This article was downloaded by: [Washington University in St Louis]On: 03 May 2013, At: 23:54Publisher: Taylor & FrancisInforma Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,37-41 Mortimer Street, London W1T 3JH, UK

Molecular SimulationPublication details, including instructions for authors and subscription information:http://www.tandfonline.com/loi/gmos20

Development of 3D-pharmacophore model followed bysuccessive virtual screening, molecular docking andADME studies for the design of potent CCR2 antagonistsfor inflammation-driven diseasesRajesh Singh a , Anand Balupuri a & M. Elizabeth Sobhia aa Department of Pharmacoinformatics, National Institute of Pharmaceutical Education andResearch (NIPER), Sector-67, S.A.S. Nagar, Punjab, 160062, IndiaPublished online: 09 Aug 2012.

To cite this article: Rajesh Singh , Anand Balupuri & M. Elizabeth Sobhia (2013): Development of 3D-pharmacophore modelfollowed by successive virtual screening, molecular docking and ADME studies for the design of potent CCR2 antagonists forinflammation-driven diseases, Molecular Simulation, 39:1, 49-58

To link to this article: http://dx.doi.org/10.1080/08927022.2012.701743

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions

This article may be used for research, teaching, and private study purposes. Any substantial or systematicreproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form toanyone is expressly forbidden.

The publisher does not give any warranty express or implied or make any representation that the contentswill be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses shouldbe independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims,proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly inconnection with or arising out of the use of this material.

Development of 3D-pharmacophore model followed by successive virtual screening, moleculardocking and ADME studies for the design of potent CCR2 antagonists for inflammation-drivendiseases

Rajesh Singh, Anand Balupuri and M. Elizabeth Sobhia*

Department of Pharmacoinformatics, National Institute of Pharmaceutical Education and Research (NIPER), Sector-67, S.A.S. Nagar,Punjab 160062, India

(Received 22 February 2012; final version received 6 June 2012)

In order to elucidate the essential structural features for CC chemokine receptor 2 (CCR2) antagonism, 3D-pharmacophorehypotheses were built based on a set of known compounds from the literature. The hypotheses were developed with the aid ofHypoGen module within Discovery Studio 2.5 program. Multiple validation approaches provided the confidence in utilisingthe predictive pharmacophore models developed in this study. The most predictive pharmacophore model (Hypo1) wasfound to be statistically significant along with its ability to predict activities of the known CCR2 antagonists in the trainingand test set with high correlation coefficient. The best model was then used as a 3D search query in the virtual screening ofchemical databases including ChemDiv and MiniMaybridge. Lipinski’s rule of five and molecular docking studies wereapplied to the screened hits for retrieving potential lead compounds. Eight hits showed better in silico CCR2-bindingaffinities than the reported CCR2 antagonists, along with good absorption, distribution, metabolism and excretion profiles.The current 3D-quantitative structure–activity relationship (QSAR) pharmacophore modelling and molecular dockingstudies attempt to elucidate QSAR for CCR2 antagonism and identify novel potent CCR2 antagonist scaffolds.

Keywords: CCR2 antagonists; pharmacophore mapping; virtual screening; ADME analysis; GPCR

1. Introduction

G-protein-coupled receptors (GPCRs) are seven-trans-

membrane receptors, engaged in many sensory functions,

neurotransmission and signal transduction processes. Their

role in important biological pathways and their localisation

on the cell membrane makes them important pharmaco-

logical targets for the development of new therapeutic

agents. Chemokine receptors belong to Class A GPCRs,

which play an important role in several diseased

conditions, viz. cancer [1], asthma [2], HIV [3], infectious

diseases [4] and inflammation [5]. CC chemokine receptor

2 (CCR2) is a crucial target for various inflammation-

driven diseases. Monocyte chemotactic protein-1 (MCP-1)

is a member of the CC chemokine subfamily that binds to

the cell surface of CCR2. Interaction between MCP-1 and

CCR2 has been implicated in the various pathological

conditions such as atherosclerosis, rheumatoid arthritis,

multiple sclerosis, nephritis, organ allograft rejection,

type II diabetes and diabetic complications, chronic

obstructive pulmonary disease, allergic asthma, Hodgkin’s

disease and various carcinomas [6–11]. Inhibition of

MCP-1 binding to CCR2 by an antagonist suppresses the

inflammatory response [12–14].

Many researchers have contributed towards the

synthesis and structure–activity relationships (SARs) of

various classes of CCR2 antagonists. Van Lommen et al.

reported the synthesis of 3,4-disubstituted analogues of

2-mercaptoimidazole as CCR2 antagonists [15]. Pinkerton

et al. reported the synthesis and SAR of diaryl-substituted

pyrazoles as CCR2 antagonists [16]. Carter et al. reported

capped diaminopropionamide–glycine dipeptides as

inhibitors of CCR2 [17]. Some researchers have found

disubstituted cyclohexanes to be active as CCR2

antagonists [18]. Furthermore, there are advancements

made in synthesising and characterising di- and tri-

substituted cyclohexanes as CCR2 antagonists [19].

A large number of synthesised CCR2 antagonists are

reported, but discussing all of them here is not possible.

As the number of synthetic antagonists of CCR2 increases,

elucidation of the SAR of these diverse compounds

becomes important.

Quantitative structure–activity relationship (QSAR),

a ligand-based approach, was successfully explored by

researchers to reveal the structural and chemical features

necessary for various biological activities [20–23], as well

as for CCR2 antagonism [24–26]. Recently, some studies

on the validation of QSAR models were reported [27,28].

ISSN 0892-7022 print/ISSN 1029-0435 online

q 2013 Taylor & Francis

http://dx.doi.org/10.1080/08927022.2012.701743

http://www.tandfonline.com

*Corresponding author. Email: mesophia@niper.ac.in

Molecular Simulation

Vol. 39, No. 1, January 2013, 49–58

Dow

nloa

ded

by [

Was

hing

ton

Uni

vers

ity in

St L

ouis

] at

23:

54 0

3 M

ay 2

013

Pharmacophore modelling has been proved as an

important tool for drug discovery over the last few

decades [29–33]. Considering the fact that CCR2 is a

potential target for inflammation-driven diseases, the

development of a potent CCR2 antagonist becomes

important. In this study, we used a combination of both

ligand- and receptor-based approaches to provide a better

understanding of the design process for CCR2 antagonists.

Firstly, we developed a ligand-based pharmacophore

model for CCR2 antagonists using HypoGen approach

with Discovery Studio 2.5 program [34]. The best

pharmacophore hypothesis was then used for virtual

screening of various commercial databases. This was

followed by molecular docking of the top hits on our in-

house 3D receptor model of CCR2.

2. Materials and methods

2.1 Training set selection and conformationalmodelling

Fifty-nine CCR2 antagonists developed by Bristol-Myers

Squibb were collected from the literature [17–19].

The training set comprising 27 compounds (1–27) was

used to generate the HypoGen hypotheses by considering

structural diversity and wide coverage of activity range.

Binding assay used for the determination of IC50 values

was the same for all the training set compounds. Biological

activity data of the compounds spans over four orders

of the magnitude, from pIC50 values of 9.43 (IC50 ¼

0.37 nM, most potent molecule) to pIC50 values of 4.92

(IC50 ¼ 12,000 nM, least potent molecule). All com-

pounds were built using the Sybyl 7.1 (Tripos, 2005)

molecular modelling package installed on a SGI work-

station running IRIX6.5. Gasteiger–Huckel partial atomic

charges were assigned to the compounds and their

conformational energies were minimised using Powell

method and Tripos force field, with a convergence

criterion for the energy gradient of 0.001 kcal/mol/A

[35]. Structures of the training set compounds are given in

Figure 1. Conformations of all the compounds were

generated by using the ‘Best conformer generation’ with

20.0 kcal/mol as energy cut-off. A maximum number of

255 conformers and an Uncert value of 3.0 were used for

all the compounds, while a default value was applied to

other parameters. Instead of using the lowest energy

conformation of each compound, all the conformational

models for each compound in the training set were used for

pharmacophore hypothesis generation.

2.2 Pharmacophore model generation

The 3D-QSAR pharmacophore generation protocol was

used to generate predictive pharmacophores. It uses the

Catalyst HypoGen algorithm [36] to derive SAR

hypothesis models (pharmacophores) from a set of ligands

with known activity values on a given biological target. 3D-

QSARs differ from typical QSAR methods in descriptors

derived from ligand alignments, or how well the ligands

fit a pharmacophore rather than the molecular features.

Often, the descriptors are concerned with the overall

molecule rather than a single substituent. Quantitative

pharmacophore models were generated by using HypoGen

module within the DS 2.5 program, based on the training

set compounds (1–27, Figure 1). Features, such as

hydrogen-bond acceptor (HBA), hydrophobic feature

(HY), hydrogen-bond donor (HBD) and five excluded

volumes (E-volumes), were included for the pharmaco-

phore generation based on the common features present

in the studied compounds. The statistical parameters

such as cost values, correlation coefficient, total cost and

root-mean-square deviation (RMSD) determine the sig-

nificance of the model. The best model was selected based

on a high correlation coefficient (r), lowest total cost,

highest cost difference and low RMSD values.

2.3 Validation of the generated pharmacophore model

Besides the cost analysis, two validation procedures were

followed, namely, CatScramble method to characterise the

quality of the best hypothesis, Hypo1 and external test set

prediction method. The validity and the predictive

character of Hypo1 were assessed by its capacity for

correct activity prediction of test set molecules. A test set

containing 32 molecules of different activity classes was

analysed. All test set molecules were built, minimised and

their conformations were analysed in a similar manner as

followed for all training set compounds. The final

assessment of the statistical confidence of Hypo1 was

done by CatScramble program from Discovery Studio.

This type of validation helps to check if there is a strong

correlation between the structures and activities.

2.4 Database screening

The validated pharmacophore model, Hypo1, was used as

a search query to retrieve molecules with novel and desired

chemical features from ChemDiv and MiniMaybridge

databases consisting of 186,000 and 2000 compounds,

respectively. Search was carried out using the ‘fast flexible

search’ approach implemented within Discovery Studio.

A total of 25,440 and 155 compounds from ChemDiv and

MiniMaybridge databases, respectively, showed a very

good mapping with the Hypo1. Lipinski’s rule of five

was used to the screened hits to identify compounds that

have desirable or ‘drug-like’ properties. By applying

this filter, the number of hits was reduced to 16,585 and

127 compounds for ChemDiv and MiniMaybridge

databases, respectively. Again, the compounds that had

R. Singh et al.50

Dow

nloa

ded

by [

Was

hing

ton

Uni

vers

ity in

St L

ouis

] at

23:

54 0

3 M

ay 2

013

HypoGen-estimated activity ,5 nM were considered as

the most active compounds. Five hundred and seventy-four

compounds from ChemDiv and 10 compounds from

MiniMaybridge databases satisfied the specified cut-off

values and were qualified for further evaluation.

2.5 Molecular docking

Molecular docking is a structure-based virtual screening

technique that generates and then according to the

calculated binding affinities scores the putative protein–

ligand complexes. It can be successfully used for

Figure 1. Chemical structures of CCR2 antagonists in training set (compounds 1–27).

Molecular Simulation 51

Dow

nloa

ded

by [

Was

hing

ton

Uni

vers

ity in

St L

ouis

] at

23:

54 0

3 M

ay 2

013

identifying active compounds by filtering out those that do

not fit into the receptor-binding site. Therefore, hits

retrieved from ChemDiv and MiniMaybridge databases

were docked into the small molecule-binding site of the

modelled CCR2 by using Glide (Grid-based Ligand

Docking with Energetics) program [37]. Because the

crystal structure of CCR2 is not yet determined, we had

previously developed homology model of human chemo-

kine CCR2 for molecular modelling studies [38]. Recently,

we developed a more refined 3D model of CCR2 using

induced fit docking of the known CCR2 antagonists

(manuscript under revision). This model is capable enough

to distinguish actives from decoys with high enrichment

values. The refined model was then used for the docking

studies.

2.6 Similarity search

Similarity search was carried out to check the novelty of

the screened hits against CCR2 using PubChem search

tool [39]. Partially similar compounds showing more than

or equal to 90% Tanimoto similarity with the screened hits

were subjected to the sequential ligand- and the receptor-

based screening procedures involving the best pharmaco-

phore hypothesis, Hypo1, based virtual screening followed

by the molecular docking studies. All the screening

parameters were the same for both potential hits and for

the partial similar compounds.

3. Results and discussion

3.1 Pharmacophore generation

Pharmacophore models were computed and top 10

hypotheses were obtained. Results of the pharmacophore

hypotheses are presented in Table 1. The first hypothesis

(Hypo1) consisting of spatial arrangement of four

chemical features (Figure 2), including one HBA, one

HY and two HDB features along with five E-volumes was

identified as the best hypothesis. It had the lowest total cost

(123.41), the least RMSD (0.85) and a strong correlation

coefficient (0.94). Other statistical values obtained were

null cost value of 168.02, fixed cost value of 108.40

and the configuration cost value of 16.49. The 3D space

and distance constraints of these pharmacophore features

are shown in Figure 2(A),(B). Again, Figure 2(C),(D)

represents the Hypo1 aligned with the most active

compound 1 (IC50: 0.37 nM, and the least active

compound 27 (IC50: 12,000 nM) in the training set,

respectively. All features of Hypo1 model were nicely

mapped with the corresponding chemical functional

groups of compound 1. In contrast, compound 27 mapped

only three features while the fourth feature of HBAwas not

mapped. All training set compounds were classified into

three categories: highly active (IC50 , 100 nM, þþþ ),

moderately active (100 nM , IC50 , 1000 nM, þþ ) and

low active (IC50 . 1000 nM, þ ) compounds. Table 2

shows the experimental and estimated inhibitory activities

of the 27 training set compounds. Most of the compounds

in the training set were predicted correctly.

Table 1. Statistical parameters of the top 10 hypotheses of CCR2 antagonists generated by HypoGen program.

Hypothesis Total costa RMSD Correlation Featuresb

1 123.418 0.857 0.942 HBA, HBD, HBD, HY, 5E2 134.789 1.335 0.833 HBA, HBD, HBD, HY, 4E3 135.480 1.368 0.822 HBA, HBD, HY, 1E4 136.909 1.365 0.827 HBA, HBD, HY, 2E5 139.688 1.465 0.794 HBA, HBD, HY, 2E6 140.452 1.477 0.791 HBA, HBD, HBD, 2E7 140.481 1.474 0.793 HBA, HBD, HY, HY8 140.639 1.404 0.825 HBA, HBD, HY9 140.676 1.458 0.801 HBA, HBD, HY, 1E10 141.362 1.483 0.791 HBA, HBD, HBD, HY, 2E

aAll cost units are in bits: null cost, 168.02; fixed cost, 108.43; configuration, 16.49. b Abbreviations used for features: HBA, hydrogen-bond acceptor;HBD, hydrogen-bond donor; HY, hydrophobic; E, excluded volumes.

Figure 2. (Colour online) Pharmacophore model of CCR2antagonists generated by HypoGen. (A) The best model Hypo1:excluded volumes are ignored for clarity. (B) 3D spatialrelationship and geometric parameters of Hypo1. (C) Hypo1mapping with the most active compound 1 (IC50: 0.37 nM). (D)Hypo1 mapping with the least active compound 27 (IC50:12,000 nM). Pharmacophoric features: HBA (green), HY (cyan),HBD (magenta) and excluded volumes (grey).

R. Singh et al.52

Dow

nloa

ded

by [

Was

hing

ton

Uni

vers

ity in

St L

ouis

] at

23:

54 0

3 M

ay 2

013

3.2 Pharmacophore model validation

Cross validation using CatScramble program available in

Discovery Studio was applied to assess the statistical

confidence of Hypo1. The difference in costs between

the HypoGen runs and the scrambled runs are shown in

Table 1S (see supplementary data). The CatScramble

mixed up activity values of all training set compounds and

generated 19 random spreadsheets for attaining the 95%

Table 2. Actual and estimated biological activities (IC50 in nM) of CCR2 antagonists in training set (compounds 1–27).

ID Actual IC50 (nM) Experimental scale Estimated IC50 (nM) Estimated scale Fit value References

1 0.37 þþþ 0.88 þþþ 10.38 [19]2 1.30 þþþ 3.10 þþþ 9.84 [19]3 1.40 þþþ 3.30 þþþ 9.81 [19]4 2.70 þþþ 3.20 þþþ 9.83 [19]5 5.10 þþþ 15.00 þþþ 9.14 [18]6 5.80 þþþ 9.60 þþþ 9.35 [19]7 11.00 þþþ 6.50 þþþ 9.51 [17]8 15.00 þþþ 65.00 þþþ 8.51 [17]9 23.00 þþþ 31.00 þþþ 8.84 [17]10 28.00 þþþ 8.80 þþþ 9.38 [19]11 46.00 þþþ 30.00 þþþ 8.85 [19]12 48.00 þþþ 36.00 þþþ 8.77 [17]13 55.00 þþþ 120.00 þþ 8.23 [17]14 87.00 þþþ 280.00 þþ 7.88 [18]15 120.00 þþ 81.00 þþþ 8.42 [17]16 150.00 þþ 570.00 þþ 7.57 [17]17 170.00 þþ 320.00 þþ 7.82 [19]18 200.00 þþ 100.00 þþ 8.31 [19]19 260.00 þþ 240.00 þþ 7.95 [17]20 400.00 þþ 230.00 þþ 7.97 [17]21 530.00 þþ 740.00 þþ 7.46 [17]22 550.00 þþ 840.00 þþ 7.40 [18]23 970.00 þþ 210.00 þþ 8.00 [18]24 1700.00 þ 1100.00 þ 7.27 [17]25 3010.00 þ 1300.00 þ 7.22 [18]26 7860.00 þ 1300.00 þ 7.20 [18]27 12000.00 þ 1500.00 þ 7.16 [18]

Notes: 111 are highly active, 11 are intermediate active and 1 are low active molecules.

Figure 3. (Colour online) Results of Fischer’s randomisation test using CatScramble implemented inDiscovery Studio 2.5 package. The95% confidence level was selected.

Molecular Simulation 53

Dow

nloa

ded

by [

Was

hing

ton

Uni

vers

ity in

St L

ouis

] at

23:

54 0

3 M

ay 2

013

Figure 4. Chemical structures of CCR2 antagonists in test set (compounds 28–59).

R. Singh et al.54

Dow

nloa

ded

by [

Was

hing

ton

Uni

vers

ity in

St L

ouis

] at

23:

54 0

3 M

ay 2

013

confidence level. A graph presenting the results of Fischer

randomisation test is shown in Figure 3. In order to

validate further the best pharmacophore model, Hypo1, a

test set including external compounds (Figure 4) with

different activity was analysed. The test set molecules

validated the predictive character of Hypo1. Table 2S (see

supplementary data) illustrates the actual and estimated

biological activities of the test set molecules. Good

correlation coefficient of 0.81 between the experimental

and predicted activity values was observed for the test set

compounds. The accuracy, precision, sensitivity and

specificity of the best model for the training set were

0.89, 0.92, 0.85 and 0.93, respectively, whereas those for

the test set were 0.84, 0.81, 1.00 and 0.54, respectively

(Table 3S, supplementary data). The statistical parameters

show the ability of the best pharmacophore model, Hypo1,

to pick highly active molecules accurately.

3.3 Virtual screening

The molecules with novel and desired chemical features

were retrieved using validated pharmacophore model,

Hypo1 as a search query from ChemDiv and MiniMay-

bridge databases consisting of 186,000 and 2000

compounds, respectively. The ‘fast flexible search’

approach implemented within Discovery Studio was used

for search process. A total of 25,440 and 155 compounds

from ChemDiv and MiniMaybridge databases, respect-

ively, showed good mapping with the Hypo1. The obtained

molecules were structurally different and included a broad

range of templates. By applying Lipinski’s rule of five, the

number of hits was reduced to 16,585 and 127 compounds

for ChemDiv and MiniMaybridge databases, respectively.

The compounds with HypoGen estimated activity ,5 nM

were considered as the most active compounds. From

ChemDiv and MiniMaybridge databases, 574 and 10

compounds, respectively, satisfied the specified cut-off

values and hence proceeded for further evaluation.

3.4 Molecular docking

In-house developed homology model of human chemo-

kine CCR2 was used in the molecular docking studies.

Glide docking score with a cut-off of 28.0 was selected as

an indicator of high ligand-binding affinity. Six

compounds from ChemDiv and two compounds from

MiniMaybridge databases were identified as the potential

lead compounds. Mapping of Hypo1 on the most active

compounds a and g from ChemDiv and MiniMaybridge

databases, respectively, is represented in Figure 5. Table 3

shows the docking scores of the eight lead compounds

along with their fit and estimated IC50 values, obtained

after mapping the compounds with Hypo1. Figures S1 and

S2 (supplementary data) illustrate the docked poses of two

of the eight lead compounds in the receptor cavity. All the

eight top hits showed high estimated IC50 values and

binding affinities towards the CCR2, and were subjected to

similarity search and absorption, distribution, metabolism

and excretion (ADME) analysis.

3.5 Similarity search analysis

Similarity search was carried out to check the novelty of

the screened hits against CCR2 and for this, PubChem

search tool was used [39]. Partially similar compounds

were not able to fit into the geometric constraints of Hypo1

and showed poor fit values. PubChem BioAssay search

confirmed that for the eight lead molecules, no CCR2

antagonistic activity was tested to date. Furthermore,

docking scores for the lead molecules were better than the

partially similar compounds. In addition, the partially

similar compounds failed to interact with the important

residues involved in ligand binding, viz. Tyr120, His121,

Arg206, Tyr259 and Glu291. Thus, similarity search

suggests the novelty of the scaffolds for CCR2

antagonism.

3.6 ADME analysis

Drug-like behaviour of the potential leads was assessed by

analysing pharmacokinetic parameters required for

ADME using QikProp 3.2 [40]. QikProp is frequently

used by researchers for ADME analysis of potential leads

[41]. We also calculated QikProp properties for the

obtained hits. The molecular weights of all the lead

molecules (a–h) were in the range of 337 – 487.

QP logPo/w shows the partition coefficient, which is

important for the estimation of absorption and distribution

of drugs within the body. Partition coefficient for the lead

compounds ranged from 1.05 to 4.74, which is in the

acceptable range of 22.0 to 6.5. Cell permeability

(QPPCaco), a key factor governing drug metabolism and its

access to biological membranes, ranged from 29.7 to

1207.5. The percentage of human oral absorption

calculated for the potential lead molecules was moderate

to high. All the compounds showed pharmacokinetic

Figure 5. (Colour online) Hypo1 mapping with the most activecompounds a (left panel) and g (right panel), from ChemDiv andMiniMayBridge databases, respectively.

Molecular Simulation 55

Dow

nloa

ded

by [

Was

hing

ton

Uni

vers

ity in

St L

ouis

] at

23:

54 0

3 M

ay 2

013

Table 3. Estimated biological activities (IC50 in nM), fit values and docking scores of the potential hits (compounds a–h).

Compound Structure Database name Estimated IC50 (nM) Fit value Glide docking score

a

HO

HO NHN

O

S

NN

ChemDiv 0.04 11.67 29.58

b

OH

OH

HO

HN

HO

O

N

O

O

ChemDiv 0.08 11.43 29.91

c

OH

N

OHOH

N

O

N

ChemDiv 0.53 10.60 29.11

d

ONH

N

OH

HOChemDiv 0.65 10.51 28.95

eO

HN

OH

NO

ChemDiv 1.76 10.08 210.12

f

OHN

OS

F

NN

O

ChemDiv 2.79 9.88 28.83

g FF

FN

NHN

NH

OH

O

N+-O

O

MiniMayBridge 0.13 11.21 28.71

hCl S

OH

NHO MiniMayBridge 3.31 9.81 29.66

R. Singh et al.56

Dow

nloa

ded

by [

Was

hing

ton

Uni

vers

ity in

St L

ouis

] at

23:

54 0

3 M

ay 2

013

parameters within the acceptable range (Table 4), which

signify potential drug-like molecules.

4. Conclusions

Three-dimensional pharmacophore models of CCR2

inhibitors have been developed with the aid of HypoGen

module in Discovery Studio 2.5 program. The best

pharmacophore model, Hypo1, consists of four pharma-

cophoric features including one HBA, one HY and two

HDB along with five E-volumes. Hypo1 was characterised

by the lowest total cost (123.41), lowest RMSD (0.85) and

the best correlation coefficient (0.94). The pharmacophore

model was validated by both cross validation and external

test set prediction methods. The statistical confidence of

Hypo1 was substantiated by Fischer randomisation test

method. The model was then used as a 3D search query to

screen ChemDiv and MiniMaybridge databases. Lipinski’s

filter was used to obtain the drug-like molecules. Again,

compounds with estimated IC50 values ,5 nM were

selected for the docking studies. Finally, eight potential

leads with good in silico CCR2-binding affinities, and

ADME profiles were obtained. The results from the

combined 3D-QSAR pharmacophore modelling and

molecular docking approach have escorted to the

prediction of novel potent CCR2 antagonist scaffolds.

Acknowledgements

The authors thank the Ministry of Chemicals and Fertilisers andthe Council of Scientific and Industrial Research (CSIR) forfinancial support.

References

[1] G. Lazennec and A. Richmond, Chemokines and chemokinereceptors: New insights into cancer-related inflammation,Trends Mol. Med. 16 (2010), pp. 133–144.

[2] Y. Bordon, Asthma and allergy: A breathtaking chemokine,Nat. Rev. Immunol. 10 (2010), pp. 810–811.

[3] E.A. Berger, P.M. Murphy, and J.M. Farber, Chemokine receptorsas HIV-1 coreceptors: Roles in viral entry, tropism, and disease,Annu. Rev. Immunol. 17 (1999), pp. 657–700.

[4] C. Murdoch and A. Finn, Chemokine receptors and their role ininflammation and infectious diseases, Blood 95 (2000),pp. 3032–3043.

[5] A. Viola and A.D. Luster, Chemokines and their receptors: Drugtargets in immunity and inflammation, Annu. Rev. Pharmacol.Toxicol. 48 (2008), pp. 171–197.

[6] M.R. Chacon, J.M. Fernandez-Real, C. Richart, A. Megıa,J.M. Gomez, M. Miranda, E. Caubet, R. Pastor, C. Masdevall, andN. Vilarrasa, Monocyte chemoattractant protein-1 in obesity andtype 2 diabetes, Insulin sensitivity study, Obesity. 15 (2007),pp. 664–672.

[7] L. Gu, S.C. Tseng, and B.J. Rollins, Monocyte chemoattractantprotein-1, Chemokines 72 (1999), pp. 7–29.

[8] C.N. Lumeng, J.L. Bodzin, and A.R. Saltiel, Obesity induces aphenotypic switch in adipose tissue macrophage polarization,J. Clin. Invest. 117 (2007), pp. 175–184.

[9] C.N. Lumeng, S.M. DeYoung, J.L. Bodzin, and A.R. Saltiel,Increased inflammatory properties of adipose tissue macrophagesrecruited during diet-induced obesity, Diabetes 56 (2007),pp. 16–23.

[10] R.C. Newton and K. Vaddi, Biological responses to C-Cchemokines, Methods Enzymol. 287 (1997), pp. 174–186.

[11] A. Rot and U.H. von Andrian, Chemokines in innate and adaptivehost defense: Basic chemokines grammar for immune cells,Annu. Rev. Immunol. 22 (2004), pp. 891–928.

[12] I.F. Charo and R.M. Ransohoff, The many roles of chemokines andchemokine receptors in inflammation, New Engl. J. Med. 354(2006), pp. 610–621.

[13] C. Gerard and B.J. Rollins, Chemokines and disease, Nat. Immunol.2 (2001), pp. 108–115.

[14] A.D. Luster, Chemokines-chemotactic cytokines that mediateinflammation, New Engl. J. Med. 338 (1998), pp. 436–445.

[15] G. Van Lommen, J. Doyon, E. Coesemans, S. Boeckx, M. Cools,M. Buntinx, B. Hermans, and J. Van Wauwe, 2-Mercaptoimida-zoles, a new class of potent CCR2 antagonists, Bioorg. Med. Chem.Lett. 15 (2005), pp. 497–500.

[16] A.B. Pinkerton, D. Huang, R.V. Cube, J.H. Hutchinson,M. Struthers, J.M. Ayala, P.P. Vicario, S.R. Patel, T. Wisniewski,and J.A. DeMartino, Diaryl substituted pyrazoles as potent CCR2receptor antagonists, Bioorg. Med. Chem. Lett. 17 (2007),pp. 807–813.

[17] P.H. Carter, G.D. Brown, S.R. Friedrich, R.J. Cherney, A.J. Tebben,Y.C. Lo, G. Yang, H. Jezak, K.A. Solomon, and P.A. Scherle,Capped diaminopropionamide–glycine dipeptides are inhibitors ofCC chemokine receptor 2 (CCR2), Bioorg. Med. Chem. Lett. 17(2007), pp. 5455–5461.

[18] R.J. Cherney, R. Mo, D.T. Meyer, D.J. Nelson, Y.C. Lo, G. Yang,P.A. Scherle, S. Mandlekar, Z.R. Wasserman, and H. Jezak,Discovery of disubstituted cyclohexanes as a new class of CCchemokine receptor 2 antagonists, J. Med. Chem. 51 (2008),pp. 721–724.

[19] R.J. Cherney, R. Mo, D.T. Meyer, M.E. Voss, Y.C. Lo, G. Yang,P.B. Miller, P.A. Scherle, A.J. Tebben, and P.H. Carter,Novel sulfone-containing di- and trisubstituted cyclohexanes as

Table 4. ADME properties of the lead compounds calculated using QikProp.

Compound Mol. wt.a QP logPo/wb QPPCaco

c Percent human oral absorptiond

a 446.52 4.74 202.10 96.01b 486.56 2.14 29.71 65.89c 361.44 1.05 126.40 70.74d 386.45 4.45 408.88 100.00e 380.48 4.29 1207.52 100.00f 415.48 4.20 1097.42 100.00g 401.34 2.29 40.06 69.03h 337.86 3.75 855.52 100.00

aMolecular weight. b Predicted octanol/water partition coefficient log P (acceptable range: 22.0 to 6.5). c Predicted Caco-2 cell permeability in nm/s(acceptable range , 25 is poor and .500 is great). d Percentage of human oral absorption (,25% is poor and .80% is high).

Molecular Simulation 57

Dow

nloa

ded

by [

Was

hing

ton

Uni

vers

ity in

St L

ouis

] at

23:

54 0

3 M

ay 2

013

potent CC chemokine receptor 2 (CCR2) antagonists, Bioorg. Med.Chem. Lett. 19 (2009), pp. 3418–3422.

[20] R. Singh and M.E. Sobhia, Synergistic application of targetstructure-based alignment and 3D-QSAR study of protein tyrosinephosphatase 1B (PTP1B) inhibitors, Med. Chem. Res. 20 (2010),pp. 714–725.

[21] A.K. Jain, V. Ravichandran, R. Singh, S. Sharma, V.K. Mourya,and R.K. Agrawal, QSAR study of disubstituted N6-cyclopentyl-adenine analogues as a adenosine A1 receptor antagonist,Digest J. Nanomater. Biostruct. 3 (2008), pp. 63–73.

[22] A.K. Jain, V. Ravichandran, R. Singh, V. Mourya, andR.K. Agrawal, QSAR study of 2,4-disubstituted phenoxyaceticacid derivatives as a CRTh 2 receptor antagonists, Chem. Pap. 63(2009), pp. 464–470.

[23] R. Singh, A. Jain, V. Ravichandran, V. Mourya, and R.K. Agrawal,Prediction of antiproliferative activity of some flavone derivatives:QSAR study, Med. Chem. Res. 18 (2009), pp. 523–537.

[24] P.C. Nair and M.E. Sobhia, Fingerprint directed scaffold hoppingfor identification of CCR2 antagonists, J. Chem. Info. Model. 48(2008), pp. 1891–1902.

[25] P.C. Nair, K. Srikanth, and M.E. Sobhia, QSAR studies onCCR2 antagonists with chiral sensitive hologram descriptors,Bioorg. Med. Chem. Lett. 18 (2008), pp. 1323–1330.

[26] M.E. Sobhia, R. Singh, P. Kare, and S. Chavan, Rational design ofCCR2 antagonists: A survey of computational studies, Expert Opin.Drug Discov. 5 (2010), pp. 543–557.

[27] P.K. Ojha, I. Mitra, R.N. Das, and K. Roy, Further exploring rm2metrics for validation of QSPR models, Chemom. Intell. Lab. Syst.107 (2011), pp. 194–205.

[28] K. Roy, I. Mitra, S. Kar, P.K. Ojha, R.N. Das, and H. Kabir,Comparative studies on some metrics for external validation ofQSPR models, J. Chem. Info. Model. 52 (2012), pp. 396–408.

[29] Y. Kurogi and O.F. Guner, Pharmacophore modeling and three-dimensional database searching for drug design using catalyst,Curr. Med. Chem. 8 (2001), pp. 1035–1055.

[30] O.F. Guner, History and evolution of the pharmacophore concept in

computer-aided drug design, Curr. Top. Med. Chem. 2 (2002),

pp. 1321–1332.

[31] T. Langer and R.D. Hoffmann, Pharmacophore modelling:

Applications in drug discovery, Expert Opin. Drug Discov. 1

(2006), pp. 261–267.

[32] K.H. Kim, N.D. Kim, and B.L. Seong, Pharmacophore-based

virtual screening: A review of recent applications, Expert Opin.

Drug Discov. 5 (2010), pp. 205–222.

[33] S.Y. Yang, Pharmacophore modeling and applications in drug

discovery: Challenges and recent advances, Drug Discov. Today 15

(2010), pp. 444–450.

[34] Discovery Studio 2.5 Guide, Accelrys Inc., San Diego, CA, 2009.

[35] Sybyl 7.1, Tripos Inc., St. Louis, MO, 2005.

[36] H. Li, J. Sutter, and R. Hoffmann, HypoGen: An automated system

for generating predictive 3D pharmacophore models,

in Pharmacophore Perception, Development, and Use in Drug

Design, O.F. Guner, ed., International University Line, La Jolla, CA,

2000, pp. 173–188.

[37] Glide, Version 5.5, Schrodinger, LLC, New York, NY, 2009.

[38] R. Singh and M.E. Sobhia, Homology modeling of human CCR2

receptor, Med. Chem. Res. 20 (2010), pp. 1704–1712.

[39] Y. Wang, E. Bolton, S. Dracheva, K. Karapetyan, B.A. Shoemaker,

T.O. Suzek, J. Wang, J. Xiao, J. Zhang, and S.H. Bryant,

An overview of the PubChem BioAssay resource, Nucleic Acids

Res. 38 (2010), pp. D255–D266.

[40] QikProp, Version 3.2, Schrodinger, LLC, New York, NY, 2009.

[41] B. Bandgar, B. Hote, R. Gangwal, and A. Sangamwar,

Synthesis, biological evaluation, and pharmacokinetic profiling

of benzo-phenone derivatives as tumor necrosis factor-alpha

and interleukin-6 inhibitors, Med. Chem. Res. (2011) (DOI:

10.1007/s00044-011-9856-1).

R. Singh et al.58

Dow

nloa

ded

by [

Was

hing

ton

Uni

vers

ity in

St L

ouis

] at

23:

54 0

3 M

ay 2

013