Current Computer - Aided Drug Design - Volume 3, Issue 4, 2007

Increasing efforts to estimate the activity or any other biological property of a given compound have been made since the advent of in silico approaches for drug design. In ligand-based QSAR/QSPR methodologies, descriptors are used to correlate samples set with the corresponding dependent variables (bioactivities, toxicity, etc.), in lieu of testing experimentally the response of a drug-like compound. A large amount of information is usually required or produced to achieve such correlation, and then multivariate analysis has been invoked to manipulate the generated data in order to investigate their variance. Accordingly, chemometric techniques for regression, data exploration and variable selection must be capable to reduce dimensionality and allow the building of predictive models. The present Hot Topic Issue of Current Computer-Aided Drug Design (CC-ADD) provides comprehensive reviews of several multivariate QSAR methods, as well as variable selection and docking studies, covering useful aspects in multivariate modeling applied to drug discovery. In addition to the methodologies used for fast developing computer-aided drug design, their applications and end results are also presented to the scientific community involved in the prediction of drug targets. In this issue, Arodz and Dudek provide an excellent overview of multivariate quantitative structure-activity relationships involving simultaneous modeling of activities toward several related endpoints, and a comparison with univariate models is also given. The authors also focus on neural networks and other non-linear methods to predict all activities simultaneously with good accuracy. In another article, Freitas reviews the various chemical descriptors used to derive QSAR models, from classical to multidimensional predictors. Special emphasis is given to MIA (multivariate image analysis) descriptors, which have shown to exhibit some operational advantages over well established protocols. Remarks and applications of the MIA-QSAR method are presented, and its potentialities and limitations are discussed. Variable selection plays an important role when building a significant QSAR model by selecting important descriptors from descriptors pool. The article of Funatso et al. is focused on this topic, such as genetic algorithm and programming, simulated annealing, and so on. In addition, 3D-QSAR and related methods are presented, and details about alignment and new methods using molecular surface properties are also taken into account. In the fourth article, Mager highlights the use of simultaneous one-regression/one-observation leaving-out resampling regression analysis (SimR) to the selection of suitable chemical descriptors from a pool of variables. This method selects well and shows advantages when compared to GRNN, a combination of generalized regression (GR) and artificial neural networks (ANNs). Additionally, the prediction of protein subdomains as potential molecular drug targets is demonstrated and the protein-ligand docking study exemplified to human P2X7 (h-P2X7) receptor subunit and a series of novel 4,5-diarylimidazoline inhibitors. Overall, I hope that readers would enjoy reading these amazing contributions, expectedly valuable to those involved in the area of drug design and, particularly, in multivariate QSAR.

This review describes an overview of multivariate QSAR methods, from classical analysis to 3D approaches and new perspectives. Data exploration, multivariate regression and molecular descriptors are some topics also appraised here. Special emphasis is given to a recently developed 2D image-based approach, known as MIA-QSAR, which is an improved method in many aspects, namely computing cost, simplicity and prediction performance. Remarks on the MIA-QSAR technique, numerical examples and comparison with traditional methodologies, in addition to a description of limitations and potentialities of this method, are also discussed.

Multivariate quantitative structure-activity relationship (QSAR) modeling, involving simultaneous modeling of activities towards several related endpoints, has emerged recently as an alternative to creating a group of separate models of each activity. The development of multivariate QSAR modeling has been driven by three factors. First, the number of aspects considered vital at earlier stages in the drug development pipeline has increased. Second, advanced screening technology has shifted the rate limiting step of drug discovery and development to other areas. Screening compounds for multiple properties has resulted in the availability of multi-endpoint datasets. Finally, the statistical and computational methods used in data analysis have evolved to allow for handling an increased complexity associated with multi-task prediction. In this review, we outline the justifications for the use of multivariate QSAR modeling. We review the techniques used for developing such models and their applications in drug discovery. We also summarize the methods for visual analysis of multivariate datasets. We focus on neural networks and other advanced, non-linear methods gaining popularity in the QSAR community, while also describing established linear techniques.

This review deals with three problems: the selection of suitable chemical descriptors from a pool of variables by a simultaneous one-regression/one-observation leaving-out resampling, the comparison of the results with a generalized-regression artificial-neural network, using an unconstrained genetic algorithm (GRNN), and the prediction of protein subdomains as potential molecular drug targets. As an example, the human P2X7 (h-P2X7) receptor subunit and a series of novel 4,5-diarylimidazoline inhibitors [Merriman et al., Bioorg. Med. Chem. Lett., 15, 435 (2005)] is used. GRNN ignores relevant and add noisy descriptors although the goodness-of-fit criterion is large. Therefore, GRNN is considered as supplementary tool which cannot replace the traditional QSAR methodology. Simultaneous one-regression/one-observation leaving-out resampling shows that the h-P2X7 inhibitory activity of 4,5- diarylimidazolines depends on electronic, steric and hydrogen-bonding properties of the substituents. Diagnostic statistic examines the validity of the results. The inhibitors are probably bound to sites that are located mainly in the subdomains 344-347 and 370-375 of h- P2X7.

Quantitative structure-activity relationships (QSAR) are one of the most important methodologies for rational drug design. In QSAR, compounds are represented by chemical structure descriptors, and then statistical models are built to predict biological activities of candidate structures. In this paper, two principal topics in QSAR, variable selection and 3D-QSAR, are picked up and are reviewed in recent trend. The aim of variable selection is to construct a significant QSAR model by selecting important descriptors among from descriptor pool. Until now, many variable selection methods have been developed and proposed. On the other hand, molecular alignment is important factor of 3D-QSAR analysis because appropriate alignment is usually required to construct proper 3D-QSAR models. In addition, we review new QSAR methods using molecular surface properties, alignment independent QSAR methods, and 4D-QSAR methods.

Recently, a new promising nonlinear method, the support vector machine (SVM), was proposed by Vapnik. It rapidly found numerous applications in chemistry, biochemistry and pharmacochemistry. Several attempts using SVM in drug design have been reported. It became an attractive nonlinear approach in this field. In this review, the theoretical basis of SVM in classification and regression is briefly described. Its applications in QSPR/QSAR studies, and particularly in drug design are discussed. Comparative studies with some linear and other nonlinear methods show SVM's high performance both in classification and correlation.

Computer-aided molecular design (CAMD) is becoming increasingly important to the drug discovery process. Although molecular mechanics (MM) has traditionally been the computational method of choice in medicinal chemistry, the MM method has significant deficiencies when used to study electron-based properties within the drug-receptor microenvironment. Quantum mechanical methods represent a solution to this problem, but QM methods are frequently too computationally intensive to be useful for molecular systems of interest to medicinal chemists. However, over the past five years, density functionally theory (DFT) has emerged as a QM method that is both sufficiently rigorous and efficient to be used for pharmaceutical problems. DFT is a popular method for accurately describing biologically relevant molecular systems at a reasonable computational cost. In this review, the potential applications of DFT to drug discovery are systematically discussed. First, the basis of DFT is reviewed. Subsequently, the accuracy of DFT for the study of molecular properties specific to drug design are reviewed in comparison to experimental results as well as other ab initio methods. The use of DFT for molecular modeling in medicinal chemistry is also reviewed. Finally, practical considerations for beginning DFT users and a summary of DFT performance are presented.

The identification of the molecular components involved in the aberrant processes that control proliferation, differentiation and apoptosis, is necessary for the development of chemotherapeutic interventions to restore or to destroy selectively the transformed cells. The discovery of new chemotherapeutic agents is probably one of the most reliable ways to improve our success against cancer, and intelligent drug design is a key factor to achieve this goal. Thus, the identification of novel targets for anticancer drug discovery is needed. Here we provide evidence that choline kinase (ChoK) is a novel target for the design of antitumor drugs. In this review we present the evolution of ChoK inhibitors by using the Hansch approach, starting from hemicholinium-3 (HC-3) as a lead compound. To start with we synthesized and evaluated ten bis-quaternary derivatives, in which the modifications affect both the spacer and the two cationic heads of the prototype. In the second phase 56 biscationic dibromides with distinct polar heads [bis(4-substituted)pyridinium, bis(4-substituted)quinolinium, and bisisoquinolinium moieties] and several spacers were synthesized and assayed for biological activity. This oriented synthesis produced 45 inhibitors of ChoK with antitumor activity against the HT-29 cell line. Finally, 40 bisquinolinium compounds were prepared and the corresponding QSAR equation was obtained for the whole set of compounds for the antiproliferative activity, the electronic parameter σR of R4, the molar refractivity of R8, and the lipophilic parameters clog P and πlinker. The most potent antiproliferative agent so far described shows IC50 = 0.20 μM, while its theoretical value is 0.45 μM.