Application of MALDI-TOF Mass Spectrometry

During the last years, mass spectrometry has revolutionised protein biochemistry and has advanced to a
superior tool for the identification and detailed analysis of peptides and proteins. The high throughput allowed by some
mass spectrometry platforms has enabled the important step from analysis of individual proteins to proteomics.
Recently, an additional field of mass spectrometry applications has emerged - namely screening and diagnostic research.
In contrast to protein identification, screening applications have to detect analyte molecules of defined molecular weights
which can be calculated beforehand, for example by means of chemical structures. Here, the accuracy and sensitivity of
mass spectrometry has to be combined with the requirements of high-throughput analyses, in particular speed and
automation. These criteria are especially fulfilled by state of the art matrix-assisted laser desorption/ionisation time-offlight
mass spectrometry (MALDI-TOF MS) instruments. The first high throughput screening (HTS) application proved
to be genotyping of single nucleotide polymorphisms. The same principle was later applied for several quality control
issues, for example for oligonucleotides, peptide or compound libraries. This development has culminated in the screening
and profiling of complex biomarker patterns in clinical proteomics to detect a molecular fingerprint for specific diseases
in biological samples. Thus, mass spectrometry based methods are expected to enable a very early diagnosis of diseases
with minimally invasive methods of investigation. This type of high end screening application has the potential to
revolutionise the early diagnosis of many diseases.
Here, we give an overview of the application of mass spectrometry in the fields of screening and diagnostic research.
Key Words: Serum profiling, biomarkers, SNP genotyping, quality control, screening, diagnosis, MALDI-TOF.

INTRODUCTION

In recent years hardly any other technology has revolutionised
the life sciences as much as mass spectrometry. This
development was enabled primarily by the invention of the
gentle ionisation techniques of matrix-assisted laser
desorption/ionisation (MALDI) [1-3] and electrospray
ionisation (ESI) [4].
The success story was catalysed by features that are
unique to mass spectrometry (MS) - at least in this
combination. In contrast to many other techniques, MS is not
dependent on an indirect detection of analytes, for example
by using fluorescent dyes, reporter enzymes or radioactive
labels. MS detects the analyte molecules by means of an
intrinsic physical property, namely their molecular weights.
This direct detection makes a considerable part of the charm
of this technique. This advantage becomes even more
prominent when MS is applied as a supportive diagnostic
tool in clinical applications. Classical ELISA tests are
dependent on antibody detection. Accordingly, an individual
marker molecule must have been identified beforehand and
specific monoclonal antibodies must be (commercially)
available. In addition, antibody specificity is an issue leading
to a significant reduction of analysis security. Using MS, for
example, as a complementary tool in profiling approaches,
there is neither a need for the identification of individual
*Address correspondence to this author at the Fahrenheitstraße 4, D-28359
Bremen, Germany; E-mail: WPU@bdal.de
biomarkers nor for the availability of antibodies. In this case,
a complex signal pattern is used for the classification of
samples and the molecular masses of the markers are directly
identified.
Some MS techniques like MALDI-TOF MS are very
suitable for a high sample throughput up to several 10, 000
samples per day and instrument [5]. They can be applied in
an unattended 24/7 operation. And the instruments can be
well integrated into existing robotic lines as microtitre plate
(MTP) formats, the standard for liquid handling in the
laboratory, are supported.
Mass spectrometry enables high-resolution analyses of
the samples. Accordingly, when searching for analytes
of defined molecular weights, multiple analyte molecules
can be detected simultaneously in a multiplexed sample
preparation.
All these advantages make MS a superior tool for in
depth biomolecule analysis in a research setting. Additionally,
they suggest the use of MS instruments as workhorses
in screening and diagnostic research applications. The use of
MS is not limited to protein and peptide analysis. The
technique is currently used for proteins [6], peptides [7],
nucleic acids [8], sugars [9-12] and small molecules [13]
thus generating a real functional genomics platform.
With this development MALDI-TOF MS recently
entered a “terra incognita” – the medical sciences. If current
promises can be kept, mass spectrometry should be able to
2578 Current Pharmaceutical Design, 2005, Vol. 11, No. 20 Pusch and Kostrzewa
revolutionise medical sciences and clinical diagnostics in a
similar way as it has revolutionised the life sciences some
years ago.
MALDI-TOF AS MS PLATFORM OF CHOICE FOR
SCREENING AND DIAGNOSTIC RESEARCH
Looking at the mass spectrometry equipment from different
vendors, a very large variety of instruments is offered,
including different ionisation techniques and different mass
analysers.
Moreover, the degree of automation and support of
robotics has to be considered as well as sample preparation
and data analysis issues.
In addition to single instruments several dedicated
solutions exist for specialized purposes. For example, wellestablished
solutions exist in the field of analysis of singlenucleotide
polymorphisms (SNPs). Here, for example,
Sequenom (San Diego, CA, USA) offers a system suitable
for industrial scale throughput (Mass Array), whereas Bruker
Daltonik (Bremen, Germany) offers a similar system for the
low-to-medium throughput (GenoLink). Moreover, dedicated
systems for clinical proteomics and biomarker analysis
are commercially available from Ciphergen (ProteinChip
System, Ciphergen Biosystem, Fremont, CA, USA) and
Bruker Daltonik (ClinProt System, Bruker Daltonik,
Bremen, Germany).
All commercially available dedicated screening systems
are currently based on MALDI-TOF technology. And
indeed, several properties make MALDI-TOF MS a superior
analytical tool for various screening and diagnostic
applications, like SNP genotyping, oligonucleotide quality
control or biomarker screening. First of all, as an intrinsic
property of the analyte molecules is measured – namely their
molecular weights – there is neither a need for indirect
detection methods, like hybridisation, labelling with
fluorescent dyes, reporter enzymes or radioactive isotopes in
the case of nucleic acid detection, nor for antibody-based
detection in the case of profiling applications in clinical
proteomics. The technique is very fast, as the acquisition of
an individual spectrum takes only a few seconds, and it can
be highly automated allowing even an unattended 24/7
operation when using high throughput instruments (Fig. (1)).
Accordingly, this review article is focused on the MALDITOF
MS technology. However, some applications of other
MS techniques in screening will also be addressed.

SNP GENOTYPING

The analysis of single nucleotide polymorphisms was one
of the first screening applications, which have been
successfully adapted to MALDI-TOF mass spectrometers. In
particular the speed of MALDI measurements and its
automation capabilities were main drivers to use this highthroughput
technology to substitute standard gel-based
methods. Nowadays, while genotyping is entering the fields
of pharmacogenetic risk prediction [14-16], forensics [17,
18] and clinical diagnostics [19-22], the accuracy of mass
spectrometry for the direct and label-free analysis of the
molecular products of genotyping reactions is getting
increasing importance.
Sample Preparation
Because MALDI measurement of DNA typically has to
be combined with molecular amplification reactions and
sample conditioning, complex sample preparation protocols
are necessary before mass spectrometric analysis. DNA as a
polyphosphate contains numerous negative charges. The
large polyanion has the tendency to bind positively charged
ions, e.g. alkali ions, and this characteristic increases
dramatically with the number of DNA building blocks
(nucleotides). Buffers used for molecular biological
reactions typically contain high concentrations of salts, e.g.
of magnesium and potassium. The adducts which these metal
ions build with nucleic acid molecules shift the molecular
Fig. (1). The benefits of MALDI-TOF mass spectrometry as a platform for screening and diagnostic research applications.
Application of MALDI-TOF Mass Spectrometry in Screening Current Pharmaceutical Design, 2005, Vol. 11, No. 20 2579
masses and thereby lead to peak broadening and a decrease
of signal intensity in the mass spectra. In addition, further
components of the reaction buffers are incompatible with the
MALDI process, e.g. detergents that prohibit proper
crystallisation of the matrix on the sample target. Therefore,
MALDI analysis of DNA requires rigid purification of the
samples prior to measurement. On the other hand, to meet
the requirements of a high-throughput screening application,
sample preparation methods for MALDI analysis do not only
have to result in highly purified analyte molecules but also,
in addition, have to be cost-effective and suitable for
automation. Furthermore, the handling of small sample
volumes in the μl-range is required.
Several approaches for solid phase purification have been
reported. The PROBE assay (Sequenom, San Diego) uses
biotinylated primers to capture PCR products by streptavidin-
coated magnetic microparticles [23]. Purification is
performed, based on the immobilisation of the PCR
products. Magnetic bead handling is automated and works
with small sample volumes. One drawback is that biotinylated
primers, which are not consumed during the PCR,
compete with the reaction products in streptavidin-binding
and thereby reduce the capacity of the particles. Another
magnetic bead based purification, which removes residual
primers after PCR has been shown to result in good quality
MALDI SNP typing results [24]. Nevertheless, although
these methods are compatible with liquid handling robotics,
automation of magnetic bead handling results in time
consuming protocols and at present is restricted to a 96 well
format. Faster and cheaper 384 well magnetic bead handling
is currently not available. For streptavidin/biotin based
purification methods, streptavidin-coated microtitre plates
which are available in the 96 and 384 well format offer an
alternative approach, This enables an elegant and fast
automation of the liquid handling.
A further approach, the so-called GOOD assay [25],
circumvents any purification of DNA molecules by
chemically changing the nature of DNA molecules. This
assay utilizes derivative nucleic acid molecules in an allelespecific
primer extension reaction (see below).

Molecular Assay

Analysis of SNPs relies on amplification of the DNA
region surrounding the polymorphism, typically by PCR.
The most convenient way to analyse the polymorphic
information contained in the generated PCR products would
be by direct determination of their molecular masses.
Practically, this has not been achievable. The minimum size
of an amplicon containing the SNP region and the sequences
of both PCR primers is about 40 base pairs, while the mass
difference between two DNA bases is only 9 to 40 Dalton
(Fig. (2)). Unfortunately, DNA molecules of this size cannot
be measured with the resolution and accuracy that is
necessary to distinguish between the respective molecular
masses. Reasons are the extreme tendency of binding metal
ions and fragmentation of DNA molecules with increasing
length. Therefore, numerous secondary reactions have been
established to determine the polymorphic status of the PCR
products created in a primary amplification cycle. The aim of
all these secondary reactions is the translation of genotype
information from the PCR products to small, MALDI
compatible molecules. In addition to size reduction, reactions
like primer extension offer the possibility to include a second
dimension of specificity in the process.
Fig. (2). Mass shifts due to altered nucleotides in MALDI-TOF
MS based SNP genotyping.
The four different DNA nucleotides dATP, dCTP, dGTP and dTTP
have slightly different molecular weights due to the different
included nucleobases, whereas the sugar-phosphate backbone is
identical. When investigating short, single-stranded DNA
molecules, the modest mass variations that can be observed even if
only a single nucleotide differs between two classes of analyte
molecules is sufficient for an automated allele calling when using a
MALDI-TOF MS detection.
Several assays have been developed which are using
alternative reactions to create small molecules containing
SNP genotype information. The invader assay uses the allele
specific 5’-nuclease cleavage of an oligonucleotide probe,
which is triggered through a so-called invader oligonucleotide
in the case of a perfect hybridisation [26]. Although
intended to be a PCR-free approach for SNP genotyping,
robust results have been generated in combination with PCR
amplification [27, 28]. A further method called MALDImonitored
nuclease selection utilises the nuclease resistance
of oligonucleotides with a perfect hybridisation match with
one specific allele [29]. The increased selectivity and
stability of peptide nucleic acids (PNA) probes have been
used to combine allele-specific hybridisation with MALDITOF
analysis [30]. By far the most successful strategy for
MALDI-TOF genotyping is its combination with allele
specific primer extension reactions. At least two methods
combining primer extension and MALDI-TOF MS analysis
of the generated molecules have been published. After
removing residual PCR components by purification or
enzymatic treatment these approaches are using a primer that
is annealed adjacent to the polymorphic site contained in the
PCR product. This primer is extended in a sequence-dependent
manner to generate allele specific oligonucleotides for
the subsequent MALDI-TOF measurement.
2580 Current Pharmaceutical Design, 2005, Vol. 11, No. 20 Pusch and Kostrzewa
The primer extension can be performed in two variations
(Fig. (3)). If only terminating dideoxynucleotides (ddNTPs)
are used in the extension reaction the primers must be
annealed directly adjacent to the SNP [31]. The resulting
genotype-specific molecules are distinguished by the mass of
the incorporated ddNTP. This approach has the advantage of
easy assay design and high molecular multiplexing capability
[32]. On the other hand, in the mass range of a regular 17 to
25-mer extension oligonucleotide (about 5000-8000 Da) the
automated routine determination and resolution of the mass
difference between A and T, which is only 9 Dalton, is errorprone.
Therefore, a combination of deoxynucleotides
(dNTPs) and terminator ddNTPs can be used to generate
specific extension products, which differ by about a minimum
of one nucleotide in length for the respective alleles
(about 300 Da). This difference can be easily determined for
the 20 to 30mers generated in such extension reactions.
Disadvantage of this approach is that the combination of
different molecular reactions for multiplexing is more
complicated and not always affordable. Alternatively, the
utilisation of mass-labelled ddNTPs to increase the mass
difference of the genotype specific products has been
proposed [33].
Other approaches achieve an accurate molecular mass
determination of the allele-specific products by shortening
the molecules. Because about 20mers are necessary for a
proper hybridisation to the target sequence shortening has to
be performed after the extension reaction. The GOOD assay
is using primers, which contain several phosphorthioate
bonds at the 3’ end and a preformed positive charge together
with alpha-thio-ddNTPs. Following the extension reaction
the 5’ part is removed by exonuclease digestion. The
negative charges of the residual exonuclease-resistant
molecules are neutralised by alkylation of the phosphorthioate
moieties. Thereby, small single-positively charged
ions are generated which do not form adducts with metal
ions. A variation of the GOOD assay preventing the
alkylation step by using methyl phosphonates has been
described recently [34].
Fig. (3). Alternative methods to generate allele-specific products by primer extension reactions.
Different variations of a primer extension reaction are most widely used to generate allele-specific analyte molecules after PCR
amplification. Shown here are two alternatives, “short sequencing” and “single-base extension”. In both cases the primer aligns adjacent to
the polymorphic site. In single-base extension only dideoxynucleotide terminators are applied in the extension reaction. The allele-specific
nucleotide is incorporated and the reaction is immediately terminated. The setup of such a reaction is straightforward and multiplex analysis
of more than one SNP can be easily performed. However, the mass difference of the resulting products is limited. In contrast, using a “short
sequencing” approach a combination of normal dNTPs and ddNTP terminators is applied. The allele-specific reaction products differ in
length by at least one nucleotide. Accordingly, the mass difference is much higher, thus making the data interpretation easier. However, the
multiplexing capabilities of this approach are limited.
Application of MALDI-TOF Mass Spectrometry in Screening Current Pharmaceutical Design, 2005, Vol. 11, No. 20 2581
Alternatively, size reduction of the molecules can be
achieved by the incorporation of a cleavable site into the
primer. For this purpose, a method using a chemically
cleavable nucleotides as part of the extension oligonucleotide
has been used [35]. Alternatively, the GenoSNIP assay uses
a photocleavable moiety, which spans the same distance as a
nucleoside in the molecule [36]. This building block does not
build a Watson-and-Crick base pair and does not inhibit the
polymerase activity. Subsequent to primer extension 5’
biotinylated allele-specific products are bound to streptavidin-
coated cavities of microtitre plates. After washing
steps, the 3’ end, which is containing the genotype information,
is cleaved off by UV irradiation and transferred to a
MALDI sample plate. The method not only creates small
molecules for precise measurement but is also highly
compatible with automation.
Further Screening Applications in Nucleic Acid Analysis
Initially, MALDI-TOF MS was believed to be an alternative
tool for high-throughput sequencing in the human
genome project. Unfortunately, adduct-forming and fragmentation
of large DNA ions restricted its usage to
resequencing applications for short DNA stretches [37, 38].
Although the utilisation of an infrared laser enabled the
measurement of PCR products of several hundred base pairs
in length [39], to date this could not be used for any routine
application.
Recently, new approaches to apply MALDI-TOF to
further areas of DNA analysis have been presented. One very
attractive goal will be the search for unknown polymorphisms
in a high-throughput environment. Methods, which
generate small fragments by base-specific cleavage, have
been described by several authors [40-42]. Other developments
are dedicated to SNP quantitation in pooled samples
[43, 44], haplotyping [45], CpG methylation analysis [46]
and expression profiling [47].
QUALITY CONTROL
Oligonucleotides
Oligonucleotide quality control is besides SNP genotyping
the predominant application of MS in nucleic acid
analysis. Because molecular biology technologies have
entered veterinary, forensic and medical diagnostics during
the last years, requirements for accuracy, fidelity and
robustness of the adopted methods have been increased
significantly. Success of modern nucleic acid technologies
using polymerase chain reaction (PCR), primer extension
reactions or microarray related methods is dependent on
superior quality of the oligonucleotides used in the respective
assays. Several failures in synthesis and purification of
nucleic acid primers or probes can dramatically decrease the
efficiency and quality of molecular assays or may even lead
to a drop out of reactions. Therefore, an efficient synthesis
quality control is required, in particular in modern highthroughput
facilities which produce hundreds to thousands of
oligonucleotides per day.
Traditional nucleic acid quality control methods like gel
electrophoresis or HPLC are constricted in resolution and
accuracy, and thereby limited in their capabilities for
oligonucleotide quality monitoring. In contrast, MALDITOF
mass spectrometry is able to detect almost all synthesis
errors of nucleic acid molecules up to approximately 70mers.
In addition to its remarkable analytical power, the method is
appropriate for automation and inexpensive. Thus, MALDITOF
MS is the perfect tool for modern oligonucleotide
quality control.
Among synthesis errors, the generation of an oligonucleotide
with a sequence mistake is one of the most
profound. But also other mistakes may lead to severe effects.
These include depurination of adenosin or guanosin residues,
shortened fragments and incomplete removal of protection
groups. For a PCR reaction, poor oligonucleotide synthesis
may lead to a decreased reaction yield, prevent the reaction
at all, or may even lead to the erroneous amplification of the
wrong DNA region. Sequencing or primer extension
reactions may be deteriorated or give incorrect results.
Spotted on microarrays, aberrant nucleic acid fragments will
cause wrong interpretation of hybridisation results which
may lead to false interpretation of e.g. transcription profiling
experiments.
Accordingly, commercial suppliers of synthetic oligonucleotides,
as well as their customers are aiming at
appropriate quality controls measures.
Both ESI [48] and MALDI [49-51] ionisation techniques
have been used for this purpose. However, due to the
automation capabilities, MALDI-TOF MS is currently
dominating this market segment. The expected molecular
weight of each individual oligonucleotide can be directly
concluded from its nucleotide sequence. Modifications can
be considered accordingly. The aim of quality control is to
detect aberrations of oligonucleotides from the expected
sequences, as well as contaminating by-products that may
lead to failure of the oligonucleotide when applied for
molecular biological assays.
Strategy
The strategy for the automated data interpretation for
quality control issues, which is used in the commercial
genotools 2.0 software (Bruker Daltonik, Bremen, Germany)
will be described as an example.
The aim of the quality control is to detect all possible
contaminations in the preparation. Some of these contaminations
can be predefined as they are due to the synthesis, for
example non-cleaved protection groups. Others may be due
to unexpected effects.
As the control step must not miss any contamination,
peak detection is performed over the complete mass range of
the spectrum. The first step of the analysis is the detection of
the respective oligonucleotide. This is automatically done
within a specified acceptable mass tolerance window. The
signal/noise ratio and intensity of the detected oligonucleotide
is calculated and compared with the specified rejection
threshold. Subsequently, an internal calibration procedure
can be performed using the product peak. Potential shifts,
which may be due to an inhomogeneous sample preparation
on the MALDI target, are thus eliminated. Finally,
contamination signals are investigated. Three different
classes of contaminating peaks are distinguished: default
2582 Current Pharmaceutical Design, 2005, Vol. 11, No. 20 Pusch and Kostrzewa
contaminations, user-defined contaminations and unexpected
(unspecified) contaminations. They include for example nonhydrolysed
protection groups, nucleobase cleavages,
nucleotide cleavages or adduct ions. These contaminations
can be predefined (Table 1). The analysis of additional userdefined
contaminations may become necessary for specialised
syntheses using different chemistries. Such user-defined
contaminations can for example be specified in an appropriate
sample input table.
Finally, there may be totally unexpected contaminations
present in the sample, which can be detected and evaluated
by simply screening for contaminations in the complete mass
range of a spectrum.
Rejection Criteria
The sample may be rejected because of insufficient
intensity and/or signal/noise ratio of the oligonucleotide due
to an insufficient yield of the synthesis. Additionally, the
synthesis can be rejected if at least the intensity of one of the
contaminating peaks exceeds the defined respective
threshold. These thresholds are specified as a percentage of
the target oligonucleotide intensity. Independent thresholds
can be adjusted for the different classes of contaminations.
Result Display and Reporting
For a convenient inspection of the results a graphical
display should be provided as well as an appropriate output
Table 1. Example for Pre-Definable Contamination Peaks
Type of contamination Group Mass [Da]
Non-hydrolised protection groups Acetyl + 42.04
Isobuturyl + 70.09
Benzoyl + 104.11
4,4´-Dimethoxytrityl + 302.37
Adducts Ammonium + 17.03
Sodium + 21.98
Potassium + 38.09
Iron + 54.84
Nucleobase cleavage - 1x Adenine - 133.11
- 2x Adenine - 266.22
- 1x Cytosine -110.09
- 2x Cytosine - 220.18
- 1x Guanine -149.11
- 2x Guanine - 298.22
- 1x Thymidine -124.10
- 2x Thymidine - 248.20
Nucleotide cleavage - 1x Desoxyadenosine - 313.21
- 2x Desoxyadenosine - 626.42
- 1x Desoxycytidine - 289.19
- 2x Desoxycitidine - 578.38
- 1x Desoxyguanosine - 329.11
- 2x Desoxyguanosine -658.22
- 1x Desoxythymidine - 304.20
- 2x Desoxythymidine - 608.40
Thioathydrolysis - 1x Thioat - 16.07
- 2x Thioat - 32.13
- 3x Thioat - 48.20
Application of MALDI-TOF Mass Spectrometry in Screening Current Pharmaceutical Design, 2005, Vol. 11, No. 20 2583
table. The graphical display showing the analysis quality for
a complete analysis set, for example in a 384 well MTP
format, can be used to get an immediate and efficient overview
of a large data set after synthesis (Fig. (4)). In contrast,
an output table can be used for the long-term administration
of quality control results.
Fig. (4). Graphical overview displaying the results derived from
a complete 384 well microtitre sample plate.
A: Color-coded allele calling results from SNP genotyping. Homozygous
genotypes are displayed as full cycles. Heterozygous
genotypes are displayed using hemi-circles. The according allele
peaks can be tracked by the respective colour-coded bars in the
spectrum view.
B: Internal quality control of results using a traffic light colourcode.
Results shown here are from oligonucleotide quality control.
Green colour showing high quality, yellow showing medium
quality and red showing low quality.
COMPOUND LIBRARIES
In recent years combinatorial chemistry has made a rapid
evolution leading to significantly improved synthesis capabilities
of structurally diverse compounds. Such libraries can
be used in the pharmaceutical industry as a valuable tool in
the process of drug development. Once potential drug target
proteins have been identified, such compound libraries can
be applied to screen for appropriate drug leads that exert a
physiological function via an interaction with the target
protein. All the compounds contained in such a library have
to be tested for structural integrity and purity. The
requirements for this purpose are very similar to other
screening approaches. First of all, the procedure has to be
sufficiently sensitive in order to be applied efficiently, as the
individual compounds are typically synthesised in a
nanomole scale [52]. Moreover, it has to be suitable for high
throughput screening (HTS) by easy automation and it has to
be cost-effective. Different MS techniques have been applied
for this purpose, like MALDI-TOF MS [53-54] or LC-MS
using ion traps, single quadruple or triple quadruple
instruments [55-57].
In order to minimise the failure rate of identified drug
leads from such a compound library it is also necessary to
address the physicochemical and pharmacological properties
of the substances as early in development as possible. In
particular absorption, distribution, metabolism and elimination
(ADME) have to be monitored. Even for these advanced
screening applications LC-MS has proven a useful tool [58].
A major critical point that has led to the success of MS
also in this field of analysis is the possibility to couple MS
with a variety of separation techniques, like capillary
electrophoresis (CE) and high performance liquid chromatography
(HPLC) [59].
All these features make MS not only a key player in the
quality control of such compound libraries, but also a
valuable tool for the identification of useful drug leads out of
such libraries. Furthermore, the stability of peptide and
protein drugs over time may be controlled in the course of
drug development.
BIOMARKERS AND PROFILING
Recently, a further new application area that is beyond
the classical expression proteomics has arisen with the use of
mass spectrometry for so-called profiling approaches for
diagnostic research purposes [60]. The idea behind such a
profiling is an earlier, more specific and more secure
diagnosis of certain diseases in comparison to classical
ELISA approaches. Diseases should be detectable by the
molecular traces that they leave in easily accessible
biological fluids, like serum, saliva or urine (Fig. (5)).
Proteins, peptides or fragments derived therefrom may be
shed into the fluid by necrosis or apoptosis of, for example,
tumor cells, or they may be even actively secreted. This
leads to a complex pattern of biomarker molecules. These
patterns can be used holistically to identify the disease by
means of a multivariate analysis. In a supervised approach,
such a biomarker pattern can be established using a training
data set derived from patient samples with very well defined
clinical diagnosis. This training data set is divided into a case
2584 Current Pharmaceutical Design, 2005, Vol. 11, No. 20 Pusch and Kostrzewa
and a control group. Subsequently, the quality of the
determined biomarker pattern can be evaluated using an
independent test data set. Finally, the biomarker pattern can
be used for the classification of new samples with unknown
clinical diagnosis (Fig. (6)).
The sensitivity and specificity achieved with this
approach has been reported from several studies to be
superior to conventional ELISA assays using an antibodybased
detection of a single biomarker protein. The idea
behind the MS-based approach is to identify the disease at a
time point when the chances of a successful treatment are
still much higher than at an advanced stage of the disease.
This has recently been shown for the diagnosis of ovarian
cancer [61]. Moreover, such an approach is also possible
without prior identification of the biomarker peptides. Nor is
there a need for commercially available antibodies. However,
in many cases it is favourable to know the individual
biomarkers of a pattern, either to use them to raise antibodies
against the protein for a use in established antibody-based
approaches or to achieve simply a better acceptance of such
a profiling test (Fig. (7)).
Definition of Sample Cohorts
A major task for the broad establishment of biomarker
profiling approaches is a very thorough definition of case
and control cohorts. As in the case of SNP genotyping using
non-matching groups can rapidly lead to classification results
that have a bias, which is not due to biological or clinical
effects [62]. Non-matching case and control cohorts may in
fact be the major drawback of this new technological
approach. The clinical diagnoses of the patients that are
included in the training data set have to be very thorough.
Besides a very good classification of the disease of interest,
for example including the various stages of cancer
development, all clinical indications not directly related to
the disease of interest have to be taken into consideration as
well. Infections and inflammation may alter biomarker
patterns in a similar way to a cancer. Furthermore, both
training data subsets (cases and controls) have to be
matching in terms of gender and age [63]. Accordingly, if
the available number of samples is limited, the result of the
classification approach may be misleading due to nonmatching
case and control cohorts.
Fig. (5). The principle of serum profiling.
Profiling is a technique to get information on the physiological condition of a patient by means of investigating easily accessible body fluids.
Serum is widely used, but other samples like urine, saliva or cerebral spinal fluid are suitable as well. The idea behind profiling is that a
specific disease leaves traces of biomarkers in the serum. In contrast to current antibody-based technologies, serum profiling does not rely on
a single marker, but performs a multivariate analysis of several biomarker signals. A purification and fractionation is performed to get rid of
disturbing contaminations and to enrich specific subfractions of the proteome according to their physical and biochemical properties.
MALDI-TOF MS spectra are acquired. Subsequently, different classes of these profile spectra are used to determine biomarker patterns that
may be used to distinguish different sample classes, for example by means of a cluster analysis. New and unknown samples can be classified
using such biomarker patterns. As a result an answer concerning the clinical condition of a patient may be generated.
Application of MALDI-TOF Mass Spectrometry in Screening Current Pharmaceutical Design, 2005, Vol. 11, No. 20 2585
Sample Preparation/Sample-to-Sample Variation
A major point that has to be addressed very thoroughly is
sample-to-sample variation due to different pre-treatments of
samples by different laboratories or persons or due to
different sample storage. Accordingly, in order to get
reproducible results, so-called standard operation procedures
(SOPs) have to be in place to ensure comparable sample
treatment during the complete procedure. Variation of the
protein profile spectra may be caused by two independent
mechanisms, namely the sample preparation and the
MALDI-TOF measurement itself.
The current strategy of most profiling approaches is an
enrichment of subfractions of the proteome of a clinical
sample based on physical and/or biochemical properties of
the individual proteins. For this purpose various surface
functionalities are currently available, for example
hydrophobic interaction, cation exchange, anion exchange,
immobilised metal-affinity chromatography and others. Each
of these functionalities enriches a subfraction of the total
proteome of the sample (Fig. (8)). However, there may be
some overlap in this fractionation behaviour, which means
that a protein may not only be enriched with one but with
two or more of the available functionalities. These surface
functionalities can be applied via different methodologies;
so-called protein chips (Ciphergen Biosystem, Fremont, CA,
USA) or magnetic beads (Bruker Daltonik, Leipzig,
Germany) are currently state of the art. Both systems have
their individual advantages. Protein chips for 8 or 16 sample
positions are easy to handle. They are well established and
broadly accepted by clinicians. However the throughput is
limited and the life cycle costs caused by consumables are
considerable. Magnetic bead based approaches are gaining
visibility as well as offering a cost-effective alternative. The
automation of magnetic bead handling has been established
for a number of years, thus enabling a considerable sample
throughput. In addition to throughput, an appropriate robotic
solution also offers the convenience of a standardised
protocol thus ensuring a maximum of reproducibility (Fig.
(9)). In initial optimisation studies, variation introduced by
the sample preparation and variation introduced by the
MALDI-TOF process can be investigated independently.
The former can be addressed by multiple preparations of the
same sample, whereas the latter can be addressed by multiple
MALDI-TOF measurements of the same preparation.
Magnetic beads also offer convenient scalability of the
sample preparation process. The volume of bead suspension
can be simply adjusted to the individual requirements for
protein fractionation.
Fig. (6). The principle of supervised pattern discovery approaches.
In a supervised pattern discovery approach a training data set containing subsets of case and control cohorts with well-defined clinical
diagnoses is used to determine a discriminating biomarker pattern (discovery phase). Subsequently, the pattern can be evaluated using an
independent test data set of well-defined case and control samples. The sensitivity and specificity of the pattern can thus be determined
(evaluation phase). Finally, high quality patterns can be used to classify patient samples with unknown clinical diagnoses (class prediction
phase).
2586 Current Pharmaceutical Design, 2005, Vol. 11, No. 20 Pusch and Kostrzewa
Fig. (7). The application of biomarker analysis for identification and screening/diagnostic research.
Biomarker analysis is a valuable tool for the investigation of various diseases. After the generation of complex protein profiles, biomarker
candidates can be identified, validated and used for screening and early diagnosis of a disease. The results of such an approach are not only
useful tools for early diagnosis of the disease, but also for prognosis, as well as for drug safety evaluation. However, to give the biomarker
patterns a higher credibility, identification of individual biomarker peptides or proteins may become necessary. Additionally, by raising
antibodies with appropriate specificity, identified biomarker proteins may be used to improve traditional antibody-based ELISA diagnostic
assays.
Fig. (8). Different surface functionalities provide diversity of biomarker patterns.
Profiling studies in clinical proteomics make use of readily accessible body fluids, for example serum, plasma, cerebral spinal fluid, saliva or
urine, to generate complex profiles of the protein content. To get rid of contaminating salt ions and other compounds that may interfere with
the MALDI-TOF process, a purification and fractionation step has to be performed. Using different surface functionalities allows the
enrichment of different fractions of the respective proteome. Different functionalities may have an overlapping protein fractionation
specificity. Multiple functionalities may be used in parallel to generate a diversity of biomarker patterns during the optimization steps of the
procedure for a specific disease.
Application of MALDI-TOF Mass Spectrometry in Screening Current Pharmaceutical Design, 2005, Vol. 11, No. 20 2587
Fig. (9). Reproducibility of magnetic bead sample preparation.
a) Manual Preparation: Spectra derived from the same sample, which has been prepared 7-fold using magnetic beads with a hydrophobic
coating (C8). The sample preparation has been manually performed.
b) Automated Preparation: Spectra derived from the same sample, which has been prepared 20-fold using magnetic beads with a
hydrophobic coating (C8). The sample preparation has been automatically performed using a robotic solution (Bruker Daltonik, Leipzig,
Germany).
c) Automated Preparation 384x: Density plot in a grey scale derived from spectra from the same sample, which has been prepared 384-fold
using magnetic beads with a hydrophobic coating (C8). The sample preparation has been automatically performed using a robotic solution
(Bruker Daltonik, Leipzig, Germany).
2588 Current Pharmaceutical Design, 2005, Vol. 11, No. 20 Pusch and Kostrzewa
Moreover, sophisticated multidimensional fractionations
can be applied by combinations of multiple surface functionalities.
Last but not least, the offline processing using a
magnetic bead system offers the advantage, that the user has
an additional sample volume available. During the process
the purified protein fractions are eluted from the beads. From
this elution only a small fraction is necessary to generate the
respective MALDI-TOF MS profile. The rest of the sample
is available to repeat the profiling experiment. But what is
more, the remaining sample can be applied for complementary
techniques, like LC-MS or tandem MS. This is
especially important for the identification of individual
biomarker species.
Visualisation and Establishment of Biomarker Patterns
After the acquisition of protein profiles from larger data
sets, discriminating biomarker patterns have to be generated
and evaluated. As a first step for this process an efficient
visualisation of a complete data panel is crucial, as the comparison
of individual spectra does not promise significant
success.
Visualisation is necessary to get a first impression
concerning the data quality and any obvious differences
between the two groups. Subsequently, algorithms for
automated pattern determination have to be applied to get
statistically evaluated results. In regard to these algorithms,
there does not seem to be a single solution, which is suitable
for each purpose. Different algorithms may prove useful for
different diseases or for different biological samples.
Currently three different approaches have a very good
visibility, genetic algorithms [47], decision trees [64-66] and
the “unified maximum separability algorithm (UMSA)”,
which is a modified support vector machine [67]. Genetic
algorithms use an iterative, multiple-cycle optimisation,
which is similar to the principles of DNA recombination in
biological evolution. However, instead of genes, which are
arranged sequentially on a chromosome, in this case, signal
masses whose respective intensities may be used in a
discriminating pattern are arranged sequentially on so-called
logical chromosomes. These signal masses on the logical
chromosome are modified iteratively (Fig. (10)). The fitness
of the resulting signal pattern to distinguish between the two
classes is determined. Alterations that lead to an improvement
of the classification are kept and used for a second
cycle of mutation and crossing over. Those alterations that
lead to a worse classification are discarded. The mechanisms
to vary the included peak masses in each cycle of the process
are similar to biological mechanisms as well. Similar to a
single point mutation, individual peak masses are exchanged
Fig. (10). The principle of pattern determination based on genetic algorithms.
Using genetic algorithms, the optimization of a biomarker pattern occurs similar to the mechanisms driving biological evolution. Similar to
biological chromosomes, signal masses of the profile spectra are arranged on linear “logical chromosomes”. The signal information included
in the various logical chromosomes is altered in an iterative process, using multiple cycles of survival of the fittest. Alterations are introduced
by “mutation” and “crossing over” events. Like a biological point mutation individual peak masses are exchanged by different masses in
mutation events. Complete logical chromosomes exchange their signal information in a crossing over event. After each cycle of mutation and
crossing over a fitness test is performed. Such alterations that lead to an improved fitness are kept. Those that lead to decreased fitness are
discarded.
Application of MALDI-TOF Mass Spectrometry in Screening Current Pharmaceutical Design, 2005, Vol. 11, No. 20 2589
in each cycle. Moreover, crossing over events between
different logical chromosomes occur, which lead to largescale
modifications. Once a pattern is established, it can be
used for a validation using an independent test data set.
Subsequently, a classification of unknown data is also
possible. For such a classification, standard cluster analysis
algorithms can be applied, for example a k-nearest neighbour
[68] or a centroid approach [69-71].
A further well-established approach is decision tree
analysis. Once a biomarker pattern is established, the
decisions that lead to a classification of a new sample is
easily pursuable as they are related to the principles of
differential diagnosis applied by clinicians. For example, if
the biomarker pattern consists of five independent peak
masses, the intensity of a new profile spectrum at each of the
indicated masses is evaluated. Each peak analysis leads to a
yes/no answer. Is the intensity higher than the specified
threshold – yes or no? All these decisions lead to a tree-like
path of individual decisions [64].
A third approach which has recently been successfully
used for profiling analysis of clinical samples is the “unified
maximum separability algorithm (UMSA)” which is a
modified support vector machine [72]. Support vector
machines have the advantage that they avoid the effect of
over-training or over-fitting of the data [73-75]. Moreover,
classical peak statistics, like a Welch’s t-test, provides
powerful tools to select signals with a very high power to
discriminate between different sample groups. In conclusion,
it may be advisable to apply different pattern determination
approaches in parallel, compare the results and select the
best model for each individual diagnostic purpose.
OUTLOOK
With screening and diagnostic research, MS has entered a
completely new application field. Proteomics will presumably
dominate the life science for at least the next 10 years
and mass spectrometry is the key technology without which
larger proteomics projects would not be feasible. However,
besides the general need for sophisticated technological
solutions for proteomics, a further general trend leads from
separated life science disciplines with separated data to an
integrated systems biology approach. The purpose here is the
integration of raw data and results from completely different
technologies, more importantly, from the analysis of
different classes of biomolecules. The era of the separated
disciplines genomics and proteomics will soon be past. In
larger projects, proteomics data have to be set into relation
with gene expression data from microarrays and from the
analyses of SNP genotyping.
This new way of data interpretation will eventually lead
to a more holistic picture, especially in the analysis of
diseases. In drug development the analysis of small molecules
and their metabolites are further major tasks.
MS is one of a few technologies – if not the only one –
that supports such an approach on the same instrument
platform. SNP genotyping, protein identification, protein indepth
analysis of post-translational modifications, serum
profiling, classification of clinical samples, small molecule
analysis and metabolite screening are all possible using MS.
Most of these approaches are even possible with one defined
instrument type, namely MALDI-TOF MS. This makes MS
not only the dominating technology in proteomics research,
but also a key player in future developments of system
biology platforms (Fig. (11)).
Especially MALDI-TOF MS will presumably expand its
role as a workhorse for various existing and further emerging
applications in high-throughput screening (HTS).
Fig. (11). MALDI-TOF mass spectrometry as a system biology
platform?
In the post-genomic era a major task of future research and
development will be the integration of various information sources
and data types derived from the analysis of different analyte
molecule classes. A holistic system biology approach addresses not
only genotypes, not only gene expression and not only protein
identification. Instead, all of these types of information have to be
integrated. MALDI-TOF MS is one of a few technologies, if not the
only one, that enables the analysis of SNPs, gene expression,
protein identification and screening/diagnostic research based on
profiling of complex clinical samples. Thus, MALDI-TOF MS has
the technological potential to evolve to an advanced system biology
platform in forthcoming years.
Additionally, with the rise of screening and diagnostic
applications in the field of medical research, MS has the
potential to revolutionise medical diagnostics in a similar
way, to what has occurred in basic life sciences in recent
years.
MS will not substitute for existing and well-established
diagnostic tools, such as enzyme-linked immunosorbent
assays (ELISA), but it will be an enabling complementary
technology, which may be very helpful for diseases for
which ELISA assays display only a limited sensitivity or
specificity. Thus the number of missed disease cases may be
reduced. Moreover, the number of false-positive diagnoses
may be reduced as well, thus eliminating the need for further
investigations of those patients. Such principally unnecessary
examinations are a burden for the public health care system,
as well as for the individual patients, as they include in many
2590 Current Pharmaceutical Design, 2005, Vol. 11, No. 20 Pusch and Kostrzewa
cases more invasive examination techniques, like biopsies
and subsequent histological diagnosis. However, the way for
these promising approaches, from clinical research to the
bedside, is still very long. Future developments on the part of
the vendors to further simplify the technology for clinical
routine applications will be necessary, as well as research
from the respective users concerning standardisation of
sample handling, sample preparation and definition of
meaningful case/control cohorts.
ACKNOWLEDGEMENTS
The authors would like to thank Prof. Dr. Richard Ivell
for proof-reading of the manuscript.
REFERENCES
References 76-78 are related articles recently published in
Current Pharmaceutical Design.
[1] Karas M, Hillenkamp F. Laser desorption ionization of proteins
with molecular masses exceeding 10,000 daltons. Analytical
Chemistry 1988; 60(20): 2299-2301.
[2] Hillenkamp F, Karas M. Mass spectrometry of peptides and
proteins by matrix-assisted ultraviolet laser desorption/ionization.
Methods Enzymol 1990; 193: 280-295.
[3] Hillenkamp F, Karas M, Beavis RC, Chait BT. Matrix-assisted
laser desorption/ionization mass spectrometry of bioploymers.
Analytical Chemistry 1991; 63(24): 1193A-1203A.
[4] Fenn JB, Mann M, Meng CK, Wong SF, Whitehouse CM.
Electrospray ionization for mass spectrometry of large
biomolecules. Science 1989; 246(4926): 64-71.
[5] Lechner D, Lathrop GM, Gut IG. Large-scale genotyping by mass
spectrometry: experience, advances and obstacles. Curr Opin Chem
Biol 2002; 6(1): 31-8.
[6] Sickmann A, Mreyen M, Meyer HE. Mass spectrometry--a key
technology in proteome research. Adv Biochem Eng Biotechnol
2003; 83: 141-76.
[7] Ashcroft AE. Protein and peptide identification: the role of mass
spectrometry in proteomics. Nat Prod Rep 2003; 20(2): 202-15.
[8] Pusch W, Wurmbach JH, Thiele H, Kostrzewa M. MALDI-TOF
mass spectrometry-based SNP genotyping. Pharmacogenomics
2002; 3(4): 537-548.
[9] Grant GA, Frison SL, Yeung J, Vasanthan T, Sporns P.
Comparison of MALDI-TOF mass spectrometric to enzyme
colorimetric quantification of glucose from enzyme-hydrolyzed
starch. J Agric Food Chem 2003; 51(21): 6137-44.
[10] Dell A, Morris HR. Glycoprotein structure determination by mass
spectrometry. Science 2001; 291(5512): 2351-6.
[11] Harvey DJ. Identification of protein-bound carbohydrates by mass
spectrometry. Proteomics 2001; 1(2): 311-28.
[12] Harvey DJ. Matrix-assisted laser desorption/ionization mass
spectrometry of carbohydrates. Mass Spectrom Rev 1999; 18(6):
349-450.
[13] Cohen LH, Gusev AI. Small molecule analysis by MALDI mass
spectrometry. Anal Bioanal Chem 2002; 373(7): 571-86.
[14] Guzey C, Spigset O. Genotyping of drug targets: a method to
predict adverse drug reactions? Drug Saf 2002; 25(8): 553-60.
[15] Izawa H, Yamada Y, Okada T, Tanaka M, Hirayama H, Yokota M.
Prediction of genetic risk for hypertension. Hypertension 2003;
41(5): 1035-40.
[16] Rose CM, Marsh S, Ameway MM, McLeod HL. Pharmacogenetic
analysis of clinically relevant genetic polymorphisms. Methods
Mol Med 2003; 85: 225-37.
[17] Ye J, Parra EJ, Sosnoski DM, Hiester K, Underhill PA, Shriver
MD. Melting curve SNP (McSNP) genotyping: a useful approach
for diallelic genotyping in forensic science. J Forensic Sci 2002;
47(3): 593-600.
[18] Linacre A, Graham D. Role of molecular diagnosticcs in forensic
science. Expert Rev Mol Diagn 2002; 2(4): 346-53.
[19] Leushner J, Chiu NH. Automated mass spectrometry: a
revolutionary technology for clinical diagnostics. Mol Diagn 2000;
5(4): 341-8.
[20] Humeny A, Bonk T, Berkholz A, Wildt L, Becker CM. Genotyping
of thrombotic risk factors by MALDI-TOF mass spectrometry. Clin
Biochemistry 2001; 34(7): 531-6.
[21] Kato N. Genetic analysis in hypertension. Hypertension Research
2002; 25(3): 319-27.
[22] Ross JS, Ginsburg GS. Integration of molecular diagnostics with
therapeutics: implications for drug discovery and patient care.
Expert Rev Mol Diag 2002; 2(6): 531-40.
[23] Little DP, Braun A, Darnhofer-Demar B, Koster H. Identification
of apolipoprotein E polymorphisms using temperature cycled
primer oligo base extension and mass spectrometry. Eur J Clin
Chem Clin Biochem 1997; 35(7): 545-8.
[24] Pusch W, Kraeuter KO, Froehlich T, Stalgies Y, Kostrzewa M.
genotools “SNP manager“: A new software for automated highthroughput
MALDI-TOF mass spectrometry SNP genotyping.
Biotechniques 2001; 30(1): 210-5.
[25] Sauer S, Lechner D, Berlin K, Plancon C, Heuermann A, Lehrach
H, et al. Full flexibility genotyping of single-nucleotide polymorphisms
by the GOOD assay. Nucleic Acids Research 2000;
28(33): E100.
[26] Griffin TJ, Hall JG, Prudent JR, Smith LM. Direct genetic analysis
by matrix-assisted laser desorption/ionization mass spectrometry.
Proc Natl Acad Sci USA 1999; 96(11): 6301-6.
[27] Mein Ca, Barratt BJ, Dunn MG, Siegmund T, Smith AN, Esposito
L, et al. Evaluation of single nucleotide polymorphism typing with
invader on PCR amplicons and its automation. Genome Res 2000;
10(3): 330-43.
[28] Fei Z, Smith LM. Analysis of single nucleotide polymorphisms by
primer extension and matrix-assisted laser desorption/ionization
time-of-flight mass spectrometry. Rapid Commun Mass Spectrom
2000;14(11): 950-9.
[29] Stoerker J, Mayo JD, Tetzlaff CN, Sarracino DA, Schwope I,
Richert C. Rapid genotyping by MALDI-monitored nuclease
selection from probe libraries. Nature Biotechnology 2000; 18(11):
1213-1216.
[30] Ross PL, Lee K, Belgrader P. Discrimination of single-nucleotide
polymorphisms in human DNA using peptide nucleic acid probes
detected by MALDI-TOF mass spectrometry. Anal Chem 1997;
69(20): 4197-202.
[31] Krook A, Stratton IM, O'Rahilly S. Rapid and simultaneous
detection of multiple mutations by pooled and multiplex single
nucleotide primer extension: application to the study of insulinresponsive
glucose transporter and insulin receptor mutations in
non-insulin-dependent diabetes. Hum Mol Genet 1992;1(6): 391-5.
[32] Ross P, Hall L, Smirnov I, Haff L. High level multiplex genotyping
by MALDI-TOF mass spectrometry. Nature Biotechnology 1998;
16: 1347-1351.
[33] Fei Z, Ono T, Smith LM. MALDI-TOF mass spectrometric typing
of single nucleotide polymorphisms with mass-tagged ddNTPs.
Nucleic Acids Res 1998; 26(11): 2827-8.
[34] Sauer S, Gelfand DH, Boussicault F, Bauer K, Reichert F, Gut IG.
Facile method for automated genotyping of single nucleotide
polymorphisms by mass spectrometry. Nucleic Acids Res 2002;
30(5): e22.
[35] Li J, Butler JM, Tan Y, Lin H, Royer S, Ohler L, et al. Single
nucleotide polymorphism determination using primer extension and
time-of-flight mass spectrometry. Electrophoresis 1999; 20(6):
1258-65.
[36] Wenzel T, Elssner T, Fahr K, Bimmler J, Richter S, Thomas I, et
al. Genosnip: SNP genotyping by MALDI-TOF MS using
photocleavable oligonucleotides. Nucleosides Nucleotides Nucleic
Acids 2003; 22(5-8): 1579-81.
[37] Nordhoff E, Luebbert C, Thiele G, Heiser V, Lehrach H. Rapid
determination of short DNA sequences by the use of MALDI-MS.
Nucleic Acids Res 2000; 28(20): E86.
[38] Fu DJ, Broude NE, Koster H, Smith CL, Cantor CR. Efficient
preparation of short DNA sequence ladders potentially suitable for
MALDI-TOF DNA sequencing. Genet Anal 1996;12(3-4): 137-42.
[39] Berkenkamp S, Kirpekar F, Hillenkamp F. Infrared MALDI mass
spectrometry of large nucleic acids. Science 1998; 281(5374): 260-
2.
[40] Rodi CP, Darnhofer-Patel B, Stanssens P, Zabeau M, van den
Boom D. A strategy for the rapid discovery of disease markers
using the MassARRAY system. Biotechniques 2002; Suppl 62-6:
68-9.
Application of MALDI-TOF Mass Spectrometry in Screening Current Pharmaceutical Design, 2005, Vol. 11, No. 20 2591
[41] Hartmer R, Storm N, Boecker S, Rodi CP, Hillenkamp F, Jurinke
C, et al . RNase T1 mediated base-specific cleavage and MALDITOF
MS for high-throughput comparative sequence analysis.
Nucleic Acids Res 2003; 31(9): e47.
[42] Krebs S, Medugorac I, Seichter D, Forster M. RNaseCut: a MALDI
mass spectrometry-based method for SNP discovery. Nucleic Acids
Res 2003; 31(7): e37.
[43] Ross P, Hall L, Haff LA. Quantitative approach to singlenucleotide
polymorphism analysis using MALDI-TOF mass
spectrometry. Biotechniques 2000; 29(3): 620-629.
[44] Buetow KH, Edmonson M, MacDonald R, Clifford R, Yip P,
Kelley J, et al. High-throughput development and characterization
of a genomewide collection of gene-based single nucleotide
polymorphism markers by chip-based matrix-assisted laser
desorption/ionization time-of-flight mass spectrometry. Proc Natl
Acad Sci USA 2001; 98(2): 581-584.
[45] Tost J, Brandt O, Boussicault F, Derbala D, Caloustian C, Lechner
D, et al. Molecular haplotyping at high throughput. Nucleic Acids
Res 2002; 30(19): e96.
[46] Tost J, Schatz P, Schuster M, Berlin K, Gut IG. Analysis and
accurate quantification of CpG methylation by MALDI mass
spectrometry. Nucleic Acids Res 2003; 31(9): e50.
[47] Ding C, Cantor CR. A high-throughput gene expression analysis
technique using competitive PCR and matrix-assisted laser
desorption ionization time-of-flight MS. Proc Natl Acad Sci USA
2003; 100(6): 3059-64.
[48] De Bellis G, Salani G, Battaglia C, Pietta P, Rosti E, Mauri P.
Electrospray ionization mass spectrometry of synthetic
oligonucleotides using 2-propanol and spermidine. Rapid Commun
Mass Spectrom 2000; 14(4): 243-9.
[49] Hathaway GM. Characterization of modified and normal
deoxyoligonucleotides by MALDI, time-of-flight mass
spectrometry. Biotechniques 1994; 17(1): 150-5.
[50] Ball RW, Packman LC. Matrix-assisted laser desorption ionization
time-of-flight mass spectrometry as a rapid quality control method
in oligonucleotide synthesis. Anal Biochem 1997; 246(2): 185-94.
[51] Butler JM, Devaney JM, Marino MA, Vallone PM. Quality control
of PCR primers used in multiplex STR amplification reactions.
Forensic Sci Int 2001;119(1): 87-96.
[52] Triolo A, Altamura M, Cardinali F, Sisto A, Maggi CA. Mass
spectrometry and combinatorial chemistry: a short outline. J Mass
Spectrom 2001; 36(12): 1249-59.
[53] Shah N, Gao M, Tsutsui K, Lu A, Davis J, Scheuermann R, et al. A
novel approach to high-throughput quality control of parallele
synthesis libraries. J Comb Chem 2000; 2(5): 453-60.
[54] Mathur S, Hassel M, Steiner F, Hollemeyer K, Hartmann RW.
Development of a new approach for screening combinatorial
chemistry libraries using MALDI-TOF-MS and HPLC-ESIMS/
MS. J Biomol Screen 2003; 8(2): 136-48.
[55] Haring N, Salama Z, Jaeger H. Triple quadrupole mass
spectrometric determination of bromocriptine in human plasma
with negative ion chemical ionization. Arzneimittelforschung 1988;
38(19): 1529-32.
[56] Isbell J, Xu R, Cai Z, Kassel DB. Realities of high-throughput
liquid chromatography/mass spectrometry purification of large
combinatorial libraries: a report on overall sample throughput using
parallel purification. J Comb Chem 2002; 4(6): 600-11.
[57] Leister W, Strauss K, Wisnoski D, Zhao Z, Lindsley C.
Development of a custom high-throughput preparative liquid
chromatography/mass spectrometry platform fort he preparative
purification and analytical analysis of compound libraries. J Comb
Chem 2003; 5(3): 322-9.
[58] Xu R, Nemes C, Jenkins KM, Rourick RA, Kassel DB, Liu CZ.
Application of parallel liquid chromatography/mass spectrometry
for high throughput microsomal stability screening of compound
libraries. J Am Soc Mass Spectrom 2002; 13(2): 155-65.
[59] Sussmuth RD, Jung G. Impact of mass spectrometry on
combinatorial chemistry. J Chromatogr B Biomed Sci Appl 1999;
725(1): 49-65.
[60] Pusch W, Flocco MT, Leung SM, Thiele H, Kostrzewa M. Mass
spectrometry-based clinical proteomics. Pharmacogenomics 2003;
4(4): 463-476.
[61] Petricoin EF, Ardekani AM, Hitt BA, Levine PJ, Fusaro VA,
Steinberg SM, et al. Use of proteomic patterns in serum to identify
ovarian cancer. The Lancet 2002; 359: 572-577.
[62] Sorace JM, Zhan M. A data review and re-assessment of ovarian
cancer serum proteomic profiling. BMC Bioinformatics 2003; 4(1):
24. Epub 2003 Jun 09.
[63] Pritzker KPH. Cancer biomarkers: Easier said than done. Clinical
Chemistry 2002; 48(8): 1147-1150.
[64] Fung ET, Enderwick C. ProteinChip clinical proteomics:
Computational challenges and solutions. Biotechniques 2002;
Suppl 34-8: 40-1.
[65] Adam B, Qu Y, Davis JW, Ward MD, Clements, MA, Cazares,
LH, et al. Serum protein fingerprinting coupled with a patternmatching
algorithm distinguishes prostate cancer from benign
prostate hyperplasia and healthy men. Cancer Research 2002; 62:
3609-3614.
[66] Qu Y, Adam BL, Yasui Y, Ward MD, Cazares LH, Schellhammer
PF, et al. Boosted decision tree analysis of surface-enhanced laser
desorption/ionization mass spectral serum profiles discriminates
prostate cancer from noncancer patients. Clinical Chemistry 2002;
48(10): 1835-1843.
[67] Li J, Zhang Z, Rosenzweig J, Wang YY, Chan D. Proteomics and
bioinformatics approaches for identification of serum biomarkers to
detect breast cancer. Clinical Chemistry 2002; 48(8): 1296-1304.
[68] Cuzick J, Edwards R. Methods for investigating localized
clustering of disease. Clustering methods based on k nearest
neighbour distributions. IARC Sci Publ 1996;(135): 53-67.
[69] Carroll PG, DeWolf WC, Mehta CR, Rohan JE, Yunis EJ. Centroid
cluster analysis of the primed lymphocyte test. Transplant Proc
1979; 11(4): 1809-12.
[70] Whitaker D, Walker H. Centroid evaluation in the vernier
alignment of random dot clusters. Vision Res 1988; 28(7): 777-84.
[71] Kia SJ, Coghill GG. Unsupervised clustering and centroid
estimation using dynamic competitive learning. Biol Cybern 1992;
67(5): 433-43.
[72] Zhang Z, Chan DW. Unified Maximum Separability Analysis a
supervised approach for biological expression data analysis. http:
//www.aacc.org/meetings/oakridge/2002/abstracts/OCRabs26.pdf.
[73] Pitt MA, Myung IJ. When a good fit can be bad. Trends Cogn Sci
2002; 6(10): 421-425.
[74] Cai YD, Liu XJ, Xu XB, Chou, KC. Support vector machines for
the classification and prediction of beta-turn types. J Pept Sci 2002;
8(7): 297-301.
[75] Hua S, Sun Z. A novel method of protein secondary structure
prediction with high segment overlap measure: support vector
machine approach. Journal of Molecular Biology 2001; 308(2):
397-402.
[76] Aune TM, Maas K, Moore JH, Olsen NJ. Gene expression profiles
in human autoimmune disease. Curr Pharm Design 2003; 9(23):
1905-17.
[77] Liu S, Jiang S. High throughput screening and characterization of
HIV-1 entry inhibitors targeting gp41: theories and techniques.
Curr Pharm Design 2004; 10(15): 1827-43.
[78] Wolfl S, Burchert A, Kroll TC. Monitoring therapy with gene
expression profiling reveals physiological differences in drug
action. Curr Pharm Design 2004; 10(16): 1959-68.