Drug Discovery Today: Technologies Vol. 1, No.

1 2004

Editors-in-Chief
Kelvin Lam – Pfizer, Inc., USA
Henk Timmerman – Vrije Universiteit, The Netherlands
DRUG DISCOVERY
TODAY
TECHNOLOGIES Target identification

Proteomics: how to control highly

dynamic patterns of millions of
molecules and interpret changes
correctly?
André Schrattenholz∗
ProteoSys AG, Carl-Zeiss-Str. 51, 55129 Mainz, Germany

“Proteomics” is essentially protein analysis and, until

recently, could be described as an umbrella for a set of
technology and bioinformatic platforms aimed at the
comprehensive molecular description of the actual pro-
tein complement of a given sample. Today, it is typi-
cally associated with systems biology. In this context,
powerful new technologies for differential complexity
reduction promise to solve some of the most pressing
problems in drug development. The resulting analyti-
cal challenges of unprecedented complexity are emerg-
ing as one of the last frontiers of molecular biology.
Considerable progress has been made in characterizing
rapid post-translational protein modifications in highly
complex molecular signatures as key disease-related
biomarkers from experimental model systems or clin-
ical samples.
Introduction
There is disappointment with results from pharmacoge-
nomics [1]. Why? Largely, because the effects of drugs are
manifest at the level of proteins. Roughly 30,000 human
genes are translated into several million highly dynamic pro-
∗
Corresponding author. (A. Schrattenholz) andre.schrattenholz@proteosys.com
URLs: http://www.proteosys.com.
Section Editors:
Wolfgang Fischer, Rob Hooft, and Michael Walker
Identifying a potential protein drug target within a cell is a major
challenge in modern drug discovery; techniques for screening the
proteome are, therefore, an important tool. Major difficulties for
target identification include the separation of proteins and their
detection. The latest developments in techniques that master these
challenges, such as SDS–PAGE and liquid chromatography in
combination with isotopic labeling and staining techniques, are
highlighted in this review. The authors evaluate the applicability of
these approaches for specific tasks.
tein isoforms, which moreover, can range in abundance over
8–15 orders of magnitude [2,3]. To cope with these num-
bers, rigorous complexity reduction, based on quantitative
and differential displays, is an absolute requirement. Then,
affinity-based fractionations, or depletion of specific highly
abundant proteins, in combination with appropriate pooling
strategies, are crucial to describe key molecular details.
Highly automated reproducible procedures with reason-
able throughput have been established, forming a flexible an-
alytical base for comprehensive molecular characterization of
drug effects. The most powerful protein separation technolo-
gies are two-dimensional polyacrylamide gel electrophoresis
(2D-PAGE; see Glossary) and liquid chromatography, on the
identification side there is a set of mass spectrometry meth-
ods with ever increasing accuracy and sensitivity. In both
the separation and identification of proteins, reliable quan-

1740-6749/$ © 2004 Elsevier Ltd. All rights reserved. DOI:10.1016/j.ddtec.2004.06.001 www.drugdiscoverytoday.com 1

Drug Discovery Today: Technologies Target identification
titative, and differential detection remains the major chal-
lenge.
The current situation in proteomics appears to be charac-
terized by differing opinions and strategic uncertainty in a
climate requiring imperative large investment, whereas the
mainstream strategic thinking of investors is still completely
dominated by daunting experiences from the genomics era.
Given the inherent enormous complexity and dynamics of
protein expression, current high-throughput screening (HTS)
methods bear the risk of producing poorly controlled data
morasses. We anticipate a phase of high intensity experimen-
tal efforts ahead, requiring the application of rigorous differ-
ential, quantitative and statistical procedures to provide the
necessary high quality data content (post-translational mod-
ifications of key biomarker proteins) to permit the next gen-
eration of relevant HTS processes for proteomic research in
drug discovery and development.
Key technologies for proteomics
There are two main requirements for a meaningful pro-
teomics strategy. (a) Going broad: pattern control, as com-
prehensive as possible, to exactly find the spot, where it is
worthwhile to drill deeper. (b) Going deep at the right coor-
dinate: once a suitable biomarker has been found with suffi-
cient statistical significance, it is necessary to characterize its
molecular features and interactions in more depth.
This review attempts to cover the fast developing field
of proteomics in terms of five key technologies. Separation,
detection, identification of proteins and subsequent data
mining, data management, and interpretation, would each
require separate discussions, with some of the major chal-
lenges (all worthy of particular considerations) at the inter-
faces of various platforms. This review, therefore, is focussed
on differential and quantitative pattern control in pro-
teomics. This implies coverage of three labeling techniques
(fluorescent cyanine dyes, radioactive, and stable isotopes),
the detection of labeled proteins in 2D gels and liquid chro-
matography (LC; see Glossary) and, most importantly, the
quantification of identified differential proteins. The corre-
sponding pattern analysis and interpretation will be treated as
well.
Progress in identification of proteins by mass spectrometry
has been covered recently in related reviews [4–8], and will
be mentioned here only in the context of differential pro-
tein quantification by stable isotope techniques. Briefly, the
most recent developments involve tandem mass spectrom-
eters with advanced MS/MS capabilities, like MALDI–TOF/
TOF–MS (matrix-assisted laser desorption ionization/time
of flight; Bruker, Applied Biosciences http://www.bdal.com;
http://www.appliedbiosystems.com, see Glossary) and LCQ–
FT–MS (http://www.thermofinniganmat.de). The ‘‘TOF/TOF’’
technique provides the possibility to sequence selected pep-
tides, by analyzing fragment ions generated by the dissoci-
2 www.drugdiscoverytoday.com
Vol. 1, No. 1 2004
ation of precursor ions. MALDI–TOF/TOF allows highly au-
tomated and reliable identification of thousands of proteins
per week with sensitivities around 1–20 fmol. The LCQ–FT
instrument is a hybrid MS system with high mass accuracy
and resolution (approximately 1 ppm). However, regardless
of the ever increasing sophistication of mass spectrometry,
the first step of a successful proteomics study is still the
effective separation and detection of proteins [4–8].
Resolution
There are essentially two separation technologies for complex
samples: 2D-PAGE and multidimensional LC [9] (see Glos-
sary). 2D-PAGE remains the most powerful tool for compar-
ative protein profiling studies. Modern large zoom gels can
reproducibly and reliably resolve tens of thousands of differ-
ent proteins from complex mixtures, which is superior to all
other methods [10,11]. LC methods resolve some hundreds
of peaks, at best a few thousand. A high-throughput variant
of the LC approach is provided by chromatographic surface
arrays, interfaced directly to tandem-MS (SELDI, 12; Cipher-
gen, http://www.ciphergen.com; see Glossary). This method
has a low resolution and is not very sensitive for identifica-
tion. It has been covered in recent reviews [4,12].
LC and 2D-PAGE have complementary advantages and
disadvantages, as discussed below. In terms of resolution,
gels are far superior, with patterns confined to a molecu-
lar weight range of 10–250 kDa. Chromatographic methods
have much poorer resolution, but have advantages at lower
molecular weights (1–20 kDa), and on-line-MS capability.
Moreover, 2D-PAGE has a single important advantage, which
most LC-based methods do not have, and which is rarely
discussed: a new gel is used for each experiment; they cannot
be used again, in contrast to HPLC columns or chromato-
graphic arrays. Anyone taking the pains to investigate, for
example, reversed phase or ion exchange resins will discover
that after a while they residually accumulate certain pep-
tides, gradually changing adsorptive properties. The use of a
new gel overcomes these problems.
Dynamic range and sensitivity of detection
Standard protein staining in gels (silver, colloidal Coomassie
or fluorescent dyes) has moderate sensitivities (approxi-
mately 1 fmol) and clear disadvantages in terms of the dy-
namic range of protein concentrations. The two to three or-
ders of magnitude of these procedures do not match the pro-
tein reality (8–15 orders of magnitude) and thus, a consider-
able amount of information is lost. The same is true for detec-
tion of protein profiles from LC or chromatographic arrays,
which is usually achieved directly in the mass spectrometer,
with also moderate dynamic ranges of approximately three
orders of magnitude and moderate sensitivities in the range
of 5 fmol. New methods in MS, like Fourier transform–ion
cyclotron resonance (FT–ICR; see Glossary), are potentially
Vol. 1, No. 1 2004 Drug Discovery Today: Technologies Target identification

much more sensitive, but still rather with peptide standards
than with real protein mixtures from biological samples.
These dynamic range and sensitivity problems, as well as
the reliable quantification of small differences, can be solved
using radioactive isotopes. Appropriate labeling of proteins is
an art in itself, but eventually leads to detection sensitivities
at attomol levels, with a dynamic range of over six orders of
magnitude and statistically significant quantification even of
changes in the 15–20% range [13]. In our opinion there is no
alternative to using a small fraction of a sample for radioac-

Figure 1. Quantitative differential pattern control of neuronal differentiation with undifferentiated murine stem cells, early and late neurons. (a)
Approximately 1␮g protein from three different stages of neuronal differentiation of murine embryonic stem cells (S1–S3) were jointly analysed in one 2D gel,
after separate labelling with three different radioactive iodine isotopes (123 I, 125 I and 131 I). The upper right panel shows a surrogate composite color image, the
lower right panel gives an impression of dynamic range; on the left the three isotopic images obtained from this single gel are shown. The actual quantification
is performed using the radioactive data set (matrix with spot coordinates and quantitative values for each isotope, not shown here) and enables an extremely
reliable and exact quantification and differential detection of even subtle changes of protein expression. (b) This type of analysis of expression levels of
differential proteins is shown for 20 proteins. Standard statistical procedures (ANOVA) ensure the significance of respective quantitative differences: the blue
bars are associated with abundance of the particular protein in the early S1 stage, purple bars with intermediate S2 and green bars with the late S3 stage.

www.drugdiscoverytoday.com 3
Drug Discovery Today: Technologies Target identification
tive pattern analysis, in particular because quantification of
even subtle protein changes in kinetic experiments is the ba-
sis for the sequence of events in subsequent biological inter-
pretations.
Differential display
Complex protein patterns of related samples, or even of mul-
tiple aliquots of the same sample, analyzed in different gels
(or LC runs) are variable, that is matching of independent ex-
periments is a non-trivial problem. This systematic error can
only be avoided by analyzing several samples in one experi-
ment, for example in one gel.
Such a real differential display is the most important
strategy to reduce intrinsic protein expression complex-
ity reliably, and there are presently only three existing
methods to achieve this: labeling proteins with fluorescent
cyanine dyes (DIGE, Amersham Biosciences, http://www.
amershambiosciences.com; see Glossary), labeling with ra-
dioactive isotopes (ProteoSys, http://www.proteosys.com) or
labeling with stable isotopes (e.g. ICAT [isotope-coded affin-
ity tag], Applied Biosystems, http://www.appliedbiosystems.
com; see Glossary).
All three can be used with 2D gels [3,14], for dyes and ra-
dioactive labeling the differential becomes visible on the level
of the stained gel (subsequent MS is for identification only).
An example of 2D-PAGE of radio labeled proteins is shown
in Fig. 1, the dynamic quantitative range, resolution and dif-
ferential capability is able to resolve three samples in one gel
(ProteoSys). Every difference depicted as a slight change in
color hue is real, the relative quantifications in Fig. 1b are sta-
tistically significant even if they are subtle, and the overall
dynamic range is more than six orders of magnitude.
Stable isotope quantification is detectable at the level of
mass spectrometry (after protein separation by 2D-PAGE or
LC) and integrates identification and quantification, with fur-
ther potential connotations for estimating protein turnover.
As a result of this integrative property, stable isotope markers
have attracted a lot of attention. They can be introduced into
proteins in two ways, either by metabolic labeling, feeding
cellular systems or organisms with 13 C-glucose or 15 N-amino
acids or post-experimentally by covalent attachment to cys-
teine residues by, for example ICAT reagents (Applied Biosys-
tems; http://www.appliedbiosystems.com/biobeat/icat/) or
deuterated acrylamide [8,14]. Metabolically incorporated 13 C
or 15 N provide huge information content, because depend-
ing on the nature of the precursors the incorporation is not
arbitrary but follows specific metabolic routes. As shown in
Fig. 2, a very reliable differential quantification of two sam-
ples can be obtained for proteins excised from 2D gels in
the range of 10 fmol, even at levels of incomplete labeling
(e.g. a minimum 15 N enrichment of the total amino acid
pool of approximately 4% is required). Feeding a cell culture
system with a selection of 15 N-labeled Ala, Glu, Asp, Asn,
4 www.drugdiscoverytoday.com
Vol. 1, No. 1 2004
Glossary
2D-PAGE: two-dimensional polyacrylamide gel electrophoresis:
method with highest resolution for separation of protein mixtures,
combines 1. separation according to isoelectric points (IEF = isoelectric
focussing), with second according to molecular mass (SDS–PAGE).
DIGE: two-dimensional difference gel electrophoresis staining with
fluorescent cyanine dyes with narrow excitation and emission bands
for multicolor analysis.
FT-ICR: fourier transform ion cyclotron resonance, extremely
sensitive and accurate mass spectrometry method, under development
for broad use in proteomics.
ICAT: isotope-coded affinity tag, chemical reagents with stable
isotopes, like deuterium, usually for alkylation of cysteine residues.
MALDI–TOF: matrix-assisted-laser-desorption-ionisation,
time-of-flight, most widely used and integrated, highly automated mass
spectrometry for protein identification.
Multidimensional liquid chromatography: combines several
separation principles in one experiment, like ion exchange, reversed
phase, among others, some powerful monolith columns under
development.
Post-translational Modifications: enzymatic chemical
modifications of polypeptides after translation, there some hundred
different types on amino acid side chains known, moreover, functional
proteolysis is increasingly recognized as a very important mechanism.
SELDI: surface enhanced laser desorption ionisation: mass
spectrometry method used together with chromatographic arrays.
Western blots: electrophoretic transfer of proteins from a gel to a
membrane (nitrocellulose or polyvinylidene fluoride) for specific
staining procedures., for example by antibodies.
Gly, Ser and Gln (in three-letter amino acid code) labeled
>99% proteins with 15 N. Some percentage of the isotope
label is redistributed to other amino acids and suggests for
all proteins the existence of a uniform amino acid precusor
pool with at least 12 labeled amino acids, providing deep
insight into the metabolic pathways of the analyzed cells
[14,15].
Regarding the LC–MS approach, however, which due to
its high-throughput is very popular, one should realize that
next to resolution and sensitivity limitations, the experi-
mental window confined to cysteine-containing peptides of
tryptic (or other) fragmentations is very narrow. As shown in
Fig. 3, a huge amount of information inherent to the sample
(as compared to 2D gels) can be lost, because the fraction of
cysteine-containing-peptides will not cover many phospho-
rylated, glycosylated or proteolytically truncated peptides,
which nevertheless might display crucial biomarker prop-
erties. Therefore, ICAT-like post-harvest labeling, together
with an on-line LC–MS approach, would detect no func-
tional changes, although differential phosphorylation with
distinct biomarker qualities could be occurring to a very
large extent. Some examples for these circumstances are in-
cluded in Fig. 2. This serious problem of false negatives using
LC-based multidimensional protein identification technolo-
gies [16] has been covered in a recent review [4]. In general
LC–MS approaches are better with less complex samples,
for example after extensive fractionation or from bacterial
Vol. 1, No. 1 2004 Drug Discovery Today: Technologies Target identification

Figure 2. Differential quantification of stable isotope labeled peptides

using MALDI-TOF. Isotopic distributions of a selected peptide.
MALDI-TOF mass spectrum of a tryptic peptide obtained from 2D gel
electrophoresis of proteins from D3 embryonic stem cells derived from
129/sv mice. Signal intensity is plotted in arbitrary units (au) against
mass/charge (m/z) ratio. The isotope pattern of the peptide with natural
isotopic composition (N) is shown at the bottom and of peptides from
the 15 N-labeled protein (L) at the top. Peptides from calibration
mixtures with mixing ratios L/N of 0.139, 0.496, 1.991 and 7.290 are
shown in between. The method provides a quick and cheap approach for
differential quantification, if the context of the experiment is clear or
even the proteins are already known. The dynamic range is limited to
MALDI-TOF, but in contrast to other approaches protein turnover and
metabolic processes can become visible.

proteomes (these are essentially without post-translational

modifications).

Pattern analysis and interpretation

A rigorous pattern control is prerequisite for pooling strate-
gies. Extensive fractionation requires ever increasing sample
amounts; for example, phosphoproteomes constitute about
10–20% of corresponding total proteomes [17], whereas
some membrane fractions comprise less than 1%. Clinical
samples often are not available in endless quantities, laser
capture micro dissection results in a few 100 ␮g of sample,
with radioactive methods being the only option for anal-
ysis, even if pooling is possible. Even for serum samples
amounts are limited, moreover, pooling not only increases
amounts but also averages individual variations and reveals
the more general molecular features: in proteomics there is
actually only one problem with pooling: how to sort out
highly abundant single case contaminations from the pool
(low abundant contaminations are diluted). This can be
achieved by differential displays from the total pools (e.g.
each 30 patients) and a set of subpools (each group 5 × 6
patients). In this case a minimum of six differential displays
can unambiguously identify the general serum biomark-
ers and discard single fates only apparent in one of the
subpools.
Proteomic analysis reaches the most interesting stage
once mass spectrometry has delivered a final protein list. As
Figure 3. The importance of pattern resolution. (a) Theoretical
example of a peptide with phosphorylation and glycosylation sites.
Assuming cumulative addition of post-translational fractionation, at least
12 molecules, easily resolved on a 2D gel can be expected. However,
tryptic digestion would produce only two cysteine-containing peptides;
this illustrates the risk of false negatives. (b) A practical example. A list of
peptides resulting from a tryptic digest for NCBI [National Center for
Biotechnology Information, http://www.ncbi.nlm.nih.gov/] entry
gi|18765733 synaptosomal-associated protein 25 isoform SNAP25A
(Homo sapiens). Cysteine-containing peptides are marked in red; only
peptides with an m/z value between 700 and 4000 Da are listed,
assuming singly charged ions and a maximum of one missed cleavage.
Only three peptides covering less than 10% of the sequence would be
detected by LC–MS of cysteine-containing peptides.
shown in Fig. 1, the quantification of differential proteins
is absolutely essential, because is can give clues about the
sequence of events (e.g. protein 1 in Fig. 1b represents an
early event in the stem cell differentiation, because it is pre-
dominantly expressed in the S1 stage, as compared to a late

www.drugdiscoverytoday.com 5
Drug Discovery Today: Technologies Target identification
one, like protein 4, which is nearly absent in the S1 stage
and about twice as abundant in the latest S3 stage as in the
intermediate S2 stage). Data management and bioinformat-
ics should be able to automatically generate text files with
relevant Medline entries to each of the accession numbers.
Based on quantification, iterative algorithms are available to
correlate al these different data sets (biological from sample,
biophysical from separation, mass spectra, and literature),
with the aim to generate hypotheses. At this point, tran-
scriptomic and genomic data should be integrated [18].
The fast and high-throughput identification of proteins by
MALDI only uses a fraction of the information available
in the spectra (unmodified peptides), and hypotheses can
be falsified or verified using the whole detected peptide
pool of a given data set, by educated sequencing of selected
peptides.
This usually is also the stage where, guided by initial pro-
tein lists, more specific affinity tools, like antibodies (for 2D
western blots or enrichment, see Glossary) or immobilized
ligands for certain targets (‘‘Chemical proteomics’’) can be
used to address more focussed questions, relating to im-
mediate functional interactions. Chemical proteomics has
the potential to directly identifying targets, and some com-
panies offer related technologies and platforms (Serenex,
http://www.serenex.com/; Cellzome, http://www.cellzome.
com/; and ProteoSys http://www.proteosys.com/). It remains
to be seen whether these approaches can be translated di-
rectly to relevant HTS systems in drug discovery or if they will
still need to be individually adjusted in specifically designed
experimental paradigms.
Comparison of technologies
The technologies compared here include 2D-PAGE with cya-
nine dyes (DIGE), radioactively labeled proteins and stable
isotope-tagged proteins, on the one hand, and LC-based
multidimensional protein separation technologies, with on-
line MS or SELDI, on the other hand. All these methods have
advantages and disadvantages and should be regarded as
complementary. 2D-PAGE is absolutely superior for pattern
analysis of complex samples; a comprehensive control of 2D
patterns requires real differential display strategies, which
are not available with standard staining procedures. The
dynamic range and sensitivities of DIGE and stable isotope
approaches with 2D gels are moderate (approximately three
orders of magnitude, >10 fmol per spot); radioactive displays
are far superior for differential detection and in particular
quantification (approximately eight orders of magnitude,
>10 amol per spot), but are more laborious to employ. Gen-
erally, 2D gels are said to be difficult to use with certain
proteins, like membrane proteins and basic proteins, but
for these subproteomes there are always excellent alterna-
tives, for example, 16-BAC 2D-PAGE [this method avoids
chaotropic reagents and isoelectric focussing for the first
6 www.drugdiscoverytoday.com
Vol. 1, No. 1 2004
Related articles
Zolg, J.W. and Langen, H., (2004) How industry is approaching the
search for new diagnostic markers and biomarkers, Mol. Cell.
Proteomics. 28 January [Epub ahead of print]
Frank, R. Hargreaves, R. (2003) Clinical biomarkers in drug discovery
and development. Nat. Rev. Drug Discov. 7, 566–580
indsay, M.A. (2003) Target discovery. Nat. Rev. Drug Discov. 10, 831–838
Boguski, M.S. McIntosh, M.W. (2003) Biomedical informatics for
proteomics. Nature 422, 233–237
Nebert, D.W. (2003) Pharmacogenomics and “individualized drug
therapy”: high expectations and disappointing achievements. Am. J.
Pharmacogenomics 3, 361–370
dimension, but instead uses two different ionic detergents,
16-BAC, benzyldimethyl-n-hexadecylammonium chloride,
and SDS], or 1D-MS/MS for membrane proteins [19,20].
Compared to LC-based methods 2D-PAGE is slower and
more laborious (despite considerable progress in automa-
tion), but it does provide superior information. In principle,
2D-PAGE can analyze 70–80% of the proteome of a given
sample and, moreover, the residual proteome can be salvaged
and treated with alternative approaches if appropriate. Mul-
tidimensional LC–MS methods have high-throughput and
clear advantages for small proteins and peptides (1–20 kDa),
but lose a lot of relevant information because of their con-
finement to cysteine-containing peptides (in a total tryptic
digest of human peptides only 3% of the peptides con-
tain cysteine; Fig. 3), moreover, the inherent complexity of
samples might remain invisible because of limited pattern
resolution. SELDI, with chromatographic arrays, has similar
pros and cons. Resolution is inferior; in typical examples
only a few dozen peaks were detected [21,22]. Differential
analysis is also essentially static matching the resolution
of multiple experiments. Nevertheless, even on this level
(a few percent of a given proteome) interesting results can
be found.
Conclusions
Protein biomarkers enable the stratification of patient popu-
lations and promote the quantification of new drug effects
in terms of molecular targets, altered pathophysiological
Links
r Human Proteome Organization: http://www.hupo.org
r Institute for Systems Biology: http://www.systemsbiology.org/
r Swiss Institute of Bioinformatics: http://us.expasy.org/
r Nature, Supplement Proteomics:
http://www.nature.com/nature/insights/6928.html
r National Center for Biotechnology Information,
http://www.ncbi.nlm.nih.gov/
 Vol. 1, No. 1 2004 Drug Discovery Today: Technologies Target identification

Table 1. Comparison summary table

Technology 1 Technology 2 Technology 3 Technology 4 Technology 5
Name of specific type 2D-PAGEa 2D-PAGE 2D-PAGE + Cy dyesd mdLCe + s-isotopes LC arrays + SELDIf
of technology + r-isotopesb + s-isotopesc
Names of specific ProteoTope, Various suppliers and DIGEg , Amersham ICATh , Applied SELDI Ciphergen
technologies with ProteoSys companies Biosciences Biosystems
associated
companies and
company websites
Pros Resolution Resolution Resolution High throughput, low Very high throughput,
+ superior dynamic + turnover and no systematic error molecular weight low molecular weight
range and differential metabolic info; proteins proteins
quantification; no no systematic error
systematic error
Cons Medium throughput Medium throughput, Medium throughput, Systematic error, false Systematic error, false
+ radioactivity limited dynamic range limited dynamic range negatives, limited negatives, very limited
dynamic range dynamic range
References [13] [4,14] [23] [8,16] [21,22]
a
Two-dimensional polyacrylamide gel electrophoresis.
b
Radioactive isotopes.
c
Stable isotopes.
d
Cyanine dyes.
e
Multidimensional liquid chromatography.
f
Surface enhanced laser desorption ionisation.
g
Two-dimensional difference gel electrophoresis.
h
Isotope-coded affinity tag.

mechanisms and drug benefit in primary prevention or

disease-modification studies. Each one of the technological
platforms discussed here is able to deliver new biomarkers
at various levels. If comprehensive molecular overviews in
discovery projects with complex samples are required, the
start should be with 2D gels, followed by a tightly controlled
reduction of sample complexity by a suite of fractionations.
If the sample has reduced complexity (e.g. after chemical
proteomics) one-dimensional MS/MS and multidimensional
LC–MS/MS approaches are methods of choice. For fast, albeit
somewhat superficial, screening of large numbers of samples,
SELDI is suitable. The questions are always: ‘‘How precious
are the samples, which amounts are available, and can these
technologies be applied in a complementary way?’’ A com-
bined ultimate strategy could be to first achieve sufficient
control of a pattern by radioactive high-resolution 2D gels
(which can amount to near total protein display together
with reliable quantification of a maximum of differential
proteins), followed by an affinity fractionation focussed
on a few specific proteins identified in the first step. This
would generate ideal substrates for high-throughput LC–MS
or other advanced Ms methods. Currently, the widespread
exclusive use of LC–MS with precious (e.g. clinical) samples
appears to be poorly considered, always risking the loss of
considerable amounts of valuable information. A summary
of advantages and disadvantages of the different methods is
given in Table 1.
Outstanding issues
r Further education. The complexity of proteins has a completely
different quality as compared to genomics. Drug development faces
a reality of fast, redundant and pleiotropic protein interactions. Many
previous (HTS) strategies based on one-drug-one-target concepts
are probably obsolete; systems biology reveals functional protein
signatures.
r Definition and refinement of international standards for sample
preparation and storage, pooling strategies, statistics and all other
aspects of proteomic analyses. This is an ongoing worldwide effort,
which in the biomedical field is coordinated by the Human
Proteome Organization (HUPO, http://www.hupo.org).
r Further refinement of bioinformatics. Currently available software
still is not sufficiently integrating and exploiting the information and
data from the different layers of proteomic studies. Essentially new
algorithms have to be developed to track and understand recurrent
molecular events in dynamic protein signatures. In particular the
increasing sensitivity and accuracy of mass spectrometry offers
exiting new perspectives of data mining.
References
1 Nebert, D.W. et al. (2003) Pharmacogenomics and ‘‘individualized drug
therapy’’: high expectations and disappointing achievements. Am. J.
Pharmacogenomics 3, 361–370
2 Phizicky, E. et al. (2003) Protein analysis on a proteomic scale. Nature
422, 208–215
3 Hanash, S. et al. (2003) Disease proteomics. Nature 422, 226–232
4 Zolg, J.W. and Langen, H., (2004) How industry is approaching the
search for new diagnostic markers and biomarkers. Mol. Cell. Proteomics.
28 January [Epub ahead of print]

www.drugdiscoverytoday.com 7
 Drug Discovery Today: Technologies Target identification
5 Tyers, M. and Mann, M. (2003) From genomics to proteomics. Nature
422, 193–197
6 Fusaro, V.A. and Stone, J.H. (2003) Mass spectrometry-based proteomics
and analyses of serum: a primer for the clinical investigator. Clin.
Exp. Rheumatol. 21, S3–14
7 Purcell, A.W. and Gorman, J.J., (2004) Immunoproteomics: Mass
spectrometry based methods to study the targets of the immune
response. J. Proteome. Res. 3(3), 572–581.
8 Aebersold, R. and Mann, M. (2003) Mass spectrometry-based proteomics.
Nature 422, 198–207
9 Wang, H. and Hanash, S. (2003) Multi-dimensional liquid phase based
separations in proteomics. J. Chromatogr. B Analyt. Technol. Biomed. Life
Sci. 787, 11–18
10 Poland, J. et al. (2003) Isoelectric focusing in long immobilized pH
gradient gels to improve protein separation in proteomic analysis.
Electrophoresis 24, 1271–1275
11 Ravichandran, V. et al. (2004) Ongoing development of two-dimensional
polyacrylamide gel electrophoresis data standards. Electrophoresis 25,
297–308
12 Tang, N.P. et al. (2004) Current developments in SELDI affinity
technology. Mass Spectrom. Rev. 23, 34–44
13 Cahill, M.A. et al. (2003) Analysis of relative isotopologue abundances for
quantitative profiling of complex protein mixtures labelled with the
acrylamide/D3-acrylamide alkylation tag system. Rapid Commun. Mass
Spectrom. 17, 1283–1290
14 Vogt, J.A. et al. (2003) Protein abundance quantification in embryonic
stem cells using incomplete metabolic labelling with 15 N amino acids,
8 www.drugdiscoverytoday.com
Vol. 1, No. 1 2004
matrix-assisted laser desorption/ionisation time-of-flight mass
spectrometry, and analysis of relative isotopologue abundances of
peptides. Rapid Commun Mass. Spectrom. 17, 1273–1282
15 Vogt, J. et al. (2003) Quantitative protein expression analysis and
determination of amino acid precursor pool enrichment in stem cells
performed with stable isotope amino acid in vivo labelling and
MALDI-TOF-MS. Mol. Cell. Proteomics 2 (9), 795
16 Washburn, M.P. et al. (2001) Large-scale analysis of the yeast proteome by
multidimensional protein identification technology. Nat. Biotechnol. 19,
242–247
17 Sommer et al. Molecular analysis of homocysteic acid-induced neuronal
stress. J. Proteome Res. 3, [in press]
18 Boguski, M.S. and McIntosh, M.W. (2003) Biomedical informatics for
proteomics. Nature 42, 233–237
19 Marvin, L. et al. (2000) Identification of proteins from one-dimensional
sodium dodecyl sulfate-polyacrylamide gel electrophoresis using
electrospray quadrupole-time-of-flight tandem mass spectrometry. Rapid
Commun. Mass Spectrom. 14, 1287–1292
20 Hartinger, J. et al. (1996) 16-BAC/SDS–PAGE: a two-dimensional gel
electrophoresis system suitable for the separation of integral membrane
proteins. Anal. Biochem. 240, 126–133
21 Petricoin, E.F. et al. (2002) Use of proteomic patterns in serum to identify
ovarian cancer. Lancet 359, 572–577
22 Xu, Q. Lam, K.S. (2003) Protein and chemical microarrays-powerful tools
for proteomics. J. Biomed. Biotechnol. 5, 257–266
23 Somiari, R.I. et al. (2003) High-throughput proteomic analysis of human
infiltrating ductal carcinoma of the breast. Proteomics 10, 1863–1873