Biochemical Pharmacology 69 (2005) 1133–1139

Commentary www.elsevier.com/locate/biochempharm

Target identification and validation in drug discovery:

the role of proteomics
Karla K. Kopec *, Donna Bozyczko-Coyne, Michael Williams
Worldwide Discovery Research, Cephalon Inc., 145 Brandywine Parkway, West Chester, PA 19380, USA
Received 1 December 2004; accepted 12 January 2005

Abstract

Proteomics, the study of cellular protein expression, is an evolving technology platform that has the potential to identify novel proteins
involved in key biological processes in the cell that may serve as potential drug targets. While proteomics has considerable theoretical
promise, individual cells/tissues have the potential to generate many millions of proteins while the current analytical technologies that
involve the use of time-consuming two dimensional gel electrophoresis (2DIGE) and various mass spectrometry (MS) techniques are
unable to handle complex biological samples without multiple high-resolution purification steps to reduce their complexity. This can
significantly limit the speed of data generation and replication and requires the use of bioinformatic algorithms to reconstitute the parent
proteome, a process that does not always result in a reproducible outcome. In addition, membrane bound proteins, e.g., receptors and ion
channels, that are the targets of many existing drugs, are not amenable to study due, in part, to limitations in current proteomic techniques
and also to these being present in low abundance and thus disproportionally represented in proteome profiles. Subproteomes with reduced
complexity have been used to generate data related to specific, hypothesis-driven questions regarding target identification, protein-
interaction networks and signaling pathways. However progress to date, with the exception of diagnostic proteomics in the field of cancer,
has been exceedingly slow with an inability to put such studies in the context of a larger proteome, limiting the value of the information.
Additionally the pathway for target validation (which can be more accurately described at the preclinical level as target confidence
building) remains unclear. It is important that the ability to measure and interrogate proteomes matches expectations, avoiding a repetition
of the disappointment and subsequent skepticism that accompanied what proved to be unrealistic expectations for the rapid contribution of
data based on the genome maps, to biomedical research.
# 2005 Elsevier Inc. All rights reserved.
Keywords: Proteomics; Drug discovery; Target validation; Target confidence building

1. Introduction Proteomes are extremely complex with component pro-

teins undergoing extensive post-translational modification.
Proteomics, defined as the ‘‘global analysis of changes A cell expressing a third of the human genome (10,000
in the quantities, and post-translational modifications, of genes, although there is still ongoing debate as to whether
all proteins in cells’’ [1] involves the full complement of the actual number is smaller [3] or much larger) is thus
proteins expressed by a genome [2]. These proteins are cell capable of producing more than 10 MM different proteins
and tissue specific and are affected by age, disease and [1]. Traditional two dimensional gel electrophoresis
trauma. Thus, unlike the genome, proteomes (many of (2DIGE) gels are capable of separating only 2000 proteins
which can be generated from the same source) are tempo- with larger gels being required to detect upwards of 8500
rally and spatially dynamic, with each analytical profile proteins [4]. Both are still far short of being able to separate
representing a unique moment in time. Proteomics can also the hundreds of thousands of proteins that can potentially
be used to assess differences in protein expression in cells exist in a given proteome. Thus proteome complexity far
or tissues in response to drug exposure and to differentiate outweighs the technologies currently available for char-
between proteomes from diseased and normal tissues. acterization requiring the subfractionation of proteomes to
accommodate technological limitations. Similarly, the
* Corresponding author. Tel.: +1 610 738 6282; fax: +1 610 344 0065. membrane proteins present on the cell surface that repre-
E-mail address: [email protected] (K.K. Kopec). sent the major targets through which many known drugs

0006-2952/$ – see front matter # 2005 Elsevier Inc. All rights reserved.
doi:10.1016/j.bcp.2005.01.004
1134 K.K. Kopec et al. / Biochemical Pharmacology 69 (2005) 1133–1139

act [5], are not readily amenable to proteomic analysis due, referenced to other datasets; the masking of low abundance
again, to limits in the current technology [1]. proteins by more abundant and separable ‘housekeeping’
Only a fraction of the putative drug targets thought to be proteins, e.g., cytoskeletal and matrix proteins; the inability
present in the genome have been identified to date [6,7]. to amplify these low abundance proteins in the absence of a
This, together with a paucity of new drug approvals over proteomic technology analogous to the polymerase chain
the past decade [8,9] has placed a high premium on target reaction (PCR); subtle to major differences in the initial
identification with proteomics-based approaches having sample from which the proteome is extracted even when
the potential to both identify and validate disease-related using homogenous cell culture sources; the inherent com-
target proteins [7,10]. While many academic and industrial plexity of the initial proteomic sample that requires exten-
institutions have invested significantly in proteomics, the sive purification to accommodate current 2DIGE gel and
consensus to date, with the exception of clinical proteo- mass spectrometry (MS) methodologies [14] that can result
mics in the cancer area [11], is that the output from in extensive protein loss particularly of those proteins in low
proteomics initiatives has fallen short of initial expecta- abundance; the challenges in isolating hydrophobic trans-
tions [1]. This has led to considerable skepticism due to membrane proteins, e.g., receptors and ion channels; the
concerns of a repetition of the hype associated with the as availability and specificity of antibodies to the numerous
yet unrealized promise of genome-based medicine [12]. proteome components in both their native and post-transla-
More importantly, it has led to a realization that the tional forms, e.g., phosphorylated/dephosphorylated; the
logistical complexity in planning, performing and analyz- time required for data generation, analysis and integration;
ing proteome studies has been underestimated. data replication semantics, including statistical validation of
the results; and validation of the putative drug target. One
2. Proteomic technologies: separating the wheat consequence of these challenges is that the proteins that may
from the chaff be of the greatest value as potential drug targets are probably
lost as a result of the purification process used to find them,
The use of proteomics in drug target discovery is limited an excellent biological example of Heisenberg’s Principle.
by a number of conceptual and technical challenges
(Table 1). These include: the infinite number of proteomes 2.1. Protein array prioritization
that can be generated from a single source based on the
effects of age, disease and tissue manipulation; the vast The ability to identify novel disease-related proteins,
amount of protein-expression data [13] unique to a proteome when these represent only a minor part of a proteome or
that ‘‘relates to only one particular situation, in one parti- when they are not well characterized, makes it difficult to
cular tissue, in one particular gel system’’ [1] that is being prioritize proteins for further investigation. Thus, proteins
used to populate proteomic databases and cannot be cross- whose biology and function are, to some degree under-

Table 1
Proteomeics challenges and possible solutions
Challenge Technical issue Solution?
Proteome heterogeneity No single proteome—proteomes are temporally Improved experimental design. Proteome
and spatially dynamic—affected by age, disease, consortia at initiation co-coordinating
trauma and drug treatment experimentation to provide multiple data
sources for same experimental paradigm
[40]. Replication and integration of similar
studies —support of results using
non-proteomic approaches [41]
Database populating Overloaded with data that cannot easily be Additional biological insight, co-coordinating
compared or integrated (GIGO) experimentation, use of systems-based
approaches [38,40–42]. More rigorous
statistical analysis.
Low abundance protein/loss Important proteins, e.g., GPCRs and ion channels not Improved subproteome separation techniques.
of protein during analysis easily amenable to isolation and separation. Larger scale, real time separation, purification
Presence masked by ‘‘housekeeping’’ proteins and analysis techniques with supporting
(cytoskeleton etc). No protein amplification bioinformatics capabilities [37]
step similar to PCR
Data replication Not all proteins seen in apparently similar proteomes [25] Replication of full experiment not just
proteomic analysis. Co-coordinating
experimental approaches
Target validation Target validation only occurs when a drug-like compound, Recognition that preclinical ‘‘target validation’’
selective for the novel disease target shows efficacy is in reality target confidence building.
in its target patient population [9] Develop reliable paradigms via consortia
approaches [1,40]
 K.K. Kopec et al. / Biochemical Pharmacology 69 (2005) 1133–1139 1135

stood, e.g., recognizable motifs, tend to be prioritized for membrane proteins, smaller scale ‘‘subproteomes’’
analysis while those lacking in detail end up at the bottom are obtained by subcellular fractionation, affinity label-
of the list until additional information becomes available. ing and chromatography of the parent proteome and
This can result in a subjective prioritization of data akin to also by using chemical proteomics [23,24]. Combina-
the proverbial search for lost keys under a street lamp and tions of classical and newer purification procedures can
emphasizes the need for a null hypothesis-based approach reduce sample complexity, allowing the visualization of
in analyzing proteomic data sets. Without this, the value of lower abundance, yet potentially important, proteins.
such data and its timely utilization can be limited leading to Analysis of a subproteome from the synaptic plasma
erroneous conclusions based on information lacking phy- membrane using multiple purification steps including
siological context. An increase in the isoforms of protein SDS-PAGE, cation exchange and reverse phase HPLC
14-3-3 in Alzheimer’s disease and Down syndrome led to to reduce sample complexity prior to MS analysis,
the suggestion that this protein may play a role in neuro- identified a number of proteins involved in synapse
degenerative disease pathology [15]. However, this same physiology [25]. Combining information obtained from
protein has been implicated in tumorigenesis, multiple ESI (Electrospray injection)- and MALDI (Matrix-
sclerosis and ALS [16], highlighting the need to assess assisted laser desorption ionization)-MS approaches
such findings in a context broader than a single experi- enhanced protein identification as a result of greater
mental series ideally involving pathway analysis in differ- sequence coverage, demonstrating the utility of
ent cells and tissues. sample fractionation and purification and different MS
systems in elucidating components of the synaptic sub-
2.2. Hydrophobic transmembrane receptors proteome.
Organic solvents and novel surfactants can also enhance
While receptors, both G-protein coupled (GPCRs) and membrane protein isolation and identification by enhan-
ion channels, are the targets through which 70% of cur- cing solubility during membrane extraction and during
rently available drugs act [5,17], issues with sample pre- subsequent analysis. Chloroform/methanol increased the
paration and purification results in these proteins being identification of highly hydrophobic proteins isolated
underrepresented in many proteome arrays. Other proteins from chloroplast membranes while at the same time
integral to the membrane account for 20–30% of the excluding hydrophilic proteins [21]. Methanol extraction
genome and may represent a wealth of as yet unexploited combined with biotinylation of cysteine-containing pep-
drug targets [18,19]. However, membrane proteins are tides and streptavidin affinity chromatography also
positively charged, enhancing their interactions with nega- increased hydrophobic protein isolation and identification
tive charges on the membrane surface [20] and typically [26], increasing by 10% the number of proteins identified,
contain hydrophobic transmembrane domains which adds 50% of which were novel and included ATP-binding
to the difficulty of isolation. cassette (ABC) transporter proteins and others typically
Detergents are frequently used to solubilize membrane in low abundance. Proteomic analysis of a mouse fore-
proteins prior to 2DIGE or gel-free analysis using MS. brain N-methyl-D-aspartate receptor complex (NRC),
However, proteins produced in this manner are prone to using immunoaffinity, peptide affinity chromatography
precipitation in the first 2DIGE dimension and are lost to SDS PAGE and MS, resulted in the identification of 34
subsequent analysis [20,21]. Detergents can also affect gel- unknown proteins [27]. However, other NRC-associated
free approaches interfering with column purification and proteins were not found indicating that proteome analysis
suppressing ionization during the MS analysis. is not always consistent, emphasizing the need for the
However, newer zwitterionic detergents compatible with replication of well planned and executed studies. Addi-
2DIGE improved the solubilization and analysis of two tional complications of data generated from complex brain
membrane proteins, the human histamine H2 receptor proteomes come from studies of mouse [28] and human
(hH2R) and the rat P2X3 receptor (rP2X3), a GPCR and fetal [29] brain proteomes where 8500 (mouse) and 1700
ATP-gated ion channel, respectively, as assessed by immu- (human) proteins were identified. In the mouse proteome,
noblotting [22]. Neither target could be visualized using MS only identified 6% (500) of these. In the human brain
commonly used protein stains, even after increasing pro- proteome, the 1700 proteins were identified as products of
tein load on the gel. MS identified only a single peptide 437 genes [29]. In a rat brain proteome, 190 proteins were
from the P2X3 channel and failed to identify the GPCR identified, 30 of which had only been predicted from
highlighting the need for technological advances to ana- nucleic acid sequences [30]. The latter study proved the
lyze membrane subproteomes. existence of these proteins in brain; however the immedi-
ate utility of the 30 proteins in target identification beyond
2.3. Sample complexity annotation of the genome was unclear. While generating
considerable data, the utility of these brain based studies in
To address the challenges present by the multitude terms of alternate target discovery has yet to be estab-
of proteins present in a proteome and of isolating lished.
1136 K.K. Kopec et al. / Biochemical Pharmacology 69 (2005) 1133–1139

Protein-protein interaction analysis and signal pathway were interrogated with two affinity matrices containing
mapping also fall under the rubric of proteomics with yeast probes based on analogs of E7070, one active and the other
two-hybrid and gal pull down techniques adding informa- inactive. Many high abundance proteins, e.g., tubulin and
tion on binary protein and receptor-scaffold interactions chaperones, were present in the 285 proteins identified by
[27,31,32]. LC-MS/MS from both the active and inactive compound
resins making it difficult to identify proteins specific for the
2.4. Chemical proteomics active analog of E7070. By using fluorescent two-dimen-
sional differential in-gel electrophoresis (2D-DIGE) fol-
This is a technique that uses small, drug-like molecules lowing affinity purification and labeling of the proteins
either tethered to a resin [23,24] or exposed to protein eluted from the active and inactive resins with Cy5 and Cy3
chips [33]. Proteins binding the ligand are then viewed as fluorescent dyes, respectively, four proteins were visua-
potential drug targets. In addition to providing insights lized by 2D-DIGE gel. Three of these were also observed
into the selectivity and mechanism of action of a com- using a cleavable isotope coded affinity tag reagent (ICAT)
pound, chemical proteomics can also identify previously approach in combination with multidimensional LC-MS/
unknown protein targets for a compound that may aid in MS. This latter technique, while reducing sample complex-
the identification of potential side effects. Knowledge of ity, actually resulted in more peptides being identified by
the co-crystal structure of the compound and a target MS than were originally found after affinity chromatogra-
protein can significantly enhance the outcome of such phy with E7070. These results were also confirmed by
studies [34] but can also impose a priori, rather than null, DNA microarray studies demonstrating both the value of
criteria for subsequent proteome analysis. If information reducing sample complexity—even a subproteome—and
on the target protein is lacking or an in vitro bioassay for of using alternative approaches to provide context to
the tethered or biotinylated compound is unavailable, the proteomic findings.
optimization of affinity chromatography probes and the Sample complexity can also be reduced by combining
downstream validation of identified proteins becomes a chromatography and MS in tandem with the protein chip
significant challenge. Chemical proteomics was used to technology used in SELDI-MS (surface-enhanced laser
examine the selectivity of the mitogen-activated protein desorption ionization MS) [35]. Chips are available that
kinase (MAP) p38 inhibitor, SB 203580 [34] which is have surfaces with ion exchange, metal affinity, reverse and
widely used as a pharmacological tool to examine the role normal phase chromatographic properties and covalently
of p38 in inflammation and other disease states. Using an bound proteins for affinity purification, e.g., antibodies,
active analog of SB 203580 as an affinity probe, 12 streptavidin, and protein A. SELDI-MS has been used to
kinases including p38 were identified in HeLa cells where generate differential protein maps from CSF proteomes
SB 203580 had comparable IC50 values. SB 203580 was from inflammatory and neuropathic pain models and has
actually a more potent inhibitor of the ominously named been used to assess the effects of the COX-2 inhibitor,
RICK (Rip-like interacting caspase-like apoptosis-regu- nimesulide, on these profiles [36]. Chemical proteomics
latory protein (CLARP) kinase/Rip2/CARDIAK) kinase has also been used to assess drug effects on the cellular
than p38 kinase (IC50 = 16 nM versus 38 nM) and also proteome using a high throughput microscopy-based tech-
inhibited two other kinases, GAK (cyclin-G associated nique that may also have the potential to identify novel
kinase; IC50  135 nM) and CK1a (IC50  124 nM). As drug targets [37].
a result of this analysis, many studies that used SB
203580 as a selective p38 kinase inhibitor will require
re-evaluation. This study involved two high resolution 3. Technical challenges and developments
steps with affinity chromatography followed by 2DIGE to
reduce sample complexity, in this instance the HeLa cell As the technologies applied to proteome analysis evolve,
kinase subproteome. While demonstrating the potential there are tactical aspects of proteome experimentation that
of proteomics when applied to a specific question, this need to be addressed to enhance progress. Subfractionation
study was somewhat reductionistic in that it involved of a proteome requires that the data derived from the
only a single cell line. It was also in marked contrast to de subproteome is reassembled into the context of the parent
novo chemical proteomics studies that may involve the proteome using some form of bioinformatics algorithm.
analysis of many thousands of cellular proteins from This process addresses several aspects of the proteome
more complex proteomes where knowledge of target experiment: (i) replication; (ii) data integration; and (iii)
structure is totally absent. target validation.
Another example of this issue is the identification of
maleate dehydrogenase as a primary target for the novel 3.1. Replication
anticancer agent, E7070 involved both proteomics and
transcription profiling strategies with subsequent target As in all branches of science, replication of a finding,
validation [24]. Binding proteins enriched from HCT cells initially in the laboratory making the finding and, sub-
 K.K. Kopec et al. / Biochemical Pharmacology 69 (2005) 1133–1139 1137

sequently, in independent laboratories, is a key part compound, selective for the novel disease target, is shown
of the process of building confidence in new scientific to have efficacy in its target patient population [10]. Once a
knowledge [38]. In proteomics, because of: (a) technol- target has achieved some status—as yet difficult to gen-
ogy limitations (despite some of the advances outlined erically define (there are for instance currently in excess of
above); (b) the dynamic nature of a proteome and; (c) 30 candidate genes for schizophrenia)—from a list of
the complexity of the initial sample, it is often very disease-related genes and is confirmed by proteomics
difficult to replicate data from a single proteome analysis, strategies are required that can build confidence
sample, let alone from a repeat experiment in the same in the selection of the protein from a subset of potential
laboratory. Thus replication between laboratories is a targets to more effectively bridge the gap from target
major concern with the need for ‘routine, reliable and identification to clinical trial initiation [48]. This should
efficient technologies’’ for data acquisition and analysis potentially lead to fewer clinical failures allowing the
in proteomics being a high priority [1]. Added to these anticipated potential of genomic and proteomic strategies
considerations is the actual time required for analysis. to be realized in a predictive rather than retrospective
This can take from months to, in one instance where manner.
triplicate measures were required, nearly 3 years—cer- The challenge however, remains in identifying those
tainly not a process that has immediate impact in bio- methodologies that can most effectively translate geno-
medical research. mic and proteomic information into high confidence
targets. One key element is to have sufficiently high
3.2. Data integration throughput in the technologies to make the testing of
lists of potential targets rapid, reliable, reproducible and
Because of the ‘‘infinite number of proteomes’’ [1], comparable. Many of the techniques required reflect the
databases can become swamped with information that, classical tools of biochemistry and pharmacology and
because of the issues discussed above, makes the com- include affinity chromatography, immunoblotting, immu-
parison of data within and between experiments and noprecipitation, enzymology, radioligand binding and
between researchers nearly impossible—if indeed such autoradiography to which are being added newer tech-
comparisons are being routinely attempted. The ability of nologies like surface plasmon resonance and MALDI-
genomics and proteomics to generate lists of genes and MS. These technologies can provide evidence of the
proteins has given cause for concern [1,39,40] with the interaction between compounds and the protein target
suggestion that the process would benefit significantly with additional confidence being provided by various
from the biological insight and context of more integrated, cell-based functional assays, systems-based approaches
systems-based approaches [41,42], including pharmacol- [41,42,49] and targeted gene knockouts for the target of
ogy [43] and physiology, especially when the latter are interest [50].
triaged from the clinical setting [11,44,45]. While pro- Prioritization of potential new drug targets is essential in
teomics has provided interesting findings in regard to drug discovery research where only a small number of well
initial target identification [23,33,34], the ability to mean- validated, high confidence targets can be advanced to the
ingfully consolidate this information in the broader con- clinic due to the high cost of downstream studies and high
text of the total proteome has yet to be achieved. Genomic attrition rates [51]. Ignoring for one moment the polyge-
and proteomic data can indeed be valuable in hypothesis nomic nature of the majority of human diseases, to advance
generation [39], despite the latter being described as an compounds active at the 30 candidate genes for schizo-
‘‘ignorance-based approach’’ [46], as it can lead to addi- phrenia mentioned above, none of which currently appears
tional testing and validation in appropriate disease models better than any other, to clinical validation would cost
[41]. However, it may also contribute to the larger failure between $750 and $900 MM, representing a large invest-
of proteomics in that it creates bias and a drive towards ment to find perhaps the one new drug target with an
preconceived conclusions – the absence of a null hypoth- acceptable therapeutic index that will improve treatment of
esis [47]. this disease.

3.3. Target confidence building versus validation

4. Future considerations
Transitioning from the identification of a putative drug
target using genomic and proteomic approaches [38] to its For proteomics to deliver on its ability to provide viable
validation as a bona fide drug target is a daunting chal- disease-associated targets for drug discovery, solving the
lenge, representing the next rate-limiting step in the uti- technological and experimental challenges outlined above
lization of ‘‘omics’’ technologies [Table 1]. At the is a high priority as is the establishment of a viable
preclinical level, target validation can be more accurately scientific pathway for target validation. The latter should
described as target confidence building being distinct from not continue to be confused with target confidence building
target validation, as the latter only occurs when a drug-like as expectations for the two are very different. This will
1138 K.K. Kopec et al. / Biochemical Pharmacology 69 (2005) 1133–1139

require, to paraphrase Horton [52], a broader acknowl- [22] Henningsen R, Gale BL, Straub KM, DeNagel DC. Application of
zwitterionic detergents to the solubilization of integral membrane
edgement of the inherent difficulties of proteomics in order
proteins for two-dimensional gel electrophoresis and mass spectro-
to appropriately advance the science to meet its anticipated metry. Proteomics 2002;2:1479–88.
promise rather than continuing the population of databases [23] Jeffery DA, Bogyo M. Chemical proteomics and its application to drug
(and journals) with data that are lacking in context and, as a discovery. Curr Opin Biotechnol 2003;14:87–95.
result, value. [24] Oda Y, Owa T, Sato T, Boucher B, Daniels S, Yamanaka H, et al.
Quantitative chemical proteomics for identifying candidate trug tar-
gets. Anal Chem 2003;75:2159–65.
[25] Stevens SM, Zharikova AD, Prokai L. Proteomic analysis of the
References synaptic membrane fraction isolated from rat forebrain. Brain Res
Mol Brain Res 2003;117(2):116–28.
[1] Huber LA. Is proteomics heading in the wrong direction? Nat Rev Mol [26] Goshe MB, Blonder J, Smith RD. Affinity labeling of highly hydro-
Cell Biol 2003;4:74–80. phobic integral membrane proteins for proteome-wide analysis. J
[2] Wasinger VC, Cordwell SJ, Cerpa-Poljak A, Yan JX, Gooley AA, Proteome Res 2003;2:153–61.
Wilkins MR, et al. Progress with gene-product mapping for the [27] Husi H, Ward MA, Choudhary JS, Blackstock WP, Grant SGN.
Mollicutes: Mycoplasma genitalium. Electrophoresis 1995;16(7): Proteomic analysis of NMDA receptor-adhesion protein signaling
1090–4. complexes. Nat Neurosci 2000;3(7):661–9.
[3] Southan C. Has the yo-yo stopped? An assessment of human protein- [28] Klose J, Nock C, Herrmann M, Stuhler K, Marcus K, Bluggel M, et al.
coding gene number. Proteomics 2004;4(6):1712–26. Genetic analysis of the mouse brain proteome. Nat Genet 2002;30(4):
[4] Kim SI, Voshol H, van Oostrum J, Hastings TG, Cascio M, Glucksman 385–93.
MJ. Neuroproteomics: expression profiling of the brain’s proteomes in [29] Fountoulakis M, Juranville JF, Dierssen M, Lubec G. Pro-
health and disease. Neurochem Res 2004;29(6):1317–31. teomic analysis of the fetal brain. Proteomics 2002;2(11):
[5] Drews J. Strategic trends in drug discovery. Drug Disov Today 1547–76.
2003;8(9):411–20. [30] Shin JH, Yang JW, Juranville JF, Fountoulakis M, Lubec G. Evidence
[6] Hopkins AL, Groom CR. The druggable genome. Nat Rev Drug for existence of thirty hypothetical proteins in rat brain. Proteome Sci
Discov 2003;1:727–30. 2004;2(1):1–8.
[7] Lindsay MA. Target discovery. Nat Rev Drug Discov 2003;2: [31] Legrain PJ, Wojcik J, Gauthier JM. Protein-protein interactive
831–8. maps: leads towards cellular function. Trends Genet 2001;17(6):
[8] Horrobin DF. Realism in drug discovery-could Cassandra be right? 346–52.
Nat Biotechnol 2001;19:1099–100. [32] Bockaert J, Marin P, Dumuis A, Fagni L. The ‘magic tail’ of G protein-
[9] Milne GM. Pharmaceutical productivity—the imperative for new coupled receptors: an anchorage for functional protein networks.
paradigms. Ann Rep Med Chem 2003;38:383–96. FEBS Lett 2003;546:65–72.
[10] Williams M. Target validation. Curr Opin Pharmacol 2003;3: [33] Huang J, Zhu H, Haggarty SL, Spring DR, Hwang H, Jin F, et al.
571–7. Finding new components of the target of rapamycin (TOR) signaling
[11] Petricoin EF, Liotta LA. Clinical applications of proteomics. J Nutr network through chemical genetics and proteome chips. Proc Natl
2003;133:2476S–84S. Acad Sci USA 2004;101:16594–9.
[12] Collins FS, Green ED, Guttmacher AE, Guyer MS, USNHGRI. A [34] Godl K, Wissing J, Kurtenbach A, Habenberger P, Blencke S, Gutbrod
vision for the future of genomics research. Nature 2003;422: H, et al. An efficient proteomics method to identify the cellular targets
835–47. of protein kinase inhibitors. Proc Natl Acad Sci 2003;100(26):
[13] Patterson SD. Data analysis—the Achilles heel of proteomics. Nat 15434–9.
Biotech 2003;21:221–2. [35] Merchant M, Weinberger SR. Recent advancements in surface-
[14] Ferguson PL, Smith RD. Proteome analysis by mass spectrometry. enhanced laser desorption/ionization-timeof flight-mass spectrometry.
Ann Rev Biophys Biomol Struct 2003;32:399–424. Electrophoresis 2000;21(6):1164–77.
[15] Fountoulakis M, Cairns N, Lubec G. Increased levels of 14-3-3 gamma [36] Gineste C, Ho L, Pompl P, Bianci M, Pasinetti GM. High-throughput
and epsilon proteins in brain of patients with Alzheimer’s disease and proteomics and protein biomarker discovery in an experimental model
Down syndrome. J Neural Transm Suppl 1999;57:323–35. of inflammatory hyperalgesia: effects of nimesulide. Drugs 2003;
[16] Bridges D, Moorhead GB. 14-3-3 proteins: a number of functions for a 63(Suppl. 1):23–9.
numbered protein. Sci. STKE 242: re20, 2004. [37] Perlman ZE, Slack MD, Feng Y, Mitchison TJ, Wu LF, Altschuler SJ.
[17] Wu CC, Yates III JR. The application of mass spectrometry to Multidimensional drug profiling by automated microscopy. Science
membrane proteomics. Nat Biotechnol 2003;21(3):262–7. 2004;306:1194–8.
[18] Krogh A, Larsson B, von Heijne G, Sonnhammer EL. Predicting [38] Kramer R, Cohen D. Functional genomics to new drug targets. Nat
transmembrane protein topology with a hidden Markov model: Rev Drug Discov 2004;3:965–72.
application to complete genomes. J Mol Biol 2001;305(3): [39] Firestein GS, Pisetsky DS. DNA microarrays: boundless technology or
567–80. bound by technology? Guidelines for studies using microarray tech-
[19] Wallin E, von Heijne G. Genome-wide analysis of integral membrane nology. Arthritis Rheum 2002;46:859–61.
proteins from eubacterial, archaean, and eukaryotic organisms. Protein [40] Choudhary JS, Grant SGN. Proteomics in postgenomic neuroscience:
Sci 1998;7(4):1029–38. the end of the beginning. Nat Neurosci 2004;7:440–5.
[20] Schwartz R, Ting CS, King J. Whole proteome pI values correlate with [41] Butcher EC, Berg EL, Kunkel EJ. Systems biology in drug discovery.
subcellular localization of proteins for organisms within the the three Nat Biotechnol 2004;22(10):1253–9.
domains of life. Genome Res 2001;11(5):703–9. [42] Hood L, Heath JR, Phelps ME, Lin B. Systems biology and new
[21] Ferro M, Seigneurin-Berny D, Rolland D, Chapel A, Salvi D, Garin J, technologies enable predictive and preventative medicine. Science
et al. Organic solvent extraction as a versatile procedure to identify 2004;306:640–3.
hydrophobic chloroplast membrane proteins. Electrophoresis 2000; [43] Williams M. A return to the fundamentals of drug discovery. Curr Opin
21(16):3517–26. Investigational Drugs 2004;5(1):29–33.
 K.K. Kopec et al. / Biochemical Pharmacology 69 (2005) 1133–1139 1139

[44] Petricoin EF, Zoon KC, Kohn EC, Barrett JC, Liotta LA. Clinical [48] Horrobin DF. Modern biomedical research: an internally self consis-
proteomics: translating benchside promise to bedside reality. Nat Rev tent universe with little contact with medical reality. Nat Rev Drug
Drug Discov 2002;1:683–95. Discov 2003;2:151–4.
[45] Petricoin EF, Fishman DA, Conrads TP, Veenstra TD, Liotta LA. [49] Hauptschein RS, Eustace BK, Jay DG. Global high-throughput screens
Lessons from Kitty Hawk: from feasibility to routine clinical use for for cellular function. Exp Hematol 2002;30:381–7.
the field of proteomic pattern diagnostics. Proteomics 2004;4: [50] Zambrowicz BP, Sands AT. Knockouts model the 100 best-selling
2357–60. drugs—will they model the next 100. Nat Rev Drug Discov 2003;
[46] Lazo JS, Ducruet AP, Koldamova RP. Sleuthful pharmacology. Mol 2:38–51.
Pharmacol 2003;64(2):199–201. [51] Duyk G. Attrition and translation. Science 2003;302:603–5.
[47] Kenakin T. Pharmacologic analysis of drug-receptor interaction, [52] Horton R. AIDS: the elusive vaccine. New York Rev. Books, vol. 51,
second ed. New York: Raven; 1997. p. 344–346. 2004. p. 53–8.