Antiproliferative novel isoxazoles: Modeling, virtual screening, synthesis, and bioactivity evaluation

Article abstract of DOI:10.1016/j.ejmech.2014.05.011

A series of novel isoxazole derivatives were efficiently synthesized through the adaptation/modification of an in situ synthetic procedure for pyrazoles. All novel compounds were tested against four different cell lines to evaluate their antiproliferative activity. Based on the Hela cells results of this study and previous work, a classification model to predict the anti-proliferative activity of isoxazole and pyrazole derivatives was developed. Random Forest modeling was used in view of the development of an accurate and reliable model that was subsequently validated. A virtual screening study was then proposed for the design of novel active derivatives. Compounds 9 and 11 demonstrated significant cytostatic activity; the fused isoxazole derivative 18 and the virtually proposed compound 2v, were proved at least 10 times more potent as compared to compound 9, with IC50 values near and below 1 μM. In conclusion, a new series of isoxazoles was exploited with some of them exhibiting promising cytostatic activities. Further studies on the substitution pattern of the isoxazole core can potentially provide compounds with cytostatic action at the nM scale. In this direction the in silico approach described herein can also be used to screen existing databases to identify derivatives with anticipated activity.

Full text of DOI:10.1016/j.ejmech.2014.05.011

Accepted Manuscript
Antiproliferative Novel Isoxazoles: Modeling, Virtual Screening, Synthesis, and
Bioactivity Evaluation
Evangelia Tzanetou, Sandra Liekens, Konstantinos M. Kasiotis, Georgia Melagraki,
Antreas Afantitis, Nikolas Fokialakis, Serkos A. Haroutounian, Ph.D.
PII:
S0223-5234(14)00425-5
10.1016/j.ejmech.2014.05.011
EJMECH 6967
DOI:
Reference:
To appear in: European Journal of Medicinal Chemistry
Received Date: 10 January 2014
Revised Date: 16 March 2014
Accepted Date: 2 May 2014
Please cite this article as: E. Tzanetou, S. Liekens, K.M. Kasiotis, G. Melagraki, A. Afantitis, N.
Fokialakis, S.A. Haroutounian, Antiproliferative Novel Isoxazoles: Modeling, Virtual Screening,
Synthesis, and Bioactivity Evaluation, European Journal of Medicinal Chemistry (2014), doi: 10.1016/
j.ejmech.2014.05.011.
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT
Antiproliferative Novel Isoxazoles: Modeling, Virtual
Screening, Synthesis, and Bioactivity Evaluation
1
2
3
Evangelia Tzanetou, Sandra Liekens, Konstantinos M. Kasiotis, Georgia
4
4
5
1,*
Melagraki, Antreas Afantitis, Nikolas Fokialakis and Serkos A. Haroutounian
1
Agricultural University of Athens, Chemistry Laboratory, Iera odos 75, Athens 11855,
Greece
2
Rega Institute for Medical Research, KU Leuven, Minderbroedersstraat 10, B-3000 Leuven,
Belgium
3
Benaki Phytopathological Institute, Laboratory of Pesticides Toxicology, 8 St. Delta Street,
Athens, Kifissia 14561, Greece
4
Department of Chemoinformatics, NovaMechanics Ltd, John Kennedy Ave 62-64, Nicosia
1
046, Cyprus
5
Department of Pharmacognosy and Natural Products Chemistry, Faculty of Pharmacy,
University of Athens, Panepistimiopolis, Zografou 15771, Athens, Greece
Corresponding author
Serkos A. Haroutounian, Ph.D. – Professor of Chemistry
Agricultural University of Athens, Department of Sciences, Chemistry Laboratory, Iera odos
7
5, Athens 11855, Greece
Τel.: +30 210 529 4247, +30 210 529 4246
Fax: +30 210 529 4265
E-mail: sehar@aua.gr

ACCEPTED MANUSCRIPT
Abstract
A series of novel isoxazole derivatives were efficiently synthesized through the
adaptation/modification of an in situ synthetic procedure for pyrazoles. All novel
compounds were tested against four different cell lines to evaluate their
antiproliferative activity. Based on the Hela cells results of this study and previous
work, a classification model to predict the anti-proliferative activity of isoxazole and
pyrazole derivatives was developed. Random Forest modelling was used in view of
the development of an accurate and reliable model that was subsequently validated. A
virtual screening study was then proposed for the design of novel active derivatives.
Compounds 9 and 11 demonstrated significant cytostatic activity; the fused isoxazole
derivative 18 and the virtually proposed compound 2v, were proved at least 10 times
more potent as compared to compound 9, with IC50 values near and below 1ꢀM. In
conclusion, a new series of isoxazoles was exploited with some of them exhibiting
promising cytostatic activities. Further studies on the substitution pattern of the
isoxazole core can potentially provide compounds with cytostatic action at the nM
scale. In this direction the in silico approach described herein can also be used to
screen existing databases to identify derivatives with anticipated activity.
Keywords: Isoxazoles, chemical synthesis, classification model, virtual screening,
antiproliferative activity, in silico

ACCEPTED MANUSCRIPT
1
. Introduction
Cancer is considered as one of the most multifactorial diseases of the world resulting
from a combined influence of genetic and environmental factors and involving a
multitude of cells, receptors, soluble factors and extracellular matrix molecules. As
such, cancer requires multifactorial treatment, targeting different cell types and/or
signaling pathways [1, 2] Consequently, a considerable focus has recently been
devoted to the discovery and development of new, more selective anticancer agents
[
3, 4] that are capable to intervene with these pathways.
Among the diverse azaheterocyclic ring systems considered as potent antitumor
agents, the isoxazole core structure constitutes one of the most promising five-
membered heterocyclic systems [5, 6]. In particular, several
O
N
OMe
Cl
HO
HO
NHEt
O
NHEt
O
OH
O
OH
O
N
N
NVP-AUY922
VER-52296)
VER-50589
(
Figure 1. Chemical structures of resorcinylic isoxazoles VER-50589 and NVP-AUY922.
resorcinolic isoxazole amides, exemplified by the molecule of NVP-AUY922 (Figure
1
9
) have been determined as very potent inhibitors of heat shock proteins (HSP90) [7-
]. The molecule of NVP-AUY922 inhibits the human tumor cell proliferation and
migration, displaying significant anti-angiogenic properties, as was established for
the inhibition of tube formation of human endothelial cells and the reduction of
microvessels density in various tumor xenografts [7]. Thus, this compound is under
Phase II clinical trials for several years [8].
In this view, a considerable research effort has been initiated towards the thorough
investigation of the antitumor properties of compounds containing the isoxazole
residue. Indicatively and aiming to enhance their antitumor properties, various

ACCEPTED MANUSCRIPT
isoxazole derivatives have been conjugated with pharmacophore benzodiazepines
[
10], while a 3,5-diaryl isoxazole derivative was assessed to display high affinity for
the retinoic acid receptor (RAR) combined with selective apoptosis-inducing activity
11].
[
In the course of our longstanding research interest concerning the development of
bioactive azaheterocycles ([12][13]), we have relied on the abovementioned data for
the design, synthesis and bioactivity assessment of a series of novel compounds
containing in their structural framework a fused isoxazole residue. The latter was
efficiently approached through the developed and adapted herein one pot addition of
hydroxylamine to tetralone and acetophenone substrates. The anti-proliferative
activities of the novel compounds were evaluated against a panel of two tumor and
two endothelial cell lines and the respective results were elaborated into in silico
modelling. The latter is emerged as a useful tool for the identification of new leads
with improved characteristics [14-16], via the employment of different modelling
methods which minimize the time and cost associated with their identification. In this
context, a classification model was developed that combines the isoxazole and
pyrazole structures with their antiproliferative activity. The virtual screening model
development was based on the inhibition of Hela cells.The model was fully validated
and proven robust and accurate.Therefore was utilized to virtually explore novel
structures and succeeded in identifying novel potent inhibitors that were
experimentally validated.
2
. Results and Discussion
2
.1 Synthesis
The reported to date methods of constructing the isoxazole structural framework refer
to the following reactions: i) cycloaddition of alkenes/alkynes to nitrile oxides, ii)
addition of hydroxylamine to 1,3-diketones, 1,3-dialkynes or α,β-unsaturated ketones,
aldehydes or nitriles, iii) cyclization of alkynyl oxime ethers, and iv) palladium-
catalyzed four-component coupling, as reported in 2005 [17].
In this work, good to excellent yields of the target isoxazoles were obtained through
the adaptation/modification of a methodology published in 2006 for the in situ
synthesis of pyrazoles by addition of hydrazines to 1,3-diketones [18]. This method

ACCEPTED MANUSCRIPT
proceeds via a similar mechanistic pattern as the one pot addition of hydroxylamine to
various 1,3-diketones, constituting a fast and efficient synthetic route for constructing
various isoxazoles and/or the synthetically challenging isoxazole-containing fused
rings.
More specifically, in the presence of lithium hexamethyl disilazane (LHMDS) and
anhydrous conditions the easily accessible acetophenone 1-4 substrates were
converted to the corresponding anions, which were subsequently condensed with the
4
-methoxyanisoyl chloride to yield the 3-diketone derivatives (Scheme 1).
Consecutively, the latter was in situ condensed with hydroxylamine to provide in one
overall step the desired isoxazole structures (compounds 5-8, 64-80 % total yields).
Deprotection of the latter with boron tribromide (BBr ) at −78°C provided the
3
isoxazoles 9-12. Accordingly, the reaction of tetralones 13 and 14 with various
benzoyl chlorides resulted in good yields (49-67%) of the isoxazole-tetralone fused
ring system-containing compounds 15-18, 21 and 23 (Schemes 2 and 3). Compounds
1
6, 18 and 21 were also deprotected in almost quantitative yields by reaction with
BBr providing the target derivatives 19, 20 and 22.
3
The structure of the synthesized isoxazoles was elucidated using several 2D-NMR
methods, such as the COSY, HSQC and HMBC experiments. In particular, the
configuration of compounds 5-12 was assigned through the long range correlation of
C-5 (downfield > 166 ppm) with H-6’ and H-2’, which is characteristic for all similar
compounds and indicative of the proposed structures. Similarly, the correlation
between C-5 and H-7 and H-13 in the HMBC experiment for isoxazoles 15-23 was
used for the unambiguous assignment of the synthesized compounds (see supporting
information).

ACCEPTED MANUSCRIPT
p-Anisoyl chloide
O
O
O
Cl
O
O
ACOH
R2
NH OH HCl
2
R2 LHMDS
Toluene
R1
O
R1
reflux
1
2
3
4
, R = OCH , R = H
1 3 2
, R = OCH , R = CH
3
1
3
2
, R = OCH , R = CH CH
1
3
2
2
3
, R = NO , R = H
1
2
2
6''
5''
N
N
O
3
O
6'
4''
B
O
1'
5'
1''
OH
BBr3
5
4
2'' 3''
4'
2'
R2
A
R2
CH Cl2
2
R3
-
78°C to r.t..
3'
R1
9
1
1
1
, R = OH, R = H
5
6
7
8
, R = OCH , R = H
3 2
1
3
2
0, R = OH, R = CH
3
, R = OCH , R = CH
3
2
1
3
2
3
1, R = OH, R = CH CH
, R = OCH , R = CH CH
3
2
2
3
1
3
2
2
3
2, R = NO , R = H
, R = NO , R = H
3
2
2
1
2
2
Scheme 1. One pot approach of isoxazoles from acetophenones substrates

ACCEPTED MANUSCRIPT
Scheme 2. Synthesis of tetralone-isoxazole fused ring systems
Scheme 3. Synthesis of substituted fused tetralone-isoxazole derivatives
2
.2. Anti-proliferative properties of the isoxazole derivatives
The anti-proliferative activity of the novel isoxazoles 5-12 and 15-23 was evaluated in
four different cell lines: two endothelial cell lines [human microvascular endothelial
cells (HMEC-1) and mouse brain endothelial cells (MBEC)] and two tumor cell lines
[
human cervical carcinoma (HeLa) and human breast carcinoma (MCF-7) cells]. Data
are expressed as IC50 (50% inhibitory concentration), defined as the compound
concentration that reduces cell proliferation by 50%, and are shown in Table 1.
Among the flexible analogues (5-12) compounds 5-8, containing an o-CH or NO at
3
2
R showed no anti-proliferative activity in any of the cell lines tested (IC >100 µM).
1
50
However, the replacement of the o-CH by an OH resulted in a significant
3

ACCEPTED MANUSCRIPT
improvement of their activity. In particular, compounds 9 and 11, without substituent
or an ethyl at R , respectively, demonstrated a significant cytostatic activity in
2
endothelial and tumor cell lines with IC values ranging from 2.5 to 6 µM for 9 and
5
0
9
.7 to 16 ꢀM for 11. Derivative 10, holding a methyl group at R showed comparable
2
activity as 11 in MBEC and HeLa cells, but proved to be 5-fold less potent in HMEC-
and MCF-7 cells. Compound 12 (containing a NO at R and no substituent at R )
1
2
3
2
showed only moderate cytostatic activity with IC values between 27 and 57 µM.
50
Thus, among this series of compounds, an OH at R and no substituent at R are
3
2
preferred, affording anti-proliferative activity in endothelial and tumor cell lines in the
lower micromolar range.
Next, the fused tetralone-isoxazole derivatives were evaluated for their anti-
proliferative activity. Similar to the flexible compounds described above, compounds
1
5-17 with an o-CH at R were inactive, regardless of whether they contained one or
3 1
two additional o-CH at R , R and/or R . However, 18 containing methoxy groups at
3
2
3
4
R , R , R and R proved 10-fold more active than the best flexible compound 9, with
1
2
3
4
IC values between 0.36 and 1.1 µM, depending on the cell line. Demethylation of
5
0
compounds 16 and 18 provided isoxazoles 19-20, with cytostatic activity in the lower
micromolar range. Notably, compound 20 was 10-fold less active than 18.
The substituted derivatives 21-23 displayed only moderate (IC of 34-81 µM for the
5
0
deprotected compounds) or no anti-proliferative activity.

ACCEPTED MANUSCRIPT
Table 1. Antiproliferative activities of novel isoxazoles
IC50 (µM)
Compound
HMEC-1
>100
MBEC
>100
HeLa
>100
MCF-7
>100
5
6
7
8
9
>100
>100
>100
>100
>100
>100
>100
>100
>100
>100
>100
>100
6.3 ± 0.5
47 ± 1
9.7 ± 1.1
27 ± 7
>100
2.5 ± 0.2
19 ± 3
13 ± 2
42 ± 1
>100
6 ± 0.9
15 ± 3
16 ± 1
57 ± 6
>100
5.2 ± 0.3
48 ± 2
9.8 ± 0.4
53 ± 2
>100
1
0
1
2
5
6
7
8
9
0
1
2
3
1
1
1
1
1
1
1
2
2
2
2
>100
>100
>100
>100
>100
>100
>100
>100
0.56 ± 0.12
7.4 ± 0.8
4.2 ± 2.2
>100
1.1 ± 0.1
6.8 ± 0.1
10 ± 1
>100
0.59 ± 0.08
10 ± 1
9.1 ± 0.6
>100
0.36 ± 0.05
2.1 ± 0.1
2.1 ± 0.1
>100
65 ± 20
>100
67 ± 20
>100
55 ± 36
34 ± 15
81 ± 19
>100
2
.3. In silico exploration
In an effort to develop a method for the design and identification of novel compounds
with improved antiprofilerative activity, a series of 33 novel inhibitors of Hela cell
proliferation, mainly isoxazole and pyrazole derivatives, were analysed
computationally. The data set contains the compounds described in this work
(compounds 5-12, 15-23) combined with sixteen pyrazoles that have been previously
explored as anti-proliferative agents [13].
Figure 2. Structures of pyrazoles reported in the literature (P1-P8 from Tzanetou et
al. 2012 and P9-P16 from Christodoulou et al. 2010)

ACCEPTED MANUSCRIPT
N
HN
HN
N
NO2
Br
O
NO2
NO2
HO
P2
P1
N
HN
N
HN
NH2
O
P4
O
P3
HN N
HN N
F
F
HO
O
P6
P5
HO
O
N
N
N N
F
F
HO
O
P7
P8

ACCEPTED MANUSCRIPT
HO
O
H
O
H
N
N
N
N
HO
HO
P10
P9
OH
OH
HO
O
N
O
N
N
N
H
H
HO
HO
P11
P12
OH
HO
N OH
O
N
N
O
N
N
H
HO
HO
P13
P14
OH
HO
N OH
O
OH
N
N
N
N
N
O
HO
HO
P16
P15

ACCEPTED MANUSCRIPT
For preprocessing, cleansing, attribute selection, modelling and validation of our data
we have created a KNIME workflow suitable to run step by step all the computational
tasks essential for the development of a validated model [19]. In particular, we have
created a KNIME workflow that implements the development of a predictive model
following the sequence described as following: Compounds and inhibition data were
imported and pre-processed, descriptors were calculated and selected, the Random
Forest algorithm was implemented, the produced model was validated and the domain
of applicability was defined.
The original dataset of the 33 inhibitors of the Hela cells was randomly partitioned
into training and validation set in a ratio of 70:30 consisting of 24 and 9 compounds
respectively. The training set was used to develop the QSAR models as described
below whereas the test set was not involved by any means in the model development.
For each compound 294 descriptors were calculated using Chemistry Development
Kit (CDK) [20], which account for the topological, geometric and structural
characteristics of compounds. As some of the descriptors do not have any
discrimination power (i.e. they have no variation) a filter was applied for their
removal. In total 159 descriptors remained to be used as possible inputs during the
QSAR model development.
The CfsSubset variable selection with BestFirst evaluator method was then applied on
the training data to select the most significant, among the 159 available descriptors.
Four descriptors were selected as the most important for the development of the
model. The selected descriptors are BCUTp-1l, Wlambda3unity, Wnu2unity and
Weta1unity.
BCUT molecular descriptor is an eigenvalue based descriptor that have been shown to
be useful in chemical diversity metrics and have enabled implementation of cell-based
algorithms for diversity-related tasks [21]. BCUT is based on a weighted version of
the Burden matrix which takes into account both the connectivity as well as atomic
properties of a molecule [22]. BCUTp-1l corresponds to lowest atomic related
polarizability weighted value.
Wlambda3unity, Wnu2unity and Weta1unity belong to the Weighted Holistic
Invariant Molecular descriptors (WHIM descriptors) which are based on statistical
indices calculated on the projections of the atoms along principal axes [23]. WHIM

ACCEPTED MANUSCRIPT
descriptors encode information relevant to 3D molecular structure, size, shape,
symmetry and atom distribution with respect to invariant frames [24].
The training set contains 24 compounds (15 “actives” and 9 “inactive”) and the test
set 9 compounds (5 “actives” and 4 “inactive”). The cutoff value for the
discrimination between “actives” and “inactives” was set to IC > 100 ꢀM.
5
0
A classification model has been developed in order to separate active from inactive
compounds and filter out the inactive ones. Random Forest (RF) classification
technique with 15 trees, implemented in the WEKA program [25], was used to
discriminate between the different classes.
After the classification model was trained, prediction on the activity of test
compounds was performed. The experimental values and the predictions for both
training and test examples are presented in Table 2. The confusion matrix for the
cross validation method and model predictions on the external test set, are presented
in Tables 3 and 4. The performance of the model was evaluated based on validation
measurements described before. The significance, accuracy and robustness of the
model are illustrated by the corresponding statistics (see Table 5). In particular, the
application of the 10-fold cross validation method produced the following statistics:
precision = 92.9%, sensitivity = 86.725%, specificity = 88.9% and accuracy = 87.5%.
Βy applying the model to the external test set, the following statistical results were
obtained: precision = 100%, sensitivity = 83.3% , specificity = 100% and accuracy =
8
9% .
The applicability domain was defined for all compounds that constituted the test sets
as described in the Materials and Methods section. Since all validation compounds
fell inside the domain of applicability, all model predictions for the external test set
were considered reliable except for compound P1 (APD limit 1.76).
Table 2. Experimental values and predictions
Compound ID
Observed Class
Inactive
Predicted Class
Inactive
5
6
Inactive
Inactive

ACCEPTED MANUSCRIPT
7
8
9
1
1
1
1
1
1
1
1
2
2
2
2
Inactive
Inactive
Active
Active
Active
Active
Inactive
Inactive
Inactive
Active
Active
Active
Inactive
Active
Active
Inactive
Inactive
Inactive
Inactive
Active
Active
Active
Inactive
*
Inactive
Active
Active
Active
Active
Inactive
Inactive
Inactive
Active
Active
Active
Inactive
Active
Active
Inactive
Inactive
Inactive
Inactive
Active
Active
Active
0*
1
2
5
6*
7
8
9
0*
1
2
3
P1*
P2
P3
P4
P5
P6*
P7

ACCEPTED MANUSCRIPT
P8
Active
Active
Active
Active
Active
Inactive
Active
Active
Active
Active
P9
Active
Active
Active
Active
Active
Active
Active
Active
P10*
P11
P12*
P13*
P14
P15
P16
*
compounds included in the test set
Table 3. Confusion Matrix (Training Set, 10-fold Crossvalidation) Random Trees.
Positive Predicted
Negative Predicted
2
Positive Observed
Active)
Negative Observed
Inactive)
1
3
(
1
8
(
Table 4. Confusion Matrix (Test Set) Random Trees.
Positive Predicted
Negative Predicted
0
Positive Observed
5
(Active)
Negative Observed
Inactive)
1
3
(
Table 5. Specificity, Sensitivity, Precision & Accuracy Statistics (Random Trees).
Specificity
0.889
Sensitivity
0.867
Precision
0.929
Accuracy
0.875
Crossvalidation

ACCEPTED MANUSCRIPT
External
1
0.833
1
0.89
Validation
2
.4. In silico virtual screening
The proposed method, due to the high predictive ability and simplicity [26-28] can be
considered as a useful aid to the costly and time-consuming experiments for the
synthesis and the determination of inhibition of the human cervical carcinoma (Hela)
cells by isoxazole and pyrazole derivatives. The design of novel active molecules by
the insertion, deletion, or modification of substituents on different sites of the
molecule and at different positions could therefore be guided by the proposed model.
To demonstrate the usefulness of the model we conducted a virtual screen to identify
potential new active targets within the models domain of applicability threshold. The
primary objective of the in silico screen was to determine whether the developed in
silico model could classify structures that are not included in the training or test sets,
as active or inactive. The secondary objective was to identify which structural
modifications could be tolerated using the domain of applicability. The ultimate role
of the in silico screen was to be used as a guide to the identification of the most
promising new synthetic targets.
During the virtual screening procedure the following chemistry driven small
molecules (1v-7v) have been designed (Figure 3) in an effort to identify novel potent
inhibitors. The design of these molecules builds upon isoxazole framework and
previous published paper of our group on rigid pyrazole derivatives (see compounds P
5
-8) focusing on the substitution pattern of the two aromatic rings A and B. By this
way isoxazoles 1v-3v, 5v-7v were designed. Moreover a pyrazole molecule 4v was
encompassed as a comparison basis molecule with our previous pyrazoles work
incorporated in this virtual screening. Isoxazoles 1v and 2v differ only in one
hydroxyl group that B ring possesses and the model classifies them as actives.
Isoxazole 3v is designed to show the effect of a) protection of hydroxyl group and b)
replacement of a hydroxyl group with fluorine atom on the anti-proliferative activity.
Although the model classifies this compound as active the experimental validation
shows the opposite.

ACCEPTED MANUSCRIPT
Isoxazole 5v was designed as an alternative derivative to 2v. Thus the hydroxyl group
(A ring) is replaced by a sulfamic moiety (an electron withdrawing group) while the
other two hydroxyl groups were replaced by the fluorine atom. This substitution leads
to an inactive compound based on the model. Compound 6v is a methoxylated
analogue of compound 2v where the tetralonic core of 2v was substituted by a
quinoline moiety. Also this methoxylation leads to an inactive compound (6v) based
on the model. Replacement of tetralone core to a more flexible one (indanone),
provided compound 7v, in an effort to elucidate effect of change on the rigid skeleton,
on the biological activity. Compound 7v was also predicted as inactive. Compounds
4
v-7v that were predicted as inactive were not proposed for experimental
investigation.
As mentioned above, among the virtually proposed structures, only three molecules
were identified as potentially active inhibitors (1v-3v) based on the developed model.
The structures of these compounds fall within the domain of applicability of the
model and were further explored experimentally. These three proposed molecules
were synthesized as shown in Schemes 2 and 3 and described in the experimental
part. Compounds 15, 17 and 23 were deprotected almost in quantitative yields to
provide compounds 1v-3v. The correlation between C-5, H-7 and H-13 in the HMBC
experiment for isoxazoles 1v-3v was used for the unambiguous assignment of the
synthesized compounds (see supporting information). The results of this in silico
work were validated as the synthesized compounds were evaluated in four different
cell lines as described before. Results are illustrated in Table 6. As seen from the
results, two of the proposed compounds, namely compounds 1v and 2v, were
experimentally verified as actives whereas compound 3v displayed only moderate or
no anti-proliferative activity It is important to highlight the fact that the tri-substituted
derivative 2v displayed the best activity, with IC values below 1 µM,
5
0
1v
2v
3v

ACCEPTED MANUSCRIPT
4v
5v
6v
7v
Figure 3. Chemistry Driven Virtual Screening Molecules
Table 6. Experimental results for the proposed compounds
IC50 (µM)
Compound
HMEC-1
6.7 ± 0.1
0.42 ± 0.12
46 ± 15
MBEC
4.6 ± 1.7
0.96 ± 0.4
38 ± 9
HeLa
3.3 ± 0.2
0.97 ± 0.08
>100
MCF-7
3.1 ± 0.8
0.72 ± 0.14
>100
1
2
3
v
v
v
3
. Experimental Section
3
.1. Chemistry
All anhydrous reactions were carried out under argon atmosphere. Solvents were
dried by distillation prior to use. Solvent mixtures used in chromatographic
separations are reported as volume to volume ratios. Starting materials were
purchased from Aldrich as analytical reagent grades and used without further
purification. Analytical thin-layer chromatography (TLC) was conducted on Merck
glass plates coated with silica gel 60 F254 and spots were visualized with UV light
or/and an alcohol solution of anisaldehyde. Flash column chromatography was
performed using Merck silica gel 60 (230-400 mesh ASTM).
Melting points (Mp) were determined on a Büchi melting point apparatus and are
1
13
uncorrected. H NMR (600 MHz) and C NMR (300 MHz) data were recorded on a
Bruker Avance 600 spectrometer. COSY, HMQC, and HMBC NMR data were
1
obtained using standard Bruker microprograms. The coupling constants are H and
1
5
N NMR recorded in Hertz (Hz) and the chemical shifts are reported in parts per
million (δ, ppm), downfield from tetramethylsilane (TMS) that was used as an
internal standard.

ACCEPTED MANUSCRIPT
Infrared spectra were obtained on a Nicolet Magna 750, series II spectrometer. HPLC
separations were performed using an Agilent 1100 series instrument with a variable
wavelength UV detector and coupled to HP Chem–Station utilizing the
manufacturer’s 5.01 software package.
3
3
4
.2. Indicative Procedure for the Synthesis of Isoxazoles
.2.1. (3,5 Bis(4-methoxyphenyl)isoxazole) (5)
-Methoxyacetophenone (0.5 g, 3.3 mmol) was dissolved in 9 mL of anhydrous
toluene, cooled to 0 °C under argon and LiHMDS (3.5 mL, 1.0 M in THF, 3.5 mmol)
and added via a syringe in one-pot under stirring. After approximately 1 min to allow
the formation of the anion, p-anisoylchloride (0.23 mL, 1.65 mmol) was added. The
ice bath was removed and after 1 min 4 mL of AcOH was added. The reaction was
stirred for an additional 2 min and consecutively EtOH (17 mL), THF (4 mL) and an
excess of hydrochloric hydroxylamine (1.03 g, 15 mmol) were added. The reaction
mixture was refluxed for 15 min and completion of the reaction was verified by TLC.
Then, the resulting solution was added to 5 mL of 1.0 M NaOH and extracted twice
with EtOAc (2 × 20 mL). The combined organic fractions were washed with brine,
dried over Na SO , and evaporated under reduced pressure. The resulting residue was
2
4
purified by flash column chromatography using a mixture of CH Cl /n-hexane (9:1)
2
2
to provide isoxazole 6 (0.75 g, 80 %) as white crystalline needles.
1
H NMR (CDCl ): δ = 3.89 (6H, s, -OCH ), 6.67 (1H, s, H-4), 7.01 (4H, d, J 8.5 Hz,
3
3
13
H-3’, Η-5’, H-3’’, Η-5’’), 7.80 (4H, d, J 8.5 Hz, H-2΄, H-6΄, H-2΄΄, H-6΄΄). C NMR
(
CDCl ): δ = 55.3(-OCH -OCH ), 95.8 (C-4), 114.3 (C-3’, C-5’, C-3’’, C-5’’), 121.0
3 3, 3
(C-1’, C-1’’), 128.0 (C-2’, C-6’, C-2’’, C-6’’), 160.9 (C-4’,C-4’’), 162.4 (C-3), 169.9
(
C-5). Analysis for C H NO Calculated: C, 72.58; H, 5.37; N, 4.98; Found: C,
:
3
1
7
15
7
2.32; H, 5.30; N, 5.09
3
.2.2. 3,5-Bis-(4-Methoxyphenyl)-4-methylisoxazole (6)
This compound was obtained from 1-(4-methoxyphenyl)propan-1-one and p-
anisoylchloride as yellow crystalline solid (0.67 g, 76%), using the aforementioned
procedure and purified by flash column chromatography (CH Cl /n-hexane 8.5:1.5).
2
2
1
H NMR (CDCl ): δ = 1.95 (3H, s, CH ), 3.89 (6H, s, -OCH ), 7.04 (4H, m, H-3’’, Η-
3
3
3
5
’’,H-3’, Η-5’), 7.64 (2H, d, J 9.0 Hz, H-2΄΄, H-6΄΄), 7.71 (2H, d, J 9.0 Hz, H-2΄, H-

ACCEPTED MANUSCRIPT
6
7
΄). Analysis for C H NO : Calculated: C, 73.20; H, 5.80; N, 4.74; Found: C,
1
8
17
3
3.00; H, 5.94; N, 4.64
3
.2.3. 4-Ethyl-3,5-bis(4-methoxyphenyl)isoxazole (7)
Reaction of 4-methoxybutyrophenone (0.5 g, 2.8 mmol) with p-anisoylchloride (0.19
mL, 1.4 mmol) using the described procedure provided ‒after recrystallization from
Et O− an isoxazole as yellow crystalline solid (0.55 g, 64%)
.
2
1
H NMR (CDCl ): δ = 1.17 (3H, t, J 7.5 Hz, CH ), 2.73 (2H, q, J 7.5 Hz,–CH –), 3.88
3
3
2
(
3H, s, -OCH ), 3.89 (3H, s, -OCH ), 7.35 (4H, m, H-3’’, Η-5’’,H-3’, Η-5’), 7.6 (2H,
3 3
d, J 8.5 Hz, H-2΄΄, H-6΄΄), 7.7 (2H, d, J 8.5 Hz, H-2΄, H-6΄). Analysis for C H NO :
19
19
3
Calculated: C, 73.77; H, 6.19; N, 4.53; Found: C, 73.97; H, 6.29; N, 4.44.
3
.2.4. 3-(4-Methoxyphenyl)-5-(4-nitrophenyl)-isoxazole (8)
Compound 8 was prepared as a yellow crystalline solid (0.6 g, 75%) following the
above mentioned general procedure with 4-nitroacetophenone (0.5 g, 3.0 mmol) and
p-anisoylchloride (0.2 mL, 1.5 mmol) substrates and purified by flash column
chromatography (CH Cl /n-hexane 8:2).
2
2
1
H NMR (CDCl ): δ = 3.92 (3H, s, -OCH ), 6.85 (1H, s, H-4), 7.02 (2H, d, J 8.5 Hz,
3
3
H-3’, Η-5’), 8.02 (2H, d, J 8.5 Hz, H-2΄, H-6΄), 8.13 (2H, d, J 9.0 Hz, H-2΄΄, H-6΄΄),
.34 (2H, d, J 8.5 Hz, H-3’’, Η-5’’).
Analysis for C H N O : Calculated: C, 64.86; H, 4.08; N, 9.46; Found: C, 64.75; H,
8
1
6
12
2
4
4
.00; N, 9.50
.2.5. 7-Methoxy-3-(4-methoxyphenyl)-4,5-dihydronaphtho[2,1-d]isoxazole (15)
3
Compound 15 was synthesized as yellow crystalline solid (0.57 g, 67%), from 6-
methoxy-1-tetralone 13 (0.5 g, 2.8 mmol) and p-anisoyl chloride (0.19 mL, 1.4
mmol), as previously described, and purified by flash column chromatography
(
CH Cl /n-hexane 8.5:1.5).
2 2
1
H NMR (CDCl ): δ = 2.95 (2H, t, J 7.5 Hz, H-13), 3.07 (2H, t, J 7.5 Hz, H-12),
3
3
.87 (6H, m, -OCH ), 6.86 (1H, dd, J 9.0, 2.0 Hz, H-8), 6.88 (1H, d, J 2.0 Hz, H-10),
3
7
.02 (2H, d, J 8.5 Hz, H-3’, H-5’), 7.67 (1H, d, J 9.0 Hz, H-7) 7.72 (2H, d, J 8.5 Hz,
1
3
H-2’, H-6’). C NMR (CDCl ): δ =19.4 (C-13), 29.5 (C-12), 55.3 (-OCH , -OCH ),
3
3
3
1
08.9 (C-4), 111.8 (C-8), 114.2 (C-10, C-3’, C-5’), 118.7 (C-6), 122.0 (C-1’), 123.4
(C-7), 128.9 (C-2’, C-6’), 138.4 (C-11), 160.0 (C-3), 160.4 (C-9, C-4’), 166.0 (C-5).

ACCEPTED MANUSCRIPT
Analysis for C H NO : Calculated: C, 78.07; H, 5.90; N, 9.10 Found: C, 77.94; H,
1
9
17
3
5
.85; N, 9.21
3
.2.6. 7,8-Dimethoxy-3-(4-methoxyphenyl)-4,5-dihydronaphtho[2,1-d]isoxazole (16)
Compound 16 was synthesized as yellow crystalline solid (0.43 g, 51%), according to
the previously described procedure, using 6,7-dimethoxy-1-tetralone 14 (0.5 g, 2.5
mmol) and p-anisoyl chloride (0.17 mL, 1.25 mmol) substrates and purified by flash
column chromatography (CH Cl /n-hexane 9:1).
2
2
1
H NMR (CDCl ): δ = 2.95 (2H, t, J 7.5 Hz, H-13), 3.03 (2H, t, J 7.5 Hz, H-12), 3.88
3
(
3H, C-4’-OCH ), 3.94 (3H, C-9-OCH ), 3.97 (3H, C-8-OCH ), 6.83 (1H, s, H-10),
3
3
3
7
.02 (2H, d, J 8.5 Hz, H-3’, H-5’), 7.67 (1H, d, J 9.0 Hz, H-7) 7.72 (2H, d, J 8.5 Hz,
1
3
H-2’, H-6’). C NMR (CDCl ): δ =19.7 (C-13), 28.8 (C-12), 55.3 (C-4’-OCH ,),
3
3
5
6.0 (C-9-OCH ,), 56.1 (C-8-OCH ,), 108.8 (C-4), 105.3 ( C-7), 111.5 (C-10), 114.2
3 3
(C-3’, C-5’), 118.2 (C-6), 122.1 (C-1’), 128.9 (C-2’, C-6’), 129.4 (C-11), 148.1 (C-8),
1
49.7 (C-9), 159.5 (C-3,), 160. 1 (C-4’), 166.3 (C-5). Analysis for C H NO :
20 19 4
Calculated: C, 71.20; H, 5.68; N, 4.15; Found C, 71.30; H, 5.75; N, 4.18.
3
.2.7. 3-(3,4-Dimethoxyphenyl)-7-methoxy-4,5-dihydronaphtho[2,1-d]isoxazole, (17)
Compound 17 was obtained as a yellow crystalline solid (0.65 g, 67%) from 6-
methoxy-1-tetralone 13 (0.5 g, 2.8 mmol) and dimethoxybenzoyl chloride (0.23 mL,
1
.4 mmol) substrates and purified by flash column chromatography (CH Cl /n-hexane
2 2
9
:1).
1
H NMR (CDCl ): δ = 2.98 (2H, t, J 7.5 Hz, H-13), 3.07 (2H, t, J 7.5 Hz, H-12),
3
3
.86 (3H, C-4’-OCH ), 3.95 (3H, , C-9-OCH ), 3.97 (3H, C-8-OCH ), 6.86 (2H, m,
3 3 3
H-8, H-10), 6.97 (1H, d, J 8.5 Hz H-5’), 7.27 (1H, s, H-7), 7.4 (1H, s, H-6’), 7.67
(
1H, d, J 9.5 Hz H-2’). Analysis for C H NO : Calculated: C, 71.20; H, 5.68; N,
20 19 4
4
.15; Found C, 71.37; H, 5.77; N, 4.21.
.2.8. 3-(3,4-Dimethoxyphenyl)-7,8-dimethoxy-4,5-dihydronaphtho[2,1-d]isoxazole
3
(18)
Compound 18 was synthesized according to the above-mentioned procedure for the
synthesis of isoxazoles using 6,7-dimethoxy-1-tetralone 14 (0.5 g, 2.5 mmol) and
dimethoxybenzoyl chloride (0.21 mL, 1.25 mmol) as reactants. Recrystallization from
Et O provided 18 (0.58 g, 63%) as pale-yellow crystals.
2

ACCEPTED MANUSCRIPT
1
H NMR (CDCl ): δ = 2.97 (2H, t, J 7.5 Hz, H-13), 3.05 (2H, t, J 7.5 Hz, H-12),
3
3
6
8
.95 (3H, C-4’-OCH ), 3.96 (3H, C-9-OCH ), 3.98 (6H, m, C-8-OCH C-3’-OCH ),
3 3 3, 3
.85 (1H, s, H-10), 6.99 (1H, d, J 8.5 Hz, H-5’), 7.27 (1H, s, H-7), 7.28 (1H, dd, J
.5, 2.0 Hz, H-6’), 7.39 (1H, d, J 1.8 Hz, H-2’).
1
3
C NMR (CDCl ): δ =19.7 (C-13), 28.9 (C-12), 56.0 (-OCH , -OCH -OCH -
3
3
3,
3,
OCH ), 108.8 (C-4), 105.3 ( C-7), 110.4 (C-2’), 110.9 (C-5’), 111.4 (C-10), 117.9 (C-
3
6
), 120.4 (C-6’), 122.2 (C-1’), 129.0 (C-11), 148.1 (C-8), 149.0 (C-3’), 149.6 (C-4’),
50.0 (C-9), 159.5 (C-3,), 166.4 (C-5).
1
Analysis for C H NO : Calculated: C, 68.65; H, 5.76; N, 3.81; Found C, 68.77; H,
2
1
21
5
5
.81; N, 3.91.
3
.2.9.
7-Methoxy-3-(4-(trifluoromethylthio)phenyl)-4,5-dihydronaphtho[2,1-
d]isoxazole ( 21)
Compound 23 was synthesized as pale brown crystalline solid (0.52 g, 49%) from 6-
methoxy-1-tetralone 13 (0.5 g, 2.8 mmol) and 4-(trifluoromethylthio)benzoyl chloride
(0.24 mL, 1.4 mmol) substrates in accordance with the general procedure for the
synthesis of isoxazoles and purified by flash column chromatography (CH Cl /n-
2
2
hexane 8.5:1.5).
1
H NMR (CDCl ): δ = 2.98 (2H, t, H-13), 3.09 (2H, t, H-12), 3.87 (3H, -OCH ), 6.87
3
3
(1H, d, J 2 Hz, H-10), 6.88 (1H, dd, J 8.5, 2.0 Hz, H-8), 7.68 (1H, d, J 8.5 Hz, H-7),
1
3
7
.79 (2H, dd, J 8.5, 2.0 Hz, H-3’, H-5’), 7.84 (2H, dd, J 8.5, 2.0 Hz, H-2’, H-6’). C
NMR (CDCl ): δ =19.7 (C-13), 28.9 (C-12), 56.0 (-OCH ,) 108.8 (C-4), 112.0 ( C-8),
3
3
1
14.4 (C-10), 118.2 (C-6), 123.6 (C-7), 128.4 (C-2’, C-6’), 132.2 (C-1’), 136.6 (C-3’,
C-4’ C-5’), 138.2 (C-11), 158.8 (C-3), 160.9 (C-9), 167.1 (C-5).
C H F NO S: Calculated: C, 60.47; H, 3.74; N, 3.71 Found: C, 60.59; H, 3.84; N,
19
14
3
2
3
.81
.2.10. 3-(4-Fluorophenyl)-7-methoxy-4,5-dihydronaphtho[2,1-d]isoxazole, (23)
3
Compound 25 was synthesized with respect to the above-mentioned procedure for the
synthesis of isoxazoles using 6-methoxy-1-tetralone 13 (0.5 g, 2.8 mmol) and 4-
fluorobenzoyl chloride (0.17 mL, 1.4 mmol) as reactants. Purification by flash column
chromatography (CH Cl /n-hexane 8:2) provided (0.52 g, 63%) a pale yellow
2
2
crystalline solid.
1
H NMR (CDCl ): δ = 2.98 (2H, t, H-13), 3.09 (2H, t, H-12), 3.87 (3H, -OCH ), 6.87
3
3
(1H, d, J 2 Hz, H-10), 6.88 (1H, dd, J 8.5, 2.0 Hz, H-8), 7.68 (1H, d, J 8.5 Hz, H-7),
7
.79 (2H, dd, J 8.5, 2.0 Hz, H-3’, H-5’), 7.84 (2H, dd, J 8.5, 2.0 Hz, H-2’, H-6’).

ACCEPTED MANUSCRIPT
C H FNO : Calculated: C, 73.21; H, 4.78; N, 4.74; Found C, 73.31; H, 4.88; N, 4.79
18
14
2
3
3
.3. Indicative Procedure for the Demethylation of Isoxazoles
.3.1. 4,4'-(Isoxazole-3,5-diyl)diphenol (9)
This compound was obtained as pale yellow crystalline solid (yield, 88%), by
demethylation of isoxazole 5, in accordance with the previously described procedure.
1
H NMR (CD OD): δ = 6.92 (2H, d, J 8.5 Hz, H-3’’, Η-5’’), 6.96 (2H, H-3’, Η-5’),
3
7
.33 (1H, s, H-4), 7.48 (2H, J 8.5, H-2΄΄, H-6΄΄), 7.62 (2H, d, J 8.5 Hz, H-2΄, H-6΄).
C NMR (CD OD): δ = 9.0 (CH ), 98.4 (C-4), 115.6 (C-3’, C-5’, C-3’’, C-5’’),
1
3
3
3
1
1
5
21.1 (C-1’, C-1’’), 129.2 (C-2’, C-6’), 130.4 (C-2’’, C-6’’), 160.2 (C-4’, C-4’’),
64.9 (C-3), 166.5 (C-5). Analysis for C H NO : Calculated: C, 71.14; H, 4.38; N,
1
5
11
3
.24; Found: C, 71.24; H, 4.29; N, 5.32.
3
.3.2. 4,4'-(4-Methylisoxazole-3,5-diyl)diphenol (10)
To a stirred solution of isoxazole 6 (0.1 g, 0.34 mmol) in CH Cl (15 mL), cooled at
2
2
‒
78°C, BBr (4.74 mmol, 1.0 M in CH Cl , 5 equiv per bond, 4.74 mL) was added
3 2 2
dropwise. The reaction was stirred for 1 hr at ‒78°C and 17 hr at ambient temperature.
Then, the reaction was placed in an ice-bath, quenched with MeOH and water (10
mL) and HCl 1N (3 to 4 drops) was added to neutralize the pH. The organic layer was
obtained by extraction with EtOAc (2 × 15 mL). The combined organic extracts were
washed with water, brine, dried over Na SO , filtered and evaporated under reduced
2
4
pressure. The product was purified with flash column chromatography (n-
Hexane/EtOAc 4:6) or recrystallized from Et O resulting in a yellow crystalline solid
2
(87% yield).
1
H NMR (CD OD): δ = 2.27 (3H, CH ), 6.95 (2H, d, J 8.5 Hz, H-3’’, Η-5’’), 6.96
3
3
(2H, H-3’, Η-5’), 7.51 (2H, J 8.5, H-2΄΄, H-6΄΄), 7.62 (2H, d, J 8.5 Hz, H-2΄, H-6΄).
1
3
C NMR (CD OD): δ = 9.0 (CH ), 107.4 (C-4), 116.2 (C-3’, C-5’, C-3’’, C-5’’),
3
3
1
1
5
20.7 (C-1’, C-1’’), 129.3 (C-2’, C-6’), 130.4 (C-2’’, C-6’’), 160.0 (C-4’, C-4’’),
64.9 (C-3), 167.0 (C-5). Analysis for C H NO : Calculated: C, 71.90; H, 4.90; N,
1
6
13
3
.24; Found: C, 71.80; H, 4.83; N, 5.18.
3
.3.3. 4,4'-(4-Ethylisoxazole-3,5-diyl)diphenol (11)
This compound was obtained from isoxazole 7 as yellow crystalline solid, in
accordance with the previously described procedure (yield, 90%).

ACCEPTED MANUSCRIPT
1
H NMR (Acetone-d6): δ = 1.16 (3H, t, J 7.5 Hz, CH ), 2.79 (2H, q, J 7.5 Hz, -CH -),
3
2
7
.05 (2H, d, J 8.5 Hz, H-3’’, Η-5’’), 7.08 (2H, d, J 8.5, H-3’, Η-5’), 7.59 (2H, d, J 8.5
1
3
Hz, H-2΄΄, H-6΄΄) 7.68 (2H, d, J 8.5, H-2΄, H-6΄). C NMR(Acetone-d6): δ = 13.6 (-
CH3), 15.6 (-CH2-), 115.5 (C-3’’, C-5’’), 115.7 (C-3’, C-5’), 119.7 (C-1’), 121.2 ( C-
1
’’), 128.3 (C-2’, C-6’), 129.5 (C-2’’, C-6’’), 158.2 (C-4’, C-4’’), 162.7 (C-3), 165.1
(
C-5). Analysis for C H NO : Calculated: C, 72.58; H, 5.37; N, 4.98; Found: C,
1
5
11
3
7
2.44; H, 5.29; N, 4.92.
3
.3.4. 4-(5-(4-Nitrophenyl)isoxazol-3-yl)phenol (12)
This compound was obtained from nitrophenyl isoxazole 8 as yellow crystalline solid,
in accordance with the previously described procedure (yield, 95%).
1
H NMR (Acetone): δ = 7.01 (2H, d, J 8.5 Hz, H-3’, Η-5’), 7.33 (1H, s, H-4), 7.82
(2H, d, J 8.5 Hz, H-2΄, H-6΄), 8.23 (2H, J 8.5, H-2΄΄, H-6΄΄) 8.40 (2H, H-3’’, Η-5’’).
1
3
C NMR (Acetone): δ = 97.0 (C-4), 116.6 (C-3’, C-5’), 124.8 (C-3’’, C-5’’), 128.2
(C-2’, C-6’), 128.6 (C-2’’, C-6’’), 120.4 (C-1’), 137.2 ( C-1’’), 150.3 (C-4’’), 161.2
(
C-4’), 162.3 (C-3), 173.2 (C-5). Analysis for C H N O : Calculated: C, 63.83; H,
1
5
10
2
4
3
.57; N, 9.92; Found: C, 63.94; H, 3.49; N, 10.01.
3
.3.5. 3-(4-Hydroxyphenyl)-4,5-dihydronaphtho[2,1-d]isoxazol-7-ol (1v)
This compound was obtained by demethylation from isoxazole 15 as a pale green
crystalline solid, according to the general deprotection procedure (yield, 92%).
1
H NMR (DMSO): δ = 2.84 (2H, t, J 7.4 Hz, H-13), 2.96 (2H, t, J 7.4 Hz, H-12),
6
.70 (1H, dd, J 8.5, 2.0 Hz H-8), 6.79 (1H, d, J 2.0 Hz H-10), 6.95 (2H, d, J 8.5 Hz,
H-3’, Η-5’), 7.51 (2H, J 8.5 Hz, H-2΄, H-6΄), 7.48 (1H, d, J 8.5 Hz, H-7).
1
3
C NMR (DMSO): δ =18.9 (C-13), 28.5 (C-12), 108.7 (C-4), 113.3 ( C-8), 114.2 (C-
’), 115.3 (C-10), 115.7 (C-5’), 116.3 (C-6), 117.4 (C-3’) 118.7 (C-6’), 120.7 (C-1’),
22.7 (C-7), 146.2 (C-4’), , 159.4 (C-3), 159.7 (C-9), 166.3 (C-5).
2
1
Analysis for C H NO : Calculated: C, 73.11; H, 4.69; N, 5.02; Found C, 72.97; H,
1
7
13
3
4
.58; N, 4.91.
3
.3.6. 3-(4-Hydroxyphenyl)-4,5-dihydronaphtho[2,1-d]isoxazole-7,8-diol (19)
This compound was obtained from isoxazole 16 as pale yellow crystalline solid, in
accordance with the general deprotection procedure (yield, 89%).

ACCEPTED MANUSCRIPT
1
H NMR (DMSO): δ = 2.85 (2H, t, J 7.4 Hz, H-13), 2.96 (2H, t, J 7.4 Hz, H-12), 6.90
1H, d, J 2.1 Hz H-10), 6.95 (2H, d, J 8.5 Hz, H-3’, Η-5’), 7.48 (1H, s, H-7), 7.53
2H, J 8.5 Hz, H-2΄, H-6΄).
(
(
Analysis for C H NO : Calculated: C, 69.15; H, 4.44; N, 4.74; Found C, 69.30; H,
1
7
13
4
4
.37; N, 4.69.
3
.3.7. 4-(7-Hydroxy-4,5-dihydronaphtho[2,1-d]isoxazol-3-yl)benzene-1,2-diol (2v)
This compound was prepared from isoxazole 17 as pale yellow crystalline solid,
following the general deprotection procedure (yield, 92%).
1
H NMR (DMSO): δ = 2.85 (2H, t, H-13), 2.96 (2H, t, H-12), 6.73 (1H, dd, J 8.5, 2.0
Hz H-8), 6.78 (1H, d, J 2.0 Hz H-10), 6.85 (1H, d, J 8.5 Hz, H-5’), 7.03 (1H, dd, J
1
3
8
.5, 2.0 Hz, H-6’), 7.18 (1H, d, J 2.0 Hz, H-2’), 7.45 (1H, d, J 8.5 Hz, H-7). C NMR
(DMSO): δ =18.6 (C-13), 28.4 (C-12), 108.7 (C-4), 113.5 ( C-8), 114.2 (C-2’), 115.3
(C-10), 115.7 (C-5’), 116.3 (C-6), 118.7 (C-6’), 120.7 (C-1’), 122.7 (C-7), 138.6 (C-
1
1), 146.2 (C-4’), 159.4 (C-3’), 159.4 (C-3), 159.5 (C-9), 165.6 (C-5). C H NO :
17 13 4
Calculated: C, 69.15; H, 4.44; N, 4.74; Found: C, 69.25; H, 4.39; N, 4.67.
3
.3.8. 3-(3,4-Dihydroxyphenyl)-4,5-dihydronaphtho[2,1-d]isoxazole-7,8-diol (20)
This compound was obtained from isoxazole 18 as pale yellow crystalline solid,
following the general deprotection procedure (yield, 93%).
1
H NMR (DMSO): δ =2.84 (2H, t, J 7.4 Hz, H-13), 2.96 (2H, t, J 7.4 Hz, H-12), 6.87
(
1H, d, J 8.5 Hz, H-5’), 6.90 (1H, d, J 2.1 Hz H-10), 7.03 (1H, dd, J 8.5, 2.0 Hz, H-
’), 7.18 (1H, d, J 2.0 Hz, H-2’), 7.48 (1H, s, H-7).
C H NO : Calculated: C, 65.59; H, 4.21; N, 4.50; Found: C, 65.50; H, 4.19; N, 4.47.
6
17
13
5
3
.3.9. 3-(4-(Trifluoromethylthio)phenyl)-4,5-dihydronaphtho[2,1-d]isoxazol-7-ol, (22)
This compound was obtained from isoxazole 23 (0.1 g, 0.26 mmol), as pale green
crystalline solid following the general deprotection procedure (0.09 g, yield 94%).
1
H NMR (Acetone-d6): δ = 3.01 (2H, t, J 7.4 Hz, H-13), 3.07 (2H, t, J 7.4 Hz, H-12),
6
.87 (2H, m, ArH), 7.56 (1H, d, J 8.0 Hz, ArH), 7.90 (2H, d, J 8.0 Hz, ArH), 7.98
2H, d, J 8.5 Hz, ArH).
C H F NO S: Calculated: C, 60.47; H, 3.74; N, 3.71 Found: C, 60.59; H, 3.84; N,
(
19
14
3
2
3
.81

ACCEPTED MANUSCRIPT
3
.3.10. 3-(4-Fluorophenyl)-4,5-dihydronaphtho[2,1-d]isoxazol-7-ol (3v)
This compound was obtained from isoxazole 25 as pale yellow crystalline solid,
following the general deprotection procedure (yield, 94%).
1
H NMR (CD OD): δ = 2.93 (2H, t, H-13), 3.05 (2H, t, H-12), 6.78 (1H, dd, J 8.5,
3
2
.0 Hz H-8), 6.81 (1H, d, J 2.0 Hz H-10), 7.29 (2H, d, J 8.5 Hz, H-3’, H-5’), 7.53
1
3
(
1H, d, J 8.5, Hz, H-7), 7.80 (2H, dd, J 8.5, 2.0 Hz, H-2’, H-6’). C NMR (CD OD):
3
δ =18.7 (C-13), 28.8 (C-12), 108.7 (C-4), 113.6 ( C-8), 115.6/115.1 (C-3’, C-5’),
1
1
15.1 (C-10), 116.6 (C-6), 122.9/129.3 (C-2’, C-6’), 125.8 (C-1’), 159.2 (C-3),
59.5(C-9), 164.1 (C-4’), 166.4 (C-5).
C H FNO : Calculated: C, 72.59; H, 4.30; N, 4.98; Found C, 72.74; H, 4.23; N, 4.89
17
12
2
3
.4. Biological Part
.4.1. Cell cultures
3
Mouse brain endothelial cells (MBEC) were provided by Prof. M. Presta (Brescia,
Italy). Human cervical carcinoma (Hela) and human breast carcinoma (MCF-7) cells
were obtained from the American Type Culture Collection (ATCC, Middlesex, UK).
The cells were grown in Dulbecco’s modified minimum essential medium (DMEM,
Life Technologies, Inc., Rockville, MD) supplemented with 10 mM Hepes (Life
Technologies, Inc., Rockville, MD) and 10% fetal bovine serum (FBS, Harlan Sera-
Lab Ltd., Loughborough, UK). Human microvascular endothelial cells (HMEC-1)
were obtained from the CDC (Atlanta, USA) and grown in EGM-2 MV BulletKit
endothelial cell medium (Lonza, Verviers, Belgium).
3
.4.2. Cell proliferation assays
2
Cells were seeded in 48-well plates at 10,000 cells per cm . After 16 h, the cells were
incubated in fresh medium in the presence of 5-fold dilutions of the test compounds.
On day 5, cells were trypsinized and counted by a Coulter counter (Analis, Belgium).
The compound concentration that inhibits cell growth by 50 % (i.e. IC ) was
5
0
calculated based on cell counts in control cultures.
3
3
.5. Computational Part
.5.1. Molecular Descriptors - Separation into a training and a test set

ACCEPTED MANUSCRIPT
For each compound we calculated a large number of descriptors that account for their
chemical, physicochemical, electronic and quantum characteristics. Following the
optimization of the compound geometries using the PM6 method included in
MOPAC2009 [J.J.P. Stewart, J. Mol. Model. 13 (2007) 1173-1213, J.J.P. Stewart, J.
Mol. Model. 14 (2008) 499-535], the descriptors were calculated using the Chemistry
Development Kit (CDK) [20]. The PM6 method was chosen for the geometry
optimizations since it offered a good balance between speed and accuracy. A recent
paper highlighted the quality of models obtained by the PM6 method as similar to that
of models based on B3LYP32 (Density Functional Theory) [29]. After the
optimization, 294 input attributes were calculated for each compound including
topological, structural, electronic and physicochemical descriptors. The Hela
inhibition data along with the corresponding full set of descriptor values were used in
the variable selection procedure.
3
.5.2. Variable Selection
Variable selection techniques have become an apparent need in many
chemoinformatics applications and different methods have been successfully applied
as variable selection tools in QSAR problems. As some of the descriptors do not have
any discrimination power (i.e. they have no variation) a filter was applied for their
removal. In total 159 descriptors remained to be used as possible inputs during the
QSAR model development. Before running the classification methodology, the most
significant attributes among the 159 available were preselected by using CfsSubset
variable selection and BestFirst evaluator, which are included in Weka [25].
Correlation-based feature subset selection (CfsSubset) algorithm evaluated the value
of a subset of attributes by considering the individual predictive ability of each feature
along with the degree of redundancy between them. Subsets of features that were
highly correlated with the class while having low intercorrelation were preferred
3
.5.3. Modeling Methods
Random forest is a machine learning method that consists of many decision trees and
outputs a prediction combining the result given from the individual trees [Leo
Breiman (2001). Random Forests. Machine Learning. 45(1):5-32.]. Individual random
trees are grown based on samples from the original data set. These samples are used

ACCEPTED MANUSCRIPT
to construct training sets and build trees. A random number of attributes are chosen
for each tree. These attributes form the nodes and leafs of the tree. In this work, the
implementation of the random forest algorithm available in WEKA [25] was used
including 4 variables for the model development. Predictions were made by averaging
predicted activities over all trees in the final forest.
3
.5.4. Model Validation
The internal performance, as represented by goodness-of-fit and robustness, and the
predictivity of a model, as determined by external validation, needs to be evaluated.
The produced model was validated using external validation (Tropsha, 2010). The
model was internally and externally validated paying special attention to the
principles of model validation for accepting QSAR models as described by the
Organisation for Economic Cooperation and Development (OECD).
The validation of the proposed model was assessed by various validation techniques.
In particular the proposed classification models were fully validated using the
following measurements:
TP
TP+FP
Precision =
Sensitivity =
Specificity =
Accuracy =
(1)
(2)
(3)
(4)
TP
TP+FN
TN
TN+FP
TP+TN
TP+FP+FN+TN
where: TP = True Positive, FP = False Positive, TN = True Negative, FN = False
Negative
Confusion Matrix is also given as shown below:
Positive Predicted
Negative Predicted

ACCEPTED MANUSCRIPT
Positive Observed
Active)
Negative Observed
Inactive)
TP
FP
FN
TN
(
(
External validation was applied, by randomly splitting the dataset into training and
validation set in a proportion of 70:30. The separation of the data set was performed
using the Partitioning KNIME node by applying the default random seed. The use of
random seed provides reproducible results upon re-execution of the node. The 10
compounds that constituted the test set were not involved by any means in the training
procedure.
3
.5.5. Applicability Domain
In order for an in silico model to be used for screening new compounds, its domain of
application must be defined and predictions for only those compounds that fall into
this domain may be considered reliable. Similarity measurements were used to define
the domain of applicability of the two models based on the Euclidean distances
among all training compounds and the test compounds. The distance of a test
compound to its nearest neighbor in the training set was compared to the predefined
applicability domain (APD) threshold. The prediction was considered unreliable when
the distance was higher than APD. APD was calculated as follows:
APD = <d> +Zσ
(6)
Calculation of <d> and σ was performed as follows: First, the average of Euclidean
distances between all pairs of training compounds was calculated. Next, the set of
distances that were lower than the average was formulated. <d> and σ were finally
calculated as the average and standard deviation of all distances included in this set. Z
was an empirical cutoff value and for this work, it was chosen equal to 0.5. The
calculation of the domain of applicability was done using Enalos KNIME nodes
[
http://www.novamechanics.com/knime.php].
4
. Conclusions
A series of novel isoxazole derivatives were synthesized by a simple procedure based
on a modification of a methodology published in 2006 for the in situ synthesis of
pyrazoles, adapted for isoxazoles. All compounds were tested in four different cell