Current Topics in Medicinal Chemistry - Volume 10, Issue 6, 2010

The focus of this hot topic issue is to give an overview on the current status of in silico technologies widely applied in drug discovery community, including QSAR, Bioisosterism, Shape-matching, Fragment-based methodologies, and in silico approaches for GPCRs-targeted drug discovery at a point when we witness a renaissance of the computational approach in the field. In mind that the aim of the issue is to provide a deeper analysis on the specific areas interested by discovery community, the authors contributing to the reviews are all have active research projects in the specific area they are reviewing. With its existence over half a century, QSAR has been the single most-used computational technique and is growing further with newer algorithms and strategies. Sprous et al. focus on the QSAR models based on large datasets and applied to broad problems such as property modeling and target family focus library design. As a consequence of this theme, their review surveys 2D descriptors and QSAR validation methodologies as necessary background for the necessary tools used for models based on large, diverse datasets and applied not for retrospective modeling and analysis, but applied for decision making. G-Protein coupled receptors (GPCRs) are targeted by ∼30% of all marketed drugs, covering a wide range of therapeutic indications, including neurological disorders, cardiovascular diseases, metabolic diseases, inflammation and pain. The role of GPCRs in diverse signaling pathways, together with the fact that the vast majority of potential GPCR targets are still unaddressed by marketed drugs, positions this receptor family at the heart of current pharmaceutical industry efforts. Kalid et al. give us a comprehensive overview of the current state-of-the-art in silico drug-discovery in GPCR field and bring together different approaches currently being utilized for 3D structure generation with their advantages (accuracy versus speed), limitations and pitfalls. Then their attention turns to their in-house program, PREDICT™ , an ab-initio methodology for predicting the structure of the trans-membrane (TM) portion of GPCR receptors. While structure-based drug-discovery is facilitated by the recent release of new GPCR crystal structures, it is still by no means an easy task. Therefore, they concluded their review with detailed analyses of the progresses of their diverse GPCR projects. Bioisostere recognition has long been treated as the art of expertise in medicinal chemistry field since the discipline ever evolved. Popelier et al. first define the scope of bioisosterism to pairs of fragments that have similar impact on molecular activity, not whole molecules. This review provides an overall picture of the broad and increasingly varied selection of computational approaches available to find bioisosteric replacements, discusses the current status of the rational and cheminformatics approaches. Tools, such as Brood of OpenEye Scientific, the Drug Rings Database by GSK for scaffold replacement, Quantum Isostere database for linker replacement, full ab initio calculation, descriptor and fingerprint-based technologies, Drug Guru package from Abbott as a SMIRK-based expert system, are all been critically reviewed. Scaffold hopping is to identify alternative molecular scaffolds or ‘cores’ that will form the backbone of new families of compounds with similar biological activity to a known reference. Therefore, it is loosely fall into bioisostere scope. It is also reviewed with a focus on the application of SHOP and GANDI packages. Shape complementarity plays an important role in the process of molecular recognition. In recent years, molecular shape technology has become an indispensable tool in drug discovery and development for target-to-hit, hit-to-lead and lead optimization projects, as well as in ADMET modeling. The commercialization of the ROCS (Rapid Overlay of Compound Structures) technology and the description of the USR (ultrafast shape recognition) method have marked the new state-of-theart of molecular shape technology, and new tools integrating these ideas will ultimately make the shape technology a great complement to, sometimes a substitute for, structure-based docking tools. Zheng et al. give a concise review on the algorithms developed for molecular shape representation and comparison, superposition-based and superposition-free algorithms for shape matching., and their attention quickly shift to recent advances and future prospects of methods in various shape tools, including ROCS, topomer shape similarity, USR, the shape signatures method. They complete their review with a insightful summary of their practical applications. Fragment-based methods have rapidly established their prominent presence in drug discovery. Although it is too early to have yielded a marketed drug, they have resulted in a significant number of clinical candidates and many more in preclinical development. It has long been recognized that there is no universal perfect docking program to fit all the drug discovery projects, and the screening result depends on the combination of the docking program used and the target protein. Even a slight structural change around the binding site sometimes has a large effect on the docking scores. In FBDD, the compounds are smaller than those for conventional drug screening. This makes it even harder to perform the protein-compound docking calculation. Fukunishi gives a comprehensive review on various computational methods proposed for docking-pose prediction and their usefulness in the fragment-based drug design (FBDD). Readers who are directly involved in in silico FBDD will find the section of problems and solutions section especially valuable.

The field of quantitative structure activity relationships (QSAR) has evolved into an integral tool for pharmaceutical discovery. It is presently an accessible technology, as can be shown by the number papers which are easily found through PubMed literature searches. At one level, QSAR is used routinely and invisibly as an aid for the bench chemist for logP, logS, pKa/pKb, metabolic stability and other such properties. Chemoinformaticians and computational chemists develop models from scratch for less routine purposes associated with lead optimization around a single target or library design around a target family such as kinase, ion channel or GPCR inhibitors. Regardless of the differences in frequency of use and the end user, any successful QSAR is successful because it rests on appropriate mathematics linking valid data and relevant descriptors. Though success is defined by the end user, the QSAR originator is well advised to validate his model and understand how it performs in different situations. The present review will cover QSAR from the ground up as it is used in pharmaceutical research environments. It will focus towards larger dataset methodologies (a minimum 100 of compounds) and by consequence will focus on 2D descriptors. It will start with the critical base of data, descriptors, equations and validation methods. It will review the broadly used and invisible QSARs for logP, pKa/pKb and metabolic stability. The review will then present progress in QSARs of broad interest which are under active development: 1) hERG liability models, 2) modeling for 2a) drug-likeness and related properties, 2b) kinase ligand likeness and 2c) GPCR ligand likeness.

In silico drug discovery is a complex process requiring flexibility and ingenuity in method selection and a careful validation of work protocols. GPCR in silico drug discovery poses additional challenges due to the paucity of crystallographic data. This paper starts by reviewing selected GPCR in silico screening programs reported in the literature, including both structure-based and ligand-based approaches. Particular emphasis is given to library design, binding mode selection, process validation and compound selection for biological testing. Following literature review, we provide insights into in silico methodologies and process workflows used at EPIX and previously at PREDIX to drive over 20 highly successful screening and lead optimization programs performed since 2001. Applications of the various methodologies discussed are demonstrated by examples from recent programs that have not yet been published.

This article provides an overview of the broad and increasingly varied selection of computational approaches available to find bioisosteric replacements for fragments of bioactive compounds. The rapidly increasing number and diversity of methods has provided medicinal chemists with a powerful range of commercial and academic tools to aid in the optimization of lead compound activity and ADMET properties for drug design. We discuss methods with fundamentally different philosophies, ranging from evaluation of similarity in a calculated property space to cheminformatics analysis of pharmaceutical compound databases. We also discuss the incorporation, within these methods, of a whole spectrum of experimental and calculated data to describe fragment chemistry and compound activity. Despite the growing sophistication of available techniques, there remains much scope for further development and especially for deeper validation of the efficacy of different approaches in what seems set to remain an expanding field.

Shape complementarity is a critically important factor in molecular recognition among drugs and their biological receptors. The notion that molecules with similar 3D shapes tend to have similar biological activity has been recognized and implemented in computational drug discovery tools for decades. But the low computational efficiency and the lack of widely accessible software tools limited the use of early shape-matching algorithms. However, recent development of fast and accurate shape comparison tools has changed the landscape, and facilitated the wide spread use of both the ligand-based and receptor-based shape-matching technologies in drug discovery. In this article, we summarize some of the well-known shape algorithms. We first describe the computational principles for both the superposition-based and the superposition-free shape-matching methods. These include ROCS (Rapid Overlay of Chemical Structures), SQ, and the CatShape method in the former category; and the shape signatures algorithm and USR (Ultrafast Shape Recognition) that belong to the latter category. We then highlight some recent validation studies and practical applications of various shape technologies. Because of the rapid development of modern shape-matching algorithms, and the increasingly affordable computational resources and software tools, we anticipate much broader use of the molecular shape technologies in future drug discovery. They will be especially useful in chemogenomics research, where large scale associations between small molecules and protein targets are studied. Thus, molecular shape technologies, together with well-defined pharmacophore constraints, can afford both efficient and effective means for drug discovery and chemical genomics research.

For fragment-based drug development, both hit (active) compound prediction and docking-pose (protein-ligand complex structure) prediction of the hit compound are important, since chemical modification (fragment linking, fragment evolution) subsequent to the hit discovery must be performed based on the protein-ligand complex structure. However, the naive protein-compound docking calculation shows poor accuracy in terms of docking-pose prediction. Thus, postprocessing of the protein-compound docking is necessary. Recently, several methods for the post-processing of protein-compound docking have been proposed. In FBDD, the compounds are smaller than those for conventional drug screening. This makes it difficult to perform the protein-compound docking calculation. A method to avoid this problem has been reported. Protein-ligand binding free energy estimation is useful to reduce the procedures involved in the chemical modification of the hit fragment. Several prediction methods have been proposed for high-accuracy estimation of protein-ligand binding free energy. This paper summarizes the various computational methods proposed for docking-pose prediction and their usefulness in FBDD.