Computational modeling of 4-Phenoxynicotinamide and 4-Phenoxypyrimidine-5-carboxamide derivatives as potent anti-diabetic agent against TGR5 receptor

Shola Elijah Adeniji; David Ebuka Arthur; Adedirin Oluwaseye

doi:10.1016/j.jksus.2018.03.007

View/Download PDF

Buy Reprints

PDF

Translate this page into:

Original article

32 (

); 102-115

doi:

10.1016/j.jksus.2018.03.007

Computational modeling of 4-Phenoxynicotinamide and 4-Phenoxypyrimidine-5-carboxamide derivatives as potent anti-diabetic agent against TGR5 receptor

Shola Elijah Adeniji^⁎, David Ebuka Arthur, Adedirin Oluwaseye

Department of Chemistry, Ahmadu Bello University, Zaria, Nigeria

⁎Corresponding author. shola4343@gmail.com (Shola Elijah Adeniji)

Received: 2018-2-6, Accepted: 2018-3-17, Published: 2018-01-01

Disclaimer:
This article was originally published by Elsevier and was migrated to Scientific Scholar after the change of Publisher.

Peer review under responsibility of King Saud University.

Abstract

Computational study was carried out to develop a Quantitative structure-activity relationship (QSAR) model and molecular docking studies on 4-Phenoxynicotinamide and 4-Phenoxypyrimidine-5-carboxamide derivatives as potent anti-diabetic agent. Chemical structure of these molecules were optimized with Density Functional Theory (DFT) utilizing the B3LYP with 6-31G^∗ basis set. Five QSAR models were generated using Multi-Linear Regression and Genetic Function Approximation (GFA). Model one was selected as the optimum model and reported based on validation parameters which were found to be statistically significant with squared correlation coefficient (R²) of 0.9460, adjusted squared correlation coefficient (R² adj) value of 0.9352 and cross validation coefficient ( $Q_{cv}^{2}$ ) value of 0.9252. The chosen model was subjected to external validations and the model was found to have (R²test) of 0.8642. Molecular docking studies revealed that the binding affinities of the compounds correlate with their pEC₅₀ and the best compound has binding affinity of −10.4 kcal/mol which formed hydrogen bond and hydrophobic interaction and with amino acid residues of TGR5 receptor. QSAR model generated and molecular docking results propose the direction for the design of new anti-diabetic agent with better activity against TGR5 target site.

Keywords

Anti-diabetic

Applicability domain

Binding affinity

Molecular docking

QSAR

Show Related Articles from PubMed

1 Introduction

Non-insulin-dependent diabetes mellitus (NIDDM) which is usually refers to as type 2 diabetes mellitus (T2DM), is a metabolic disorder characterized by high glucose level in the blood. Wide range of anti-diabetic drugs and treatments are available for this metabolic syndrome problem. Many patients suffering from this type of diabetes are unable to get satisfactory glycemic control with these treatment (Saydah et al., 2004). This led to development and designing of novel drugs with better activities against multi-drug resistance and uncontrolled T2DM (See Fig. 1).

(A) Prepared structure of TGR5 receptor (B) 3D structure of the prepared ligand.

TGR5 is a class of G protein-coupled receptor (GPCR) for bile acids (BAs) which was first identified in 2003. Before its identification, farnesoid X receptor (FXR) was the only known receptor activated by BAs. The TGR5 level among different tissues varies significantly. The highest level of TGR5 is in the gallbladder, moderate level in the intestine, spleen and placenta and low level expression in skeletal muscle and liver (Vassileva et al., 2006). TGR5 activation stimulates Glucagon-like peptide-1 (GLP-1) secretion from intestinal enteroendocrine cells by increasing the intracellular cAMP concentration (Watanabe et al., 2006).

A novel analogue of 4-Phenoxynicotinamide and 4-Phenoxypyrimidine-5-carboxamide derivatives has been reported as potent anti-diabetic agent against TGR5 receptor (Duan et al., 2012). Synthesis of novel molecules are typically developed using a trial and error approach, which is time consuming and costly.

The advent of computational chemistry led to challenges of drug discovery (Cramer et al., 1988). QSAR establish a relationship between various molecular properties of molecules and their experimentally known activities (Ibezim et al., 2009). QSAR technique alongside with molecular docking approach were employed to predict the activities of various compounds and elucidate the specific areas where interaction (steric, electrostatic, hydrogen bond donor, hydrogen bond acceptor and hydrophobic) may decrease or increase the activity of the inhibitor molecules. Few researchers; (Amit et al., 2016; Bajpai and Malik, 2003; Dasoondi et al., 2008; Dieguez-Santana et al., 2017; Dixit and Saxena, 2008 ) have carried QSAR studies to established relationship between some inhibitory compounds and their activities against diabetic mellitus. However molecular docking study has not been emphasis to understand the binding mode and binding interactions between the inhibitory compounds and the target site.

The aim of this research was to build a QSAR model that will predict the activity of 4-Phenoxynicotinamide and 4-Phenoxypyrimidine-5-carboxamide derivatives against type 2 diabetes mellitus (T2DM) and to carry out molecular docking studies to elucidate the kind of interaction existing between the inhibitor compounds and the target site (TGR5).

2 Materials and methods

2.1

2.1 Data collection

Thirty-six (36) molecules of 4-Phenoxynicotinamide and 4-Phenoxypyrimidine-5-carboxamide derivatives as potent and orally efficacious TGR5 agonists that were used in this studies were gotten from the literature (Duan et al., 2012).

2.2

2.2 Biological activities (pEC₅₀)

The biological activities of 4-phenoxypyridine-5-carboxamide and 4-phenoxynicotinamide derivatives against TGR5 of type 2 diabetes mellitus measured in EC₅₀ ( $nM)$ were converted to logarithm unit (pEC₅₀ = −logEC₅₀) in order to increase the linearity activities values and approach normal distribution. The observed structures with their biological activities of the molecules were presented in Table 1.

2.3

2.3 Optimization

The optimizations of the compounds were achieved by employing Density Functional Theory (B3LYP B3LYP/6-31G^∗ basis set) utilizing Spartan 14 Version 1.1.4 software. (Becke, 1993; Lee et al., 1988).

2.4

2.4 Molecular descriptor calculation

The optimized structures were submitted for descriptor calculation. Molecular descriptors for all the thirty-six (36) molecules of the inhibitor compounds were calculated utilizing the PaDEL-Descriptor software V2.20. A total of 1875 molecular descriptors were calculated.

2.5

2.5 Normalization and data pretreatment

The descriptors’ value for all the molecules were normalized using Eq. (1) in order to give each variable the same opportunity at the onset to influence the model (Singh, 2013).

(1)

X = \frac{X_{1} - X_{\min}}{X_{\max} - X_{\min}}

Where Xi is the value of each descriptor for a given molecule, Xmax and Xmin are the maximum and minimum value for each column of descriptors X.

2.6

2.6 Data division

The data set was split into training set and test set by employing Kennard and Stone’s algorithm (Kennard and Stone, 1969). The training set comprises 70% of the data set which were used to build the model while the remaining 30% of the data set (test set) were used to validate the built model.

2.7

2.7 Internal validation of model

Material studio software version 8 was used to determine the internal validation parameters by employing the Genetic Function Approximation (GFA) method. The validation of the built model was evaluated by employing the Friedman formula (LOF) which measured the fitness score of the model. LOF is defined as; (Friedman, 1991) .

(2)

LOF = \frac{SEE}{{(1 - \frac{C + d \times p}{N})}^{2}}

The Standard Error of Estimation (SEE) is equivalent to the models standard deviation. It’s a measure of model quality and a model is said to be a better model if it has low SEE value. SEE is defined by equation below;

(3)

SEE = \sqrt{\frac{{(Y_{\exp} - Y_{pred})}^{2}}{N - P - 1}}

c is the number of terms in the model, N is the number of compound that made up the training set, p is the number of descriptors, d is a user-defined smoothing parameter, (Khaled, 2011).

The correlation coefficient (R²) defines the fraction of the entire variation in the model. The closer the value of R² to 1.0, the stronger the model generated. R² is expressed as:

(4)

R^{2} = 1 - [\frac{\sum {(Y_{\exp - Y_{pred}})}^{2}}{\sum {(Y_{\exp - {\bar{Y}}_{training}})}^{2}}]

Where: ${\bar{Y}}_{training}$ $Y_{exp}$ , and Y_pred are the mean experimental activity, experimental activity and the predicted activity in the training set, respectively.

R² value varies directly with the increase in number of descriptors, thus, R² is not reliable to measure the stability of the model. Therefore, R² is adjusted in order to have a reliable and stable model. The R²_adj is defined as:

(5)

R_{adj}^{2} = \frac{R^{2} - P (n - 1)}{n - p + 1}

Where p and n are number of descriptors in the model and number compounds that made up the training set.

The strength of the QSAR model to predict the activity of a new compound was determined using cross validation test. The cross-validation coefficient ( $Q_{cv}^{2}$ ) is defined as:

(6)

Q_{cv}^{2} = 1 - [\frac{\sum {(Y_{pred - Y_{\exp}})}^{2}}{\sum {(Y_{\exp - {\bar{Y}}_{training}})}^{2}}]

${\bar{Y}}_{training}$ $Y_{exp}$ , and Y_pred are the mean experimental activity, experimental activity and the predicted activity in the training set, respectively.

2.8

2.8 External validation of the model

External validation of the developed model was assessed by the value $R_{test}^{2}$ value. The $R_{test}^{2}$ is defined by as;

(7)

R_{test}^{2} = 1 - \frac{\sum {({Ypred}_{test} - Y_{\exp_{test}})}^{2}}{\sum {({Ypred}_{test} - {\bar{Y}}_{training})}^{2}}

Where ${Ypred}_{test}$ and $Y_{\exp_{test}}$ are the predicted and experimental activity test set. While ${\bar{Y}}_{training}$ is the training set mean values of the experimental activity.

2.9

2.9 Y-Randomization test

To be assured that the built QSAR model is strong, reliable and not obtained by chance, the Y-randomization test was carried out on the compound that made up the training set (Tropsha et al., 2003). For the built QSAR model to robust and reliable, the model is expected to have a low R² and Q² values for several trials. Coefficient of determination (c $R_{p}^{2}$ ) for Y-randomization is another parameter calculated which should be greater than 0.5 for passing this test.

(8)

{c R}_{p}^{2} = R \times {[R^{2} - {(R_{r})}^{2}]}^{2}

c $R_{p}^{2}$ is Coefficient of determination for Y-randomization, R is coefficient of determination for Y-randomization and Rr is average ‘R’ of random models.

2.10

2.10 Evaluation of the applicability domain of the model

The leverage approach was employed in defining and describing the applicability domain of the built QSAR models (Veerasamy et al., 2011). Leverage of a given chemical compound hi, is defined as follows:

(9)

hi = Xi {(X^{T} X)}^{- 1} X_{i}^{T}

Where Xi is training compounds matrix of i. X is the $m \times k$ descriptor matrix of the training set compound. $X^{T}$ is the transpose matrix of X and $X_{i}^{T}$ is the transpose matrix X_i used to build the mode. The warning leverage (h^∗) is the boundary of values for X outliers and is defined as:

(10)

h^{*} = 3 \frac{(d + 1)}{m}

Where m is the number of descriptors and d is the number of compounds that made up the training set.

2.11

2.11 Quality assurance of the model

The fitness, reliability, stability, and predictability of the built models were evaluated by the validation parameters. The minimum recommended value for internal and external validation parameters for a generally acceptable QSAR model (Veerasamy et al., 2011) is presented in Table 2.

Table 2 Generally accepted value for the validation parameters for a given QSAR model.

Parameter	Definition	Recommended value
R²	Coefficient of determination	≥0.6
P_(95%)	Confidence interval at 95% confidence level	<0.05
$Q_{cv}^{2}$	Cross validation coefficient	>0.5
R² - $Q_{cv}^{2}$	Difference between R² and $Q_{cv}^{2}$	≤0.3
N_{ext. test set}	Minimum number of external test set	≥5
c $R_{p}^{2}$	Coefficient of determination for Y-randomization	>0.5

2.12

2.12 Docking studies

Molecular docking study was carried between 4-Phenoxynicotinamide and 4-Phenoxypyrimidine-5-carboxamide derivatives and TGR5 receptor target site. The crystal structure of TGR5 receptor used in the study was obtained from protein data bank. The optimized structures of the 4-Phenoxynicotinamide and 4-Phenoxypyrimidine-5-carboxamide derivatives initially saved as SDF files were converted to PDB files using Spartan 14 Version 1.1.4. The prepared ligands were docked with prepared structure of TGR5 receptor using Autodock Vina incorporated in Pyrx software. The docked results were visualized and analyzed using Discovery Studio Visualizer.

3 Results and discussion

QSAR investigation was carried out to relate the structure activity relationship of 4-Phenoxynicotinamide and 4-Phenoxypyrimidine-5-carboxamide derivatives as potent inhibitor of TGR5.

Experimental and predicted activities of the inhibitors and their derivatives were reported in Table 1. The low residual value between experimental and predicted activity indicates that the model has a high predictive power.

Five descriptors were used to build a linear model for predicting the activities of the inhibitor compounds based on Multi-Linear Regression and Genetic Function Algorithm (GFA) method employed. Model one was selected as the best model due to statistical significance and prominent validation parameters.

Model 1

pEC₅₀ = 0.125973308 ^∗ ATS1p - 0.010504968 ^∗ ATSC1m + 0.097632128 ^∗ X RDF80u - 0.064185438 ^∗ RDF55i - 4.004440770

Model 2

pEC₅₀ = 0.108430945 ^∗ ATS1p - 0.759472864 ^∗ AATSC1m - 0.327839361 ^∗ nHother + 0.311565867 ^∗ RDF80p - 1.845286768

Model 3

pEC₅₀ = 0.127238717 ^∗ RDF80p - 0.010483164 ^∗ ATSC1m - 0.064741508 ^∗ RDF55i + 0.077010983 ^∗ RDF80i - 4.144260897

Model 4

pEC₅₀ = 0.127289414 ^∗ ATS1p - 0.010694802 ^∗ ATSC1m - 0.079815940 ^∗ RDF55u + 0.103816691 ^∗ RDF80u - 4.188809957

Model 5

pEC₅₀ = 0.105099406 ^∗ ATS1p + 28.698005977 ^∗ AATSC0p - 0.068763191 ^∗ RDF55i + 0.080255278 ^∗ RDF80i - 9.976240161

All the validation parameter to confirm the stability and robustness of the model were reported in Table 3 which were all in agreement with validation parameters presented in Table 2. While the calculated descriptors used in predicting the activity of each compound where reported in Tables 4 and 5.

Table 3 Validation parameters for each model using Genetic Function Approximation (GFA).

S/N	Validation parameters	Model 1	Model 2	Model 3	Model 4	Model 5	Threshold value
1	Friedman LOF	0.432202	0.433894	0.435663	0.443398	0.445466	0.5
2	R-squared	0.946014	0.945803	0.945582	0.944616	0.944357	≥0.6
3	Adjusted R-squared	0.935217	0.934963	0.934698	0.933539	0.933229	>0.6
4	Cross validated R-squared	0.928759	0.919402	0.927651	0.926561	0.923921	>0.5
5	Significant Regression	Yes	Yes	Yes	Yes	Yes
6	Replicate points	0	0	0	0	0
7	Computed experimental error	0	0	0	0	0
8	Lack-of-fit points	20	20	20	20	20
9	Min expt. error for non-significant LOF (95%)	0.249492	0.24998	0.250489	0.252702	0.253291
10	R² test	0.8642	0.7211	0.5322	0.5422	0.4542	≥0.6

Table 4 Calculated descriptors for the training set and predicted Activity.

Molecule	ATS1p	ATSC1m	RDF80u	RDF55i	Predicted Activity
10	95.03083	44.03765	15.58872	23.88487	7.493192
11	89.90462	41.47018	16.38505	23.05786	7.005227
12	101.0936	65.57105	15.7974	29.19099	7.710525
13	106.1096	73.3786	17.13241	32.66898	8.167491
14	106.1096	73.3786	10.73848	28.90777	7.784653
15	103.8825	63.66103	19.21257	33.44618	8.142226
17	111.6874	82.05282	14.40201	30.60505	8.644922
18	91.54652	64.2069	9.827982	37.90672	5.379953
2	97.12432	70.75769	22.64375	27.98552	7.901818
22	90.49977	50.7956	9.667698	42.83823	5.056796
23	81.03962	13.29221	10.41906	26.95842	5.351652
25	106.4529	69.43373	0	1	8.612203
26	101.3988	134.1153	0	1	7.296042
28	103.9259	99.87649	0	1	7.974061
29	106.3046	152.0681	0	1	7.725444
3	102.1403	79.86035	14.64496	35.07428	7.202141
30	111.2104	170.0227	0	1	8.154828
32	111.4308	150.0564	0	1	8.392342
33	111.4689	79.39107	0	1	9.139482
34	116.4468	157.0242	0	1	8.951027
35	108.98	42.78697	0	1	9.210468
36	116.4849	88.65475	0	1	9.674048
4	102.1403	79.86035	11.62847	47.06389	6.138077
5	99.91322	68.77706	18.62154	28.9224	7.821121
6	102.7021	76.71254	17.42115	24.51301	8.254908

Table 5 Calculated descriptors for the test set and predicted Activity.

Molecule	ATS1p	ATSC1m	RDF80u	RDF55i	Predicted Activity
1	95.06158	48.16925	12.16469	28.90358	−0.95502
16	106.6714	72.81042	18.2977	27.10917	2.824743
19	100.4838	43.34609	6.340206	35.53609	1.158496
20	80.4678	35.2542	10.68857	26.89952	0.425606
21	95.37356	49.46837	11.31287	34.18471	1.742684
24	82.99485	18.20398	7.579923	25.10282	−0.56217
27	103.9259	99.87649	0	1	8.056255
31	106.4148	142.449	0	1	12.69045
7	88.15629	88.29171	25.9644	31.1306	2.904864
8	95.71811	87.22781	23.39105	26.99112	3.012702
9	85.9427	56.99298	13.6134	16.56309	0.917058

Pearson’s correlation of the four descriptors employed in the QSAR Model was reported in Table 6. The correlation coefficient between each descriptor in the model is very low thus, it can be inferred that there is no significant inter-correlation among the descriptors used in building the model.

Table 6 Pearson correlation matrix for the selected descriptors.

	ATS1p	ATSC1m	RDF80u	RDF55i
ATS1p	1
ATSC1m	0.341475	1
RDF80u	−0.4939	−0.56435	1
RDF55i	−0.56598	−0.57235	0.023338	1

Y- Randomization parameter test were reported in Table 7. The low R² and Q² values for numbers of trials confirm that the built QSAR model is stable, robust and reliable. While the ${cR}_{p}^{2}$ value greater than 0.5 assured that the built model is powerful and not inferred by chance.

Table 7 Y- Randomization Parameters test.

Model	R	R^2	Q^2
Original	0.73026857	0.5332922	0.320676
Random 1	0.20860711	0.0435169	−0.412234
Random 2	0.74943813	0.5616575	0.3881064
Random 3	0.30303256	0.0918287	−0.326775
Random 4	0.33073537	0.1093859	−0.275737
Random 5	0.39593156	0.1567618	−0.191466
Random 6	0.19624781	0.0385132	−0.40262
Random 7	0.13642358	0.0186114	−0.567249
Random 8	0.35290385	0.1245411	−0.257905
Random 9	0.46995088	0.2208538	−0.062309
Random 10	0.16738226	0.0280168	−0.477468
Random Models Parameters
Average r:	0.33106531
Average r^2	0.13936872
Average Q^2	−0.25856557
cRp^2	0.47534142

The description and other statistical parameters that influence the selected descriptors were reported in Table 8. The presence of 2D and 3D descriptors in the model suggests that these types of descriptors are able to characterize better anti-diabetic activities of the compounds. The calculated Variance Inflation Factor (VIF) values for all the four descriptors in the model were all less than 4 which imply that the descriptors were orthogonal and model generated was significant. The null hypothesis says there is no significant relationship between the activities of the inhibitor molecules and the descriptors used in building the model at p > 0.05. The P-values of the descriptors in the model at 95% confidence limit shown in Table 8 are all less than 0.05. This implies that the null hypothesis is rejected. Thus we accepted the alternative hypothesis. Hence we infer that there is a significant relationship between the activities of the inhibitor molecules and descriptors used in building the model at p < 0.05.

Table 8 List of some descriptors and their statistical influence in the QSAR optimization model.

S/N	Descriptors symbol	Name of descriptor(s)	Class	Statistics
				ME	VIF	P- Value
1	ATS1p	Broto-Moreau autocorrelation - lag 1/weighted by polarizabilities	2D	−0.37338	2.4531	0.00162
2	ATSC1m	Centered Broto Moreau autocorrelation - lag 1/weighted by mass	2D	0.23434	1.3322	4.4 x 10⁻¹⁰
3	RDF80u	Radial distribution function - 080/unweighted	3D	0.43242	2.4543	7.3 x 10⁻⁸
4	RDF55i	Radial distribution function - 055/weighted by relative first ionization potential	3D	0.55430	2.3221	5.6 x 10⁻⁵

3.1

3.1 Interpretation of selected descriptors

The 2D descriptor, ATS1p which correspond to Average centered Broto-Moreau autocorrelation - lag 1/weighted by polarizabilities, have negative mean effect (MF) which means they have negative impact on the activity. ATSC1m correspond to Centered Broto Moreau autocorrelation - lag 1/weighted by mass. It has negative mean effect which indicates that an increase in the weight of molecule leads to a decrease in its anti-diabetic activity. RDF80u and RDF55i are one of the 3D-radial distribution function (RDF) descriptors which were proposed based on a radial distribution function. The radial distribution function is probability distribution to find an atom in a spherical volume of radius. RDF descriptors are independent of the size and rotation of the entire molecule. They describe the steric hindrance or the structure/activity properties of a molecule. The RDF descriptor provides valuable information about the bond distances, ring types, planar, non-planar systems and atom types. Having positive mean effect (MF) implies that they have positive impact on the activity.

Plot of predicted activity against experimental activity of training and test set where shown in Figs. 2 and 3 respectively. The R² value of 0.9460 for training set and R² value of 0.8642 for test set reported in this study was in agreement with Genetic Function Approximation (GFA) derived R² value reported in Table 2. This confirms the robustness and reliability of the model. Plot of standardized residual versus experimental activity shown in Fig. 4 indicates that there was no systematic error in the model built as the spread of standardized residual values were on both sides of zero (Jalali-Heravi and Kyani, 2004).

Plot of predicted activity against experimental activity of training set.

Plot of predicted activity against experimental activity of test set.

Plot of Standardized residual activity versus experimental activity.

The leverage values for the entire compounds in the dataset were plotted against their standardized residual values leading to discovery of outliers and influential compound in the models. The Williams plot of the standardized residuals versus the leverage value is shown in Fig. 5 which an evident that all the compounds were within the square area $\pm$ 3 of standardized cross-validated residual produced by the model. Therefore no compound is said to be an outlier. However, only one compound is said to be an influencing compound since its leverage value is greater than the warning leverage (h^∗ = 0.60). This was attributed to difference in its molecular structure compared to other compounds in the dataset.

The Williams plot of the standardized residuals versus the leverage value.

3.2

3.2 Molecular docking

Molecular docking studies were carried out in order to analysis and understand the interaction formed between the targets (TDG5) and inhibitor ligands that have the least and best pEC50. The docking results reported in Tables 9 and 10 shows that the binding affinities of the ligands with best pEC50 were greater than the binding affinity of the ligands with least pEC50 which indicates that the binding affinities of these ligands correlate with their pEC50. Ligand 27 with least binding affinity (−8.5 kcal/mol) and ligand 35 with best binding affinity (-10.4 kcal/mol) were visualized and analyzed in Discovery Studio Visualizer as shown in Figs. 6 and 7 below. Ligand 27 formed one hydrogen bond (2.15425A°) with ASN69 of the target. Hydrophobic interaction is a bond formed between the ligand and the binding pocket of the target site (receptor). It adhere the ligand to the surface of target site. Ligand 27 formed hydrophobic bond with ALA271, ALA271, ALA271, UNK1, PRO330, LEU275, LYS267, ALA271of the target site. Ligand 35 also formed two hydrogen bonds (2.43479, 2.15121A°) with ARG131 of the target. While hydrophobic interactions were observed with ALA271, LEU275, PHE332, LEU275, VAL67of the target site.

Table 9 Binding Affinity, Hydrogen bond and hydrophobic interaction formed between ligands with least PEC₅₀ and the active site of the TGR5 receptor.

Ligand	Binding Affinity (kcal/mol)	Target	Hydrogen bond		Hydrophobic
			Amino acid	Bond length (A^o)	Hydrophobic
22^a 4-(2,5-dichlorophenoxy)-N-(o-tolyl)pyrimidine-5-carboxamide	−6.5	TGR5	–	–	ALA271, ALA271, ALA271, UNK1, PRO330, LEU275, LYS267, ALA271
23^a 4-(2,5-dichlorophenoxy)-N-ethyl-N-(o-tolyl)pyrimidine-5-carboxamide	−7.2	TGR5	–	–	PHE194, ALA200, HIS178, PHE194PHE194, LYS305, VAL297, LYS305
27^b (4-cyclopropyl-3,4-dihydroquinoxalin-1(2H)-yl)(4-(2,5-dichlorophenoxy)pyridin-3-yl)methanone	−8.5	TGR5	ASN69	2.15425	ALA271, ALA271, ALA271, UNK1, PRO330, LEU275, LYS267, ALA271
28^b (4-(4-chlorophenoxy)pyridin-3-yl)(4-cyclopropyl-3,4-dihydroquinoxalin-1(2H)-yl)methanone	−8.1	TGR5	–	–	PHE194, ALA200, LYS305, HIS178, PHE194

Where superscript a and b represent 4‑Phenoxypyrimidine-5-carboxamide and 4‑Phenoxynicotinamide derivative

Table 10 Binding Affinity, Hydrogen bond interaction and hydrophobic interaction formed between ligands with best PEC₅₀ and the active site of the TGR5 receptor.

Ligand	Binding Affinity kcal/mol	Target	Hydrogen bond		Hydrophobic
			Amino acid	Bond length (A^o)	Hydrophobic
16^a (1-cyclopropyl-1,2,3,4-tetrahydroquinolin-4-yl)(4-(2,5-dichlorophenoxy)pyrimidin-5-yl)methanone	−9.2	TGR5	TRP10 SER104	1.06624 2.31227	LYS267, LYS267, ALA271, ALA271, PHE332
17^a (1-cyclopropyl-1,2,3,4-tetrahydroquinolin-4-yl)(4-(2,5-dichlorophenoxy)pyrimidin-5-yl)methanone	−8.8	TGR5	TRP10 SER104	2.81227	PHE194, ALA200, LYS305, HIS178, PHE194, VAL297
33^b (4-cyclopropyl-3,4-dihydroquinoxalin-1(2H)-yl)(4-(2,4,5-trichlorophenoxy)pyridin-3-yl)methanone	−9.7	TGR5	TYR141	2.7432 1.86624	LYS267, LYS267, ALA271, ALA271 PHE332
35^b (4-cyclopropyl-3,4-dihydroquinoxalin-1(2H)-yl)(4-(2,4,5-trimethylphenoxy)pyridin-3-yl)methanone	−10.4	TGR5	ARG131 ARG131	2.43479 2.15121	ALA271, LEU275, PHE332, LEU275 VAL67

Where superscript a and b represent 4‑Phenoxypyrimidine-5-carboxamide and 4‑Phenoxynicotinamide derivatives

(27a) and (27b) show the 3D and 2D interactions between TGR5 and Ligand 27.

(35a) and (35b) show the 3D and 2D interactions between TGR5 and Ligand 35.

Hydrogen bond between the ligand 27 and target site is shown in Fig. 8. The C=O of the ligand also act as hydrogen acceptor and formed only one hydrogen bond with ASN69 of the target. Fig. 9 shows the hydrogen bond interaction between the ligand 35 and the target site. A total of two hydrogen bonds were formed. The ether functional group (–O–) of the ligand acts as hydrogen acceptor and formed one hydrogen bond with ARG131 of the target. The C=O of the ligand also act as hydrogen acceptor and formed one hydrogen bond with ARG131 of the target. The hydrogen bond formation alongside with the hydrophobic interaction provides an evidence that ligand 35 of the inhibitor compounds is potent against TGR5 receptor.

H-bond between the ligand 27 and TGR5 receptor.

H-bond between the ligand 35 and TGR5 receptor.

4 Conclusion

In this research, QSAR model was generated with descriptor (ATS1p, ATSC1m, RDF80u, RDF55i) which were highly correlated with biological activities of 4-Phenoxynicotinamide and 4-Phenoxypyrimidine-5-carboxamide derivatives. These descriptors produced a robust model to predict the anti-diabetic activities of these compounds. The internal and external validation tests for the QSAR model generated were in agreement with recommended value of validation parameters for a generally acceptable QSAR model. The Molecular docking studies showed that the binding affinities of the inhibitors correlate with their pEC₅₀ and the best compound has binding affinity of −10.4 kcal/mol which formed H-bond and hydrophobic interactions with amino acid of the target. The QSAR technique alongside with molecular ducking study provides a valuable approach for medicinal and pharmaceutical researchers to design and synthesis new anti-diabetes agent against TGR5 receptor of type 2 diabetes mellitus (T2DM).

References

Amit C., Payal C., Baghel U.S., Aakash D., . Synthesis, cytotoxic evaluation, Docking and QSAR study of N-(4-oxo-2-(4-((5-aryl-1, 3, 4-thiadiazol-2-yl) amino) phenyl) thiazolidin-3-yl) benzamides as antitubulin agents. Curr. Top. Med. Chem.. 2016;16:2509-2520.
[Google Scholar]
Bajpai G., Malik S., . diabetic Drugs. Pharm. Bull.. 2003;51:138-151.
[Google Scholar]
Becke A.D., . Becke’s three parameter hybrid method using the LYP correlation functional. J. Chem. Phys.. 1993;98:5648-5652.
[Google Scholar]
Cramer R.D., Patterson D.E., Bunce J.D., . Comparative molecular field analysis (CoMFA). 1. Effect of shape on binding of steroids to carrier proteins. J. Am. Chem. Soc.. 1988;110:5959-5967.
[Google Scholar]
Dasoondi A.S., Singh V., Voleti S.R., Tiwari M., . Comparative molecular field analysis of benzothiazepine derivatives: Mitochondrial sodium calcium exchange inhibitors as antidiabetic agents. Indian J. Pharm. Sci.. 2008;70:186.
[Google Scholar]
Dieguez-Santana K., Pham-The H., Rivera-Borroto O.M., Puris A., Le-Thi-Thu H., Casanola-Martin G.M., . A Two QSAR Way for Antidiabetic Agents Targeting Using α-Amylase and α-Glucosidase Inhibitors: Model Parameters Settings in Artificial Intelligence Techniques. Lett. Drug Des. Discov.. 2017;14:862-868.
[Google Scholar]
Dixit A., Saxena A.K., . QSAR analysis of PPAR-γ agonists as anti-diabetic agents. Eur. J. Med. Chem.. 2008;43:73-80.
[Google Scholar]
Duan H., Ning M., Chen X., Zou Q., Zhang L., Feng Y., Zhang L., Leng Y., Shen J., . Design, synthesis, and antidiabetic activity of 4-phenoxynicotinamide and 4-phenoxypyrimidine-5-carboxamide derivatives as potent and orally efficacious TGR5 agonists. J. Med. Chem.. 2012;55:10475-10489.
[Google Scholar]
Friedman J.H., . Multivariate adaptive regression splines. Ann. Stat. 1991:1-67.
[Google Scholar]
Ibezim E.C., Duchowicz P.R., Ibezim N.E., Mullen L.M.A., Onyishi I.V., Brown S.A., Castro E.A., . Computer-aided linear modeling employing QSAR for drug discovery. Sci. Res. Essays. 2009;4:1559-1564.
[Google Scholar]
Jalali-Heravi M., Kyani A., . Use of computer-assisted methods for the modeling of the retention time of a variety of volatile organic compounds: a PCA-MLR-ANN approach. J. Chem. Inf. Comput. Sci.. 2004;44:1328-1335.
[Google Scholar]
Kennard R.W., Stone L.A., . Computer aided design of experiments. Technometrics. 1969;11:137-148.
[Google Scholar]
Khaled K.F., . Modeling corrosion inhibition of iron in acid medium by genetic function approximation method: a QSAR model. Corros. Sci.. 2011;53:3457-3465.
[Google Scholar]
Lee C., Yang W., Parr R.G., . Development of the Colle-Salvetti correlation-energy formula into a functional of the electron density. Phys. Rev. B. 1988;37:785.
[Google Scholar]
Saydah S.H., Fradkin J., Cowie C.C., . Poor control of risk factors for vascular disease among adults with previously diagnosed diabetes. Jama. 2004;291:335-342.
[Google Scholar]
Singh P., . Quantitative Structure-Activity Relationship Study of Substituted-[1, 2, 4] Oxadiazoles as S1P1 Agonists. J. Curr. Chem. Pharm. Sci.. 2013;3:64-79.
[Google Scholar]
Tropsha A., Gramatica P., Gombar V.K., . The importance of being earnest: validation is the absolute essential for successful application and interpretation of QSPR models. Mol. Inform.. 2003;22:69-77.
[Google Scholar]
Vassileva G., Golovko A., Markowitz L., Abbondanzo S.J., Zeng M., Yang S., Hoos L., Tetzloff G., Levitan D., Murgolo N.J., . Targeted deletion of Gpbar1 protects mice from cholesterol gallstone formation. Biochem. J.. 2006;398:423-430.
[Google Scholar]
Veerasamy R., Rajak H., Jain A., Sivadasan S., Varghese C.P., Agrawal R.K., . Validation of QSAR models-strategies and importance. Int. J. Drug Des. Discov.. 2011;3:511-519.
[Google Scholar]
Watanabe M., Houten S.M., Mataki C., Christoffolete M.A., Kim B.W., Sato H., Messaddeq N., Harney J.W., Ezaki O., Kodama T., . Bile acids induce energy expenditure by promoting intracellular thyroid hormone activation. Nature. 2006;439:484-489.
[Google Scholar]

Show Sections

[1] Amit C., Payal C., Baghel U.S., Aakash D., . Synthesis, cytotoxic evaluation, Docking and QSAR study of N-(4-oxo-2-(4-((5-aryl-1, 3, 4-thiadiazol-2-yl) amino) phenyl) thiazolidin-3-yl) benzamides as antitubulin agents. Curr. Top. Med. Chem.. 2016;16:2509-2520.
[Google Scholar]

[2] Bajpai G., Malik S., . diabetic Drugs. Pharm. Bull.. 2003;51:138-151.
[Google Scholar]

[3] Becke A.D., . Becke’s three parameter hybrid method using the LYP correlation functional. J. Chem. Phys.. 1993;98:5648-5652.
[Google Scholar]

[4] Cramer R.D., Patterson D.E., Bunce J.D., . Comparative molecular field analysis (CoMFA). 1. Effect of shape on binding of steroids to carrier proteins. J. Am. Chem. Soc.. 1988;110:5959-5967.
[Google Scholar]

[5] Dasoondi A.S., Singh V., Voleti S.R., Tiwari M., . Comparative molecular field analysis of benzothiazepine derivatives: Mitochondrial sodium calcium exchange inhibitors as antidiabetic agents. Indian J. Pharm. Sci.. 2008;70:186.
[Google Scholar]

[6] Dieguez-Santana K., Pham-The H., Rivera-Borroto O.M., Puris A., Le-Thi-Thu H., Casanola-Martin G.M., . A Two QSAR Way for Antidiabetic Agents Targeting Using α-Amylase and α-Glucosidase Inhibitors: Model Parameters Settings in Artificial Intelligence Techniques. Lett. Drug Des. Discov.. 2017;14:862-868.
[Google Scholar]

[7] Dixit A., Saxena A.K., . QSAR analysis of PPAR-γ agonists as anti-diabetic agents. Eur. J. Med. Chem.. 2008;43:73-80.
[Google Scholar]

[8] Duan H., Ning M., Chen X., Zou Q., Zhang L., Feng Y., Zhang L., Leng Y., Shen J., . Design, synthesis, and antidiabetic activity of 4-phenoxynicotinamide and 4-phenoxypyrimidine-5-carboxamide derivatives as potent and orally efficacious TGR5 agonists. J. Med. Chem.. 2012;55:10475-10489.
[Google Scholar]

[9] Friedman J.H., . Multivariate adaptive regression splines. Ann. Stat. 1991:1-67.
[Google Scholar]

[10] Ibezim E.C., Duchowicz P.R., Ibezim N.E., Mullen L.M.A., Onyishi I.V., Brown S.A., Castro E.A., . Computer-aided linear modeling employing QSAR for drug discovery. Sci. Res. Essays. 2009;4:1559-1564.
[Google Scholar]

[11] Jalali-Heravi M., Kyani A., . Use of computer-assisted methods for the modeling of the retention time of a variety of volatile organic compounds: a PCA-MLR-ANN approach. J. Chem. Inf. Comput. Sci.. 2004;44:1328-1335.
[Google Scholar]

[12] Kennard R.W., Stone L.A., . Computer aided design of experiments. Technometrics. 1969;11:137-148.
[Google Scholar]

[13] Khaled K.F., . Modeling corrosion inhibition of iron in acid medium by genetic function approximation method: a QSAR model. Corros. Sci.. 2011;53:3457-3465.
[Google Scholar]

[14] Lee C., Yang W., Parr R.G., . Development of the Colle-Salvetti correlation-energy formula into a functional of the electron density. Phys. Rev. B. 1988;37:785.
[Google Scholar]

[15] Saydah S.H., Fradkin J., Cowie C.C., . Poor control of risk factors for vascular disease among adults with previously diagnosed diabetes. Jama. 2004;291:335-342.
[Google Scholar]

[16] Singh P., . Quantitative Structure-Activity Relationship Study of Substituted-[1, 2, 4] Oxadiazoles as S1P1 Agonists. J. Curr. Chem. Pharm. Sci.. 2013;3:64-79.
[Google Scholar]

[17] Tropsha A., Gramatica P., Gombar V.K., . The importance of being earnest: validation is the absolute essential for successful application and interpretation of QSPR models. Mol. Inform.. 2003;22:69-77.
[Google Scholar]

[18] Vassileva G., Golovko A., Markowitz L., Abbondanzo S.J., Zeng M., Yang S., Hoos L., Tetzloff G., Levitan D., Murgolo N.J., . Targeted deletion of Gpbar1 protects mice from cholesterol gallstone formation. Biochem. J.. 2006;398:423-430.
[Google Scholar]

[19] Veerasamy R., Rajak H., Jain A., Sivadasan S., Varghese C.P., Agrawal R.K., . Validation of QSAR models-strategies and importance. Int. J. Drug Des. Discov.. 2011;3:511-519.
[Google Scholar]

[20] Watanabe M., Houten S.M., Mataki C., Christoffolete M.A., Kim B.W., Sato H., Messaddeq N., Harney J.W., Ezaki O., Kodama T., . Bile acids induce energy expenditure by promoting intracellular thyroid hormone activation. Nature. 2006;439:484-489.
[Google Scholar]

4‑Phenoxypyrimidine-5-carboxamide derivatives
S/N	Molecules	EC₅₀ (nM)	pEC₅₀	Predicted Activity	Residual
1^a		156	6.8068	6.797241	0.009559
2		20	7.6989	7.901818	−0.20292
3		30	7.5228	7.202141	0.320659
4		710	6.1487	6.138077	0.010623
5		30	7.5228	7.821121	−0.29832
6^a		2.9	8.5376	8.254908	0.282692
7^a		590	6.2291	6.710226	−0.48113
8^a		23	7.6382	7.688441	−0.05024
9^a		164	6.7851	6.489332	0.295768
10		49	7.3098	7.493192	−0.18339
11		127	6.8961	7.005227	−0.10913
12		3.1	8.5086	7.710525	0.798075
13		7.1	8.1487	8.167491	−0.01879
14		69	7.1612	7.784653	−0.62345
15		6.2	8.2076	8.142226	0.065374
16		1.5	8.8239	8.714862	0.109038
17		2.8	8.5528	8.644922	−0.09212
18		3711	5.4305	5.379953	0.050547
19^a		535	6.2716	6.536593	−0.26499
20^a		3151	5.5015	5.079	0.4225
21^a		160	6.7958	6.400758	0.395042
22		10,000	5.000	5.056796	−0.0568
23		4886	5.3110	5.351652	−0.04065
24^a		451	6.3458	5.388271	0.957529
4‑Phenoxynicotinamide derivatives
25		1.5	8.8239	8.612203	0.211697
26		47	7.3279	7.296042	0.031858
27^a		27	7.5686	7.974061	−0.40546
28		7.9	8.1023	7.974061	0.128239
29		12	7.9208	7.725444	0.195356
30		12	7.9208	8.154828	−0.23403
31^a		12	7.9208	7.840378	0.080422
32		0.72	9.1427	8.392342	0.750358
33		0.46	9.3372	9.139482	0.197718
34		0.60	9.2218	8.951027	0.270773
35		0.31	9.5086	9.210468	0.298132
36		0.72	9.1426	9.674048	−0.53145

Computational modeling of 4-Phenoxynicotinamide and 4-Phenoxypyrimidine-5-carboxamide derivatives as potent anti-diabetic agent against TGR5 receptor

Abstract

Keywords

Anti-diabetic

Applicability domain

Binding affinity

Molecular docking

QSAR

1 Introduction

2 Materials and methods

2.1 Data collection

2.2 Biological activities (pEC50)

2.3 Optimization

2.4 Molecular descriptor calculation

2.5 Normalization and data pretreatment

2.6 Data division

2.7 Internal validation of model

2.8 External validation of the model

2.9 Y-Randomization test

2.10 Evaluation of the applicability domain of the model

2.11 Quality assurance of the model

2.12 Docking studies

3 Results and discussion

3.1 Interpretation of selected descriptors

3.2 Molecular docking

4 Conclusion

References

Suggested read for related articles:

2.2 Biological activities (pEC₅₀)