Synthesis and structure-antibacterial activity relationship investigation of isomeric 2,3,5-substituted perhydropyrrolo[3,4-d]isoxazole-4,6-diones

Article abstract of DOI:10.1016/j.bmc.2007.01.029

The synthesis of 2,3,5-substituted perhydropyrrolo[3,4-d]isoxazole-4,6-diones (44 compounds) has been accomplished by the cycloaddition reaction of N-methyl-C-arylnitrones with N-substituted maleimides. The compounds were screened for their antibacterial activities and most of them exhibited activity against Enterococcus faecalis (ATCC 29212) and Staphylococcus aureus (ATCC 25923). cis-3a and cis-3d were found fairly effective against E. faecalis (ATCC 29212) and S. aureus (ATCC 25923) with MIC values of 25 and 50 μg/ml. With the changes of cis isomers of the compounds to trans, their antibacterial activities also changed against the bacteria studied. First, pharmacophoric fragments had been calculated in accordance with the rules of the electronic-topological method (ETM). Next, both active compounds and pharmacophores had been projected to the nodes of Kohonen's self-organizing maps (SOM) to obtain the weights of pharmacophore fragments as numerical descriptors, that were used after this for the associative neural networks (ASNN) training. A model for the activity prediction was developed as the result of training the ASNNs.

Full text of DOI:10.1016/j.bmc.2007.01.029

Bioorganic & Medicinal Chemistry 15 (2007) 2322–2333
Synthesis and structure–antibacterial activity relationship
investigation of isomeric 2,3,5-substituted
perhydropyrrolo[3,4-d]isoxazole-4,6-diones
Hikmet Agirbas,a,* Selahaddin Guner,a Fatma Budak,b Sema Keceli,b Fatma Kandemirli,a
Nathaly Shvets,c Vasyl Kovalishynd and Anatholy Dimogloc,e,
*
aDepartment of Chemistry, Kocaeli University, Izmit, Turkey
bDepartment of Microbiology, Faculty of Medicine, Kocaeli University, Izmit, Turkey
cGebze Institute of Technology, Cayirova, Gebze 41400, Kocaeli, Turkey
dBiomedical Department, Institute of Bioorganic Chemistry & Petrochemistry, Kyiv 02660, Ukraine
eInstitute of Applied Physics, Academy of Sciences of Moldova, MD-2028, Kishinev, Republic of Moldova
Received 8 November 2006; revised 4 January 2007; accepted 17 January 2007
Available online 19 January 2007
Abstract—The synthesis of 2,3,5-substituted perhydropyrrolo[3,4-d]isoxazole-4,6-diones (44 compounds) has been accomplished by
the cycloaddition reaction of N-methyl-C-arylnitrones with N-substituted maleimides. The compounds were screened for their anti-
bacterial activities and most of them exhibited activity against Enterococcus faecalis (ATCC 29212) and Staphylococcus aureus
(ATCC 25923). cis-3a and cis-3d were found fairly effective against E. faecalis (ATCC 29212) and S. aureus (ATCC 25923) with
MIC values of 25 and 50 lg/ml. With the changes of cis isomers of the compounds to trans, their antibacterial activities also changed
against the bacteria studied. First, pharmacophoric fragments had been calculated in accordance with the rules of the electronic-
topological method (ETM). Next, both active compounds and pharmacophores had been projected to the nodes of Kohonen’s
self-organizing maps (SOM) to obtain the weights of pharmacophore fragments as numerical descriptors, that were used after this
for the associative neural networks (ASNN) training. A model for the activity prediction was developed as the result of training the
ASNNs.
ꢀ 2007 Published by Elsevier Ltd.
1. Introduction
For structural characterization, we have recently pre-
pared isomeric 2,3,5-substituted perhydropyrrolo [3,4-
d]isoxazole-4,6-diones by the cycloaddition reactions
of N-methyl-C-arylnitrones to N-substituted malei-
mides.13,14 In view of the biological interest shown in
these compounds, we have decided to synthesize a wide
range of isoxazole derivatives to screen the antibacterial
activities against various bacteria.
Over the past decades, the frequency of resistance in
antimicrobial agents has increased dramatically.1 There-
fore, this places new emphasis on the search for alterna-
tive substances which are effective against organisms
resistant to currently available drugs. Cycloaddition
of nitrones to a variety of unsaturated systems has
been exploited to synthesize isoxazoline/isoxazolidine
ring.2–4 Recently, useful antiinflammatory,5 immuno-
suppressive6 and antibacterial7–12 properties have been
ascribed to molecules possessing such heterocyclic
functionalities.
The other goal of the present study was the search of the
most potent and selective antibacterial preparations
against Enterococcus faecalis (ATCC 29212) and
Staphylococcus aureus (ATCC 25923) by applying the
combined approach that is the known electronic-
topological method (ETM) followed by the artificial
neural networks application (in short, ETM–ANNs
approach) to the ETM results. Both ETM and ANN
approaches have already been successfully applied but
separately to a wide enough variety of tasks related to
Keywords: Isoxazole; Structure–antibacterial activity relationship;
Electronic-topological method (ETM); ETM–ANN application.
*
Corresponding authors. E-mail addresses: agirbas@kou.edu.tr;
dimoglo@gyte.edu.tr
0968-0896/$ - see front matter ꢀ 2007 Published by Elsevier Ltd.
doi:10.1016/j.bmc.2007.01.029

H. Agirbas et al. / Bioorg. Med. Chem. 15 (2007) 2322–2333
2323
the structure–activity relationship (SAR) investiga-
tion.15–21
work to be compared with the results of the method
used and to control its reliability.
ANNs are of significant interest for QSAR studies.22 If
the dependencies between analysed descriptors and
molecular parameters are non-linear, the neural net-
works can produce more accurate models than linear
regression methods do. As an example, the feed-forward
neural networks (FFNN) trained with the back propa-
gation algorithm23,24 are widely used to perform differ-
ent chemical calculations. This can be also very
important for such practical applications as design of
new compounds possessing desirable properties. In our
case, ANNs provide more fast and easy implementation
of the activity prediction for new potential drugs,
because this prediction is not fully automated in the
ETM approach. Meantime, ANNs are capable of pro-
viding estimation of a compound’s activity immediately
from the structural description of the compound.
2.2. Antibacterial activity investigation
The antibacterial activities of 44 compounds were deter-
mined by using broth microdilution susceptibility test
outlined by the National Committee for Clinical Labo-
ratory Standards.43 Minimal inhibitory concentrations
for each compound were investigated against E. faecalis
(ATCC 29212) and S. aureus (ATCC 25923). For broth
microdilution procedures, sterile, disposable, multiwell
microdilution plates (96 U-shaped wells) were used.
The stock solutions were prepared in dimethylsulfoxide
(DMSO; Sigma), and DMSO had no effect on the
microorganisms in the concentrations studied.
2.3. Dilutions of the compounds
All of the dilutions were done with Mueller–Hinton
Broth (Oxoid) in the wells of microdilution plates. The
concentrations of 30 compounds tested were 1600, 800,
400, 200, 100, 50, 25, 12.5, 6.25, 3.12, 1.56, 0.78 and
0.39 lg/ml (the concentrations are given as lmol/ml in
Table 1). The highest concentration of eleven com-
pounds (cis-3l, cis-3p, cis-3r, cis-3s, cis-3t, cis-3u, trans-
3f, trans-3d, trans-3m, trans-3o, trans-3q) was 1000
lg/ml, therefore their tested concentrations were pre-
pared as follows: 1000, 500, 250, 125, 62.5, 31.2, 15.6,
7.8, 3.9, 1.95 and 0.975 lg/ml. Ampicillin (Fako, Istan-
bul, Turkey) was used as a reference compound.
2. Results and discussion
2.1. Chemistry
1,3-Dipolar cycloaddition reactions of alkenes to nitro-
nes often give a pair of diastereomeric isoxazolidines.25
In these reactions, a small structural change in the nitrone
can lead to a significant change in the stereoselectivity of
the cycloaddition.26 Regiochemical and stereochemical
courses of 1,3-dipolar cycloaddition of N-alkyl-C-
phenylnitrones to alkenes have been described in the
literature.27–41
2.4. Inoculum preparation
After diluting the compounds, standardized inoculum of
each bacterium (0.5 Mc Farland standard unit, 1 · 108
CFU/ml; colony forming unit/ml) was prepared. Then,
the compounds were diluted once more (1/10), and final
concentrations became 1 · 107 CFU/ml. Five microliters
from each dilution was placed into each well containing
100 ll of dilutions of compounds so that each well con-
tained 5 · 105 CFU/ml of inoculum. All the inoculated
plates were incubated at 35 ꢁC for 16–20 h. The lowest
concentration of compounds that prevents visible
growth was considered to be the minimal inhibitory con-
centration (MIC). MIC values are given in Table 1.
Ampicillin was used as reference antimicrobial reagent
to compare its parameters with the data that result from
the method applied in this work and to control the reli-
ability of the latter. Data on the antibacterial activity of
the compounds demonstrate that structural modifica-
tions can affect the antibacterial activity of the tetra-
hydropyrroloisoxazole skeleton.
In this work, a wide range of substituents have been
used for the cycloaddition of N-methyl-C-substituted
phenylnitrones to N-methyl and N-phenylmaleimides
(see Scheme 1). The stereochemical assignment of the
cis- and trans-isomers is made on the basis of the mag-
nitude of the Hb–Hc coupling constants. The cis-isomers
of isoxazolidine give rise to larger Hb–Hc coupling con-
stants (J ꢀ 6–8 Hz) than those observed for the trans-
isomers (J ꢀ 2–5 Hz).28,35,42 In the NMR spectra, the
Hc protons of the cis-isomers have been observed as
multiplets together with Hb protons at about 3.8–
4.00 ppm, whereas the Hc protons of the trans-isomers
appeared as very broad peaks at about 4.3–4.5 ppm.
From the cycloaddition of N-methyl-C-substituted nit-
rones (which carry electron-releasing substituents) to
N-methylmaleimide, we obtained major cis-isomers
along with minor trans-isomers. The same reaction of
cycloaddition but for nitrones (which carry electron-
withdrawing substituents) gave about 50:50 mixtures
of cis/trans isomers. On the other hand, the cycloaddi-
tion of all nitrones with N-phenylmaleimide gave both
isomers with 50:50 mixtures. The structures of all newly
synthesized compounds were identified by 1H NMR, IR
and micro analyses. The data are given in the experi-
mental section. The data on the antibacterial activity
(MIC values) are given in Table 1, which also contains
ampicillin results for the microorganisms used in this
All tested compounds exhibited expressive antibacterial
activity against E. faecalis and S. aureus. However, these
activities were lesser than that observed with Ampicillin.
cis-3a and cis-3d were found fairly effective against
E. faecalis and S. aureus with MIC values of 25 and
50 lg/ml, respectively. It was observed that if cis-isomers
change to trans-isomers, the corresponding activities
also change against E. faecalis and S. aureus. In general,
trans-3a, d, f, g, j, m, o, q, r, t, u and cis-3a, d, f, h, l, n, r,

2324
H. Agirbas et al. / Bioorg. Med. Chem. 15 (2007) 2322–2333
O
Ha
R'
Ha
O
R'
O
H
O
N
O
N
R
R''
N
+
+
R
R
N+ R'
N
N
Hc
R''
Hb
Hc
R''
Hb
O-
O
O
O
1a (R':Me)
1b (R':Ph)
2a(R'':Me)
2b(R'':Ph)
Cis-3a-v
Trans-3a-v
Comp. 3
R
R'
R'' Cis/Trans ratio
a
b
c
d
e
f
g
h
i
j
k
l
m
n
o
p
q
r
p-Et2NC6H4
p-Me2NC6H4
p-MeOC6H4
4-BzO-3-MeOC6H3 Me
p-MeC6H4
C6H5
m-ClC6H4
m-BrC6H4
p-ClC6H4
p-BrC6H4
p-NO2C6H4
p-Et2NC6H4
p-Me2NC6H4
p-MeOC6H4
4-BzO-3-MeOC6H3 Me
p-MeC6H4
C6H5
m-BrC6H4
p-ClC6H4
p-BrC6H4
p-NO2C6H4
p-Me2NC6H4
Me
Me
Me
Me
Me
Me
Me
Me
Me
Me
Me
Me
Me
Me
Ph
Ph
Ph
Ph
Ph
Ph
Ph
Ph
Ph
2.03
1.96
2.82
2.48
1.24
0.79
1.00
1.07
1.09
1.12
1.07
1.11
0.92
1.11
1.23
0.89
0.85
0.93
0.97
1.00
0.92
1.09
Me
Me
Me
Me
Me
Me
Me
Me
Me
Me
Me
Me
Me
Me
Me
Me
Ph
s
t
u
v
Ph
Me
Scheme 1. The synthesis of compounds 3a–3v.
s, t, u and v were found effective to inhibit the growth of
the bacteria with the MIC values between 25 and 250
lg/ml.
The language reflects the discrete nature of compounds
that are viewed as consisting of atoms some of which
are chemically bonded. Labelled graphs appeared to
be the most appropriate mathematical counterparts of
chemical structures and relationships on their atoms
and bonds. As known, a graph’s representative is a
matrix of the order n · n, where n is the number of the
graph’s vertices. Therefore, the ETM proposes electron-
ic-topological matrices of contiguity (ETMC) to be its
own, very special language, for chemical compounds’
description. Bonds have no orientation, thus the matri-
ces are symmetrical relative to their left diagonal, and
it is enough to have only the right upper triangle of
any such matrix along with its diagonal.
To complete the study, the structure–activity relation-
ship study for new compounds as antibacterial agents
is done as well. This gave rise to the identification of sev-
eral potent broad-spectrum antibacterial compounds.
2.5. Brief review of the combined ETM–ANNs approach
The aim of the ETM approach is to find molecular frag-
ments (pharmacophores) common for the structures of
all active compounds (for a fixed activity) and absent
in all inactive compounds possessing similar structures.
After this, the fragments found are used to predict the
activity for newly synthesized compounds. However,
the procedure of the activity prediction is difficult en-
ough and needs the expert’s participation. To make
the procedure easier and faster, the ETM application
was followed by the ANNs application (with unsuperv-
ized and supervized learning algorithms), and, as a
whole, the calculations were named the combined
ETM–ANNs approach.
To begin the ETM-study, one must have a representa-
tive series of compounds (a few tens of compounds, at
least, but the more is the better). The measured activity
can be either qualitative (i.e., active/inactive, in which
case there are two classes for comparison) or quantita-
tive (in which case there can be more than two classes).
Ideally, a half of compounds possessing similar struc-
tures should be inactive. The main steps of the ETM-
study are given below:
Since details of the ETM can be found in the litera-
ture,44–47 we only give here the most distinguished prop-
erties of the ETM relative to other methods used in
diverse SAR studies. ETM belongs to the so-called
structural methods. So, the main part of the method is
a language for the compound structure description.
1. Calculate the values of spatial and electron character-
istics for all atoms and bonds of each compound from
the initial selection.
2. Form the corresponding ETMC (see Fig. 1) for each
molecular structure by choosing appropriate values
from the data calculated; usually, they are charges

H. Agirbas et al. / Bioorg. Med. Chem. 15 (2007) 2322–2333
2325
Table 1. Minimal antibabacterial inhibitory concentration (lmol/ml) and data on the experimental versus theoretical activities of compounds under
study
Compound
E. faecalis ATCC 29212
Exp.
Theor.
S. aureus ATCC 25923
Exp.
Theor.
Ampicillin
cis-3a
0.002
0.08
0.63
1.38
5.53
>5.79
>5.79
0.13
0.65
>6.15
6.15
1.62
1.02
2.85
1.43
0.62
1.23
1.43
1.43
1.23
0.62
0.69
5.49
0.66
4.22
>4.55
0.71
1.18
1.18
>3.6
1.13
3.10
1.24
5.19
1.62
0.65
0.52
0.73
1.17
0.65
0.52
0.71
0.28
0.28
1.14
+
+
+
ꢁ
ꢁ
ꢁ
ꢁ
+
+
ꢁ
ꢁ
ꢁ
+
ꢁ
ꢁ
+
ꢁ
ꢁ
ꢁ
ꢁ
+
+
ꢁ
+
ꢁ
ꢁ
+
ꢁ
ꢁ
ꢁ
ꢁ
ꢁ
ꢁ
ꢁ
ꢁ
+
+
+
ꢁ
+
+
+
+
+
ꢁ
+
ꢁ
ꢁ
ꢁ
ꢁ
ꢁ
ꢁ
+
ꢁ
ꢁ
ꢁ
ꢁ
+
+
ꢁ
+
ꢁ
ꢁ
ꢁ
+
+
+
ꢁ
+
ꢁ
ꢁ
ꢁ
ꢁ
ꢁ
ꢁ
ꢁ
ꢁ
ꢁ
ꢁ
ꢁ
+
+
+
ꢁ
+
+
+
+
+
ꢁ
0.002
0.08
0.63
1.38
1.38
5.79
1.45
0.13
0.33
6.15
3.07
0.81
0.51
1.43
0.71
1.23
0.62
1.43
1.43
2.46
0.31
1.37
2.75
0.33
4.22
4.55
0.36
0.59
1.18
3.60
0.28
3.10
0.62
>5.19
0.41
0.32
0.13
0.36
1.17
0.32
0.13
0.35
0.57
0.57
0.28
+
+
+
ꢁ
ꢁ
ꢁ
ꢁ
+
+
ꢁ
ꢁ
+
+
ꢁ
+
ꢁ
+
ꢁ
ꢁ
ꢁ
+
ꢁ
ꢁ
+
ꢁ
ꢁ
+
+
ꢁ
ꢁ
+
ꢁ
+
ꢁ
+
+
+
+
ꢁ
+
+
+
+
+
+
+
+
+
+
ꢁ
ꢁ
ꢁ
+
ꢁ
ꢁ
ꢁ
ꢁ
+
ꢁ
+
ꢁ
+
ꢁ
ꢁ
ꢁ
+
ꢁ
ꢁ
+
ꢁ
ꢁ
ꢁ
+
ꢁ
ꢁ
+
ꢁ
+
ꢁ
+
+
+
+
ꢁ
ꢁ
+
+
+
+
+
trans-3aa
cis-3b
trans-3b
cis-3c
trans-3ca
cis-3d
trans-3d
cis-3e
trans-3ea
cis-3fa
trans-3f
cis-3g
trans-3g
cis-3ha
trans-3h
cis-3i
trans-3i
cis-3ja
trans-3j
cis-3k
trans-3k
cis-3la
trans-3l
cis-3ma
trans-3m
cis-3n
trans-3n
cis-3o
trans-3o
cis-3p
trans-3pa
cis-3q
trans-3q
cis-3ra
trans-3r
cis-3sa
trans-3s
cis-3t
trans-3ta
cis-3u
trans-3u
cis-3v
trans-3v
a Test set compounds.
for diagonal elements (qi) and the Wiberg’s indices for
bonds (Wij); otherwise, distances (Rij) are taken for
non-bonded atoms.
5. Estimate the fragments selected (Phi) in accordance
with probabilistic criterion Pa and choose those of
them that correspond to the desired level of Pa that
has been set before calculations. If the fragments
found are not informative enough, change some ini-
tial settings (or all of them) and repeat steps 3–5.
3. Set some desirable level of probability (Pa) for the
fragment selection and some precision values for
diagonal (D1) and out-diagonal (D2) elements of
ETMCs to have ability to compare the values of cor-
responding atomic and bond characteristics (i.e., to
take into account the molecule’s flexibility).
4. By comparing all ETMCs with the ETMC of the
most active compound (taken as a template), select
those structural fragments that are common for all
active compounds only (i.e., pharmacophores, Phi).
Phi are represented by submatrices of ETMCs, or,
in short, by ETSCs.
Criterion that is commonly used in structural methods
for the probability of each structural feature occurrence
in active compounds under view is given by the equation
Pa ¼ ðn1 þ 1Þ=ðn1 þ n2 þ 2Þ;
where n1 and n2 are numbers of molecules, which possess
the feature of activity in the class of active and inactive
compounds, respectively.

2326
H. Agirbas et al. / Bioorg. Med. Chem. 15 (2007) 2322–2333
The supervised learning was performed using a variant
ETMC molecule 1
ETMC molecule N
of FFNNs known as the Associative Neural Network.49
This type of networks improves the prediction ability of
the FFNNs by explicit correction of biases. The training
of the Kohonen’s SOM is carried out in such a manner
that input vectors from the n-dimensional space
(n ꢂ 2), which possess similar properties, are mapped
to the same or nearby neurons in the two-dimensional
space. Therefore, it is possible to determine clusters of
vectors that have similar properties in the n-dimensional
space by considering all projections of input vectors to
the same SOM neurons.
W12
R13
R14R15
q1
W12
q1
R13R14R15
R16
R26
R36
R46
W23 R24 R25
q3 W35 R35
q2
W23 R24
q2
R25
W34
q4
q3
R35
W45
W45
q4
q5
q5 W56
q6
Data set formation
q1` q2 W12
q1 q3 W13
q1 q4 W14
. . . . . . . . . .
q5 q6 W56
q1` q2 W12
q1 q3 W13
q1 q4 W14
. . . . . . . . . .
q4 q5 W45
The general block-scheme of the ETM–ANNs data
analysis is presented in Figure 2.
According to this algorithm, the process of the ETM-
analysis is as follows:
(1) Form the matrices called ETMCs calculate their
common submatrices being molecular fragments (called
ETSCs) by applying ETM procedure and form the input
data set to be further projected to the SOM’s nodes. A
new table containing k input samples Xi is being formed
from the ETMCs (the value of k depends on the matrix
order for each molecule and the total number Nmol of
molecules in the initial set). Each data sample Xi is a tri-
ple (x1i,x2i,x3i), where x1i and x2i are charges for a pair
of atoms (q1 and q2) and x3i is a value that characterises
the connection between them, bond or distance (see
Fig. 1). Analogous procedure is used for the input data
SOM
ASNN
ETM
Input data set
formation (ETMCs)
Pharmacophores
calculation
Prediction of new data
SOM
Figure 1. General scheme of the data analysis.
Pharmacophores
projection on nodes
SOM map initialization
and training
Analogously, the features of inactivity, or anti-pharma-
cophores (APhj), can be calculated and estimated with
the help of the same procedure, but relative to an inac-
tive template compound. However, the further frag-
ments’ analysis cannot be done by feed forward neural
networks (FFNNs) in a straightforward manner. To
overcome this problem, a special algorithm being a com-
bination of FFNNs and the Kohonen’s self-organizing
map (SOM) has been proposed.48 The principal idea
of the combined approach is to determine the weights
of fragments represented by ETSCs and, afterwards,
to use these weights as descriptors (WDs) for the
FFNNs training. To do this, the ETMCs of molecules
and then the fragments found by the ETM software
and presented as ETSCs are being projected on the
Kohonen’s maps. In such way, the degree of each frag-
ment’s presence in the molecule (or its weight, WD) can
be determined.
Weights ( WDs, as descriptors) calculation
Partitioning input data as Training/Testing sets
ASNN
ASNN pruning of the
redundunt descriptors
ASNN training
with all descriptors
Predicting activities for new compounds
Figure 2. Block-scheme of the ETM–ANNs calculations.

H. Agirbas et al. / Bioorg. Med. Chem. 15 (2007) 2322–2333
2327
formation based on the submatrices corresponding to
the fragments, or ETSCs. A new table containing input
samples Yl for all ETSCs is being formed by means of
the procedure described at the first step for the ETMCs
(see Fig. 1).
(4) After training the ASNN, select the most informa-
tive ETMC fragments by using special pruning
methods.52,53
The table containing the calculated fragment weights
WDs is being formed for the further ASNNs training.
The number of neurons in the input layer of the network
corresponds to the number of descriptors. The hidden
layer contains five neurons. The bias neuron is presented
both on the input and hidden layers. An ensemble of
M = 100 neural networks was trained. By this, the activ-
ity value for each compound was calculated for each
ASNN and averaged over all M networks. The resulting
value was used to calculate statistical coefficients.54
Finally, the pruning methods52,53 were applied to select
the best of the ETMC fragments and obtain the most
appropriate model for the activity prediction.
(2) Initialize the Kohonen’s network parameters; calcu-
late projections of the ETMCs of all compounds on the
nodes of the SOM. The approximate number of ele-
ments in our Kohonen’s map is evaluated as
S = k*SETM, where k belongs to the interval [1.0,2.0]
and SETM is the size of the largest ETM matrix. The
value of k was found by varying it in the range of
[1,Nmol]. The researches have shown that the values
of k that are higher than 2 practically do not improve
the quality of the models obtained. The Kohonen’s net-
works realize a nonlinear projection of high-dimension-
al data set onto a low-dimensional domain. Detailed
description of Kohonen’s networks can be found in
Refs. 50,51.
After finding the best model, parameters and weight
matrices that have been found for each ASNN are to
be saved. They are used to predict activities of new po-
tential drugs (for details, see Ref. 54).
(3) For each fragment, that is either a pharmacophore or
antipharmacophore, calculate its projection on the units
of the Kohonen’s SOM and the projection error (Eij).
Then take the weight WDij of each fragment equal to
the inverse of its error Eij. Here i is the molecule’s num-
ber and j is the fragment’s number.
2.6. Analysis of pharmacophores and anti-pharmacophores
Optimized geometry data and calculated electronic char-
acteristics were used in ETMCs that had been formed
for all compounds (44 molecules), in accordance with
the main steps of the ETM-study. Conformational anal-
ysis was done for all compounds, that is, their electronic
structures were calculated from the semi-empirical AM1
method.55 Effective charges on atoms (qi) were chosen as
diagonal elements (local atomic characteristics); at place
of off-diagonal elements, either values of a bond charac-
teristic (here, Wiberg’s indices, Wij) or optimized dis-
tances (Rij, in ) were used, respectively.
The way for the weights calculation is shown at the
example of one definite submatrix, or ETSC. For the
jth ETSC, the minimal distance dl between an input
vector Yl and vectors Um is to be searched for, where
Um(u1m,u2m,u3m) is the reference vector of a map unit
that corresponds to the mth node of SOM.
It should be noted that only nodes being projections of
the X-vectors of the ETMC for the ith molecule are used
to calculate dl. In other words, we are seeking for a
molecular fragment projection to a node related to the
ETMC matrix analysed.
For E. faecalis (ATCC 29212), all compounds fall into
classes of active molecules (18 mol, MIC 6 0.6538 l-
mol/ml) and inactive ones (26 mol, IC50 > 0.6538 l-
mol/ml), in a natural way. Analogously, the classes
of active and inactive compounds relative to S. aureus
(ATCC 25923) include 24 and 20 molecules, respec-
tively. To have more stable activity features, every ac-
tive compound was used as a template for comparison
with the rest of ETMCs in the course of the ETM
study.
The projection error el for the element Yl is calculated as
the difference between Yi and Xr, where Xr is a vector for
which the error is minimal. That is, many vectors
belonging to the same matrix can be projected to the
same node, but the closest of them is selected, which is
just the projection error for this element.
The total error Eij of the ETSC projection is
After processing all ETMCs, a set of pharmacophores
and anti-pharmacophores was obtained for each type
of activity. These features form a basis for a system
capable of carrying out computer screening of new drug
prototypes and forecasting their activities. Optimal val-
ues of variations that are allowable in the process of the
matrices comparison (when testing if atoms and bonds
match) were found as D1 = 0.07 for diagonal elements
(qi) and D2 = 0.20 for off-diagonal values (Wij and Rij).
To determine the most informative activity features, the
lowest level of probabilistic estimations, Pa, was taken
as 0.80.
k
X
1
Eij ¼
el;
k
l¼1
here k is the number of samples in the jth ETSC.
Then the weight of ETSC is taken as the inverse of its
error Eij: WDij = 1 ꢁ Eij/Emax,j. Here Emax,j is the maxi-
mal error, for all j, that is, Emax,j is the maximal error
value relative to the jth column. As the result, a table
is being formed, which contains weights of all ETSCs
relative to each molecule of the initial series. Then the
data are divided into training set and test set and used
for the model development.
To form the basis of a system for the antibacterial
activity prediction, the cis-3a compound possessing the

2328
H. Agirbas et al. / Bioorg. Med. Chem. 15 (2007) 2322–2333
highest activity was taken first of all as a template for
the comparison. In Figure 3, a submatrix of this tem-
plate ETMC (or, in short, its ETSC) is given, which
corresponds to one of the pharmacophores revealed
(namely, to Ph1).
8
4
18
The given Ph1 pharmacophore consists of 8 atoms locat-
ed in different parts of the template. The nitrogen atom
N8 has a high negative charge q ¼ ꢁ0:34ꢀe; negative
charges are concentrated on the atoms N16, O7, C20
and C11, as well. The charges on the C9 and C18 carbon
atoms are close to zero. The rest of pharmacophores
were found analogously, and the probabilities of their
realization in the class of active compounds varied in
the limits of 0.86–0.95.
16
17
Cis-3e (MIC=1600)
N8 C16 C17
0.32 0.99 6.42 4.91 1.84
-0.34 7.58 6.23 2.33
-0.18 2.49 5.71
-0.13 4.46
To determine anti-pharmacophores, ETMCs of several
inactive compounds were taken as templates for com-
parison with the rest of compounds. ETSC that corre-
sponds to the APh1 anti-pharmacophore is given in
Figure 4 along with the structure of the corresponding
template from which all anti-pharmacophores have been
found.
C4
O18
-0.29
Figure 4. Submatrix (ETSC) and corresponding structure of the APh2
antipharmacophore (the template compound cis-3e).
As seen from Figure 4, APh1 (found from the cis-3e
inactive compound used as a template) consists of atoms
C4, N8, C16, C17 and O18.
the ETM. Both types of molecular fragments formed
the basis of a system for the antibacterial activities
against E. faecalis (ATCC 29212) and S. aureus (ATCC
25923) prediction. In Table 2, statistical characteristics
of six pharmacophores (P) and six anti-pharmacophores
(AP) entering the system are given.
So, a set of pharmacophores and anti-pharmacophores
was calculated (relative to the series under study) by
Figure 5 shows the dependences between frequencies of
the Ph1–Ph6 occurrence in the compounds studied and
corresponding values of their activities. As seen from
Figure 5, pharmacophores appear with high frequencies
in the class of active compounds. They are practically
absent in the class of inactive compounds. In a similar
way, maximal values are observed for the frequencies
of APh1–APh2 appearance in the class of inactive com-
pounds while for Ph1–Ph6 the frequencies are close to
zero.
9
8
7
10
11
16
Table 2. Statistical characteristics for some of pharmacophores (Phi)
and anti-pharmacophores (APhi) calculated by ETM
18
20
Type of pharmacophore/
antipharmacophore
(template compound)
E. faecalis
S. aureus
a
a
Pa
Pin
Pa
Pin
cis-3a (MIC=25)
Ph1(cis-3a)
Ph2(cis-3a)
0.92
0.92
0.86
0.90
0.91
0.88
0.19
0.16
0.16
0.08
0.12
0.14
0.08
0.08
0.14
0.10
0.09
0.12
0.81
0.84
0.84
0.92
0.88
0.86
0.91
0.90
0.87
0.90
0.91
0.86
0.18
0.14
0.15
0.08
0.11
0.13
0.09
0.06
0.13
0.10
0.09
0.14
0.82
0.86
0.85
0.92
0.89
0.87
O7
-0.20
N8 C9
3.33 4.61
-0.34 0.92
0.07
N10
0.96
3.60
4.82
-0.07
C11 N16 C18 C20
2.29 8.70 7.91 9.69
5.06 7.20 7.40 8.71
6.25 7.75 7.45 8.60
0.99 6.53 8.01 9.41
-0.16 6.82 8.33 9.65
-0.29 0.94 2.60
Ph3(cis-3d)
Ph4(cis-3d)
Ph5(trans-3u)
Ph6(cis-3v)
APh1(cis-3e)
APh2(cis-3e)
APh3(trans-3c)
APh4(trans-3c)
APh5(trans-3k)
APh6(cis-3q)
-0.02 0.98
-0.22
Figure 3. Submatrix (ETSC) and corresponding structure of the Ph1
pharmacophore (template active compound cis-3a).
a Probability estimates for inactive compounds.

H. Agirbas et al. / Bioorg. Med. Chem. 15 (2007) 2322–2333
2329
Active comp.
Inactive comp.
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
Ph(cis3a)
Ph(cis3d)
APh(cis3e)
APh(trans3k)
E. faecalis activity
Active comp.
Inactive comp.
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
Ph(cis3a)
Ph(cis3d)
S.aureus activity
Figure 5. Frequency of the pharmacophores’ (antipharmacophore’s) occurrences in the compounds studied.
When comparing the structures of the pharmacophores
and anti-pharmacophores, one can pay attention to the
differences in their spatial and electron characteristics.
Thus, both pharmacophores and anti-pharmacophores
used together definitely play an important role for the
activity prediction in the process of searching for a
new drug. Thus, the set of activity/inactivity fragments
found as the result of this study forms a basis for a
system of the antibacterial activity prediction.
and 12 randomly selected compounds (2, 6, 10, 11, 15,
19, 23, 25, 32, 35, 37, 40) were used for the model vali-
dation (see Table 1) for both E. faecalis and S. aureus.
As seen from the table, classes of active and inactive
compounds against these two types of activity are very
similar to each other.
For the E. faecalis, 170 fragments were selected from the
ETM study. The ASNNs recognized correctly 71.9%, or
23 from 32 compounds, for the training set and 50% for
the test set (6 compounds from 12) (Table 3).
2.7. Results of the ETM–ANNs approach application
A data set containing 44 molecules was used. Thirty-two
of the compounds were used for the model development,
The importance of the detected fragments for the ob-
served activity was evaluated as the result of using
Table 3. The results based on data obtained from the pharmacophores set
Sets
All pharmacophores
Molecule
Pharmacophores selected by pruning methods
WDsa number
WDs number
Molecule
Predicted (%)
Amount
Predicted (%)
Amount
Data set 1 (Enterococcus faecalis)
Training set
Test set
170
170
32
12
23 (71.9%)
6 (50.0%)
3
3
32
12
28 (87.5%)
10 (83.3%)
Data set 2 (Staphylococcus aureus)
Training set
Test set
119
119
32
12
25 (78.1%)
11 (91.6%)
5
5
32
12
28 (87.5%)
11 (91.6%)
a WDs, weight descriptors.

2330
H. Agirbas et al. / Bioorg. Med. Chem. 15 (2007) 2322–2333
pruning methods at the last step of the algorithm. The
most part of the ETMC descriptors were found to be
non-significant for the system of prognosis and elimi-
nated by the pruning algorithms. As the result, only
three ETMC fragments from 170 were selected as the
most important ones. By this, ASNNs classified cor-
rectly 87.5%, or 28 compounds, from 32 for the train-
ing set and 83.3%, or 10 compounds, from 12 for the
test set (Table 3).
4.1. Synthesis of 3-(4-diethylaminophenyl)-2,5-dimethyl-
perhydropyrrolo[3,4-d]isoxazole-4,6-dione, cis-3a and
trans-3a (General procedure)
A mixture of N-methyl-C-(4-diethylaminophenyl)nit-
rone 1a (3 mmol, 0.630 g) and N-methylmaleimide 2a
(3.3 mmol, 0.370 g) was dissolved in 50 ml benzene. The
reacting mixture was refluxed for 6–12 h. During this
time, the reaction was monitored by TLC. The products
were separated by dry-column flash chromatography.34
The mixture of ethylacetate and petroleum ether was
used as an eluent (the polarity of eluent was increased
to Rf = 0.5). The cis-3a and trans-3a isomers were recrys-
tallized separately from THF/n-hexane mixture (1:6).
For the S. aureus 119 fragments were selected. The re-
sults were correspondingly 78.1% (25 from 32 com-
pounds) for the training set, and 91.6% (11
compounds from 12) for the test set. After applying
the pruning methods, only five fragments from 119 were
selected for further use (Table 3). After this, ASNNs
classified correctly 87.5%, or 28 compounds from 32,
for the training set and 91.6%, or 11 compounds from
12, for the test set (Table 3).
Spectroscopic and analytical data of 30 new compounds
are given below.
4.1.1. 3-(4-Diethylaminophenyl)-2,5-dimethylperhydropy-
rrolo[3,4-d]isoxazole-4,6-dione, cis-3a and trans-3a. cis-
1
3a: Yield: 61%, mp 129 ꢁC. H NMR: d (ppm); 1.08 (t,
6H, 2CH3, J ꢀ 7.0 Hz), 2.53 (s, 3H, CH3), 2.95 (s, 3H,
CH3), 3.19–3.31 (m, 4H, 2CH2), 3.55 (dd,1H, Hb,
J ꢀ 8.2/7.5 Hz), 3.63 (d, 1H, Hc, J ꢀ 8.59 Hz), 4.78 (d,
1H, Ha, J ꢀ 7.16 Hz), 6.54 (d, 2H Ar-H, J ꢀ 8.7 Hz),
3. Conclusion
Peculiarities of conformational and electronic struc-
tures of compounds belonging to a large series of
2,3,5-substituted perhydropyrrolo[3,4-d]isoxazole-4,6-
diones which possess antibacterial activity against E.
faecalis (ATCC 29212) and S.aureus (ATCC 25923)
have been studied. The results of the study agree sat-
isfactorily with data obtained by other researchers rel-
ative to the same classes of compounds. A system for
the antibacterial activity prediction was developed on
the base of the pharmacophores and anti-pharmaco-
phores revealed. The obtained system allows for the
screening of active compounds and design of potent
drugs. The results of the ETM–ANNs application to
the test sample have shown that 75% of the com-
pounds are classified correctly (Table 3). Correspond-
ingly, the prediction made immediately from the
results of the ETM can give much higher percentage,
but this work is very tedious and needs the experts’
participation, while ANNs can be used even by non-
experts.
6.87(d, 2H Ar-H, J ꢀ 8.6 Hz). IR (KBr): m (cmꢁ1
)
1708.8 (C@O). Anal. Calcd For C17H23N3O3: C,
64.33; H, 7.30; N, 13.24. Found: C, 64.46; H, 7.58; N,
13.83.
1
trans-3a. Yield: 30%, mp 117 ꢁC. H NMR: d (ppm);
1.10 (t, 6H, 2CH3, J ꢀ 7.0 Hz), 2.34 (br, 3H, CH3),
2.98 (s, 3H, CH3), 3.26–3.31 (m, 4H, 2CH2), 3.60 (dd,
1H, Hb, J ꢀ 3.7/3.7 Hz), 4.39 (very broad, 1H, Hc),
3.84 (d, 1H, Ha, J ꢀ 6.97 Hz), 6.58 (d, 2H, Ar-H,
J ꢀ 8.50 Hz), 7.02 (br, 2H, Ar-H). IR (KBr): m (cmꢁ1);
1705.0 (C@O). Anal. Calcd For C17H23N3O3: C,
64.33; H, 7.30; N, 13.24. Found: C, 63.99; H, 7.30; N,
13.08.
4.1.2. 3-(4-Dimethylaminophenyl)-2,5-dimethylperhydro-
pyrrolo[3,4-d]isoxazole-4,6-dione, trans-3b. Yield: 25%,
1
mp 140 ꢁC. H NMR:d(ppm); 2.42 (br, 3H, CH3), 2.90
(s, 6H, 2CH3), 2.98 (s, 3H, CH3), 3.58 (dd, 1H, Hb,
J ꢀ 4.0/4.6 Hz), 4.30 (very broad, 1H, Hc), 4.84 (d,
1H, Ha, J ꢀ 7.21 Hz), 6.81–7.19 (m, 4H, Ar-H). IR
(KBr): m (cmꢁ1); 1705.0 (C@O). Anal. Calcd For
C15H19N3O3: C, 62.27; H, 6.62; N, 14.52. Found: C,
62.31; H, 6.93; N, 14.92.
4. Experimental
All preparative experiments were carried out in ben-
zene, which had been dried with the standard meth-
ods.56
Melting
points
were
determined
on
4.1.3. 3-(4-Methoxyphenyl)-2,5-dimethylperhydropyrrol-
o[3,4-d]isoxazole-4,6-dione, cis-3c and trans-3c. cis-3c.
Yield: 65%, mp 135 ꢁC. 1H NMR: d (ppm); 2.54 (s,
3H, CH3), 2.93 (s, 3H, CH3), 3.57 (dd, 1H, Hb,
J ꢀ 8.2/8.8 Hz), 3.71 (d, 1H, Hc, J ꢀ 8.84 Hz) 3.73 (s,
3H, CH3), 4.80 (d, 1H, Ha, J ꢀ 7.25 Hz), 6.83 (d, 2H,
Ar-H, J ꢀ 8.0 Hz), 7.03 (d, 2H, Ar-H, J ꢀ 14.2 Hz).
IR (KBr): m (cmꢁ1); 1712.7 (C@O). Anal. Calcd for
C14H16N2O4: C, 60.86; H, 5.84; N, 10.14. Found: C,
60.94; H, 6.09; N, 10.12.
Electrothermal 9200 apparatus and uncorrected. IR
spectra were recorded on Shimadzu FTIR-821 PC
Fourier Transform IR Spectrometer; they are given
in wavenumbers (cmꢁ1). 1H NMR spectra were
recorded on Bruker DPX-400 (400 MHz) High Per-
formance Digital FT NMR Spectrometer using
CDCl3 with Me4Si as an internal standard. Elemental
analyses were performed on Carlo Erba-1106 instru-
ment. Preparative chromatography was performed
using Silica gel (Merck). N-Alkyl(aryl)-C-substituted
phenylnitrones were prepared by the literature
method.57
1
trans-3c. Yield: 23%, mp 193 ꢁC. H NMR: d (ppm);
2.37 (br, 3H, CH3), 2.99 (s, 3H, CH3), 3.59 (dd, 1H,

H. Agirbas et al. / Bioorg. Med. Chem. 15 (2007) 2322–2333
2331
Hb, J ꢀ 3.4/3.7 Hz), 3.75 (s, 3H, CH3), ꢃ4.30 (very
broad, 1H, Hc), 4.85 (d, 1H, Ha, J ꢀ 7.24 Hz), 6.85 (d,
2H, Ar-H, J ꢀ 8.52 Hz), 7.15 (br, 2H, Ar-H). IR
(KBr): m (cmꢁ1); 1701.1 (C@O). Anal. Calcd For
C14H16N2O4: C, 60.86; H, 5.84; N, 10.14. Found: C,
60.56; H, 6.27; N, 10.11.
for C13H13ClN2O3: C, 55.62; H, 4.67; N, 9.98. Found: C,
55.55; H, 4.73; N, 10.46.
1
trans-3g. Yield: 38%, mp 126 ꢁC. H NMR: d (ppm);
2.42 (br, 3H, CH3), 3.00 (s, 3H, CH3), 3.61 (dd, 1H,
Hb, J ꢀ 3.7/3.7 Hz), ꢃ4.0 (very broad,1H, Hc) 4.86 (d,
1H, Ha, J ꢀ 7.30 Hz), 7.05–7.27 (m, 4H, Ar-H). IR
(KBr): m (cmꢁ1); 1701.1 (C@O). Anal. Calcd for
C13H13ClN2O3: C, 55.62; H, 4.67; N, 9.98. Found: C,
55.17; H, 4.67; N, 10.02.
4.1.4. 3-(4-Benzyloxy-3-methoxyphenyl)-2,5-dimethylper-
hydropyrrolo[3,4-d]isoxazole-4,6-dione, trans-3d. Yield:
25%, mp 100 ꢁC. 1H NMR: d (ppm); 2.42 (br, 3H,
CH3), 2.98 (s, 3H, CH3), 3.60 (dd, 1H, Hb, J ꢀ 4.0/
4.2 Hz), 3.83 (s, 3H, CH3), ꢃ4.50 (very broad, 1H,
Hc), 4.84 (d, 1H, Ha, J ꢀ 7.21 Hz),5.08 (s, 2H, CH2),
6.73-6.81 (m, 3H Ar-H), 7.20–7.35 (m, 5H, Ar-H). IR
(KBr): m (cmꢁ1); 1705.0 (C@O); Anal. Calcd For
C21H22N2O5: C, 65.96; H, 5.80; N, 7.33. Found: C,
64.71; H, 5.57; N, 6.92.
4.1.8. 3-(3-Bromophenyl)-2,5-dimethylperhydropyrrolo-
[3,4-d]isoxazole-4,6-dione, cis-3h and trans-3h. cis-3h.
1
Yield: 42%, mp 186 ꢁC. H NMR: d (ppm); 2.57 (s, 3H,
CH3), 2.93 (s, 3H, CH3), 3.66 (dd, 1H, Hb J ꢀ 8.6/
8.3 Hz), 3.72 (d, 1H, Hc, J ꢀ 8.67 Hz), 4.82 (d, 1H, Ha,
J ꢀ 7.14 Hz), 7.05 (d, 1H, Ar-H, J ꢀ 7.6 Hz), 7.18 (dd,
1H, Ar-H, J ꢀ 7.8/6.5 Hz), 7.28 (s, 1H, Ar-H), 7.39 (d,
1H, Ar-H, J ꢀ 7.8 Hz). IR (KBr): m (cmꢁ1); 1705.0
(C@O). Anal. Calcd for C13H13BrN2O3: C, 48.02; H,
4.03; N, 8.62. Found: C, 47.80; H, 4.05; N, 8.56.
4.1.5. 2,5-Dimethyl-3-(4-methylphenyl)perhydropyrrolo-
[3,4-d]isoxazole-4,6-dione, cis-3e and trans-3e. cis-3e.
Yield: 46%, mp 161 ꢁC. 1H NMR: d (ppm); 2.26 (s,
3H, CH3), 2.55 (s, 3H, CH3), 2.92 (s, 3H, CH3), 3.61
(dd, 1H, Hb, J ꢀ 8.2/7.6 Hz), 3.72 (d, 1H, Hc,
J ꢀ 8.58 Hz), 4.78 (d, 1H, Ha, J ꢀ 7.22 Hz), 6.99 (d,
2H, Ar-H, J ꢀ 7.8 Hz), 7.07 (d, 2H, Ar-H, J ꢀ 7.8 Hz).
IR (KBr): m (cmꢁ1); 1705.0 (C@O). Anal. Calcd for
C14H16N2O3: C, 64.60; H, 6.20; N, 10.76. Found: C,
64.13; H, 6.00; N, 10.75.
1
trans-3h. Yield: 39%, mp 129. H NMR: d (ppm); 2.42
(br, 3H, CH3), 2.99 (s, 3H, CH3), 3.61 (dd, 1H, Hb,
J ꢀ 3.8/3.8 Hz), ꢃ4.0 (very broad,1H, Hc) 4.87 (d, 1H,
Ha, J ꢀ 7.30 Hz), 7.20–7.41 (m, 4H, Ar-H). IR (KBr):
m
(cmꢁ1); 1705.0 (C@O). Anal. Calcd for
C13H13BrN2O3: C, 48.02; H, 4.03; N, 8.62. Found: C,
48.14; H, 4.22; N, 8.87.
1
trans-3e. Yield: 37%, mp 131 ꢁC. H NMR: d (ppm);
2.38 (s, 3H, CH3), 2.45 (br, 3H, CH3), 3.08 (s, 3H,
CH3), 3.71 (dd, 1H, Hb, J ꢀ 6.6/6.7 Hz), ꢃ4.30 (very
broad, 1H, Hc) 4.94 (d, 1H, Ha, J ꢀ 7.20 Hz), 7.20 (br,
4H, Ar-H). IR (KBr): m (cmꢁ1); 1705.0 (C@O). Anal.
Calcd For C14H16N2O3: C, 64.60; H, 6.20; N, 10.76.
Found: C, 64.72; H, 5.68; N, 10.15.
4.1.9. 3-(4-Bromophenyl)-2,5-dimethylperhydropyrrolo[3,
4-d]isoxazole-4,6-dione, cis-3j and trans-3j. cis-3j. Yield:
1
45%, mp 195 ꢁC. H NMR: d (ppm); 2.55 (s, 3H, CH3),
2.92 (s, 3H, CH3), 3.63 (dd, 1H, Hb, J ꢀ 8.3/7.5 Hz),
3.71 (d, 1H, Hc, J ꢀ 8.57 Hz), 4.81 (d, 1H, Ha,
J ꢀ 7.22 Hz), 7.00 (d, 2H, Ar-H, J ꢀ 8.4 Hz), 7.41 (d,
2H, Ar-H, J ꢀ 8.4 Hz). IR (KBr): m (cmꢁ1); 1708.8
(C@O). Anal. Calcd for C13H13BrN2O3: C, 48.02; H,
4.03; N, 8.62. Found: C, 48.02; H, 3.88; N, 8.71.
4.1.6. 2,5-Dimethyl-3-phenylperhydropyrrolo[3,4-d]isox-
azole-4,6-dione, cis-3f and trans-3f. cis-3f. Yield: 34%,
1
mp 127 ꢁC. H NMR: d (ppm); 2.57 (s, 3H, CH3), 2.92
(s, 3H, CH3), 3.64 (dd, 1H, Hb, J ꢀ 8.4/7.5 Hz), 3.75
(d, 1H, Hc, J ꢀ 8.62 Hz), 4.81 (d, 1H, Ha, J ꢀ 7.25 Hz),
7.10–7.12 (m, 2H, Ar-H), 7.20-7.33 (m, 3H, Ar-H). IR
(KBr): m (cmꢁ1); 1705.0 (C@O). Anal. Calcd For
C13H14N2O3: C, 63.40; H, 5.73; N, 11.38. Found: C,
63.60; H, 5.66; N, 11.31.
trans-3j. Yield: 40%, mp 163 ꢁC. 1H NMR: d (ppm); 2.40
(br, 3H, CH3), 2.99 (s, 3H, CH3), 3.58 (dd, 1H, Hb,
J ꢀ 3.7/3.6 Hz), ꢃ4.10 (very broad, 1H, Hc) 4.82 (d,
1H, Ha, J ꢀ 7.30 Hz), 7.13 (br, 2H, Ar-H), 7.45 (d,
2H, Ar-H, J ꢀ 8.4 Hz). IR (KBr): m (cmꢁ1); 1701.1
(C@O). Anal. Calcd for C13H13BrN2O3: C, 48.02; H,
4.03; N, 8.62. Found: C, 48.05; H, 4.07; N, 8.78.
1
trans-3f. Yield: 43%, mp 140 ꢁC. H NMR: d (ppm);
2.39 (br, 3H, CH3), 3.00 (s, 3H, CH3), 3.65 (dd, 1H,
Hb, J ꢀ 3.6/3.4 Hz), ꢃ4.20 (very broad, 1H, Hc) 4.87
(d, 1H, Ha, J ꢀ 7.28 Hz), 7.24–7.34 (m, 5H, Ar-H). IR
(KBr): m (cmꢁ1); 1703.0 (C@O). Anal. Calcd For
C13H14N2O3: C, 63.40; H, 5.73; N, 11.38. Found: C,
63.69; H, 5.86; N, 11.49.
4.1.10. 3-(4-Diethylaminophenyl)-2-methyl-5-phenylper-
hydropyrrolo[3,4-d]isoxazole-4,6-dione, cis-3l and trans-
1
3l. cis-3l. Yield: 48%, mp 142 ꢁC. H NMR: d (ppm);
1.08 (t, 6H, 2CH3, J ꢀ 7.0 Hz), 2.60 (s, 3H, CH3),
3.18–3.28 (m, 4H, 2CH2), 3.70 (dd, 1H, Hb, J ꢀ 8.4/
7.5 Hz), 3.77 (d, 1H, Hc, J ꢀ 8.70 Hz), 4.93 (d, 1H,
Ha, J ꢀ 7.22 Hz), 6.53 (d, 2H, Ar-H, J ꢀ 8.5 Hz), 7.01
(d, 2H, Ar-H, J ꢀ 8.5 Hz), 7.20–7.37 (m, 5H, Ar-H).
IR (KBr): m (cmꢁ1); 1708.8 (C@O). Anal. Calcd for
C22H25N3O3: C, 69.64; H, 6.64; N, 11.07. Found: C,
69.59; H, 6.70; N, 11.39.
4.1.7. 3-(3-Chlorophenyl)-2,5-dimethylperhydropyrrolo-
[3,4-d]isoxazole-4,6-dione, cis-3g and trans-3g. cis-3g.
Yield: 38%, mp 184 ꢁC. 1H NMR: d (ppm); 2.58 (s,
3H, CH3), 2.93 (s, 3H, CH3), 3.64 (dd, 1H, Hb,
J ꢀ 8.3/7.5 Hz), 3.72 (d, 1H, Hc, J ꢀ 8.63 Hz), 4.81 (d,
1H, Ha, J ꢀ 7.18 Hz), 7.02 (d, 2H, Ar-H, J ꢀ 17.3 Hz),
7.12 (s, 1H, Ar-H), 7.22 (dd, 1H Ar-H, J ꢀ 9.6/
7.6 Hz). IR (KBr): m (cmꢁ1); 1708.8 (C@O). Anal. Calcd
trans-3l. Yield: 43%, 143 ꢁC. 1H NMR: d (ppm); 1.10 (t,
6H, 2CH3), 2.41 (br, 3H, CH3), 3.26–3.31 (m, 4H,
2CH2), 3.77 (dd, 1H, Hb, J ꢀ 3.6/3.6 Hz), ꢃ4.39 (very

2332
H. Agirbas et al. / Bioorg. Med. Chem. 15 (2007) 2322–2333
broad, 1H, Hc), 4.98 (d, 1H, Ha, J ꢀ 7.28 Hz), 6.60 (d,
2H, Ar-H, J ꢀ 7.8 Hz), 7.06 (br, 2H, Ar-H), 7.28–7.41
(m, 5H, Ar-H). IR (KBr): m (cmꢁ1); 1716.5 (C@O).
Anal. Calcd for C22H25N3O3: C, 69.64; H, 6.64; N,
11.07. Found: C, 69.31; H, 6.75; N, 10.96.
(d, 1H, Hc, J ꢀ 8.84 Hz), 4.98 (d, 1H, Ha, J ꢀ 7.25 Hz),
7.19–7.25 (m, 4H, Ar-H), 7.36–7.44 (m, 5H, Ar-H). IR
(KBr): m (cmꢁ1); 1716.5 (C@O). Anal. Calcd For
C18H15BrN2O3: C, 55.83; H, 3.90; N, 7.23. Found: C,
55.84; H, 3.69; N, 6.93.
1
4.1.11. 3-(4-Methoxyphenyl)-2-methyl-5-phenylperhydro-
pyrrolo[3,4-d]isoxazole-4,6-dione, cis-3n and trans-3n. cis-
3n. Yield: 40%, mp 165 ꢁC. H NMR: d (ppm); 2.61 (s,
3H, CH3), 3.72 (s, 3H, CH3), 3.75 (dd, 1H, Hb,
J ꢀ 7.3/7.4 Hz), 3.83 (d, 1H, Hc, J ꢀ 8.74 Hz), 4.94 (d,
1H, Ha, J ꢀ 7.38 Hz), 6.85 (d, 2H, Ar-H, J ꢀ 8.7 Hz),
7.13–7.19 (m, 5H, Ar-H), 7.27–7.39 (m, 2H, Ar-H). IR
(KBr): m (cmꢁ1); 1712.7 (C@O). Anal. Calcd for
C19H18N2O4: C, 67.44; H,5.36; N,8.28. Found: C,
67.38; H, 5.44; N, 8.21.
trans-3r. Yield: 46%, mp 149 ꢁC. H NMR: d (ppm);
2.49 (s, 3H, CH3), 3.75 (dd, 1H, Hb, J ꢀ 3.7/3.6 Hz),
ꢃ4.20 (very broad, 1H, Hc), 4.98 (d, 1H, Ha,
J ꢀ 7.46 Hz), 7.18-7.27 (m, 4H, Ar-H), 7.33–7.44 (m,
5H, Ar-H). IR (KBr): m (cmꢁ1); 1716.5 (C@O). Anal.
Calcd for C18H15BrN2O3: C, 55.83; H, 3.90; N, 7.23.
Found: C, 55.81; H, 3.98; N, 7.15.
1
4.1.15. 3-(4-Bromophenyl)-2-methyl-5-phenylperhydro-
pyrrolo[3,4-d]isoxazole-4,6-dione, cis-3t and trans-3t.
cis-3t. Yield: 42%, mp 136 ꢁC. H NMR: d (ppm); 2.62
(s, 3H, CH3), 3.78 (dd, 1H, Hb, J ꢀ 7.86/8.0 Hz), 3.84
(d, 1H, Hc, J ꢀ 8.71 Hz), 4.98 (d, 1H, Ha, J ꢀ 7.23 Hz),
7.13-7.40 (m, 9H, Ar-H). IR (KBr): m (cmꢁ1); 1716.5
(C@O). Anal. Calcd for C18H15BrN2O3: C, 55.83; H,
3.90; N, 7.23. Found: C, 55.77; H, 3.75; N, 7.13.
1
trans-3n. Yield: 36%, mp 166 ꢁC. 1H NMR: d (ppm); 2.42
(br, 3H, CH3), 3.75 (s, 4H, Hb and CH3), ꢃ4.5
(very broad, 1H), 4.98 (d,1H, Ha, J ꢀ 7.38 Hz), 6.85
(d, 2H, Ar-H, J ꢀ 8.6 Hz), 7.28 (d, 2H, Ar-H,
J ꢀ 7.4 Hz), 7.39–7.43 (m, 2H, Ar-H). IR (KBr): m
(cmꢁ1); 1712.7(C@O). Anal. Calcd for C19H18N2O4: C,
67.44; H, 5.36; N, 8.28. Found: C, 67.35; H, 5.34; N, 8.43.
1
trans-3t. Yield: 42%, mp 182 ꢁC. H NMR: d (ppm);
2.48 (br, 3H, CH3), 3.74 (dd, 1H, Hb, J ꢀ 3.6/3.7 Hz),
ꢃ4.20 (very broad, 1H, Hc) 4.99 (d, 1H, Ha,
J ꢀ 7.44 Hz), 7.20–7.43 (m, 9H, Ar-H). IR (KBr): m
(cmꢁ1); 1716.5 (C@O). Anal. Calcd for C18H15BrN2O3:
C, 55.83; H, 3.90; N, 7.23. Found: C, 55.93; H, 3.74; N,
7.12.
4.1.12. 2-Methyl-3-(4-methylphenyl)-5-phenylperhydro-
pyrrolo[3,4-d]isoxazole-4,6-dione, cis-3p and trans-3p.
cis-3p. Yield: 44%, mp 138 ꢁC. 1H NMR: d (ppm);
2.25 (s, 3H, CH3), 2.61 (s, 3H, CH3), 3.73 (dd, 1H,
Hb, J ꢀ 7.7.6/7.7 Hz) ,3.81 (d, 1H, Hc, J ꢀ 8.74 Hz),
4.92 (d, 1H, Ha, J ꢀ 7.35 Hz), 7.10–7.30 (m, 9H, Ar-
H). IR (KBr): m (cmꢁ1); 1716.5 (C@O). Anal. Calcd
for C19H18N2O3: C, 70.79; H, 5.63; N, 8.69. Found: C,
70.80; H, 5.65; N, 8.81.
4.1.16.
3-(4-Dimethylaminophenyl)-5-methyl-2-phenyl-
perhydropyrrolo[3,4-d]isoxazole-4,6-dione, cis-3v and
trans-3v. cis-3v. Yield: 45%, mp 186 ꢁC. 1H NMR: d
(ppm); 2.87 (s, 6H, 2CH3), 2.89 (s, 3H, CH3), 3.78 (dd,
1H, Hb, J ꢀ 7.9/8.6 Hz), 4.55 (d, 1H, Hc, J ꢀ 9.02 Hz),
4.99 (d, 1H, Ha, J ꢀ 7.51 Hz), 6.58 (d, 2H, Ar-H,
J ꢀ 8.5 Hz), 6.93-7.19 (m, 7H, Ar-H). IR (KBr): m
(cmꢁ1); 1708.8 (C@O). Anal. Calcd for C20H21N3O3:
C, 68.36; H, 6.02; N, 11.96. Found: C, 68.15; H, 5.97;
N, 11.89.
1
trans-3p. Yield: 49%, mp 149 ꢁC. H NMR: d (ppm);
2.29(s, 3H, CH3), 2.43 (br, 3H, CH3), 3.76 (dd, 1H,
Hb, J ꢀ 3.6/3.5 Hz), ꢃ4.50 (very broad, 1H, Hc,), 4.97
(d, 1H, Ha, J ꢀ 7.41 Hz), 7.13–7.33 (m, 9H, Ar-H). IR
(KBr): m (cmꢁ1); 1716.5 (C@O). Anal. Calcd for
C19H18N2O3: C, 70.79; H, 5.63; N, 8.69. Found: C,
71.11; H, 5.68; N, 8.97.
1
trans-3v. Yield: 41%, mp 178 ꢁC. H NMR: d (ppm);
4.1.13. 2-Methyl-3,5-diphenylperhydropyrrolo[3,4-d]isox-
azole-4,6-dione, cis-3q and trans-3q. cis-3q. Yield: 41%,
mp 146 ꢁC. 1H NMR: d (ppm); 2.64 (s, 3H, CH3),
3.79(dd, 1H, Hb, J ꢀ 8.6/7.5 Hz), 3.89 (d, 1H, Hc,
J ꢀ 8.77 Hz), 4.98 (d, 1H, Ha, J ꢀ 7.39 Hz), 7.09–7.47
(m, 10H, Ar-H). IR (KBr): m (cmꢁ1); 1720.4 (C@O).
Anal. Calcd for C18H16N2O3: C, 70.12; H, 5.23; N,
9.09. Found: C, 69.93; H, 5.54; N, 9.07.
2.64 (s, 3H, CH3), 2.86 (s, 6H, 2CH3), 3.72 (d, 1H, Hb,
J ꢀ 7.15 Hz), 4.92 (d, 1H, Ha, J ꢀ 7.21 Hz), 5.32 (s,
1H, Hc), 6.62 (d, 2H, Ar-H, J ꢀ 7.5 Hz), 6.80–6.83 (m,
1H, Ar-H), 6.93 (d, 2H, Ar-H, J ꢀ 7.5 Hz), 7.11 (t,
2H, Ar-H, J ꢀ 8.0 Hz), 7.21 (t, 2H, Ar-H, J ꢀ 8.50 Hz).
IR (KBr): m (cmꢁ1); 1705 (C@O). Anal. Calcd for
C20H21N3O3: C, 68.36; H, 6.02; N, 11.96. Found: C,
68.26; H, 6.32; N, 12.28.
1
trans-3q. Yield: 48%, mp 182 ꢁC. H NMR: d (ppm);
Spectroscopic and analytical data of cis-3b, cis-3d, cis-3i,
Trans-3i, cis-3k, trans-3k, cis-3m, trans-3m, cis-3o, trans-
3o, cis-3s, trans-3s, cis-3u, trans-3u were reported
previously.13,14
2.47 (br, 3H, CH3), 3.80 (dd, 1H, Hb, J ꢀ 3.8/3.6 Hz),
ꢃ4.25 (very broad, 1H, Hc), 5.01 (d, 1H, Ha,
J ꢀ 7.42 Hz), 7.19-7.44 (m, 10H, Ar-H). IR (KBr): m
(cmꢁ1); 1712.7 (C@O). Anal. Calcd for C18H16N2O3: C,
70.12; H, 5.23; N, 9.09. Found: C, 70.48; H, 5.55; N, 9.53.
References and notes
4.1.14. 3-(3-Bromophenyl)-2-methyl-5-phenylperhydro-
pyrrolo[3,4-d]isoxazole-4,6-dione, cis-3r and trans-3r.
cis-3r. Yield: 43%, mp 197 ꢁC. H NMR: d (ppm); 2.65
(s, 3H, CH3), 3.78 (dd, 1H, Hb, J ꢀ 7.3/8.8 Hz), 3.85
1. Davies, J. Nature 1996, 383, 219–220.
2. Black, D. S. C.; Crozier, R. FV.; Davis, C. Synthesis 1975,
205–221.
1

H. Agirbas et al. / Bioorg. Med. Chem. 15 (2007) 2322–2333
2333
3. Tuffariello, J. J. In Nitrones in 1,3-Dipolar Cycloaddition
Chemistry; Padwa, A., Ed.; Wiley Interscience: NewYork,
1984; Vol. 2, p 1.
4. Carruthers, W. R. In Cycloaddition in Organic Synthesis;
Pergamon Press: London, 1990; p 269.
29. Heaney, F.; Rooney, O.; Cunningham, D.; Mc Ardle, P.
J. Chem. Soc. Perkin Trans. 2 2001, 373–378.
30. Jaroskova, L.; Konopikova, M.; Fisera, L. Chem. Pap. -
Chem. Zvesti 1994, 48, 35–39.
31. Signal, K. K.; Kaur, J. Indian J. Heterocycl. Chem. 1999,
9, 103–106.
32. ASC Symp. Ser. 1992, 494, 158–171.
5. Kwon, T.; Heimann, A. S.; Oriaku, E. T.; Yoon, K.; Lee,
H. J. J. Med. Chem. 1995, 38, 1048–1051.
6. Gordaliza, M. G.; Faircloth, T.; Castro, M. A.; Corral, J.
M. M.; Lopez Vazqueng, M. L. J. Med. Chem. 1996, 39,
2865–2866.
7. Kumar, K. R. R.; Mallesha, H.; Rangappa, K. S. Arch.
Pharm. Pharm. Med. Chem. 2003, 336, 159–164.
8. Kang, Y. K.; Shin, K. J.; Yoo, K. H.; Seo, K. J.; Hong, C.
Y.; Lee, C. S.; Park, S. Y.; Kim, D. J.; Park, S. W. Bioorg.
Med. Chem. Lett. 2000, 10, 95–99.
33. Altimari, U. A. R.; Fisera, L.; Pronayova, N. Collect.
Czech. Chem. Commun. 1992, 57, 1982–1992.
34. Fisera, L.; Altimari, U. A. R.; Ertl, P.; Pronayova, N.
Monatsh. Chem. 1993, 1019–1029.
35. Gother, K. V.; Jorgensen, K. A. J. Org. Chem. 1994, 59,
5687–5691.
36. Bernotas, R. C.; Sabol, J. S.; Sing, L.; Friedrich, D.
Synlett 1999, 653–655.
9. Popat, K. H.; Nimavat, K. S.; Kachhadia, V. V.; Joshi, H.
S. J. Indian Chem. Soc. 2003, 80, 707–708.
37. Jensen, K. B.; Hazell, R. G.; Jorgensen, K. A. J.Org.
Chem. 1999, 64, 2353–2360.
10. Rajanarendar, E.; Afzal, M.; Karunakar, D. Indian
J.Chem. B 2003, 42, 353–357.
38. Minakata, S.; Ezoe, T.; Nakamura, N.; Ryu, I.; Komatsu,
M. Tetrahedron Lett. 1998, 39, 5205–5208.
11. Desai, J. M.; Shah, V. H. Indian J. Chem. B 2003, 42, 382–
385.
39. Jensen, K. B.; Gothelf, K. V.; Hazell, R. G.; Jorgensen, K.
A. J. Org. Chem. 1997, 62, 2471–2477.
12. Brom, N. J. P.; Elder, J. S.; Hannan, P. C. T.; Pons, J. E.;
Ohanlon, P. J.; Walker, G.; Wilson, J.; Woodall, P.
J. Antibiot. 1995, 48, 1336–1344.
40. Chakraborty, B.; Debritto, P. J.; Ghosh, L. A. R. Indian
J. Heterocycl. Chem. 1996, 6, 77–78.
41. Gothelf, K. V.; Jorgensen, K. A. Acta Chem. Scand. 1996,
50, 652–660.
13. w w Agirbas, H.; Guner, S. Heterocycl. Commun. 2000, 6,
¨
557–562.
14. Pihlaja, K.; Tahtinen, P.; Shaikhutdinov, R.; Hartikainen,
42. Huisgen, R.; Hauck, H.; Seidl, H.; Burger, M. Chem. Ber.
1969, 102, 1117–1128.
H.; Ovcharenko, V.; Agirbas, H.; Guner, S. J. Heterocycl.
Chem. 2004, 41, 741–746.
43. NCCLS, Methods for diluation antimicrobial susceptibil-
itytests for bacteria that grow aerobically, Approved Std.
M-7, A-4, USA, 1997.
44. Dimoglo, A. S. Chim. Pharm. Zh. (Russ.) 1985, 19, 438–
444.
45. Shvets, N. M. Comp. Sci. J. Moldova 1993, 1, 101–110.
46. Shvets, N. M. Comp. Sci. J. Moldova 1997, 5, 301–310.
47. Shvets, N. M.; Terletskaya, A.; Dimoglo, A. S.; Chuma-
kov, Yu. M. THEOCHEM (J. Mol. Str.) 1999, 463, 105–
110.
48. Dimoglo, A.; Kovalishyn, V.; Shvets, N.; Ahsen, V. Mini-
Rev. Med. Chem. 2005, 5, 879–892.
49. Tetko, I. V. J. Chem. Inf. Comput. Sci. 2002, 42, 717–728.
50. Kohonen, T. Self-organisation Maps; Springer-Verlag:
Berlin, 1995.
¨
15. Shahapurkar, S.; Pandya, T.; Kawathekar, N.; Chaturv-
edi, S. C. Eur. J. Med. Chem. 2004, 39, 383–388.
16. Hye-Jung, K.; Chong, H. C.; Kyu Yang, Y.; Kyung-Lae,
P.; Sung-eun, Y. Bioorg. Med. Chem. 2004, 12, 1629–1641.
17. Shvets, N. M.; Dimoglo, A. S. Nahrung-Food 1998, 42,
364–370.
18. Dimoglo, A. S.; Beda, A. A.; Shvets, N. M.; Gorbachov,
M. Yu.; Kheifits, L. A.; Aulchenko, I. S. New. J. Chem.
1995, 19, 149–155.
19. Dimoglo, A. S.; Vlad, P. F.; Shvets, N. M.; Coltsa, M. N.
New J. Chem. 2001, 25, 283–288.
20. Dimoglo, A. S.; Shvets, N. M.; Tetko, I. V.; Livingstone,
D. J. QSAR 2001, 20, 31–45.
21. Dimoglo, A. S.; Chumakov, Yu. M.; Dobrova, B. N.;
Shvets, N. M.; Saracoglu, M. Arzneim.-Forsch./Drug Res.
1997, 47, 415–419.
22. Manallack, D. T.; Livingstone, D. J. Eur. J. Med. Chem.
1999, 34, 195–208.
23. Zupan, J.; Gasteiger, J. In Neural Networks for
Chemistry and Drug Design. An Introduction; VCH:
Weinheim, 1999.
24. Tetko, I. V.; Luik, A. I.; Poda, G. I. J. Med. Chem. 1993,
36, 811–814.
25. Frederickson, M. Tetrahedron 1997, 53, 403–425.
26. Al-Timari, U. A. R.; Fisera, L.; Ertl, P.; Goljer, I.;
Pronayova, N. Monatsh. Chem. 1992, 123, 999–1013.
27. Patwa, A.; Kline, D. N.; Koehler, K. F.; Matzinger, M.;
Venkatramanan, M. K. J. Org. Chem. 1987, 52, 3909–
3917.
51. Kohonen, T.; Hinninen, J.; Kangas, J.; Laaksonen, J.
SOM_PAK: The self-organizing map program package is
obtainable from the internet address http://www.cis.hut.fi/
research/som_pak/.
52. Tetko, I. V.; Villa, A. E. P.; Livingstone, D. J. J. Chem.
Inf. Comput. Sci. 1996, 36, 794–803.
53. 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.
54. Tetko, I. V.; Livingstone, D. J.; Luik, A. I. J. Chem. Inf.
Comput. Sci. 1995, 35, 826–833.
55. Dewar, M. J. S.; Zoebisch, E. G.; Healy, E. F.; Stewart, J.
J. P. J. Am. Chem. Soc. 1985, 107, 3902–3909.
56. Furniss, B. S.; Hannaford, A. J.; Smith, P. W. G.;
Tatchell, A. R.. Vogel’sTextbook of Practical Organic
Chemistry, 5th ed.; Longman: England, 1998, p 168.
57. Rundel, W. A Herstellung von Nitronen. Methoden der
Organische Chemie, Stutgart, 1968; X/4, pp 316.
28. Gothelf, K. V.; Hazell, R. G.; Jorgensen, K. A. J. Org.
Chem. 1996, 61, 346–355.