Predicting the AntiInflammatory Activity of Novel 5Phenylsulfamoyl2(2Nitroxy) (Acetoxy) Benzoic acid derivatives using 2D and 3DQSAR (kNNMFA) Analysis
Kiran Madhawai*, Dr. Dinesh Rishipathak, Dr. Santosh Chhajed, Dr. Sanjay Kshirsagar
Department of Pharmaceutical Chemistry, Bhujbal Knowledge City, MET’S IOP, Adgaon, Nashik422003, India.
*Corresponding Author Email: kiranmadhawai777@gmail.com
The quantitative structure–activity relationship (QSAR) analyses were carried out for a series of new side chain modified 5phenylsulfamoyl2(2nitroxy) (acetoxy) benzoic acid derivatives to find out the structural requirements of their antiinflammatory activities. The statistically significant best 2D QSAR models for antiinflammatory activity having correlation coefficient (r^{2}) = 0.897 and cross validated squared correlation coefficient (q^{2}) = 0.701 with external predictive ability (pred_r^{2}) = 0.390 were developed by multiple linear regression coupled with genetic algorithm (GA–MLR) and stepwise (SW–MLR) forward algorithm, respectively. The results of the present study may be useful on the designing of more potent analogues as antimalarial agents.
KEYWORDS: QSAR, 5phenylsulfamoyl2(2nitroxy) (acetoxy) benzoic acid, Antiinflammatory, MLR, Genetic algorithm.
1. INTRODUCTION:
Most NSAIDs are weak organic acids. Once absorbed, they get bound to serum albumin .Due to increased vascular permeability in localized sites of inflammation; this high degree of protein binding may result in the delivery of higher levels of NSAIDs. They are a chemically diverse group.
The primary effect of NSAIDs is to inhibit COX enzyme, thereby blocking the transformation of arachidonic acid to prostaglandins, prostacyclin and thromboxanes. These result in complex effects on vascular permeability and platelet aggregation, undoubtfully contributing to the overall clinical effects of these compounds (Abadi at all 2005).
COX 1, or prostaglandin synthase H, is a house keeping enzyme that regulates normal cellular functions and is stimulated by hormones and growth factors. It is constitutively expressed in most tissues and is inhibited by NSAIDs in varying degrees. COX 1 is important in maintaining the integrity of the gastric and duodenal mucosa and many of the side effects of NSAIDs on the gastrointestinal tract are attributed to its inhibition (Furst and Ulrich, 2007).
COX 2 or prostaglandin synthase H_{2 }is an inducible enzyme and is usually not detectable in most tissues. Its expression is increased during states of inflammation or experimentally in response to mitogenic stimuli. Its expression is inhibited by glucocorticoids. COX2 is also inhibited by all of the presently used NSAIDs, to a lesser or greater degree. Thus, differences in the effectiveness with which a particular NSAID inhibits an isoform of COX may affect with its activity and its potential toxicity. It has been proposed that the ideal NSAID would inhibit the inducible COX2 alone, without having any effect on COX1.
Our interest is in developing new effective medications as antiinflammatory agent because of less toxicity in GIT. Pharmacological evaluation shows the NO2 release of compound will decrease the GIT toxicity (Bandarage, et al, 2000).
The purpose of the present study is to investigate the physicochemical parameters responsible for the antiinflammatory effect of 5phenylsulfamoyl2(2(nitrooxy) (acetoxy) benzoic acid derivatives, explore the correlation between them and to obtain more information for designing novel substituted 5phenylsulfamoyl2(2(nitrooxy) (acetoxy) benzoic acid derivatives with potent protective activity. In the present investigation, three widely used techniques, viz. stepwise forward variable selection method, Genetic algorithm and simulated annealing have been applied for descriptor optimization and multiple linear regression analysis, principal component regression and partial least square has been applied for two and threedimensional QSAR models development. The generated models provide insight into the influence of various interactive fields on the activity and, thus, can help in designing and forecasting the protecting effect of novel 5phenylsulfamoyl2(2(nitrooxy) (acetoxy)benzoic acid molecules (Bhandari, et al, 2009).
A series of 5phenylsulfamoyl2(2(nitrooxy) (acetoxy)benzoic acid analogues which were reported are chosen for QSAR study in order to establish quantitative relationship between physiochemical properties and biological activities of the compounds using MDS software (VlifeScience)
All molecular modelling studies (2D and 3D) were performed using the Molecular Design Suite (VLife MDS software package, version 4.6; from VLife Sciences, Pune, India), on a HP computer with a Pentium IV processor and a Windows 7 operating system. Structures were sketched using the 2D draw application and converted to 3D structure
Fig 1: Parent Chemical structure of 5phenylsulfamoyl2(2(nitrooxy) (acetoxy) benzoic acid drivatives
Table 1: Substituent and Biological activity
Sr. No. 
Substitution 
%NO Release 
logIC_{50} 
1 
2methoxy 
20.86 
1.319 
2 
3methoxy 
11.23 
1.050 
3 
4methoxy 
14.31 
1.155 
4 
2nitro 
11.26 
1.651 
5 
3nitro 
13.02 
1.114 
6 
4nitro 
11.51 
1.060 
7 
2fluro 
18.20 
1.260 
8 
4fluro 
17.82 
1.250 
9 
2,4dichloro 
16.34 
1.213 
10 
H 
15.74 
1.197 
11 
4bromo 
12.14 
1.084 
12 
4chloro 
11.25 
1.051 
13 
6methyl 
9.00 
0.954 
14 
5methyl 
14.40 
1.158 
15 
4methyl 
15.95 
1.202 
16 
Naphthalene 
19.90 
1.298 
17 
Hydrazine 
14.80 
1.170 
18 
4carboxyl 
12.30 
1.089 
19 
2carboxyl 
11.31 
1.053 
2.1 Biological data:
The antiinflammatory activity of 5phenylsulfamoyl2(2(nitrooxy) (acetoxy) benzoic acid derivative were taken from the reported work. The total set of compounds was divided into a training set for 2d and 3D QSAR models and a test for validating the quality of the model. Selection of training and test set was done on the basis of structural diversity and wide range of activity such that the test set molecules represent a range of biological activity similar to that of the training set; thus, the test set is truly representative of the training set. The biological activity values [IC_{50} (µM)] reported in micromolar units were converted to their molar units and then further to negative logarithmic scale and subsequently used as the dependent variable for the QSAR analysis. The values of IC50 along with the structure of the compounds in the series are listed in Table 1.
2.2. Molecular modeling for 2D QSARS:
In 2D QSAR analysis, significant methods Multiple linear regression, principle component regression and partial least square were applied to generate the2DQSAR model. The 2D structures were converted to 3D structures by sending them to MDS software. Each compound was energy minimized and batch optimized by using Merck Molecular Force Field and charges followed by Austin Model1. Hamiltonian method was available in MOPAC module with the convergence criterion 0.001 kcal/mol A fixing Root Mean Square Gradients (RMS) to 0.001 kcal/mol.2D descriptors (physicochemical and alignment independent) were calculated for the optimized compounds on QSAR pus worksheet. The invariable descriptors (the descriptors that are constant for all the molecules) were removed, as they do not contribute to QSAR .Most stable structure for each compound was generated after energy minimization and used for calculating various physicochemical descriptors like thermodynamic, steric and electronic. The energyminimized geometry was used for the calculation of the various 2D descriptors (individual, Chi, ChiV, pathcount, chi Chain, ChiV Chain, Chain path count, cluster, path cluster, kappa, Element count, Estate contributions, semiempirical, Hydrophilichydrophobic, polar surface area and Alignment independent) and was considered as independent variables in the present study.
QSAR analysis was performed after removal of all the invariable columns, as they do not contribute to the QSAR. The optimal test and training data set were generated using the manual as well as random data selection method. Sphere exclusion method is used for creating training and test set from the data. All 19 molecules were subjected to regression analysis using Multiple linear regression analysis, as model building methods coupled with stepward forward backward variable selection method. Regression analysis was carried out for treatment of drug abuse disorders and the best model was crossvalidated. Best two dimensional QSAR results obtained by multiple linear regression analysis (using random and manual data selection method), partial least squares regression are obtained by the following table.
Table 2; 2D QSAR model optimization by Multiple Linear Regression analysis
Trials 
r^{2} 
q^{2} 
r^{2}se 
q^{2}se 
Pred_r^{2} 
Pred_r^{2}se 
F test 
1(Model13) 
0.897 
0.701 
0.047 
0.019 
0.309 
0.1262 
14.19 
2(Model14) 
0.878 
0.764 
0.040 
0.085 
0.129 
0.114 
18.84 
3(Model15) 
0.799 
0.741 
0.094 
0.147 
0.292 
0.106 
20.94 
The selection of the best model is based on the values of r^{2} (squared correlation coefficient), q^{2} (crossvalidated correlation coefficient), pred_r^{2} (predicted correlation coefficient for the external test set), F (Fisher ratio) value. High values of the Ftest indicated that the model was statistically significant. r^{2}se, q^{2}se and pred_r^{2}se are the standard errors terms for r^{2}, q^{2} and pred_r^{2} respectively.
2.3. Three dimensional (3D) QSAR studies:
In the kNNMFA method three models were generated for the selected members of training and test sets, and the corresponding best two models are reported here in. VLife Molecular Design Suite 4.6 allows user to choose probe, grid size and grid interval for the generation of descriptors. The variable selection methods along with the corresponding parameters are allow to be choosen, and optimum models are generated by maximizing q^{2.}. KNearest neighbor molecular field analysis requires suitable alignment of given set of molecules. To derive the kNNMFA descriptor fields, a 3D cubic lattice with grid spacing of 2A^{0 }and a charge of + 1.0 with default but energy 30Kcal/mole to generate stearic field, electrostatic, hydrophobic fields. The 3D QSAR studies were performed by kNNMFA using stepwise forward backward, simulated annealing selection method and genetic algorithm method. The software produced more than 1568 descriptors are prior to module development descriptors having zero values or same values were removed which result in more than total 2500 descriptors for all the compounds in separate columns. This algorithm allows constructing training sets covering all descriptors space areas occupied by representative points. KNNMFA with simulated annealing and stepwise variable selection was employed for selection of variables to obtain the QSAR models.
2.4. kNearest neighbor molecular field analysis (kNNMFA):
The kNN methodology relies on a simple distance learning approach whereby an unknown member is classified according to the majority of its kNN in training set. The nearness is measured by an appropriate distance metric. The 3D QSAR studies were performed by kNNMFA using stepwise forward backward variable selection method. In this method the crosscorrelation limit set to 0.5 and term selection criterion as q^{2}. Ftest ‘in’ was set to 4.0, and F test ‘out’ to 3.99. As some additional parameters, variance cutoff was set at 0 kcal/mol and scaling to auto scaling; additionally, kNN parameter setting was done within the range of 25 and prediction method was selected as the distancebased weighted average.
The model was derived by clicking OK, after all the parameters have been set. Once the significant model is obtained, its fitness plot and contour plot is saved.
Table 3 : Best results of 3DQSAR models obtained by kNNMFA method
Trials 
kNN 
DOF 
q^{2} 
q^{2}_se 
pred_r^{2} 
pred_r^{2}se 
1(Model16) 
3 
10 
0.8170 
0.2667 
0.7773 
0.2815 
2(Model17) 
2 
9 
0.7121 
0.2951 
0.6754 
0.3759 
3(Model18) 
2 
10 
0.7979 
0.2638 
0.5119 
0.4780 
2.5. Development and validation of QSAR models:
The developed QSAR models are evaluated using the following statistical measures: n, (the number of compounds in regression); k, (number of variables); DF, (degree of freedom); optimum component, (number of optimum PLS components in the model); r^{2}, (the squared correlation coefficient); r^{2}se, (standard error of squared correlation coefficient); F test, (Fischer’s value) for statistical significance; q^{2}, (crossvalidated correlation coefficient); q^{2}_se, (standard error of crossvalidated square correlation coefficient); pred_r^{2}, (r^{2} for external test set); pred_r^{2}se, (standard error of predicted squared regression); Z score, (Z score calculated by the randomization test); best_ran_q^{2}, (highest q^{2} value in the randomization test); best_ran_r^{2}, (highest r^{2} value in the randomization test). The regression coefficient r^{2} is a relative measure of fit by the regression equation. It represents the part of the variation in the observed data that is explained by the regression. However, a QSAR model is considered to be predictive, if the following conditions are satisfied: r^{2} > 0.6, q^{2} > 0.6 and pred_r^{2} > 0. The Ftest reflects the ratio of the variance explained by the model and the variance due to the error in the regression. High values of the Ftest indicate that the model is statistically significant. The low standard error of r^{2} (r^{2}_se), q^{2} (q^{2}_se) and pred_r^{2} (Pred_r^{2}se) shows absolute quality of fitness of the model. Internal validation was carried out using ‘leaveoneout’ (q^{2}, LOO) method. The crossvalidated coefficient, q^{2}, was calculated using the following equation:
Where y_{i}, and are the actual and predicted activity of the ith molecule in the training set, respectively, and y_{mean} is the average activity of all molecules in the training set.
However, a high q^{2} value does not necessarily give a suitable representation of the real predictive power of the model for antimalarial ligands. So, an external validation was also carried out in the present study. The external predictive power of the model was assessed by predicting pIC_{50} value of the nine test set molecules, which were not included in the QSAR model development. The predictive ability of the selected model was also confirmed by pred_r^{2}.
Where y_{i} and ŷ_{i} are the actual and predicted activity of the ith molecule in the test set, respectively, and y_{mean} is the average activity of all molecules in the training set.
To evaluate the statistical significance of the QSAR model or an actual data set, we have employed a onetail hypothesis testing.
3. RESULTS AND DISCUSSION:
The importance and utility of the new 2D and 3D QSAR method discussed has been established by applying it to known sets of molecules as described above. The importance and utility of the new 2D and 3D QSAR method discussed has been established by applying it to known sets of molecules as described above.
All the calculated descriptors were considered as independent variable and biological activity as dependent variable. In 2D QSAR analysis, significant methods like Multiple linear regression analysis, Partial Least Square (PLS) and Principal Component Regression (PCR) were applied to generate the model having good q2 and pred_r2 values, one of which was selected having good internal and external predictivity. Selection of training and test set was by Manual data selection and random data selection method. Training and test set were selected if they follow the unicolumn statistics, i.e. maximum of the test is less than maximum of training set and minimum of the test set is greater than of training set, which is prerequisite for further QSAR analysis. This result shows that the test is interpolative i.e., derived from the minmax range of training set. The mean and standard deviation of the training and test set provides insight to the relative difference of mean and point density distribution of the two sets the QSAR models developed by kNNMFA include both the electrostatic, steric descriptors along with their range to indicate their importance for interaction in molecular field. QSAR investigations of the substituted 5Nphenylsulfamoyl2(2(nitrooxy) (acetoxy) benzoic acid derivatives series resulted in several QSAR equations.
3.1 2DQSAR model:
Model 13:
By Random data selection method;
Model13 (Test set: 2, 3, 4, 8, 16, and 19)
pIC_{50} = 0.1801 (SaasC count)  0.2101 (T_N_N_3) + 0.2741(T_N_N_7) + 1.906
Statistics:
[n= 13; Degree of freedom= 9; r^{2} =0.897; q^{2}=0.701; F test=14.19; r^{2}se=0.047; q^{2}se= 0.019; pred_r^{2}= 0.309; pred_r^{2}se = 0.126]
From model 13 explains 89.7% (r^{2 }= 0.897) of the total variance in the training set as well as it has internal (q^{2}) and external (pred_r^{2}) predictive ability of 70.16 % and 12.6 % respectively. The F test shows the statistical significance of 99.99 % of the model which means that probability of failure of the model is 1 in 10000. In addition, the randomization test shows confidence of 99.9999 (Alpha Rand Pred R^2 = 0.00000) that the generated model is not random and hence chosen as the QSAR model. The Ftest=14.19 which is greater than the tabulated value 2.93766 (Bolton 2004).
From QSAR model 13 it is revealed that,
1. Negative coefficient value of T_N_N_3 [This is the count of number of Nitrogen atoms (single double or triple bonded) separated from any other Nitrogen atom (single double or triple bonded) by 3 bonds in a molecule.] on the biological activity indicated that lower values leads to good inhibitory activity while higher value leads to reduced inhibitory activity.
2. Positive coefficient value of T_N_N_7 [This is the count of number of nitrogen atoms (single double or triple bonded) separated from any nitrogen atom (single or double bonded) by 7 bond distance in a molecule.] on the activity indicated that higher value leads to better inhibitory activity whereas lower value leads to decrease inhibitory activity.
3. Negative coefficient value of SaasC count [This descriptor signifies the total no of carbon connected with one single bond along with two aromatic bonds.]On the biological activity indicated that lower values leads to good activity while higher value leads to reduced activity.
Contribution chart Data fitness plot and activity of training and test set for model 13 is represented in Fig.2, 3, 4 and 5 respectively.
Fig. 2 contribution plot
Fig.3 Data fitness plot
Fig.4 Training set activity
Fig.5 Test set activity (2, 3, 4, 8, 16, 19)
3.2.3DQSAR model:
Model16 (Test set: 11, 12, 14, 18, and 19)
pIC_{50}= E_397 (2.4827,0.4193) + S_469 (0.0643, 0.1358)
Statistics: [kNN= 3; n = 14; DOF= 10; q^{2}= 0.8170; q^{2}_se = 0.2667; pred_r^{2 }= 0.7773; pred_r^{2}se = 0.2815]
From QSAR model 16 it is revealed that,
The model 16 explains values of k (3), q^{2} (0.8170), pred_r^{2} (0.7773), q^{2}_se (0.2667), and pred_r^{2} se (0.2815) prove that QSAR equation so obtained is statistically significant and shows the predictive power of the model is 81.70% (internal validation).
The data fitness plot for model 16 is shown in Fig.7.The plot of observed vs. predicted activity Fig. 8,9 provides an idea about how well the model was trained and how well it predicts the activity of the external test set.
Fig. 6 Plot of contribution chart
Fig. 7 Data fitness plot
Fig. 8 Training Set
Fig. 9 Test Set (11, 12, 14, 18, 19)
Steric field, S_469 (0.0643, 0.1358) positive steric potential is present around the field which indicates the favorability of bulky groups on the salicylic acid ring to increase the activity.
Electrostatic field, E_937 (2.4827,0.4193) negative coefficient suggested that electronegative substituent may be favorable on the position of salicylic acid ring for better activity.
4. CONCLUSION:
In the present investigation, all proposed QSAR models were statistically significant, thus, from above QSAR investigations it could be concluded that 2D/3D descriptors properties of substituted 5phenylsulfamoyl2(2(nitrooxy) (acetoxy)benzoic acid derivatives are mainly involved in treatment of drug abuse disorders. The good correlation between experimental and predicted biological activity for compounds in the test set further highlights the reliability of the constructed QSAR model. The requirements for the more potent biological activity are explored with 2D, 3D and group based QSAR studies. The 2D technique indicates the importance of XlogP, SssCH3Eindex, SssCH3count, and SlogP of the compounds on the activity. The 3D QSAR analysis makes it possible to relate chemical structures of ligands and their binding affinity with respect to different bio targets by using the kNNMFA techniques. Thus it provides a direct view of factors expressed in terms of molecular fields (electrostatic, steric) affecting the binding affinity. This in turn could give the reasonably good prediction of binding affinity. The location and range of function values at the field points selected by the models provide clues for the design of new molecules. Hence, this method is expected to provide a good alternative for the drug design. The 3DQSAR model showed that electrostatic effects dominantly determine the binding affinities and these QSAR models developed in this study would be useful for the development of new drugs as a medicament for the drug abuse disorder. The 2DQSAR studies revealed that alignment independent descriptors were the major contributing descriptors. The descriptor values obtained in this study helped in quantification of the structural features of 5phenylsulfamoyl2 (2(nitrooxy) (acetoxy) benzoic acid derivative.
After successful QSAR studies, attempts were made to predict the activities of the newly designed analogues of these reported compounds. Above series of 5phenylsulfamoyl2(2nitroxy) (acetoxy) benzoic acid derivatives and some newer compounds were suggested for synthesis on the basis of selected model.
In future we can synthesize these designed compounds using the selected scheme and confirm their activity by carrying out in vivo evaluation.
Table 4: Newly Designed 5phenylsulafamoyl2(2nitroxy) (acetoxy) benzoic acid molecules with predicted activities
Sr. No. 
Newly Designed Molecules 
logIC_{50} 
1 

1.15715

2 

1.15714

3 

1.5715

4 

1.1221

5. ACKNOWLEDGMENT:
The author wishes to express gratitude to Vlife sciences Technologies Pvt. Ltd. for providing the software for the study. Also the authors are thankful to the trustee, Bhujbal Knowledge City for providing the necessary facilities to carry out the research work.
6. REFERENCES:
1. H. Abadi, et al. (2005), Synthesis of novel 4substituted7trifluoromethylquinoline derivatives with nitric oxide releasing properties and their evaluation as analgesic and antiinflammatory agents, Bioorganic and Medicinal Chemistry , 13, pp. 5759–5765.
2. U. K. Bandarage, et al. (2000) Nitrosothiol esters of diclofenac synthesis and pharmacological characterization as gastrointestinalsparing prodrugs, Journal of Medicinal Chemistry, 43, pp. 400516.
3. S. V. Bhandari, et al. (2009) Design, synthesis and pharmacological screening of novel nitric oxide donors containing 1,5diarylpyrazolin3one as nontoxic NSAIDs, European Journal of Medicinal Chemistry, 44, pp.4622–4636
4. Ajmani, S., Jadhav, K., and Kulkarni, S. (2006), Threedimensional QSAR using the knearest neighbor method and its interpretation. Journal of chemical information and modeling, 46, pp. 2431.
5. BOLTON, S. and BON, C. (2004), Pharmaceutical Statistics Practical and Clinical Application 4th Ed. New York, USA: Marcel Dekker Inc.
6. C. Cena, et al. (2003) Antiinflammatory, Gastro sparing, and Antiplatelet Properties of New NODonor Esters of Aspirin, Journal of Medicinal Chemistry, 46, pp. 747 54.
7. M. A. Chowdhury, et al. (2010), Celecoxib analogs possessing a N(4nitrooxybutyl) piperidin4yl or N(4 nitrooxybutyl)1,2,3,6tetrahydropyridin4yl nitric oxide donor moiety: Synthesis, biological evaluation and nitric oxide release studies, Bioorganic and Medicinal Chemistry Letters , 20, pp.1324–1329
8. M. A. Chowdhury, et al. (2008), Synthesis of new 4[2(4methyl (amino) sulfonylphenyl)5trifluoromethyl2Hpyrazol3yl]1, 2, 3, 6tetrahydropyridines: A search for novel nitric oxide donor antiinflammatory agents, Bioorganic and Medicinal Chemistry, 16, pp.8882–88
9. B. C. Daniel, et al. (2008), Nonsteroidal Antiinflammatory Agents – Benefits and New Developments for Cancer Pain, Supportive Oncology, pp. 1822.
10. J. E Downing, et a.l (2005) Gastric and thymic assay of acute oral treatment of rats with nitric oxide esters of ibuprofen or indomethacin, Biochem. Biophys. Res. Commun.
11. O. O. Fadeyi, et al (2004) Antipyretic, analgesic, antiinflammatory and cytotoxic effects of four derivatives of salicylic acid and anthranilic acid in mice and rats, African Journal of Biotechnology, 3(8), pp. 426431
12. S. Fiorucci, J. L. Wallace, R. N. Dubois, et al.(2000) New antiinflammatory agents: NONSAIDS and COX2 inhibitors. International conference on “Advances in Prostaglandin and Leukotriene Research: Basic Science and New Clinical Applications”, June (48), pp. 2.
13. S. Fioruccia, et al. (2007), NSAIDs, coxibs, CINOD and H2Sreleasing NSAIDs: What lies beyond the horizon, Digestive and Liver Disease, 39, pp. 1043–1051.
Received on 08.06.2017 Accepted on 12.08.2017
© Asian Pharma Press All Right Reserved
Asian J. Res. Pharm. Sci. 2017; 7(4): 227234.
DOI: 10.5958/22315659.2017.00036.4