View metadata, citation and similar papers at core.ac.uk brought to you by CORE provided by Archivio istituzionale della ricerca - Università di Palermo A. M. Almerico et al. & Combinatorial Science A Multivariate Analysis on Non-nucleoside HIV-1 Reverse Transcriptase Inhibitors and Resistance Induced by Mutation Anna Maria Almerico*, Antonino Lauria, Marco Tutone, Patrizia Diana, Paola Barraja, Alessandra Montalbano, Girolamo Cirrincione and Gaetano Dattolo Dipartimento Farmacochimico, Tossicologico e Biologico, Universita¡ degli Studi, Via Archirafi 32, 90123 Palermo (Italy) Full Paper This paper describes the use of multivariate statistical results were obtained in the case of L100I and K103N procedure PCA as a tool to explore the inhibitory activity mutants for which the higher number of assignments was of classes of NNRTIs against HIV-1 viruses (wild type and found when the principal components derived from the more frequent mutants, Y181C, V106A, K103N, L100I) descriptors were used. On this basis this statistical and against RT enzyme. The analysis of correlations approach is proposed as a reliable method for the between biological activity and molecular descriptors or prediction of the activity of NNRTIs, for which the data similarity indexes allowed a reliable classification of the against mutant strains have not been reported. fifty five derivatives considered in this study. The best 1 Introduction mutations of the reverse transcriptase enzyme associated with NNRTI therapy represent a major problem in devel- The treatment regimens for the human immunodeficiency oping resistance to current drugs regimens and they limit virus (HIV-1) have included both HIV protease and reverse enormously the effectiveness of the treatment. A single transcriptase (RT) enzyme inhibitors. All antiretroviral mutation in the NNRTI-binding pocket may result in high- drugs currently approved for clinical use are directed against level resistance to one or more NNRTIs. Therefore a high one of these targets. The first RT inhibitors approved in priority for medical research remains the discovery of USA and in Europe were nucleoside derivatives (NRTIs) antiviral agents effective against mutant HIV strains. which compete with normal nucleoside substrates for In this paper we propose the multivariate statistical incorporation into the viral genome, thus behaving as chain procedure, called PCA (principal component analysis), as a terminators [1]. Unlike nucleoside analogs, non-nucleoside tool to exploit the enormous amount of information reverse transcriptase inhibitors (NNRTIs) bind in a non- available on the inhibitory activity against HIV-1 viruses competitive manner to a specific ™pocket∫ of the HIV-1 RT, (wild type and the more frequent mutants, Y181C, V106A, which is closely associated with, but distinct from the K103N, L100I) and against RT enzyme. substrate binding site, altering its ability to function [2]. Biological problems have an intrinsic multivariate nature, Nevirapine, Delavirdine, and Efavirenz are the only involving many variables at the same time, and in general NNRTIs that have received regulatory approval, whereas the relation between these variables and the biological several other inhibitors (MKC-442, PNU-142721, and so on) response is hidden and no useful information can easily be are currently undergoing clinical trials. extracted. In order to simplify the data set in a multivariate Although current NNRTIs have demonstrated potent problem and to obtain an informative picture of the data antiviral activity in vitro and in HIV-1 infected patients, tendencies, a chemometric multivariate analysis can be used. In the past multivariate data analysis has been applied in * To receive all corrispondence. Tel 390916161606, Fax many fields of science demonstrating to be most suitable in 390916169999, e-mail: [email protected] handling complex data sets and allowing to investigate relationships among all objects and all variables simulta- Key words: NNRTIs, PCA, DA, resistance, mutation neously. In particular PCA is able to detect similarities Abbreviations: NNRTIs, non nucleoside reverse transcriptase among variables and is used to reduce the number of inhibitors; PCA, principal components analysis; DA, discriminant variable thus preparing the data for further analysis. The analysis; hitÀrate; chi-square. easier mathematical way to represent a multivariate prob- 984 QSAR Comb. Sci. 22 (2003) DOI: 10.1002/qsar.200330834 ¹ 2003 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim A Multivariate Analysis on Non-nucleoside HIV-1 Reverse Transcriptase Inhibitors and Resistance ... & Combinatorial Science lem is to build a matrix relating variables and objects. In our Our selection of structural variables was made regarding case the objects are the NNRTIs, the variables are selected several reported x-ray crystallographic structures of HIV-1 chemicals descriptors, and the biological response are IC50 NNRT inhibitors complexes which supported the already against RT and susceptibility/resistance data. well acknowledged binding mode of the different classes of Each object is now placed in an n-dimensional space NNRTIs. In particular ellipsoidal volume, which is defined (where n is the number of variables). However is more by the moments of inertia, and accessible surface area give practical for us work in two or three dimensions. The PCA information about steric properties of a molecule, log P is a permits the projection of higher order space in two or three typical QSAR variable, related to hydrophobic/hydrophilic dimensions with a minimal loss of statistic information. The profile of the inhibitor and in our case can be related to this coordinate axes of the original n-order space are rotated kind of interaction in the active site, total lipole is a measure until the direction of maximum variance is coincident with of the lipophilic distribution in a 3D space and it is calculated one of the rotated axes (called the first principal component from the summed atomic log P values, the number of H- axis). The second principal component and so on give the bond donors and acceptors give other information about the orthogonal direction of the maximum residual variance. ability of the inhibitor to stabilize its interaction with the Another interesting aspect in a chemometric multivariate ™binding pocket∫, total dipole gives information about the analysis is the possibility of classification as done in the electronic features, ELUMO and EHOMO are energetic varia- discriminant analysis (DA). The derived classification rule bles that classify the set of the inhibitors in terms of their describes a surface which separates the classes and it may be ability to act as electrophiles and nucleophiles, as expected used to predict class membership. for the inhibition process. Other energy variables are heat of It is our aim to develop a simple but efficient method to formation, which classifies the set of the inhibitors in terms evaluate, on the basis of chemical-physical descriptors and of relative thermodynamic stability and is widely used in structural similarity, new NNRTIs that are less likely to chemometric studies, and total energy in water which trigger resistance or are effective against mutant HIV classifies the behaviour of the inhibitors in the physiological strains. solvent. The whole set of 11 descriptors and 55 compounds was used to perform the PCA. Table 3 reports the matrix of 2 Materials and Methods eleven PCs with their composition in terms of original variables, together with the fraction of variance explained, More than 30different classes of NNRTIs have been the total fraction of variance explained, and the eigenvalue described to date. In this study fifty five NNRTI derivatives, of the covariance matrix corresponding to each component the structures of which are depicted in Figure 1, were that is equal to the fraction of variance explained by the utilized. The selection includes most of the derivatives number of variable used. The data were standardized by currently present into the database of the National Institute mean/sd (standard deviation). The first four PCs which have of Allergy and Infectious Diseases (NIAID) [3], and the eigenvalue > 1 and explaining 73.7% of variance, were literature derivatives for which the inhibitory concentration selected for further calculations. m against RT enzyme (IC50, M) was reported [4]. The most In the first PC the variables that have major importance active compounds of each class were considered. Table 1 are ellipsoidal volume ( 0.39), surface area ( 0.45), and shows these activity values, to be included in the calcu- total energy in water ( À 0.46). These variables clearly lations, together with the available data on resistance/ reflect the importance of steric approach to the binding susceptibility to the more frequent RT single mutant strains pocket and the behaviour of the molecule within the which might confer NNRTI-resistance, reported in litera- cytoplasm. In the second PC electronics features have the ture [5]. greater weight: ELUMO ( 0.45), EHOMO ( 0.45), total dipole The compounds were divided into two classes of activity: moment ( À 0.47), reflecting the importance of p stacking m > high [(H), (IC50 up to 0.095 M)] and low [(L), (IC50 during the interaction drug-receptor. Log P and heat of 0.095 mM)]. The 3D structures of all the derivatives were formation have great importance in the third and in the constructed and optimized by semiempirical methods fourth PC ( 0.59, 0.44 and 0.38,