Expression profiling reveals differences in metabolic gene
expression between exercise-induced cardiac effects and
maladaptive cardiac hypertrophy
Claes C. Strøm
1
, Mark Aplin
1
, Thorkil Ploug
2
, Tue E. H. Christoffersen
1
, Jozef Langfort
3
,
Michael Viese
2
, Henrik Galbo
2
, Stig Haunsø
1
and Søren P. Sheikh
1
1 CHARC (Copenhagen Heart Arrhythmia Research Center), Department of Medicine B, H:S Rigshospitalet, University of Copenhagen
Medical School, Denmark
2 Copenhagen Muscle Research Centre, Department of Medical Physiology, Panum Institute, University of Copenhagen, Denmark
3 Laboratory of Experimental Pharmacology, Polish Academy of Science, Warsaw, Poland
Keywords
adaptive; DNA microarray; gene expression;
hypertrophy; maladaptive
Correspondence
S. P. Sheikh, Laboratory of Molecular

including uncoupling protein 2 (UCP2) and fatty acid translocase (CD36).
DNA microarray analysis of gene expression changes in exercise-induced
cardiac hypertrophy suggests that a set of genes involved in fatty acid and
glucose metabolism could be fundamental to the beneficial phenotype of
exercise-induced hypertrophy, as these changes are absent or reversed in
maladaptive hypertrophy.
Abbreviations
ACE, angiotensin converting enzyme; ALP, actinin a2 associated LIM protein; EST, expressed sequence tag; FABP4, fatty acid binding
protein 4; FACL, fatty acid CoA ligase; FDR, false discovery rate; GCKR, glucokinase regulatory protein; HR, heart rate; LVEDP, left
ventricular end diastolic pressure; MAP, mean arterial pressure; MBE, model based expression; MYL, fast myosin alkali light chain; PCA,
principal component analysis; PDC, pyruvate dehydrogenase complex; PDP, pyruvate dehydrogenase phosphatase; Slc27a1, fatty acid
transport protein 4; UCP2, uncoupling protein 2.
2684 FEBS Journal 272 (2005) 2684–2695 ª 2005 FEBS
Heart disease is a leading cause of death in the West-
ern world and is commonly associated with cardiac
hypertrophy. Sustained cardiac hypertrophy leads to
cardiac dysfunction, heart failure, arrhythmia and
sudden death. As a result, cardiac hypertrophy is an
independent risk factor for cardiac morbidity and
mortality [1].
Exercise-induced cardiac hypertrophy is distinct
from the hypertrophy seen in different pathological
settings, as it is not accompanied by cardiac dysfunc-
tion or increased morbidity [2,3]. This intriguing dis-
tinction has led to the concepts of maladaptive and
adaptive forms of cardiac hypertrophy. While gene
expression changes in maladaptive cardiac hypertrophy
have been extensively investigated, much less is known
about transcriptional regulation in exercise-induced
hypertrophy. Identification of a set of genes unique to

induced cardiac hypertrophy, however, is lacking. Such
an approach may identify shared and divergent mole-
cular networks between adaptive and maladaptive
hypertrophy and point to new therapeutic strategies.
The microarray technology allows simultaneous
analysis of the expression level of thousands of genes
making this technology well suited for comprehensive
analysis of gene expression changes in response to phy-
siological challenges. DNA microarrays have been use-
ful in analysis of cellular responses to stimuli, animal
models of human disease and cancer classification
[13,14].
We used DNA microarrays to define gene expression
changes that characterize exercise-induced cardiac
hypertrophy. We identified 305 genes with differential
expression in response to cardiac exercise, the majority
of which have not previously been associated with
exercise. The most directly interpretable and poten-
tially biologically important finding was a reversed
metabolic shift in response to exercise suggesting that
genes involved in fatty acid and glucose metabolism
are key regulatory points that distinguish adaptive
beneficial hypertrophy from more adverse maladaptive
forms elicited by pathological stimuli.
Results
Physiological response to exercise
Several pieces of data indicated that exercised rats had
cardiac hypertrophy as compared to the sedentary con-
trol animals. First, training resulted in an  25%
increase in left and right ventricular masses when nor-

reduced. A less intense training protocol resulted in
significant body weight reductions but no increase in
ventricular weights (data not shown). Secondly, echo-
cardiographic examination of the cardiac phenotype
revealed that exercised rats had increases in both left
ventricular wall thickness and left ventricular cavity
dimensions (Table 2). Anterior and posterior wall
thicknesses were both increased by 14% and left ven-
tricular area indexed to lean body mass increased 9%.
Cardiac function at rest, as determined by fractional
area of change (Table 2), left ventricular end diastolic
pressure (LVEDP), and maximal rate of isovolumetric
pressure development and decay (Table 3), was identi-
cal in the two groups, which is consistent with pre-
vious findings [15]. Mean arterial pressure showed
no differences between exercised and sedentary rats
(Table 3), but resting heart rate decreased 10% in
response to exercise. The decrease in resting heart rate
probably results from an increase stroke volume and
increased parasympathetic tone.
Thus, our exercise protocol resulted in a phenotype
of eccentric hypertrophy without impairment of car-
diac function.
Distinct global gene expression profiles between
exercised and sedentary animals
We first analysed the data for differences in global
gene expression patterns between exercised and con-
trol animals using a principal component analysis
(PCA). This type of analysis serves to reduce the
number of variables in multivariate data with mini-

AWTd
(cm)
PWTd
(cm)
LVAd ⁄ BW
(mm
2
Æg
)0.78
)
FAC
(%)
Exercised 0.205 ± 0.007* 0.195 ± 0.007* 4.97 ± 0.12* 77 ± 1
Sedentary 0.180 ± 0.003 0.171 ± 0.004 4.57 ± 0.09 78 ± 1
*P < 0.05.
Table 3. Left ventricular pressures. Values are mean ± SEM.
dP ⁄ dt-max, Maximal rates of isovolumetric pressure development;
dP ⁄ dt-min, maximal rates of isovolumetric pressure decay.
LVEDP
(mm Hg)
dP ⁄ dt-max
(m HgÆs
)1
)
dP ⁄ dt-min
(mHgÆs
)1
)
MAP
(mm Hg)

trophy.
Expression of selected genes in maladaptive
hypertrophy
To compare expression of CD36 and UCP2 in adap-
tive and maladaptive hypertrophy we analysed expres-
sion of CD36 and UCP2 in the noninfarcted region of
the left ventricle 3 weeks after myocardial infarction
as compared to sham-operated animals. Contrary to
adaptive hypertrophy, where CD36 was upregulated
and UCP2 downregulated, CD36 expression was
unchanged and UCP2 expression increased (26%) in
maladaptive hypertrophy (Fig. 5).
Discussion
In this work, we present a comprehensive analysis of
transcriptional changes in response to exercise-induced
cardiac hypertrophy, thereby for the first time provi-
ding an overview of molecular clues to the adaptive
cardiac phenotype. We identified a distinct global gene
expression pattern of myocardium adapting to the
physiological challenge of exercise, and statistical ana-
lysis identified 267 upregulated and 62 downregulated
gene transcripts, providing a host of potential novel
diagnostic and therapeutic targets for further investiga-
tion.
The exercise resulted in a relatively small increase in
left ventricular mass (6%), which was in the same
range as that found by others after isotonic exercise
[10,16]. When normalized to body weight or tibial
length the increase in left ventricular mass was larger
and significant. Taken together with the fact that the

Fatty acid binding protein 4 AI169612 3.8 1.2 3.34E-03
Fatty acid Coenzyme A ligase, long chain 4 AI236284 5.1 1.3 5.17E-04
Cd36 AA925752 4.5 1.3 1.21E-03
Fatty acid transport protein U89529 3.2 1.2 1.88E-02
Uncoupling protein 2, mitochondrial AB010743 ) 4.7 0.7 8.25E-04
Glucokinase regulatory protein AA945442 4.4 1.1 1.72E-03
Pyruvate dehydrogenase phosphatase AF062740 3.3 1.4 8.97E-03
Solute carrier 16 (monocarboxylic acid transporter), member 1 D63834 4.3 1.4 1.78E-03
Hexokinase 1 AI012593 3.3 1.2 8.41E-03
Phosphofructokinase, liver, B-type X58865 3.8 1.1 3.61E-03
Extracellular matrix
Biglycan U17834 3.3 1.3 1.74E-02
Matrix Gla protein AI012030 3.2 1.3 1.01E-02
Integrin alpha 7 X65036 3.4 1.2 7.93E-03
Laminin receptor 1 D25224 5.3 1.2 3.67E-04
Cystatin B AI008888 3.9 1.2 3.25E-03
Cystatin C AI231292 5.9 1.4 1.54E-04
Cathepsin L AI176595 3.8 1.1 5.88E-03
Cathepsin S L03201 3.5 1.5 7.97E-03
Cytoskeletal
Sarcosin AI639444 3.2 1.4 8.74E-03
Fast myosin alkali light chain (MYL1) L00088 3.4 1.5 1.03E-02
Talin AA800962 3.3 1.2 8.57E-03
Actinin alpha 2 associated LIM protein AF002281 3.2 1.2 1.13E-02
Arg ⁄ Abl-interacting protein (ArgBP2) AF026505 4.1 1.3 2.48E-03
Myosin light chain alkali, smooth-muscle isoform (MYL6) AA875523 4.4 1.3 1.29E-03
Non-muscle myosin alkali light chain, new-born, heart ventricle (MYL4) S77858 4.1 1.2 2.08E-03
Actin-related protein complex 1b AF083269 3.5 1.3 5.85E-03
Growth
Eukaryotic translation elongation factor 1 alpha 1 AI008852 4.4 1.2 3.83E-03

sarcomeric proteins resembled those of maladaptive
hypertrophy. The most prominent difference from the
maladaptive response was differential expression of a set
of metabolic genes not previously associated with exer-
cise-induced cardiac hypertrophy. While downregula-
tion of genes involved in lipid oxidation is typical of
maladaptive hypertrophy, we found upregulation of sev-
eral of these genes in adaptive hypertrophy. Expression
levels of glycolytic enzymes indicated both enhanced
glycolysis and glucose oxidation to contrast the impair-
ments of glucose oxidation in maladaptive hypertrophy.
We also identified several differences in expression of
Fig. 3. A scatter plot of the number of differentially expressed
genes compared to the number of false-positive genes at different
levels of delta. The black line represents the actual data while the
three grey lines represent data from three random divisions of
samples into two groups. The dotted black line represents unity,
where the number of called genes is identical to the number of
false positives.
Fig. 4. Expression of selected metabolic genes by quantitative PCR
confirming the microarray data. Expression was normalized to
GAPDH. Bars represent SEM and *P < 0.05.
Table 4. (Continued).
Gene Accession number Score(d) FC P-value
Cathechol-O-methyltransferase M93257 4.2 1.3 5.89E-03
Guanine nucleotide binding protein, alpha inhibiting polypeptide 3 AI228247 3.3 1.1 8.47E-03
N-myristoyltransferase 1 AA859942 4.0 1.2 3.58E-03
ADRBK1 (GRK2) M87854 ) 4.3 0.9 2.83E-03
MAP-kinase phosphatase (cpg21) AF013144 ) 7.8 0.9 1.87E-04
Calcium ⁄ calmodulin-dependent protein kinase 1 D86556 ) 4.3 0.9 3.47E-03

[21]. Thus the differential expression of CD36 between
maladaptive and adaptive hypertrophy might be of key
importance for the difference in clinical outcome in the
two conditions.
Glucose utilization through glycolysis is enhanced in
hypertrophic hearts [22,23]. However, there is no cor-
responding increase in rates of glucose oxidation
[22,23]. The consequent low coupling of glucose oxida-
tion to glycolysis is functionally relevant, as it contri-
butes to the contractile dysfunction in hypertrophic
hearts [23]. The multienzyme pyruvate dehydrogenase
complex (PDC) catalyses the oxidative decarboxylation
of pyruvate and contributes strongly to flux control of
myocardial glucose oxidation. The activity of PDC is
continuously regulated by balance of inhibiting pyru-
vate dehydrogenase kinase and activating pyruvate
dehydrogenase phosphatase (PDP) reactions [24]. We
found upregulation of the PDP gene, thus, suggesting
an increased glucose oxidation in exercise-induced
hypertrophy. GCKR was upregulated; this has been
shown to increase both glucokinase protein and enzy-
matic activity levels, leading to improved glucose toler-
ance and lowered plasma glucose in diabetic mice [25].
In accordance with these data, we found upregulation
of glucokinase (hexokinase 1) in hearts of exercised
rats. Further evidence of enhanced glycolysis came
from the upregulation of 6-phosphofructo-2-kinase ⁄
fructose-2,6-bisphosphatase that stimulates 6-phospho-
fructo-1-kinase [26], a key enzyme of glycolysis, which
was also upregulated in our experiments. Collectively,

upregulation of UCP2 seems a general feature of mal-
adaptive cardiac remodelling, and the well documented
beneficial effects of ACE-inhibitors and b-adrenergic
receptor-blockade are accompanied by decreased UCP2
expression. These findings indicate that the downregula-
tion of UCP2 in adaptive hypertrophy constitutes a
molecular feature of ‘adaptiveness’ and that upregula-
tion of UCP2 may be a key factor underlying defective
energetics in diseased hearts.
In accordance with previous reports we did not find
activation of the typical neonatal gene expression pat-
tern found in pathological hypertrophy, which includes
uprelation of atrial natriuretic peptide, B-type natriure-
tic peptide, a-skeletal and smooth muscle actin, and
b-myosin heavy chain [16]. However, exercise-induced
hypertrophy was accompanied by a marked upregula-
tion of genes involved in extracellular matrix remode-
ling (biglycan, matrix gla protein, cathepsins, cystatins,
integrin a7 and laminin receptor). These genes are con-
sistently upregulated in pathological models of cardiac
hypertrophy indicating that these genes are necessary
to the cardiac growth response [18,31,32]. In contrast
to pathological models of cardiac hypertrophy we
found no increase in collagen mRNA expression.
We found upregulation of a number of cytoskeletal
genes. Several of these genes were previously described
to be upregulated in pathological hypertrophy
(MYL 1, 4 and 6, sarcosin, talin, actin-related protein
complex 1b and ArgBP2) [31,33]. Upregulation of acti-
nin a2 associated Lim11/rat Isl-1/Mec3 (LIM) protein

nucleotide 3¢-phosphodiesterase) [18,32,33]. Adrenergic
signalling is important in cardiac hypertrophy and we
found differential expression of several genes involved
in adrenergic signal transduction (catechol-O-methyl
transferase, GRK2, AKAP4 and Gai3). GRK2 desen-
sitizes G-protein coupled receptors and is upregulated
[37] in maladaptive hypertrophy. We found downregu-
lation of GRK2 in adaptive hypertrophy pointing to a
potentially important difference in adrenergic signal-
ling between maladaptive and adaptive hypertrophy.
In conclusion, we have used DNA microarrays to
map gene expression in adaptive hypertrophy. While
expression of extracellular matrix proteins and sarco-
meric proteins was similar to the changes known to
occur in maladaptive hypertrophy, we found striking
differences in expression of genes involved in metabo-
lism between adaptive and maladaptive hypertrophy.
Experimental procedures
Animal handling and training procedure
Twenty-four male Wistar rats (Taconic M & B, Ejby,
Denmark) weighing 285 ± 10 g (mean ± SD; n ¼ 24) were
randomly assigned to either a seven-week treadmill running
program (n ¼ 12) or served as sedentary controls (n ¼ 12).
The animals had free access to food (standard rodent pel-
lets) and water. Rats in the running group were exercised
on a custom-built 12-lane treadmill with an 8° inclination
for 2 hÆday
)1
, 5 daysÆweek
)1

)80 °C until mRNA extraction.
As exercise resulted in a significantly reduced body
weight (BW) when compared to sedentary controls, normal-
ization of organ weights to BW would result in apparent
hypertrophy of all organs in the training animals. For valid
comparison of experimental groups, organ weights were
instead normalized to lean body mass, estimated as BW
0.78
[38], rendering only weights of total heart and left and right
ventricles different between groups, while lung, kidney and
stomach were not.
Echocardiography
Echocardiography was performed during anesthesia with
1–1.5% isoflurane using a Vivid Five Echocardiograph
(GE Medical Systems Ultrasound, Little Chalfont, UK).
Recordings were stored digitally for off-line analysis. Left
ventricular cavity and wall dimensions were measured in 2D
short axis recordings at the level of the papillary muscles.
Hemodynamic examination
A microtip transducer catheter (Millar Instruments, Hous-
ton, TX, USA) was introduced from the right carotid
artery and placed in the left ventricle for measurements of
LVEDP and maximal rates of isovolumetric pressure devel-
opment (dP ⁄ dt
max
) and decline (dP ⁄ dt
min
). After retraction
from the left ventricle, mean arterial pressure (MAP) was
measured. Simultaneous elecrocardiography was performed

revstro90, password: revstro90).
Array data analysis
Array data were normalized using the nonlinear invariant
rank fitting method of Li and Wong available at http://
www.dchip.org [40]. Model based expression (MBE) values
were calculated for each gene using dChip (perfect
match only model). Differentially expressed genes were
identified using SAM available at http://www-stat.
stanford.edu ⁄ tibs ⁄ SAM ⁄ [41]. Briefly, SAM is a statistical
approach to identify differentially expressed genes by con-
trolling the FDR. The FDR is the percentage of genes iden-
tified by chance. SAM identifies the differentially regulated
genes by assimilating a set of gene specific t-tests. Each
gene is assigned a score by dividing the average difference
in gene expression between groups by the pooled SD.
Genes with scores greater than threshold delta (Fig. 2, grey)
are deemed potentially significant. By permutation of the
Table 5. Primer sets used in quantitative PCR. Sequences are shown in the 5¢)3¢ orientation.
Gene Forward Reverse Target position Product size
GAPDH GTCGGTGTGAACGGATTTG CTTGCCGTGGGTAGAGTCAT 859–1008 150
FABP4 GGAAAGTGAAGAGCATCATAACC ATGACACATTCCACCACCAG 289–412 124
CD36 GCAAAGAAGGGAAACCTGTG TCCAGTTATGGGTTCCACATC 1071–1207 137
FACL4 CCTGGATTAGGACCAAAGGA ATTTTGCTGGACTGGTCAGA 981–1126 146
GCKR TGCAGAGGTTCTCTGGACAGT GTGGGGATCACCTTTTCCTT 1589–1739 151
Slc27a1 CCACTCAGCAGGGAACATCA GGCATATTTCACCGATGTACTGC 950–1098 149
UCP2 GAAAGGGACCTCTCCCAATG GGAGGTCGTCTGTCATGAGG 872–987 116
Gene expression in exercise-induced cardiac hypertrophy C. C. Strøm et al.
2692 FEBS Journal 272 (2005) 2684–2695 ª 2005 FEBS
samples and recalculation of the scores, the FDR is estima-
ted at different values of delta. Log

ume of 20 lL with 1 lL of template. Standard curves in
duplicate were included in every run, and quantification of
individual samples performed by normalization to GAP-
DH. Constant GADPH expression between exercised and
sedentary animals was confirmed by northern blotting (data
not shown). At least three independent runs were per-
formed for every target transcript. The primer sets used in
quantitative PCR are shown in Table 5.
Statistical analysis
Array data were analysed as described above. All other
comparisons were made by an unpaired Student’s t-test.
P-values ¼ 0.05 were considered significant.
Acknowledgements
We thank the staff at the Microarray Center, Rigshos-
pitalet, Denmark, for performing the microarray hy-
bridizations and scannings. We thank Peter Schjerling
for Northern blots of GAPDH and Pernille Gundelach
and Katrine Kastberg for technical assistance. The
work was supported by the John and Birthe Meyer
Foundation, the Danish Heart Foundation (01-1-2-59-
22907, 99-1-2-31-22684), the Villadsen Family Founda-
tion, the Foundation of 17.12.1981, the University of
Copenhagen, Rigshospitalet, the Novo-Nordisk Foun-
dation and the Danish National Research Foundation.
References
1 Agabiti-Rosei E & Muiesan ML (1997) Prognostic
significance of left ventricular hypertrophy regression.
Adv Exp Med Biol 432, 199–205.
2 Ehsani AA, Hagberg JM & Hickson RC (1978) Rapid
changes in left ventricular dimensions and mass in

R2036.
11 Calderone A, Murphy RJ, Lavoie J, Colombo F &
Beliveau L (2001) TGF-beta(1) and prepro-ANP
mRNAs are differentially regulated in exercise-induced
cardiac hypertrophy. J Appl Physiol 91, 771–776.
12 Wisloff U, Loennechen JP, Falck G, Beisvag V, Currie
S, Smith G & Ellingsen O (2001) Increased contractility
and calcium sensitivity in cardiac myocytes isolated
from endurance trained rats. Cardiovasc Res 50, 495–
508.
C. C. Strøm et al. Gene expression in exercise-induced cardiac hypertrophy
FEBS Journal 272 (2005) 2684–2695 ª 2005 FEBS 2693
13 Golub TR, Slonim DK, Tamayo P, Huard C,
Gaasenbeek M, Mesirov JP, Coller H, Loh ML,
Downing JR, Caligiuri MA, Bloomfield CD & Lander
ES (1999) Molecular classification of cancer: class
discovery and class prediction by gene expression
monitoring. Science 286, 531–537.
14 Roberts CJ, Nelson B, Marton MJ, Stoughton R,
Meyer MR, Bennett HA, He YD, Dai H, Walker WL,
Hughes TR, Tyers M, Boone C & Friend SH (2000)
Signaling and circuitry of multiple MAPK pathways
revealed by a matrix of global gene expression profiles.
Science 287, 873–880.
15 Buttrick PM & Scheuer J (1987) Physiologic, biochem-
ical, and coronary adaptation to exercise conditioning.
Cardiol Clin 5, 259–270.
16 Jin H, Yang R, Li W, Lu H, Ryan AM, Ogasawara
AK, Van Peborgh J & Paoni NF (2000) Effects of exer-
cise training on cardiac function, gene expression, and

trophied hearts. Am J Physiol 267, H742–H750.
23 Wambolt RB, Lopaschuk GD, Brownsey RW & Allard
MF (2000) Dichloroacetate improves postischemic
function of hypertrophied rat hearts. J Am Coll Cardiol
36, 1378–1385.
24 Lydell CP, Chan A, Wambolt RB, Sambandam N,
Parsons H, Bondy GP, Rodrigues B, Popov KM,
Harris RA, Brownsey RW & Allard MF (2002) Pyru-
vate dehydrogenase and the regulation of glucose oxi-
dation in hypertrophied rat hearts. Cardiovasc Res 53,
841–851.
25 Slosberg ED, Desai UJ, Fanelli B, St Denny I, Connelly
S, Kaleko M, Boettcher BR & Caplan SL (2001) Treat-
ment of type 2 diabetes by adenoviral-mediated overex-
pression of the glucokinase regulatory protein. Diabetes
50, 1813–1820.
26 Beauloye C, Marsin AS, Bertrand L, Vanoverschelde JL,
Rider MH & Hue L (2002) The stimulation of heart
glycolysis by increased workload does not require
AMP-activated protein kinase but a wortmannin-sensi-
tive mechanism. FEBS Lett 531, 324–328.
27 Nicholls DG & Locke RM (1984) Thermogenic
mechanisms in brown fat. Physiol Rev 64, 1–64.
28 Gaussin V, Tomlinson JE, Depre C, Engelhardt S,
Antos CL, Takagi G, Hein L, Topper JN, Liggett SB,
Olson EN, Lohse MJ, Vatner SF & Vatner DE (2003)
Common genomic response in different mouse models
of beta-adrenergic-induced cardiomyopathy. Circulation
108, 2926–2933.
29 Murakami K, Mizushige K, Noma T, Tsuji T, Kimura

36 Strom CC, Kruhoffer M, Knudsen S, Stensgaard-Hansen F,
Jonassen TEN, Orntoft TF, Haunso S & Sheikh SP
Gene expression in exercise-induced cardiac hypertrophy C. C. Strøm et al.
2694 FEBS Journal 272 (2005) 2684–2695 ª 2005 FEBS
(2004) Identification of a core set of genes that signifies
pathways underlying cardiac hypertrophy. Comp Funct
Genom 5, 459–470.
37 Choi DJ, Koch WJ, Hunter JJ & Rockman HA (1997)
Mechanism of beta-adrenergic receptor desensitization
in cardiac hypertrophy is increased beta-adrenergic
receptor kinase. J Biol Chem 272, 17223–17229.
38 Batterham AM, George KP & Mullineaux DR (1997)
Allometric scaling of left ventricular mass by body
dimensions in males and females. Med Sci Sports Exerc
29, 181–186.
39 Theilade J, Strom C, Christiansen T, Haunso S &
Sheikh SP (2003) Differential G protein receptor kinase
2 expression in compensated hypertrophy and heart
failure after myocardial infarction in the rat. Basic Res
Cardiol 98, 97–103.
40 Li C & Hung Wong W (2001) Model-based analysis of
oligonucleotide arrays: model validation, design issues
and standard error application. Genome Biol 2,
RESEARCH0032 Epub.
41 Tusher VG, Tibshirani R & Chu G (2001) Significance
analysis of microarrays applied to the ionizing radiation
response. Proc Natl Acad Sci USA 98, 5116–5121.
Supplementary material
The following material is available from http://www.
blackwellpublishing.com/products/journal s/suppmat/