D-QSAR Analysis of Oxadiazole Substituted α-Isopropoxy phenylpropionic Acids as PPAR-α & PPAR-γ Agonists

A quantitative structure activity relationship study on a series of oxadiazole substituted α-isopropoxy phenylpropionic acids with activity on PPAR-α and PPAR-γ was made using combination of various physiochemical descriptors. Several statistical regression expressions were obtained using stepwise multiple linear regression analysis. The best quantitative structure activity relationship model was further validated by leave one out cross validation method. Steric parameter (molar refractivity) was found to have significant correlationship with PPAR-γ agonist activity and hydrophobic (Hansch substituent constant), electronic parameter (field effect) were found to have significant correlationship with PPAR-α agonist activity. The increment in the number of carbon atom (indicative variable) between the oxadiazole tail and central phenoxy moiety increases the PPAR-γ agonist activity whereas decreases the PPAR-α agonist activity.


Introduction Introduction
Resistance of peripheral tissue to the action of insulin is a characteristic feature of non-insulin dependent diabetes mellitus (NIDDM).A number of prospective studies have shown that the development of insulin resistance is an early event in the natural progression of the disease 1 .Untreated NIDDM leads to several chronic diseases such as retinopathy, nephropathy, neuropathy, and cardiovascular disease 2 .The latter lead to increase in mortality.Treatment of NIDDM currently centers on biguanides, sulphonylureas, α-glucosidase inhibitor, insulin sensetizer, and insulin secretagogues.Moreover search for new agents is ongoing.One novel class of antidiabetic agents that appear to be effective as a treatment for diabetes is peroxisome proliferator-activated receptor (PPAR) agonists such as thiazolidine 2,4-diones (TZDs) represented by ciglitazone, englitazone, pioglitazone and rosiglitazone 3 .PPARs are the orphan members of the nuclear receptor gene family of ligand activated transcription factors.Three subtypes of PPARs have been cloned from mouse and human: PPAR-α, PPAR-γ, and PPAR-δ.The PPARs are belived to play a physiological role in regulation of glucose and lipid metabolism.TZD as PPAR-γ agonists have been shown to enhance sensitivity of targeted tissue to insulin and to reduce plasma glucose, and insulin level in animal models of type 2 diabetes as well as human.The hypolipidemic fibrate drugs are an important class of PPAR-α ligands, which show the role of this receptor in mediating the lipid lowering activity 4 .This is of interest since there is also a high frequency of concurrent dyslipidemia (high levels of triglyserides and low level of HDL-c) with type 2 diabetes, which may contribute to accelerated coronary atherosclarosis, the leading cause of death in these individuals 5,6 .Treatment of insulin resistance and dyslipidemia would provide the great benefit of type 2 diabetic patient.
Despite the growing evidence of positive effects of fibrate (PPAR-α agonist) treatment in type 2 diabetic patients, most reports have been on the identification of selective PPAR-γ agonists.Only a few reports have dealt with selective PPAR-α agonists 7 , and even fewer compounds have been reported to have both PPAR-γ and PPAR-α agonist activity [8][9][10] , e.g.KRP-297 and (-) DRF2725 (Fig. 1).We therefore decided to study quantitative structure activity relationship (QSAR) of such dual acting receptor agonists [11][12][13] .The aim of this work was therefore to identify the associated molecular properties and exploit to optimize PPAR-α and PPAR-γ agonist activity.

Experimental Experimental
The transient transfection data of PPAR agonists oxadiazole substituted α-isopropoxy phenylpropionic acids with activity on PPAR-α and PPAR-γ (Table 1) were taken from the reported work of Liu et al 13 .The biological activity data (EC 50 in nM) was converted to negative logarithmic dose (pEC 50 ) for QSAR analysis.The correlation were sought between PPAR agonists activity and various physicochemical parameters such as hydrophobic (ð), steric (Molar refractivity or MR), hydrogen acceptor (HA), hydrogen donar (HD) and electronic (field effect or F, resonance or R, Hammett's constant or ó), taking into account the comments of Hansch et al 14 .
Stepwise multiple linear regression analysis method 15,16 was used to perform QSAR analysis.Following statistical parameters were considered to select the statistical significance QSAR models: correlation coefficient (r), standard deviation (s), F-test, and leave one out (LOO) cross-validated squared correlation coefficient (Q 2 ) 17 .

Results and Discussion
The series has two different subsets, subset A (compound no. 1 to 15) and subset B (comp.No. 16 to 23), these are different only in the number of carbon atoms (one for A two for B) between oxadiazole tail and the central phenoxy moiety.In subset A only 14 compounds could be subjected to 2D-QSAR analysis because of non-availability of physicochemical parameter for substituent of comp.no.15.Hence, in series, the combined form of 2 subsets, only 22 compounds could be subjected to 2D-QSAR analysis excluding comp no. 15 because of the above reason.When the subsets of compounds were subjected to stepwise multiple linear regression analysis, the following significant equation was obtained for PPAR-γ agonist activity of subset A. Inter-correlation within the parameter is considerably vary less i.e. 0.1(table 2) pEC 50 = 0.087 * MR -0.376 * H a -2.104 n = 14, r = 0.895, r 2 = 0.802 , F=22.248, s = 0.256 Model-1 The subset B gave the various equations but no single equation is significant for the PPAR-γ agonist activity although the model that shows highest co-relation among them is pEC 50 = 0.773 * σ -1.338 n = 8, r = 0.325, r 2 = 0.106 , F = 0.708, s = 0.645 Model-2 The models, which were considered for PPAR-α agonist activity for subset A and B, are given below respectively.Inter-correlation within the parameter is less than 0.2 (table 2) The two subsets were combined together by using indicative variable (I v ) and resultant series of 22 compounds was subjected to stepwise multiple linear regression analysis, the di and tri-parametric models gave statistically significant correlations.Based on correlation coefficient (r) the following model was selected for PPAR-γ agonist activity.
pEC 50 = 0.0753( ± 0.044)*MR +0.650( ± 0.432)*I v -2.055( ± 0.391) n=22,r=0.701,r 2 =0.492,s=0.450,F=9.192 Model-5 The model is unable to explain the activity of two compounds based on z-score values > |2.5| hence after removing these compounds the better correlation equation was obtained with correlation coefficient r>0.874 (model 6).The inter-correlation within the parameters is also less than 0.2 , which suggested the non-dependency of the parameter.pEC 50 = 0.0850(±0.028)*MR+0.692(±0.296)*Iv -2.120(±0.249)n=20,r=0.874,r 2 =0.765,s=0.282,F=27.625 Model-6 Leave one out cross validation method was used for predictivity of model-6.The value for Q 2 in the biological activity test revealed that the results were not based on chance correlation.The predicted activity data and graph of observed verses predicted activities of the compounds for model-6 are shown in table 4  The study revealed that steric parameter (molar refractivity) contributed positively to the activity and increase in carbon chain length from one carbon to two carbon i.e. presence of indicator variable are very important for PPAR-γ agonist activity.The data showed overall significance level better than 99.9% as it exceeded the tabulated F (2,17 α 0.001) = 18.4.
The model, which was considered for PPAR-α agonist activity was having better correlation coefficient r>0.941 (model 7) as well as inter-correlation within the parameters is also less than 0.220 suggested the non-dependency (Table -3 Leave one out cross validation method was used for predictivity of model-7.The value for Q 2 in the biological activity test revealed that the results were not based on chance correlation.The correlation matrix and predicted activity data for model-7 are given in table-3 and 4 where as fig. 3 display the graph of observed verses predicted activities of the compounds.The result of LOO crossvalidated method is: Q 2 : 0.807 S PRESS : 0.309 S DEP : 0.280 The study revealed that hydrophobicity (hansch substituent constant (π)) and electronic (field effect (F)) parameters positively contribute to the biological activity whereas decrease in carbon chain length from two carbon to one carbon i.e. absence of indicator variable (I v ) are responsible for PPAR-α agonist activity.The data showed overall significance level better than 99.9% as it exceeded the tabulated F (3,18 α 0.001) = 9.69.
Study reveled that the increment in the number of carbon atom between the oxadiazole tail and central phenoxy moiety affects the activity.The increment in the carbon atom increases the PPAR-γ agonist activity whereas decreases the PPAR-α agonist activity.