MODELING ANTIBACTERIAL ACTIVITY OF 4-THIOZOLIDONE DERIVATIVES

Received on 12 February, 2014; received in revised form, 23 April, 2014; accepted, 13 June, 2014; published 01 August, 2014

MODELING ANTIBACTERIAL ACTIVITY OF 4-THIOZOLIDONE DERIVATIVES

K. Anita*¹, V.K. Agrawal ², B. Shaik ², S. Sharma ¹

Department of Chemistry, Career College ¹, Bhopal, Madhya Pradesh, India

Department of Applied Science, National Institute of Technical Teachers Training and Research ², Shamla  Hills, Bhopal-462002, Madhya Pradesh, India

ABSTRACT: In the present work, efforts have been made to model the antibacterial activity of 4-thiazolidone derivatives against pathogenic bacteria P. aeruginosa by using QSAR (quantitative structure-activity relationship) methodology. Multivariate analysis gave excellent model which was tested using cross validation using leave one out method. The cross- validation method was applied to the data set in order to prove the predictive power of statistically significant QSAR models, which help to explore some expectedly potent compounds. The best model predicting the antibacterial activity indicated that the 2D auto-correlation, 3D and WHIM parameters such as MATS8p, RDF070u, RDF035e, Mor30v (3D) and E1u (WHIM Parameter) are very effective in describing the antibacterial activities of these compounds. The study revealed that E1u, MATS8p, RDF035e and RDF070u contribute positively whereas contribution of Mor30v contributes negatively to the antibacterial activity. The compounds with improved antibacterial potential can be successfully designed with selected quantitative structure activity relationship model.

Antibacterial activity, 4- thiazolidone, Multivariate analysis, Cross validation

INTRODUCTION: Heterocyclic compounds like thiazolidones, are very good antibacterial drugs. Thiazolidones and their derivatives are known to have antimicrobial, antiviral, antitumor, antihypertensive and anti-inflammatory properties ^1-2. However, Searching of more potent and efficient antibacterial agents is one of the major tasks of clinical practice due to the antibiotic resistant strains. Quantitative structure activity relationship (QSAR) analysis has been found to be a good tool for prediction of biological activity of novel compounds including antibacterial and antiviral agents ^3-7.

Computer technologies based on model development provide good insight for modification of molecular structures to get new organic molecules giving useful properties (including antibacterial activity). The present work is focused on the modeling of antibacterial activity of 4-thiazolidone derivatives.

MATERIALS AND METHODS: QSAR has been widely used for years to provide a quantitative correlation between chemical structure and biological activity ⁸. Agrawal and coworkers have used many physicochemical and topological parameters to predict the antibacterial activity of many compounds ^9-12.

In the present study, 25 thiazolidone derivatives have been considered having biological activity in terms of its log values. The compounds are taken from the literature ¹³.

Dragon 6 software is used for the calculation of descriptors. The structural details of the compounds used in the present study are given in Table 1. The clinical isolates of the pathogenic bacteria P. aeruginosa has been used as the test microorganisms.

TABLE 1: STRUCTURES OF COMPOUNDS USED IN THE PRESENT STUDY

The calculated descriptors from Dragon software are reported in Table 2. Useful descriptors were selected by variable selection procedure and multiple regression analysis was performed using NCSS software ¹⁴. The models obtained were subjected to cross validation by leave one out procedure ¹⁵. The parameters which have been calculated for modeling are:

TABLE 2: CALCULATED VALUES OF PARAMETERS ALONG WITH THEIR BIOLOGICAL ACTIVITY

1. E1u: 1st component accessibility directional WHIM index / unweighted ^16-20. WHIM descriptors are based on the statistical indices calculated on the projections of atoms along principal axes. They are built in such a way as to capture relevant molecular 3D information regarding the molecular size, shape, symmetry and atom distribution with respect to invariant reference frames. The algorithm consists of performing a Principal Components Analysis on the centered Cartesian coordinates of a molecule by using a weighted covariance matrix obtained from different weighing schemes for the atoms.

The following weighting schemes are used for computing the weighted covariance matrix, S^w:

• Unweighted (u), that is the weight w_i= 1 for each i.
• Atomic masses (w_i = m_i )
• Atomic van der waals volumes (w_i = v_i)
• Atomic Sanderson electronegativities (w_i = e_i)
• Atomic polarizabilities (w_i = p_i)
• Atomic electro topological states (w_i = s_i)

2. MATS8p: Moran autocorrelation of lag 8 weighted by polarizability ^21-22. Moran Autocorrelation Descriptors is labeled as MATS. The symbol for each of the autocorrelation descriptors is followed by two indices d and w where d stands for the lag and w stands for the weight. The lag is defined as the topological distance d between pairs of atoms. The topological distance between a pair of atoms (i, j) is given in the ij^thentry in the Topological Level Matrix. The lag can have any value from the set {0, 1, 2, 3, 4, 5, 6, 7, 8}.The weight can be m (relative atomic mass), p (polarizability), e (Sanderson electronegativity) and v (Van der Waals volume). Relative mass is defined as the ratio of atomic mass of an atom to that of carbon. Similarly, the other three weights p, e and v are scaled by the corresponding values for Carbon.

Let n be the number of atoms in the molecule. For any chosen value for lag d and any chosen weight w, we compute the Autocorrelation Descriptors using the following formulae.

Where, w_i and w_j are the weights of the atoms i and j, , and δ_ij is Kronecker delta, that is,  δ_ij =1  if the ij^th entry in the Topological Level Matrix is = d, and  δ_ij = 0 otherwise.

3. RDF035e: Radial Distribution Function-035 /weighted by Sanderson electronegativity
4. RDF070u: Radial Distribution Function - 070 / unweighted ²³. The radial distribution function (RDF) descriptors are based on the distance distribution in the molecule. The radial distribution function of an ensemble of n atoms can be interpreted as the probability distribution of finding an atom in a spherical volume of radius R. A typical RDF descriptor is denoted by RDFsw where s and w take the values 10 ≤ s ≤ 155 in units of 5 and, and it is defined as follows:

Where, f is a scaling factor, r_ij is the Euclidean distance between the atoms i and j, w_i and w_j  are the weights of the atoms i and j  respectively, n is the total number of atoms, β is the smoothing parameter which defines the probability distribution of the individual inter-atomic distance. β can be interpreted as the temperature factor that defines the movement of the atoms.

5. Mor30v: 3D-MoRSE descriptors Weighted by van der Waals volume ^24-26. 3D MoRSE descriptors (3D Molecule Representation of Structures based on Electron diffraction) are derived from Infrared spectra simulation using a generalized scattering function (Soltzberg and Wilkins, 1977). A typical MoRSE descriptor is denoted by M orsw where s and w take the values 1≤ s ≤ 32 and, where, u is unweighted, m is weighted by mass, v is weighted by van der Waals volume, e is weighted by electronegativity and p is weighted by polarizability

The MoRSE descriptor is defined as follows:

Where, r_ij is the Euclidean distance between the atoms i and j, and w_i and w_jare the weights of the atoms i and j respectively.

RESULTS AND DISCUSSION: The parameters calculated have been summarized in Table 2 which includes MATS8p, RDF070u, RDF035e, Mor30v and E1u. A correlation matrix has been obtained which shows correlation among the selected parameters and activity and is reported in Table 3. This table reveals that E1u is the only parameter which may be useful in one-parametric modeling. However, the combination of different parameters may result better models. The data was subjected to regression analysis and many statistically significant models have been obtained which are summarized in Table 4.

TABLE 3: CORRELATION MATRIX

TABLE 4: REGRESSION PARAMETERS AND QUALITY OF CORRELATIONS

Here some statistically significant models having R² more than 0.7 have been discussed.

Three variable model: When Mor30v and RDFo35e is added with Elu a three parametric model is resulted with R² = 0.7488. The adjusted R²_A (0.7129) for this model also shows significant improvement. The model is as below:

log BA =0.8288(±0.1072) E1u-0.1942(±0.0524) Mor30v+0.0059(±0.0010) RDF035e+0.2633

N=25, R² = 0.7488, R²_A= 0.7129, Se = 0.0317, F=20.869, Q = 27.2975

Addition of RDF070u to above model yielded a four parametric model with better statistics. The R² value changes from 0.7488 to 0.8292 and R²_A change from 0.7129 to 0.7951. Change in adjusted R² clearly indicates that the added parameter has its fair share in the model. The model is as below:

Four variable model: log BA =0.9343(±0.0969) E1u-0.2595(±0.0492) Mor30v+0.0047(±0.0010) RDF035e+0.0022(±0.0007) RDF070u+ 0.2132

N=25, R² = 0.8292, R²_A= 0.7951, Se= 0.0268, F=24.276, Q = 33.9778

Further improvement was observed when E1u, MATS8p,Mor30v, RDF035e and RDF070u have been taken together resulting into a five-parametric model (model 9, Table 4). The values of R² and R²_A have come out to be 0.8981 and 0.8713 and the Q value ^27-28 has come out to be 44.4921. The model is as under:

Five variable model: log BA =0.9059(±0.0772) E1u+0.1161(±0.0324) MATS8p-0.2826(±0.0395) Mor30v+0.0048(±0.0008) RDF035e+0.0023(±0.0006) RDF070u+ 0.2342

N=25, R² = 0.8981, R²_A= 0.8713, Se= 0.0213, F=33.488, Q = 44.4921

No higher order model is permitted as Rule of Thumb restricts that (No of compounds are 25 hence maximum permitted no of parameters is 5.) Therefore, the five parametric model is the best model for modeling the anti-bacterial activity (log BA) of compounds used in the present study. Further confirmation is obtained by plotting observed activity against estimated activity and such a comparison is demonstrated in figure 1. The predictive power of the model comes out to be 0.8981. The biological activity (log BA) of all the compounds have been estimated using model- 9 (Table 5). The estimated log BA values are in good agreement with the observed values showing that the proposed model is best suited for estimating log BA values of present set of compounds.

To validate the model cross validation parameters have been calculated and they are reported in Table 6. We know that PRESS is a good estimate of the real predictive power of the model. If this value is smaller than SSY, the model predicts better than chance and can be considered statistically significant. Table 6 shows that all the proposed models are better than chance and are statistically significant. The ratio PRESS ⁄ SSY can be used to calculate the approximate confidence interval of the prediction of new compounds. If this ratio should be smaller than 0.4 the model is reasonably good. In the proposed model, this ratio is smaller than 0.4 and therefore, the model-9 has excellent predictive power.

TABLE 5: OBSERVED AND ESTIMATED BIOLOGICAL ACTIVITY AND RESIDUAL VALUES USING MODEL 9

TABLE 6: CROSS VALIDATION PARAMETERS FOR PROPOSED MODELS

FIG. 1: CORRELATION BETWEEN OBSERVED AND ESTIMATED BIOLOGICAL ACTIVITY VALUES   USING   MODEL 9

The developed models are cross-validated by leave-one-out method. Another cross-validated parameter related to uncertainty of prediction, the PSE, has also been calculated. The lowest value of PSE for model 9 supports its highest predictive potential (power).

The low value of PSE and S_PRESS and high value of R²_CV suggest that the five-parametric model is most appropriate in predicting the log BA values of present set of compounds.

There is no colinearity among the used parameters which has been established by ridge analysis as well as various inflation factors calculated from the model 9 (Table 7, figures 2 and 3).

TABLE 7: RIDGE ANALYSIS FOR FIVE VARIABLE MODEL (MODEL 9)

VIF= Variance inflation factor; T= Tolerance; λ      = Eigen value; k= Condition number

FIG. 2: RIDGE TRACE FOR FIVE VARIABLE MODEL (MODEL 9)

FIG. 3: VIF PLOT FOR FIVE VARIABLE MODEL (MODEL 9)

CONCLUSIONS: On the basis of above discussion following conclusions can be drawn.

1. E1u along with MATS8p, RDF035e, RDF070u and Mor30v are suitable parameters for  modeling the anti-bacterial activity of present set compounds
2. Coefficients for E1u, MATS8p, RDF035e and RDF070u are positive suggesting that higher value of these parameters will favour the biological activity. Negative coefficient of Mor30v suggests that it has retarding effect towards log BA values, hence in future designing of potent compounds its lower value will give better results.

ACKNOWLEDGEMENT: One of the authors (Anita K.) is thankful to UGC for providing financial assistance (F. MS - 132/102003/11-12/ CRO (2)) to this work.

REFERENCES:

1. Vidal Reference Book, Medicines in Russia, Nikolaeva N. B., Al perovich B. R., and Sozinov V. N. (eds.), Moskow: AstraFarmServis 1996 (In Russian).
2. Negwer M., Scharnow H-G., Organic-chemical drugs and their synonyms, Wiley-VCH, Weinheim 2002.
3. Yuan H., Parrill A. L., J. Mol. Struct.-Theochem. 2000, 1–3,273 – 282.
4. Lesyk R. B., Zimenkovsky B. S., Curr. Org. Chem. 2004, 8, 1547 – 1577.
5. Lesyk R. B., Zimenkovsky B. S., Kutsyk R. V., Atamanyuk D. V., Semenciv H. M., Pharm. J. 2003, 2, 52 – 56 (in Ukrainian).
6. Kuz_min V. E., Artemenko A. G., Polischuk P. G., J. Mol.Mod. 2005, 11, 457 – 467.
7. Kuz_min V. E., Artemenko A. G., Muratov E. N., Volineckaya I. L., Makarov V. A., Riabova O. B., Wutzler P., Schmidtke M., J. Med. Chem. 2007, 17, 4205 – 4213.
8. Oleg, A. Costescu, M.V. Diudea, B. Parv, QSAR modeling of antifungal activity of some heterocyclic compounds, CROATICA CHEMICA ACTA, 2006,79, (3), 483, 17-20.
9. Khadikar P. V., Karmarkar S. and Agrawal V. K., A Novel PI index and its applications to QSPR/QSAR studies, J. Chem. Inf. Comput. Sci., 2001, 41, 934-949.
10. Srivastava A.K., Pathak V.K., Archana, Jaiswal M., Agrawal V.K. Qsar Analysis of Mur B Inhibitors with Antibacterial Properties Discussing Role of Physico-chemical Parameters Med. Chem. Res., 2011, 20(9), 1713-1723.
11. Louis B., Agrawal V. K., Quantitative structure-pharmacokinetic relationship (QSPkR) analysis of volume of distribution values of anti-infective agents from J group of the ATC classification in humans, Acta Pharma, 2012, 62, 305-323.
12. Gupta D., Agrawal V. K., Singh J., Shaik B., QSAR Study On 3-Azolylmethylindoles as Anti-Leishmanial Agents, J. Eng. Sci. Mangt. Edu., 2010 1, 62-69.
13. Anatoliy G. Artemenkoa, Eugene N. Muratova, b, Dmytro V. Atamanyukc, Victor E. Kuz_mina, Alexander I. Hromova, Roman V. Kutsykd and Roman B. Lesykc QSAR Analysis of Antimicrobial Activity of 4-thiazolidone Derivatives QSAR Comb. Sci. 28, 2009, No. 2, 194 – 205.
14. NCSS, Kaysville Utah, www.ncss.com
15. Chaterjee S., Hadi A.S., Price B, (2000), Regression Analysis by Examples, 3rd Ed.  Wiley: New York.
16. R. Todeschini, M. Lasagni, E. Marengo, J. Chemom. 1994, 8, 263-273; R. Todeschini, P. Gramatica, 3D QSAR in Drug Design - Vol. 2, H. Kubinyi, G. Folkers, Y. C. Martin (Eds.), Kluwer/ESCOM, Dordrecht (The Netherlands), 1998, 355-380.
17. Gramatica, P.,Navas, N. &Todeschini, R.Chemom. Intell. Lab. Syst., 1998, 40, 53-63.
18. Gramatica, P., Consonni, V. & Todeschini, R. Chemosphere, 1999, 38, 1371 -1378.
19. Gramatica, P., Navas, N. & Todeschini, R. TRAC. 1999, 18, 461-471.
20. Gramatica, P., Corradi, M. & Consonni, V. Chemosphere, 2000, 41,763-777.
21. Broto,P., Moreau,G.&Vandicke,C., Eur.J.Med.Chem, 1984, 19, 79-84.
22. Karcher,W., Devillers, J.Eds.,Kluwer Academicpublishers, Dordrecht, 1990; 105-127.
23. Hemmer, M. C., Steinhauer, V. & Gasteiger, J. Vibrat.Spect., 1999, 19, 151 -164.
24. Schuur J.H., Selzer P., Gasteiger J., J. Am. Chem. Soc. 1996, 36, 334-344.
25. Gasteiger, J., Sadowski, J., Schuur, J., Selzer, P., Steinhauer, L. & Steinhauer, V.,
    J .Chem Inf .Comput. Sci., 1996, 36, 1030-1037.
26. Schuur, J. & Gasteiger, J., Anal.Chem., 1997, 83, 2398-2405.
27. Pogliani L., Structure property relationships of amino acids and some dipeptides, Amino Acids, 1994, 6, 141-153.
28. Pogliani L., Modeling with Special Descriptors Derived from a Medium-Sized Set of Connectivity Indices, J. Phys. Chem., 1996,100, 18065-18077.
    

    ---

    How to cite this article:

    Anita K, Agrawal VK, Shaik B and Sharma s: Modeling antibacterial activity of 4-thiozolidone derivatives. Int J Pharm Sci Res2014; 5(8): 3333-41.doi: 10.13040/IJPSR.0975-8232.5(8).3333-41

    ---

    All © 2014 are reserved by International Journal of Pharmaceutical Sciences and Research. This Journal licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 3.0 Unported License.