Drug Discovery Today  Volume 15, Numbers 1/2  January 2010 REVIEWS Molecular fields in drug discovery: getting old or reaching maturity? Simon Cross 1 and Gabriele Cruciani 2 1 Molecular Discovery Limited, 215 Marsh Road, Pinner, Middlesex, London HA5 5NE, United Kingdom 2 Laboratory for Chemometrics and Cheminformatics, Chemistry Department, University of Perugia, Via Elce di sotto 10, I-06123 Perugia, Italy With GRID first published 23 years ago, and CoMFA 20 years ago, the two most widely known methods that apply molecular fields to drug discovery are now into their third decade. Are molecular-field-based methods still applicable to modern drug discovery? Are they old and outdated? Or are they maturing into their full potential? Introduction Drug discovery is a complex process that is costly and takes many years [1]. Typically, once a therapeutic target is identified, and appropriate assays developed to enable the in vitro testing of potential drugs, the identification of potential drugs can begin. Millions of compounds are available commercially or in-house, and subsets of these can be assayed to find ‘hits’ that show activity in the assay, using high-throughput screening (HTS). After con- firmation through secondary assays (e.g. dose–response) they are prioritised according to various criteria, including chemical tract- ability, intellectual property, physicochemical properties and potency. Through an iterative follow-up process, analogues of the best ‘leads’ are made with the aim of improving potency, reducing off-target effects, obtaining favourable pharmacokinetic and metabolism profiles and avoiding toxicity. Lead compounds that exhibit favourable pharmacodynamic and pharmacokinetic profiles can then be prioritised according to their efficacy in vivo and one or more ‘drug candidates’ chosen for clinical testing. Many years of clinical testing follow before successful candidates are approved for use. In addition to the experimental in vitro and in vivo approaches, computer simulation (coined as in silico) is now routinely used as a tool to prioritise experiments at each stage of the process [2]. The later that a compound fails in the discovery process, the more costly it has been, hence predicting this failure earlier in the process is highly desirable. Virtual compounds can be filtered according to calculated and predicted physicochemical properties to increase their chances of exhibiting favourable pharmacoki- netic properties [3] and chemical tractability. Virtual screening can be performed with many different methods, to prioritise com- pounds in terms of potency and selectivity so that the more promising compounds are tested [4]. Focused libraries around the most promising hits are then designed; once these analogues have been tested, structure–activity relationships (SAR) may be found in silico and statistical methods used to build quantitative models (QSAR), enabling future virtual analogues to be prioritised before synthesis [5]. Structure-based design methods may be used directly on experimentally determined ligand–protein complexes or apo-structures [6]; docking potential ligands into target struc- tures is one method that can be used for virtual screening, and knowledge of co-crystallised ligands can improve this method [7]. In the absence of experimentally determined structures compara- tive modelling of the target structure may also be possible if suitable homologues are available [8]. Perhaps more simplistically, pure in silico visualisation of a ligand–receptor complex can help in understanding the SAR of the system by identifying accessible binding pockets, especially if they are unique to the target and hence provide an opportunity for improving the selectivity profile. Alongside potency and selectivity, pharmacokinetic properties can be optimised using in silico quantitative structure–property rela- tionship (QSPR) models [9]. More specifically, methods are avail- able to enable the optimisation of metabolic stability, metabolite prediction, and also to predict toxicity [10]. Whilst in silico meth- ods have been developed to support many aspects of the drug discovery process, it is important to stress that they are no sub- stitute for experiment and have limitations; they are a guide to help prioritise the vast number of experiments available and understand the results. Reviews  INFORMATICS Corresponding author: Cross, S. (Simon@moldiscovery.com) 1359-6446/06/$ - see front matter ß 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.drudis.2008.12.006 www.drugdiscoverytoday.com 23