Current Topics in Medicinal Chemistry - Volume 2, Issue 12, 2002

There exists a huge attrition rate of molecules in clinical trials. It was expected that high-throughput screening and combinatorial chemistry would make the task of producing drugs easier. However, the efforts of the past decade have not been an unvarnished success. As a result, a lot of experimental and computational efforts are currently being directed at determining the basic requirements for a molecule to become a drug. Here we will review the physiological, structural, and other requirements for obtaining a molecule that will be successful in the clinic. Following this we will provide a description, analysis, and commentary on the computational efforts in this direction. We will focus both on the traditional computational chemistry perspective of starting from the structure of the molecule as well as the traditional computational pharmaceutical scientist's perspective of physiologically based simulations. We end with a few comments about the future and some ideas on re-organzing the pharmaceutical enterprise.

The high attrition rate of drug candidates during clinical trials for poor pharmacokinetic and metabolic properties has created a need to do these studies as early as it is possible during the drug discovery process.In addition the most successful drug is often not the most potent one but the one that has the suitable level of potency, safety, and pharmacokinetics. Science and technology development during the last few years and the generation of last databases and information has created the basis for doing early experimental PK and ADME studies in addition to eADME. Similarly, testing safety features as early as possible is key to affordable drug discovery and development.Throughput and cost are crucial for early application. In silico methods have by far the highest throughput, followed by the in vitro and in vivo approaches. On the other hand, with regard to relevance and reliability of data the ranking is the opposite. The great challenge for in silico methods is generation of models that correlate more closely with in vivo systems. For the in vitro assays increasing the throughput is an absolute must.Ex silico methods that combine in silico predictions with experimental methods are new additions to the scientific repertoire (e.g. Chromatographic Hydrophobicity Index that is deduced from the reverse phase HPLC data can be used for calculation of lipophilicity). The emerging new approaches have clear impact on the design of early stage screening and combinatorial libraries. In addition to the Lipinski's rules descriptors such as number of rotatable bonds, number of aromatic rings, branching behavior and polar surface area (PSA) are commonly used is the drug design process.

We review the prominent technologies in virtual screening, and their applications in drug discovery.

With computer-aided drug design established as an integral part of the lead discovery and optimization process, pharmacophores have become a focal point for conceptualizing and understanding receptor-ligand interactions. In the structure-based design process, pharmacophores can be used to align molecules based on the threedimensional arrangement of chemical features or to develop predictive models (e.g., 3DQSAR) that correlate with the experimental activities of a given training set. Pharmacophores can be also used as search queries for retrieving potential leads from structural databases, for designing molecules with specific desired attributes, or as fingerprints for assessing similarity and diversity of molecules. This review article presents a historical perspective on the evolution and use of the pharmacophore concept in the pharmaceutical, biotechnology, and fragrances industry with published examples of how the technology has contributed and advanced the field.

Whole-molecule descriptors are obtained computationally from molecular structures using a variety of programs. Their applications are reviewed in the areas of solubility, bioavailability, bio- and nonbio-degradability and toxicity.

In this review, the evolution of QSAR is traced from the insightful observations of Crum-Brown and Frazier to Hammett's critical equations and finally Hansch's seminal contributions on hydrophobicity and modelling of biological activity based on extrathermodynamic principles. Today's QSAR models can stand alone, augment other graphical approaches or be examined in tandem with equations of a similar mechanistic genre to truly reveal the power of the paradigm. This review will focus on the three standard classifications routinely used in QSAR analysis - electronic, hydrophobic, and steric, as well as topological indices.Electronic parameters will focus on Hammett sigma constants and their numerous variations. Dipole moments, hydrogen bond descriptors and quantum chemical indices as well as applications of their utilization will be described. The hydrophobicity parameter will be examined by tracing its early history, its operational definition and its determination by either experimental methods or computational calculations. Steric parameters, which run the gamut from size to shape, will be described by Taft's, Hancock's, Charton's, Fujita's, Verloop's and Simon's contributions. Topological effects, delineated by connectivity indices, kappa shape and electrotopological indices of Kier and Hall are also described. Examples of QSAR models incorporating most of these parameters are reviewed. In cases where the 95% confidence intervals of variables are available, they are listed in parentheses. A brief Comparative QSAR analysis of non-nucleoside reverse transcriptase inhibitors (NNRTI's) is outlined and various models obtained by different groups examining 4, 5, 6, 7-tetrahydro-5- methylimidazo [4, 5,1-j,k][1,4] benzodiazepin-2(1H)-ones (TIBO) and 1-[(2-hydroxyethoxy)methyl]-6- (phenylthio)-thymine (HEPT) derivatives are compared for mechanistic insight that could be useful in the process of inhibitor design.

Quantitative structure-activity relationships (QSAR) have played an important role in the design of pharmaceuticals and agrochemicals. All QSAR techniques assume that all the compounds used in analyses bind to the same site of the same biological target. However, each method differs in how it describes structural properties of compounds and how it finds the quantitative relationships between the properties and activities. The Hansch-Fujita approach, the so-called classical QSAR, is a representative of QSAR methods. Despite the usefulness, classical QSAR techniques cannot be applied to all datasets due to the lack of availability of physicochemical parameters of the whole molecule or its substituents and often it is difficult to estimate those values. In addition, molecular properties based on the three dimensional (3D) structure of compounds may be useful in describing the ligand-receptor interactions. Recently, a variety of ligand-based 3D-QSAR methods such as Comparative Molecular Field Analysis (CoMFA) have been developed and widely used in medicinal chemistry. This review describes different 3D-QSAR techniques and indicates their advantages and disadvantages. Several studies about 3D-QSAR of ADME-toxicity and perspective of 3D-QSAR are also described in this review.