In silico identification of novel lead compounds with AT1 receptor antagonist activity: successful application of chemical database screening protocol
© Pal and Paliwal; licensee Springer. 2012
Received: 21 October 2011
Accepted: 1 March 2012
Published: 1 March 2012
AT1 receptor antagonists are clinically effective drugs for the treatment of hypertension, cardiovascular, and related disorders. In an attempt to identify new AT1 receptor antagonists, a pharmacophore-based virtual screening protocol was applied. The pharmacophore models were generated from 30 training set compounds. The best model was chosen on the basis of squared correlation coefficient of training set and internal test set. The validity of the developed model was also ensured using catScramble validation method and external test set prediction.
The final model highlighted the importance of hydrogen bond acceptor, hydrophobic aliphatic, hydrophobic, and ring aromatic features. The model satisfied all the statistical criteria such as cost function analysis and correlation coefficient. The result of estimated activity for internal and external test set compounds reveals that the generated model has high prediction capability. The validated pharmacophore model was further used for mining of 56000 compound database (MiniMaybridge). Total 141 hits were obtained and all the hits were checked for druggability, this led to the identification of two active druggable AT1 receptor antagonists with diverse structure.
A highly validated pharmacophore model generated in this study identified two novel druggable AT1 receptor antagonists. The developed model can also be further used for mining of other virtual database.
The renin-angiotensin system plays a fundamental role in blood pressure and fluid and electrolyte homeostasis . Angiotensin II (AII), an octapeptide produced by the renin-angiotensin system, is a powerful endogenous vasopressor. Angiotensin converting enzyme inhibitors work by blocking the production of angiotensin II from angiotensin I. An alternative and possibly superior approach would be to block the action of AII at the level of its receptor. Two distinct subtypes of AII receptors [type 1 (AT1) and type 2 (AT2)] have been identified, and both belong to the G protein-coupled receptors super family (GPCRs) [2, 3]. Most of the biological actions of AII are mediated by the AII receptors of the AT1 subtype. The AT1 receptor subtype mediates virtually all the known physiological actions of AII in cardiovascular, neuronal, endocrine, and hepatic cells as well as in other ones. Since AT1 receptor is GPCR the interaction of AII with the AT1 receptor induces a conformational change, which promotes the coupling with the G protein(s) and leads to the signal transduction via several effector systems (phospholipases C, D, A2, adenyl cyclase, etc.). The AT1 receptors play a major role in the pressor and trophic actions of the AII, and much effort has been spent in developing nonpeptide antagonists for this receptor for the treatment of hypertension and congestive heart failure .
Like other GPCR families, AT1 receptors are transmembrane proteins and such macromolecules are not easily crystallized for structural analysis by X-ray crystallography . In the absence of three-dimensional (3D) structure for AT1 receptor, a rational design of antagonists using a structure-based approach is not feasible . For this reason, 3D pharmacophore models from the ligand-based approach are very useful for analyzing the ligand-receptor interactions. Moreover, a pharmacophore can also be used as a query in a 3D database search to identify new structural classes of potential lead compounds. In the recent years, the development of a 3D-pharmacophore and its use in the virtual screening of the chemical databases appear to be a more relevant and time-saving approach. Thus, the construction of an accurate pharmacophore is a key objective in many drug discovery efforts.
The pharmacophore generation methods of the Catalyst software have been successfully used in drug discovery research and toxicology [6–8] as evident from pharmacophore-based development of protein farnesyl transferase, human immunodeficiency virus (HIV) protease, and HIV reverse transcriptase inhibitors [9, 10].
In this study, our approach of pharmacophoric exploration via set of diverse 3D structures has resulted in development of a highly validated and predictive pharmacophore model for AT1 receptor antagonists. The developed phamacophore was subsequently used for virtual screening of chemical databases for identification of novel lead compounds with nanomolar activity range.
2. Results and discussion
2.1. HypoGen model
Pharmacophore models were generated using 30 training set compounds representing two series of structurally diverse compounds with AT1 receptor antagonist activity. All the generated pharmacophore hypotheses were evaluated for their statistical fitness on the basis cost difference values, correlation coefficients (r), and rms deviations. The pharmacophoric features and statistical data for a set of ten chosen hypothesis are listed in Additional file 1.
Out of ten, hypothesis1 was identified as best pharmacophore model, since this hypothesis showed a cost difference of 20.17 between null cost 148.75 and total cost 128.58 satisfying the range recommended in the cost analysis of the catalyst procedure. Hypothesis1 had total cost close to fixed cost (124.52), lower error cost (103.409), lowest root-mean-square (RMS) divergence (0.408), best correlation (r = 0.977), and good internal test set prediction (r test-set = 0.93). The configuration cost of the hypothesis exceeded the limit of 17 bits but can be accepted as the model achieves other validation criterion [11, 12].
The chosen hypothesis comprised of one hydrogen-bond acceptor (HBA), hydrophobic aliphatic region, and hydrophobic (HY) and one ring aromatic (RA) sites in a specific 3D orientation. The results of tolerance and weight fit to the features of the training set compounds are given in Additional file 2. The pharmacophore model mapped well to the training and test set compounds. The values of actual and predicted activity for the training and internal test set compounds are given in Additional files 3 and 4. The model was found to be quite good in predicting the activity of external test set compounds  with correlation co-efficient value of 0.71 and the values of actual and predicted activity are given in Additional file 5.
2.2. Fisher's cross validation test
2.3. Mapping of training set compounds
2.4. Mapping of test set compounds
2.5. Database search
List of hits obtained from the MiniMaybridge and NCI database with their corresponding fit value and estimated activity
3. Materials and methods
Disco very studio, version 2.0, Accelrys Software Inc., San Diego, CA, was used to develop pharmacophore hypothesis for structurally diverse series of triazolinone derivatives reported in the literature [16, 17] with activity range from 0.072 to 250 nM. Chemical structures of various N2-aryltriazolinone biphenylsulfonamides with their experimental IC50 values for the AT1 receptor subtype are listed in Additional file 6.
3.1. Selection of the training set and test set
The most important aspect of the hypothesis generation in HypoGen is the selection of the training set of molecules. The selection has to follow some basic requirements; such as a minimum of 16 structurally diverse compounds should be selected to avoid any chance correlation, most active compound should be included and the activity data should have a range of 3.5-5 orders of magnitude .
On the basis of above criteria, the dataset was divided into training set and test set. The training set comprised of 30 compounds, whereas internal test set was composed of 27 compounds. The most active compounds were included in the training set so that they would provide critical information for pharmacophore requirements. Several moderately active and inactive compounds were also included to spread the activity ranges as wide as possible. The important aspect of such selection scheme is that each active compound should teach something new to the HypoGen module to help it uncover as much critical information as possible for predicting biological activity.
3.2. Generation of pharmacophores
Details of the pharmacophore development procedures have been described in the literature [9, 18]. In brief, conformational models of all training set molecules with AT1 receptor antagonist activity were generated using the best quality conformational search option in Catalyst employing a constraint of a 20 kcal/mol energy threshold above the global energy minimum using CHARMm force field. A maximum of 250 conformations were generated to ensure maximum coverage in the conformational space . Instead of using just the lowest energy conformation of each compound, all conformational models for molecules in each training set were used in for pharmacophore hypothesis generation. The Catalyst software can generate pharmacophore hypotheses consisting of a maximum of five features. An initial analysis revealed that four chemical feature types such as HBA, HY, hydrophobic aliphatic (HY-ALI), and ring aromatic (RA) could effectively map all critical chemical features of all molecules in the training set. The minimum and maximum counts for HBA, HY, HY-ALI, and RA were set to 0 and 3, respectively. These four feature types were used to generate ten pharmacophore hypothesis from the training set. The uncertainty value was defaulted to 3 which is a ratio range of uncertainty in the activity value and MinPoints and MinSubset-Points were 4 (default value). The MinPoints parameter controls the minimum number of location constraints required for any hypothesis. The MinSubsetPoint parameter defines the number of chemical features that a hypothesis must match in all the compounds set .
3.3. Evaluation of the HypoGen model
3.3.1. Cost function analysis
All the hypotheses generated were subjected to cost function analysis which is considered as stringent quality check tool. Two important theoretical cost calculations that determine the success of any pharmacophore hypothesis are "fixed cost" and "null cost". Fixed cost represents the simplest model that fits all data perfectly, and the second null cost represents the highest cost of a pharmacophore with no features and which estimates activity to be the average of the activity data of the training set molecules. The null cost value is equal to the maximum occurring error cost. The greater the difference between null cost and total cost and closer the total cost of the generated hypothesis to the fixed cost, the more statistically significant is the generated hypothesis. Another important cost is the overall cost of HypoGen model which consist of three cost components, the weight cost, the error cost, and the configuration cost. The quality of each hypothesis can be judged on the basis of total cost which is sum of error cost, weight cost, and configuration cost. The configuration cost, which is also known as the entropy cost, depends on the complexity of the pharmacophore hypothesis space. The error cost is dependent on the RMS differences between the estimated and the actual activities of the training set molecules. In standard HypoGen model, the configuration should not be greater than 17.0. The RMS deviations represent the quality of the correlation between the estimated and the actual activity data. The error cost is the most important part of the total cost and increases as the RMS difference between the estimated and the actual affinity for the training set increases. The RMS value is related to the quality of prediction of the hypothesis. Error cost provides the highest contribution to total cost and it is directly related to the capacity of the particular pharmacophore as 3D QSAR model, i.e., in correlating the molecular structures to the corresponding biological responses. The weight cost is a value that increases in a gaussian form as the difference between the actual and ideal weights of the features deviates. According to the documentation, the ideal value of the weight is 2 because higher weight values tend to force unrealistic conformations of the compounds to fit such features .
3.3.2. Test set prediction
The ability of the models to predict the biological activity of compounds outside the model development procedure is a common method of validation . Internal test set of 27 and external test set of 46 compounds were employed to assess statistical significance of the developed model. All test set molecules were built and minimized as well as used in conformational analysis like the training set molecules. Predictions were made to evaluate the level of similarity between actual and predicted activity.
3.3.3. Statistical validation
where x is the total number of hypotheses having a total cost lower than best significant hypothesis and y the number (HypoGen runs initial + random runs). To obtain a 95% confidence level, 19 random spreadsheets are generated (y = 20) and every generated spreadsheet is submitted to HypoGen using the same experimental conditions (functions and parameters) as the initial run.
4. Database mining
The generated validated pharmacophore was used as query to search the virtual chemical compound database (NCI and MiniMaybridge) to identify new lead compounds with AT1 receptor antagonist activity.
The quantitative pharmacophore models were developed using the training set of molecules with the help of HypoGen module implemented in the Catalyst. The best pharmacophore model provided a statistically significant correlation and well-estimated AT1 activities for the test set compounds. Pharmacophore models generated for AT1 antagonists in this study highlight the structural requirements for antagonistic activity. This study also helped in the identification of five structurally diverse AT1 receptor antagonists.
This study was supported by grants from the Council of Scientific and Industrial Research, New Delhi. The authors also thank the Vice Chancellor, Banasthali University, for extending all the necessary facilities.
- Tuccinardi T, Calderone V, Rapposelli S, Martinelli A: Proposal of a new binding orientation for non-peptide AT 1 antagonists. J Med Chem 2006, 49: 4305–4316. 10.1021/jm060338pView ArticleGoogle Scholar
- Ellis ML, Patterson JH: A new class of antihypertensive therapy: angiotensin II receptor antagonists. Pharmacotherapy 1996, 16: 849–860.Google Scholar
- Israili ZH, Hall WD: Cough and angioneurotic edema associated with angiotensin-converting enzyme inhibitor therapy. A review of the literature andpathophysiology. Ann Intern Med 1992, 117: 234–242.View ArticleGoogle Scholar
- Bondensgaard K, Ankersen M, Thogersen H, Hansen BS, Wulff BS, Bywater RP: Recognition of privileged structures by G-protein coupled receptors. J Med Chem 2004, 47: 888–899. 10.1021/jm0309452View ArticleGoogle Scholar
- Sprague PW: Automated chemical hypothesis generation and database searching with catalyst. Perspect Drug Discov Des 1995, 3: 1. 10.1007/BF02174464View ArticleGoogle Scholar
- Langer T, Krovat EM: Chemical feature based pharmacophores and virtual library screening for discovery of new-leads. Curr Opin Drug Discovery Dev 2003, 6: 370–376.Google Scholar
- Clark DE, Westhead DR, Sykes RA, Murray CW: Active site-directed 3D database searching: pharmacophore extraction and validation of hits. J Comput Aided Mol Des 1996, 10: 397–416. 10.1007/BF00124472View ArticleGoogle Scholar
- Al-Sha'er MA, Taha MO: Elaborate ligand-based modeling reveals new nanomolar heat shock Protein 90α Inhibitors. J Chem Inf Model 2010, 50: 1706–1723. 10.1021/ci100222kView ArticleGoogle Scholar
- Mukherjee S, S Mullick S, Mukherjee A, Saha A: Pharmacophore mapping of selective binding affinity of estrogen modulators through classical and space modeling approaches: exploration of bridged-cyclic compounds with diarylethylene linkage. J Chem Inf Model 2007, 47: 475–487. 10.1021/ci600419sView ArticleGoogle Scholar
- Chang LL, Ashton WT, Flanagan KL, Chen TB, O'Malley SS, Zingaro GJ, Siegl PKS, Kivlighn SD, Lotti VJ, Chang RSL, Greenlee WJ: Triazolinone biphenylsulfonamides as angiotensin II receptor antagonists with high affinity for both the AT 1 and AT 2 subtypes. J Med Chem 1994, 37: 4464–4478. 10.1021/jm00052a006View ArticleGoogle Scholar
- Ashton WT, Chang LL, Flanagan KL, Hutchins SM, Naylor EM, Chakravarty PK, Patchett AA, Greenlee WJ, Chen TB, Faust KA, Chang RSL, Lotti VJ, Zingaro GJ, Schorn TW, Siegl PKS, Kivlighn SD: Triazolinone biphenylsulfonamide derivatives as orally active Angiotensin II antagonists with potent AT 1 receptor affinity and enhanced AT2 affinity. J Med Chem 1994, 37: 2808–2824. 10.1021/jm00043a020View ArticleGoogle Scholar
- Fischer R: The principle of experimentation illustrated by a psychophysical experiment. 8th edition. Hafner Publishing Co., New York; 1966.Google Scholar
- Manallack DT: Getting that hit: 3D database searching in drug discovery. Drug Discover Today 1996, 1: 231–238. 10.1016/1359-6446(96)88990-2View ArticleGoogle Scholar
- Chang LL, Ashton WT, Flanagan KL, Chen TB, O'Malley SS, Zingaro GJ, Kivlighn SD, Siegl PKS, Lotti VJ, RSL Greenlee WJ: Potent and orally active angiotensin II receptor antagonists with equal affinity for human AT 1 and AT 2 subtype. J Med Chem 1995, 38: 3741–3758. 10.1021/jm00019a004View ArticleGoogle Scholar
- Chang LL, Ashton WT, Flanagan KL, Strelitz RA, MacCoss M, Greenlee WJ, Chang RSL, Lotti VJ, Faust KA, Chen TB, Bunting P, Zingaro GJ, Kivlighn SD, Siegl PKS: Triazolinones as nonpeptide angiotensin II antagonists. 1. Synthesis and evaluation of potent 2,4,5-trisubstituted triazolinones. J Med Chem 1993, 36: 2558–2568. 10.1021/jm00069a015View ArticleGoogle Scholar
- Smellie A, Teig S, Towbin P: Poling: promoting conformational variation. J Comput Chem 1995, 16: 171–187. 10.1002/jcc.540160205View ArticleGoogle Scholar
- Mason JS, Good AC, Martin EJ: 3-D pharmacophores in drug discovery. Curr Pharm Des 2001, 7: 567–597. 10.2174/1381612013397843View ArticleGoogle Scholar
- Greenidge PA, Weiser J: A comparison of methods for pharmacophore generation with the Catalyst software and their use for 3DQSAR: application to a set of 4-aminopyridine thrombin inhibitors. Mini Rev Med Chem 2001, 1: 79–87. 10.2174/1389557013407223View ArticleGoogle Scholar
- Brooks BR, Bruccoleri RE, Olafson BD, States DJ, Swaminathan S, Karplus M: CHARMM: a program for macromolecular energy minimization, and dynamics calculations. J Comput Chem 1983, 4: 187–217. 10.1002/jcc.540040211View ArticleGoogle Scholar
- Paliwal S, Pal M, Yadav D, Singh S, Yadav R: Ligand-based drug design studies using predictive pharmacophore model generation on 4H-1,2,4-triazoles as AT1 receptor antagonists. Med Chem Res 2011. DOI 10.1007/s00044–011–9756–4Google Scholar
- Toba S, Srinivasan J, Maynard AJ, Sutter J: Using pharmacophore models to gain insight into structural binding and virtual screening: An application study with CDK2 and human DHFR. J Chem Inf Model 2006, 46: 728–35. 10.1021/ci050410cView ArticleGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.