The recent IBC conference on Proteomics:Delivering New Routes to Drug Discovery(14–17 May 2001, Philadelphia, PA,USA) provided a broad coverage of theapplication of innovative proteomictechnologies for drug discovery. It wouldbe difficult to review the large numberof presentations during the four daysand details of technologies are describedelsewhere1. The first part of this con-ference review deals with the role of proteomics in drug development andstructural genomics.

Functional genomics andmolecular profilingLothar Germeroth (Cytos ProteomeTherapeutics, Konstanz, Germany) de-scribed the company’s protein-discoverytechnology as an alternative to conven-tional approaches to decipher genefunction on a genomic scale. By apply-ing highly sophisticated in vitro and invivo assays, a virally reproduced library israpidly screened for drug candidates.This functional genomics approach enables the reproduction of the entirehuman proteome as a large library of therapeutic human proteins, which enables target identification, target vali-dation and small-molecule profiling.

The pitfalls of proteomic analysiscaused by polymorphism mis-annotationand chimeric mRNAs were described byChristopher Southan (Gemini Genomics,Cambridge, UK). These anomalies canbe uncovered by a detailed manual examination of all the sequences and literature links.

Proteomic technologies allow theidentification of proteins associated with

phagosomes – organelles that enablemacrophages to participate in tissue remodelling, clear apoptotic cells and restrict the spread of intracellularpathogens2. The advantages of an organelle-proteomic approach is that it provides functional insights into signalling pathways and protein orien-tation. Similarly, John Bergeron (CaprionPharmaceuticals, Montreal, Canada) de-scribed the use of subcellular proteomicsfor identification of novel post-transcrip-tional drug targets, which can be usedfor target validation, determination oflead mechanism of action and toxicityprofiling.

Target validation in the post-genomic eraStrategic requirements for drug-targetvalidation in the post-genomic era requirean increase in speed as well as quality.High-quality cell models can facilitatethis process but, frequently, cell lines aredifficult to obtain in large numbers, diffi-cult to transfect and undergo replicativequiescence.

Mark Lindsay (AstraZeneca, Macclesfield,UK) described developmental cell mod-els for target identification and valida-tion, in which conditional immortaliza-tion of genes can stimulate proliferationwithout differentiation, but when theproliferation stimulus is turned off thecells can differentiate again. Such im-mortalized cells are obtainable in largenumbers and are easier to transfect ortransform than primary cells, makingthem useful for target identification andprioritization. Specific targeting of agene, protein or mRNA using antisense

technology can modulate gene expres-sion. However, these approaches are still‘hit-and-miss’, and there are problemswith toxicity and cell delivery. Alternativeapproaches include the use of DNA,peptide nucleic acid (PNA) or locked nu-cleic acid (LNA), which can be applied togene targeting (by using triple-helix-forming oligonucleotides and zinc-fingerproteins) and in vivo target validation (by acute application of oligonucleotides or genetic manipulation of ribozymes).Delivery to tissues is still a problem, butfusion of a protein to the protein trans-duction domain (PTD) of certain organ-isms results in direct delivery of active fusion protein to all tissues; in vivodelivery of a β-galactosidase–PTD fusion protein from the HIV protein, TAT, hasbeen demonstrated3.

High-throughput protein validationwas discussed by Stewart Chipman(Immunex Corporation, Seattle, WA, USA).The protein-function screening processrequires a panel of robust bioassays thatare capable of quick, accurate andreproducible prediction of the immune function of novel proteins. An essentialelement is an inventory of purified, novelproteins as a consistent supply of bio-logically active material for screening. Tothis end, an information managementsystem has been built to track projectworkflow, facilitate data analysis and disseminate scientific information.

High-throughput protocols for identi-fying highly active antisense moleculesthat inhibit gene expression were pre-sented by Nicholas Dean (GeneTrove,Carlsbad, CA, USA), who highlighted potential applications of this approach

update conferences DDT Vol. 6, No. 15 August 2001

1359-6446/01/$ – see front matter ©2001 Elsevier Science Ltd. All rights reserved. PII: S1359-6446(01)01913-4772

Proteomics: delivering new routes todrug discovery – Part 1Kewal K. Jain, Bläsiring 7, CH-4057 Basel, Switzerland, tel/fax: +4161 692 4461, e-mail: jain@pharmabiotech.ch, URL: http://pharmabiotech.ch

such as the in vitro and in vivo inhibitionof interleukin-5 (IL-5)-mediated eosino-poiesis by murine IL-5 receptor-α (IL-5R-α) antisense oligonucleotide4. The use ofantisense affords speed, selectivity andflexibility to gene functionalization andtarget validation. GeneTrove’s drug dis-covery platform can be applied to anygene, provides predictable pharmaco-kinetics and enables a rapid transitionfrom gene targeting to human therapy.

Post-genomic drug developmentIntegrated technologies – chemical, bio-logical and robotic – and their role inpost-genomic drug development werediscussed by James Lorens (Rigel, SouthSan Francisco, CA, USA). Vast retrovirallibraries and functional screening assaysare often used in drug discovery pro-grammes. Rigel uses its rapid targetidentification technology to identify andcharacterize the potential drug targetsthat arise from such assays. Proteomicsand two-hybrid screening technologiesare then used to study the interactionsof these targets, which provides immedi-ate validation of the target in the contextof a disease-specific cellular response.Rigel’s post-genomics combinatorial-biology technology addresses many ofthe limitations of traditional genomics-based drug discovery by bypassing theneed to know the identity or sequenceof the genes to discover new drug targetsand, ultimately, the small-molecule drugsthat can modulate them5.

The nematode, Caenorhabditis ele-gans, can be used as a live test-tube toidentify the targets of compounds withunknown modes of action, find newchemical drug leads and discover thefunction of human genes, Thierry Bogaert(deVGen, Ghent, Belgium) explained. Byfeeding with bacteria expressing an ap-propriate gene, genes in disease-relevantpathways in C. elegans can essentially besilenced. This technology is used to vali-date candidate targets discovered byproteomics, gene-expression profilingand protein-linkage mapping. Validation

is based on the restoration of selecteddisease pathways to normal function.This Function Factory™ technology isused to analyze reduced gene functionor the effect of compounds, as well as to profile phenotypes of novel genes orcompounds. This technology is appli-cable to many target classes includingion channels and scaffold kinases.

The Human Genome Project and microbial sequencing have provided several new targets for drug discovery,which include many proteins of un-known function. Approximately one-thirdof the genes in Streptococcus pyogeneshave no identifiable function6. To meetthese challenges, David Nelson (AnadysPharmaceuticals, San Diego, CA) pre-sented an approach to drug discoverythat uses integrated biology and medicinal chemistry to advance from proteomics to drugs. The proprietary technology of the company is GATE(Genetics Assisted Target Evaluation), inaddition to technologies for HTS of chal-lenging targets for drug developmentthat include: ATLAS (Any Target LigandAffinity Screen) for protein targets; SCAN(Screen for Compounds with Affinity forNucleic Acids) for RNA targets; and ribo-proteomics (systematic annotation ofRNA–protein interactions that affect RNAmetabolism: transport, splicing, trans-lation and decay). Anadys is using thesetechnologies to translate human and microbial genomic information intosmall-molecule therapeutics.

Structural genomicsThe goal of structural genomics is to obtain useful, three-dimensional (3D)models of all proteins using a combinationof experimental structure-determinationand comparative model building. Theshapes of most protein sequences aremodelled on their similarity to experi-mentally determined protein structures7.Dennis Vitkup (MIT Genome Center/Harvard University, Cambridge, MA, USA)described target-selection strategies instructural genomics. Structural-genomic

approaches include the organization ofknown protein sequences into domainfamilies and the selection of family representatives as targets for structuredetermination by X-ray crystallographyand NMR. The strategy that maximizesstructural coverage requires about seventimes fewer structure determinationscompared with the strategy in whichrandom target selection. With a choiceof reasonable model-quality and thegoal of 90% structural coverage, itwould take ~16,000 carefully selectedstructure-determinations to constructatomic models for the majority of pro-teins8. This task could be accomplishedwithin a decade provided that selectionof targets is highly co-ordinated andsignificant funding is available. One ofthe examples given was the Pfam pro-tein family database, a collection of pro-tein family domains, the latest version ofwhich contains 2000 families, 260,000domains and 180,000 sequences. InEscherichia coli, 52% of the sequenceshave at least one Pfam.

Mark Gerstein (Yale University, NewHaven, CT, USA) focused on analyzingfunctional genomic data as a finite list ofprotein ‘parts’ that can be either a pro-tein fold or a protein family. A new webresource called PartsList (http://www.partslist.org) enables comparative pro-tein-fold surveys9. It will rank the ~420protein folds based on more than 180 structural attributes. The integration of whole-genome expression and pro-tein–protein interaction data with struc-tural information is a novel feature of thissystem. The ‘parts’ can then be used tointerpret genomes and mine expressiondata.

The simultaneous high-throughputcrystallographic and NMR spectroscopicanalysis of pharmaceutically and med-ically important proteins, as used at the Berlin Protein Structure Factory(http://www.fu-berlin.de/psf), was de-scribed by Konrad Büssow (Max PlanckInstitute for Molecular Genetics, Berlin,Germany). This approach will close

DDT Vol. 6, No. 15 August 2001 updateconferences

773

the gap between genomic and struc-tural information. Cheryl Arrowsmith(Integrative Proteomics, Toronto, Canada)also emphasized the role of NMR instructural proteomics and regards it as a link between proteins and chemistrythat has the potential to be high-throughput.

Many structural genomic initiativeshave focused initially on solving thestructures of readily expressed proteins,but high-throughput protein productionremains a challenge. Shane Teremi(Schering-Plough Research Institute,Kenilworth, NJ, USA) described how tobuild the protein-production infrastruc-ture that addresses this problem. Such‘directed structural genomics’ involvesstrategies in which various high-throughput technologies, including multiple expression-vectors, engineeredhost-cells and solubility sparse matrixes,have been incorporated at both the expression and purification stages.Schering-Plough’s proprietary GATE-WAY™ technology uses converted, des-tination-vectors (incorporating cleavableN-termini and green fluorescent proteinC-termini) for protein expression and en-gineering. A protocol was presented forthe entire process from target genecloning and protein expression to grow-ing crystals; it is possible to automatethis process. The benefits of such an approach are that it can assess more targets simultaneously and rapidly deter-mine the suitability of targets for struc-tural studies.

The New York Structural GenomicsResearch Consortium has yielded struc-tural information on five enzymes in the cholesterol biosynthesis pathway.Stephen Burley (Rockefeller University,New York, NY, USA) discussed the rapidexpression, purification, crystallizationand structural characterization of two ofthese enzymes. Comparative protein-structure modelling with these structuresgave reliable homology models for >210different proteins, giving an insight

into biochemical function and proteinevolution. The problem of target selec-tion from >500,000 protein sequences isreduced to selection from ~30,000 fami-lies at 30% sequence identity, therefore,the families can be prioritized.

Structure-based drug designRaymond Salemme (3-DimensionalPharmaceuticals, Exton, PA, USA) de-scribed the DirectedDiversity™ technol-ogy, a combinatorial chemistry approachwith structure-based drug-design tech-nology that uses X-ray crystallography to directly visualize how drugs bind to atarget receptor. This approach enablesthe simultaneous investigation andoptimization of multiple drug propertiesthroughout the process of drug design.

David Bailey (De Novo Pharmaceuticals,Cambridge, UK) described the use ofstructural genomics to expand the medi-cinal chemistry vision and predicted thathuman therapeutic targets defined bygenomics could yield small-moleculetherapeutics by 2005. The next wave ofdrug discovery will be driven by chemi-cal genomics linked to exploratory ligand-design.

Assigning function from structureJeffrey Skolnick (Danforth Plant ScienceCenter, St Louis, MO, USA) described a novel method for protein function prediction based on a sequence-to-structure-to-function paradigm. Thestructure-based approaches to predictprotein function from amino-acid se-quence can be applied to entire genomesand can be extended to ligand dockingand quaternary structure prediction.

New technologies for merging struc-ture-based drug design and structuralgenomics were discussed by Peter Rose(Pfizer La Jolla Laboratories, San Diego,CA, USA). Comparative modelling enables the generation of 3D models oftarget proteins from closely related homologues of known structure. Thesetechniques can be used to evaluate

putative targets for drugability, to facili-tate the structure solution of new targetsand to design selective inhibitors forhuman targets or broad-spectrum inhibitors for anti-infective targets.

Concluding remarksThis was a comprehensive proteomicsconference, the size of which reflects theexpansion in proteomics technologiesand applications. In addition to the pre-sentations highlighted here, there wereparallel sessions on sample preparationand strategic corporate alliances. Thesecond part of this conference report willbe published in the 15th August 2001issue of Drug Discovery Today and willdeal with structure-based drug designand analysis of proteomic information.

References1 Jain, K.K. (2001) Proteomics: technologies and

commercial opportunities (7th edn), JainPharmaBiotech Publications

2 Garin, J. et al. (2001) The phagosomeproteome: insight into phagosome functions.J. Cell Biol. 152, 165–180

3 Schwarze, S.R. et al. (1999) In vivo proteintransduction: delivery of a biologically activeprotein into the mouse. Science 285,1569–1572

4 Lach-Trifilieff, E. et al. (2001) In vitro and invivo inhibition of interleukin (IL)-5-mediatedeosinopoiesis by murine IL-5R alphaantisense oligonucleotide. Am. J. Respir. CellMol. Biol. 24, 116–122

5 Peelle, B. et al. (2001) Intracellular proteinscaffold-mediated display of random peptidelibraries for phenotypic screens inmammalian cells. Chem. Biol. 8, 521–534

6 Ferretti, J.J. et al. (2001) Complete genomesequence of an M1 strain of Streptococcuspyogenes. Proc. Natl. Acad. Sci. U. S. A. 98,4658–4663

7 Sanchez, R. et al. (2000) Protein structuremodeling for structural genomics. Nat. Struct.Biol. 7 (Suppl.), S986–990

8 Vitkup, D. et al. (2001) Completeness instructural genomics. Nat. Struct. Biol. 8,559–566

9 Qian, J. et al. (2001) PartsList: a web-basedsystem for dynamically ranking protein foldsbased on disparate attributes, includingwhole-genome expression and interactioninformation. Nucleic Acids Res. 29, 1750–1764

update conferences DDT Vol. 6, No. 15 August 2001

774