Current Topics in Medicinal Chemistry - Volume 6, Issue 15, 2006

In silico (computational) simulation of chemical-biological interactions underpins efforts to remedy the escalating average cost and timeframe required to develop a marketable pharmaceutical. Chemicals with poor drug metabolism properties in humans are unfavourable for medicinal use, typically presenting problems with bioavailability, half life, inter-individual variability, and drug-drug interactions. As a result, it is increasingly seen as prudent to screen chemicals for these properties as early as possible in the drug discovery and development process. In silico methods are generally orders of magnitude faster and less expensive than the best in vitro methods, enabling screening of drug metabolism properties on much larger numbers of chemicals and much earlier in the drug discovery pipeline. Existing in silico screens span a range of different metabolic properties for an array of different phase I and phase II enzymes. The Cytochrome P450 (CYP) family of phase I enzymes plays the greatest role in drug metabolism in humans and consequently the majority of research has focused on this enzyme. Nevertheless, Phase 2 enzymes are increasingly recognized as important and research in this area is growing. One of the most common goals of in silico screens is the prediction of the ability of a chemical to inhibit a drug metabolizing enzyme. Other important metabolic properties studied include regioselectivity of metabolism, the ability of a chemical to be metabolised, metabolic stability, and induction of drug metabolising enzymes. A variety of in silico methodologies have been applied to predict these properties, including two- and threedimensional quantitative structure activity relationships (2D- and 3D-QSAR), pharmacophore modelling, quantum chemistry, protein modelling, and docking simulations. The major current challenges of in silico screens include validation, accuracy and interpretability. In this issue of Current Topics in Medicinal Chemistry, titled ‘Predicting drug metabolism in silico’, the current major issues, directions, techniques, and applications of this area are reviewed in detail. Although each review has a different focus, there is sufficient overlap to appreciate the variety of perspectives existing in the field today. Chohan, Paine and Waters begin the issue with an in-depth review of the contemporary 2D QSAR, 3D QSAR, and pharmacophore approaches that have been applied to gain insight into the molecular features influencing binding and metabolism by the major human phase 1 and phase 2 drug metabolising enzymes Fox and Kriegl follow on by reviewing a variety of machine learning techniques that commonly underlie recent global QSAR studies on drug metabolism properties. The application and recent progress of these methods for the prediction of drug metabolism properties are considered in detail. This leads into a more specific review by Yap, Xue, Li and Chen on the use of support vector machines (SVM) - one of the most popular and powerful machine learning methods of the moment - for the prediction of CYP substrates and inhibitors. The discussion of SVM methodology, performance, difficulties and future prospects are a must-read for any researcher considering using machine learning methods for the prediction of drug metabolism. 1568 Current Topics in Medicinal Chemistry, 2006, Vol. 6, No. 15 Editorial Arimoto details the key findings of recent pharmacophore, QSAR and structure-based modeling undertaken to understand and predict metabolic properties of the major human CYP isoforms; CYP1A2, 2A6, 2C9, 2D6 and 3A4. Subsequently, Marechal and Sutcliffe review the structure-based modeling of human CYPs, focusing particularly on CYP2D6. The recent crystallization of a number of mammalian and human CYPs has been a significant step forward in using structure-based methods to understand the active site of CYPs and predict chemical binding and metabolism. The use of in silico methods for the prediction of drug metabolism induction is reviewed in detail by Schuster, Steindl and Langer. Such models are complementary to those described in other reviews in this issue. In this review there is a focus on the modelling methods used, their applicability, limitations and recent applications. Westwood, Kawamura, Fullan, Russell and Sim conclude the issue with a review on ligand- and structure-based modelling of N-acetyltransferases for the prediction of chemical binding and metabolism. N-acetyltransferase is an important phase 2 drug metabolising enzyme family and the variety and depth of research described herein indicates the potential for future work on the in silico prediction of drug metabolism by phase 2 enzymes..

This review of 61 references delineates contemporary computation quantitative structure activity relationship (QSAR) approaches that have been used to elucidate the molecular features that influence the binding and metabolism of a compound by the major phase 1 and phase 2 metabolising enzymes; Cytochrome P450 (CYP) and UDPglucuronosyltransferase (UGT), respectively. Contemporary studies are applying 2D and 3D QSAR, pharmacophore approaches and nonlinear techniques (for example: recursive partitioning, neural networks and support vector machines) to model drug metabolism. Furthermore, this review highlights some of the challenges and opportunities for future research; the need to develop ‘global’ models for CYP and UGT metabolism and to extend QSAR for other important metabolising enzymes.

The computational assessment of drug metabolism has gained considerable interest in pharmaceutical research. Amongst others, machine learning techniques have been employed to model relationships between the chemical structure of a compound and its metabolic fate. Examples for these techniques, which were originally developed in fields far from drug discovery, are artificial neural networks or support vector machines. This paper summarizes the application of various machine learning techniques to predict the interaction between organic molecules and metabolic enzymes. More complex endpoints such as metabolic stability or in vivo clearance will also be addressed. It is shown that the prediction of metabolic endpoints with machine learning techniques has made considerable progress over the past few years. Depending on the procedure used, either classification or quantitative prediction is possible for even large and diverse compound sets. Together with the expanding experimental data basis, these in silico methods have become valuable tools in the drug discovery and development process.

Cytochrome P450 enzymes are responsible for phase I metabolism of the majority of drugs and xenobiotics. Identification of the substrates and inhibitors of these enzymes is important for the analysis of drug metabolism, prediction of drug-drug interactions and drug toxicity, and the design of drugs that modulate cytochrome P450 mediated metabolism. The substrates and inhibitors of these enzymes are structurally diverse. It is thus desirable to explore methods capable of predicting compounds of diverse structures without over-fitting. Support vector machine is an attractive method with these qualities, which has been employed for predicting the substrates and inhibitors of several cytochrome P450 isoenzymes as well as compounds of various other pharmacodynamic, pharmacokinetic, and toxicological properties. This article introduces the methodology, evaluates the performance, and discusses the underlying difficulties and future prospects of the application of support vector machines to in silico prediction of cytochrome P450 substrates and inhibitors.

Cytochrome p450 (CYP) enzymes are predominantly involved in Phase 1 metabolism of xenobiotics. As only 6 isoenzymes are responsible for ∼90% of known oxidative drug metabolism, a number of frequently prescribed drugs share the CYP-mediated metabolic pathways. Competing for a single enzyme by the co-administered therapeutic agents can substantially alter the plasma concentration and clearance of the agents. Furthermore, many drugs are known to inhibit certain p450 enzymes which they are not substrates for. Because some drug-drug interactions could cause serious adverse events leading to a costly failure of drug development, early detection of potential drug-drug interactions is highly desirable. The ultimate goal is to be able to predict the CYP specificity and the interactions for a novel compound from its chemical structure. Current computational modeling approaches, such as two-dimensional and three-dimensional quantitative structure-activity relationship (QSAR), pharmacophore mapping and machine learning methods have resulted in statistically valid predictions. Homology models have been often combined with 3D-QSAR models to impose additional steric restrictions and/or to identify the interaction site on the proteins. This article summarizes the available models, methods, and key findings for CYP1A2, 2A6, 2C9, 2D6 and 3A4 isoenzymes.

The cytochromes P450 (CYPs) comprise a vast superfamily of enzymes found in virtually all life forms. In mammals, xenobiotic metabolising CYPs provide crucial protection from the harmful effects of exposure to a wide variety of chemicals, including environmental toxins and therapeutic drugs. Elucidating the structural features of CYPs that contribute to their metabolism of structurally diverse substrates impacts on the rational design of improved therapeutic drugs and specific inhibitors. Models capable of predicting the possible involvement of CYPs in the metabolism of drugs or drug candidates are thus important tools in drug discovery and development. Ideally, functional information would be obtained from crystal structures of all the CYPs of interest. Initially only crystal structures of distantly related bacterial CYPs were available - comparative modelling techniques were used to bridge the gap and produce structural models of human CYPs, and thereby obtain some useful functional information. A significant step forward in the reliability of these models came six years ago with the first crystal structure of a mammalian CYP, rabbit CYP2C5, followed by the structures of five human enzymes, CYP2A6, CYP2C8, CYP2C9, CYP2D6 and CYP3A4, and a second rabbit enzyme, CYP2B4. The evolution of a CYP2D6 model, leading to the validation of the model as an in silico tool for predicting binding and metabolism, is presented as a case study.

The inducibility of drug-metabolizing enzymes and transporters by numerous xenobiotics has become a vital issue to be considered in the drug development process. Activation of so-called orphan nuclear receptors has been identified to result in increased expression of these detoxifying systems and consequently altered drug levels in the human body. In order to anticipate such mechanisms already in early stages of drug design and to avoid dangerous drug-drug interactions, reliable in silico simulation tools are highly desirable. This review aims to give an introduction into induction of drug metabolism and transport and focuses on computer-assisted molecular modeling prediction techniques, on their applicability and limitations, on recent case studies, and on success stories.

Arylamine N-acetyltransferases (NATs) are a family of phase II drug-metabolising enzymes which are important in the biotransformation of various aromatic and heterocyclic amines and hydroxylamines, arylhydrazines and arylhydrazides. NATs are present in a wide range of eukaryotes and prokaryotes. Humans have two functional NAT isoforms, both of which are highly polymorphic. The pharmacogenetics of NATs is an area which has been extensively studied. The determination of the X-ray crystal structure of NAT from Salmonella typhimurium led to the identification of the catalytically essential triad of residues: Cys-His-Asp, which is present in all functional NAT enzymes. Recent cocrystallisation data and in silico docking studies of NAT from Mycobacterium smegmatis with substrates and inhibitors have aided the identification of important contact residues within the active site. The X-ray crystal structures of four prokaryotic NAT proteins have now been determined, and these have been used to generate structural models of eukaryotic NATs, providing valuable insight into their active-site architecture. In addition to aiding crystallographic experiments, recent progress in the production of recombinant prokaryotic and eukaryotic NATs has allowed comparative studies of the kinetics and activity profiles of these enzymes.In this review we present an overview of recent structural and activity studies on NAT enzymes, and we outline how in silico methods may be used to predict NAT protein-ligand interactions based on the current knowledge.