Available on line www.jocpr.

com

Journal of Chemical and Pharmaceutical Research

__________________________________________________

J. Chem. Pharm. Res., 2010, 2(6):344-350

ISSN No: 0975-7384
CODEN(USA): JCPRC5

Computational QSAR analysis of some physiochemical and topological

descriptors of Curcumin derivatives by using different statistical methods

Raghunath Satpathy, Rajesh Ku. Guru, Rashmiranjan Behera

Department of Biotechnology, MIRC LAB, MITS Engineering College, Rayagada, Odisha, India
_____________________________________________________________________________

ABATRACT
Curcumin is produced from the rhizomes of Curcuma longa plant and having various medicinal
and pharmaceutical applications.Here in this work a QSAR study has been performed by taking
the 23 analogs of Curcumin.Various structural and physiochemical descriptors were generated.
The effect was calculated for each type of descriptors by taking the Andrews coefficient as
dependent variable. Multiple regression analysis was performed by Minitab 14 tool. Good
correlation R-sq value 0.78 was obtained from the physiochemical descriptors in comparison to
structural descriptor calculation. The statistics were also further verified by using SVM (Support
vector machines) and ANN (artificial neural networks) based calculation. The results obtained
were consistent with MLR statistics and the ANN based method show R-sq value as 0.88 in case
of physiological descriptor which is observed to be the highest among above three methods of
analysis.

Key words: Descriptors, QSAR, Curcumin, Multiple linear regression, artificial neural network
_____________________________________________________________________________

INTRODUCTION

Curcumin is an alkaloid produced from the turmeric plant Curcuma longa, which is a member of
the ginger family (Zingiberaceae). Historically the turmeric has been used as a major component
of Indian Ayurvedic medicine to treat a wide variety of health problems [1]. Current research
has also identified the Curcumin as responsible molecule for most of the biological activity of
turmeric. The Curcumin molecules are chemically polyphenols and are responsible for the
yellow color of turmeric and can exist in at least two tautomeric forms, keto and enol
[2].Curcumin incorporates several functional groups and the aromatic ring systems the carbonyl
groups form a diketone [3].Recently numerous clinical trials in humans are going on,
investigating the effect of Curcumin on various diseases including multiple myeloma, pancreatic
cancer, myelodysplastic syndromes, colon cancer, psoriasis, and Alzheimer's disease, and also
deadliest Swine flu [4-5-6-7].

344
Raghunath Satpathy et al J. Chem. Pharm. Res., 2010, 2(6):344-350
_____________________________________________________________________________

To analyse different potential drug molecules the quantitative structure-activity relationship

(QSAR) method is a useful approach.QSAR is basically used to study the biological activities
with various properties associated with the structures, which is helpful to explain how structural
features in a drug molecule influence the biological activities. The analysis also gathers
information that is very much useful for molecular drug design and medicinal Chemistry.
Therefore correlating the physicochemical properties or structural features of the important
compounds with their biological activity is essential. In addition to this a successful in silico
based QSAR analysis also provides the advantages of higher speed and lower costs for
bioactivity evaluation of drug as compared to experimental testing [8].

EXPERIMENTAL SECTION

The molecular structure of Curcumin derivatives were collected from Pubchem database
available in the NCBI server (http://pubchem.ncbi.nlm.nih.gov/).The structure were drawn by
Marvin sketch 5.0 tool (http://www.chemaxon.com/marvin/sketch/index.jsp) and corresponding
3D structure were obtained. The molecules were then energy minimised by PRODRG server [9].
Prodrg is an on line tool where the energy minimization of the molecule was performed by using
Gromos 96 force field. Then the energy minimised molecules were fed to Preadmet server
(http://preadmet.bmdrc.org/preadmet/index.php) for the calculation of descriptors. Two types of
descriptors were chosen physiological and topological types under which the selected descriptors
were calculated (Table 1).

Table 1: The various physiochemical and Topological descriptors considered in the study

Serial number Physiochemical Topological

1 Molecular weight (MW) Quadratic index (QI)
2 2D Vander walls volume (2DVWV) Edge based molecular topological
index (EMTI)
3 Water solvation free energy (WSE) Kier symmetry index (KSI)
4 Hydrophobic surface area saturated (HSAS) Ring degree distance index (RDI)
5 Hydrophobic surface area un-saturated Eccentric connectivity index (ECI)
(HSAU)
6 LogP Wiener index (WI)

For the two types of descriptor sets, MLR (Multiple linear regressions) analysis was performed
by using the MINITAB 14 tool [10].The Andrews affinity was chosen as dependent variable
.Andrews affinity is calculated based on the drug receptor binding affinity [11]. For the best
model selection the statistical parameters like F value, R-Sq value and mean square deviation etc
were considered. The above MLR calculations were also further verified by ANN (Artificial
Neural Network) and SVM (Support Vector Machine) based approach by using Molegro tool
[12].

RESULTS AND DISCUSSION

Overall 23 Curcumin analogs were retrieved from Pubchem data base and the same were used
for the QSAR analysis (Table 2).

345
Raghunath Satpathy et al J. Chem. Pharm. Res., 2010, 2(6):344-350
_____________________________________________________________________________

Table 2: Structure of Curcumin analogs considered for the experiment

Curcumin Andrews Molecular

Structure
analogs affinity weight

4.76372
1 332.182

4.83701
2 368.385

3.51782
3 368.381

4.69043
4 366.413

4.76372
5 370.401

4.69043
6 368.385

3.88426
7 338.359

2.85823
8 308.333

6.44934
9 452.459

3.0781
10 336.387

2.85823
11 448.515

346
Raghunath Satpathy et al J. Chem. Pharm. Res., 2010, 2(6):344-350
_____________________________________________________________________________

22.2063
12 692.667

22.2063
13 694.683

13.485
14 532.542

13.5583
15 425.437

13.485
16 530.526

3.51782

17 396.439

10.1137
18 588.659

25.4309
19 566.651

33.9323
20 626.615

22.4261
21 482.489

347
Raghunath Satpathy et al J. Chem. Pharm. Res., 2010, 2(6):344-350
_____________________________________________________________________________
OH OH

O O
H C CH
3 3

HO O

O
O OH 22.2063
22 HO
O O 694.683
OH
HO OH

OH

19.348
23 744.749

The predicted Andrews affinity on various physiochemical and topological descriptors were
calculated by MLR analysis and the regression equations were generated from Mintab 14 tool.
Andrews affinity (physiochemical descriptors) = - 27.0 - 5.11 log p - 0.0291 MW + 0.159
2DVWV + 0.123 WSE - 0.0395 HSAS + 0.156 HSAU
Andrews affinity (topological descriptors) = - 20.5 - 0.18 QI - 0.000022 EMTI + 0.828 KSI -
1.62 RDM + 0.0014 ECI - 0.00108 WI

Figure 1: The predicted and calculated affinity relationship in case of physiochemical descriptors by ANN
method.

Figure 2: The predicted and calculated affinity relationship in case of topological descriptors by ANN
method.

The data set was further verified by Molegro software. In addition to the MLR analysis the
SVM and ANN method was used to calculate the statistical variables because the comparative
mode of statistical analysis (Multiple linear regression analysis, SVM and ANN based approach)
are more reliable to analyse the statistical parameters [13].Default parameter set up was
considered in Molegro tool for the ANN and SVM based calculation. In case of ANN based
calculation single hidden layer with 3 neurons were chosen. Among all 3 method of analysis

348
Raghunath Satpathy et al J. Chem. Pharm. Res., 2010, 2(6):344-350
_____________________________________________________________________________

the application of artificial neural networks show the maximum Pearsons coefficient
0.940,Pearsons coefficient square 0.88 and minimum mean square deviation 10.033 in case of
physiochemical descriptors that signifies the statistical analysis (Table-3).

Table 3: Comparative statistical parameter calculations by using Molegro tool

Spearman Mean Cross

Pearson Pearson’s
Methods of Rank Squared validated
Descriptor Type Correlatio coefficient
analysis Correlation Deviation squared
n (r) square (r2)
(ρ) (MSD) (q2)
MLR 0.886 0.7844 0.777 17.081 0.784
SVM 0.862 0.742 0.805 22.046 0.721
Physiochemical
ANN 0.940 0.884 0.77 10.033 0.873

MLR 0.799 0.637 0.656 28.697 0.637

Topological SVM 0.838 0.703 0.719 23.96 0.697
ANN 0.905 0.818 0.758 14.366 0.818

The artificial neural network based calculation provides the improved QSAR model for the
effect topological (Figure 1) and physiochemical descriptors (Figure 2) with the Andrews
affinity. The successful application of ANN methods to QSAR analysis also has been confirmed
for other drug molecules in medicinal chemistry [14]. So the ANN could be used as a promising
tool for a good statistical approximation thereby solving complex problems. In general the
topological and structural descriptors are very important type of molecular descriptor for
bioactivity prediction [15]. Here the results in this work indicate in comparison to topological
parameters the physiochemical parameters are more responsible for receptor binding activity of
Curcumin.

Acknowledgement:
We are thankful to Chief executive, Majhighariani Institute of Technology & Science, Ryagada
for his encouragement and to provide us MIRC lab for computing facility.

REFERENCES

[1]Aggarwal BB; Sundaram C; Malani N; Ichikawa H. Adv. Exp. Med. Biol. 2007, 595: 1–75.
[2]Tsonko M. Kolev; Evelina A.; Velcheva; Bistra A; Stamboliyska; Michael Spiteller.
International Journal of Quantum Chemistry, 2005, 102 (6), 1069–79.
[3]Albena T.; Dinkova-Kostova; Paul Talalay Carcinogenesis, 1999, 20 (5), 911-914.
[4]Aggarwal BB; Shishodia S. Biochemical Pharmacology, 2006, 71 (10), 1397–1421.
[5]Hyunsung Choi; Yang-Sook Chun; Seung-Won Kim; Myung-Suk Kim; Jong-Wan Park.
Molecular Pharmacology, 2006, 70 (5), 1664–71.
[6]Hatcher H; Planalp R; Cho J; Torti FM; Torti SV. Cell. Mol. Life Sci., 2008, 65 (11), 1631–
1652.
[7]Da-Yuan Chen ; Jui-Hung Shien ; Laurence Tiley ; Shyan-Song Chiou ; Sheng-Yang
Wang; Tien-Jye Chang ; Ya-Jane Lee ; Kun-Wei Chan ; Wei-Li Hsu. FoodChemistry,
2010, 119 (4), 1346–1351.
[8] David A. Winkler. Briefings in Bioinformatics, 2002, 3(1), 73-86.
[9]Schuettelkopf AW; D.M.F VanAalten. Acta Crystallographica, 2004, D60: 1355-1363.
[10]David J.; Wild J. Chem. Inf. Model, 2005, 459(1), 212
[11]Andrews PR; Craik DJ; Martin JL. J Med Chem. 1984, 27(12), 1648-1657.
[12] www.molegro.com
[13]Shahlaei M; Fassihi A; Saghaie L. Eur J Med Chem., 2010, 45(4), 1572-1582.

349
Raghunath Satpathy et al J. Chem. Pharm. Res., 2010, 2(6):344-350
_____________________________________________________________________________

[14]Zahra Garkani-Nejad; Fereshteh Saneie. Bull. Chem. Soc. Ethiop., 2010, 24(3), 317-325.
[15]Du Q, Mezey PG; Chou KC. J Comput Chem., 2005, 26(5):461-470.

350