Search

When a series of compounds of known biological activity is available and the structure of the receptor is unknown, the only rational drug design approaches applicable are Quantitative Structure-Activity Relationships (QSAR). On the other hand, when the structure of the receptor is known it is possible to use Structure-Based Drug Design (SBDD) techniques in order to gain insight into the ligand interaction. However, an increasingly common situation is that in which a full set of ligand-receptor complex structures is at hand.

Heterocyclic aromatic amines (HCA) present in cooked food, exert a genotoxic activity after metabolism (N-oxidation) by cytochrome P450 1A21. Two different 3D-QSAR approaches (COMBINE2 and GRID/GOLPE3) have been applied to a series of 12 HCAs showing different degrees of mutagenic activity (Figure 1).

3D-QSAR is an interesting and expanding discipline1–2. Nowadays software for 3D-QSAR methodologies and efficient algorithms to describe molecules and to predict biological activity are more accessible and easy to use3–5. Progress were done on the numerical description of the biological systems which now are more precise and detailed6.

Scientific interest in cannabinoids increased after the isolation of ∆9-tetrahydrocannabinol (∆9-THC) and prompted a systematic re-evaluation of their use as therapeutical agents. The pharmacological activity of cannabinoids is mediated by two recently identified cannabinoid receptors: the CB1 receptor localized in specific brain areas and the peripheral receptor CB2. In 1992 anandamide, identified as an endogenous ligand for cannabinoid receptors, was shown to share with THC most pharmacological properties in both CNS and peripheral systems1.

The optimization of pharmacokinetic properties is still one of the greatest challenges in lead optimization, and for the most part it is based on trial and error. As pharmacokinetics is closely linked with physicochemical properties, experimental design and quantitative structure-property modeling are key factors to systematically explore physicochemical property space and to establish stable, predictive models for lead optimization. However, experimental measurements of relevant parameters are often time-consuming, difficult and expensive. Furthermore, in vitro/in vivo approaches require the synthesis of compounds and cannot be used for the priorization of synthesis targets.

A method for the modeling and prediction of pharmacokinetic properties based on computed molecular interaction fields and multivariate statistics has been investigated in different experimental datasets. The program VolSurf was used to correlate 3D molecular structures with physico-chemical and pharmacokinetic properties. In membrane partitioning, VolSurf produced a two-component model explaining 94% of the total variation with a predictive q(2) of 0.90. This result was achieved without conformational sampling and without any quantum-chemical calculation. For the prediction of blood-brain barrier penetration the VolSurf model was able to predict the BBB profile for most of the drugs in the external prediction set. In Caco-2 and MDCK permeation experiments, VolSurf was used with success to establish statistical models and to predict the behaviour of new compounds. The method thus appears as a valuable new property filter in virtual screening and as a novel tool in optimizing the pharmacokinetic profile of pharmaceutically relevant compounds.

The importance of lipophilicity in QSAR and drug design demands for the availability of quick, precise and reproducible experimental approaches to quantify this physico-chemical property. The Hansch group introduced the determination of log P in the octanol-water system as the standard. The need to derive lipophilicity data for steadily increasing numbers of compounds initiated the search for both experimental and computational alternatives to octanol-water partitioning. Calculation approaches are either atom-based or use fragments; in recent time attention is paid to the impact of 3D-aspects on lipophilicity. Application of calculation approaches demands a validity check with experimental data. In this study 2 experimental (log Poct, RMW) and 17 calculation approaches (fragmental, atom-based, based on molecular properties) are investigated by regression and principal component analysis (PCA) for 159 molecules including simple structures and more complex drugs.

A large number of substituted chalcones have been synthesized and tested for antileishmanial and lymphocyte-suppressing activities. A subset of the chalcones was designed by using statistical methods. 3D-QSAR analyses using 67 (antileishmanial activity) and 63 (lymphocyte-suppressing activity) of the compounds for the training sets and 9 compounds as an external validation set were performed by using the GRID/GOLPE methodology. The Smart Region Definition procedure with subsequent region selection as implemented in GOLPE reduced the number of variables to approximately 1300 yielding 3D-QSAR models of high quality (lymphocyte-suppressing model, R2 = 0. 90, Q2 = 0.80; antileishmanial model, R2 = 0.73, Q2 = 0.63). The coefficient plots indicate that steric interactions between the chalcones and the target are of major importance for the potencies of the compounds. A comparison of the coefficient plots for the antileishmanial effect and the lymphocyte-suppressing activity discloses significant differences which should make it possible to design chalcones having a high antileishmanial activity without suppressing the proliferation of lymphocytes.

A series of [1,3]benzothiazino[3,2-a]quinoline- (5) and [3,1]benzothiazino[1,2-a]quinoline-6-carboxylic acids (10) were synthesized and evaluated for their in vitro antibacterial activity. The activity is discussed in terms of their structural features revealed by molecular orbital correlation.

The effects of data pretreatment, data scaling, and variable selection on three-dimensional quantitative structure-activity relationships derived by comparative molecular field analysis (CoMFA) using the GRID energy function were studied in detail for a set of inhibitors of the human synovial fluid phospholipase A2 (HSF-PLA2). The quality of the models was evaluated for predictive power and ability to map the receptor binding site by (a) comparison of predicted and experimental activities using cross-validation and external validation sets and (b) comparison of the regions selected in space in the CoMFA models with a crystal structure of a HSF-PLA2-inhibitor complex, with optimized comparative binding energy analysis (COMBINE) models (Ortiz et al., 1995) and with structure-activity relationships derived previously for different sets of compounds. It is found that (1) data scaling and dielectric modeling strongly influence CoMFA results. Unscaled data and a uniform dielectric constant of 4 are well suited to GRID-CoMFA studies for the present compound set. (2) The GOLPE and Q2-GRS variable selection methods select variables in roughly the same regions in Cartesian space, but they produce different models in chemometric space and differ in their sensitivity to data scaling and pretreatment and their tendency to overfitting. (3) CoMFA models are consistent with COMBINE models in that they identify approximately the same intermolecular interactions as relevant for activity. Our study provides support for the qualitative receptor-mapping properties of CoMFA models and for the validity of variable selection when applied with care and also provides guidelines for how to evaluate the quality of CoMFA models.

This report describes a new methodology aimed at grouping 3D-QSAR interaction energy descriptors into regions of neighbor variables bearing the same chemical and statistical information. These regions represent the structural variability of the series better than individual descriptor variables and can advantageously replace them in the chemometric analysis. The algorithm used to generate such regions is described, together with their application for improving the quality of GOLPE variable selection. The method is illustrated on a series of 47 glucose analogues, inhibitors of glycogen phosphorylase b, and is shown to improve both the predictive ability and the interpretability of the 3D-QSAR models obtained, comparing favorably with other methods previously described.

A major desirable characteristic of many drugs is their ability to interact specifically with only one variety of the target receptor among many others. It is remarkable that, even when accurate three dimensional structures for the target biomolecules are available, there is no well-established methodology to describe their differences and use them for the design of selectively-interacting compounds. This work presents a novel method that uses multivariate GRID descriptors and principal component analysis (PCA) with the aim of revealing the most relevant structural and physicochemical differences between biomacromolecules related to receptor selectivity. The methodology is described through an example involving the study of bacterial (Escherichia coli) and recombinant human varieties of the dihydrofolate reductase (EC 1.5.1.3, DHFR) enzyme. This analysis easily unveils the most important regions on these biomolecules which should be taken into consideration for the design of selectively interacting compounds.

Several inhibitors of the large regulatory enzyme glycogen phosphorylase (GP) have been studied in crystallographic and kinetic experiments. GP catalyses the first step in the phosphorylysis of glycogen to glucose-l-phosphate, which is utilized via glycolysis to provide energy to sustain muscle contraction and in the liver is converted to glucose. alpha-D-Glucose is a weak inhibitor of glycogen phosphorylase form b (GPb, K(i) = 1.7 mM) and acts as a physiological regulator of hepatic glycogen metabolism. Glucose binds to phosphorylase at the catalytic site and results in a conformational change that stabilizes the inactive T state of the enzyme, promoting the action of protein phosphatase 1 and stimulating glycogen synthase. It has been suggested that in the liver, glucose analogues with greater affinity for glycogen phosphorylase may result in a more effective regulatory agent. Several N-acetyl glucopyranosylamine derivatives have been synthesized and tested in a series of crystallographic and kinetic binding studies with GPb. The structural results of the bound enzyme-ligand complexes have been analysed together with the resulting affinities in an effort to understand and exploit the molecular interactions that might give rise to a better inhibitor. Comparison of the N-methylacetyl glucopyranosylamine (N-methylamide, K(i) = 0.032 mM) with the analogous beta-methylamide derivative (C-methylamide, K(i) = 0.16 mM) illustrate the importance of forming good hydrogen bonds and obtaining complementarity of van der Waals interactions. These studies also have shown that the binding modes can be unpredictable but may be rationalized with the benefit of structural data and that a buried and mixed polar/non-polar catalytic site poses problems for the systematic addition of functional groups. Together with previous studies of glucose analogue inhibitors of GPb, this work forms the basis of a training set suitable for three-dimensional quantitative structure-activity relationship studies. The molecules in the training set are void of problems and potential errors arising from the alignment and bound conformations of each of the ligands since the coordinates were those determined experimentally from the X-ray crystallographic refined ligand-enzyme complexes. The computational procedure described in this work involves the use of the program GRID to describe the molecular structures and the progam GOLPE to obtain the partial least squares regression model with the highest prediction ability. The GRID/GOLPE procedure performed using 51 glucose analogue inhibitors of GPb has good overall predictivity [standard deviation of error predictions (SDEP) = 0.98 and Q(2) = 0.76] and has shown good agreement with the crystallographic and kinetic results by reliably selecting regions that are known to affect the binding affinity.

A series of quinolone- and 1,8-naphthyridone-3-carboxylic acids, designed by previous QSAR studies and characterized by an amino group at the C-6 position instead of the usual fluorine atom, were synthesized for the first time and evaluated for in vitro antibacterial activity. All of the synthesized compounds maintain good activity against Gram-negative bacteria (Pseudomonas aeruginosa excluded), and those compounds having a thiomorpholine group as the C-7 substituent also have good activity against Gram-positive bacteria. Some aspects of structure-activity relationships associated with the C-1, C-5, C-7, and C-8 substituents are also discussed. Derivatives 18g and 38g displayed the best activity with geometric mean MICs of 0.45 and 0.66-0.76 micrograms/mL against Gram-negative and Gram-positive bacteria, respectively. This antimicrobial activity reflects their ability to inhibit bacterial DNA-gyrase. The results of this study show that, while the C-6 fluorine is still the preferred substituent, good activity can still be obtained by replacing it with an amino group.

A primary goal in any drug design strategy is to predict the activity of new compounds. Comparative molecular field analysis (CoMFA) has been used in drug design and three-dimensional quantitative structure/activity relationship (3D-QSAR) methods. The CoMFA approach permits analysis of a large number of quantitative descriptors and uses chemometric methods such as partial least squares (PLS) to correlate changes in biological activity with changes in chemical structure. One of the characteristics of the 3D-QSAR method is the large number of variables which are generated in order to describe the nonbonded interaction energies between one or more probes and each drug molecule. Since it is difficult to know a priori which variables affect the biological activity of the compounds, much effort has been devoted to developing methods that optimize the selection of only those variables of importance. This work focuses on some of the aspects involved in the selection of such variables, applied to a series of glucose analogue inhibitors of glycogen phosphorylase b, using the program GRID to describe the molecular structures and using a method of generating optimal partial least squares estimations (program GOLPE) as the chemometric tool. This data set, consisting of over 30 compounds in which the three-dimensional ligand-enzyme bound structures are known, is well suited to study the effect of different data pretreatment procedures on the final model used for the prediction of new drug molecules. By relying on our knowledge of the real physical problem (i.e., using the combined crystallographic and kinetic results), it has been shown that suitable data pretreatment and variable selection have been found that does not result in a significant loss of relevant information. Moreover, by using an appropriate scaling procedure, GOLPE variable selection minimizes the risk of overfitting and overpredicting.

Statistical design in principal properties based on D-optimality criteria are particularly appropriate for selecting the most informative molecules to be synthesized and tested in the framework of QSAR studies. Selection by D-optimal designs are better than those based on fractional factorial designs since they allow one to reduce the number of required structures, to reduce polysubstitution, to exclude molecules too difficult to synthesize and to include molecules already available and/or tested.

Tosato, M.L., Piazza, R., Chiorboli, C., Passerini, L., Pino, A., Cruciani, G. and Clementi, S., 1992. Application of chemometrics to the screening of hazardous chemicals. A case study. Chemometrics and Intelligent Laboratory Systems , 16: 155–167. The key aspects of a priority setting process for risk assessment of chemicals are reviewed. The process is based on the assumption that chemical and biological properties can be modelled within series of closely similar compounds. It rests heavily on multivariate methodologies for the development of predictive models. Statistical design, in particular, is essential for optimizing the selection of the most informative objects and variables to be used in the calibration of the models. A case study is presented where the process is applied to all (EEC-wide) commercially available halo-methanes, -ethanes and -propanes with the objective of modelling their tropospheric reactivity and, hence, estimating their lifetimes. A training set of just nine, properly informative haloalkanes was first selected from a series of 142; their reactivity was expressed in terms of the rate constant of their reaction with the OH radical, k (OH); their structure was parametrized by a selected set of variables that, by using a new statistical method, were found to optimize the k (OH) model. The experimentally validated model provided k (OH) estimates, and hence allowed the formulation of a persistency ranking for the non-tested compounds in the series, which also included several hydro-chlorofluorocarbons, H-CFCs, presently considered as possible replacement chemicals for the extremely stable, environmentally incompatible ozone-depleting CFCs.

A chemometric approach based on multivariate calibration and design in latent variables permits one to rationalize the antibacterial activity of quinolones against gram negative bacteria, ranking the relative importance of individual structural features, and suggesting a new active structure.

Experimental data on PCDD and PCDF emissions in raw gases from municipal solid waste combustion in a pilot plant were processed by Response Surface models to gain information on process parameter values which minimize both the total PCDD/PCDF concentrations and those of a selected congener group.

Chemometric methods suitable for QSAR modelling, namely partial least squares (PLS) and multiple regression analysis (MRA), are compared in terms of their prediction abilities by means of the appropriate statistical parameter called standard deviation of errors of predictions (SDEP). The results show the reliability of the PLS approach and point out the risk of overfitting with MRA, which may give rise to unreliable predictions.

This paper shows the advantages of using chemometric strategies in QSAR of NKA analogues containing D-Tryptophan and behaving as highly selective NK-2 receptor antagonists: a) detecting the sequence with optimal antagonist activity, and b) having a strategy for selecting a few most informative sequences. The data set is included on disk as RVD .DAT.

The available chemometric tools for response surface optimization in food chemistry and biotechnology are discussed, with particular reference to the treatment of sensory profile data and food and beverage quality. The review is illustrated with two examples; one concerned with the quality of Chianti, and the second with wine produced by immobilized yeast.

A set of a hundred aromatic substituents were multivariately characterized by nine descriptor variables taken from the literature. From the 9*100 data set were calculated four principal properties for the aromatic substituents as the four first dimensions in a principal components analysis, PCA. The first three principal properties were used to develop a strategy for selecting substituents from eight subgroups according to a factorial design.

The paper presents a procedure to obtain response surfaces with non-designed data. The method is based on PLS modelling of the expanded X-matrix followed by transformation of the PLS loadings into polynomial coefficients and detection of the co-ordinates of the best response within the experimental domain. The results are presented both graphically and numerically. The procedure is validated on an optimization study of the yield of an organic reaction.

The philosophy of the applications of the partial least-squares method in chemistry is outlined. The method is compared with multiple regression and illustrated by means of examples in various fields of chemistry.

Since bioavailability, activity, toxicity, distribution, and final elimination all depend on metabolic biotransformations, it would be extremely advantageous if this information to be produced early in the discovery phase. Once obtained, researchers can judge whether or not a potential candidate should be eliminated from the pipeline, or modified to improve chemical stability or safety. The use of in silico methods to predict the site of metabolism in Phase I cytochrome-mediated reactions is a starting point in any metabolic pathway prediction. This paper presents a new method, which provides the site of metabolism for any CYP-mediated reaction acting on unknown substrates. The methodology can be applied automatically to all the cytochromes whose Xray 3D structure is known, but can be also applied to homology model 3D structures. The fully automated procedure can be used to detect positions that should be protected in order to avoid metabolic degradation, or to check the suitability of a new scaffold or pro-drug. Therefore the procedure is also a valuable new tool in early ADME-Tox, where drug-safety and metabolic profile patterns must be evaluated as soon, and as early, as possible.