Current Computer-Aided Drug Design, 2010, 6, 37-49 37

Pharmacophore Based Drug Design Approach as a Practical Process in

Drug Discovery
Qingzhi Gao*,1, Lulu Yang2 and Yongqiang Zhu*,3

1
Tianjin Key Laboratory for Modern Drug Delivery & High-Efficiency, School of Pharmaceutical Science and
Technology, Tianjin University, 92 Weijin Road, Nankai District, Tianjin 300072, China
2
Auriver Inc., 1567 Quebec Ct. #2, Sunnyvale, CA 94087, USA
3
Jiangsu Simcere Pharmaceutical Research Institute, No. 699-18 Xuan Wu Avenue, Xuan Wu District, Nanjing 210042,
China

Abstract: This review summarizes the background and updated progress of pharmacophore based drug design and provides the
fundamental approach strategies on both structure based and ligand based pharmacophore approaches. The different programs and
methodologies enable the implementation of more accurate and sophisticated pharmacophore model generation and application in drug
discovery. This review will discuss and illustrate their advantages in pharmacophore based virtual screening and exemplify the detailed
application workflow, which can be easily utilized by pharmaceutical bench work medicinal chemists. Pharmacophore based drug design
process includes pharmacophore modeling and validation; pharmacophore based virtual screening, virtual hits profiling and lead
identification. Strategies and proven methodologies for pharmacophore modeling are described including common feature and 3D QSAR
based pharmacophore generation as well as structure based pharmacophore development. Different virtual screening strategies will be
described in this review with detailed case studies for supporting practical applications. Representative success examples of
pharmacophore based virtual screening for lead generation will be collected to demonstrate capabilities.
Keywords: Pharmacophore, drug design, virtual screening, ligand based pharmacophore, structure based pharmacophore, pharmacophore
based drug design.

INTRODUCTION
Computer-Aided Drug Design (CADD) has become an integral
part of drug discovery and development efforts in the
pharmaceutical and biotechnology industry. QSAR techniques have
been used for this purpose for over 50 years [1]. However, since the
1980’s, the structure-based design technology has evolved, and
today, these techniques are being widely employed and credited for
the discovery and design of most of the recent drug products in the
market. Due to rapid technological progress in chemistry and
bioinformatics, structural biology and computer technology, CADD
approaches in molecular docking studies, library design and
profiling, high throughput and virtual screening, along with
target/structure based de novo design, have also become powerful
tools and are routinely used in the multi-step process of drug
discovery. As an emerging technology, CADD accelerates drug
development by making use of the accumulated information of
existing drugs and biological targets, combined with
interdisciplinary information from different fields.
The process of pharmaceutical R&D is time consuming and
challenging. In principle, the drug discovery process involves a
sequence of distinct activities (Fig. 1). Target identification and
validation, identification of primary hits and lead compounds
followed by lead optimization, these comprise the drug discovery
process and make up the research phase of a prospective medicine's
life. As one of the most time consuming discovery stages, the
average time from target identification to filing of an
Investigational New Drug (IND) application (after completion of
*Address correspondence to these authors at the Tianjin Key Laboratory for
Modern Drug Delivery & High-Efficiency, School of Pharmaceutical
Science and Technology, Tianjin University, 92 Weijin Road, Nankai
District, Tianjin 300072, China; Tel: +1-408-750-6961; Fax: +1-408-720-
1836; E-mail: qingzhi@gmail.com and Jiangsu Simcere Pharmaceutical
Research Institute, No. 699-18 Xuan Wu Avenue, Xuan Wu District,
Nanjing 210042, China; Tel: 86-25-85566666-1726; Fax: 86-25-85566666-
1835; E-mail: zhyqscu@hotmail.com
the preclinical development) can take from 5 to 7 years. The total
R&D cost for new active and safe drugs after all clinical trials
(Phase I to Phase III) is higher than ever and continues to increase.
Today it takes over ten years and approximately from $500 million
to more than $2,000 million depending on the therapy to bring a
new drug to market [2-7].
The use of CADD technologies has the ability to accomplish
several goals especially in early research stages to improve the
efficiency of each process as well, thus reducing costs. Namely, the
growing effort to apply CADD technologies to above mentioned
each process is to expedite and facilitate target validation, hit
identification, lead optimization, and more extensively to predict
and optimize the absorption, distribution, metabolism, excretion
and toxicity profile and avoid safety issues.
On the other hand, academic scientists, software companies as
well as pharmaceutical industry are actively involved in
development of computational tools that will improve effectiveness
and efficiency of drug discovery and development process. They
integrate algorithms of drug discovery, bioinformatics,
cheminformatics, computational biology, computational chemistry,
system biology, and research informatics with visualization tools,
increase the reliability, predictability, and user-friendly nature of
the programs. Skillful application of such CADD products in
pharmaceutical R&D definitely streamline drug design,
optimization and development, increase the efficiency, decreased
use of labor resource and animals, and therefore reduce the time
and cost been used for R&D. It is expected that the power of
CADD approach will grow as the technology continues to evolve.
Commonly used CADD technologies include ligand-based drug
design (pharmacophore based approach), structure-based drug
design (molecular docking based approach) [8, 9], and quantitative
structure-activity and quantitative structure-property relationships
[10, 11]. At the times when the structure of the protein target is not
available, pharmacophore perception and use in drug design has
proven to be one of the most productive approaches [12]. On the
other hand, in the recent years, pharmacophore based drug design

1573-4099/10 $55.00+.00 © 2010 Bentham Science Publishers Ltd.

38 Current Computer-Aided Drug Design, 2010, Vol. 6, No. 1
Fig. (1). The drug development process: from discovery to commercialization.
has considerably extended its range of applications, spanning
almost all stages in the drug discovery process, from lead
identification by pharmacophore-based high-throughput and in
silico screening to drug discovery with structure based
pharmacophore approach and pharmacophore based molecular
docking. These methods help predict and retrieve potential active
molecules from rational design and screening processes, identify
common aspects of the active compounds, extract and identify the
recognition mechanisms between ligands and receptors.
Recent pharmacophore related CADD techniques could be
classified into two categories that are structure-based
pharmacophore approaches and ligand-based pharmacophore
approaches. These pharmacophore based concepts and applications
will be the focus of this review.
PHARMACOPHORE
A pharmacophore was first defined by Paul Ehrlich in 1909 as
“a molecular framework that carries (phoros) the essential features
responsible for a drug’s (pharmacon’s) biological activity” [13]. In
1977, this definition was updated by Peter Gund to “a set of
structural features in a molecule that is recognized at a receptor site
and is responsible for that molecule's biological activity” [14]. The
IUPAC definition of a pharmacophore is “an ensemble of steric and
electronic features that is necessary to ensure the optimal
supramolecular interactions with a specific biological target and to
trigger (or block) its biological response” [15].
In modern computational chemistry, pharmacophores are used
to define the essential features of one or more molecules with the
same biological activity. A database of diverse chemical
compounds can then be searched for more molecules which share
the same features located a similar distance apart from each other
From the definition of pharmacophore, it’s easy to imagine the
importance of another concept about molecular recognition.
Molecular recognition plays an important role in biological systems
and is observed in between receptor-ligand,antigen-antibody, DNA-
protein, sugar-lectin, RNA-ribosome, etc. The basic model in
understanding molecular recognition was given by the “grandfather
of bioorganic chemistry”, Emil Fischer [16, 17]: the substrate is
recognized by the receptor in forming a complex (supramolecule)
[18]. An important example of molecular recognition is
the antibiotic vancomycin that selectively binds with
the peptides with terminal D-alanyl-D-alanine in bacterial cells
through five hydrogen bonds. The vancomycin is lethal to the
bacteria since once it has bound to these particular peptides they are
unable to be used to construct the bacteria’s cell wall. Molecular
recognition is not only important in understanding the selectivity of
organic molecules, it is an essential concept for drug discovery as
well as our life processes. Most of the biological communication
within living systems is based on molecular recognition.
Gao et al.
There are a couple of most important aspects of molecular
recognition we need to be aware about: it is related to
conformational change on both receptor and substrate. It is very
easy to understand that different drug molecules may bind to the
same binding site of a specific receptor. From biological study, we
now believe that the conformational change of the receptor induced
by the binding of a ligand also results in the specific effect. On the
one hand, for most of the substrates, they may have some or high
degree of flexibility. Hence, the conformational changes of the
ligand during binding to the receptor also have to be considered for
the “induced fit”. And finally, it is known, for example, from the
opioid receptor studies (e.g. enkephalins) [19] that different
receptors can bind the same molecule in different conformations.
According to these basic knowledge on drug-receptor
relationship, we may easily figure out that a pharmacophore as an
ensemble of steric and electronic features for lidang-receptor
binding, it can be multiple and changeable for a specific drug and a
specific receptor to ensure the optimal supramolecular interactions
and to trigger (or block) its biological responses. With the same
protein target, different drugs may have the same binding mode
thus share the same pharmacophore. Different drugs may also take
the same or a different binding mode with the same protein target,
this includes: 1) binding to different location in the active site, 2)
interact with different feature set in the same location of the active
site [20] (Fig. 2).
Typical pharmacophore features are for where a molecule is
hydrophobic, aromatic, a hydrogen bond acceptor, a hydrogen bond
donor, a cation, or an anion. The features represent chemical feature
complementarities to the receptor in the 3D space and need to
match different chemical groups with similar properties, in order to
identify biologically active ligands. Ligand-receptor interactions are
typically “polar positive”, “polar negative” or “hydrophobic”. A
well-defined pharmacophore model includes both hydrophobic
volumes and hydrogen bond vectors. Further enhancement of a
pharmacophore can be obtained by combining its pharmacophoric
features with shape and exclusion volumes (steric) constraints [21,
22]. These enhancements decrease likelihood of candidate substrate
with a suitable 3D arrangement of functional groups but wrong
shape that could prevent them from fitting into the binding site of
the target.
Pharmacophore requires knowledge either from active ligands
and/or active site of the target receptor. There are a number of ways
to build a pharmacophore. It can be done by common feature
analysis to find the chemical features shared by a set of active
compounds that seem commonly important for receptor interaction.
Alternately, diverse chemical structures for certain numbers of
training set molecules along with the corresponding IC50 or Ki
values can be used [23, 24] to correlate the three-dimensional
arrangement of their chemical features with the biological activities
of training set molecules. Structure based pahrmacophore
generation from active site of a known protein target or a ligand-
Pharmacophore Based Drug Design Approach
Fig. (2). Binding modes and drug-receptor relationship.
receptor complex have also been a well-validated and very useful
method [25-28].
VIRTUAL SCREENING
Virtual screening is a widely accepted method in lead
discovery, because it is advantageous in the elimination of
undesired molecules from compound libraries and the reduction of
cost and time in drug discovery projects [29-36]. Compare with ten
years ago, computer aided virtual screening nowadays has become
a mature and well established tool for routine drug research and
discovery. An increasing number of success stories are
communicated in different drug projects and most of them have
succeeded to retrieve micro- to nanomolar hits or lead compounds
in a truly predictive manner. With the following section, we will
briefly outline and discuss by means of some instructive examples
from recent research where, the successful applications with
detailed techniques will be beneficial to our medicinal chemists and
research scientists.
Fig. (3). Workflow for pharmacophore based virtual screening and lead
generation.
Current Computer-Aided Drug Design, 2010, Vol. 6, No. 1 39
As illustrated in Fig. (3), in general, virtual screening proceeds
through a number of stages to deliver novel hit compounds and
desired lead molecules.
STAGE-1: Pharmacophore Generation
When the chosen target has three-dimensional structural
information, such as X-ray crystallography co-ordinates associated
with an active ligand, the pharmacophore can be generated by
performing a structure based techniques. In this case, simply extract
the binding points from the ligand-protein complex, with their
corresponding interaction features. For the case where no 3D
structure could be experimentally obtained, for example,
transmembrane receptors, homology modeling can be used to find
the correct structure and the use of molecular docking combine with
mutagenesis study will help to define the ligand-receptor interaction
and binding mode. There are a couple of commercial tools can be
used to generate pharmacophore models from protein structures:
DS Structure Based Pharmacophore [37-39], MOE [40],
LigandScout [41, 42], and Phase [43, 44] are well validated
software packages for the automated generation of pharmacophore
models.
When the protein structure is unknown, the ligand-based
approach can be useful to suggest possible pharmacophore queries
based on a set of aligned active compounds. Basically to create
optimal ligand based pharmacophore models, as above mentioned,
we can use two methods: common feature alignment, and activity-
based 3D QSAR modeling. DS Catalyst Hypothesis, Phase, and
MOE are commonly used for this purpose.
STAGE-2: Virtual Screening
Virtual screening (VS) use well-validated pharmacophores, if
applied appropriately, could easily find the desired virtual hits with
a reasonable hit rate. This process is based upon a similarity search
of molecule’s chemical feature in 3D space to the selected
pharmacophore. Recently, the version of 3D pharmacophore-based
VS has received increased attention. This is due to two major
factors: first, the public availability of the pharmacophore
searchable chemical database include ZINC [45] has provided a
large selection pool with high-quality and purchasable small
molecules; second, new computational technology has enabled a
more accurate and flexible definition of pharmacophore models
coupled with an efficient search speed. The major goal of this step
is to achieve improved specificity and sensitivity as well as
accuracy of pharmacophore based VS by optimizing the variables
used to generate conformations of small molecules and those
additional features used to construct pharmacophore models from
known inhibitors or from inhibitor-protein complex structures
(Stage 1). The impact of the key things, including the database
40 Current Computer-Aided Drug Design, 2010, Vol. 6, No. 1
selection, involvement of excluded volumes (EV), the tolerance
radius of excluded volumes, method for conformational search,
energy windows, and the maximum number of conformers in
conformation generation [46].
STAGE-3: Generation of Primary Hits
For the pharmaceutical industry, the purpose of the initial
screen for finding new drug candidates is to limit the number of
chemicals that proceed to the next (more expensive) phase of
testing, therefore reduce the numbers of false positives is very
important and collected virtual hits must be a reasonably small
numbers. Therefore, the strategy and method of filtering of the
virtually collected molecules are of critical for this stage. Bioassay
for those virtual hits from commercial or in-house compound
database will directly provide information for false positives, thus
the experimentally selected true hits we usually call it as primary
hits. For the case of virtually designed compound library, chemists
may face some challenges in order to synthesize the collected
virtual hits. The interaction and cycle between bench chemistry and
biochemistry is important to obtain the initial SAR information
from such active molecules of primary hits.
STAGE-4: Generate Lead Compounds
Sometimes the obtained primary hits can be a well-tolerated
lead compound. But in most cases the lead compounds are derived
from so-called identification and optimization process. The
concepts of greater drug-likeness are used routinely to refine
primary hits for lead generation. On the other hand, the compound’s
IP position is more important to the organization. Challenges also
remain to develop chemistry amenable to the introduction of
multiple diversity components while creating molecules of lead-like
and drug-like properties. Despite the common capability to
synthesize and screen thousands of molecules, the current focus has
shifted to smaller libraries targeted with greater computational
sophistication according to putatively smarter target-drug premises
[47]. Throughout the lead identification and optimization step,
many CADD approaches are useful to accelerate and improve the
process. QSAR analysis of the obtained biological activities of the
primary hits, rational design and modification based on the
pharmacophore mapping analysis, molecular docking study on the
selected primary hits are usually combined with the lead generation
process.
LIGAND BASED PHARMACOPHORE GENERATION
As scientists continue to search for new targets of therapeutic
interest, transmembrane and G-protein coupled receptors are of
ever-increasing importance. However, crystal structures for these
targets may be impossible to resolve, posing great challenges in
rational drug design. Structure based virtual screening is not
directly an option when the active site geometry is unknown, but
assaying an entire library for hits is an inefficient and expensive
proposition. In this case, ligand based pharmacophore approach is
the major strategy and focus.
Pharmacophore generation from active ligands can be divided
into two categories: shared feature pharmacophore (common
feature pharmacophore) and QSAR based pharmacophore. For the
basic algorithm of shared feature pharmacophore generation, we
can take the example from LigandScout (Inte:Ligand ) [25, 41, 48-
50] to explain the concept: LigandScout creates a shared feature
pharmacophore using a set of selected elements and aligns them
using the alignment algorithm [51]. The resulting alignment of all
selected elements will be searched for valid feature overlappings.
This means that one feature of a fixed element is considered and all
its valid matchings during the alignment. If this feature was validly
matched by one feature of each of the other elements these features
are considered as a valid matched feature set. If every feature lies
within the tolerance of all the other features in this valid matched
Gao et al.
feature set, a new feature will be derived by combining all the
original features. The resulting feature consists of the mean of all
positions and the mean of all tolerances. This combination is done
for every set of overlapping features and results in a shared feature
pharmacophore. The QSAR based pharmacophore method, namely,
the pharmacophore models generated from SAR data of a set of
active ligands.
Programs for Ligand Based Pharmacophore
Several companies have developed computational tools that
provide automated ligand based pharmacophore generation. DS
Catalyst, from Accelrys, includes algorithms for both shared feature
pharmacophore (Catalyst/HipHop) and QSAR based
pharmacophore generation (Catalyst/HypoGen). The Catalyst
software suite can be used for 3D pharmacophore modeling,
pharmacophore-based alignment of molecules and generation of
pharmacophore hypotheses based on common feature and SAR
data. As shared feature pharmacophore method, Catalyst/HipHop
identifies pharmacophoric hypotheses common across a series of
active compounds. These models are also used to perform feature-
based alignments that can be used subsequently in the development
of 3D QSAR and other interpretive models. During pharmacophore
generation, Catalyst/HipHop matches chemical features, such as
surface-accessible hydrophobes, surface-accessible hydrogen bond
donors/acceptors, charged/ionizable groups, and features defined by
users.
There has been some comparison studies reported the
differences between Catalyst/HypoGen from Accelrys [52] and
Phase program from Schrodinger [43, 53]. Both programs provide
3D QSAR based pharmacophore generation method from ligands
and their biological activities (SAR data). They search for common
3D pharmacophores amongst compounds in the training set, and
aim to quantitatively predict activity using a training set of
compounds with known activity. Catalyst/HypoGen uses the
distance between the pharmacophore features (site points) in the
hypotheses and matching feature in the compounds [54] while
Phase scores compounds using a grid-based 3D QSAR model built
from the compounds aligned on the hypotheses [43, 53]. The study
results using eight public data sets demonstrated that the
performance of Phase is better than or equal to that of
Catalyst/HypoGen in terms of the predictivity of obtained
pharmacophores.
Chemical Computing Group’s MOE pharmacophore modeling
methodology comprises tools for scaffold replacement that identify
substitution points on a scaffold molecule and the location of
potential R-group substituents, as well as tools for pharmacophore
elucidation and searching [55-59]. Pharmacophore elucidation
generate pharmacophore queries and induced molecular alignments
from a collection of input compounds with activity data by
considering all possible discrete geometries and all possible
combinations of feature query expressions. A build-up strategy is
used to avoid combinatorial explosion. Enforce limits on feature
counts and add custom query expressions. Score queries based on
known active compound coverage, statistical activity enrichment
and atomic overlap of matching conformations. The
Pharmacophore Consensus application that suggests
pharmacophore queries based on a set of aligned active compounds.
A consensus calculation requires a set of aligned input molecules, a
tolerance radius, the consensus score threshold and the consensus
score mode. Pharmacophore consensus is particularly useful when
starting from a few highly active compounds similar with
Catalyst/HipHop.
Recently, Schneider and colleagues described an approach that
combines a 3D alignment-free pharmacophore descriptor with
artificial neural networks (ANNs). The ANNs were trained for
receptor selectivity using known antagonists with IC50 < 1
m as
reference structures. Self-organizing maps (SOM) helped select
Pharmacophore Based Drug Design Approach
structurally diverse compounds for bioactivity testing via virtual
screening. Several hits were successfully identified using this novel
methodology, include the new GluR5 antagonist generation [60].
Ligand-based pharmacophore elucidation starting from 3D
conformational search of bio-active molecules. For this purpose,
different programs share the same or have their specialties in terms
of conformational search algorithm and to generate and present
molecular data including geometries (bond lengths, bond angles,
torsion angles), energies (heat of formation, activation energy, etc.),
electronic properties (moments, charges, ionization potential,
electron affinity), and bulk properties (volumes, surface areas, etc.).
Any conformation generation algorithm [61] for the purposes of
pharmacophore modeling needs to address two important issues:
adequate coverage of the energy landscape and diversity
represented among the conformational models of the compound
under consideration. DS Catalyst [62] using the Poling algorithm
[63], which attempts to generate diverse conformers in feature
space [64-66]. In a recently published study [67], scientists
investigated Catalyst’s ability to generate biologically relevant
conformers using 510 protein-ligand complexes obtained from the
Protein Data Bank (PDB) [68]. The result demonstrated that the
conformational search implemented in Catalyst, is significantly
more effective in sampling the full range of conformational space
compared to the other methods. This work represents the most
comprehensive external review focused on the quality of
conformers generated in Catalyst and the effect of several
parameters (FAST vs BEST methods, maximum numbers of
conformers generated, and the choice of energy threshold). The
study reveals that the recommended settings of generating 255
conformers using the BEST method and a 20 kcal/mol produces
conformations with RMS of less than 1 Å to the X-ray structure of
the ligand. Both FAST and BEST methods provide best fitting
conformers with RMS < 1.50 in more than 80% of all cases and
with RMS < 2 in more than 93% cases. Commonly used force
fields estimate that the energy of biologically active conformations
are, in most of the cases, at a higher level than the computed global
energy minimum and that Catalyst does well in predicting such
conformers when the appropriate energy threshold values are
chosen. These high quality conformers can be subsequently used
for pharmacophore modeling and ligand based drug design.
STRAUCTURE BASED PHARMACOPHORE GENERATION
From the historically grown archive of protein-ligand
complexes in the PDB, small organic ligands could be extracted and
interpreted in terms of their chemical characteristics and features,
especially in terms of bioactive conformation. Subsequently,
structure based pharmacophores can be generated representing
ligand-receptor interaction which derived from each of these small
molecules and its surrounding amino acids. Based on a defined set
of about six types of chemical features (hydrogen bond
donor/acceptor, anion/cation, aromatic ring, hydrophobic group)
and volume constraints, three-dimensional pharmacophore models
can be constructed, which can be used to identify the ligand binding
mode, differentiate compounds and are thus proved to be a useful
tools for in-silico screening for lead generation.
Programs for Structure Based Pharmacophore
Recently, more intuitive and computationally more efficient
structure based pharmacophore methods have been reported that
seek to find effective means to utilize experimental structure
information without employing detailed docking calculations.
These tools can (should) be coupled with efficient, experimental
screening technologies to improve the probability of success in the
discovery process. For example, LigandScout has been successfully
applied in several virtual and experimental HTS projects [69]. They
are used as effective virtual screening tools and provide testable
hypotheses for medicinal chemists to study experimentally [25].
Current Computer-Aided Drug Design, 2010, Vol. 6, No. 1 41
Other commercial tools that have similar functionality include DS
Structure Based Pharmacophore [37-39], MOE [40] and Phase [43,
44]. DS Structure Based Pharmacophore (formerly C2•SBF) from
Accelrys is a user-friendly program package for medicinal chemists
[70]. This module extends the structure based design capabilities of
the Cerius2 environment by leveraging validated technologies from
Ludi [26, 71-81] and Catalyst. C2•SBF derives Ludi maps directly
from receptor site features, and converts this information into
pharmacophore models suitable for rapid 3D database searching.
Active site maps can be edited and clustered by medicinal chemists
so that only essential information is retained, and these interaction
points then can be converted to 3D pharmacophores for virtual
screening [25-27]. Another benefit from this program that
medicinal chemists will be able to use the extracted 3D
pharmacophores for virtual screening through direct access to DS
Library Design, in which user can design focused compound library
by using their chemistry expertise and build their own compound
database and searched with the structure based pharmacophores.
C2•Ludi and C2•LigandFit modules in this program will also help
chemists to filter and prioritize the virtual hits by Ludi score and
perform extended flexible docking studies [82-94].
Shape-Constrained Pharmacophore
The concept of a structure based pharmacophore is based on a
crystal structure of a complexed ligand (or a homology model of the
protein). In this case, the shape of the molecule, as indicated by the
crystal structure, can be also used as a part of the pharmacophore
query. This combination of pharmacophore features and the shape
of a molecule can provide a highly selective 3D search query
capable of retrieving much fewer virtual hits than ordinary queries
but with substantial enrichment. Generally speaking, 3D shape
query generation can be considered from two different ways: from
protein active site and from active ligand. The active site based 3D
shape query is easily constructed from a receptor binding surface
and can be used as a rough filter in the virtual screening or combine
with both the ligand or structure based 3D pharmacohores for
virtual screening [95]. Ligand based shape pharmacophore
approach belongs to 3D similarity search, which is one of the most
popular database mining techniques, utilizing the concept that
compounds that are similar to each other may possess similar
activities [95-97]. It retrieves compounds with more diversity when
compared to the traditional substructure searches. A substructure or
feature search is “exact”, it retrieves compounds that contain the
exact substructure or features presented in the query. However, a
shape search is “fuzzy”, it retrieves compounds that have some
resemblance to the defined spatial query. Therefore, as
complementary tool, the use of active ligand based shape
pharmacophore combine with above mentioned feature
pharmacophore becoming a significant strategy for virtual
screening.
DS Catalyst Shape in this purpose is useful to perform a virtual
screening by adding a shape query into the ligand or structure based
feature pharmacophores. The successful use of such a query was
exemplified by researchers at Biogen Idec Inc. Using transforming
growth factor type 1 (TGF1) receptor as a novel drug target for the
potential treatment of diseases such as scleroderma, Singh and co-
workers developed a receptor-based pharmacophore model, and
using the X-ray structure of a weak inhibitor of TGF1 receptor, a
combined shape and pharmacophore model was developed. A
multi- conformational database of commercially available
compounds was then built and screened with the pharmacophore +
shape model, resulting in the discovery of a highly potent inhibitor,
the binding of which was also validated by X-ray crystallography.
This research has received significant publicity due to researchers at
Eli Lilly identifying the same lead molecule as that discovered at
Biogen Idec but using purely experimental high-throughput
screening assays [98, 99].
42 Current Computer-Aided Drug Design, 2010, Vol. 6, No. 1
An idea has been tried in our group, to use a merged shape-
pharmacophore query based on the bound conformation of several
active binding ligands instead of the single active compound, but
currently it was not successful due to the non-mergeable attribute of
the shape queries [100-105].
Most of the aforementioned commercial programs also provide
complementary tools to add excluded volumes during
pharmacophore generation process to impose geometric and
topographic constraints to avoid ligand receptor clashes. It can be
achieved simply with the programs for example during structure
based pharmacophore generation by adding such volumes from
residues on the protein surface.
Zheng and Liang’s group recently reported a new structure
based shape pharmacophore method for virtual screening based
on the shape technology developed by Open Eye Scientific. This
new method uses special shape analysis to detect the binding site
atoms and to construct the negative image (pseudoligand) of a
target binding site and was used as shape query to achieve fast
shape overlays between the negative image and database
molecules [106]. This is a new method using accurate shapes
derived from rigorous computational geometry analysis of the
binding site to construct shape-based pharmacophore queries and
use them for large-scale virtual screening.
APPLICATIONS OF PHARMACOPHORE BASED DRUG
DESIGN
Pharmacophore based drug design has become an essential tool
in medicinal chemistry research and pharmaceutical industry. It is
widely applied and created a lot of success stories. The advantages
of using pharmacophore based drug design approach for medicinal
chemists compare with molecular docking and structure based de
novo design is the “simplicity”. Medicinal chemists will easily to be
more familiar with feature based pharmacophore and use them to
map chemical functional groups and design new structures. In this
section, several practical topics related to pharmacophore based
virtual screening and lead generation will be examined in detail.
Some other case studies and successful applications have been
summarized in Table 1 following the key features of each study.
Discovery of Novel CB 2 Receptor Ligands by a Step-4: Filtering by Structural Similarity Analysis Method
Pharmacophore Based Virtual Screening [107]
Cannabinoid receptor 2 (CB2 receptor) ligands are potential
candidates for the therapy of chronic pain, inflammatory disorders,
atherosclerosis, and osteoporosis. This study reported the
development of pharmacophore models for CB2 receptor ligands,
as well as a pharmacophore based virtual screening workflow,
which resulted in 14 virtual hits after several filtering steps for
experimental bioassay. Seven compounds were identified with Ki
values below 25 μM. One of the pyridine tetrahydrocannabinol
derivative was identified with Ki = 1.78 μM as a CB2 partial
agonist. Two acetamides derivatives with Ki = 1.35 μM, and 2.1
μM represent new scaffolds were obtained as CB2 receptor-
selective antagonists and inverse agonists, respectively. Overall,
this ligand based pharmacophore virtual screening process yielded
three novel scaffolds for the chemical development of CB2 receptor
ligands. The workflow for the entire process was illustrated in Fig.
(4) followed by detailed explanation for each step.
Step-1: Ligand Based Pharmacophore Generation
A Catalyst/HipHop common feature-based pharmacophore
generation method was used from 5 training set compounds resulted
in 10 pharmacophore models. The two models with the largest
number of features were selected. Both models included one
hydrogen bond acceptor feature and four hydrophobic features.
From each of the five compounds of the CB2 receptor ligand
training set, a Catalyst shape that represents the spatial information
of the corresponding compound was derived and combined with
Gao et al.
both the selected models. Total 10 merged models (feature
combined with shape) were resulted and validated before virtual
screening.
Step-2: Pharmacophore Validation Method
In order to validate the discriminatory power of the 10 merged
models, this study performed a screening test using two validation
databases that included 15 CB2 receptor ligand test set compounds
and 67,046 marketed and developmental drugs from the Derwent
World Drug Index 2005 (Derwent WDI) [108], which served as
decoys. The CB2 receptor ligand enrichment among the validation
database compounds was expressed as enrichment factor (EF),
which quantifies the number of times the enrichment of ligands
improves using pharmacophore models compared to a random-
based selection of compounds [109, 110]. The best two
pharmacophore models (feature + shape) were selected from this
validation results, which retrieved 73% and 67% of the test set CB2
receptor-selective agonists respectively and showed the highest EF
value (32 and 28).
Step-3: Virtual Screening and Physicochemical Filtering
Six chemical databases containing 922,944 compounds in total
were screened utilizing selected two pharmacophore models, and
29,565 unique structures (3% of all virtually screened compounds)
were obtained from the first virtual screening step. Catalyst Best
Flexible Search method was used. To narrow down the number of
virtual hits, subsequently, these virtual screening hits retrieved from
the chemical databases were filtered using a physicochemical
property filter. To determine the physicochemical filter criteria, 20
compounds from the CB2 receptor ligand training and test set were
analyzed using the software package MOE. An analysis of the
physicochemical property distribution resulted in the following
filter criteria: number of heavy atoms,
35; molecular weight,

500; number of hydrogen bond donors,
2; number of hydrogen
bond acceptors,
6; AlogP
8. Finally, a Pipeline Pilot script
(Accelrys) was created to automatically filter the first 29,565 hits
according to these filter criteria and resulted 22,253 compounds that
was about 2% according to the whole chemical database.
To further narrow down the numbers of the virtual hits, this
study adopted the same script as physicochemical property analysis
and performed a structural similarity clustering analysis. From the
remaining 22,253 virtual hits, this analysis yielded 605 clusters.
The 605 cluster centers (0.07%) were selected and inspected
visually with respect to chemical stability and toxicity. For
example, compounds that could be cleaved in aqueous solution
such as hydrolysable esters as well as structures containing
electrophilic warheads that are known having toxicities, were
discarded [111]. Overall, 14 compounds (0.002%) fulfilled all filter
criteria and were purchased for real biological assay.
Step-5: Biological Assay and True Hits Collection
The displacement assay using radioligand by the 14 selected
compounds was measured to determine their Ki values. As the
result, seven compounds showed binding affinity for the CB2
receptor, and 4 compounds were determined as weak CB2 receptor
ligands with binding affinities lower than 25 μM, whereas 3 high
binding affinity compounds were found with Ki values: 1.78 μM,
1.35 μM, and 2.1 μM, which possess a 27-fold, 37-fold, and 23-fold
selectivity for the CB2 receptor, respectively.
Probing the Pharmacophore of Ginkgolides as Glycine
Receptor Antagonists [112]
Ginkgolides have been in the remarkable spotlight in CNS
related new drug discovery. Ginkgolides are antagonists of the
inhibitory ligand-gated ion channels for the neurotransmitters
Pharmacophore Based Drug Design Approach
Fig. (4). Workflow for pharmacophore based virtual screening in CB2 receptor ligand discovery.
glycine and
-aminobutyric acid (GABA). In this study the
ginkgolide structure was modified in order to investigate the
minimum structural requirements for glycine receptor antagonism.
The five native ginkgolides and a series of 29 ginkgolide
derivatives were characterized at the three glycine receptor
subtypes R1, R1, and R2, which revealed that only minor changes
in the ginkgolide skeleton were allowed for maintaining glycine
receptor antagonism. A pharmacophore model was generated and
applied in a virtual screening of a compound database (300,000
compounds), resulting in the identification of 31 hits. Twenty-seven
of these virtual hits were purchased or synthesized and screened for
biological activity, but none displayed desired antagonist activity at
the glycine receptors.
STEP-1: Ligand Base Pharmacophore Generation:
Due to the very rigid structure of active ginkdolides, the
pharmacophore model was generated by simply overlaying the X-
ray crystallographic structures of two native ginkgolides GA and
GB with another two computationally minimized structures of
compounds GC and GM. The superimposition of these structures
showed very minimal deviation and were placed in very similar
regions. The pharmacophore model was generated from the
superimposed 4 conformers using the built-in function in MOE
(Fig. 5), and four donor-acceptor interactions and a single
hydrophobic interaction feature were selected according to the
QSAR results [113].
Step-2: Pharmacophore Based Virtual Screening:
Virtual screening was conducted on a database of ca. 300,000
commercially available compounds using MOE and resulted in 31
hits. The structures of these virtual hits were characterized by
having several hydroxyl groups and/or carbonyl groups, thus
Current Computer-Aided Drug Design, 2010, Vol. 6, No. 1 43
mimicking the pharmacophore elements of the ginkgolides. Twenty
seven of these virtual hits were purchased and examined for GlyR
activity in biological assay, which all showed IC50 values >100
μM.
As the author has notified in the paper that this negative result
obtained from ligand based virtual screening can be interpreted as
the following reasons: (1) the pharmacophore model does not fully
contain the biologically relevant information for inhibition of the
desired GlyRs, (2) the virtual hits from the screening process were
much more structurally flexible than the ginkgolides. As a
conclusion from this study, the author emphasized that the rigidity
of the ginkgolide structure is crucial for their antagonistic activities
at GlyRs.
Structure Based Pharmacophore of COX-2 Selective
Inhibitors and Identification of Original Lead Compounds
from 3D Database Searching method [114]
This application performed a ligand based common feature
analysis using Catalyst/HipHop program for COX-2 selective
inhibitors. In order to take into account topology of the human
COX-2 binding site, 44 exclusion volume spheres were generated
using ligand-COX-2 enzyme complex and combined with the
ligand based feature pharmacophore. The validated pharmacophore
models were used as 3D query to screen the Maybridge 3D
database. Several filtering techniques are applied to reduce the
number of the virtual hits and the resulted small number of hit
molecules are purchased for real biological assay. Five collected
virtual hits preferentially inhibit COX-2 and two of them have a
pharmacological profile similar to rofecoxib, a marketed COX-2
selective drug, and have an original scaffold. The study
demonstrated that, the ligand based pharmacophore approach
combining with structure based pharmacophore modification are
useful strategy in the lead generation process (Fig. 6).
44 Current Computer-Aided Drug Design, 2010, Vol. 6, No. 1
Fig. (5). Ligand crystal structure based pharmacophore generation for Ginkgolides.
Step-1: Ligand Based Common Feature Pharmacophore
Generation
In this study, the 3D pharmacophore model was generated from
Catalyst/HipHop program, using a training set of 16 COX-2
selective inhibitors. Several chemical features are used including
hydrogen bond acceptor/donor (A/D), hydrophobic group (H), ring
aromatic (R) and negative/positive ionizable (N/P). One crystal
structure was chosen as a reference compound (Principal = 2;
MaxOmitFeat = 0). It was allowed to map all features and the
others were allowed to map partially on the hypotheses. 10 common
feature pharmacophores are automatically generated from
Catalyst/HipHop.
Step-2: Pharmacophore Validation Method
The 10 generated models consist of three or four points. The
three-featured hypotheses are less specific than the other five four-
featured pharmacophores and were not retained as the first
validation. The remaining four-featured hypotheses were
characterized by performing a 2D mapping with the crystal
structure of the active COX-2 inhibitor. Based on the SAR data and
the binding mode information obtained from the crystal complex of
the COX-2 enzyme, each pharmacophore models are validated by
mapping to the inhibitor crystal conformation and the agreement
between pharmacophore feature and the true binding mode based
on the crystal structure are checked. Best fit values have been then
calculated to further evaluate the robustness of the pharmacophores
toward all training set molecules. As the result, a four-featured
pharmacophore of COX-2 selective inhibitors, it consists of a H
bond acceptor, two hydrophobic groups and an aromatic ring, in
accordance with SAR data of the training set compounds and
Fig. (6). Pharmacophore based virtual screening for selective COX-2 inhibitor.
Gao et al.
structure based binding mode is selected for further modification.
Step-3: Combination with Topological Constraints
In order to mimic the boundary of the human COX-2 active site
and to take into consideration flexibility of the side-chains, as well
as the spatial requirement of the active binding site, exclusion
(ligand-inaccessible) volume spheres of 1.0 Å diameter were added
to the selected feature pharmacophore, based on the human COX-2,
modeled from crystallized murine COX-2 in complex with active
ligand. The spheres were set on the positions occupied by the side
chains of amino acids located at 10 A° from the active site centre.
This operation method has been used to reduce the number of hits
and also the number of false positives in virtual screening [115].
The resulted 3D pharmacophore combined with exclusion volume
spheres representing important residues of the COX-2 binding site,
was then used to virtually screen the Maybridge database.
Step-4: Virtual Screening and Virtual Hits Filtering Method
The Maybridge database was searched using the “Fast Flexible
Search Databases” algorithm which only uses the precomputed
conformations of compounds during the search. The 44 exclusion
volume spheres enabled to shrink the Maybridge hit list from
12,675 to 2,577 i.e. a factor of 5. The author picked the first 300
hits fitting the model for further filtering and analyses. In order to
reduce the number of molecules to be tested biologically, several
filtering criteria were used. An initial filtering to satisfy the Lipinski
rules was first applied to the compounds: MW< 500, log P < 5,
number of H acceptors <10 and number of H donors <5 [116]. As a
next step, hits with asymmetric carbon(s) were omitted to simplify
the problem. Moreover, to avoid too flexible compounds, those
with more than seven rotatable bonds were rejected [117].
Pharmacophore Based Drug Design Approach Current Computer-Aided Drug Design, 2010, Vol. 6, No. 1 45

Table 1. Summary of Some Representative Pharmacophore Based Drug Design Applications

Case Results Approaches & Tools Ref.

Catalyst/HypoGen
5-HT7 Antagonists 3 Novel 5-HT7 antagonist identified [118]
Ligand based pharmacophore

Selective MCP Inhibitor
Non-steroidal 5
-Reductase
Inhibitors
Tyrosine Kinase Inhibitors
Inhibitors of Parainfluenza 1 Virus
Selective COX-2 Inhibitors
MAC DHFR Inhibitors

41 Integrin Antagonists
Inhibitors of 17
-Hydroxylase-
17,20-lyase
Novel Dopamine 3
Human Aldose Reductase Inhibitor
Catalyst/HipHop
Ligand based pharmacophore
41 Novel MCPI were identified At 100nM [119]
Pharmacophore based VS
Maybridge database
Catalyst/HypoGen
New Isoflavonoids identified
Ligand based pharmacophore [120]
IC50=6.9-48.7μM
Pharmacophore based VS
Catalyst/HipHop
CoMFA/CoMSIA
New lead scaffold identified [121]
Ligand based pharmacophore
Pharmacophore based rational design
New lead scaffold identified Catalyst/HypoGen
[122]
IC50=0.38-193μM Ligand based pharmacophore
Catalyst/HipHop
New lead scaffold identified Ligand based pharmacophore with excluded volume
[123]
IC50=0.65μM Structure based docking
Pharmacophore based rational design
Catalyst/HypoGen
New lead scaffold identified Ligand based pharmacophore [124]
Pharmacophore based VS
Catalyst/HypoGen
New lead scaffold identified Ligand based pharmacophore
[125]
IC50=0.6-67nM Pharmacophore based VS
8,824 ACD database
Catalyst/HipHop
New lead scaffold identified
Ligand based pharmacophore [126]
IC50=56nM
Pharmacophore based VS
Ligand/Fit
New lead scaffold identified Ligand/Score
[127]
4 compounds Ki<100nM Structure based pharmacophore
Pharmacophore based VS
FlexX
New lead scaffold identified
Structure based pharmacophore [128]
6 compounds IC50<1μM
Pharmacophore based VS

In a final step, the collected 135 remaining hits after the above
mentioned filtering steps were first docked into the human COX-2
active site using GOLD program in order to take into account
molecular complementarities between the compounds and the
COX-2 enzyme. The binding modes were ranked according to the
scoring function (called “Fitness”) of GOLD. Secondly, the
Catalyst “Best Fit” scoring was calculated for each hit, reflecting
the mapping of the compound in the 3D pharmacophore. The
compounds with Fitness < 40 and Best Fit < 1 were rejected.
From these analyses, 40 best-scored molecules still remained.
Visual inspection of the suggested binding modes with GOLD
docking analysis, together with the high scoring values of GOLD
and Catalyst, were used to select eight structures for in vitro
biological testing.
Step-5: Biological Assay and True Hits Collection
Among eight selected virtual hits, after an in vitro enzymatic
assay, three test candidates are inactive, five of them showed COX-
2 inhibition close to that of nimesulide and rofecoxib, two reference
COX-2 selective inhibitors. Two of obtained primary hits have a
pharmacological profile similar to rofecoxib, a marketed COX-2
selective inhibitor drug, and have an original scaffold in
comparison with reference inhibitors and all others proposed in
literature. Both compounds were finally docked in the human COX-
2 binding site (using GOLD program) and showed a good
complementarity with the isoform. The new structures are original
and are not members of known families of COX-2 inhibitors.
Therefore, this ligand and structure based pharmacophore virtual
screening approach afforded promising lead molecules for further
structural modifications, which could lead to more potent COX-2
selective inhibitors.
CONCLUSIONS
Virtual screening using 3D pharmacophores has evolved into an
important and successful method for drug discovery over the last
ten years. In this review we focused on recent progression in
pharmacophore based drug design and virtual screening by
description of the basic pharmacophore generation concept, newly
developed CADD programs and their applications. In order to
provide with practical techniques to our application scientists and
medicinal chemists, several case studies have been exemplified
with detailed workflow and process illustrations. Successful usage
of these programs and techniques has become a mainstream in drug
discovery for pharmaceutical industry. The newly developed
methodologies for more accurate pharmacophore model and shape
query generation are expected to be more frequently adopted and
46 Current Computer-Aided Drug Design, 2010, Vol. 6, No. 1 Gao et al.

validated by scientists. The incorporation and integration of novel [13] Ehrlich, P. Present status of chemotherapy. Ber. Dtsch. Chem. Ges.,
algorithms from continually addressing and overcoming the 1909, 42, 17-47.
flexibility issue of protein and ligand as well as the interfacial [14] Gund, P. Three-dimensional pharmacophoric pattern searching.
solvation, will be main direction and more challenging in the near Prog. Mol. Subcell. Biol., 1977, 5, 117-143.
[15] Wermuth, C.G.; Ganellin, C.R.; Lindberg, P.; Mitscher, L.A.
future. Nevertheless, recent efforts provide compelling evidence Glossary of terms used in medicinal chemistry (IUPAC
that pharmacophore based drug design and virtual screening are Recommendations 1997). Annu. Rep. Med. Chem., 1998, 33, 385-
successful when practiced in an integrated fashion involving 395.
synthetic, structural and computational expertise. The more [16] Fischer, E. Einflusse der Konfiguration auf die wirkung der
applications with the use of focused libraries with a team enzyme. Beriche Chem Gesellschaft, 1894, 27, 2985-2993.
encompassing expertise from medicinal chemists and incorporate [17] Hruby, V.J. Structure and conformation related to the activity of
information from an initial scaffold will be of a great direction in peptide hormones. In Perspectives in Peptide Chemistry, Eberle,
the future drug R&D. A.; Geiger, R.; Wieland, T. Eds. Karger, Basel, New York, 1981,
pp. 207-221.
[18] Lehn, J.M. Supramolecular chemistry: receptors, catalysts and
ABBREVIATIONS carriers. Science, 1985, 227, 849-856.
ACD = Available chemicals directory [19] Chang, K.J.; Cuatrecasas, P. Multiple opiate receptors. Enkephalins
and morphine bind to receptors of different specificity. J. Biol.
CADD = Computer-Aided Drug Design Chem., 1979, 254, 2610-2618.
CB2 = Cannabinoid receptor 2 [20] Wermuth, C.G.; Ganellin, C.R.; Lindberg, P.; Mitscher, L.A.
Glossary of terms used in medicinal chemistry. Pure Appl. Chem.,
CNS = Central Nervous System 1998, 70, 1129-1143.
COX-2 = Cyclooxygenase-2 [21] Pandit, D.; So, S.S.; Sun, H. Enhancing specificity and sensitivity
of pharmacophore-based virtual screening by incorporating
DS = Discovery Studio chemical and shape features-a case study of HIV protease
EV = Excluded volumes inhibitors. J. Chem. Inf. Model., 2006, 46, 1236-1244.
[22] Toba, S.; Srinivasan, J.; Maynard, A.J.; Sutter, J. Using
GABA = Gammaaminobutyric acid pharmacophore models to gain insight into structural binding and
virtual screening: an application study with CDK2 and human
GlyR = Glycine receptor DHFR. J. Chem. Inf. Model., 2006, 46, 728-735.
HTS = High-throughput screening [23] Funk, O.F.; Kettmann, V.; Drimal, J.; Langer, T. Chemical function
based pharmacophore generation of endothelin-A selective receptor
IND = Investigational New Drug antagonists. J. Med. Chem., 2004, 47, 2750-2760.
MOE = Molecular Operating Environment [24] Krovat, E.M.; Langer, T. Non-peptide angiotensin II receptor
antagonists: chemical feature based pharmacophore identification.
NCI = National Cancer Institute J. Med. Chem., 2003, 46, 716-726.
PDB = Protein Data Bank [25] Wolber, G.; Langer, T. LigandScout: 3-D pharmacophores derived
from protein-bound ligands and their use as virtual screening
QSAR = Quantitative structure-activity relationship filters. J. Chem. Inf. Model., 2005, 45, 160-169.
SAR = Structure-activity relationship [26] Hoffren, A.M.; Murray, C.M.; Hoffman, R.D. Structure based
focusing using pharmacophores derived from the active site of 17b-
SBF = Structure Based Focusing hydroxysteroid dehydrogenase. Curr. Pharm. Des., 2001, 7, 547-
566.
TGF1 = Transforming growth factor type 1 [27] Böhm, H.J. The computer program Ludi: a new method for the De
VS = Virtual screening Novo design of enzyme inhibitors. J. Comp. Aided Mol. Des., 1992,
6, 61-78.
REFERENCES [28] Pisabarro, M.T.; Ortiz, A.R.; Palomer, A.; Cabré, F.; García, L.;
Wade, R.C.; Gago, F.; Mauleón, D.; Carganico, G. Rational
[1] Hopfinger, A.J. Computer-assisted drug design. J. Med. Chem., modification of human synovial fluid phospholipase A2 inhibitors.
1985, 28, 1133-1139. J. Med. Chem., 1994, 37, 337-341.
[2] Dickson, M.; Gagnon, J.P. The cost of new drug discovery and [29] Rester, U. From virtuality to reality - virtual screening in lead
developement. Discov. Med., 2004, 4, 172-179. discovery and lead optimization: a medicinal chemistry
[3] DiMasi, J.A.; Hansen, R.W.; Grabowski, H.G. The price of perspective. Curr. Opin. Drug. Discov. Dev., 2008, 11, 559-568.
innovation: new estimates of drug development costs. J. Health [30] Rollinger, J.M.; Stuppner, H.; Langer, T. Virtual screening for the
Econom., 2003, 22, 151-185. discovery of bioactive natural products. Prog. Drug. Res., 2008, 65,
[4] Dickson, M.; Gagnon, J. P. Key factors in the rising cost of new 213-249.
drug discovery and development. Nat. Rev. Drug Discov., 2004, 3, [31] Walters, W.P.; Stahl, M.T.; Murcko, M.A. Virtual screening - an
417-429. overview. Drug Discov. Today, 1998, 3, 160-178.
[5] Adams, C.P.; Brantner, V.V. Estimating the cost of new drug [32] McInnes, C. Virtual screening strategies in drug discovery. Curr.
development: is it really $802 million? Health Affairs, 2006, 25, Opin. Chem. Biol., 2007, 11, 494-502.
420-428. [33] Sun, H. Pharmacophore-based virtual screening. Curr. Med. Chem.,
[6] DiMasi, J.A.; Hansen, R.W.; Grabowski, H.G. The price of 2008, 15, 1018-1024.
innovation: new estimates of drug development costs. J. Health [34] Kroemer, R.T. Structure-based drug design: docking and scoring.
Econ., 2003, 22, 151-185. Curr. Protein Pept. Sci., 2007, 8, 312-328.
[7] Rawlins, M.D. Cutting the cost of drug development. Nat. Rev. [35] Cavasotto, C.N.; Orry, A.J. Ligand docking and structure-based
Drug Discov., 2004, 3, 360-364. virtual screening in drug discovery. Curr. Top. Med. Chem., 2007,
[8] Brooijmans, N.; Kuntz, I.D. Molecular recognition and docking 7, 1006-1014.
algorithms. Annu. Rev. Biophys. Biomol. Struc., 2003, 32, 335-373. [36] Krovat, E.M.; Steindl, T.; Langer, T. Recent advances in docking
[9] Whittle, P.; Blundell, T. Protein structure-based drug design. Annu. and scoring. Curr. Comput.-Aided Drug Des., 2005, 1, 93-102.
Rev. Biophys. Biomol. Struct., 1994, 23, 349-375. [37] Available at: http://accelrys.com/products/discovery-studio/referen
[10] Gupta, S.P. QSAR studies on enzyme inhibitors. Chem. Rev., 1987, ces.html [Accessed: 06. 04. 09].
87, 1183-1253. [38] Güner, O.F.; Clement, O.; Kurogi, Y. Pharmacophore modeling
[11] Gupta, S.P.; Scient, J. QSAR in design of bioactive compounds. and three dimensional database searching for drug design using
Indus. Res., 1991, 50, 301-310. catalyst: recent advances. Curr. Med. Chem., 2004, 11, 2991-3005.
[12] Güner, O.F. History and evolution of the pharmacophore concept in [39] Hoffren, A.M.; Murray, C.M.; Hoffman, R.D. Structure based
computer aided drug design. Curr. Top. Med. Chem., 2002, 2, focusing using pharmacophores derived from the active site of 17b-
1321-1332.
Pharmacophore Based Drug Design Approach
hydroxysteroid dehydrogenase. Curr. Pharm. Des., 2001, 7, 547-
566.
[40] MOE; Chemical Computing Group: Montreal, Quebec, Canada,
2006.
[41] Wolber, G.; Dornhofer, A.; Langer, T. Efficient overlay of small
molecules using 3- D pharmacophores. J. Comput. - Aided Mol.
Design, 2006, 20, 773- 788.
[42] Wolber, G.; Langer, T. LigandScout: 3- D pharmacophores derived
from protein - bound ligands and their use as virtual screening
filters. J. Chem. Inf. Model., 2006, 45, 60- 69.
[43] Dixon, S.L.; Smondyrev, A.M.; Knoll, E.H.; Rao, S.N.; Shaw,
D.E.; Friesner, R.A. PHASE: a new engine for pharmacophore
perception, 3D QSAR model development, and 3D database
screening: 1. Methodology and preliminary results. J. Comput.
Aided Mol. Des., 2006, 20, 647-671.
[44] Schneider, W.; Neidhart, T.; Giller, G. Schnmid, "Scaffold-
Hopping" by topological pharmacophore search: a contribution to
virtual screening. Angew. Chem. Int. Ed. Engl., 1999, 38, 2894-
2896.
[45] Irwin, J.J; Shoichet, B.K. ZINC--a free database of commercially
available compounds for virtual screening. J. Chem. Inf. Model.,
2005, 45, 177-182.
[46] Deepangi, P.; So, S.; Sun, H. Enhancing specificity and sensitivity
of pharmacophore-based virtual screening by incorporating
chemical and shape features: a case study of HIV protease
inhibitors. J. Chem. Inf. Model., 2006, 46, 1236-1244.
[47] Goodnow, R., Jr. Small molecule lead generation processes for
drug discovery. Drugs Fut., 2002, 27, 1165.
[48] Barreca, M.L.; De Luca, L.; Iraci, N.; Rao, A.; Ferro, S.; Maga, G.;
Chimirri, A. Structure-based pharmacophore identification of new
chemical scaffolds as non-nucleoside reverse transcriptase
inhibitors. J. Chem. Inf. Model., 2007, 47, 557-562.
[49] Kastenholz, M.A.; Pastor, M.; Cruciani, G.; Haaksma, E.E.J.; Fox,
T. GRID/CPCA: A new computational tool to design selective
ligands. J. Med. Chem., 2000, 43, 3033-3044.
[50] Lauri, G.; Bartlett, P.A.; CAVEAT: a program to facilitate the
design of organic molecules. J. Comp. Aided Mol. Des., 1994, 8,
51-66.
[51] Jones, G.; Willett, P.; Glen, R.C. A genetic algorithm for flexible
molecular overlay and pharmacophore elucidation. J. Comp. Aided
Mol. Des., 1995, 9, 532-549.
[52] Li, H.; Sutter, J.; Hoffmann, R. Pharmacophore preception,
development and use in drug design, International University Line,
La Jolla, California, 2000, pp. 171-189.
[53] Evans, D.A.; Doman, T.N.; Thorner, D.A.; Bodkin, M.J. 3D QSAR
methods: phase and catalyst compared. J. Chem. Inf. Model., 2007,
47, 1248-1257.
[54] Langer, T.; Hoffmann, R.D. On the use of chemical function-based
alignments as input for 3D-QSAR. J. Chem. Inf. Comput. Sci.,
1998, 38, 325-330.
[55] Berman, H.M.; Westbrook, J.; Feng, Z.; Gilliland, G.; Bhat, T.N.;
Weissig, H.; Shindyalov, I.N.; Bourne, P.E. The protein data bank.
Nucleic Acids Res., 2000, 28, 235-242.
[56] Hoffren, A.M.; Murray, C.M.; Hoffman, R.D. Structure-based
focusing using pharmacophores derived from the active site of
17beta-hydroxysteroid dehydrogenase. Curr. Pharm. Des., 2001, 7,
547-566.
[57] Makela, S.; Poutanen, M.; Lehtimaki, J.; Kostian, M.L.; Santti, R.;
Vihko, R. Estrogen-specific 17beta-hydroxysteroid oxidoreductase
type 1 (E.C. 1.1.1.62) as a possible target for the action of
phytoestrogens. Proc. Soc. Exp. Biol. Med., 1995, 208, 51-59.
[58] Makela, S.; Poutanen, M.; Kostian, M.L.; Lehtimaki, J.; Strauss, L.;
Santti, R.; Vihko, R. Inhibition of 17beta-hydroxysteroid
oxidoreductase by flavonoids in breast and prostate cancer cells.
Proc. Soc. Exp. Biol. Med., 1998, 217, 310-316.
[59] Poutanen, M.; Isomaa, V.; Peltoketo, H.; Vihko, R. Regulation of
estrogen action: role of 17beta-hydroxysteroid dehydrogenases.
Ann. Med., 1995, 27, 675-682.
[60] Renner, S.; Hechenberger, M.; Noeske, T.; B_cker, A.; Jatzke, C.;
Schmuker, M.; Parsons, C.G.; Weil, T.; Schneider, G. Searching
for drug scaffolds with 3D pharmacophores and neural network
ensembles. Angew. Chem. Int. Ed., 2007, 46, 5336-5339.
[61] Putta, S.; Landrum, G.A.; Penzotti, J.E. Conformation mining: an
algorithm for finding biologically relevant conformations. J. Med.
Chem., 2005, 48, 3313-3318.
[62] Available at: http://www.accelrys.com/products/catalyst/
Current Computer-Aided Drug Design, 2010, Vol. 6, No. 1 47
[63] Smellie, A.; Teig, S.L. Poling: promoting conformational variation.
J. Comp. Chem., 1995, 16, 171-187.
[64] Smellie, A.; Kahn S.D.; Teig, S.L. Analysis of conformational
coverage. 1. Validation and estimation of coverage. J. Chem. Inf.
Comp. Sci., 1995, 35, 285-294.
[65] Smellie, A.; Kahn, S.D.; Teig, S.L. Analysis of conformational
coverage. 2. Application of conformational models. J. Chem. Inf.
Comp. Sci., 1995, 35, 295-304.
[66] Brooks, B.R.; Brucolleri, R.E.; Olafson, B.D.; States, D.J.;
Swaminathan, S.; Karplus, M. A program for macromolecular
energy, minimization, and dynamics calculations. J. Comput.
Chem., 1983, 4, 187-217.
[67] Agrafiotis, D.K.; Gibbs, A.C.; Zhu, F.; Izrailev, S.; Martin, E.
Conformational sampling of bioactive molecules: a comparative
study. J. Chem. Inf. Model., 2007, 47, 1067-1086.
[68] Kirchmair, J.; Laggner, C.; Wolber, G.; Langer, T. Comparative
analysis of protein-bound ligand conformations with respect to
catalyst’s conformational space subsampling algorithms. J. Chem.
Inform. Mod., 2005, 45, 422.
[69] Steindl, T.M.; Schuster, D.; Laggner, C.; Langer, T. Parallel
screening: a novel concept in pharmacophore modeling and virtual
screening. J. Chem. Inf. Model., 2006, 46, 2146-2157.
[70] Available at: http://accelrys.com/industry/pharma-biotech/lead-
discovery/structure-based design.html
[71] Grembecka, J.; Mucha, A.; Cierpicki, T.; Kafarski, P. The most
potent organophosphorus inhibitors of leucine aminopeptidase.
structure-based design, chemistry, and activity. J. Med. Chem.,
2003, 46, 2641-2655.
[72] Ji, H.; Zhang, W.; Zhang, M.; Kudo, M.; Aoyama, Y.; Yoshida, Y.;
Sheng, C.; Song, Y.; Yang, S.; Zhou, Y.; Lu, J.; Zhu, J. Structure-
based de novo design, synthesis, and biological evaluation of non-
azole inhibitors specific for lanosterol 14r-demethylase of fungi. J.
Med. Chem., 2003, 46, 474-485.
[73] Jorgensen, F.S.; Christensen, I.T.; Johansen, B.N.; Terp, G.E. A
new concept for multidimensional selection of ligand
conformations (multiselect) and multidimensional scoring
(multiscore) of protein-ligand binding affinities. J. Med. Chem.,
2001, 44, 2333-2343.
[74] Kubinyi, H. Chance favors the prepared mind-from serendipity to
rational drug design. J. Recept. Signal Transduc., 1999, 19, 15-35.
[75] Muegge, I.; Martin, Y.C. A general and fast scoring function for
protein-ligand interactions: a simplified potential approach. J. Med.
Chem., 1999, 42, 791-804.
[76] Böhm, H.J. Prediction of binding constants of protein ligands: a
fast method for the prioritization of hits obtained from de novo
design or 3D database search programs. J. Comput. Aided Mol.
Des., 1998, 4, 309-323.
[77] Böhm, H.J. Towards the automatic design of synthetically
accessible protein ligands: peptides, amides and peptidomimetics.
J. Comput. Aided Mol. Des., 1996, 4, 265-272.
[78] Böhm, H.J. On the use of ludi to search the fine chemicals
directory for ligands of proteins of known three-dimensional
structure. J. Comput. Aided Mol. Des., 1994, 5, 623-632.
[79] Böhm, H.J. The development of a simple empirical scoring
function to estimate the binding constant for a protein-ligand
complex of known three-dimensional structure. J. Comput. Aided
Mol. Des., 1994, 3, 243-256.
[80] Böhm, H.J. A novel computational tool for automated structure-
based drug design. J. Mol. Recognit., 1993, 3, 131-137.
[81] Böhm, H.J. Ludi: rule-based automatic design of new substituents
for enzyme inhibitor leads. J. Comput. Aided Mol. Des., 1992, 6,
593-606.
[82] Taha, M.O.; AlDamen, M.A. Effects of variable docking conditions
and scoring functions on corresponding protein-aligned
comparative molecular field analysis models constructed from
diverse human protein tyrosine phosphatase 1B inhibitors. J. Med.
Chem., 2005, 48, 8016-8034.
[83] Taha, M.O.; Qandil, A.M.; Zaki, D.D.; AlDamen, M.A. Ligand-
based assessment of factor Xa binding site flexibility via elaborate
pharmacophore exploration and genetic algorithm-based QSAR
modeling. Eur. J. Med. Chem., 2005, 40, 701-727.
[84] Triballeau, N.; Acher, F.; Brabet, I.; Pin, J.P.; Bertrand, H.O.
Virtual screening workflow development guided by the 'receiver
operating characteristic' curve approach. application to high-
throughput docking on metabotropic glutamate receptor subtype 4.
J. Med. Chem., 2005, 48, 2534-2547.
48 Current Computer-Aided Drug Design, 2010, Vol. 6, No. 1
[85] Kontoyianni, M.; McClellan, L.M.; Sokol, G.S. Evaluation of
docking performance: comparative data on docking algorithms. J.
Med. Chem., 2004, 47, 558-565.
[86] Kroemer, R.T.; Vulpetti, A.; McDonald, J.J.; Rohrer, D.C.; Trosset,
J.Y. Assessment of docking poses: interactions-based accuracy
classification (IBAC) vs crystal structure deviations. J. Chem. Inf.
Comput. Sci., 2004, 44, 871-881.
[87] Asano, T.; Yoshikawa, T.; Usui, T.; Yamamoto, H.; Yamamoto, Y.
Benzamides and benzamidines as specific inhibitors of epidermal
growth factor receptor and v-Src protein tyrosine kinases. Bioorg.
Med. Chem., 2004, 12, 3529-3542.
[88] Gouldson, P.R.; Kidley, N.J.; Bywater, R.P.; Psaroudakis, G.;
Brooks, H.D. Toward the active conformations of rhodopsin and
the ß2-adrenergic receptor. Proteins, 2004, 56, 67-84.
[89] Krovat, E.M.; Langer, T. Impact of scoring functions on
enrichment in docking-based virtual screening: an application study
on renin inhibitors. J. Chem. Inf. Comput. Sci., 2004, 44, 1123-
1129.
[90] Varady, J.; Wu, X.; Fang, X.; Min, J.; Hu, Z. Molecular modeling
of the three-dimensional structure of dopamine 3 (D3) subtype
receptor: discovery of novel and potent D3 ligands through a
hybrid pharmacophore- and structure-based database searching
approach. J. Med. Chem., 2003, 46, 4377-4392.
[91] Wang, R.; Lu, Y.; Wang, S. Comparative evaluation of 11 scoring
functions for molecular docking. J. Med. Chem., 2003, 46, 2287-
2303.
[92] Venkatachalam, C.M.; Jiang, X.; Oldfield, T.; Waldman, M.
LigandFit: a novel method for the shape-directed rapid docking of
ligands to protein active sites. J. Mol. Graph. Model., 2003, 21,
289-307.
[93] Bertrand, H.O.; Bessis, A.S.; Pin, J.P.; Acher, F.C. Common and
selective molecular determinants involved in metabotopic
glutamate receptor agonist activity. J. Med. Chem., 2002, 45, 3171-
3183.
[94] Xu, R.X.; Hassell, A.M.; Vanderwall, D.; Lambert, M.H.; Holmes,
W.D. Atomic structure of PDE4: insights into phosphodiesterase
mechanism and specificity. Science, 2000, 288, 1822-1825.
[95] Hahn, M. Three-dimensional shape-based searching of
conformationally flexible compounds. J. Chem. Inf. Comput. Sci.,
1997, 37, 80-86.
[96] Brown, R.D.; Martin, Y.C. Use of structure-activity data to
compare structure-based clustering methods for use in compound
selection. J. Chem. Inf. Comput. Sci., 1996, 36, 572-84.
[97] Brown, R.D.; Martin, Y.C. The information content of 2D and 3D
structural descriptors relevant to ligand-receptor binding. J. Chem.
Inf. Comput. Sci., 1997, 37, 1-9.
[98] Singh, J.; Chuaqui, C.E.; Boriack-Sjodin, P.A.; Lee, W.; Pontz, T.;
Corbley, M.J.; Cheung, H.K.; Arduini, R.M.; Mead, J.N.; Newman,
M.N.; Papadatos, J.L.; Bowes, S.; Josiah, S.; Ling, L.E.
Corrigendum to “successful shape-based virtual screening: the
discovery of a potent inhibitor of the type I TGF receptor kinase.
Bioorg. Med. Chem. Lett., 2003, 13, 4355-4359.
[99] Singh, J.; Ling, L.E.; Sawyer, J.S.; Lee, W.; Zhang, F.; Yingling,
J.M. Transforming the TGF beta pathway: convergence of distinct
lead generation strategies on a novel kinase pharmacophore for
TbetaRI (ALK5). Curr. Opin. Drug Discov. Deve., 2004, 7, 437-
445.
[100] Wei, J.; Wang, S.; Gao, S.; Dai, X.; Gao, Q. 3D-pharmacophore
models for selective A2A and A2B adenosine receptor antagonists.
J. Chem. Inf. Model., 2007, 47, 613-625.
[101] Ye, Y.; Wei, J.; Dai, X. Gao, Q. Computational studies of the
binding modes of A2A adenosine receptor antagonists. Amino
Acids, 2008, 35, 389-396.
[102] Wei, J.; Yu, W.; Gao, Q. Advances in the study of A2B adenosine
receptor antagonists. Acta Pharm. Sin., 2008, 43, 2-7.
[103] Dong, A.; Huo, J.; Gao, Q.; Zhao, K.; Wei, J. A three-dimensional
pharmacophore model for RXRa agonists. J. Mol. Struct., 2009,
920, 252-263.
[104] Dong, A.; Wei, J.; Gao, Q. 3D-pharmacophore model for RXRg
agonists. Neurochem. Int., 2009, 54, 286-291.
[105] Zhu, Y.; Lei, M.; Lu, A.; Zhao, X.; Yin, X.; Gao, Q. 3D-QSAR
studies of boron-containing dipeptides as proteasome inhibitors
with CoMFA and CoMSIA methods. Eur. J. Med. Chem., 2009, 44,
1486-1499.
[106] Ebalunode, J.O.; Ouyang, Z.; Liang, J.; Zheng, W. Novel approach
to structure-based pharmacophore search using computational
Gao et al.
geometry and shape matching techniques. J. Chem. Inf. Model.,
2008, 48, 889-901.
[107] Markt, P.; Feldmann, C.; Rollinger, J.M.; Raduner, S.; Schuster,
D.; Kirchmair, J.; Distinto, S.; Spitzer, G.M.; Wolber, G.; Laggner,
C.; Altmann, K.; Langer, T.; Gertsch, J. Discovery of novel CB2
receptor ligands by a pharmacophore-based virtual screening
workflow. J. Med. Chem., 2009, 52, 369-378.
[108] Thomson Scientific. Derwent World Drug Index. London, UK:
Derwent Publications Ltd., 2003.
[109] Jacobsson, M.; Liden, P.; Stjernschantz, E.; Bostroem, H.;
Norinder, U. Improving structure-based virtual screening by
multivariate analysis of scoring data. J. Med. Chem., 2003, 46,
5781-5789.
[110] Diller, D.J.; Li, R.K. Homology models, and high throughput
docking. J. Med. Chem., 2003, 46, 4638-4647.
[111] Ward, Y.D.; Thomson, D.S.; Frye, L.L.; Cywin, C.L.; Morwick, T.;
Emmanuel, M.J.; Zindell, R.; McNeil, D.; Bekkali, Y.; Giradot, M.;
Hrapchak, M.; DeTuri, M.; Crane, K.; White, D.; Pav, S.; Wang,
Y.; Hao, M.-H.; Grygon, C.A.; Labadia, M.E.; Freeman, D.M.;
Davidson, W.; Hopkins, J.L.; Brown, M.L.; Spero, D.M. Design
and synthesis of dipeptide nitriles as reversible and potent
cathepsin S inhibitors. J. Med. Chem., 2002, 45, 5471-5482.
[112] Jensen, A.A.; Begum, N.; Vogensen, S.B.; Knapp, K.M.;
Gundertofte, K.; Dzyuba, S.V.; Ishii, H.; Nakanishi, K.;
Kristiansen, U.; Strømgaard, K. Probing the pharmacophore of
ginkgolides as glycine receptor antagonists. J. Med. Chem., 2007,
50, 1610-1617.
[113] Jaracz, S.; Nakanishi, K.; Jensen, A.A.; Strømgaard, K.
Ginkgolides and glycine receptors: a structure-activity relationship
study. Chem. Eur. J., 2004, 10, 1507-1518.
[114] Michaux, C.; Leval, X.; Jule´mont, F.; Dogne, J.; Pirotte, B.;
Durant, F. Structure-based pharmacophore of COX-2 selective
inhibitors and identification of original lead compounds from 3D
database searching method. Eur. J. Med. Chem., 2006, 41, 1446-
1455.
[115] Greenidge, P.A.; Carlsson, B.; Bladh, L.G.; Gillner, M.
Pharmacophores incorporating numerous excluded volumes
defined by X-ray crystallographic structure in three-dimensional
database searching: application to the thyroid hormone receptor. J.
Med. Chem., 1998, 41, 2503-2512.
[116] Lipinski, C.A.; Lombardo, F.; Dominy, B.W.; Feeney, P.J.
Experimental and computational approaches to estimate solubility
and permeability in drug discovery and development settings. Adv.
Drug Deliv. Rev., 2001, 46, 3-26.
[117] Veber, D.F.; Johnson, S.R.; Cheng, H.Y.; Smith, B.R.; Ward,
K.W.; Kopple, K.D. Molecular properties that influence the oral
bioavailability of drug candidates. J. Med. Chem., 2002, 45, 2615-
2623.
[118] López-Rodríguez, M.L.; Porras, E.; Benhamú, B.; Ramos, J.A.;
Morcillo, M.J.; Lavandera, J.L. First pharmacophoric hypothesis
for 5-HT7 antagonism. Bioorg. Med. Chem. Lett., 2000, 10, 1097-
1100.
[119] Kurogi, Y.; Miyata, K.; Okamura, T.; Hashimoto, K.; Tsutsumi, K.;
Nasu, M.; Moriyasu, M. Discovery of novel mesangial cell
proliferation inhibitors using a three-dimensional database
searching method. J. Med. Chem., 2001, 44, 2304-2307.
[120] Chen, G.S.; Chang, C.S.; Kan, W.M.; Chang, C.L.; Wang, K.C.;
Chern, J.W. Novel lead generation through hypothetical
pharmacophore three-dimensional database searching: discovery of
isoflavonoids as nonsteroidal inhibitors of rat 5
-reductase. J. Med.
Chem., 2001, 44, 3759-3763.
[121] Zhu, L.L.; Hou, T.J.; Chen, L.R.; Xu, X.J. 3D QSAR analyses of
novel tyrosine kinase inhibitors based on pharmacophore
alignment. J. Chem. Inf. Comput. Sci., 2001, 41, 1032-1040.
[122] Saladino, R.; Crestini, C.; Palamara, A.T.; Danti, M.C.; Manetti, F.;
Corelli, F.; Garaci, E.; Botta, M. Synthesis, biological evaluation,
and pharmacophore generation of uracil, 4(3H)-pyrimidinone, and
uridine derivatives as potent and selective inhibitors of
parainfluenza 1 (Sendai) virus. J. Med. Chem., 2001, 44, 4554-
4562
[123] Palomer, A.; Cabré, F.; Pascual, J.; Campos, J.; Trujillo, M.A.;
Entrena, A.; Gallo, M.A.; García, L.; Mauleón, D.; Espinosa, A.
dentification of novel cyclooxygenase-2 selective inhibitors using
pharmacophore models. J. Med. Chem., 2002, 45, 1402-1411.
Pharmacophore Based Drug Design Approach
[124] Debnath, A.K. Pharmacophore Mapping of a Series of 2,4-
Diamino-5-deazapteridine inhibitors of mycobacterium avium
complex dihydrofolate reductase. J. Med. Chem., 2002, 45, 41-53.
[125] Singh, J.; Vlijmen, H.; Liao, Y.; Lee, W.; Cornebise, M.; Harris,
M.; Shu, I.; Gill, A.; Cuervo, J.H.; Abraham, W.M. Identification
of potent and novel a4b1 antagonists using in silico screening. J.
Med. Chem., 2002, 45, 2988-2993.
[126] Clement, O.O.; Freeman, C.M.; Hartmann, R.W.; Handratta, V.D.;
Vasaitis, T.S.; Brodie, A.M.H.; Njar, V.C.O. Three dimensional
pharmacophore modeling of human CYP17 inhibitors. potential
Received: June 23, 2009 Revised: September 2, 2009
Current Computer-Aided Drug Design, 2010, Vol. 6, No. 1 49
agents for prostate cancer therapy. J. Med. Chem., 2003, 46, 2345-
2351.
[127] Varady, J.; Wu, X.; Fang, X.; Min, J.; Hu, Z.; Levant, B.; Wang, S.
Molecular modeling of the three-dimensional structure of
dopamine 3 (D3) subtype receptor: discovery of novel and potent
D3 ligands through a hybrid pharmacophore- and structure-based
database searching approach. J. Med. Chem., 2003, 46, 4377-4392.
[128] Kraemer, O.; Hazemann, I.; Podjarny, A.D.; Klebe, G. Virtual
screening for inhibitors of human aldose reductase. Proteins, 2004,
55, 814-823.
Accepted: September 25, 2009