554 Letters in Drug Design & Discovery, 2009, 6, 554-562
1570-1808/09 $55.00+.00 © 2009 Bentham Science Publishers Ltd.
QSAR Study of Androstenedione Analogs as Aromatase Inhibitors
Aline A. Oliveira, Teodorico C. Ramalho and Elaine F.F. da Cunha*
Departamento de Química. Universidade Federal de Lavras, Campus Universitário, 3037, CEP, 37200-000, Lavras,
MG, Brazil
Received April 23, 2009: Revised August 06, 2009: Accepted August 06, 2009
Abstract: The aromatase enzyme is responsible for conversion of androgens to phenolic estrogens. The five-dimensional
quantitative structure-activity relationships (5D-QSAR) of a series of androstenedione analogs developed as aromatase in-
hibitors were studied using the Raptor program. The best model (N=47, q
2
=0.660, R
2
=0.719) showed contributions of the
hydrophobic, hydrogen-bond-donating and hydrogen-bond-accepting fields to the activity. The model was also externally
validated using 12 compounds (test set) not included in the model generation process. The statistical parameters from the
model indicate that the data are well fitted and have good predictive ability. Thus it was possible to generate and to vali-
date aromatase receptor surrogates through the prediction of relative free energies of aromatase inhibitors binding in re-
ceptor-modeling studies.
Keywords: 5D-QSAR, Aromatase, Androstenedione, Medicinal chemistry, Cancer.
INTRODUCTION
Cancer is responsible for one in eight deaths in the world
[1]. It is estimated that by 2020, fifteen million new cases of
the disease will be registered per year, lung and prostate tu-
mors being more common in men and breast and cervix uteri
tumors in women [2]. Breast cancer is the predominant type
of cancer among women [1,3]. The growth of these tumors is
stimulated by estrogen in approximately one third of ad-
vanced breast cancers [4]. The estrogen level decreases sig-
nificantly after menopause. In breast cancer tissues, signifi-
cant differences in levels of estrogen at pre- and post-
menopause are not observed [5]. This fact indicates that in-
tratumoral production of estrogen is highly responsible for
the high level of estrogen at post-menopause [6,7]. Thus the
inhibition of estrogenic stimulation becomes the main target
in endocrine treatment of breast cancer and may involve
blocking estrogen receptors or inhibiting the production of
estrogen [8].
One way to inhibit the production of estrogen involves
the cytochrome P-450. Aromatase is a cytochrome P-450
enzyme that catalyzes the synthesis of estrone and estradiol
(estrogens) from testosterone and androstenedione (andro-
gens) [8-10]. The chemical structure of aromatase inhibitors
(AIs) allows to classify them as steroid (type I), such as ex-
emestane, or non-steroidal (type II), such as anastrozole and
letrozole [11,12]. But there are different mechanisms by
which the two classes of AIs prevent the biosynthesis of es-
trogens. Steroidal inhibitors, analogs to androstenedione,
irreversibly bind to the enzyme while the non-steroidal bind
reversibly, both competitively inhibiting the enzyme [12].
Several laboratories describe efficient aromatase inhibi-
tors, analogs to the natural substrate androstenedione, since
these inhibitors are of great interest in the treatment of breast
cancer associated with the production of estrogens [13,14]. It
is very important for the AIs to inhibit more than 90% of the
*Address correspondence to this author at the Departamento de Química.
Universidade Federal de Lavras, Campus Universitário, 3037, CEP: 37200-
000, Lavras, MG, Brazil; E-mail: elaine_cunha@ufla.br
estrogen production [15]. But the use of AIs is restricted to
women with ovarian function stopped due to menopause or
surgery for removal of ovaries [16,17].
Based on the androstenedione structure, Numazawa and
coworkers [13,18-22] synthesized and evaluated a series of
4-ene, 5-ene, 1,4-diene, 4,6-diene and 1,4,6-triene deriva-
tives as competitive inhibitors of aromatase activity in hu-
man placental microsomes. In order to develop quantitative
structure–activity relationship (QSAR) models for ligands of
the aromatase receptor, we have selected this series of an-
drostenedione analogs as a case study.
QSAR is a mathematical methodology, statistically vali-
dated, and mostly used to correlate experimental or calcu-
lated properties derived from chemical structures with bio-
logical activities [23,24]. QSAR also may be applied to pre-
dict the activity values of non-synthesized compounds struc-
turally related to training sets. With the advent of molecular
modeling, it is possible to represent each molecule by an
ensemble of conformations, orientations, protonation states
(4D-QSAR) and to consider an ensemble of different in-
duced-fit models (5D-QSAR). Both ensembles are available
throughout the entire simulation, and genetic algorithms are
used for selecting the most predictive combination [24]. The
compounds used as aromatase inhibitors can be analyzed
through their structure and biological activity generating
aromatase binding-site models that contribute to the devel-
opment of new drugs against cancer.
METHODOLOGY
Biological Data and Compound Model Building
A data set of 59 steroidal aromatase inhibitors was taken
from published results [13,18-22] and the QSAR model was
developed using a training data set of 47 compounds (2–5, 7-
9, 11, 12, 14–17, 19-21, 23-27, 29-37, 40-45, 47–51, 53, 54
and 56–59), selected from the original 59 compounds. The
model was also externally validated using a test data set of
12 compounds (1, 6, 10, 13, 18, 22, 28, 38, 39, 46, 52 and
55), selected from the original 59 compounds. The 59