REVIEWbph_1127 1239..1249Principles of early drugdiscoveryJP Hughes1, S Rees2, SB Kalindjian3and KL Philpott31MedImmune Inc, Granta Park, Cambridge, UK,2GlaxoSmithKline, Gunnels Wood Road,Stevenage, Hertfordshire, UK, and3Kings College, Guys Campus, London, UKCorrespondenceDr Karen Philpott, HodgkinBuilding, Kings College, GuysCampus, London SE1 1UL, UK.E-mail: karen.philpott@kcl.ac.uk----------------------------------------------------------------Keywordsdrug discovery; high throughputscreening; target identication;target validation; hit series; assaydevelopment; screening cascade;lead optimization----------------------------------------------------------------Received2 August 2010Revised7 October 2010Accepted8 November 2010Developing a new drug from original idea to the launch of a nished product is a complex process which can take 1215years and cost in excess of $1 billion. The idea for a target can come from a variety of sources including academic and clinicalresearch and from the commercial sector. It may take many years to build up a body of supporting evidence before selectinga target for a costly drug discovery programme. Once a target has been chosen, the pharmaceutical industry and morerecently some academic centres have streamlined a number of early processes to identify molecules which possess suitablecharacteristics to make acceptable drugs. This review will look at key preclinical stages of the drug discovery process, frominitial target identication and validation, through assay development, high throughput screening, hit identication, leadoptimization and nally the selection of a candidate molecule for clinical development.AbbreviationsADME, absorption, distribution, metabolism and excretion; DMPK, drug metabolism pharmacokinetics; DMSO,dimethyl sulphoxide; GPCRs, G-protein-coupled receptors; HTS, high throughput screening; mAbs, human monoclonalantibodies; PD, pharmacodynamic; PK, pharmacokinetic; SAR, structureactivity relationshipIntroductionAdrug discovery programme initiates because there is adiseaseorclinicalconditionwithoutsuitablemedicalprod-ucts available and it is this unmet clinical need which is theunderlyingdrivingmotivationfor the project. The initialresearch, often occurring in academia, generates data todevelopa hypothesis that the inhibitionor activationofaproteinorpathwaywillresultinatherapeuticeffectinadisease state. The outcome of this activity is the selection of atarget which may require further validation prior to progres-sion into the lead discovery phase in order to justify a drugdiscoveryeffort(Figure 1). Duringleaddiscovery, aninten-sive searchensues to nd a drug-like small molecule orbiologicaltherapeutic,typicallytermedadevelopmentcan-didate, thatwill progressintopreclinical, andifsuccessful,into clinical development (Figure 2) and ultimately be a mar-keted medicine.Target identicationDrugs fail in the clinic for two main reasons; the rst is thatthey do not work and the second is that they are not safe. Assuch,oneofthemostimportantstepsindevelopinganewdrug is target identication and validation. A target is a broadtermwhichcanbeappliedtoarangeofbiologicalentitieswhich may include for example proteins, genes and RNA. Agoodtarget needstobeefcacious, safe, meet clinical andcommercial needs and, aboveall, bedruggable. Adrug-gable targetisaccessibletotheputativedrugmolecule, bethat a small molecule or larger biologicals and upon binding,elicitabiologicalresponsewhichmaybemeasuredbothinvitroandinvivo.Itisnowknownthatcertaintargetclassesare more amenable tosmall molecule drugdiscovery, forexample, G-protein-coupled receptors (GPCRs), whereas anti-bodies are good at blocking protein/proteininteractions.GoodtargetidenticationandvalidationenablesincreasedBJPBritish Journal ofPharmacologyDOI:10.1111/j.1476-5381.2010.01127.xwww.brjpharmacol.orgBritish Journal of Pharmacology (2011) 162 12391249 1239 2011 The AuthorsBritish Journal of Pharmacology 2011 The British Pharmacological Societycondence in the relationship between target and disease andallowsustoexplorewhethertargetmodulationwillleadtomechanism-based side effects.Dataminingof availablebiomedical datahas ledtoasignicantincreaseintargetidentication. Inthiscontext,data mining refers to the use of a bioinformatics approach tonot only help in identifying but also selecting and prioritiz-ing potential disease targets (Yang et al., 2009). The datawhich are available come froma variety of sources butincludepublications andpatent information, geneexpres-sion data, proteomics data, transgenic phenotyping and com-pound proling data. Identication approaches also includeexaminingmRNA/proteinlevelstodeterminewhethertheyare expressed in disease and if they are correlated with diseaseexacerbation or progression. Another powerful approach is tolookfor genetic associations, for example, is there a linkbetweenageneticpolymorphismandtheriskofdiseaseordiseaseprogressionoristhepolymorphismfunctional. Forexample, familial Alzheimers Disease (AD) patients com-monlyhavemutationsintheamyloidprecursorproteinorpresenilin genes which lead to the production and depositionin the brain of increased amounts of the Abeta peptide, char-acteristicof AD(BertramandTanzi, 2008). Therearealsoexamples of phenotypes inhumans wheremutations cannullify or overactivate the receptor, for example, the voltage-gated sodium channel NaV1.7, both mutations incur a painphenotype, insensitivity or oversensitivity respectively (Yanget al., 2004; Cox et al., 2006).An alternative approach is to use phenotypic screening toidentifydiseaserelevant targets. Inanelegant experiment,Kurosawa et al. (2008) used a phage-display antibody libraryto isolate human monoclonal antibodies (mAbs) that bind tothe surface of tumour cells. Clones were individually screenedby immunostaining and those that preferentially and stronglystained the malignant cells were chosen. The antigens recog-nized by those clones were isolated by immunoprecipitationandidentiedby mass spectroscopy. Of 2114 mAbs withunique sequences they identied 21 distinct antigens highlyexpressed on several carcinomas, some of which may be usefultargets for the corresponding carcinoma therapy and severalmAbs which may become therapeutic agents.Target validationOnce identied, the target then needs to be fully prosecuted.Validationtechniquesrangefrominvitrotoolsthroughtheuse of whole animal models, to modulation of a desired targetFigure 1Drug discovery process from target ID and validation through to ling of a compound and the approximate timescale for these processes. FDA,Food and Drug Administration; IND, Investigational New Drug; NDA, New Drug Application.in vitro & ex vivosecondary assays(mechanistic)Selectivity & liabilityassaysHTS & selectivelibrary screens;structure based designReiterative directedcompound synthesis toimprove compoundpropertiesGenetic, cellularand in vivoexperimentalmodels to identifyand validate targetCompoundpharmacologyDisease efficacymodelsEarly safety &toxicity studiesCandidatesecondary assays In vivo analysis Compound screeningTargetvalidationPreclinicalsafety & toxicitypackageEnzyme/Receptor% inhibitionFigure 2Overview of drug discovery screening assays.BJPJP Hughes et al.1240 British Journal of Pharmacology (2011) 162 12391249indiseasepatients.Whileeachapproachisvalidinitsownright, condenceintheobservedoutcomeis signicantlyincreased by a multi-validation approach (Figure 3).Antisense technology is a potentially powerful techniquewhich utilizes RNA-like chemically modied oligonucleotideswhicharedesignedtobecomplimentarytoaregionof atarget mRNA molecule (Henning and Beste, 2002). Binding oftheantisenseoligonucleotidetothetargetmRNApreventsbinding of the translational machinery thereby blocking syn-thesis of the encoded protein. A prime example of the powerofantisensetechnologywasdemonstratedbyresearchersatAbbottLaboratorieswhodevelopedantisenseprobestotherat P2X3 receptor (Honore et al., 2002). When given byintrathecal minipump, toavoidtoxicities associatedwithbolus injection, the phosphorothioate antisense P2X3 oligo-nucleonucleotideshadmarkedanti-hyperalgesicactivityintheCompleteFreundsAdjuvant model, demonstratinganunambiguous role for this receptor in chronic inammatorystates. Interestingly, after administration of the antisense oli-gonucleonucleotides was discontinued, receptor functionand algesic responses returned. Therefore, in contrast to thegene knockout approach, antisense oligonucleotide effectsarereversibleandacontinuedpresenceof theantisenseisrequired for target protein inhibition (Peet, 2003). However,thechemistryassociatedwithcreatingoligonucleotideshasresultedinmoleculeswithlimitedbioavailabilityandpro-nounced toxicity, making their in vivo use problematic. Thishas beencompounded by non-specic actions, problemswithcontrols for these tools anda lackof diversityandvarietyinselectingappropriatenucleotideprobes(Henningand Beste, 2002).Incontrast, transgenicanimalsareanattractivevalida-tiontoolastheyinvolvewholeanimalsandallowobserva-tionof phenotypic endpoints toelucidate the functionalconsequence of gene manipulation. In the early days of genetargetinganimalsweregeneratedthatlackedagivengenesfunctionfrominceptionandthroughout their lives. Thiswork yielded great insights into the in vivo functions of a widerange of genes. One such example is through use of the P2X7knockout mouse to conrm a role for this ion channel in thedevelopmentandmaintenanceof neuropathicandinam-matorypain(Chessell et al., 2005). Inmice lackingP2X7receptors,inammatoryandneuropathichypersensitivityiscompletelyabsenttobothmechanicalandthermalstimuli,while normal nociceptive processing is preserved. Thesetransgenic animals were also used to conrm the mechanismof action for this ablation in vivo as the transgenic mice wereunable to release the mature pro-inammatory cytokineIL-1beta from cells although there was no decit in IL-1betamRNA expression. An alternative to gene knockouts are geneknock-ins, whereanon-enzymaticallyfunctioningproteinreplacestheendogenousprotein.Theseanimalscanhaveadifferent phenotypetoaknockout, for examplewhentheproteinhasstructuralaswellasenzymaticfunctions(Abellet al., 2005) andthesemiceshouldostensiblymimicmorecloselywhathappensduringtreatmentwithdrugs, thatis,the protein is there but functionally inhibited.More recently, the desire to be able to make tissue-restrictedand/orinducibleknockoutshasgrown. Althoughthese approaches are technically challenging, the mostobviousreasonforthisistheneedtoovercomeembryoniclethality of the homozygous null animals. Other reasonsinclude avoidance of compensatory mechanisms due tochronic absence of a gene-encoded function and avoidance ofdevelopmental phenotypes. However, theuseoftransgenicanimals is expensive andtime-consuming. Soinorder tocircumvent some of these issues, the use of small interferingRNA (siRNA) has become increasingly popular for target vali-dation. Double-stranded RNA (dsRNA) specic to the gene tobe silenced is introduced into a cell or organism, where it isrecognizedasexogenousgeneticmaterial andactivatestheRNAi pathway. TheribonucleaseproteinDicer is activatedwhich binds and cleaves dsRNAs to produce double-strandedfragments of 2125 base pairs with a few unpaired overhangDiseaseassociation:genetics andexpression dataOver-expression,transgenics,RNAi, antisenseRNAComparativegeneticsExpressionprofile: taqman,IHC, westernblottingTarget id andvalidationLiterature survey& competitorinformationTool compounds& bioactivemoleculesMolecularpharmacology ofvariantsAnalysis ofmolecularsign allingpath waysInteractions:immunoprecip;yeast 2 hybridCell-based and invivo diseasemodelsFigure 3Target ID and validation is a multifunctional process. IHC, immunohistochemistry.BJPPrinciples of early drug discoveryBritish Journal of Pharmacology (2011) 162 12391249 1241bases on each end. These short double-stranded fragments arecalledsiRNAs. ThesesiRNAsarethenseparatedintosinglestrands and integrated into an active RNA-induced silencingcomplex (RISC). After integration into the RISC, siRNAs base-pair to their target mRNA and induce cleavage of the mRNA,therebypreventingitfrombeingusedasatranslationtem-plate (reviewedinCastanottoandRossi, 2009). However,RNAitechnologystillhasthemajorproblemofdeliverytothe target cell, but many viral and non-viral delivery systemsarecurrentlyunderinvestigation(forreviewseeWhiteheadet al., 2009).Monoclonal antibodies are an excellent target validationtoolastheyinteractwithalargerregionofthetargetmol-ecule surface, allowing for better discriminationbetweeneven closely related targets and often providing higher afn-ity. Incontrast, small molecules aredisadvantagedbytheneed to interact with the often more conserved active site ofatarget,whileantibodiescanbeselectedtobindtouniqueepitopes. This exquisite specicity is the basis for their lack ofnon-mechanistic (or off-target) toxicity a major advantageover small-molecule drugs.However, antibodies cannot cross cell membranes restrict-ingthetargetclassmainlytocellsurfaceandsecretedpro-teins. One impressive example of the efcacy of a mAb in vivois that of the function neutralizing anti-TrkA antibodyMNAC13, which has been shown to reduce both neuropathicpain and inammatory hypersensitivity (Ugolini et al., 2007),therebyimplicatingNGFintheinitiationandmaintenanceofchronicpain. Finally, theclassictargetvalidationtoolisthesmall bioactivemoleculethat interactswithandfunc-tionally modulates effector proteins.More recently, chemical genomics, a systemic applicationof tool molecules to target identication and validation hasemerged. Chemical genomics can be dened as the study ofgenomicresponsestochemicalcompounds.Thegoalistherapid identication of novel drugs and drug targets embrac-ing multiple early phase drug discovery technologies rangingfromtarget identicationandvalidation, over compounddesign and chemical synthesis to biological testing. Chemicalgenomicsbringstogetherdiversity-orientedchemicallibrar-ies and high-information-content cellular assays, along withtheinformaticsandminingtoolsnecessaryforstoringandanalysing the data generated (reviewed inZanders et al.,2002). The ultimate goal of this approach is to provide chemi-cal tools against every protein encoded by the genome. Theaim is to use these tools to evaluate cellular function prior tofull investment in the target and commitment to a screeningcampaignThe hit discovery processFollowing the process of target validation, it is during the hitidentication and lead discovery phase of the drug discoveryprocess that compoundscreeningassays aredeveloped. Ahit moleculecanvaryinmeaningtodifferentresearchersbutinthisinreviewwedeneahitasbeingacompoundwhichhas thedesiredactivityinacompoundscreenandwhose activity is conrmed uponretesting. Avariety ofscreening paradigms exist to identify hit molecules (seeTable 1). High throughput screening (HTS) involves thescreening of the entire compound library directly against thedrugtarget or inamorecomplexassaysystem, suchas acell-based assay, whose activity is dependent upon the targetbut which would then also require secondary assays toconrmthesiteofactionofcompounds(Foxet al., 2006).This screening paradigm involves the use of complex labora-tory automationbut assumes noprior knowledge of thenature of the chemotype likely to have activity at the targetprotein. Focused or knowledge-based screening involvesselectingfromthechemical librarysmallersubsetsof mol-eculesthat arelikelytohaveactivityat thetarget proteinbasedonknowledgeof thetargetproteinandliteratureorpatent precedents for the chemical classes likely to have activ-ityat thedrugtarget (Boppanaet al., 2009). This typeofknowledgehasgivenrise, morerecently, toearlydiscoveryparadigmsusingpharmacophoresandmolecularmodellingto conduct virtual screens of compound databases (McInnes,2007). Fragment screeninginvolvesthegenerationof verysmall molecular weight compound libraries which arescreened at high concentrations and is typically accompaniedby the generation of protein structures to enable compoundprogression(Lawet al., 2009). Finally, a more specializedfocused screening approach can also be taken, physiologicalscreening. This is atissue-basedapproachandlooks for aresponse more aligned with the nal desired in vivo effect asopposed to targeting one specic molecular component.High throughput and other compound screens are devel-opedandruntoidentifymoleculesthat interact withthedrug target, chemistry programmes are run to improvethe potency, selectivity and physiochemical properties of themolecule, and data continue to be developed to support thehypothesis that interventionat the drugtarget will haveefcacy in the disease state. It is this series of activities that arethe subject of intense activity withinthe pharmaceuticalindustry and increasingly within academia to identify candi-date molecules for clinical development. Pharmaceuticalcompanies have built large organizations with the objectiveof identifying targets, assembling compound collections andtheassociatedinfrastructuretoscreenthosecompoundstoidentifyinitiallyhitmoleculesfromHTSorotherscreeningparadigms and to optimize those screening hits into clinicalcandidates.Inrecentyearstheacademicsectorhasbecomeincreasingly interested in the activities traditionally per-formed within the lead discovery phase in the pharmaceuti-cal industry. Academicscientistsarenowformattingassaysfor drug discovery whichare passedontoacademic drugdiscovery centres for compound screening. These centres, asexemplied by the NIH Roadmap initiative in the USA (Frear-son and Collie, 2009), have established compound libraries,screeninginfrastructureandtheappropriateexpertisetradi-tionallyfoundwithintheindustrial sectortoscreentargetproteins to identify so-called chemical probes for use in targetvalidationanddiseasebiologystudies andincreasinglytoidentify chemical start points for drug discovery programmes.The success of these efforts has been facilitated by the transferof skills between the industrial and academic sectors.A typical programme critical path within the lead discov-ery phase consists of a number of activities and begins withthe development of biological assays to be used for the iden-tication of molecules with activity at the drug target. Oncedeveloped, such assays are used to screen compound librariesBJPJP Hughes et al.1242 British Journal of Pharmacology (2011) 162 12391249to identify molecules of interest. The output of a compoundscreenis typicallytermedahit molecule, whichhas beendemonstratedtohavespecicactivityatthetargetprotein.Screening hits form the basis of a lead optimization chemistryprogramme to increase potency of the chemical series at theprimary drug target protein. During the lead discovery, phasemolecules are also screened in cell-based assays predictive ofthe disease state and in animal models of disease to charac-terize both the efcacy of the compound and its likely safetyprole (Figure 2). The following paragraphs describe in moredetail the requirements and application of compound screen-ing assays within hit and lead discovery.Assay developmentInthe recombinant era the majority of assays inuse withintheindustry rely upon the creation of stable mammalian cell linesover-expressing the target of interest or upon the over-expression and purication of recombinant protein to estab-lishso-calledbiochemical assays althoughinrecent yearsthere has been an increase in the number of reports describingthe use of primary cell systems for compoundscreening(Dunneet al., 2009). Generally, cell-basedassayshavebeenappliedtotarget classes suchas membranereceptors, ionchannels and nuclear receptors and typically generate a func-tional read-out as a consequence of compound activity (Mich-elini et al., 2010). In contrast, biochemical assays, which havebeenappliedtobothreceptor andenzyme targets, oftensimply measure the afnity of the test compound for the targetprotein. Therelativemerits of biochemical andcell-basedassays have been debated extensively and have been reviewedelsewhere (Moore and Rees, 2001). Both assay paradigms havebeenusedsuccessfully to identify hit andcandidate molecules.A plethora of assay formats have been enabled to supportcompoundscreening. Thechoiceofassayformatisdepen-dent upon the biology of the drug target protein, the equip-ment infrastructure in the host laboratory, the experience ofthe scientists inthat laboratory, whether aninhibitor oractivator molecule is sought and the scale of the compoundTable 1Screening strategiesScreen Description CommentsHigh throughput Large numbers of compounds analysed in a assaygenerally designed to run in plates of 384 wellsand aboveLarge compound collections often run by big pharma butsmaller compound banks can also be run in eitherpharma or academia which can help reduce costs.Companies also now trying to provide coverage across awide chemical space using computer assisted analysis toreduce the numbers of compounds screened.Focused screen Compounds previously identied as hittingspecic classes of targets (e.g. kinases) andcompounds with similar structuresCan provide a cheaper avenue to nding a hit molecule butcompletely novel structures may not be discovered andthere may be difculties obtaining a patent position in awell-covered IP areaFragment screen Soak small compounds into crystals to obtaincompounds with low mM activity which canthen be used as building blocks for largermoleculesCan join selected fragments together to t into thechemical space to increase potency. Requires a crystalstructure to be availableStructural aided drugdesignUse of crystal structures to help design molecules Often used as an adjunct to other screening strategieswithin big pharma. In this case usually have docked acompound into the crystal and use this to help predictwhere modications could be added to provide increasedpotency or selectivityVirtual screen Docking models: interogation of a virtualcompound library with the X-ray structure ofthe protein or, if have a known ligand, as abase to develop further compounds onCan provide the starting structures for a focused screenwithout the need to use expensive large library screens.Can also be used to look for novel patent space aroundexisting compound structuresPhysiological screen A tissue-based approach fordetermination of the effects of a drug at thetissue rather than the cellular or subcellularlevel, for example, muscle contractilityBespoke screens of lower throughput. Aim to more closelymimic the complexity of tissue rather than just looking atsingle readouts. May appeal to academic experts indisease area to screen smaller number of compounds togive a more disease relevant readoutNMR screen Screen small compounds (fragments) by soakinginto protein targets of known crystal or NMRstructure to look for hits with low mM activitywhich can then be used as building blocks forlarger moleculesUse of NMR as a structure determining toolNMR, nuclear magnetic resonance.BJPPrinciples of early drug discoveryBritish Journal of Pharmacology (2011) 162 12391249 1243screen. For examplecompoundscreeningassays at GPCRshavebeenconguredtomeasurethebindingafnityof aradio- or uorescently labelledligandtothe receptor, tomeasure guanine nucleotide exchange at the level of theG-protein, tomeasurecompound-mediatedchangesinoneof a number of second messenger metabolites includingcalcium, cAMPor inositiol phosphates or tomeasure theactivation of downstream reporter genes. Whatever the assayformat that is selected, it is a requirement that the followingfactors are considered:1. Pharmacological relevance of the assay. If available, studiesshould be performed using known ligands with activity atthe target under study, to determine if the assay pharma-cology is predictive of the disease state and to show thattheassayiscapableof identifyingcompoundswiththedesired potency and mechanism of action.2. Reproducibilityoftheassay.Withinacompoundscreen-ing environment it is a requirement that the assay is repro-ducible across assay plates, across screen days and, withinaprogrammethat mayrunforseveral years, acrosstheduration of the entire drug discovery programme.3. Assay costs. Compound screening assays are typically per-formedinmicrotitreplates.Withinacademiaorforrela-tivelysmall numbersof compoundsassaysaretypicallyformatted in 96-well or 384-well microtitre plates whereasin industry or in HTS applications assays are formatted in384-well or 1536-well microtire plates in assay volumes assmall as a few microlitires. In each case assay reagents andassayvolumesareselectedtominimizethecostsof theassay.4. Assay quality. Assay quality is typically determined accord-ing to the Z factor (Zhang et al., 1999). This is a statisticalparameter that in addition to considering the signalwindowintheassayalsoconsidersthevariancearoundboth the high and low signals in the assay. The Z factor hasbecometheindustrystandardmeansofmeasuringassayquality on a plate bases. The Z factor has a range of 0 to 1;an assay with a Z factor of greater than 0.4 is consideredappropriatelyrobust for compoundscreeningalthoughmany groups prefer to work with assays with a Z factor ofgreater than 0.6. In addition to the Z factor assay quality isalso monitored through the inclusion of pharmacologicalcontrols within each assay. Assays are deemed acceptable ifthe pharmacology of the standard compound(s) fallswithin predened limits. Assay quality is affected by manyfactors. Generally, high-quality assays are created throughimplementing simple assay protocols with few steps, mini-mizing wash steps or plate to plate reagent transfers withinthe assay, through the use of stable reagents and biologi-cals, andthroughensuringthat all theinstrumentationused to perform the assay is performing optimally. This istypically achieved through developing quality controlpractices for all items of laboratory automation (see http://www.ncgc.nih.gov/guidance/section2.html#replicate-experiment-study-summary-acceptance).5. Effectsofcompoundsintheassay.Chemicallibrariesaretypicallystoredinorganic solvents suchas ethanol ordimethyl sulphoxide (DMSO). Thus, assays needtobecongured that are not sensitive to the concentrations ofsolvents used in the assay. Typically, cell-based assays areintolerant tosolvent concentrationsof greater than1%DMSOwhereasbiochemical assayscanbeperformedinsolvent concentrations of up to 10%DMSO. Studies are alsoperformed to establish the false negative and false positivehit rates inthe assay. If these are unacceptably highthentheassay will need to be recongured. Finally some consider-ationshouldbe made tothe screening concentration. Com-pound screening assays for hit discovery are typically runat110 mMcompoundconcentration. At theseconcentra-tionscompoundswithactivitiesof upto40 mMcanbeidentied. The test concentration can be varied to identifycompounds with higher or lower activity.One example of an HTS technology implemented for theidentication of hit molecules with activity at GPCRs is theaequorinassay(Stableset al., 2000). Aequorinisacalcium-sensitivebioluminescent proteinclonedfromthejellyshAequoreavictorea. Stablemammaliancell lines havebeencreated transfected to express the GPCR drug target and theaequorinbiosensoer. For receptors capable of couplingtoheterotrimeric G-proteins of the Gaq/11 family, ligand activa-tion results in an increase in intracellular calcium concentra-tion. Whenaequorinis expressedinthe same cells, thisincrease in intracellular calciumconcentration is detected as aconsequenceofcalciumbindingtotheaequorinphotopro-tein, whichinthepresenceof thecofactorcoelenterazine,results in the generation of a ash of light that can be detectedwithina microtitre plate-basedluminometer suchas the Lumi-luxplatform(PerkinElmer, Waltham, MA, USA). The aequo-rinassay has a very simple protocol andhas beendevelopedforHTS in 1536-well plate format in assay volumes of 6 mL and forcompound proling activities in 384-well plate format.When developing any HTS assay, which can involve thescreening of several million molecules over several weeks, it isbestpracticetoscreentrainingsetsofcompoundstoverifythattheassayisperformingacceptably. Figure 4showsthescreening of a 12 000compoundtraining set against thehistamineH1receptorexpressedinChinesehamsterovarycells in a 1536-well format HTS assay. The training set is typi-cally run on two or three occasions to identify the hit rate inthe assay, the reproducibility of the assay and the false positiveand false negative hit rates in the assay. Typically, statisticalpackageshavebeendevelopedtoidentifytheseparameters.When screened to detect agonist ligands the hit rates in theaequorinassayaretypicallyless than0.5%of compoundsscreenedwithastatisticalassaycut-offof5%orlessoftheagonist signal seenwitha standard agonist ligand. Inthis assayformat false positive and false negative hit rates are very low.For antagonist screening the hit rate in the aequorin assay istypicallyof 23%of compoundsscreenedwithanactivitycut-off of greater than25%inhibition. This is acommonphenomenon of all screening assays. Hit rates in antagonist orinhibitorformattendtobehigherthanhitratesinagonistassays as antagonist assays, which are dened by detection ofadecreaseinassaysignal, will alsodetectcompoundsthatinterfere insignal generation. Following completionof robust-ness testing an assay moves into HTS. During HTS, up to 200assay plates are screened each day, often using complex labo-ratoryautomation.Duringthescreen,assayperformanceismeasuredaccordingtotheZ ontheassayplateandthevariance in the pharmacology of a standard compound, withBJPJP Hughes et al.1244 British Journal of Pharmacology (2011) 162 12391249assay plates being failed and rescreened if these quality controlmeasures fall outside predened limits (Figure 5).Dening a hit seriesCompoundlibrarieshavebeenassembledtocontainsmallmolecularweightmoleculesthatobeychemicalparameterssuchastheLipinskiRuleofFive(Lipinskiet al.,2001),andmore oftenhave molecular weights of less than400andclogP(ameasureof lipophilicitywhichaffects absorptionintothebody)oflessthan4.Moleculeswiththesefeatureshave been termed drug-like, in recognition of the fact thatthemajorityofclinicallymarketeddrugshaveamolecularweight of less than350andacLogPof less than3. It iscriticallyimportanttoinitiateadrugdiscoveryprogrammewith a small simple molecule as lead optimization, toimprove potency and selectivity, typically involves anincrease in molecular weight which in turn can lead to safetyandtolerabilityissues.Figure 4Aequorin high throughput screening: validation testing GPCR antagonist assay (1536-well). Assay validation of a GPCR drug screening assay forthe identication of agonist and antagonist ligands. Cells expressing the histamine H1 receptor and the calcium-sensitive photoprotein aequorinwere dispensed into 1536-well microtitre plates. A total of 12 000 compounds were screened in duplicate to detect agonist ligands (left panel)andantagonist ligands(right panel). Intheagonist assay(left panel), nodrugresponseisrepresentedinred, theresponsetoamaximalconcentration of the ligand histamine in blue and compound data in yellow. As is typically seen in agonist assays, the hit rate is very low due tothe absence of false positives. In the antagonist assay (right panel), the response to histamine in the absence of test compound is represented inred (basal response), the response to a maximal concentration of a histamine antagonist in blue (100% inhibition) and compound data in yellow.As is typically seen in a cell-based inhibitor assay, there is signicant spread of the compound data due to a combination of assay interference andcompound activity. True actives correlate in the range 40% to 100% inhibition. Both assays have excellent Z. GPCR, G-protein-coupled receptor.Figure 5Quality control (QC) in high throughput screening. To ensure the control of screening data in compound screening campaigns each assay platetypically contains a number of pharmacological control compounds. (A) Each 384-well plate contains 16 wells containing a low control and afurther 16 wells containing an EC100 concentration of a pharmacological standard which are used to calculate the Z factor (reference Zhanget al., 1999). PlatesthatgenerateaZ factorbelow0.4arerescreened. (B)Eachplatealsocontains16wellsof anEC50concentrationof apharmacological standard to monitor the variance in the assay (diamonds). (C) A heat map is generated for all plates that pass the pharmacologicalstandard QC to monitor the distribution of activity across the assay plate. One would expect to see a random distribution of activity across thescreening plate. A plate such as the one presented would be failed and rescreened due to the active wells clustering in the centre of the plate.BJPPrinciples of early drug discoveryBritish Journal of Pharmacology (2011) 162 12391249 1245Onceanumberofhitshavebeenobtainedfromvirtualscreening or HTS, the rst role for the drug discovery team istotrytodenewhichcompoundsarethebesttoworkon.Thistriagingprocessisessential as, fromalargelibrary, ateamwilllikelybeleftwithmanypossiblehitswhichtheywill need to reduce, conrm and cluster into series. There areseveral steps to achieving this. First, although this is less of aproblemas the quality of libraries have improved, com-poundsthatareknownbythelibrarycuratorstobetobefrequent hitters in HTS campaigns need to be removed fromfurther consideration. Second, anumber of computationalchemistry algorithms have beendevelopedto grouphitsbased on structural similarity to ensure that a broad spectrumof chemical classes are represented on the list of compoundstaken forward. Analysis of the compound hit list using thesealgorithms allows the selection of hits for progression basedon chemical cluster, potency and factors such as ligand ef-ciency which gives an idea of how well a compound binds forits size (log potency divided by number of heavy atoms i.e.non-hydrogen atoms, in a molecule).Thenext phaseintheinitial renement process is togenerate doseresponse curves in the primary assay for eachhit, preferably with a fresh sample of the compound.Showing normal competitive behaviour in hits is important.Compoundswhichgiveanallornothingresponsearenotacting in a reversible manner and indeed may not be bindingto the target protein at all, with the activity at high concen-trations arising from an interaction between the sample andanother component of the assay system. Reversible com-pounds are favoured because their effects can be more easilywashed-outfollowingdrugwithdrawal,animportantcon-sideration when using in patients. Obtaining a doseresponsecurve allows the generationof ahalf maximal inhibitoryconcentration which is used to compare of the potencies ofcandidatecompounds.Sourcingandusingfreshsamplesofcompounds for this exerciseis highlydesirable. NearlyallHTSlibrariesarestoredasfrozenDMSOsolutionswiththeresult that, after some time, the compound canbecomedegraded or modied. Virtually anyone who has worked withlibraries of this typehas got anecdotes about howpotentactivity has disappeared when the compound was resynthe-sizedandusedinre-testing, althoughoccasionallyidenti-cation of potent impurities has allowed progress to be made.With reliable doseresponse curves generated in theprimaryassayforthetarget,thestageissettoexaminethesurvivinghitsinasecondaryassay, if oneisavailable, forthetarget of choice. This neednot beanassayinahighthroughputformatbutwillinvolvelookingattheaffectofthecompoundsinafunctional response, forexampleinasecond messenger assay or in a tissue-or cell-based bioassay.Activity in this setting will give reassurance that compoundsare able to modulate more intact systems rather than simplyinteractingwiththeisolatedandoftenengineeredproteinused inthe primary assay. Throughout the conrmationprocess, medicinal chemists would be looking to cluster com-pounds into groups which could form the basis of lead series.As part of this process, considerationwill begiventotheproperties of each cluster such as whether there is an identi-ablestructureactivityrelationship(SAR) evolvingover anumberofcompounds, thatis, identicationofagroupofcompoundswhichhavesomesectionorchemicalmotifincommonandtheadditionof different chemical groupstothis core structure results indifferent potencies. Issues ofchemical synthesis wouldalsobeexamined. Thus, easeofpreparation, potentialamenabilitytoparallelsynthesisandthe ability to generate diversity from late-stage intermediateswould be assessed.Withdenedclustersinplaceanexercisecannowtakeplace on several groups of compounds in parallel. This phasewill includetherapidgenerationof rudimentarySARdataanddeningtheessential elementsinthestructureassoci-ated with activity. At the same time, representative examplesof each of these mini-series will be subjected to various in vitroassays designed to provide important information withregard to absorption, distribution, metabolism and excretion(ADME)propertiesaswellasphysicochemicalandpharma-cokinetic (PK) measurements (see Table 2). Selectivity prol-ing,especiallyagainstthetypesoftargets,ifany,forwhichthe compounds were originally made, is also useful to carryout at this time. For example you may want to inhibit kinaseX but avoid kinase Y to reduce unwanted in vivo side effects.This exercise will reveal the strengths and aws of each seriesandallowadecisiontobetakenaboutthemostpromisingseries of compounds to be progressed. The numbers of seriestaken forward at this stage will depend on the resource avail-ablebutideallyseveralshouldbetakenintothehit-to-leadstage to allow for attrition in the coming phase.Whateverthescreeningparadigm,theoutputofthehitdiscovery phase of a lead identication programme is aso-called hit molecule, typically with a potency of 100 nM5 mM at the drug target. A chemistry programme is initiatedto improve the potency of this molecule.Hit-to-lead phaseThe aim of this stage of the work is to rene each hit series totry to produce more potent and selective compounds whichpossessPKpropertiesadequatetoexaminetheirefcacyinany in vivo models that are available.Typically,theworknowconsistsofintensiveSARinves-tigationsaroundeachcorecompoundstructure, withmea-surements being made to establish the magnitude of activityandselectivityofeachcompound. Thisneedstobecarriedoutsystematicallyand, wherestructural informationaboutthetargetisknown,structure-baseddrugdesigntechniquesusing molecular modelling and methodologies such as X-raycrystallography and NMR can be applied to develop the SARfasterandinamorefocusedway. Thistypeofactivitywillalso often give rise to the discovery of new binding sites onthe target proteins.Ascreeningcascadeatthistimewouldgenerallyconsistof a relatively high throughput assay establishing the activityof eachmolecule onthe molecular target, together withassays in the same format for sites where selectivity might beknown,orexpectedtobe,anissue(Figure 6).Acompoundmeeting basic criteria at this stage would be escalated into afurther bankof assays. Theseshouldincludehigher orderfunctional investigations against themolecular target andalso whether the compounds were active in primary assays indifferentspecies. TheHTSassayisgenerallycarriedoutonproteinencodedbyhumanDNAsequences but as animalBJPJP Hughes et al.1246 British Journal of Pharmacology (2011) 162 12391249models are used to validate the activity of compounds in invivo disease models, in pharmacodynamic (PD)/PK modellingand in preclinical toxicity studies, it is important to have dataonactivityinvitroonorthologues. Thisisalsoparticularlyimportant as it will assist inminimizingdosinglevels intoxicology studies which are chosen on the basis multiples ofthe pharmacologically effective doses.Attention in this phase has to also turn to more detailedprolingof physicochemical andinvitroADMEpropertiesandthisseriesofstudiesiscarriedoutinparallel,withkeycompounds being selected for assessment. The sort of assaystobeconsidered, withtargetsthat havebeenfoundtobeappropriate are shown in Table 2.Solubility and permeability assessments are crucial inrulinginoroutthepotentialofacompoundtobeadrug,thatis,drugsubstanceoftenneedsaccesstoapatientscir-culation and therefore may be injected or more generally hasto be adsorbed in the digestive system. Deciency in one orotherparameterinamoleculecansometimesbeputright.Forexampleformulationstrategiescanbeusedtodesignatablet such that it dissolves in a particular region of the gut ata pH in which the compound is more soluble. A compoundthat lacks both these properties is very unlikely to become adrugnomatter howpotent it is intheprimaryscreeningassay. Microsomal stability is a useful measure of the ability ofinvivometabolizingenzymestomodifyandthenremoveacompound. Hepatocytesaresometimesusedinthissortofstudyinsteadandthesewillgivemoreextensiveresultsbutare not used routinely as they need to be prepared freshly onaregular basis. CYP450inhibitionis examinedas, amongotherthings,itisanimportantpredictorofwhetheranewcompound might have an inuence on the metabolism of anexisting drug with which it may be co-administered.If one or more of these properties is less than ideal, thenit might be necessary to screen many more compounds spe-cicallyfor thoseproperties. Eachprogrammewill endupsubtlydifferent inthis regard. For exampleinonerecentproject toidentifynovel GPCRantagonists, a number ofsub-micromolar hit compounds wereidentied. Themainissues associated with these molecules was that they showedsome speciationwithpoorer receptor afnities inrodentTable 2Key in vitro assays in early drug discoveryAssays Targetvalue CommentsAqueous solubility >100 mM Important for running in vitro assays and for in vivo delivery of drugLog D7.403 (for BBB penetration ca 2) A measure of lipophilicity hence movement across membranesMicrosomal stability Clint10 mM Main enzymes in body which metabolize drugs and their inhibitioncan cause toxicityCaco-2 permeability Papp>1 10-6cm-1(asymmetry 10 10-6cm-1(asymmetry 50% inhibition at10 mMat 30out of 63GPCRsandtransporterstestedinacross-screening panel as well as broadCYP450inhibitoryactivity.Itwasfeltthatanumberofthesedeciencieswereassociatedwiththenatureof thebasecommontoall theinitial structures. Modication of the basic residue resulted ina number of compounds which were as potent as the initialhits at the principal receptor but which were more selective intheir actions. In common with many programmes, aspotency at the principal target improved selectivity issues inthis series were left behind.Keycompoundswhicharebeginningtomeetthetargetpotency and selectivity, as well as most of the physicochemi-cal and ADME targets, should be assessed for PK in rats. Hereone would normally be aiming for a half-life of >60 min whenthe compound is administered intravenously and a fraction inexcess of 20% absorbed following oral dosing although some-times, differenttargetsrequireverydifferentPKproles. Inlarge pharma with inhouse drug metabolism pharmacokinet-ics (DMPK) departments numerous compounds might be pro-led while in academic environments there may be funds foronlyapredenednumberoftheseexpensiveinvestigationsAs the receptor antagonist programme, described above,advancedthroughthehit-to-leadphase, anumberofcom-pounds were prepared which had potency in the nanomolarrange and a benign selectivity prole except for some potencyat the hERG channel, a potassium voltage-gated ion channelimportantforcardiacfunctionandinhibitionatwhichcancause cardiac liability. Ideally for hERG we were aiming for anactivity over 30 uM or at least a 1000-fold selectivity for thetarget. A number of these compounds were examined in PKstudies and were found to have a reasonable half-life follow-ingintravenousdosingbut poor plasmalevelswerenotedwhen the compound was given orally to rats. It was felt thatsome of these compounds, representing the end of the hit-to-lead phase of the project were, although not likely themselvestobeprogressed, capableofansweringquestionsindiseasemodels. Thus, compounds were administered intra-peritoneally and results from the experiments gave substan-tial credence to the developing programme.Lead optimization phaseTheobjectofthisnaldrugdiscoveryphaseistomaintainfavourable properties in lead compounds while improving ondecienciesintheleadstructure.Continuingwithexampleabove, theaimof theprogrammewasnowtomodifythestructure tominimize hERGliabilityandtoimprove theabsorptionofthecompound. Thus, moreregularchecksofhERG afnity and CACO2 permeation were undertaken andcompounds were soon available which maintained theirpotency and selectivity at the principal target but which hada much reduced hERG afnity and a better apparent perme-ationthaninitialleadcompounds. WhenexaminedforPKproperties in rat one of these compounds, with 8 nM afnityat the receptor of interest, had an oral bioavailability of over40% in rats and about 80% in dogs.Compounds at this stage may be deemed to have met theinitial goals of the lead optimization phase and are ready fornal characterization before being declared as preclinical can-didates. Discovery work does not cease at this stage. The teamhas to continue to explore synthetically in order to producepotential back up molecules, in case the compound undergo-ingfurtherpreclinical orclinical characterizationfailsand,more strategically, to look for follow-up series.Thestageatwhichthevariouselementsthatconstitutefurther characterization are carried out will vary fromcompany to company and parts of this process may be incor-porated into the lead optimization phase. However, ingeneralmoleculesneedtobeexaminedinmodelsofgeno-toxicity such as the Ames test and in in vivo models of generalbehaviour such as the Irwins test. High-dose pharmacology,PK/PD studies, dose linearity and repeat dosing PK looking fordrug-induced metabolism and metabolic proling all need tobecarriedoutbytheendof thisstage. Considerationalsoneedstobegiventochemicalstabilityissuesandsaltselec-tion for the putative drug substance.Alltheinformationgatheredaboutthemoleculeatthisstage will allowfor the preparationof a target candidateprole which with together with toxicological and chemicalmanufacture and control considerations will form the basis ofaregulatorysubmissiontoallowhumanadministrationtobegin.The process of hit generationtopreclinical candidateselection often takes a long time and cannot in any way beconsidered a routine activity. There are rarely any short cutsand signicant, intellectual input is required from scientistsfrom a variety of disciplines and backgrounds. The quality ofthehit-to-leadstartingpointandtheexpertiseoftheavail-able team are the key determinants of a successful outcome ofthis phase of work. Typically, within industry for each project200 000 to >106compounds might be screened initially andduringthefollowinghit-to-leadandleadoptimizationpro-grammes 100s of compounds are screened to hone down toone or two candidate molecules, usually fromdifferentchemical series. In academia screens are more likely to be ofa focused nature due to the high cost of an extensive HTS orcompounds are derived froma structure-based approach.Only 10% of small molecule projects within industry mightmakethetransitiontocandidate,failingatmultiplestages.These caninclude the (i) inabilitytocongure a reliableassay; (ii) nodevelopablehitsobtainedfromtheHTS; (iii)compounds do not behave as desired in secondary or nativetissue assays; (iv) compounds are toxic in vitro or in vivo; (v)compoundshaveundesirablesideeffectswhichcannot beeasily screened out or separated from the mode of action ofthe target; (vi) inability to obtain a good PK or PD prole inline with the dosing regeme required in man, for example, ifrequire a once a day tablet then need the compound to havea half-life in vivo suitable to achieve this; and (vii) inability tocross the blood brain barrier for compounds whose target lieswithinthe central nervous system. The attritionrate forproteintherapeutics, oncethetargethasbeenidentied, ismuch lower due to less off target selectivity and prior expe-rience of PK of some proteins, for example, antibodies.Although relatively less costly than many processescarriedout later oninthedrugdevelopment andclinicalphases, preclinical activity is sufciently high risk and remotefromnancial returntooftenmakefundingit aproblem.Ensuring transparency of the cost of each stage/assay withinlarge pharma may help reduce some of their costs and thereBJPJP Hughes et al.1248 British Journal of Pharmacology (2011) 162 12391249aresomemovementstowardsthisascompaniesinstigateabiotech mentality and accountability for costs.Onceacandidateisselected, theattritionrateof com-poundsenteringtheclinical phaseisalsohigh, againonlyone in 10 candidates reaching the market but at this stage thenancial consequences of failure are much higher. There hasbeenconsiderabledebateinindustryastohowtoimprovethe success rate, by failing fast and cheap. Once a candidatereaches the clinical stage, it can become increasingly difculttokill theproject, asat thisstagetheproject hasbecomepublic knowledge and thus termination can inuence con-denceinthecompanyandshareholdervalue.Carryingoutmore studies prior to clinical development such as improvedtoxicology screens (using failed drugs to inform these assays),establishing predictive translational models based on a thor-ough disease understanding and identifying biomarkers mayhelpinthisendeavour. Itisparticularlyintheselatertwoareas where academic-industry partnerships could really addvaluepreclinicallyandeventuallyhelpbringmoreeffectivedrugs to patients.AcknowledgementsKaren Philpott is supported by the Medical Research Counciland Guys and St Thomas Charity.S. Barret Kalindjian is supported by a Seeding Drug DiscoveryWellcome Trust grant.Conict of interestJane Hughes is employed by MedImmune, Steve Rees isemployed by GSK and Karen Philpott was previouslyemployed by GSK.ReferencesAbell AN, Rivera-Perez JA, Cuevas BD, Uhlik MT, Sather S,Johnson NL et al. (2005). Ablation of MEKK4 kinase activity causesneurulation and skeletal patterning defects in the mouse embryo.Mol Cell Biol 25: 89488959.Bertram L, Tanzi RE (2008). Thirty years of Alzheimers diseasegenetics: the implications of systematic meta-analyses. Nat RevNeurosci 9: 768778.Boppana K, Dubey PK, Jagarlapudi SARP, Vadivelan S, Rambabu G(2009). Knowledge based identication of MAO-B selectiveinhibitors using pharmacophore and structure based virtualscreening models. Eur J Med Chem 44: 35843590.Castanotto D, Rossi JJ (2009). The promises and pitfalls ofRNA-interference-based therapeutics. Nature 457: 426433.Chessell IP, Hatcher JP, Bountra C, Michel AD, Hughes JP, Green Pet al. (2005). Disruption of the P2X7 purinoceptor gene abolisheschronic inammatory and neuropathic pain. Pain 114: 386396.Cox JJ, Reimann F, Nicholas AK (2006). An SCN9A channelopathycauses congenital inability to experience pain. Nature 444:894898.Dunne A, Jowett M, Rees S (2009). Use of primary cells in highthroughput screens. Meth Mol Biol 565: 239257.Fox S, Farr-Jones S, Sopchak L, Boggs A, Nicely AW, Khoury R et al.(2006). High-throughput screening; Update on practices andsuccess. J Biol Screen 11: 864869.Frearson JA, Collie IT (2009). HTS and hit nding in academia from chemical genomics to drug discovery. Drug Discov Today 14:11501158.Henning SW, Beste G (2002). Loss-of-function strategies in drugtarget validation. Curr Drug Discov May: 1721.Honore P, Kage K, Mikusa J, Watt AT, Johnston JF, Wyatt JR et al.(2002). Analgesic prole of intrathecal P2X3 antisenseoligonucleotide treatment in chronic inammatory andneuropathic pain states. Pain 99: 1119.Kurosawa G, Akahori Y, Morita M, Sumitomo M, Sato N,Muramatsu C et al. (2008). Comprehensive screening for antigensoverexpressed on carcinomas via isolation of human mAbs thatmay be therapeutic. Proc Natl Acad Sci U S A 105: 72877292.Law R, Barker O, Barker JJ, Hesterkamp T, Godemann R,Andersen O et al. (2009). The multiple roles of computationalchemistry in fragment-based drug design. J Comput Aided Mol Des23: 459473.Lipinski CA, Lombardo F, Dominy BW, Feeney PJ (2001).Experimental and computational approaches to estimate solubilityand permeability in drug discovery and development settings. AdvDrug Deliv Rev 46: 326.McInnes C (2007). Virtual screening strategies in drug discovery.Curr Opin Chem Biol 11: 494502.Michelini E, Cevenini L, Mezzanotte L, Coppa A, Roda A (2010).Cell Based Assays: fuelling drug discovery. Anal Biochem 397: 110.Moore K, Rees S (2001). Cell-based versus isolated target screening:how lucky do you feel? J Biomol Scr 6: 6974.Peet NP (2003). What constitutes target validation? Targets 2:125127.Stables J, Mattheakis LC, Chang TR, Rees S (2000). Recombinantaequorin as a reporter of changes in intracellular calciumconcentration in mammalian n cells. Meth Enzymol 327: 456471.Ugolini G, Marinelli S, Covaceuszach S, Cattaneo A, Pavone F(2007). The function neutralizing anti-TrkA antibody MNAC13reduces inammatory and neuropathic pain. Proc Natl Acad Sci U SA 104: 29852990.Whitehead KA, Langer R, Anderson DG (2009). Knocking downbarriers: advances in siRNA delivery. Nature Rev Drug Discov 8:129138.Yang Y, Wang Y, Li S (2004). Mutations in SCN9A, encoding asodium channel alpha subunit, in patients with primaryerythermalgia. J Med Genet 41: 171174.Yang Y, Adelstein SJ, Kassis AI (2009). Target discovery from datamining approaches. Drug Discov Today 14: 147154.Zanders ED, Bailey DS, Dean PM (2002). Probes for chemicalgenomics by design. Drug Discov Today 7: 711718.Zhang JH, Chung DY, Oldenberg KR (1999). A simple statisticalparameter for use in evaluation and validation of high throughputscreening assays. J Biomol Scr 4: 6773.BJPPrinciples of early drug discoveryBritish Journal of Pharmacology (2011) 162 12391249 1249