.
37. Adhikari, M. V.; Samant, S. D. Ultrason. Sonochem. 2002,
9, 107.
38. Camps, P.; Farres, X. Tetrahedron Asymmetry 1995, 6,
1283–1294.
39. Sket, B.; Zupan, M. Synth. Commun. 1989, 19, 2481–2487.
14. Oruc¸, E.; Rollas, S.; Kandemirli, F.; Shvets, N.; Dimoglo,
A. S. J. Med. Chem. 2004, 47, 6760–6767.