C Basic & Clinical Pharmacology & Toxicology 2005, 96, 151–155.Printed in Denmark . All rights reserved

Copyright C

ISSN 1742-7835

MiniReview

Molecular Modelling of Drug Targets: The Past, the Presentand the Future

Svein G. Dahl and Ingebrigt Sylte

Department of Pharmacology, Institute of Medical Biology, University of Tromsø, N-9037 Tromsø, Norway

(Received July 12, 2004; Accepted October 19, 2004)

Abstract: Most currently used therapeutic drugs have an enzyme or a membrane-bound receptor as site of action. Thesequencing of the human and other genomes has provided a potential to identify many hitherto unknown proteins thatmight serve as new drug targets. To achieve this, knowledge about three-dimensional protein structures is crucial for theunderstanding of their functional mechanisms, and for a rational drug design. Over the last decade atomic resolutioncrystal structures of soluble proteins have been reported in a rapidly increasing number, but the detailed three-dimensionalstructures are still unknown for the majority of membrane proteins since their membrane association makes experimentalstructure determinations complicated. Computerized modelling of protein structures, based on experimentally determinedstructures of homologue proteins, may be a useful methodological alternative, especially for membrane proteins. In thepast, molecular modelling of transporters and G-protein-coupled receptors was based on low-resolution structural dataobtained by cryo-electron microscopy. Recent high-resolution crystal structure determinations of a G-protein-coupledreceptor, rhodopsin, and several different transporter proteins and ion channels have enabled construction of more accu-rate receptor and transporter models. For the future, collaborative structural genomics initiatives aim at determining thethree-dimensional structure of all known proteins, based on a combination of experimental structure determination andmolecular modelling. Development of still more powerful computer hardware and software will enable extensive studiesof the protein structure and dynamics of new potential drug targets, but raises a new challenge in the validation andcalibration of computerized methods of biosimulations.

From receptors to drug targets

In 1878 the British physiologist John Newport Langley pos-tulated from his studies on the action of pilocarpin andatropine, that in order to produce a therapeutic effect, drugmolecules must reach tissue cells in the target organ (Lang-ley 1878). He later concluded that nicotine and curare acton a ‘‘receptive substance’’ in muscle cells (Langley 1905).In 1907 Paul Ehrlich postulated the existence of ‘‘chemo-receptors’’ from his studies of the selective affinity of dyesfor biological tissues (Drews 2000; Maehle et al. 2002). Thereceptor concept, based on reversible binding and governedby the law of mass action, was further developed by AlfredJoseph Clark (Clark 1933), and is often illustrated by ex-pressions such as

D (Drug)πR (Receptor) ↔ DR (Drug-Receptor complex)» Effect

Over a century the postulated existence of drug receptorsproved to be a useful concept for explaining quantitativerelationships between drug concentrations and pharmaco-

Author for correspondence: Svein G. Dahl, Department of Phar-macology, I.M.B., University of Tromsø, N-9037 Tromsø, Norway(fax π47 77 64 53 10, e-mail sgd/fagmed.uit.no).

logical effects, but was still nothing more than a theoreticalconcept.

This changed with the cloning and amino acid sequencedetermination of the nicotinic acetylcholine receptor (Nodaet al. 1982, 1983a & 1983b; Numa et al. 1983), whichstarted a new era. During the two decades since then, recep-tors have materialized from a theoretical concept into pro-teins with a known amino acid sequence. What was pre-viously considered as drug ‘‘receptors’’ as a general term isnow referred to as drug targets, and classified into enzymes,receptors of various types, ion channels, transporters, andother targets. A recent survey of the 483 drugs described ina modern textbook of pharmacology (Goodman et al.1996), showed that G-protein-coupled receptors (GPCRs)constituted the largest subgroup with 45% of all targets,while 28% of the compounds had an enzyme as site of ac-tion (Drews 2000).

Structural genomics and the protein-folding problem

A recent report by the U.S. Food and Drug Administration(FDA 2004), analyses the pipeline problem – the recent slow-down, instead of the expected acceleration, in innovativemedical therapies reaching patients. As pointed out by theFDA report, basic research provides both an understanding

152 SVEIN G. DAHL AND INGEBRIGT SYLTE MiniReview

of biology and disease processes and the foundation for dis-covery and development of new drugs. The completion ofthe human genome sequencing project in April 2003, open-ed the possibility of identifying all proteins produced bythe organism, which could substantially extend the currentnumber of some 500 drug target molecules that are beingexploited (Terwilliger 2004). However, knowing the aminoacid sequence of a protein is not the same as knowing itsfunction, and the protein-folding problem represents onemajor obstacle on the pathway from DNA sequences tonovel drug targets.

In order to understand the functional mechanisms of aprotein, it is crucial to know its three-dimensional molecu-lar structure. While it is quite straightforward to deduce theamino acid sequence of a protein from the DNA sequenceof the gene encoding it, determining the three-dimensionalmolecular structure of proteins has proven to be more diffi-cult, especially for membrane proteins that represent abouthalf of all currently exploited drug targets (Drews 2000).The average time it takes to solve an eukaryotic proteintarget from clone to three-dimensional structure has beenone to three years for soluble proteins and even longer, witha higher risk of failure, for membrane proteins (Stevens2004).

Three-dimensional protein structures may be determinedexperimentally by X-ray crystallography, NMR spectros-copy and cryo-electron microscopy, or predicted by struc-tural bioinformatics and molecular modelling techniquesbased on the homology with proteins of known three-dimen-sional structure. A table of membrane proteins with knowncrystal structures is found at http://blanco.biomol.uci.edu/Membrane_Proteins_xtal.html. While atomic resolutioncrystal structures of soluble proteins have been reported in arapidly increasing number over the last decade, such progresshas not been made in terms of membrane proteins, whichhave proven extremely difficult to crystallize for two mainreasons. One is related to the amphipathic nature of their sur-face, with a hydrophobic area in contact with membranephospholipids and polar surface areas in contact with theaqueous phases on both sides of the membrane. Secondly, themajority of medically important membrane proteins arepresent in tissues at very low concentrations, making over-ex-pression a prerequisite for structural studies, and this hasoften proven to be very difficult. Molecular modelling pro-vides a useful methodological alternative to experimentalstructure determination, especially for membrane proteins.

As indicated in fig. 1, homology modelling of a proteinstructure is based on three elements: a three-dimensionalstructural template, an amino acid sequence alignment ofthe modelled molecule with the template molecule andother related molecules, and specially adapted computersoftware. The accuracy and reliability of the protein modeldepends on the accuracy of the structural data template,and the resemblance of the modelled structure with the oneor those used as a template for the initial protein model.

So far, de novo modelling from amino acid sequences,without using any 3-dimensional template, has not pro-

duced reliable, validated 3-dimensional protein models, andmust be regarded as much less accurate than modellingbased on a 3-dimensional template.

Compared to modelling of drug targets, which usuallyare macromolecules and in most cases proteins with severalhundred amino acids, modelling of drug molecules is rela-tively straightforward. The challenge in modelling of drug-receptor interactions lies in modelling of the target mol-ecule, and in the docking of ligand molecules into postu-lated binding sites.

This review focuses on membrane-bound receptors andtransporters, where crystal structures are sparse. For en-zymes, which are the targets of action for nearly one thirdof all existing drugs, many more 3-dimensional structuresare available from x-ray crystallographic experiments.

Modelling of receptors and transporters – past and present

G-protein coupled receptors. In the past, molecular modellingof G-protein-coupled receptors was based on low-resolutionprojection maps from cryo-electron microscopy experiments.In a ground-braking paper published in 1975, R. Hendersonand P.N.T. Unwin demonstrated that bacteriorhodopsin, aprotein in the purple membrane from halobacterium halobi-um that serves as a light-driven proton pump, has seven mem-brane spanning domains that were interpreted as a-helices(Henderson & Unwin 1975). This was the structural basis forthe first molecular models of G-protein-coupled receptors,constructed shortly after the cloning of several G-protein-coupled receptors in the period from 1986 to 1989 Dahl et al.1989, 1991a & 1991b; Hibert et al. 1991).

The concept of a seven transmembrane a-helical (TMH)architecture of GPCRs, initially based on bacteriorhodop-sin which does not act via a G-protein, was well establishedand had already been used in more than 100 scientificpapers proposing GPCR models, when strong support forthe 7 TMH architecture of GPCRs came from R. Hender-son’s group in 1993. An electron projection map of visualrhodopsin, a chromoprotein in the retina that acts via aG-protein, transducin, showed seven membrane spanningdomains that appeared to be a-helices, arranged slightly dif-ferent from those in bacteriorhodopsin (Schertler et al.1993). The final proof of the 7 TMH architecture of rho-dopsin came when an atomic-resolution crystal structurewas reported seven years later (Palczewski et al. 2000).

The rhodopsin crystal structure represented a break-through both in that it demonstrated beyond any doubt the7 TMH structure of a G-protein-coupled receptor, assumedseveral years earlier, and in providing a much more accuratetemplate for molecular modelling of GPCRs than the pre-vious projection maps. At the present, the rhodopsin crystalstructure is the generally used template for modelling ofGPCRs, and it has been demonstrated that such modelsmay be successfully used to identify both agonists and anta-gonists by virtual screening of compound libraries (Bissantzet al. 2003).

It is interesting to note that there is less than 10% amino

153MODELLING DRUG TARGETSMiniReview

acid sequence identity between bacteriorhodopsin and vis-ual rhodopsin in the transmembrane helices, although thetwo proteins have a similar overall architecture with sevenmembrane spanning helices. The initial structural data onbacteriorhodopsin from 1975 therefore provided a valid, al-though less precise, template for modelling of G-protein-coupled receptors than the more recent rhodopsin structure,in spite of the low sequence identity.

Transporters. Transporter proteins in biological mem-branes may be divided into channels and carriers (Saier2000). Channels function as selective pores that open in re-sponse to a chemical or electrophysiological stimulus,allowing movement of a solute down an electrochemicalgradient, while active carrier proteins use an energy produc-ing process to translocate a substrate against a concentrationgradient. It is interesting to note that two of the most widelyprescribed drugs in the world, fluoxetine and omeprazole,have a carrier transporter protein as site of action.

It had long been assumed from cloning experiments andamino acid sequence analysis that many carrier trans-porters have a 12 THM structure with the N- and C-ter-minal parts localized intracellularly (Edvardsen & Dahl1994), when a three-dimensional electron density projectionmap of the Escherichia coli Naπ/Hπ antiporter (NhaA) wasreported from cryo-electron microscopy experiments (Willi-ams 2000). The NhaA electron density map showed twelvecylindrical features, assumed to represent membrane span-ning domains, but did not indicate the order of the corre-sponding domains in the protein sequence, nor which sidesof the map that represented the periplasmic and cyto-plasmic sides of the transporter. Combined with resultsfrom other biochemical studies, the low-resolution NhaA

Fig. 1. Main elements of homology-based modelling of proteins.

structure still provided a useful template for molecularmodelling of the transporter (Ravna et al. 2001).

In the three years following the publication of the NhaAelectron density projection map, six high-resolution crystalstructures of carrier transporters were reported fromsynchrotron X-ray crystallographic studies (Toyoshima et al.2000; Chang & Roth 2001; Locher et al. 2002; Murakami etal. 2002; Abramson et al. 2003; Huang et al. 2003), in ad-dition to seven other low-resolution structures from cryo-electron microscopy experiments (Rosenberg et al. 2001a &2001b; Chami et al. 2002; Hirai et al. 2002; Lee et al. 2002;Ferreira-Pereira et al. 2003; ; Ubarretxena-Belandia et al.2003).

At the present, the high-resolution crystal structures ofthe six carrier transporters and of several ion channels(Doyle et al. 1998; Dutzler et al. 2002; Jiang et al. 2003)and water channels (Agre & Kozono 2003) have providednew opportunities for more accurate modelling of these andother homologue transporters, than what was previouslypossible with electron density maps as template.

However, also the accuracy of homology-based models islimited by the validity of the sequence alignment of themodeled molecule to the template, the resemblance of themodeled structure with the one used as a template, and bythe accuracy of the modelling of inserted or non-alignedregions. This may be particularly important when a bac-terial protein structure is used as template to model a mam-malian membrane protein that has substantially larger in-tra- and extracellular loops between the membrane-span-ning a-helices (Dahl et al. 2004). Five out of the six high-resolution carrier transporter structures reported up to nowwere from Escherichia coli.

154 SVEIN G. DAHL AND INGEBRIGT SYLTE MiniReview

Future challenges

In the future as in the past and at the present, molecularmodelling of receptors and other potential drug targets willprobably still rely on structural templates, sequence align-ments and specially adapted software. However, accelerat-ing advances in synchrotron X-ray crystallography, NMRspectroscopy, structural genomics, molecular biology, bioin-formatics, computer hardware and software development israpidly changing the field.

The ‘‘Protein Structure Initiative’’, initiated in 1999, hasas long-range goal to make three-dimensional atomic-levelstructures of most proteins easily obtainable from knowl-edge of their corresponding DNA sequences (PSI 2004).The project, which plans to enter into a ‘‘Production phase’’in 2005, aims at organising all known protein sequencesinto families, select representative targets from each family,carry out experimental three-dimensional structure deter-mination of targets by X-ray crystallography or NMR spec-troscopy, and build models of all other proteins by homo-logy to solved three-dimensional structures.

The ‘‘Joint Center for Structural Genomics’’ has as objec-tive to create new technologies to drive high-throughputstructure determination (JCSG 2004). As of May 21, 2004,2461 different targets had been cloned and expressed, andamong 80 structures that had been solved and deposited inthe Protein Data Bank, 11 represented new folds. The pro-ject has as a 5-year goal to generate three-dimensional struc-tures of approximately two thousand proteins.

In addition to the results from improved experimentalstructure determination techniques, future modelling ofdrug-target interactions will benefit from more powerfulcomputer hardware that may permit molecular dynamicssimulations of longer periods that the nanosecond intervalsthat represent the limits of most present simulations. Drug-target simulations are likely to strongly benefit from ad-vances already underway in confocal microscopy, fluor-escence labeling and other experimental techniques to studyprotein dynamics and cell signaling. A wealth of new soft-ware for biosimulation has become available, also for drugabsorption, distribution and elimination (van de Waterbe-emd & Gifford 2003), and this evolution is likely to con-tinue at an increasing pace.

The hundreds of software programmes for biological ap-plications already available and those likely to appear inthe future, obviously raises the need for a better integrationbetween the experimental and computational sides of struc-tural biology, that hopefully will be achieved within the nextdecade (Stevens 2004). As pointed out in a recent commen-tary (Bottomley 2004), it seems that, while software is in-creasingly relied on as a biological tool, it is not given thesame careful and critical consideration as other tools usedin biological research. The need to calibrate, characterizeand standardize software, and to understand how appliedsoftware works, is a major future challenge. ‘‘We should notbe so impressed by the perceived efficacy of software thatwe are willing to forgo the strict scientific procedures and

quality control that we apply to other laboratory tools’’(Bottomley 2004).

When the Protein Data Bank (PDB) was set up at Brook-haven National Laboratory in 1971 it held seven structures.Today it has more than 22,000 protein and peptide struc-tures, and contains an increasing amount of ‘meta-data’ inaddition to atomic coordinates, describing how the proteinwas produced and purified, and how its structure wassolved. Structural genomics and the rise in high-throughputprojects are expected to double the number of data itemsattached to each protein submitted to the PDB (Buckingh-am 2004). The need of integration of different data types,with possibilities of queries across distributed databases ofdifferent types, poses a major challenge to software devel-opers and administrators of structural databases.

Above all, the mutual integration of different theoretical/computational and experimental approaches is a major pre-requisite for successful drug design.

Concluding remarks

During the last years genome sequencing has provided andwill still continue to provide important information aboutthe genetic map of different organisms. At the same timenew technologies (e.g. microarray technology, 2D electro-phoresis/Mass Spectrometry) have provided insight into themechanisms of regulations at the DNA, RNA and proteinlevel. In the post-genomic area, the focus will be on theunderstanding of the cellular machinery for regulation andcommunication, and how proteins and other gene productscooperate on a detailed atomic level. Such information isproviding a better understanding of both biological mech-anisms and disease processes, and is of immense importancefor the discovery and development of new drugs. In orderto understand this machinery, theoretical calculations andmodelling will be an important support to experimentalstudies. Due to the increase in computational power, theor-etical modelling may give an insight into processes andmechanisms that might be impossible or too expensive ortime-consuming to study by experimental methods.

References

Abramson, J., I. Smirnova, V. Kasho, G. Verner, H. R. Kaback &S. Iwata: Structure and mechanism of the lactose permease ofEscherichia coli. Science 2003, 301, 610–615.

Agre, P. & D. Kozono: Aquaporin water channels: molecular mech-anisms for human diseases. FEBS Lett. 2003, 555, 72–78.

Bissantz, C., P. Bernard, M. Hibert & D. Rognan: Protein-basedvirtual screening of chemical databases. II. Are homology modelsof G-Protein Coupled Receptors suitable targets? Proteins 2003,50, 5–25.

Bottomley, S.: Bioinformatics: smartest software is still just a tool.Nature 2004, 429, 241.

Buckingham, S.: Bioinformatics: data’s future shock. Nature 2004,428, 774–777.

Chami, M., E. Steinfels, C. Orelle, J. M. Jault, A. Di Pietro, J. L. Ri-gaud & S. Marco: Three-dimensional structure by cryo-electronmicroscopy of YvcC, an homodimeric ATP-binding cassette trans-porter from Bacillus subtilis. J. Mol. Biol. 2002, 315, 1075–1085.

155MODELLING DRUG TARGETSMiniReview

Chang, G. & C. B. Roth: Structure of MsbA from E. coli: a homo-log of the multidrug resistance ATP binding cassette (ABC)transporters. Science 2001, 293, 1793–1800.

Clark, A. J.: The mode of action of drugs on cells. E. Arnold Co.,London, 1933.

Dahl, S. G., Ø. Edvardsen, E. Heimstad & I. Sylte: Three dimen-sional structure of the dopamine D2 receptor. VIII World Con-gress of Psychiatry. Athens, Greece, 1989, pp. 489.

Dahl, S. G., Ø. Edvardsen & I. Sylte: Molecular dynamics of dop-amine at the D2 receptor. Proc. Natl. Acad. Sci. U S A 1991a,88, 8111–5.

Dahl, S. G., Ø. Edvardsen & I. Sylte: Molecular modeling of anti-psychotic drugs and G protein coupled receptors. Therapie1991b, 46, 453–9.

Dahl, S. G., A. W. Ravna & I. Sylte: Structures and models of trans-porter proteins. J. Pharmacol. Exp. Therap. 2004, 309, 853–860.

Doyle, D. A., J. Morais Cabral, R. A. Pfuetzner, A. Kuo, J. M.Gulbis, S. L. Cohen, B. T. Chait & R. MacKinnon: The structureof the potassium channel: molecular basis of Kπ conduction andselectivity. Science 1998, 280, 69–77.

Drews, J.: Drug discovery: a historical perspective. Science 2000,287, 1960–1964.

Dutzler, R., E. B. Campbell, M. Cadene, B. T. Chait & R. MacKin-non: X-ray structure of a ClC chloride channel at 3.0 A revealsthe molecular basis of anion selectivity. Nature 2002, 415, 287–294.

Edvardsen, O. & S. G. Dahl: A putative model of the dopaminetransporter. Brain Res. Mol. Brain Res. 1994, 27, 265–274.

FDA: Innovation or Stagnation? Challenge and opportunity onthe critical path to new medical products. In: U.S. Departmentof Health and Human Services. Food and Drug Administration,2004 (http://www.fda.gov/oc/initiatives/criticalpath/whitepaper.html)

Ferreira-Pereira, A., S. Marco, A. Decottignies, J. Nader, A. Gof-feau & J. L. Rigaud: Three-dimensional reconstruction of theSaccharomyces cerevisiae multidrug resistance protein Pdr5p. J.Biol. Chem. 2003, 278, 11995–11999.

Goodman, L. S., L. E. Limbird, P. B. Milinoff, R. W. Ruddon,A. G. Gilman & J. G. Hardman In: Goodman and Gilman’s ThePharmacological Basis of Therapeutics. McGraw-Hill, New York,1996.

Henderson, R. & P. N. Unwin: Three-dimensional model of purplemembrane obtained by electron microscopy. Nature 1975, 257,28–32.

Hibert, M. F., S. Trumpp-Kallmeyer, A. Bruinvels & J. Hoflack:Three-dimensional models of neurotransmitter G-binding pro-tein-coupled receptors. Mol. Pharmacol. 1991, 40, 8–15.

Hirai, T., J. A. Heymann, D. Shi, R. Sarker, P. C. Maloney & S.Subramaniam: Three-dimensional structure of a bacterial oxalatetransporter. Nat. Struct. Biol. 2002, 9, 597–600.

Huang, Y., M. J. Lemieux, J. Song, M. Auer & D. N. Wang: Struc-ture and mechanism of the glycerol-3-phosphate transporterfrom Escherichia coli. Science 2003, 301, 616–620.

JCSG: Joint Center for Structural Genomics, 2004 (http://www.jcsg.org/).

Jiang, Y., A. Lee, J. Chen, V. Ruta, M. Cadene, B. T. Chait & R.MacKinnon: X-ray structure of a voltage-dependent Kπ chan-nel. Nature 2003, 423, 33–41.

Langley, J. N.: On the reaction of cells and of nerve endings tocertain poisons, chiefly as regards the reaction of striated muscleto nicotine and to curari. J. Physiol. 1905, 33, 374–413.

Langley, J. N.: On the physiology of salivary secretion. Part II. Onthe mutual antagonism of atropin and pilocarpin, having especialreference to their relations in the sub-maxillary gland of the cat.J. Physiol. 1878, 1, 339–369.

Lee, J. Y., I. L. Urbatsch, A. E. Senior & S. Wilkens: Projectionstructure of P-glycoprotein by electron microscopy. Evidence for

a closed conformation of the nucleotide binding domains. J. Biol.Chem. 2002, 277, 40125–40131.

Locher, K. P., A. T. Lee & D. C. Rees: The E. coli BtuCD structure:a framework for ABC transporter architecture and mechanism.Science 2002, 296, 1091–1098.

Maehle, A. H., C. R. Prull & R. F. Halliwell: The emergence of thedrug receptor theory. Nat. Rev. Drug Discov. 2002, 1, 637–641.

Murakami, S., R. Nakashima, E. Yamashita & A. Yamaguchi:Crystal structure of bacterial multidrug efflux transporter AcrB.Nature 2002, 419, 587–593.

Noda, M., H. Takahashi, T. Tanabe, M. Toyosato, Y. Furutani, T.Hirose, M. Asai, S. Inayama, T. Miyata & S. Numa: Primarystructure of alpha-subunit precursor of Torpedo californica ace-tylcholine receptor deduced from cDNA sequence. Nature 1982,299, 793–797.

Noda, M., H. Takahashi, T. Tanabe, M. Toyosato, S. Kikyotani, Y.Furutani, T. Hirose, H. Takashima, S. Inayama, T. Miyata & S.Numa: Structural homology of Torpedo californica acetylcholinereceptor subunits. Nature 1983a, 302, 528–532.

Noda, M., H. Takahashi, T. Tanabe, M. Toyosato, S. Kikyotani, T.Hirose, M. Asai, H. Takashima, S. Inayama, T. Miyata & S.Numa: Primary structures of beta- and delta-subunit precursorsof Torpedo californica acetylcholine receptor deduced fromcDNA sequences. Nature 1983b, 301, 251–255.

Numa, S., M. Noda, H. Takahashi, T. Tanabe, M. Toyosato, Y.Furutani & S. Kikyotani: Molecular structure of the nicotinicacetylcholine receptor. Cold Spring Harb Symp Quant Biol 1983,48 Pt 1, 57–69.

Palczewski, K., T. Kumasaka, T. Hori, C. A. Behnke, H. Motoshi-ma, B. A. Fox, I. Le Trong, D. C. Teller, T. Okada, R. E. Stenk-amp, M. Yamamoto & M. Miyano: Crystal structure of rhodop-sin: A G protein-coupled receptor. Science 2000, 289, 739–45.

PSI: Protein Structure Initiative, 2004 (http://www.structuralgenom-ics.org/).

Ravna, A. W., I. Sylte & S. G. Dahl: Molecular model of the Escher-ichia coli Naπ/Hπ antiporter NhaA. Receptors Channels 2001, 7,319–328.

Rosenberg, M. F., Q. Mao, A. Holzenburg, R. C. Ford, R. G. De-eley & S. P. Cole: The structure of the multidrug resistance pro-tein 1 (MRP1/ABCC1). crystallization and single-particle analy-sis. J. Biol. Chem. 2001a, 276, 16076–16082.

Rosenberg, M. F., G. Velarde, R. C. Ford, C. Martin, G. Berridge,I. D. Kerr, R. Callaghan, A. Schmidlin, C. Wooding, K. J. Lin-ton & C. F. Higgins: Repacking of the transmembrane domainsof P-glycoprotein during the transport ATPase cycle. Embo. J.2001b, 20, 5615–5625.

Saier, M. H., Jr.: A functional-phylogenetic classification system fortransmembrane solute transporters. Microbiol. Mol. Biol. Rev.2000, 64, 354–411.

Schertler, G. F., C. Villa & R. Henderson: Projection structure ofrhodopsin. Nature 1993, 362, 770–772.

Stevens, R. C.: Long live structural biology. Nat. Struct. Mol. Biol.2004, 11, 293–295.

Terwilliger, T. C.: Structures and technology for biologists. Nat.Struct. Mol. Biol. 2004, 11, 296–297.

Toyoshima, C., M. Nakasako, H. Nomura & H. Ogawa: Crystalstructure of the calcium pump of sarcoplasmic reticulum at 2.6Å resolution. Nature 2000, 405, 647–655.

Ubarretxena-Belandia, I., J. M. Baldwin, S. Schuldiner & C. G.Tate: Three-dimensional structure of the bacterial multidrugtransporter EmrE shows it is an asymmetric homodimer. Embo.J. 2003, 22, 6175–6181.

van de Waterbeemd, H. & E. Gifford: ADMET in silico modelling:towards prediction paradise? Nat. Rev. Drug. Discov. 2003, 2,192–204.

Williams, K. A.: Three-dimensional structure of the ion-coupledtransport protein NhaA. Nature 2000, 403, 112–115.