Molecular Docking and 3D-QSAR CoMFA Studies on Indole Inhibitors of GIIA
Secreted Phospholipase A2
Varnavas D. Mouchlis, Thomas M. Mavromoustakos,* and George Kokotos
Laboratory of Organic Chemistry, Department of Chemistry, University of Athens, Panepistimiopolis,
Athens 15771, Greece
Received June 2, 2010
Automated docking allowing a “protein-based” alignment was performed on a set of indole inhibitors of the
GIIA secreted phospholipase A2(GIIA sPLA2). A correlation between the binding scores and the experimental
inhibitory activity was observed (r2) 0.666, N ) 34). All the indole inhibitors were docked in the active
site of the GIIA sPLA2enzyme, and the best score docking pose of each inhibitor was used for the “protein-
based” alignment of the compounds. A three-dimensional quantitative structure-activity relationship (3D-
QSAR) model was then established using the comparative molecular field analysis (CoMFA) method. The
set of 34 indole inhibitors was divided into two subsets: the training set, composed of 26 compounds, and
the test set, consisting of eight compounds. The robustness and the predictive ability of the generated CoMFA
model were examined by using the test set. A good correlation (r2) 0.997) between predicted and
experimental inhibitory activity data allows the validation of the CoMFA model. Finally, the generated
CoMFA model was used for the design and evaluation of new compounds. The new designed compounds
exert improved predicted inhibitory activity and may be a target for the synthesis of new GIIA sPLA2
Phospholipase A2corresponds to a superfamily of enzymes
that to date include 15 separate, identifiable groups and
numerous subgroups of PLA2enzymes.1,2These enzymes
catalyze the hydrolysis of the ester bond of membrane
phospholipids at the sn-2 position. The five main categories
of PLA2enzymes are: the secreted sPLA2s, the cytosolic
cPLA2s, the Ca2+-independent iPLA2s, the PAF acetylhy-
drolases, and the lysosomal PLA2s.
The products of the sn-2 ester bond hydrolysis of phos-
pholipids by the PLA2 enzymes are free fatty acids and
lysophospholipids. The action of the PLA2enzymes on the
phospholipids is of high importance when the esterified fatty
acid at the sn-2 position is arachidonic acid (AA). AA is
converted by different downstream metabolic enzymes (such
as COX-1, COX-2 and 5-LO) to several bioactive lipid
mediators called eicosanoids, including prostaglandins (PGs)
and leukotrienes (LTs).3,4The eicosanoids participate in
many pathological inflammatory conditions, such as athero-
sclerosis5and ischemia diseases.6The lysophospholipids are
precursors of other bioactive mediators, such as the platelet
activating factor (PAF).7
PLA2enzymes are membrane-bound enzymes, and as a
result they have an i-face that has been proposed to make
contact with the substrate interface.8The i-face of the sPLA2
enzymes is a relatively flat surface of 1600 A2, which binds
tightly to the phospholipid bilayers (Kd< 10-13M).9There
are some polar and hydrophobic residues on the flat surface,
through which the sPLA2s bind to the ionic bilayers. The
sPLA2enzymes are interfacial enzymes, and their active site
is localized near the substrate binding i-face.10
There are 10 known members of sPLA2enzymes that have
been identified in mammals, which are numbered and
classified in groups according to the chronological order of
their discovery. The 10 groups of the sPLA2enzymes are:
IB, IIA, IIC-F, III, V, X, and XII.11The characterization
of their molecular structure, the classification, the genome
localization, and the details of their catalytic mechanism
attracted the research interest of many scientists.11-15In the
class of the low molecular weight secreted PLA2enzymes,
the group IIA secreted PLA2(GIIA sPLA2) is of paramount
importance since it is involved in several inflammatory
diseases, such as rheumatoid arthritis, and was cloned in
1989.16,17In vitro studies using recombinant GIIA sPLA2
on phospholipid substrates have provided important informa-
tion about the biochemistry of the enzyme. For instance, the
GIIA sPLA2(known as human nonpancreatic sPLA2) shows
biological activity on ionic phospholipids, such as phos-
phatidylglycerol (PG), phosphatidylserine (PS), and phos-
phatidylethanolamine (PE), but it is inactive on phosphati-
GIIA sPLA2 is a disulfide-linked enzyme, with seven
disulfide bonds, which contribute to the folding and stability
of the enzyme structure. In addition, it has a Ca2+-binding
loop and a His/Asp catalytic dyad. The mechanism of
substrate hydrolysis begins by the activation of a water
molecule by the catalytic histidine (His47). Beside this
histidine, there is an aspartate residue (Asp48), which
together with three other residues (Gly29, Gly31, and His27)
construct the conserved Ca2+-binding loop, where the Ca2+
ion is bound.19The hepta-coordinated Ca2+ion provides two
positions for the substrate binding, one axial and one
* Corresponding author. E-mail: email@example.com. Telephone: +30
J. Chem. Inf. Model. 2010, 50, 1589–1601
2010 American Chemical Society
Published on Web 08/26/2010
equatorial.20The high-resolution crystal structures of the
GIIA sPLA2enzyme have defined an enclosed active site
with a hydrophobic region which is located near the
N-terminal helix.21,22This hydrophobic region contributes
to the binding of a phospholipid molecule and to the
interfacial binding of the enzyme to the phospholipid
The GIIA sPLA2enzyme is an attractive target for the
development of new inhibitors, which might lead to thera-
peutic drugs for diseases where the enzyme is involved. It
has also been crystallized with or without different ligands,
and this fact renders the enzyme a suitable target for drug
design using computational methods.23-28Many different
classes of synthetic and natural inhibitors are known to
date.29-33However, in order to enhance the design of new
GIIA sPLA2inhibitors, the requirements are: (i) the knowl-
edge of the exact location of the active site and the
understanding of the inhibitor-enzyme interactions; and (ii)
the establishment of drug design computational protocols to
predict the activity of new designed molecules.
The main computational methods used in the rational drug
design can be divided into two groups: (i) protein-based
studies, which include molecular docking studies23,24and
molecular dynamics simulations25,26(MD) (wherein the
receptor-ligand interactions model is simulated); and (ii)
quantitative structure-activity relationship (QSAR) ana-
lysis,27,28which does not require a priori hypothesis about
the receptor structure. The QSAR analysis is based on the
pharmacophore alignment of the ligands and implies a
common biochemical mechanism.
Molecular docking is an extensively used computational
method in rational drug design,34and the main principals
that govern this technique have been described in recent
review articles.35-37For any docked pose, the binding score
is calculated generally as the sum of the electrostatic, van
der Waals and hydrophobic interactions, and hydrogen
bonding. Some scoring functions include also metal-binding
and solvation terms. Methodologies have been developed to
pick up the pose that simulates best the biological molecular
The 3D-QSAR comparative molecular fields analysis
(CoMFA)39requires the alignment of all the studied mol-
ecules in a three-dimensional space. In conventional CoMFA
studies, the molecules are fitted to a reference molecule,
which is the most rigid or constrained molecular structure
among the most active compounds. The hypothesis in this
case is that the conformation of the reference compound is
supposed to correspond to the “biologically active” confor-
mation. In the case of a known X-ray crystal structure with
high resolution, a structure-based design protocol reduces
the uncertainty about the determination of the “bioactive
conformation” of the reference compound. A protocol of
automated molecular docking of all the available compounds
in the receptor active site creates an indubitable “receptor-
based” alignment for the CoMFA analysis.
The present study is consisted of molecular docking
calculations and the generation of a CoMFA model on a set
of GIIA sPLA2indole inhibitors reported in the literature.40
All the GIIA sPLA2indole inhibitors were docked in the
enzyme active site using GLIDE 5.5.41-43The extra-
precision44(XP) mode of GLIDE was used for the docking
calculations. The best score docking pose of each indole
inhibitor was used in a “protein-based” alignment for the
CoMFA procedure. The quality of the CoMFA analysis
depends greatly on the alignment of the studied compounds.
“Protein-based” alignment via molecular docking allows the
development of a high-quality CoMFA model.
2. COMPUTATIONAL METHODS
2.1. Preparation of the GIIA sPLA2Enzyme File. Four
crystal structures of the GIIA sPLA2 enzyme which are
deposited in the RCSB protein data bank were downloaded
(PDB IDs: 1DB4 holo form 2.20 Å X-ray resolution,451DB5
holo form 2.80 Å X-ray resolution,451KVO holo form 2.00
Å X-ray resolution,46and 1J1A holo form 2.20 Å X-ray
resolution47). The objective was to judge which one is
sufficient for docking the indole inhibitors. The procedure
for this determination is consisted by the following steps:
(i) using “superposition panel” of Maestro 9.048all the crystal
structures were superimposed based on all the backbone
atoms including beta carbons. The crystal structure with PDB
ID: 1DB4 was chosen as the reference structure for the
superposition (rmsd between: 1DB4-1DB5: 0.147 Å,
1DB4-1J1A: 0.746 Å, 1DB4-1KVO: 0.487 Å). No sig-
nificant structural differences were observed (see Figure 1
in Supporting Information); (ii) the active site region was
examined to determine if the superimposed ligands can fit
into the reference site without steric clashes. No significant
steric clashes were observed; (iii) the active site region of
all the crystal structures, in turn, was examined in order to
determine whether any residues in the superimposed protein
differ appreciably in position or conformation from those in
the reference site. No significant differences were observed.
Thus, the 1DB4.pdb file has been chosen for the molecular
docking calculations. This file contains a single unit of the
GIIA sPLA2 enzyme cocrystallized with a native indole
inhibitor, which is structurally similar with the indole
inhibitors used in this study. The 1DB4.pdb crystal structure
was prepared using the “Protein Preparation Wizard” panel49
of Schro ¨dinger 2009 molecular modeling package. In par-
ticular, using the “preprocess and analyze structure” tool,
the bond orders were assigned, all the hydrogen atoms were
added, the calcium ion was treated in order to have the
correct geometry and formal charge (+2), the disulfide bonds
were assigned, and all the water molecules in a distance
greater than 5 Å from any heterogroup were deleted. Using
Epik 2.0,50,51a prediction of the heterogroups ionization and
tautomeric states was performed. An optimization of the
hydrogen-bonding network was performed using the “H-bond
assignment” tool. Finally, using the “impref utility”, the
positions of the hydrogen atoms were optimized by keeping
all the heavy atoms in place.
2.2. Preparation of Ligands Files. All the indole ligands
were built and adjusted using the Maestro 9.0 molecular
builder. All the hydrogen atoms were added, and the ligands
were submitted in full structure optimization, using the
minimization procedure of MacroModel 9.7.52For the
minimization, a standard molecular mechanics energy func-
tion (OPLS_200553force field) and the Polak-Ribiere
conjugated gradient method (5000 iterations with gradient
0.01 kJ/mol·Å)54were used. Solvent effects were modeled
with the generalized Born/surface area (GB/SA) implicit
solvent model55using water as solvent and normal non-
J. Chem. Inf. Model., Vol. 50, No. 9, 2010
MOUCHLIS ET AL.
bonded cutoffs. The constant dielectric electrostatic treatment
was selected with a dielectric constant of 1. The minimized
structures of the indole ligands were subsequently prepared
using LigPrep 2.3.56In particular, using LigPrep 2.3 all the
structures were checked to be correct, ionization states were
generated for all the indole inhibitors by taking into account
the metal mode of Epik 2.0 ionizer for the metalloproteins,
and the OPLS_2005 force field was used for the structure
optimization of all the indole ligands.
2.3. Receptor Grid Generation for the GIIA sPLA2
Enzyme. Glide is a grid-based ligand docking with energetics
approach and searches for favorable interactions between
ligands and receptors. The shape and properties of the
receptor are represented on a grid by different sets of fields
that provide progressively more accurate scoring of the ligand
pose (by “pose” we mean a complete specification of the
ligand: position and orientation relative to the receptor, core
conformation, and rotamer-group conformations). These
fields are generated as preprocessing steps in the calculation
and hence need to be computed only once for each receptor.
For the grid generation of the GIIA sPLA2 enzyme, the
binding site was defined using the native indole ligand
cocrystallized with the enzyme. The default value (1.00) for
the van der Waals radii scaling factor was chosen, which
means no scaling for the nonpolar atoms was performed (no
flexibility was simulated for the receptor). Glide uses two
boxes to organize the calculation: (i) the grids are calculated
within the space defined by the enclosed or outer box. This
is also the box in which all the ligand atoms must be
contained; (ii) acceptable positions for the ligand center
during the site point search lie within the ligand diameter
midpoint box or the inner box. This box gives a more precise
measure of the effective size of the search space. However,
ligands can move outside this box during grid minimization.
The native indole inhibitor has a very similar size with the
indole ligands of this study, and for defining the size of the
inner box, the option “docking ligands similar in size to
the workspace ligand” was chosen. The Cartesian coordinates
of the inner box, X, Y, and Z length were set to 12 Å.
2.4. Protocol of Grid-Based Ligand Docking with
Energetics (Glide). Glide uses a series of hierarchical filters
to search for possible locations of the ligand in the active
site region of the receptor. For the grid-based ligand docking
with energetics, the receptor grid generation file was used
(see Section 2.3). The next step is to define the conforma-
tional space of the ligand. Glide uses a sophisticated fashion
described by Friesner and co-workers to choose six lowest-
energy poses.42Finally, the three to six lowest-energy poses
obtained in this fashion are subjected to a Monte Carlo
procedure that examines nearby torsional minima. This
procedure is needed in some cases to properly orient
peripheral groups and occasionally alters internal torsion
angles. To soften the potential for nonpolar parts of the
ligands the “scaling of van der Waals radii” was used with
“scaling factor” of 0.8 and “partial charge cutoff” of 0.15.
Glide uses the GlideScore in order to rank poses. GlideScore
is a modified and expanded version of ChemScore57scoring
function that is used for predicting binding affinity and rank-
ordering ligands in database screens. By using GlideScore
as the scoring function, a composite Emodel score is then
used to rank the poses of each ligand and to select the poses
to be reported to the user. Emodel combines GlideScore, the
nonbonded interaction energy, and for flexible docking, the
excess internal energy of the generated ligand conformation.
In the present study the extra-precision44(XP) mode of
GlideScore scoring function was used. XP mode can be used
when the active site of the receptor contains a metal and
often works well. Glide assigns a special stability to ligands
in which anions coordinate to the metal center. To benefit
from this assignment, groups such as carboxylates, hydrox-
amates, and thiolates must be anionic. The protein residues
that line the approach to the metal center need to be
protonated in a manner compatible with the coordination of
an anionic ligand, such as a carboxylate or hydroxamates.
With respect to the GIIA sPLA2enzyme, the calcium ion
has charge of +2, but the effective charge which GLIDE
uses to dock the ligands is +1 because the ionic carboxylate
group of Asp48, that bears a charge -1, binds to the calcium
ion. The GlideScore XP scoring function also includes terms
which assign penalties to structures where statistical results
indicate that one or more groups is inadequately solvated.
A large database of cocrystallized structures has been used
by the creators of the scoring function, to optimize the
parameters associated with the penalty terms. No penalties
on the docked structures were observed. The force field used
for the docking was the OPLS_2001.
2.5. CoMFA. 2.5.1. Data Set (Training and Test Sets).
The 34 indole inhibitors and the corresponding biological
data used in this study have been selected from the
literature.40The molecular structures and the GIIA sPLA2
inhibitory activity data for these 34 indole inhibitors are
summarized in Table 1. All the collected biological data (%
inhibition) were measured in vitro under the same experi-
These 34 compounds were divided into two subsets: 26
of them were used as a training set, and 8 were used as a
test set. The compounds were separated to training and test
sets in order for the two sets to have a good molecular
diversity. Both, training and test sets included active and
inactive compounds to a scale from 0 to 100% inhibition
against the GIIA sPLA2enzyme.
2.5.2. Data Set Alignment. In the 3D-QSAR CoMFA
analysis, biological conformation and alignment rule selection
are two important factors to construct a reliable model. In
the present study, the poses from the automated molecular
docking calculations by GLIDE were used for the alignment
method. Compound 1 was selected as a template molecule
for the alignment. Considering the structural similarities of
the studied indole inhibitors, 17 common atoms of com-
pounds 1-34 were selected for the alignment. The alignment
of the molecules was performed using the module “database
alignment” in SYBYL 8.0 molecular modeling package. The
alignment results are shown in Figure 2 in the Supporting
Information with the common atoms highlighted in magenta
color in compound 1.
2.5.3. CoMFA Analysis. In the present study, a “protein-
based” alignment was used (the conformations of the studied
compounds used for the alignment were chosen by docking
these compounds in the GIIA sPLA2enzyme active site).
Atomic charges for the aligned molecules were calculated
using the Gasteiger-Hu ¨ckel method, which is a combination
of two other charge computational methods: the Gasteiger-
Marsili58method to calculate the σ component of the atomic
charge and the Hu ¨ckel59method to calculate the π compo-
COMFA STUDIES ON INDOLE INHIBITORS
J. Chem. Inf. Model., Vol. 50, No. 9, 2010 1591
nent of the atomic charge. The total charge is the sum of
the charges calculated by the two methods. The CoMFA
fields are generated by creating a grid around the molecule
and calculating the steric and electrostatic potentials at each
point on the grid using a charged probe atom. The CoMFA
calculations were performed using Tripos Advance CoMFA60
module in SYBYL 8.0. The steric and electrostatic field
energies were calculated using the Lennard-Jones and
coulomb potentials, respectively with 1/r2distance-dependent
dielectric constant in all intersections of regularly spaced (0.2
nm) grid. The sp3carbon atom with radius of 1.53 Å and
charge +1.0 was used as a probe to calculate the steric and
electrostatic energies between the probe and the molecules
using the standard Tripos force field.61The truncation for
both the steric and electrostatic energies was set to 30 kcal
mol-1. This indicates that any steric or electrostatic field
value that exceeds this value will be replaced with 30 kcal
mol-1, thus makes a plateau of the fields close to the center
of any atom.
Validations. The initial PLS analysis is used to correlate the
experimental inhibitory activity of the indole inhibitors
against the GIIA sPLA2enzyme with the CoMFA values
containing magnitude of steric and electrostatic potentials.
CoMFA standard scaling was applied to all the CoMFA
analysis. The full PLS analysis was run with a column
filtering of 2.0 kcal mol-1to reduce the noise and to speed
up the calculation. In CoMFA analysis, descriptors were
treated as independent variables, whereas the % inhibition
against the GIIA sPLA2 enzyme values were treated as
dependent variables in the PLS regression analyses to derive
the 3D-QSAR model. The model was assessed by their cross-
The final model (non-cross-validated conventional analysis)
was developed from the model with the highest cross-
sessed by the r2conventional correlation coefficient, the s
standard error of prediction, and the F value (Fisher test).
To obtain confidence limits and test the stability of
obtained PLS model, for the conventional CoMFA run,
bootstrapping64was performed (100 runs, column filtering:
2.00 kcal mol-1). The idea is to simulate a statistical sampling
procedure by assuming that the original data set is the true
population and generating many new data sets from it. These
new data sets (called bootstrap samplings) are of the same
size as the original data set and are obtained by randomly
choosing samples (rows) from the original data, with repeated
selection of the same row being allowed. The statistical
calculation is performed on each of these bootstrap sam-
plings, with new values being calculated for each of the
parameters to be estimated. The difference between the
parameters calculated from the original data set and
the average of the parameters calculated from the many
bootstrap samplings is a measure of the bias of the original
2) using leave-one-out (LOO) procedure.62,63
2). The non-cross-validated model was as-
3. RESULTS AND DISCUSSION
3.1. Molecular Docking Results. In this section, the
results derived from the docking of the indole inhibitors in
the GIIA sPLA2enzyme active site will be discussed. The
attention has been focused on the most characteristic
receptor-ligand interactions for a few key inhibitor structures
and especially for compound 1, the most active one in the
The established automated molecular docking was first
applied to two indole inhibitors already cocrystallized with
the GIIA sPLA2enzyme and described by Schevitz et al.45
These two compounds (see Table 1 in the Supporting
Information) are structurally very similar to the indole
inhibitors in this study. By docking these compounds, it was
possible to examine if GLIDE is able to reproduce crystal-
lographic experimental data for structural similar compounds.
Afterward, the 34 indole inhibitors in this study were docked
in the GIIA sPLA2enzyme active site, and the inhibitory
activity was compared with the XP binding scores calculated
by GLIDE. The purpose of this docking was to examine if
there is a correlation between the experimental % inhibitory
activity and the theoretical XP binding scores.
The crystallographic data reveal the following interactions
of indole845and indole645with the GIIA sPLA2enzyme
Table 1. In Vitro % Inhibition of the Indole Inhibitors40Against
the GIIA sPLA2Enzyme and the XP Binding Scores Calculated by
J. Chem. Inf. Model., Vol. 50, No. 9, 2010
MOUCHLIS ET AL.
active site: (i) two interactions between two oxygen atoms
of the ligands and the calcium ion and (ii) one hydrogen
bond with His47. In the case of indole8, one oxygen atom
of the phosphonate group participates in a hydrogen bond
with Lys62 through a water molecule placed near the enzyme
active site. The compounds indole8 and indole6 were docked
successfully by GLIDE. The interactions aforementioned
were reproduced, and the conformation of each ligand was
almost identical to the crystallographic one. In addition, three
other hydrogen bonds were observed between one hydrogen
atom of the amide group and Asp48, the oxygen atom of
the amide group and Gly29 and one of the oxygen atoms of
the carboxylate or phosphonate group with Gly31 (see
Figure 3 and Table 1 in the Supporting Information).
The docking calculations were then applied to the 34
indole inhibitors in this study. The computed XP binding
scores of the indole inhibitors from the docking calculations
with GLIDE were compared with their experimental inhibi-
tory activity against the GIIA sPLA2enzyme (Table 1). The
linearity of the plot (r2) 0.666, N ) 34) presents a good
correlation between the XP binding scores and the inhibitory
activity of the indole inhibitors (Figure 1).
In Figure 2 the binding of the most active compound 1 is
presented and reveals the main interactions with the enzyme
active site. The observed interactions are: (i) two interactions
of two oxygen atoms (the oxygen atom of the amide group
and one of the oxygen atoms of the carboxylate group) with
the calcium ion (COO-···Ca2+2.30 and NHCO···Ca2+2.70
Å); (ii) one hydrogen bond of one of the hydrogen atoms of
the amide group with the nitrogen atom of His47 (H· · ·N,
1.90 and N· · ·N 2.90 Å); (iii) one hydrogen bond of one of
the hydrogen atoms of the amide group with the oxygen atom
of Asp48 (H···O 1.80 and O···O 2.70 Å); (iv) one hydrogen
bond of the oxygen atom of the amide group with the
hydrogen atom of the amine group of Gly29 (O.. .H-N 2.10
and O· · ·N 2.80 Å); and (v) one hydrogen bond of one of
the oxygen atoms of the carboxylate group with the hydrogen
atom of the amine group of Lys62 through a water molecule
placed near the active site of the enzyme (O···H-O···H-N
2.50, 1.90 and O· · ·O· · ·N 3.00, 2.90 Å). The phenyl ring
of compound 1 participates in aromatic (π-π) stacking
interactions with the residues Phe5 (Rcen ) 6.49 Å, θ )
9.40°) and His6 (Rcen ) 5.76 Å, θ ) 70.77°), and the indole
ring participates also in aromatic (π-π) stacking interactions
with Phe5 (Rcen ) 5.88 Å, θ ) 84.86°) and His47 (Rcen )
6.64 Å, θ ) 70.20°). Aromatic (π-π) stacking interactions
are one of the forces governing molecular recognition. Burley
and Petsko have reported that aromatic (π-π) stacking
interactions in proteins operate at distances (Rcen) of 4.5-7.0
Å between the center mass of the rings, the rings’ centroids.65
The angle (θ) between normal vectors of interacting aromatic
rings is typically between 30° and 90°, producing a “tilted-
T” or “edge-to-face” arrangement of interacting rings. Hunter
and co-workers66have reported that aromatic (π-π) parallel
stacking interactions (θ < 30°) between phenylalanine
residues in proteins are also favorable, if the rings are offset
from each other. The two chlorine atoms participate in van
der Waals contacts with the carbon atoms of the site chains
of Phe5, Ile9, Ala17, Ala18, and Tyr21.
In the 34 indole inhibitors described in this study, the main
structural changes are in the R group. This position is
governed mainly by steric effects in the active site of the
GIIA sPLA2enzyme. In Figure 3 the binding of the inactive
compound 25 is presented. The main interactions of com-
pound 25 in the active site of the enzyme are: (i) two
interactions of the two oxygen atoms (the oxygen atom of
the amide group and one of the oxygen atoms of the
carboxylate group) with the calcium ion (COO-···Ca2+2.30
and NHCO· · ·Ca2+2.60 Å); (ii) one hydrogen bond of one
of the hydrogen atoms of the amide group with the nitrogen
Figure 1. Plot of the experimental % inhibition against the GIIA sPLA2enzyme versus the XP binding scores calculated by GLIDE. The
linear regression statistics for the data set are also presented.
COMFA STUDIES ON INDOLE INHIBITORS
J. Chem. Inf. Model., Vol. 50, No. 9, 2010 1593
atom of His47 (H· · ·N 2.00 and N· · ·N 3.00 Å); (iii) one
hydrogen bond of one of the hydrogen atoms of the amide
group with the oxygen atom of Asp48 (H· · ·O 1.70 and
O· · ·O 2.70 Å); and (iv) one hydrogen bond of the oxygen
atom of the amide group with the hydrogen atom of the
amine group of Gly29 (O· · ·H-N 2.20 and O· · ·N 3.00 Å).
Our attention has been focused on understanding the ste-
reoelectronic factors that affect the inhibitory activity of the
indole inhibitors. As presented in Figure 3, the methoxy
group at the two position of the phenyl ring is bulky enough
and cannot be accommodated inside the hydrophobic cavity
of the active site. A comparison with the binding of
compound 1 reveals that a substituent with the same size as
chlorine atom is better for this position. On the other hand,
Figure 2. The binding of compound 1 in the active site of the GIIA sPLA2enzyme. The enzyme-ligand complex was obtained by automated
molecular docking of the indole inhibitor in the enzyme active site using GLIDE.
Figure 3. The binding of compound 25 in the active site of the GIIA sPLA2enzyme. The enzyme-ligand complex was obtained by
automated molecular docking of the indole inhibitor in the enzyme active site using GLIDE.
J. Chem. Inf. Model., Vol. 50, No. 9, 2010
MOUCHLIS ET AL.
the nitro group participates in a hydrogen bond with Val30
(O···H 2.20 and O···N 3.20 Å). It seems that this hydrogen
bond contributes to lock the phenyl ring in the position
presented in Figure 3. As a result, the phenyl ring is not
accommodated inside the hydrophobic cavity of the enzyme
active site, and it does not participate in aromatic (π-π)
stacking interactions or in aromatic/aliphatic interactions with
the residues of the hydrophobic region.
By comparing compounds 6 and 28 with compound 1, it
can be concluded that the replacement of the chlorine atom
by the bulkier trifluoromethyl group at the two position of
the phenyl ring does not affect significantly the inhibitory
activity. On the other hand, compound 19 which possesses
a nitro group at the three position of the phenyl ring shows
almost the half inhibitory activity compared to compound
1. Another remark can be made for compounds 2-5 and 29
which possess at the three position of the phenyl ring a
halogen or a bulky group. These substituents do not decrease
significantly the activity compared to compound 1, as the
nitro group does. It seems that the polar nitro group is not
an appropriate substituent on the phenyl ring of the indole
inhibitors, since it is placed in the hydrophobic region of
the enzyme active site. The combination of the nitro group
with the bulky methoxy group resulted to the inactivity of
Another remarkable point emerges by comparing com-
pounds 10, 22, and 32 with compound 1. When the
substituents on the phenyl ring are very polar groups, the
inhibitory activity decreases. In the case of compounds 10
and 22, the cyano group is a polar substituent and affects
negatively the inhibitory activity. Among the two com-
pounds, 22 is less active because the four position of the
phenyl ring is directed to the hydrophobic region, but the
three position is directed toward the exterior of the enzyme
active site (Figure 2). The five fluorine atoms in the case of
compound 32 give to the phenyl ring extra polarization,
which is not appropriate for the hydrophobic region of the
GIIA sPLA2enzyme active site.
The comparison of compound 27 with compound 1
indicates that the naphthalene ring may be an appropriate
replacement for the phenyl ring. However, the low inhibitory
activity of compound 23 reveals that the naphthalene group
is not able to accept substituents, because the presence of
only one bulky group in the eight position of the naphthalene
group decreases the inhibitory activity dramatically. The low
inhibitory activity presented by compounds 24, 26, and 34
indicates that coumarinyl and phthalimido groups are not
suitable replacements for the phenyl ring.
The replacement of the phenyl ring by aliphatic chains
(compounds 8, 15, 17, 20, 21, 30, 31, and 33) decreases the
inhibitory activity in comparison with compound 1. The
phenyl ring is able to participate in aromatic (π-π) stacking
interactions, but the aliphatic chain does not. It seems that
the phenyl ring itself is essential for the binding. This is
validated by the fact that compound 13, which has only the
phenyl ring, has higher inhibitory activity in comparison to
compounds having aliphatic chains and possesses almost the
same inhibitory activity with compound 8 with the longest
aliphatic chain. On the other hand, the increment of the
inhibitory activity of compounds 9 and 12 relative to
compound 13 shows that a bulky substituent at the four
position of the phenyl ring affects positively the inhibitory
activity. This may be due to the fact that the methyl group
can participate in van der Waals contacts with residues, such
as Leu2, Val3, and His6, lining the hydrophobic region of
Table 2. Statistics and Cross-Validation Results of the CoMFA
Model Generated from the Protein-Based Alignmenta
0.998 ( 0.001
1.321 ( 0.982
aPLS components: the optimal number of principal components
in the PLS model; rcv
after LOO procedure on the training set of compounds; rtraining set
correlation coefficient between predicted and experimental values
for the training set of compounds; SEE: the standard error of
estimate; F: the value of Fisher test; electrostatic-steric: the
contribution of the electrostatic and steric fields in the established
PLS model; rbootstrapping
: the average of correlation coefficient for 100
samplings using bootstrapping procedure; SEEbootstrapping: the average
standard error of estimate for 100 samplings using bootstrapping
procedure; and rtest set
: the correlation coefficient between predicted
and experimental values for the test set of compounds.
2: the cross-validated correlation coefficient
Table 3. Summary of the Experimental % Inhibition Against the
GIIA sPLA2Enzyme and the CoMFA-Predicted % Inhibition for
the Training and Test Sets of Indole Inhibitors
COMFA STUDIES ON INDOLE INHIBITORS
J. Chem. Inf. Model., Vol. 50, No. 9, 2010 1595
the enzyme active site (Figure 2). The replacement of the
methyl group by the fluorine atom decreases the inhibitor
activity (compound 14) due to the fact that the van der Waals
radii of the fluorine atom is smaller than that of the methyl
group, and it is not able to participate in very strong van der
Waals contacts like the second one. This is also validated
by the fact that the replacement of the two chlorine atoms
of compound 1 with two fluorine atoms (compounds 7, 11,
and 16) decreases the inhibitory activity.
The R group of the indole inhibitors corresponds to a key
position for governing the inhibitory activity and especially
the steric effects. The substituents at the benzyl group are
important for the inhibitory activity. The question that arises
is which combination of these substituents improves the
3.2. 3D-QSAR CoMFA Model. The present CoMFA
model was generated using a “protein-based” alignment of
the 34 indole inhibitors (see Figure 2 in the Supporting
Information). The geometries of the inhibitors were deter-
mined by their interactions with the GIIA sPLA2enzyme
active site. For each of the 34 indole inhibitors GLIDE
yielded a single conformation, among the several possible
Figure 4. Plot of the experimental % inhibition against the GIIA sPLA2enzyme versus the CoMFA-predicted one for the training set of
Figure 5. The CoMFA steric and electrostatic fields SD × coeff contour maps for compound 1. Bulky groups in the green region favor the
inhibitory activity, but bulky groups in the yellow region are not favorable for the inhibitory activity. Negative potentials in the red region
favor the inhibitory activity, but negative potentials in the blue region are not desirable.
J. Chem. Inf. Model., Vol. 50, No. 9, 2010
MOUCHLIS ET AL.
low-energy ones derived in the theoretical calculations, for
the ligand-enzyme complex.
The predictive ability of the CoMFA model depends on
the alignment of the studied compounds. The “protein-based”
alignment suggests that the indole inhibitors in the real
biological system are aligned in accordance with the
ligand-enzyme interactions obtained by the molecular
docking using GLIDE. A derived successful CoMFA model
suggests the validity of the proposed model of the ligand-
enzyme interactions by GLIDE.
The separation of the data set into a training and test set
was performed so that the two sets contain a significant
diversity of inhibitory activity. Both sets included active and
inactive compounds in a range of 0-100% inhibition against
the GIIA sPLA2enzyme. For better understanding of the
stereoelectronic factors underlying the activity, a CoMFA
model was generated taking into account both steric and
electrostatic fields. The results of this analysis are presented
in Table 2, which underlines the contribution of the steric
and electrostatic fields in the CoMFA model. The analysis
shows that the relative contributions are 14.7 and 85.3% for
the electrostatic and steric fields, respectively. This analysis
underlines the important role of CoMFA steric field for the
R group of indole inhibitors compared to the electrostatic
one. The CoMFA model, in the present study, was used to
predict the inhibitory activity for new compounds which
potentially can be new GIIA sPLA2inhibitors.
A non-cross-validated PLS analysis was performed using
the five principal components, which have given the higher
cross-validated correlation coefficient after the LOO proce-
dure on the training set of the compounds in order to generate
the CoMFA contour maps. This value was selected as the
addition of any new component adds less than 5% to the
cross-validated correlation coefficient. Table 3 summarizes
the experimental and predicted % inhibition and the differ-
ence between them. Figure 4 shows the correlation between
the experimental and CoMFA-predicted % inhibition values
of the non-cross-validated analysis for the training set.
Figure 6. (a) The CoMFA contour maps within the active site of the GIIA sPLA2enzyme for compound 1 (MOLCAD lipophilic potential
surface was calculated for the receptor with the Connolly method; brown color denotes the most lipophilic areas, and blue color denotes
the most hydrophilic areas); and (b) the match of the contour maps with the residues of the GIIA sPLA2enzyme active site.
Figure 7. The CoMFA steric and electrostatic fields SD × coeff contour maps for compound 25. Bulky groups in the green region favor
the inhibitory activity, but bulky groups in the yellow region are not favorable for the inhibitory activity. Negative potentials in the red
region favor the inhibitory activity, but negative potentials in the blue region are not desirable.
COMFA STUDIES ON INDOLE INHIBITORS
J. Chem. Inf. Model., Vol. 50, No. 9, 2010 1597
The CoMFA contour maps outline a statistic field express-
ing the relationship between the variation of the steric and
electrostatic fields and the variation of the inhibitory activity
against the GIIA sPLA2enzyme. The values of the fields
are calculated at each lattice intersection and are equal to
the product descriptor coefficient multiplied by the corre-
sponding standard deviation (SD × coeff). Hence, a low
value of SD × coeff indicates that the presence of the
corresponding steric or electrostatic field at this point
decreases the activity, whereas a high value means that the
presence of fragments producing such a field favors the
activity. The SD × coeff contour maps for the electrostatic
and steric fields are presented in Figure 5 for the more active
compound 1. The contour maps of compound 1 within the
MOLCAD surface of the active site are represented in
Figure 6a. The MOLCAD surface was developed to display
the lipophilic potential. Figure 6b represents the match of
the contour maps with the residues of the GIIA sPLA2
Compounds 1 (Figure 5) and 25 (Figure 7) illustrate the
main features of the CoMFA contour maps. The colored
regions on the contour maps mark the favorable and
unfavorable characteristics that should have the substituents
on the phenyl ring. In Figure 5, the hydrogen atom at the
four position of the phenyl ring is extended near the green
region, which means that bulky groups are desirable and
affect the inhibitory activity in a favorable way. The green
contour around the hydrogen atom at the four position of
the phenyl ring approaches the lipophilic region (Figure 6a)
near Leu2, Val3, and His6 (Figure 6b), suggesting that a
bulkier group will increase the inhibitory activity, an
observation that is in agreement with the CoMFA model.
Near the green region, there is a blue region which indicates
that negative potentials are not desirable and affect the
inhibitory activity in an unfavorable way. The position two
of the phenyl ring is extended near a yellow region indicating
that the bulky groups are not desirable and affect also the
inhibitory activity in an unfavorable way. The yellow contour
is on the lipophilic region of the active site (Figure 6a),
suggesting that bulkier groups at this position will clash with
the residues, such as Ala18 and Tyr21 (Figure 6b). Near the
yellow region there is a red region which shows that negative
potentials are desirable for the inhibitory activity. The
chlorine atom seems to have the suitable properties for this
region because it is not as bulky to approach the yellow
contour region, but it has enough size to approach the red
contour region. By examining the position of compound 25
in the contour maps, it is possible to understand why this
inhibitor is inactive. The positively charged nitrogen atom
of the nitro group at the five position of the phenyl ring is
in vicinity with the red contour region, and the methoxy
group at the two position of the phenyl ring is in spatial
Figure 8. Plot of the experimental % inhibition against the GIIA sPLA2enzyme versus the CoMFA-predicted one for the test set of indole
Table 4. Structures Designed Based on the CoMFA Model and
Possess Higher Predicted Inhibitory Activity than the Most Active
J. Chem. Inf. Model., Vol. 50, No. 9, 2010
MOUCHLIS ET AL.
proximity to the yellow contour region. Thus, the two groups
are characterized by unfavorable interactions. This is in
accordance with the molecular docking calculations in which
compound 25 indicates low XP binding score.
3.3. Validation of the CoMFA Model. To further assess
the robustness and statistical confidence of the generated
CoMFA model bootstrapping analysis for 100 runs was
performed (Table 2). The higher value of rbootstrapping
) 0.998 ( 0.001) indicates the robustness and
statistical confidence of the generated CoMFA model.
The sensitivity of the CoMFA results toward the training
set selection has been also examined by repeating the analysis
on the training set with different ligand partitioning on the
ligand set. The rcv
values range between 0.767-0.792,
confirming the LOO results on the first partition and the
stability of the CoMFA model.
To further validate the stability and predictive ability of
the CoMFA model, eight compounds not included in the
training set were used as a test set. The predicted results for
the test set are listed in Table 3. Figure 8 indicates that the
predicted values of % inhibition by the CoMFA model are
in agreement with the experimental ones in a tolerable error
range with rtest set
The conclusion derived from the validation part of this
study is that the CoMFA model could be used reliably to
design new inhibitors with improved inhibitory activity
against the GIIA sPLA2 enzyme. The tight correlation
between experimentally observed and predicted values in this
study suggests that the present model is able to provide
reliable predictions of GIIA sPLA2inhibitory capacities in
a set of new inhibitors.
3.4. Design of New Indole Inhibitors. According to the
detailed contour analyses of the CoMFA model, useful
information on the structural requirements for the observed
inhibitory activities is obtained. We have employed this
information to design several analogues showing improved
inhibitory activity. The most potent molecule (compound 1)
was used as a reference structure to design new molecules.
Initially, the two chlorine atoms at the positions two and
six of the phenyl ring were replaced with the fluorine,
bromine, and iodine atoms. The predicted inhibitory activities
for these compounds were lower than the inhibitory activity
of compound 1. Chlorine atom is the best substituent for
these positions and especially for the position two, which is
in spatial vicinity with the red and yellow contours (Figure
5). Substituents at the five position of the phenyl ring do
not affect significantly the predicted inhibitory activity. The
three position approaches the yellow contour, and substituents
bulkier than the hydrogen atom may contribute to the
decrease of the inhibitory activity. The new analogues that
indicated higher predicted inhibitory activity than compound
1 are listed in Table 4. The changes on the new designed
compounds have been made at the four position of the phenyl
ring. The % inhibitory activity in the in vitro test was
measured using ligand concentration 0.33 µM.40The %
inhibitory activity higher than 100, calculated for the new
compounds, means that the new compounds might present
100% inhibitory activity against the GIIA sPLA2enzyme at
a lower concentration than 0.33 µM.
The new designed compounds have been subsequently
docked in the GIIA sPLA2enzyme active site in order to
see how the substituent at the four position of the phenyl
ring interacts with the enzyme active site. The XP binding
scores calculated for the compounds (Table 4) are higher
than the one of the template compound 1 (Table 1). The
binding of these compounds is very similar with the binding
of compound 1 and also with each other. The binding of
compound N1 is presented in Figure 9 (for the binding of
Figure 9. The binding of compound N1 in the active site of the GIIA sPLA2enzyme. The enzyme-ligand complex was obtained by
automated molecular docking of the compound in the enzyme active site using GLIDE.
COMFA STUDIES ON INDOLE INHIBITORS
J. Chem. Inf. Model., Vol. 50, No. 9, 2010 1599
compounds N2 and N3 see Figure 4 and Figure 5 in the
Supporting Information). The polar interactions shown in
compound N1 are identical to those shown in compound 1
and are the same for all the compounds in Table 4. The
phenyl ring of the new compounds orients in a similar way
in the active site as the phenyl ring of compound 1 and
participates in aromatic (π-π) stacking interactions with the
residues Phe5 and His6. The indole ring participates in
aromatic (π-π) stacking interactions with residues Phe5 and
His47. The substituent at the four position of the phenyl ring
of compound N1 participates in van der Waals contacts with
the residues Leu2, Val3, and His6 (Figure 9) and might be
the reason that the XP binding score is increased, in
comparison with the XP binding score of compound 1. The
substituent at the four position of the phenyl ring accom-
modated at the same hydrophobic region in all the new
designed compounds and interacts with the residues Leu2,
Val3, and His6.
A combination of automated docking calculations and
three-dimensional quantitative structure-activity relationship
studies using the comparative molecular field analysis method
(3D-QSAR CoMFA) was performed on a set of 34 GIIA
secreted phospholipase A2(GIIA sPLA2) indole inhibitors
for designing new compounds with improved inhibitory
activity. The docking of two crystallographic indole inhibitors
in the enzyme active site using GLIDE showed that the
algorithm can reproduce experimental crystallographic data,
and thus a reliable docking was performed for all the 34
indole inhibitors which have been studied using the extra-
precision (XP) mode.
The binding of the crystallographic compounds, after their
molecular docking with GLIDE, revealed three new hydrogen
bonds with the residues Gly29, Gly31, and His47. The XP
binding scores calculated by the molecular docking of the
34 indole inhibitors with GLIDE were compared with their
experimental inhibitory activity against the GIIA sPLA2
enzyme. The linearity of the plot (r2) 0.666, N ) 34)
presents a good correlation between the XP binding scores
and the experimental inhibitory activity of the indole
The best score docking pose for each indole inhibitor was
then used to generate the CoMFA model. The data set of
the 34 compounds was divided into two subsets, one with
26 compounds to construct the CoMFA model (training set)
and the other with 8 compounds for the validation of the
model (test set). The CoMFA model was created using a
“protein-based” alignment, and according to the cross-
validation test (rcv
The robustness and the statistic confidence of the generated
CoMFA model (rbootstrapping
predictive ability (rtest set
) 0.997) have shown that the model
can be used to design new compounds and further evaluate
how the structural changes affect the inhibitory activity.
This robust and predictive model was then used to design
new compounds presenting improved inhibitory activity. The
new compounds were subsequently docked in the GIIA
sPLA2enzyme active site to check how they interact with
the enzyme active site. The combination of the molecular
docking calculations and the three-dimensional QSAR
2) 0.793) has a good predictive capacity.
) 0.998 ( 0.001) and its
CoMFA studies gave important information about the
binding of the 34 indole inhibitors and the structural changes
that affect the inhibitory activity of these compounds against
the GIIA sPLA2enzyme. This model can be used to guide
the rational design of new compounds presenting improved
inhibitory activity against the GIIA sPLA2enzyme.
This work was supported in part by the University of
Supporting Information Available: Superimposition of
the four crystal structures of the GIIA sPLA2 enzyme,
alignment of the data set used in the 3D-QSAR CoMFA
model, and binding of the crystallographic inhibitors indole8
and indole6. This material is available free of charge via
the Internet at http://pubs.acs.org.
REFERENCES AND NOTES
(1) Burke, J. E.; Dennis, E. A. Phospholipase A2biochemistry. CardioVasc.
Drugs Ther. 2009, 23, 49–59.
(2) Schaloske, R. H.; Dennis, E. A. The phospholipase A2superfamily
and its group numbering system. Biochim. Biophys. Acta 2006, 1761,
(3) Funk, C. D. Prostaglandins and leukotrienes: advances in eicosanoid
biology. Science 2001, 294, 1871–5.
(4) Kudo, I.; Murakami, M. Phospholipase A2enzymes. Prostaglandins
Other Lipid Mediators 2002, 68-69, 3–58.
(5) Oestvang, J.; Johansen, B. PhospholipaseA2: a key regulator of
inflammatory signalling and a connector to fibrosis development in
atherosclerosis. Biochim. Biophys. Acta 2006, 1761, 1309–16.
(6) Back, M. Leukotriene signaling in atherosclerosis and ischemia.
CardioVasc. Drugs Ther. 2009, 23, 41–8.
(7) Moolenaar, W. H.; van Meeteren, L. A.; Giepmans, B. N. The ins
and outs of lysophosphatidic acid signaling. Bioessays 2004, 26, 870–
(8) Gelb, M. H.; Min, J. H.; Jain, M. K. Do membrane-bound enzymes
access their substrates from the membrane or aqueous phase: interfacial
versus non-interfacial enzymes. Biochim. Biophys. Acta 2000, 1488,
(9) Zhou, L.; Fang, C.; Wei, P.; Liu, S.; Liu, Y.; Lai, L. Chemically
induced dimerization of human nonpancreatic secretory phospholipase
A2by bis-indole derivatives. J. Med. Chem. 2008, 51, 3360–6.
(10) Gelb, M. H.; Jain, M. K.; Hanel, A. M.; Berg, O. G. Interfacial
enzymology of glycerolipid hydrolases: lessons from secreted phos-
pholipases A2. Annu. ReV. Biochem. 1995, 64, 653–88.
(11) Lambeau, G.; Gelb, M. H. Biochemistry and physiology of mammalian
secreted phospholipases A2. Annu. ReV. Biochem. 2008, 77, 495–520.
(12) Boyanovsky, B. B.; Webb, N. R. Biology of secretory phospholipase
A2. CardioVasc. Drugs Ther. 2009, 23, 61–72.
(13) Winget, J. M.; Pan, Y. H.; Bahnson, B. J. The interfacial binding
surface of phospholipase A2s. Biochim. Biophys. Acta 2006, 1761,
(14) Gelb, M. H.; Valentin, E.; Ghomashchi, F.; Lazdunski, M.; Lambeau,
G. Cloning and recombinant expression of a structurally novel human
secreted phospholipase A2. J. Biol. Chem. 2000, 275, 39823–6.
(15) Valentin, E.; Lambeau, G. Increasing molecular diversity of secreted
phospholipases A2and their receptors and binding proteins. Biochim.
Biophys. Acta 2000, 1488, 59–70.
(16) Kramer, R. M.; Hession, C.; Johansen, B.; Hayes, G.; McGray, P.;
Chow, E. P.; Tizard, R.; Pepinsky, R. B. Structure and properties of
a human non-pancreatic phospholipase A2. J. Biol. Chem. 1989, 264,
(17) Seilhamer, J. J.; Pruzanski, W.; Vadas, P.; Plant, S.; Miller, J. A.;
Kloss, J.; Johnson, L. K. Cloning and recombinant expression of
phospholipase A2present in rheumatoid arthritic synovial fluid. J. Biol.
Chem. 1989, 264, 5335–8.
(18) Valentin, E.; Ghomashchi, F.; Gelb, M. H.; Lazdunski, M.; Lambeau,
G. On the diversity of secreted phospholipases A2. Cloning, tissue
distribution, and functional expression of two novel mouse group II
enzymes. J. Biol. Chem. 1999, 274, 31195–202.
(19) Dennis, E. A. Diversity of group types, regulation, and function of
phospholipase A2. J. Biol. Chem. 1994, 269, 13057–60.
J. Chem. Inf. Model., Vol. 50, No. 9, 2010
MOUCHLIS ET AL.
(20) Scott, D. L.; White, S. P.; Otwinowski, Z.; Yuan, W.; Gelb, M. H.;
Sigler, P. B. Interfacial catalysis: the mechanism of phospholipase
A2. Science 1990, 250, 1541–6.
(21) Scott, D. L.; White, S. P.; Browning, J. L.; Rosa, J. J.; Gelb, M. H.;
Sigler, P. B. Structures of free and inhibited human secretory
phospholipase A2 from inflammatory exudate. Science 1991, 254,
(22) Wery, J. P.; Schevitz, R. W.; Clawson, D. K.; Bobbitt, J. L.; Dow,
E. R.; Gamboa, G.; Goodson, T., Jr.; Hermann, R. B.; Kramer, R. M.;
McClure, D. B.; Mihelich, E. D.; Putnam, J. E.; Sharp, J. D.; Stark,
D. H.; Teater, C.; Warrick, M. W.; Jones, N. D. Structure of
recombinant human rheumatoid arthritic synovial fluid phospholipase
A2at 2.2 A resolution. Nature 1991, 352, 79–82.
(23) Mouchlis, V. D.; Mavromoustakos, T. M.; Kokotos, G. Design of new
secreted phospholipase A2inhibitors based on docking calculations
by modifying the pharmacophore segments of the FPL67047XX
inhibitor. J. Comput.-Aided Mol. Des. 2010, 24, 107–15.
(24) Li, B.; Liu, Z.; Zhang, L. Multiple-docking and affinity fingerprint
methods for protein classification and inhibitors selection. J. Chem.
Inf. Model. 2009, 49, 1725–33.
(25) Tomoo, K.; Yamane, A.; Ishida, T.; Fujii, S.; Ikeda, K.; Iwama, S.;
Katsumura, S.; Sumiya, S.; Miyagawa, H.; Kitamura, K. X-ray crystal
structure determination and molecular dynamics simulation of pro-
phospholipase A2inhibited by amide-type substrate analogues. Bio-
chim. Biophys. Acta 1997, 1340, 178–86.
(26) Hariprasad, V.; Kulkarni, V. M. A molecular dynamics study of the
three-dimensional model of human synovial fluid phospholipase A2-
transition state mimic complexes. J. Mol. Recognit. 1996, 9, 95–102.
(27) Ortiz, A. R.; Pastor, M.; Palomer, A.; Cruciani, G.; Gago, F.; Wade,
R. C. Reliability of comparative molecular field analysis models:
effects of data scaling and variable selection using a set of human
synovial fluid phospholipase A2inhibitors. J. Med. Chem. 1997, 40,
(28) Ortiz, A. R.; Pisabarro, M. T.; Gago, F.; Wade, R. C. Prediction of
drug binding affinities by comparative binding energy analysis. J. Med.
Chem. 1995, 38, 2681–91.
(29) Magrioti, V.; Kokotos, G. Phospholipase A2inhibitors as potential
therapeutic agents for the treatment of inflammatory diseases. Expert
Opin. Ther. Pat. 2010, 20, 1–18.
(30) Garcia-Garcia, H. M.; Serruys, P. W. Phospholipase A2inhibitors.
Curr. Opin. Lipidol. 2009, 20, 327–332.
(31) Magrioti, V.; Kokotos, G. Synthetic Inhibitors of Group IVA and
Group VIA Phospholipase A2. Curr. Med. Chem.: Anti-Inflammatory
Anti-Allergy Agents 2006, 5, 189–203.
(32) Reid, R. C. Inhibitors of secretory phospholipase A2group IIA. Curr.
Med. Chem. 2005, 12, 3011–26.
(33) Meyer, M. C.; Rastogi, P.; Beckett, C. S.; McHowat, J. Phospholipase
A2inhibitors as potential anti-inflammatory agents. Curr. Pharm. Des.
2005, 11, 1301–12.
(34) Kuntz, I. D.; Meng, E. C.; Shoichet, B. K. Structure-Based Molecular
Design. Acc. Chem. Res. 2002, 27, 117–123.
(35) Tuccinardi, T. Docking-based virtual screening: recent developments.
Comb. Chem. High Throughput Screening 2009, 12, 303–14.
(36) Halperin, I.; Ma, B.; Wolfson, H.; Nussinov, R. Principles of docking:
An overview of search algorithms and a guide to scoring functions.
Proteins 2002, 47, 409–43.
(37) Taylor, R. D.; Jewsbury, P. J.; Essex, J. W. A review of protein-
small molecule docking methods. J. Comput.-Aided Mol. Des. 2002,
(38) Kauvar, L. M.; Higgins, D. L.; Villar, H. O.; Sportsman, J. R.;
Engqvist-Goldstein, A.; Bukar, R.; Bauer, K. E.; Dilley, H.; Rocke,
D. M. Predicting ligand binding to proteins by affinity fingerprinting.
Chem. Biol. 1995, 2, 107–18.
(39) Cramer III, R. D.; Patterson, D. E.; Bunce, J. D. Comparative molecular
field analysis (CoMFA). 1. Effect of shape on binding of steroids to
carrier proteins. J. Am. Chem. Soc. 1988, 110, 5959–5967.
(40) Smart, B. P.; Pan, Y. H.; Weeks, A. K.; Bollinger, J. G.; Bahnson,
B. J.; Gelb, M. H. Inhibition of the complete set of mammalian secreted
phospholipases A2 by indole analogues: a structure-guided study.
Bioorg. Med. Chem. 2004, 12, 1737–49.
(41) Glide, version 5.5; Schro ¨dinger, LLC: New York, 2009.
(42) Friesner, R. A.; Banks, J. L.; Murphy, R. B.; Halgren, T. A.; Klicic,
J. J.; Mainz, D. T.; Repasky, M. P.; Knoll, E. H.; Shelley, M.; Perry,
J. K.; Shaw, D. E.; Francis, P.; Shenkin, P. S. Glide: a new approach
for rapid, accurate docking and scoring. 1. Method and assessment of
docking accuracy. J. Med. Chem. 2004, 47, 1739–49.
(43) Halgren, T. A.; Murphy, R. B.; Friesner, R. A.; Beard, H. S.; Frye,
L. L.; Pollard, W. T.; Banks, J. L. Glide: a new approach for rapid,
accurate docking and scoring. 2. Enrichment factors in database
screening. J. Med. Chem. 2004, 47, 1750–9.
(44) Friesner, R. A.; Murphy, R. B.; Repasky, M. P.; Frye, L. L.;
Greenwood, J. R.; Halgren, T. A.; Sanschagrin, P. C.; Mainz, D. T.
Extra precision glide: docking and scoring incorporating a model of
hydrophobic enclosure for protein-ligand complexes. J. Med. Chem.
2006, 49, 6177–96.
(45) Schevitz, R. W.; Bach, N. J.; Carlson, D. G.; Chirgadze, N. Y.;
Clawson, D. K.; Dillard, R. D.; Draheim, S. E.; Hartley, L. W.; Jones,
N. D.; Mihelich, E. D.; Olkowski, J. L.; Snyder, D. W.; Sommers,
C.; Wery, J.-P. Structure-based design of the first potent and selective
inhibitor of human non-pancreatic secretory phospholipase A2. Nat.
Struct. Biol. 1995, 2, 458–65.
(46) Cha, S. S.; Lee, D.; Adams, J.; Kurdyla, J. T.; Jones, C. S.; Marshall,
L. A.; Bolognese, B.; Abdel-Meguid, S. S.; Oh, B. H. High-resolution
X-ray crystallography reveals precise binding interactions between
human nonpancreatic secreted phospholipase A2and a highly potent
inhibitor (FPL67047XX). J. Med. Chem. 1996, 39, 3878–81.
(47) Hansford, K. A.; Reid, R. C.; Clark, C. I.; Tyndall, J. D. A.;
Whitehouse, M. W.; Guthrie, T.; McGeary, R. P.; Schafer, K.; Martin,
J. L.; Fairlie, D. P. D-Tyrosine as a chiral precusor to potent inhibitors
of human nonpancreatic secretory phospholipase A2(IIa) with anti-
inflammatory activity. ChemBioChem 2003, 4, 181–185.
(48) Maestro, version 9.0; Schro ¨dinger, LLC: New York, 2009.
(49) Schro ¨dinger Suite 2009 Protein Preparation Wizard, Epik version 2.0,
Impact version 5.5, Prime version 2.1; Schro ¨dinger, LLC: New York,
(50) Epik, version 2.0; Schro ¨dinger, LLC: New York, 2009.
(51) Shelley, J. C.; Cholleti, A.; Frye, L. L.; Greenwood, J. R.; Timlin,
M. R.; Uchimaya, M. Epik: a software program for pKa prediction
and protonation state generation for drug-like molecules. J. Comput.-
Aided Mol. Des. 2007, 21, 681–91.
(52) MacroModel, version 9.7; Schro ¨dinger, LLC: New York, 2009.
(53) Kaminski, G. A.; Friesner, R. A.; Tirado-Rives, J.; Jorgensen, W. L.
Evaluation and reparametrization of the OPLS-AA force field for
proteins via comparison with accurate quantum chemical calculations
on peptides. J. Phys. Chem. B 2001, 105, 6474–6487.
(54) Polak, E.; Ribiere, G. Note sur la convergence de me ´thodes de
directions conjugue ´es. ReV. Fr. Informat. Recherche Operationelle Ser.
Rouge 1969, 16, 35–43.
(55) Still, W. C.; Tempczyk, A.; Hawley, R. C.; Hendrickson, T. Semi-
analytical treatment of solvation for molecular mechanics and dynam-
ics. J. Am. Chem. Soc. 1990, 112, 6127–6129.
(56) LigPrep, version 2.3; Schro ¨dinger, LLC: New York, 2009.
(57) Eldridge, M. D.; Murray, C. W.; Auton, T. R.; Paolini, G. V.; Mee,
R. P. Empirical scoring functions: I. The development of a fast
empirical scoring function to estimate the binding affinity of ligands
in receptor complexes. J. Comput.-Aided Mol. Des. 1997, 11, 425–
(58) Gasteiger, J.; Marsili, M. Iterative partial equalization of orbital
electronegativity-a rapid access to atomic charges. Tetrahedron 1980,
(59) Purcell, W. P.; Singer, J. A. A brief review and table of semiempirical
parameters used in the Hueckel molecular orbital method. J. Chem.
Eng. Data 1967, 12, 235–246.
(60) SYBYL/QSAR and COMFA, version 8.0; Tripos Inc.: St. Louis, MO,
(61) Clark, M.; Cramer, D. R., III; Van Opdenbosch, N. Validation of the
general purpose tripos 5.2 force field. J. Comput. Chem. 1989, 10,
(62) Stone, M. Cross-validatory choice and assessment of statistical
predictions. J. Roy. Stat. Soc. B Stat. Meth. 1974, 36, 111–147.
(63) Cramer, R. D., III; Bunce, J. D.; Patterson, D. E.; Frank, I. E.
Crossvalidation, bootstrapping, and partial least squares compared with
multiple regression in conventional QSAR Studies. Quant. Struct.-
Act. Relat. 1988, 7, 18–25.
(64) Shao, J. Bootstrap model selection. J. Am. Stat. Assoc. 1996, 91, 655–
(65) Burley, S. K.; Petsko, G. A. Aromatic-aromatic interaction: a mech-
anism of protein structure stabilization. Science 1985, 229, 23–8.
(66) Hunter, C. A.; Singh, J.; Thornton, J. M. Pi-pi interactions: the
geometry and energetics of phenylalanine-phenylalanine interactions
in proteins. J. Mol. Biol. 1991, 218, 837–46.
COMFA STUDIES ON INDOLE INHIBITORS
J. Chem. Inf. Model., Vol. 50, No. 9, 2010 1601