Advanced Computer-Assisted Techniques in Drug Discovery edited by Han van de Waterbeemd Methods and Principles in Medicinal Chemistry Edited by R. Mannhold P. Krogsgaard-Larsen H. Timmerman Volume 1 Hugo Kubinyi, QSAR: Hansch Analysis and Related Approaches Volume 2 Han van de Waterbeemd (ed.), Chemometric Methods in Molecular Design Volume 3 Han van de Waterbeemd (ed.), Advanced Computer- Assisted Techniques in Drug Discovery Methods and Principles in Medicinal Chemistry edited by R. Mannhold, P. Krogsgaard-Larsen, H. Timmerman This practice-oriented series of handbooks and monographs introduces the reader to basic principles and state-of-the-art methods in medicinal chemibtry. Topics treated in-depth include W chemical propertiesofdrugs W characterization of biological activity W advanced techniques in QSAR W physiological and biochemical understanding of diseases Volume I In prepamtion: Kubinyi,H. H D. Holtje, G. Folkers QSAR: Hansch Analysis and Related Approaches Drug Design Molecular Modeling and An Introductory Handbook - Winter l995/6 - 1993. XI], 240 pages with 60 figures and 32 tables. Hardcover. DM 164.00. ISBN 3-527-30035-X (VCH, Weinheim) Volume 2 V. Pliska, B. Testa, H. van de Waterbeemd (eds.) Lipophilicity in Drug Research and Toxicology van de Waterbeemd, H. (ed.) Chemometric Methods in Molecular Design -Winter 1995/6 - 1995. Ca 300 pages. Hardcover. Ca DM 178.00. (VCH, Weinheim) ISBN 3-527-30044-9 VCH 4b Advanced Computer- Assisted Techniques in Drug Discovery edited by Han van de Waterbeemd Weinheim New York Base1 Cambridge Tokyo Volume editor: Dr. Han van de Waterbeemd F. Hoffmann - La Roche Ltd. Pharma Research New Technologies CH-4002 Basel Switzerland Editors: Prof. Raimund Mannhold Biomedical Research Center Molecular Drug Research Group Heinrich- Heine-Universitat UniversitatsstraRe 1 D-40225 Diisseldorf Germany Prof. Povl Krogsgaard-Larsen Dept. of Organic Chemistry Royal Danish School of Pharmacy DK-2100 Copenhagen Denmark Prof. Hendrik Timmerman Faculty of Chemistry Dept. of Pharmacochemistry Free University of Amsterdam De Boelelaan 1083 NL-1081 HV Amsterdam The Netherlands I I This book was carefully produced. Nevertheless, authors, editors and publisher do not warrant the information contained therein to be free of errors. Readers are advised to keep in mind that statements, data, illustrations, procedural details or other items may inadvertently be inaccurate. Published jointly by VCH Verlagsgesellschaft mbH, Weinheim (Federal Republic of Germany) VCH Publishers, Inc., New York, NY (USA) Editorial Director: Dr. Thomas Mager Production Manager: Dip1 Ing. (FH) Hans Jorg Maier Library of Congress Card No. applied for. British Library Cataloguing-in-Publication Data: A catalogue record for this book is available from the British Library. Deutsche Bibliothek Cataloguing-in-Publication Data: Advanced computer assisted techniques in drug discovery I ed. by Han van de Waterbeemd. - Weinheim ; New York ; Basel ; Cambridge ; Tokyo : VCH, 1994 NE: Waterbeemd, Han van de [Hrsg.]; GT 0 VCH Verlagsgesellschaft mbH. D-6945 I Weinheim (Federal Republic of Germany), 1995 Printed on acid-free and chlorine-free paper. All rights reserved (including those of translation in other languages). No part of this book may be reproduced in any form - by photoprinting, microfilm, or any other means - nor transmitted or translated into machine language without written permission from the publishers. Registered na- mes, trademarks, etc. used in this book, even when not specifically marked as such, are not to be considered unprotected by law. Composition:K+V Fotosatz GmbH, D- 64743 Beerfelden. Printing: betz-druck gmbh, D-64291 Darmstadt. Printed in the Federal Republic of Germany. (Methods and principles in medicinal chemistry ; Vol. 3) ISBN 3-527-29248-9 Distribution: VCH, P.O. Box 10 11 61, D-69451 Weinheim (Federal Republic of Germany) Switzerland: VCH, P.O. Box, CH-4020 Basel (Switzerland) United Kingdom and Ireland: VCH (UK) Ltd., 8 Wellington Court, Cambridge CBI 1HZ (England) USA and Canada: VCH, 220 East 23rd Street, New York, NY 10010-4606 (USA) Japan: VCH, Eikow Building, 10-9 Hongo 1-chome, Bunkyo-ku, Tokyo 113 (Japan) Preface The main objective of this series is to offer a practice-oriented survey of techniques currently used in Medicinal Chemistry. Following the volumes on Hansch analysis and related approaches (Vol. 1) and multivariate analyses (Vol. 2), the present hand- book focuses on some new, emerging techniques in drug discovery; emphasis is plat- ed on showing users how to apply these methods and to avoid time-consuming and costly errors. Four major topics are covered. The first centers on three-dimensional QSAR, and some of the enormous progress achieved in this field is summarized. Both the various 3D-QSAR methods available as well as the chemometric tools for handling the statis- tical problems involved in 3D-QSAR studies are covered. Intimately coupled with 3D-QSAR is the current trend in pharmaceutical industry to establish chemical structure databases as a tool for identifying new leads. Cor- respondingly, in the second section, problems encountered in our understanding of molecular similarity and aspects of compound selection by clustering databases are treated. The third section covers advanced statistical techniques in drug discovery. Inter alia the approach of Svante Wold to apply PLS to non-linear structure-activity rela- tions deserves to be mentioned here. Last but not least, the use of neural networks for data analysis in QSAR problems is discussed. Advantages and disadvantages are critically analysed by comparing net- works versus statistics. The editors would like to thank all contributors and VCH publishers for their fruitful cooperation. Summer 1994 Diisseldorf Kopenhagen Amsterdam Raimund Mannhold Povl Krogsgaard-Larsen Hendrik Timmerman A Personal Foreword It is no coincidence that the first three volumes of Methods and Principles in Medicinal Chemistry deal with computer-assisted medicinal chemistry. After the classical Hansch method in Volume 1 and applications of chemometric methods in Volume 2, the present volume of the series contains a number of emerging new tech- niques. Of course, all approaches using molecular modeling techniques, such as structure-based design and de novo design, rely on computers as well. These will be treated separately in a forthcoming volume. This volume is a logical continuation of Volume 2. In fact, after analyzing the methods that have been developed following the Hansch method, we came to the conclusion that a number of these techniques have now matured, while others still require further developments. This criterion was used to select the chapters for Volumes 2 and 3. In reviewing the contents of the first three volumes in this series, it is evident that highly specialized tools have become available for the analysis of complex biological and chemical data sets in order to unravel quantitative structure-activity relation- ships. It has not become easier for the bench chemist to select the ideal method for dealing with the analysis of structure-activity relationships using chemical and bio- logical data. Specialist support is required to validate and apply statistical or chemo- metric and other computer-assisted tools. Volume 3 focusses very much on the newest methods employed by the chemometrician. We hope that, in an indirect way, some of the methods discussed will be of use to molecular design on a day to day basis. I am grateful to, and would like to thank all the contributing authors for their efforts in compiling this volume. February 1994, Base1 Han van de Waterbeemd Contents Preface V A Personal Foreword VI 1 Introduction 1 1.1 3D QSAR 1 H . van de Waierbeemd 1.2 Databases 4 1.3 Progress in Multivariate Data Analysis 4 1.4 Scope of this Book 5 References 6 2 3D QSAR: The Integration of QSAR with Molecular Modeling 9 2.1 Chemometrics and Molecular Modeling 9 D . Piiea. c! Cosenfino. G . Moro. L . Bonaii. E . Fraschini. M . Lasagni and R . Todeschini 2.1 . 1 Introduction 10 2.1.2 QSAR Methodology using Molecular Modeling and Chemometrics 11 2.1.2.1 Search for the Geometric Pharmacophore 13 2.1.2.2 Quantitative Correlation between Molecular Properties and Activity 16 2.1.2.3 Computer Programs 18 2.1.3 Illustrative Examples 18 2.1.3.1 Amnesia-Reversal Compounds 18 2.1.3.2 Non-Peptide Angiotensin I1 Receptor Antagonists 21 2.1.3.3 HMG-CoA Reductase Inhibitors 25 2.1.3.4 Antagonists at the 5-HT3 Receptor 28 2.1.3.5 Polychlorinated Dibenzo-p-dioxins 32 2.1.4 Conclusions 35 References 36 X Contents 2.2 3D QSAR Methods 39 A.M. Davis 2.2.1 Introduction 39 2.2.2 3D QSAR of a Series of Calcium Channel Agonists 41 2.2.2.1 Molecular Alignment 43 2.2.2.2 Charges 45 2.2.2.3 Generating 3D Fields 2.2.2.4 Compilation of GRID Maps 47 2.2.2.5 Inclusion of Macroscopic Descriptors with 3D Field Data 2.2.3 Statistical Analysis 49 2.2.3.1 Results of the Analysis 51 2.2.3.2 Testing the Model 56 2.2.4 Conclusions 57 45 48 References 59 2.3 GOLPE Philosophy and Applications in 3D QSAR 61 G . Cruciani and S . Clementi 2.3.1 Introduction 61 2.3.1.1 3D Molecular Descriptors and Chemometric Tools 2.3.1.2 Unfolding Three-way Matrices 2.3.2 The GOLPE Philosophy 65 2.3.2.1 Variable Selection 68 2.3.3 Applications 70 2.3.3.1 PCA on the Target Matrix 2.3.3.2 PCA on the Probe Matrix 73 2.3.3.3 PLS Analysis on the Target Matrix 76 to Ascertain the Active Conformation 78 2.3.3.5 GOLPE with Different 3D Descriptors 81 2.3.4 Conclusions and Perspectives 82 References 87 63 64 71 2.3.3.4 PLS on Target Matrix as a Strategy 3 Rational Use of Chemical and Sequence Databases 89 3.1 Molecular Similarity Analysis: Applications in Drug Discovery 89 M . A . Johnson. G . M . Muggiora. M . S . Lujiness. J. B . Moon. J D . Petke and D . C . Rohrer 3.1.1 Introduction 89 3.1.2 Similarity-Based Compound Selection 91 3.1.2.1 Similarity Measures and Neighborhoods 91 Contents XI 3.1.2.2 Application of 2D and 3D Similarity Measures 3.1.3 Structure-Activity Maps (SAMs) 96 3.1.3.1 A Visual Analogy 96 3.1.3.2 Representing Inter-Structure Distances 3.1.3.4 Coloring a Structure Map 101 3.1.4 Field-Based Similarity Methods 102 3.1.4.1 Field-Based Similarity Measures 103 3.1.4.2 Field-Based Molecular Superpositions 104 3.1.4.3 An Example of Field-Based Fitting: Morphine and Clonidine 105 3.1.5 Conclusions 108 References 109 94 Screening 95 3.1.2.3 Application of Dissimilarity-Based Compound Selection for Broad 97 3.1.3.3 Structure Maps 99 3.2 Clustering of Chemical Structure Databases for Compound Selection 111 G.M. Downs and I? Willett 3.2.1 Introduction 111 3.2.2 Review of Clustering Methods 114 3.2.2.1 Hierarchical Clustering Methods 115 3.2.2.2 Non-Hierarchical Clustering Methods 118 3.2.3 Choice of Clustering Method 121 3.2.3.1 Computational Requirements 121 3.2.3.2 Cluster Shapes 122 3.2.3.3 Comparative Studies 123 3.2.4 Examples of the Selection of Compounds from Databases by Clustering Techniques 125 3.2.4.1 The Jarvis-Patrick Method 125 3.2.4.2 The Leader Method 126 3.2.5 Conclusions 127 References 128 3.3 Receptor Mapping and Phylogenetic Clustering 131 3.3.1 G-protein Coupled Receptors 132 3.3.3 Principle Coordinates Analysis of 26 Receptor Subtypes 144 3.3.4 Phylogenetic Clustering 148 3.3.5 Discussion 157 References 161 I?J Lewi and H . Moereels 3.3.2 Principal Coordinates Analysis of 71 Receptor Sequences 135 [...]... Fabrocini, F., L Cornput.-Aid.Mol Des 5, 617-628 (1991) Advanced Computer- Assisted Techniques in Drug Discovery edited by Han van d e Waterbeemd Copyright 0 VCH Vedagsgesellschaft mbH, 1995 2 3D QSAR The Integration of QSAR with Molecular Modeling 2.1 Chemometrics and Molecular Modeling Demetrio Pitea, Ugo Cosentino, Giorgio Moro, Laura Bonati, Elena Fraschini, Murina Lasagni, and Roberto Todeschini... Field-Hartree Fock Standard Deviation Tetrachloro-dibenzo-p-dioxins Trichloro-dibenzo-p-dioxins 2.1.1 Introduction The role of computational simulations of molecular mechanisms in understanding biological processes is emphasized in this sentence by Weinstein [I]: “The rapid growth in our mechanistic understanding of biological systems and processes combined with the recently developed technological capabilities... within the active class LDCT consists in the combined use of Linear Discriminant Analysis, LDA, [16, 171 see Volume 2, and tree classification methods [18, 191 The tree structure is composed of nodes and leaves, the starting node being the initial estimate At each step, two new nodes can be generated from each node by binary splitting When a node is no longer split, it becomes a leaf At each splitting... considered as being in its infant years The major problem being that CoMFA models are based on an alignment of compounds in a series, which is far from being a trivial problem [8] Some progress has been made using genetic algorithms [9] and 3D ACC transforms (based on autocorrelation and cross-covariance of field descriptors [ 101 In summary, computers in molecular design are used in the following ways:... 2) In Sec 3 of this book we deal with similarity criteria for rational clustering and searching through chemical databases Furthermore, it is illustrated how clustering techniques can be used to extract information from protein sequence databases As stated above, recent developments in the understanding of certain data analysis problems may have applications in the field of molecular design This involves,... activities observed in a specific class of compounds, finding a hypothesis for the molecular mechanisms and design of new molecular structures with a more specific and enhanced activity One way to achieve these goals can be through the development of computational methodologies based on the combined use of molecular modeling and chemometrics Molecular modeling provides considerable molecular information on... Principles in Medicinal Chemistry, Vol l), VCH, Weinheim, 1993 [4] Tute, M S., History and Objectives of Quantitative Drug Design In: Quantitative Drug Design (Comprehensive Medicinal Chemistry, Vol 4 Hansch, C., Sammes, P G and Taylor, J B., eds Pergamon Press, Oxford, 1990, p 1 - 3 1 [5] Topliss, J.G., Perspect Drug Disc Des 1, 253-268 (1993) [6] Van de Waterbeemd, H., (ed.) Chemometric Methods in. .. each splitting in our procedure, the conformations present in the node are partioned into two groups and classified using LDA, obtaining two new nodes During this process, a node containing conformations of only one class, or a percentage of conformations of a particular class which is greater than a threshold value, is no longer split, i.e it becomes a leaf: the conformations included in the leaf are... rotor model and systematically increasing the conformationally relevant torsional angles to generate the starting geometries Then each of these geometries was fully optimized, relaxing all the geometric degrees of freedom All the stationary 14 D Pitea et al points localized on the Potential Energy Surface (PES) were characterized as minima or saddle points by a second derivative analysis Finally, in the... van de Waterbeemd have led to the unraveling of many details about the structure and function of macromolecules, such as nucleic acids and proteins The second revolution, developed in parallel and which is now indispensable, concerns the use of computers in molecular design and in the lead discovery process The present series “Methods and Principles in Medicinal Chemistry” compiles the progress made . Advanced Computer-Assisted Techniques in Drug Discovery edited by Han van de Waterbeemd Methods and Principles in Medicinal Chemistry Edited by R. Mannhold P. Krogsgaard-Larsen. Hugo Kubinyi, QSAR: Hansch Analysis and Related Approaches Volume 2 Han van de Waterbeemd (ed.), Chemometric Methods in Molecular Design Volume 3 Han van de Waterbeemd (ed.), Advanced. Library. Deutsche Bibliothek Cataloguing -in- Publication Data: Advanced computer assisted techniques in drug discovery I ed. by Han van de Waterbeemd. - Weinheim ; New York ; Basel ;