Mini Reviews in Medicinal Chemistry - Volume 3, Issue 8, 2003

The concept of Quantitative Structure-Activity Relationship (QSAR) has been widely revisited and extended in recent years. The original dependent variable (A) represented the desired therapeutic effect of the system when exposed to a bioactive compound. Today A can be replaced by P (property, e.g. lipophilicity), D (disposition, e.g. permeation), Pk (pharmacokinetics, e.g. distribution volume) or T (toxicology, e.g. LD50). It is therefore more appropriate (as Prof. Kubinyi has suggested) to see QSAR as a subset of the more general term SPC (Structure-Property Correlations), which covers all statistical mathematical methods used to correlate any molecular property (intrinsic, chemical or biological) to any other property, using statistical regression or pattern recognition techniques. However, despite the evolution of the original significance of A, activity remains the field of biologists, biochemists and pharmacologists and not of medicinal chemists. On the contrary, the latter are used to working with molecular descriptors, which can be defined as “numbers extracted by a well defined algorithm from a molecular representation of a complex system, i.e. the molecule” (Prof. Todeschini at http: / / www.disat.unimib.it / chm / Dragon.htm). Over the past two decades the increased demand for in silico tools to model new biological targets, combined with the development of informatics, has led to a very rapid increase in the number of molecular descriptors available. This issue brings together a number of papers that illustrate examples of the design, development, understanding and application of some widely-used molecular descriptors of pharmaceutical interest. In the first paper Prof. Gasteiger discusses the fundamental prerequisite for the determination of molecular descriptors: the various levels of structure representation. He then introduces his own theoretical approach to obtain descriptors. In the second paper, Dr. Petrauskas and his co-workers present alternative ways to generate descriptors and describe the potential of fragmental methods to obtain molecular determinants from experimental data, and their use in many areas of medicinal chemistry. Dr. Tetko then illustrates the use of Internet to access chemical information and software packages. In particular he discusses the Internet Java-based technology, which makes traditional software available worldwide. The logical consequence of this technological improvement is the ease with which a large database of compounds can now be submitted for calculation, in a few seconds obtaining an impressive spreadsheet of descriptors. Once obtained, descriptors should be used in the most productive way. By definition, molecular descriptors are a source of chemical information. The comparison between calculated and experimental data (if available, of course) besides validating the calculations, can provide information derived from their differences. In the fourth paper we describe this new way of obtaining structural information in the case of descriptors related to lipophilicity. Dr. Migliavacca then explains how chemometric techniques are able to handle a plethora of data and reduce redundant tables to a limited series of useful numbers and / or plots. As evidenced in the first paper, the calculation of molecular descriptors requires a clear representation of molecular structure, which is far from obvious in the case of metallo-organic compounds. In the sixth paper Dr. Maiocchi discusses the use of modeling techniques in designing gadolinium(III)-based contrast agents, the most widely used compounds in Magnetic Resonance Imaging (MRI). ADME prediction (Absorption, Distribution, Metabolism and Excretion) is one of the today's biggest challenges for medicinal chemists, and thus great efforts have been made in the application of computational technology to this field. In the last paper Dr. Lombardo and his co-workers review the use of molecular descriptors to determine in silico ADME properties for potential leads. I am very grateful to all contributors for the enthusiasm they have shown in accepting my invitation and in preparing their papers but I also would like to apologize to many other important scientists working in the field for my having been unable to contact all of them. I finally thank my colleague and friend Sonja Visentin for encouraging me to accept this exciting task.

After the identification of a biological target, drug design is to analyze the relationships between the structure of potential ligands and their biological activity. A hierarchy of structure representation is presented here considering either the constitution of a molecule, its 3D structure, or the molecular surface. At each level, a variety of physicochemical effects can be accounted for. Furthermore, the special requirements of learning algorithm, such as neural networks, are taken into consideration. Application to problems from combinatorial chemistry, lead identification, high-throughput screening, and prediction of ADME-Tox properties are given.

The current mini-review explains how fragmental methods (FMs) can be used in the analysis and prediction of physicochemical properties and biological activities. The considered properties include log P, solubility, pKa, intestinal permeability, P-gp substrate specificity and toxicity. The focus will be a description of a “mechanistic” approach, which implies a gradual reduction of alternative explanations for any property or activity. This means a flexible construction of fragmental parameters using large amounts of experimental data. Since biological activities involve multiple (unknown) target macromolecules with multiple binding modes, a stepwise classification (C-SAR) analysis is most useful. It involves the following procedures: (i) construction of physicochemical profiles using parameters that can be reliably predicted, (ii) identification of reactive functional groups and the largest active skeletons, (iii) generalization of these groups and skeletons in terms of “site-specific physicochemical profilingrdquo;. This entails a dynamic construction of 2D pharmacophores that can be converted into 3D models.

This article analyses molecular property calculation resources available on the Internet. The first section summarizes the on-line database resources that could be useful to search molecular and biological properties of chemicals, and indicates some principal databases with physicochemical, thermochemical, toxicity, cancer and HIV data. The second section overviews popular standalone programs for calculation of molecular descriptors. Some of these programs can be downloaded for free and used as standalone applications for calculation of molecular descriptors. The third section describes on-line tools for the prediction of molecular properties, activities and calculation of molecular descriptors. Analysis of emerging tools that can be useful to developing new on-line servers for the prediction of molecular parameters and properties is also given.

This review is organized in three parts: firstly there is a general overview of recent developments in lipophilicity written to induce medicinal chemists to question what they want to obtain from this kind of study; secondly, the state-of-the-art of experimental and computational determination of log P is briefly reviewed; finally, some applications are discussed to illustrate how much information can be extracted from lipophilicity, and to highlight the difficulty of obtaining a reliable, general method to work with.

The number of articles concerning optimization and applications of multivariate techniques in drug discovery testifies the growing importance attributed to these methods. This mini review focuses on some of the basic and most employed multivariate techniques in drug discovery research. Examples from the literature were selected to illustrate a number of potential applications.

Nuclear Magnetic Resonance Imaging (MRI) is a very useful tool in modern medical diagnostics, especially when gadolinium(III)-based contrast agents are administered to the patient with the aim of increasing the image contrast between normal and diseased tissues. The main purpose of this review is to show that a new generation of these contrast agents could be developed by making greater use of soft modelling techniques such as QSAR / QSPR after a suitable description of their molecular structure.

A critical review of a very recent work in the field of in silico ADME prediction is presented with emphasis on the work published during the period 2000-2002, and several other review articles are mentioned in order to offer a broader view of the field. We find that not much progress has been made in developing robust and predictive models, and that the lack of accurate data, together with the use of questionable modeling end-points, has greatly hindered the real progress in defining generally applicable models. Due to the largely empirical nature of QSAR / QSPR approaches, general and truly predictive models for complex phenomena, such as absorption and clearance, may still be chimeric. The development of local models for use within focused chemical series may be the most appropriate way of utilizing in silico ADME predictions, once experience and data have been gained on a given project and / or structural class.