Quoc-Tuan Do1

CØcile Lamy2
Isabelle Renimel2
Nancy Sauvan2
Patrice AndrØ2 Reverse Pharmacognosy: Identifying Biological
Franck Himbert1
Luc Morin-Allory3
Properties for Plants by Means of their Molecule
Philippe Bernard1 Constituents: Application to Meranzin

Abstract
Reverse pharmacognosy aims at finding biological targets for
natural compounds by virtual or real screening and identifying
natural resources that contain the active molecules. We report
Original Paper
by an extract containing this compound in a suitable concentra-
tion. These results demonstrate that reverse pharmacognosy and
its inverse docking component is a powerful tool to identify bio-
logical properties for natural molecules and hence for plants
containing these compounds.

Downloaded by: Jadhavpur University. Copyrighted material.

herein a study focused on the identification of biological proper-
ties of meranzin, a major component isolated from Limnocitrus Key words
littoralis (Miq.) Swingle. Selnergy, an in silico biological profil- Inverse docking ´ Limnocitrus littoralis ´ Rutaceae ´ cyclooxygen-
ing software, was used to identify putative binding targets of ase ´ meranzin ´ Selnergy ´ reverse pharmacognosy
meranzin. Among the 400 screened proteins, 3 targets were se-
lected: COX1, COX2 and PPARg. Binding tests were realised for Supporting information available online at
these 3 protein candidates, as well as two negative controls. The http://www.thieme-connect.de/ejournals/toc/plantamedica
predictions made by Selnergy were consistent with the experi-
mental results, meaning that these 3 targets can be modulated

1235

Introduction products, we have introduced and explored the usefulness of the

Pharmacognosy is the study of the pharmacochemistry of natur-
al raw materials, mainly, but not exclusively extracted from
plants [1], for pharmaceutical, dietary and cosmetic purposes. It
leads to bioactive molecules after extraction, purification, char-
acterisation and bioassays.
After more than a century of research in pharmacognosy, a huge
amount of data has been acquired on plants and on their mol-
ecules. Nevertheless, in most of the cases, little is known about
the biological properties of these natural products. In order to ex-
ploit these data and to find biological properties for these natural
concept of ªreverse pharmacognosyº (RPn) [2].
RPn is similar to ªreverse pharmacologyº [3] as small molecules
are used as probes to evaluate their effects on a biological sys-
tem, but differs from reverse pharmacology by its final goal. RPn
aims at finding applications for natural molecules and their sour-
ces (mostly plants), but does not focus only on identifying new
targets or new biological routes. RPn allies chemoinformatics
tools and traditional knowledge in the search for plants aimed
at the development of botanicals, pharmaceuticals or cosmetics.
In a first step, the biological properties of a molecule are
screened in silico and/or in vitro, then, in a second step, it is pos-

Affiliation
1
GREENPHARMA S. A., OrlØans, France
2
GIE LVMH RECHERCHE, 45804 Saint Jean de Braye cedex, France
3
Institut de Chimie Organique et Analytique, OrlØans, France

Correspondence
Dr Philippe Bernard ´ GREENPHARMA S. A. ´ 3 allØe du Titane ´ 45100 OrlØans ´ France ´
Phone: +33-2-3825-9980 ´ Fax: +33-2-3825-9965 ´ E-mail: philippe.bernard@greenpharma.com

Received June 22, 2007 ´ Revised July 16, 2007 ´ Accepted July 19, 2007
Bibliography
Planta Med 2007; 73: 1235±1240  Georg Thieme Verlag KG Stuttgart ´ New York
DOI 10.1055/s-2007-990216 ´ Published online September 13, 2007
ISSN 0032-0943
 sible to retrieve the plants containing the active compounds
thanks to a plant/molecule relational database. Thus, reverse
pharmacognosy is complementary to pharmacognosy.
Two key components are essential for implementing reverse
pharmacognosy: (i) a virtual screening tool such as Selnergy or
a real screening platform; (ii) a database linking plants and mol-
ecules. Selnergy is an inverse docking tool and consists in the as-
sociation of a docking sofware and a 3-dimensional (3D) protein
target database. Instead of searching for compounds that can
bind to a specific protein, Selnergy will identify putative targets
that can interact with a molecule. Virtual screening (or docking
compound database) is now a well-established method that ap-
peared more than ten years ago [4]. It has also been applied suc-
Original Paper
cessfully to natural compounds [5], [6], [7]. The potential of in-
verse docking was evaluated in 2001 by Chen et al. [8]. These au-
thors used Dock [4] as the docking tool. This software unfortu-
nately suffers from a lack of speed and accuracy so is not suitable
for a systematic screening of a huge target database. Similar
works have been published that used other docking tools [9]. Sel-
nergy [10] has been developed with FlexX that is a balance be-
tween accuracy and speed. Moreover, we brought particular at-
tention to the quality of the protein models we have included
into our target database by implementing a selection procedure
(see Materials and Methods). A visual analysis is finally applied
to the best hits from Selnergy to discard false positives.
Herein we report the study of meranzin (Fig. 1), a coumarin deri-
vative characterised by an epoxide group. The goal is to find an
application for this product and consequently an application for
its source. Meranzin was isolated from Limnocitrus littoralis
(Miq.) Swingle, Rutaceae. Little is known about the biological
1236 properties of meranzin. As this molecule can be purified in huge
quantities from a readily available sources, a rapid industrialisa-
tion can be expected with a possible sustainable development
with local population.
Materials and Methods
Virtual screening
1. Selnergy implementation: Selnergy is implemented in the Sy-
byl 6.9 package [11] in SPL language and consists of a target li-
brary from crystallography (structures retrieved from Protein
Data Bank [12] http://www.rcsb.org/pdb/º or from homology
modelling. Selnergy is based on FlexX [13]. Homology modelling
was performed with Biopolymer and Composer modules within
the Sybyl package. A protein may be represented with several 3D
structures (if available) in order to reflect protein flexibilities. So
far, we have more than 400 proteins and 300 olfactory receptor
structures. A target is considered as ªvalidatedº and added to
Fig. 1 The structure of meranzin.
Do Q-T et al. Reverse Pharmacognosy: Identifying ¼ Planta Med 2007; 73: 1235 ± 1240
the target library if it fulfills the following conditions. From a set
of molecules consisting of active ligands known for their target
affinity and molecules randomly selected, the virtual screening
on this target will rank the actives among the best scored struc-
tures [2].
2. Molecule drawing: Meranzin was sketched in Sybyl. The three-
dimensional structure was calculated with Concord 4.0.4 [14]
and minimised with Tripos force field with 25 Simplex [15] steps
coupled with Powell method [16]. The maximum number of
iterations was set to 1000 with a termination gradient of 0.05
kcal ´ mol±1´ Š±1.
3. Docking of meranzin: Meranzin was considered flexible in the
docking procedure. A rigid core with a maximum of interaction
groups (e. g., donors and acceptors of hydrogen bonds) is first se-
lected, then placed inside the active site of a protein. After that,
the flexible parts of the molecule are iteratively grown with re-
spect to steric and electrostatic constraints of the protein pocket
[13]. Dissimilar molecule poses are scored and ranked. By de-
fault, the 30 best scored placements are retained.
Downloaded by: Jadhavpur University. Copyrighted material.
4. Hit selection procedure: A spatial fitting is taken into account to
determine whether meranzin is ªcorrectlyº docked into a protein
active site. This is achieved by comparing the distance between
the centroid defined in the active site of the studied protein and
the centroid defined in meranzin (distance centroid to centroid:
dC-C). A meranzin placement with a dC-C > 4 Š is considered out of
the protein active site, and the protein/ligand pair is discarded
from further analysis. As Selnergy can propose several poses for
the same molecule, the placement with the shortest dC-C distance
is selected. A macro written in SPL and distributed by Tripos was
implemented to automatise this procedure.
Hit identification is based on an estimated interaction energy of
meranzin with a target divided by the number of meranzin
atoms (E1). This energy is referred as ªnormalised interaction en-
ergyº. This normalised energy is then compared with a reference
ligand normalised interaction energy (ligand co-crystallised
with the protein) referred as E2. E1 should be less or equal to E2
for a solution to be accepted. We used the interaction energy nor-
malised by atom number because huge molecules have the tend-
ency to be better scored by docking and scoring tools, thus bias-
ing the results.
Plant material and molecule
Leaves of Limnocitrus littoralis (Miq.) Swingle were collected in
the province of Binh Thuan, Vietnam, in July 2003 and identified
by Prof. Tran Hop of University of Ho Chi Minh City. A voucher
specimen (NCY008913) is deposited at the Conservatory and the
Botanical Gardens of Nancy, France.
The extraction, purification and characterisation of meranzin are
described in the Supporting Information section.
COX-1 and COX-2 binding assays
The COX inhibitor screening assay (Catalogue No. 560131) by
Cayman Chemical (Ann Arbor, MI, USA) was used to identify
COX inhibitors and to assess the selectivity between COX-1 and
COX-2 of the inhibitors. The protocol [17] was modified as de-
scribed in the kit booklet accompanying the product. All meas-
urements were done in duplicates. Table 1 In silico prediction of affinity by Selnergy. ªPDB idº column
contains Protein Data Bank codes that identifies a protein 3D
Elastase binding assays structure
In our study, pancreatic elastase (Sigma Aldrich; Saint-Quentin-
Rank PDB id Targets E1-E2
Fallavier, France) was used. The assay was based on the hydroly-
sis of N-succinyl-L-ala-L-ala-L-ala-p-nitroanilide, a synthetic sub- 1 1eqh Cyclooxygenase 1 ±1.793
strate developed by Bieth et al. [18]. Hydrolysis of N-succinyl-L- 2 1fm9 Peroxisome proliferators activated ±1.178
ala-L-ala-L-ala-p-nitroanilide by elastase at 25 8C and pH 8.0 Rreceptor g
could be followed by reading the absorbance at 410 nm. 3 1rbp Retinol binding protein ±1.017
4 1oyn Phosphodiesterase 4D ±0.816
PPARg cell-based assays 5 1soj Phosphodiesterase 3B ±0.783
HEK293 cells (ACACC; Salisbury, UK) were transiently transfect- 6 4cox Cyclooxygenase 2 ±0.756
ed with the PPAR responsive human CPT-1b promoter-luciferase 7 1qkt Estrogen receptor b ±0.661

Original Paper
reporter plasmid [19] with or without expression vectors har- 8 1ela Elastase ±0.532
9 1cx2 Cyclooxygenase 2 ±0.490
bouring human PPARg. Cells were co-transfected with a b-galac-
10 3pgh Cyclooxygenase 2 ±0.466
tosidase expression vector to correct for possible differences in
11 1hgi Hemagglutinin ±0.421
transfection efficiency. Rosiglitazone (Axxora LLC; San Diego,
12 1hgh Hemagglutinin ±0.411
CA, USA) was used as a reference compound. Luciferase activity
13 1e8z Phosphoinositide 3-kinase gamma ±0.360
was normalised to b-galactosidase activity [19].
14 1acj Acetylcholinesterase ±0.324
15 1hvr HIV-1 protease ±0.310
Prostaglandins production by keratinocytes 16 1m7q p38 Map kinase ±0.222

Downloaded by: Jadhavpur University. Copyrighted material.

Keratinocytes were grown in supplemented keratinocyte serum-
free medium (KSFMc) (Gibco ref : 17 005-034 + 37 000-015;
Grand Island, NY, USA). The keratinocytes were seeded in 96-
well microplates (10,000 cells per well). The first day of culture
was considered as D0. After 24 hours (D1) of incubation, the
medium was replaced by KSFM-c containing meranzin. 24 hours
later, PGE2 released was measured in cultured human keratino-
cytes medium using ELISA tests (R&D Systems, Ref DE0100; Lille,
France). The positive control, indomethacin (Sigma Aldrich) and
meranzin (LVMH Recherche; Saint-Jean-de-Braye, France) were
tested at 2mM.
Cell viability
A cell viability test was conducted with the 2,3-bis(2-methoxy-
4-nitro-5-sulfophenyl)-5-[(phenylamino)carbonyl]-2H-tetrazo-
lium hydroxide (XTT) assay (after 24 hours of treatment) to veri-
fy that the treatments at the concentrations studied were not
toxic for the keratinocytes. The Cell Proliferation Kit II (XTT test)
was purchased from Roche Diagnosis (Mannheim, Germany).
Supporting information
The isolation procedure for meranzin and the physico-chemical
data of this molecule are available as Supporting Information.
Results and Discussions
The pairs of molecule/protein with a negative E1-E2 are selected.
Table 1 lists all the protein structures that fulfill this criterion.
We will focus on the first 10 as these proteins are most likely to
interact with meranzin.
COX1 is the ªbestº scored protein. Interestingly, there are three
structures of COX2 that are ranked within the top 10, in position
6, 9 and 10. This family of enzymes seems to have a high affinity
to meranzin. The second protein selected is the nuclear receptor
PPARg, and the third one the retinal binding protein, a transport
protein. Phosphodiesterases (PDE) come in the second place of
17 4phv HIV-1 protease ±0.177
18 4hmg Hemagglutinin ±0.152
19 1rkp Phosphodiesterase 5A ±0.080
20 1aq1 Cyclin-dependent protein kinase 2 ±0.068
21 1agw Fibroblast growth factor receptor 1 ±0.066
this ranking list with respect to enzyme families. PDE4D and
PDE3B have been well scored and topped at position 4 and 3 (no-
teworthy, PDE5A is on the extended list at position 19). The other 1237
putative targets are oestrogen receptor b and elastase.
Because the number of interaction sites can be high in a protein
active site, docking software can sometimes place a ligand in a
wrong pose. In order to discard false positives, we have to exam-
ine visually the docking results. We will detail below our analysis
on the 10 best pairs in order to make a final ranking according to
knowledge on proteins and their preferred binding modes. The
pro and con arguments are listed. The final selections will be
submitted to experimental validations.
The active site of COX1 is a deep close pocket surrounded mostly
by hydrophobic residues [20]. Meranzin is a hydrophobic mol-
ecule, it is docked into the active site and forms hydrophobic in-
teractions with residues of COX1. The coumarin fragment is locat-
ed in the cleft formed by MET113, VAL116, ALA527, LEU531,
LEU534, VAL349, TYR355 and LEU359 (the residues are num-
bered according to [20]), while the epoxide part is at the vicinity
of PHE381, TYR385, TRP387, TYR348, PHE518, LEU532, GLY526
and SER530. The oxygen of the meranzin epoxide may form a hy-
drogen bond with the hydroxy group of SER530 (Fig. 2). Meran-
zin is docked in a similar fashion in the COX2 active site, except
that no hydrogen bond is detected with SER530 (Fig. 2). By over-
laying COX1 and COX2 active sites and docked meranzin in the
respective pocket as shown in Fig. 2, we find a low RMS value of
1.7Š, which denotes that the poses of meranzin in both sites are
quite similar. The only change in the active sites of the two sub-

Do Q-T et al. Reverse Pharmacognosy: Identifying ¼ Planta Med 2007; 73: 1235 ± 1240
Original Paper
Fig. 2 Superpossition of COX1 (in green) and COX2 (in white). The pla-
cements of merazin are colour-coded as follows: merazin in COX1 in
green and in COX2 in white. The hashed line represents a hydrogen
bond. Residues of interest such as VAL523 and ILE523 are highlighted in
capped sticks. These amino acids are responsible for the selectivity be-
tween the subtypes. Colour codes for atoms: carbon in white or green,
nitrogen in blue, oyxygen in red, sulphur in yellow and hydrogen in cyan.
types of COX is that ILE523 in COX1 is replaced by VAL523 in
COX2. This substitution is enough to confer selectivity [21].
Thus, we cannot discriminate which docking is better in the light
of the placement and the selectivity data. Both subtypes will be
kept in our selection.
Meranzin seems to be well docked into PPARg. The polar group of
the coumarin is positioned close to four residues that form hy-
1238 drogen and/or ionic interactions with the carboxylic group of
some ligands [22]. The carboxyl group of meranzin may form hy-
drogen bonds with SER289, HIS323, HIS449 and TYR473 (num-
bering according to [22]). At the vicinity of these polar residues,
a hydrophobic cleft can be found in which the hydrophobic moi-
ety of meranzin can fit in. An argument in disfavour of meranzin
is the large size of PPARg active site. By considering the co-crys-
tallised ligand GI262570 and meranzin ªbestº docked pose, the
overlapping volume represents 66 % of meranzin's volume and
31 % of that of GI262570. The total volume of meranzin corre-
sponds to 47 % of that of GI262570 (Fig. 3). PPARg should be a
good candidate although meranzin just occupied a small fraction
of the binding pocket.
The third putative protein retained by Selnergy is the retinol
binding protein. The ligands of this transporter are retinoids
[23] that consist of a large apolar part and a polar head. So the
binding cleft is composed of hydrophobic residues. Retinol in
the retinol binding protein is docked with its hydroxy group at
the entrance of the cleft whereas the apolar part is buried into
the site. Meranzin adopts a different binding pattern: the mol-
ecule's polar groups are deeply placed into the pocket. These
data considered, retinol binding protein appears a less likely
binding partner for our compound.
Phosphodiesterases (PDE) were not retained after reviewing the
docking results. The PDE site possesses a conserved glutamine in
orientations which purportedly discriminate between cAMP and
Do Q-T et al. Reverse Pharmacognosy: Identifying ¼ Planta Med 2007; 73: 1235 ± 1240
Fig. 3 Merazin docked in the PPARg active site. It is represented in a
green ball-and-stick fashion. The co-crystallised ligand is represented
in magenta capped sticks. The volumes occupied by both molecules
are highlighted by colour-coded volumes. Important residues of the
protein are displayed in capped sticks. Hashed lines symbolise hydro-
gen bonding. Colour codes for atoms: carbon in white or green, nitro-
gen in blue, oyxygen in red, sulphur in yellow and hydrogen in cyan.
Downloaded by: Jadhavpur University. Copyrighted material.
cGMP. This residue intervenes in the binding of the cyclic nucleo-
tides and inhibitors [24]. In our case, meranzin did not interact
with this crucial residue either in PDE4D or in PDE3B. For that
reason, the affinity of meranzin to both PDE may be question-
able.
In the case of estrogen receptor b, the molecule is situated near
but out of the active site. So we have discarded this target from
our final selection.
In the case of elastase, the co-crystallised ligand shares hydrogen
bonds with VAL224, SER222 and possibly SER203. None of these
interactions are observed for our compound, or other polar inter-
actions [25]. It seems that the predicted binding of our molecule
relies only on the apolar interaction. In the light of these data, we
rejected elastase as a potential target. The rest of the targets
ranked from 11 to 21 were also reviewed and none of them satis-
fied our binding mode analysis.
We now present results from in vitro evaluations for the best
ranked protein partners of meranzin, namely COX1, COX2 and
PPARg. As negative controls, we tested meranzin with elastase
and 5-lipoxygenase (LOX). Some chemical series can bind to
COX and LOX as dual inhibitors [26]. So this enzyme was also ad-
ded to the list of proteins to be evaluated although it was not se-
lected by Selnergy.
Table 2 shows the results of inhibition assays for COX1 and COX2.
The experimental data nicely fit with our prediction for COX2.
Meranzin inhibits COX2 in a dose-dependent manner with an in-
hibition of 56.2 % at 0.4 mM. Therefore the IC50 of meranzin on
COX2 is in a submicromolar range, which is an outstanding activ-
ity for a natural compound. Meranzin is less active on COX1 than
on COX2. Moreover, the inhibition on COX1 is not dose-depen-
dent.
 intended to rank affinities of different molecules on the same
Table 2 Binding assays of meranzin on cyclooxygenase 1, cyclooxy- target.
genase 2 and 5-lipoxygenase. The inhibition values are the
mean of 2 measures To conclude, we have validated in vitro that meranzin can bind to
COX1, COX2 and PPARg as predicted by Selnergy. Moreover, it in-
COX1 COX2 LOX
hibits prostaglandin release in human keratinocytes. Our results
Meranzin % inhibition % inhibition Meranzin % inhibition
(mM) (mM) demonstrate that inverse docking can be useful in finding new
applications for compounds and hence natural sources, e. g.,
0.4 37.1 56.2 1 8.3 plants that contain such compounds. Selnergy can accelerate
4 19.1 71.1 10 ±0.9 this process of known target identification through its indexed
40 64.4 79.5 100 4.9 target library. In the case of meranzin, the prediction was in
good agreement with experimental data in respect to the inter-
action with COX1, PPARg and particularly with COX2. Our in
silico tool is obviously beneficial for decision making in the pro-

In the case of COX, Selnergy was able to detect activity at micro-
molar ranges for meranzin but failed to correctly rank its affini-
ties for COX subtypes. A possible explanation is the structural si-
milarity of the two active sites which differ by a single residue.
We also verified that there is no significant activity of meranzin
on LOX at 100 mM.
Original Paper
cess of finding new applications for molecules and their related
sources. Nevertheless, it cannot predict selectivity for subtypes
with very close active sites such as COX that differs by only one
apolar residue. Moreover one has to bear in mind that simple
models are used here to simulate complex biological processes
so that experimental validations will still be necessary.

We also measured the inhibition of prostaglandin release by hu- With the fast evolution of analytical chemistry, information on

Downloaded by: Jadhavpur University. Copyrighted material.

man keratinocytes. The inhibition rate by meranzin at 2 mM is 25
 0.9 % (n = 4) compared to the reference molecule indometha-
cin at 2 mM, for which the inhibition rate is 17  0.3 % (n = 6).
Therefore, our molecule is slightly more active. Thus meranzin
appears a potent inhibitor of the inflammatory process.
The meranzin/PPARg interaction was validated in a cell-based
assay with transfected PPAR. The reference compound was rosi-
glitazone. Meranzin activates PPARg in a dose-dependent man-
ner (see Table 3). It is active above 10 mM. At 100 mM, its activity
is comparable to that of rosiglitazone at 10 mM.
We verified that meranzin is not active on pancreatic elastase at
a concentration of 1 mM (I% = 2  0.01 %), compared to ellagic
acid (our positive control) with an inhibition rate is 98  15 % at
1 mM on the same test.
the composition of natural material is becoming more and more
accessible in quantity and in quality. Our approach can help to
find new applications for plants based on Western medicinal
concepts of molecules and protein targets, and, more focused on
symptoms. It complements traditional cures which are more fo-
cused on the origin of disease. Bearing in mind that toxicity also
constitutes a crucial issue, we hope, however, that our approach
will widen access to treatments for populations that cannot af-
ford Western medicine drugs.
1239
Acknowledgements
Our warm thanks to Prof. T. Hop and to the Spectropole of Mar-
seille for NMR analysis. Special thanks to Dr. M. van Bilsen and
Dr. A. Gilde from University of Maastricht.

These results demonstrate the ability of Selnergy to select the

most probable protein partners for a molecule within its top References
2 % of the best scored proteins. This reduced number of hits al-
1
lows for human supervisions to retain the most promising can- Bruneton J. Pharmacognosie, Phytochimie, Plantes MØdicinales. Paris:
Lavoisier; 1993: viii.
didate and to discard false positive proteins. Ranking accurately 2
Do QT, Bernard P. Pharmacognosy and reverse pharmacognosy : a new
the affinity proteins and ligands is a difficult challenge. This is concept for accelerating natural drug discovery. IDrugs 2004; 7:
due to the imprecision of current scoring functions which are 1017 ± 27.
3
Harrigan GG, Brackett DJ, Boros LG. Medicinal chemistry, metabolic
profiling and drug target discovery: a role for metabolic profiling in re-
verse pharmacology and chemical genetics. Mini Rev Med Chem
2005; 5: 13 ± 20.
4
Table 3 PPARg cell-based assay. Measures are in triplicates. Activities Shoichet BK, Stroud RM, Santi DV, Kuntz ID, Perry KM. Structure-based
are expressed relative to vehicle control (DMSO) arbitrarily discovery of inhibitors of thymidylate synthase. Science. 1993; 259:
set to 1.0 1445 ± 50.
5
Rollinger JM, Haupt S, Stuppner H, Langer T. Combining ethnopharma-
Meranzin (mM) Activating rate cology and virtual screening for lead structure discovery: COX-inhibi-
tors as application example. J Chem Inf Comput Sci 2004; 44: 480 ± 8.
6
Liu H, Li Y, Song M, Tan X, Cheng F, Zheng S et al. Structure-based dis-
1 1.0  0.1 covery of potassium channel blockers from natural products: virtual
10 1.1  0.0 screening and electrophysiological assay testing. Chem Biol 2003; 10:
100 1.4  0.2 1103 ± 13.
7
DMSO 1.0  0.0 Bernard P, Scior T, Didier B, Hibert M, Berthon JY. Ethnopharmacology
and bioinformatic combination for leads discovery: application to
Rosiglitazone 10 mM 1.6  0.0
phospholipase A(2) inhibitors. Phytochemistry 2001; 58: 865 ± 74.

Do Q-T et al. Reverse Pharmacognosy: Identifying ¼ Planta Med 2007; 73: 1235 ± 1240
 8 19
Chen YZ, Zhi DG. Ligand-protein inverse docking and its potential use
in the computer search of protein targets of a small molecule. Proteins
2001; 43: 217 ± 26.
9
Paul N, Kellenberger E, Bret G, Muller P, Rognan D. Recovering the true
targets of specific ligands by virtual screening of the protein data
bank. Proteins 2004; 54: 671 ± 80.
10
Do QT, Renimel I, Andre P, Lugnier C, Muller CD, Bernard P. Reverse
pharmacognosy: application of selnergy, a new tool for lead discovery.
The example of epsilon-viniferin. Curr Drug Discov Technol 2005; 2:
161 ± 7.
11
Tripos Inc; 1699 South Hanley Rd.; St. Louis, Missouri, 63144, USA.:
12
Berman HM, Westbrook J, Feng Z, Gilliland G, Bhat TN, Weissig H et al.
The protein data bank. Nucleic Acids Res 2000; 28: 235 ± 42.
13
Rarey M, Kramer B, Lengauer T, Klebe G. A fast flexible docking meth-
od using an incremental construction algorithm. J Mol Biol 1996; 261:
470 ± 89.
14
Pearlman RS. CONCORD 4.0.7A, ªConcord User's Manual,º. distribut-
Original Paper
ed by Tripos Inc; St. Louis, MO.:
15
Press WH, Flannery BP, Teukolsky SA, Vetterling WT. Simplex. Numer-
ical recipes in C, the art of scientific computing. Cambridge: University
Press; 1988: 312 ± 27.
16
Powell MJD. An efficient method for finding the minimum of a func-
tion of several variables without calculating derivatives. Comp J
1964; 7: 155 ± 62.
17
Maclouf J, Grassi J, Pradelles P. Development of enzyme-immunoassay
techniques for the measurement of eicosanoids. In: Walden TL,
Hughes HN; editors. Prostaglandin and lipid metabolism in radiation
injury New York: Plenum Press; 1987: 355 ± 64.
18
Bieth J, Spiess B, Wermuth CG. The synthesis and analytical use of a
highly sensitive and convenient substrate of elastase. Biochem Med
1974; 11: 350 ± 7.
1240
Do Q-T et al. Reverse Pharmacognosy: Identifying ¼ Planta Med 2007; 73: 1235 ± 1240
Gilde AJ, van der Lee KA, Willemsen PH, Chinetti G, van der Leij FR, van
der Vusse GJ et al. Peroxisome proliferator-activated receptor (PPAR)
alpha and PPARbeta/delta, but not PPARgamma, modulate the expres-
sion of genes involved in cardiac lipid metabolism. Circ Res 2003; 92:
518 ± 24.
20
Garavito RM, Picot D, Loll PJ. Preliminary X-ray investigations into
NSAID-binding to cyclooxygenase-1. Am J Ther 1995; 2: 611 ± 5.
21
Gierse JK, McDonald JJ, Hauser SD, Rangwala SH, Koboldt CM, Seibert
K. A single amino acid difference between cyclooxygenase-1 (COX-1)
and -2 (COX-2) reverses the selectivity of COX-2 specific inhibitors. J
Biol Chem 1996; 271: 15 810 ± 4.
22
Xu HE, Lambert MH, Montana VG, Plunket KD, Moore LB, Collins JL et
al. Structural determinants of ligand binding selectivity between the
peroxisome proliferator-activated receptors. Proc Natl Acad Sci U S A
2001; 98: 13 919 ± 24.
23
Zanotti G, Berni R. Plasma retinol-binding protein: structure and inter-
actions with retinol, retinoids, and transthyretin. Vitam Horm 2004;
69: 271 ± 95.
24
Zoraghi R, Corbin JD, Francis SH. Phosphodiesterase-5 Gln817 is criti-
cal for cGMP, vardenafil, or sildenafil affinity: its orientation impacts
cGMP but not cAMP affinity. J Biol Chem 2006; 281: 5553 ± 8.
25
Mattos C, Rasmussen B, Ding X, Petsko GA, Ringe D. Analogous inhibi-
tors of elastase do not always bind analogously. Nat Struct Biol 1994;
1: 55 ± 8.
26
Fiorucci S, Meli R, Bucci M, Cirino G. Dual inhibitors of cyclooxygenase
and 5-lipoxygenase. A new avenue in anti-inflammatory therapy? Bio-
chem Pharmacol 2001; 62: 1433 ± 8.
Downloaded by: Jadhavpur University. Copyrighted material.
27
Hito C, Furukawa H. Constituents of Murraya exotica L. Structure elu-
cidation of new coumarins. Chem Pharm Bull 1987; 35: 4277 ± 85.