Article

A comprehensive analysis of prognostic signatures reveals the high predictive capacity of the Proliferation, Immune response and RNA splicing modules in breast cancer

Department of Pathology, The Netherlands Cancer Institute, Plesmanlaan 121, 1066 CX Amsterdam, The Netherlands.
Breast cancer research: BCR (Impact Factor: 5.49). 12/2008; 10(6):R93. DOI: 10.1186/bcr2192
Source: PubMed
ABSTRACT
Several gene expression signatures have been proposed and demonstrated to be predictive of outcome in breast cancer. In the present article we address the following issues: Do these signatures perform similarly? Are there (common) molecular processes reported by these signatures? Can better prognostic predictors be constructed based on these identified molecular processes?
We performed a comprehensive analysis of the performance of nine gene expression signatures on seven different breast cancer datasets. To better characterize the functional processes associated with these signatures, we enlarged each signature by including all probes with a significant correlation to at least one of the genes in the original signature. The enrichment of functional groups was assessed using four ontology databases.
The classification performance of the nine gene expression signatures is very similar in terms of assigning a sample to either a poor outcome group or a good outcome group. Nevertheless the concordance in classification at the sample level is low, with only 50% of the breast cancer samples classified in the same outcome group by all classifiers. The predictive accuracy decreases with the number of poor outcome assignments given to a sample. The best classification performance was obtained for the group of patients with only good outcome assignments. Enrichment analysis of the enlarged signatures revealed 11 functional modules with prognostic ability. The combination of the RNA-splicing and immune modules resulted in a classifier with high prognostic performance on an independent validation set.
The study revealed that the nine signatures perform similarly but exhibit a large degree of discordance in prognostic group assignment. Functional analyses indicate that proliferation is a common cellular process, but that other functional categories are also enriched and show independent prognostic ability. We provide new evidence of the potentially promising prognostic impact of immunity and RNA-splicing processes in breast cancer.

Full-text

Available from: Andrew E Teschendorff
Open Access
Available online http://breast-cancer-research.com/content/10/6/R93
Page 1 of 15
(page number not for citation purposes)
Vol 10 No 6
Research article
A comprehensive analysis of prognostic signatures reveals the
high predictive capacity of the Proliferation, Immune response
and RNA splicing modules in breast cancer
Fabien Reyal
1,2,3
, Martin H van Vliet
4,5
, Nicola J Armstrong
4
, Hugo M Horlings
1
, Karin E de Visser
6
,
Marlen Kok
1
, Andrew E Teschendorff
7
, Stella Mook
1
, Laura van 't Veer
1
, Carlos Caldas
7
,
Remy J Salmon
3
, Marc J van de Vijver
1,8
and Lodewyk FA Wessels
4,5
1
Department of Pathology, The Netherlands Cancer Institute, Plesmanlaan 121, 1066 CX Amsterdam, The Netherlands
2
Department of Surgery, Institut Curie, 6 rue d'Ulm, 75005 Paris, France
3
UMR 144, CNRS-Institut Curie, Molecular Oncology Team, 12 rue Lhomond, 75005 Paris, France
4
Bioinformatics and Statistics Group, The Netherlands Cancer Institute, Plesmanlaan 121, 1066 CX Amsterdam, The Netherlands
5
Faculty of EEMCS, Delft University of Technology, Mekelweg 4, 2628 CD Delft, The Netherlands
6
Department of Molecular Biology, The Netherlands Cancer Institute, Plesmanlaan 121, 1066 CX Amsterdam, The Netherlands
7
Cancer Research UK, Cambridge Research Institute and Department of Oncology, Li Ka Shing Centre, University of Cambridge, Cambridge CB2
ORE, UK
8
Department of Pathology, Academic Medical Center, Meibergdreef 9, 1100 DD Amsterdam The Netherlands
Corresponding author: Lodewyk FA Wessels, l.wessels@nki.nl
Received: 29 Mar 2008 Revisions requested: 1 May 2008 Revisions received: 31 Jul 2008 Accepted: 13 Nov 2008 Published: 13 Nov 2008
Breast Cancer Research 2008, 10:R93 (doi:10.1186/bcr2192)
This article is online at: http://breast-cancer-research.com/content/10/6/R93
© 2008 Reyal et al.; licensee BioMed Central Ltd.
This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0
),
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Abstract
Introduction Several gene expression signatures have been
proposed and demonstrated to be predictive of outcome in
breast cancer. In the present article we address the following
issues: Do these signatures perform similarly? Are there
(common) molecular processes reported by these signatures?
Can better prognostic predictors be constructed based on
these identified molecular processes?
Methods We performed a comprehensive analysis of the
performance of nine gene expression signatures on seven
different breast cancer datasets. To better characterize the
functional processes associated with these signatures, we
enlarged each signature by including all probes with a
significant correlation to at least one of the genes in the original
signature. The enrichment of functional groups was assessed
using four ontology databases.
Results The classification performance of the nine gene
expression signatures is very similar in terms of assigning a
sample to either a poor outcome group or a good outcome
group. Nevertheless the concordance in classification at the
sample level is low, with only 50% of the breast cancer samples
classified in the same outcome group by all classifiers. The
predictive accuracy decreases with the number of poor
outcome assignments given to a sample. The best classification
performance was obtained for the group of patients with only
good outcome assignments. Enrichment analysis of the
enlarged signatures revealed 11 functional modules with
prognostic ability. The combination of the RNA-splicing and
immune modules resulted in a classifier with high prognostic
performance on an independent validation set.
Conclusions The study revealed that the nine signatures
perform similarly but exhibit a large degree of discordance in
prognostic group assignment. Functional analyses indicate that
proliferation is a common cellular process, but that other
functional categories are also enriched and show independent
prognostic ability. We provide new evidence of the potentially
promising prognostic impact of immunity and RNA-splicing
processes in breast cancer.
Introduction
Breast cancer is composed of distinct diseases with different
outcomes. Clinical and pathological factors are currently
employed to determine the prognosis of patients. The Saint-
BCSS: breast cancer-specific survival; DMFS: distant metastasis-free survival; ER: estrogen receptor; HR: hazard ratio; LN: lymph node; PCR:
polymerase chain reaction; RT: reverse transcriptase.
Page 1
Breast Cancer Research Vol 10 No 6 Reyal et al.
Page 2 of 15
(page number not for citation purposes)
Gallen guidelines [1], National Institute of Health guidelines
[2] and Nottingham Prognostic Index guidelines [3] as well as
the AdjuvantOnline! decision-making tools [4] use a combina-
tion of these prognostic factors to adapt adjuvant treatment
based on the prognosis prediction . Owing to insufficiently
accurate prognosis predictions, however, a substantial pro-
portion of breast cancer patients with breast cancer of inher-
ently good outcome receive adjuvant systemic therapy without
gaining any benefit [5].
High-throughput technologies such as gene expression micro-
arrays can offer new opportunities to improve the ability to
determine individual prognosis. After confirmation of their per-
formance in validation studies, classifiers such as the 70-gene
signature developed at the Netherlands Cancer Institute [6,7]
and the 76-gene signature developed at the Erasmus Medical
Center [8,9] may become of use in clinical practice. Similarly,
Paik and colleagues built a 16-gene classifier for paraffin-
embedded tissues (OncotypeDX
©
; Genomic Health, Red-
wood, California, USA) using quantitative RT-PCR [10]. The
ongoing MINDACT trial [11] and TAILORx trial [12] aim to
confirm the performance of the 70-gene and the OncotypeDX
signatures on large populations of 6,000 patients and 10,000
patients, respectively. In the meantime, it has become com-
monplace for research groups to define new classifiers with a
potentially higher level of performance than existing classifiers
[13-21].
Publications have raised several concerns about microarray
studies that would potentially impair the use of gene expres-
sion classifiers in clinical routine [22-24]. We performed a
comprehensive analysis of gene-expression-based classifiers
on a collection of 1,127 breast cancer samples all hybridized
on the Affymetrix
©
platform. To the best of our knowledge, this
represents the largest, single platform dataset employed to
evaluate prognostic classifiers to date.
Nine gene expression signatures derived using diverse meth-
odological approaches and focusing on various aspects
thought to be associated with breast cancer outcome were
selected: the 76-gene signature employed to compute the
Relapse score [8]; the 52-gene and 17-gene signatures
employed to compute the Molecular Prognostic Index (T52
and T17) [20]; the Intrinsic/UNC gene set used to define the
molecular subtypes [15]; the 70-gene and 25-gene Chromo-
somal Instability signatures (CIN70 and CIN25) [13]; the Core
Serum Response signature [14]; the Invasiveness Gene Sig-
nature [16]; and the 97-gene signature used to derive the
Gene expression Grade Index [25]. (Note that there are nine
signatures in total, but T17 and CIN25 are smaller versions of
T52 and CIN70, respectively. T17 and CIN25 were therefore
not included in the enlargement and enrichment analyses.)
Two well-known signatures, the 21-gene Genomic Health sig-
nature [10] and the 70-gene signature from the Netherlands
Cancer Institute [6,7], were not included. The Genomic Health
signature was not included since it is not a micro-array-based
gene expression signature but is designed for RT-PCR assays.
The 70-gene signature was not included since the 295 sample
dataset on which this signature was developed was employed
as a completely independent validation set. Throughout the
text we use signature to refer to both the gene set as well as
the classifier, which, based on this gene set, assigns a sample
to an outcome group.
While some signatures were designed for specific subgroups
of patients, we applied the signatures to the complete, unstrat-
ified set of breast cancer samples. We followed this approach
for several reasons. First, a large dataset maximizes power to
detect common prognostic factors, while stratification into
subtypes results in smaller groups and, therefore, in reduced
power. Second, there is mounting evidence that classifiers
designed within a specific subgroup also perform well in other
subgroups. For example, the 70-gene classifier developed by
van 't Veer and colleagues was developed for lymph-node-
negative patients [6], but performs very well on lymph-node-
positive patients as demonstrated by van de Vijver and col-
leagues [7]. In fact, patients with up to three positive lymph
nodes are now included in the MINDACT trial [26]. Similarly,
the 21-gene OncotypeDX signature, which was developed for
node-negative disease, has also been shown to be a good pre-
dictor of disease outcome in node-positive patients [27].
These facts indicate that breast cancer subgroups as defined
by classical markers are currently being re-evaluated, and that
the signatures have some common core of biological/molecu-
lar processes involved in the development of breast cancer
that is critical for predicting good outcome or poor outcome.
Finally, an analysis performed on the complete dataset pro-
vides a benchmark against which subtype specific signature
performance can be gauged and also reveals whether predic-
tors designed on the whole cohort have any predictive value in
subgroups.
We addressed the following issues: Do these nine separate
signatures perform similarly? Are there (common) molecular
processes reported by these signatures? Can better prognos-
tic predictors be constructed based on (a combination of)
these identified molecular processes?
Materials and methods
Figure 1 depicts a graphical overview of the analysis proce-
dure followed. A detailed Materials and methods section is
available in Additional data file 1.
Data preprocessing
Six breast cancer datasets [9,17-19,28,29] – all arrayed on
the same platform (HGU-133A Affymetrix
©
Santa Clara, Cali-
fornia, USA) to avoid cross-platform discrepancies – for which
the raw data (.CEL files) were publicly available were down-
loaded from the Gene Expression Omnibus and ArrayExpress
repository websites [30,31] . This resulted in a total of 1143
Page 2
Available online http://breast-cancer-research.com/content/10/6/R93
Page 3 of 15
(page number not for citation purposes)
microarrays. Of these, 1,127 were deemed to be of sufficient
quality and were kept in the current study. Sixteen micro-arrays
were rejected due to major artifacts in the original .CEL files.
To ensure comparability between the different datasets, they
were all subjected to the same preprocessing procedure.
Microarray quality-control assessment was carried out using
the AffyPLM R-package [32]. Selected arrays were normalized
using the RMA expression measure algorithm [33]. For 724
samples, distant metastasis-free survival (DMFS) data were
available; while for 403 samples, breast cancer-specific sur-
vival (BCSS) data were available. For some samples both
DMFS and BCSS information is available. For the perform-
ance analyses, where the results per endpoint are reported
separately, the overlapping samples were analyzed twice
(once for each endpoint).
For the enlargement of the signatures and the construction of
the functional classifiers, only the Chin and Loi datasets were
used – resulting in a total of 450 samples. We will refer to this
combined dataset as the Chin–Loi training set. Clinical and
histopathological features for all samples are summarized in
Table 1.
Figure 1
Detailed overview of the datasets, gene sets and analysis steps employedDetailed overview of the datasets, gene sets and analysis steps employed. PPV, positive predictive value; NPV, negative predictive value; PA,
predictive accuracy.
Page 3
Breast Cancer Research Vol 10 No 6 Reyal et al.
Page 4 of 15
(page number not for citation purposes)
Signature validation
To the complete dataset of 1,127 samples, we applied the 76-
gene signature [8], the Intrinsic/UNC signature [15], the Chro-
mosomal Instability signatures (CIN70 and CIN25) [13], the
Core Serum Response signature [14], the Invasiveness Gene
Signature [16], the Molecular Prognosis Index signatures (T52
and T17) [20], and the Gene expression Grade Index signa-
ture [25]. Survival analyses were performed using the Kaplan–
Meier estimate of the survival function. Comparison between
survival curves was performed using the log-rank test. Hazard
ratios were estimated using the Cox proportional hazard
model. P values were considered significant when <0.05.
Only variables with a significant P value in univariate analyses
were included in a multivariate model. Time-censoring analy-
ses were performed using right censoring of the events from 1
to 12 years. The performance analysis (sensitivity, specificity,
positive predictive value, negative predictive value, and predic-
tive accuracy) of the signatures was carried out using the
ROCR package [32] . For these analyses, the outcome was
dichotomized into a poor outcome group (samples with an
event before 5 years of follow-up) and a good outcome group
(samples with no event and a follow-up of at least 5 years). The
analyses were performed using R software [32] and Matlab
software [34].
Enlarged signatures
We created enlarged signatures by including all probes with
an absolute Spearman rank order correlation >0.7 with at least
one of the genes in the original signatures. This calculation
was performed on the Chin–Loi training set.
Table 1
Clinical and histopathological features of 1,127 patients presenting breast carcinoma and included in the survival analyses
Distant metastasis-free survival Breast cancer-specific survival
Patients (n) 724 403
Tumor size
Number (%) of T1 ( 20 mm) 292 (43%) 271 (69%)
Mean (standard deviation) size (mm) 25.3 (13) 22.2 (10.4)
Histological grading: Elston Ellis
Grade I 104 (15.4%) 88 (23%)
Grade II 253 (37.5%) 176 (45%)
Grade III 195 (28.9%) 111 (28%)
Not available 121 (17.9%) 13 (3%)
Genomic grade index
High grade 331 (49%) 179 (46%)
Estrogen receptor status: immunohistochemistry
Positive 488 (72.5%) 197 (51%)
Not available 6 (0.8%) 160 (40%)
Estrogen receptor status: gene expression
Positive 496 (74%) 297 (76%)
Her2 status: gene expression
Positive 86 (13%) 50 (13%)
Lymph node metastasis
Positive 197 (28%) 76 (20%)
Not available 8 (1%) 165 (41%)
Distant metastasis or death from breast cancer
Positive 177 (26%) 83 (21%)
Within 5 years 131 (19%) 59 (15%)
Mean (standard deviation) follow-up (months) 85.6 (53.4) 89.6 (39.8)
Datasets used (Ref) [9,17,19,29] [18,28]
Sixty-six patients out of 1,127 samples were excluded from the survival analysis due to missing information (time or event censoring).
Page 4
Available online http://breast-cancer-research.com/content/10/6/R93
Page 5 of 15
(page number not for citation purposes)
Enrichment analyses of signatures
For all signatures, we evaluated whether specific gene sets
(that is, functional groups), are overrepresented. We gathered
a collection of 5,480 gene sets from four databases: Gene
Ontology [35], Kyoto Encyclopedia of Genes and Genomes
[36], Reactome [37] and the Molecular Signatures Database
[38] (see Materials and methods in Additional data file 1) . The
hypergeometric test was employed to test the significance of
the overlap between each signature and gene set. Multiple
testing adjustment was performed by applying the Benjamini
Hochberg correction (per signature) [39].
Discovery and validation of module classifiers
The gene sets that were enriched in at least one signature
were grouped into modules based on common functional
annotation. The only exception was the group of proliferation-
related gene sets. This group was split into two groups, the
first containing gene sets common to either five or six enlarged
signatures while the second contained the remaining prolifer-
ation gene sets. This resulted in 11 functional modules. For
each of these modules we constructed a nearest mean classi-
fier using the genes common between the enlarged signatures
and the group of genes in the module. The classifier was
trained using the Chin–Loi training set. For each module, the
poor outcome centroid was derived from samples with a met-
astatic event before 5 years of follow-up; the good outcome
centroid was derived from samples with no metastatic events
and a follow-up longer than 5 years. Based on the genes asso-
ciated with a given module, the Spearman rank correlation
between each sample and the centroid values of the poor out-
come and the good outcome centroids were determined.
Each sample in the training set was then assigned to the cen-
troid with which it showed the highest correlation (the nearest
centroid).
We then created classifiers by combining the output of every
pair of module classifiers. For each pair of modules, this results
in a three-class classifier: samples with two poor (good) group
labels are assigned to the poor (good) group, while samples
with discordant labels are assignment to the intermediate
group. The performance of these classifiers was evaluated,
based on the log-rank test, on the time-to-event data on the
Chin–Loi training set. The best module-pair classifier was
selected to enter the validation stage.
We validated the best module-pair classifier on independent
data. Validation was performed for DMFS on the datasets of
Desmedt and colleagues [9] and Minn and colleagues [19],
while for BCSS the datasets of Pawitan and colleagues [28]
and Miller and colleagues [18] were used . In addition, we val-
idated the best module-pair classifier on the dataset of van de
Vijver and colleagues [7]. The gain in predictive accuracy of
the classifier, as compared with common clinical staging sys-
tems, was investigated on the dataset of van de Vijver and col-
leagues using the method of Schemper and Henderson [40],
implemented in the R package software [41] . Dunkler and col-
leagues have also recently used this approach in the context
of gene signatures [42]. Briefly, predictive inaccuracy is calcu-
lated as the average of the absolute difference between
observed outcomes and the model predictions. The explained
variation is also computed and represents a measure equiva-
lent to R
2
in linear regression. Standard errors were obtained
by bootstrapping 200 resamples.
Results
Prognostic performance
Cox proportional hazard analyses using clinical and histologi-
cal features with DMFS and BCSS as endpoints are summa-
rized in Additional data file 1. Nine gene signatures were
applied to the complete dataset of 1,127 breast cancer sam-
ples (Figure 2). Each of the signatures was used to define a
poor outcome group and a good outcome group. For both clin-
ical endpoints (DMFS, BCSS), the Kaplan–Meier curves
obtained for each of the nine signatures were very similar – the
exception being the Intrinsic/UNC gene set, which, per defini-
tion, defines five molecular subtypes (Figure 2a; see also Addi-
tional data file 2, Figure S1a). The log-rank test for differences
in survival between the different groups as classified by the
individual signatures was significant (P < 0.05) in all cases.
In addition, multivariate Cox proportional hazard models were
fitted with lymph node (LN) status, size, estrogen receptor
(ER) status (DMFS only) and each signature separately (see
Additional data file 1, Table S2). All nine signatures remained
as significant variables in the multivariate models for DMFS.
Tumor size was the most significant, in terms of P value, in all
models except for those including the Intrinsic/UNC gene set
signature and the 76-gene signature. When considering
BCSS, three signatures remained significant: Intrinsic/UNC
gene set (P = 0.01), Invasiveness Gene Signature (P = 0.02)
and Core Serum Response signature (P = 0.039). The clinical
variables LN and size were significant in all models
considered.
The data were then divided into subsets according to LN sta-
tus, ER status, grade and origin, and the signatures were com-
pared in terms of their association with survival in these clinical
subgroups (Figure 2b (top panel); see also Additional data file
2, Figure S1b (top panel)). With DMFS as the endpoint, no
signature was significantly associated with survival in the ER-
negative subgroup. In the LN-positive subgroup, some signa-
tures were significantly associated with survival – while in the
high-grade subgroup, only a single signature was significantly
associated with DMFS. The sample sizes in these subgroups
are relatively small (<200 patients); however, the individual
hazard ratios (HRs) for the signatures within these subgroups
are of similar magnitude. For BCSS, only the ER-positive sub-
group showed association and was the largest subgroup avail-
able for this endpoint (n = 297). Other clinical subgroups in
this analysis have small sample sizes (between 80 and 160
Page 5
Breast Cancer Research Vol 10 No 6 Reyal et al.
Page 6 of 15
(page number not for citation purposes)
patients). We also applied the signatures to the individual
datasets, and found that no signature showed association with
survival in the Minn dataset (Figure 2b (top panel); see also
Additional data file 2, Figure S1b (top panel)). For this dataset,
only 80 samples had survival information; however, all signa-
tures had HRs close to 1.
The sensitivity of the signatures to variations in the time after
which all events are censored was also investigated. Over a
censoring range of 1 year to 12 years, each of the nine signa-
tures showed a very similar pattern (Figure 2b (bottom panel);
see also Additional data file 2, Figure S1b (bottom panel)) spe-
cific to the clinical endpoint of interest. With DMFS as the end-
point, the performance measure reached a maximal value at
approximately 5 years. This might be explained by the fact that
some of the signatures were generated using a 5-year cutoff
point to define the good and poor outcome groups. Two vali-
dation series of the 76-gene signature [9] and of the 70-gene
signature [43] also showed strong time-dependent perform-
ance variability. The hazard ratio adjusted for clinical parame-
ters, however, was still significant for the censoring time
ranging from 1 year to 10 years.
Classification concordance
For each signature, every sample was labeled either good out-
come or poor outcome. Figure 3a provides an overview of the
assignments of the samples to either outcome group by all sig-
natures for DMFS (left panel) and for BCSS (right panel).
Figure 2
Kaplan–Meier analysis of distant metastasis-free survivalKaplan–Meier analysis of distant metastasis-free survival. (a) Analysis performed for 724 patients. In all cases except the Intrinsic/UNC subtyp-
ing gene set (HU) the blue (red) curve represents the good (poor) outcome group. Each panel depicts the result obtained after applying one of the
classifiers (indicated in the heading) as described in Materials and methods. Represented signatures: T17 and T52, Molecular Prognostic Index sig-
nature [20]; IGS, Invasiveness Gene signature [16]; HU, Intrinsic/UNC gene set [15] – resulting in the Luminal A (blue), Normal-like (coral), Luminal
B (red), Basal (grey) and Her2 (purple) subtypes – CSR, Core Serum Response signature [14]; CIN70 and CIN25, Chromosomal Instability signa-
ture [13]; GGI, Gene expression Grade Index [25]; and Wang, 76-gene signature [8]. (b) Top panel: performance of the signatures on subgroups of
the patient population. Vertical axis, -log(P value) of the log-rank test from the Kaplan–Meier analysis for a particular subgroup with distant metasta-
sis-free survival (DMFS) as the endpoint. Analyzed subgroups: lymph node (LN)-negative, LN-positive, estrogen receptor (ER)-positive, ER-negative,
low grade (Elston Ellis I), high grade (Elston Ellis III), individual Chin, Desmedt, Loi and Minn datasets. Horizontal line, P = 0.05. Bottom panel: time-
censoring performance analysis of the signatures. Horizontal axis, time at which right censoring was applied to all samples; vertical axis, -log(P value)
of the log-rank test from the Kaplan–Meier analysis for a given time-censoring and a particular signature with DMFS as the endpoint. Horizontal line,
P = 0.05.
Page 6
Available online http://breast-cancer-research.com/content/10/6/R93
Page 7 of 15
(page number not for citation purposes)
We investigated the concordance of the classification labels
for the nine signatures. Considering DMFS as the endpoint,
only 322 out of 724 samples (44%) had perfect concordance:
127 samples (17.5%) were consistently classified as good
outcome, while 195 samples (26.9%) were classified as poor
outcome by all signatures. These samples are indicated by the
shaded regions in Figure 3a. A total of 114 samples (15.7%)
proved difficult to classify since more than one-third of the sig-
natures assigned discordant labels to these samples. When
BCSS was taken as the endpoint (n = 403), similar results
were found: 198 (49%) had perfect concordance, with 86
samples (21.3%) always classified as good outcome and 112
samples (27.8%) always classified as poor outcome by the
signatures. A number of samples (46 samples, 11.4%) were
again difficult to classify, with more than one-third of the signa-
tures assigning discordant labels to them. In order to investi-
Figure 3
Classification and stratification of poor outcome for distant metastasis-free survival and breast cancer-specific survivalClassification and stratification of poor outcome for distant metastasis-free survival and breast cancer-specific survival. (a) Classification
results of the nine gene signatures on the collection of 1,127 samples with distant metastasis-free survival (DMFS) (left panel) and breast cancer-
specific survival (BCSS) (right panel) as the endpoints. In both panels, breast cancer samples are depicted in columns, and signatures in rows. Each
cell colored according to the outcome of the signature in the corresponding row for the patient in the associated column: blue, poor outcome; yel-
low, good outcome. Top three lanes represent clinical parameters. HER2, HER2 status with red representing amplified cases and green nonampli-
fied cases (as determined from gene expression); ER, estrogen receptor status with red representing ER-negative cases and green ER-positive
cases as determined by gene expression; Meta or Death, DMFS or BCSS with events indicated in black. Represented signatures: HU, Intrinsic/UNC
gene set [15]; IGS, Invasiveness Gene signature [16]; CIN70 and CIN25, Chromosomal Instability signatures [13]; GGI, Gene expression Grade
Index [25]; T17 and T52, Molecular Prognostic Index signature [20]; CSR, Core Serum Response signature [14]; and Wang, the 76-gene signature
[8]. For the subtyping based on the Intrinsic/UNC gene set, assignment to the Luminal B, Basal and HER2 subtypes was scored as poor outcome
while assignment to the Luminal A and Normal types was scored as good outcome. Shaded regions highlight those patients classified as good out-
come or poor outcome by all signatures. (b) Kaplan–Meier curves for the population of 1,127 samples stratified according to the number of times a
sample is assigned to the poor outcome category with DMFS (left panel) and BCSS (right panel) as the endpoints. Graphs depict the curves for four
subgroups: all good group (no poor outcome assignments; green), mainly good group (one to three poor outcome assignments; blue), mainly poor
group (four to six poor outcome assignments; purple), and poor group (seven to nine poor outcome assignments; black).
Page 7
Breast Cancer Research Vol 10 No 6 Reyal et al.
Page 8 of 15
(page number not for citation purposes)
gate the effect of sample heterogeneity on the discordance,
we repeated this analysis for subgroups based on ER status,
LN status, and their combinations. The results are depicted in
Additional data file 3. This analysis revealed that the concord-
ance/discordance percentages within these subgroups are
always similar. That is, the discordance ranged from 52% to
62% for DMFS and from 44% to 60% for BCSS, leading to
the conclusion that the concordance/discordance is not due
to these clinical parameters.
Several clinical parameters are associated with the number of
poor outcome assignments. The results for DMFS are
depicted in Figure 4, while the results for BCSS are presented
in Additional data file 4. In both cases, the proportion of events
was significantly correlated with the number of times each
sample was assigned to the poor outcome group (DMFS, chi-
square test, P = 1.9 × 10
-4
; BCSS, chi-square test, P = 6.1 ×
10
-10
). Similarly, the proportion of patients with ER-negative
status (DMFS, P = 2.2 × 10
-16
; BCSS, P = 2 × 10
-14
), with
HER2-positive status (DMFS, P = 8.5 × 10
-5
; BCSS, P =
0.0001), in the high-grade subgroup (DMFS, P = 2.2 × 10
-16
;
BCSS, P = 2.2 × 10
-16
) and with a mean tumor size (DMFS,
P = 0.0004; BCSS, P = 2.4 × 10
-5
) were all correlated to the
number of times a sample was classified as poor outcome by
the nine signatures.
To further investigate the impact of the number of poor out-
come classifications, the samples were divided into four cate-
gories: all good (no poor outcome), mainly good (one to three
poor outcomes), mainly poor (four to six poor outcomes), and
poor (seven or more poor outcomes). Using this grouping,
Kaplan–Meier analysis was performed using DMFS as the
endpoint (Figure 3b (left panel), log-rank, P = 5.1 × 10
-7
). A
univariate Cox proportional hazard model showed significantly
increased HRs when comparing the all good group with the
other groups (versus mainly good group, HR = 2.15 (1.11 to
4.14); versus mainly poor group, HR = 3.49 (1.73 to 7.01);
versus poor group, HR = 4.12 (2.27 to 7.49)). Using BCSS as
the endpoint, the log-rank test was again significant (P = 9.4
× 10
-7
; Figure 3b (right panel)) and the HRs increased signifi-
cantly when comparing the all good group with the other
groups (data not shown).
The sensitivity, specificity, positive predictive value, negative
predictive value and predictive accuracy were very similar for
all signatures – except for the Core Serum Response signa-
ture, which showed higher sensitivity and lower specificity
(see Additional data file 1). The predictive accuracy (fraction
of correctly assigned samples) of each signature was analyzed
as a function of the number of times a sample was assigned to
the poor outcome group. For both endpoints, the predictive
accuracy is very high for the samples always classified as
good outcome (DMFS, 87%; BCSS, 96%) and decreases
dramatically with the number of poor outcome assignments.
The predictive accuracy for the samples with nine poor out-
come labels was only 45% for DMFS and 36% for BCSS (see
Figure 4 and Additional data file 4).
Enlarged gene signature analysis
The intersection of seven gene signatures showed that no
probe set was present in more than four signatures and that
most of the probe sets were found in only one signature. In
order to better reveal the common processes associated with
the signatures, each of the signatures was enlarged by aug-
menting them with highly correlated probes (absolute Spear-
man correlation >0.7). These enlarged signatures contain, on
average, three times as many probes as the original signa-
tures. We identified a group of 72 probe sets present in at
least five enlarged gene signatures. This overlap is highly sig-
nificant, since observing an overlap of 72 or more probes
between the two enlarged signatures with the smallest
number of genes (164 genes, enlarged Invasiveness Gene
Signature; and 189 genes, enlarged Gene expression Grade
Index signature) is already highly significant (P <2.2 × 10
-16
,
hypergeometric test). The chance of observing an overlap of
72 or more probes among all seven enlarged signatures will
be even smaller, and thus even more significant. An enrich-
ment analysis of these 72 probe sets showed an overrepre-
sentation of genes related to DNA replication, cell cycle, and
proliferation.
Constructing and validating module classifiers
Functionally related gene sets were merged to define func-
tional modules. This resulted in the following 11 modules:
Immune, KRAS, Proliferation1 (defined by genes common to
two to four enlarged signatures), Proliferation2 (defined by
genes common to five or more enlarged signatures), RNA
splicing, Rb pathway, Sterol biosynthesis, Extracellular matrix
constituent (ECM), Focal adhesion, Negative regulation of
proliferation, and Apoptosis (Figure 5a). For each of these
modules we determined a nearest centroid classifier on the
Chin–Loi training set. On this training set, with DMFS as the
endpoint, all modules except Sterol biosynthesis had a signifi-
cant association with outcome (log-rank test P value ranging
from 4.4 × 10
-3
to 6.8 × 10
-9
; Figure 5b).
We then set out to construct a classifier that combines the
binary class assignments of the module classifiers into a sin-
gle, three-valued (low-risk, intermediate-risk and high-risk)
classification output. We restricted the search to pairs of mod-
ules, since the size of the discordant groups (where at least
one module classifier output differs from the others) grows as
the number of modules combined increases, leading to diffi-
culty in interpretability. The separate classifiers were com-
bined in a pairwise fashion (55 possible combinations) and the
resulting classifiers were evaluated on the Chin–Loi training
set. The classifier with the most significant log-rank test (low-
est P value) on the training data was selected. This classifier
was based on the Immune and RNA splicing modules (see
Additional data files 5 and 6). Samples that are assigned to the
Page 8
Available online http://breast-cancer-research.com/content/10/6/R93
Page 9 of 15
(page number not for citation purposes)
good outcome (poor outcome) category by both the Immune
and the RNA splicing module classifiers are classified as low
(high) risk, while the discordant cases are assigned to the
intermediate risk category. On independent data, this classifier
was a significant prognosis predictor (data not shown) of
DMFS (dataset of Desmedt and colleagues [9]) and BCSS
(datasets of Pawitan and colleagues [28] and Miller and col-
leagues [18]).
The Immune and RNA splicing classifier was then validated
using the dataset of van de Vijver and colleagues [7]. For both
DMFS and BCSS, we found a highly significant association
between the classifier and outcome (log-rank test: DMFS, P =
2.28 × 10
-8
; BCSS, P = 1.8 × 10
-10
) (Figure 5c). For DMFS,
the HR for the intermediate-risk group versus the low-risk
group is 2.25 (1.3 to 3.9), while the HR for the high-risk group
versus the low-risk group is 4.35 (2.57 to 7.36). Similarly, for
BCSS the HR of the intermediate-risk group versus the low-
risk group was 2.31 (1.2 to 4.5) and the HR for the high-risk-
Figure 4
Overlap and performance analysis of 724 samples with distant metastasis-free survival as the endpointOverlap and performance analysis of 724 samples with distant metastasis-free survival as the endpoint. (a) Distribution of the samples as a
function of the number of times a sample was classified as of poor prognosis by the gene signatures. (b) Distribution of metastasis events as a func-
tion of the number of times a sample was classified as of poor prognosis by the gene expression signatures. (c) Distribution of estrogen receptor
(ER)-negative samples as a function of the number of times a sample was classified as of poor prognosis by the gene signatures. (d) Distribution of
HER2 amplified samples as a function of the number of times a sample was classified as of poor prognosis by the gene signatures. (e) Tumor size
(mean, 5th, 25th, 75th and 95th percentiles) as a function of the number of times a sample was classified as of poor prognosis by the gene signa-
tures. (f) Distribution of grade I, grade II and grade III samples (Elston Ellis) as a function of the number of times a sample was classified as of poor
prognosis by the gene signatures. Dark shading, grade I; grey shading, grade II; pale shading, grade III. (g) Average predictive accuracy (percentage
of samples well classified) as a function of the number of times a sample was classified as of poor prognosis by the gene signatures.
Page 9
Breast Cancer Research Vol 10 No 6 Reyal et al.
Page 10 of 15
(page number not for citation purposes)
group versus the low-risk group is 6.1 (3.3 to 11.4). The
known clinical predictors Elston Ellis grading, age at diagno-
sis, size (mm) and ER status, as well as the Immune and RNA
splicing classifier, were then included in a multivariate analysis.
For both endpoints, the combined classifier was the most sig-
nificant variable in the model (see Additional data file 1).
The gain in predictive accuracy from adding the RNA splicing/
Immune signature to each of the three common clinical staging
systems is presented in Table 2. The largest decrease in pre-
dictive inaccuracy is seen when the RNA splicing/Immune sig-
nature is added to the St Gallen index, as this has the worst
individual performance for both DMFS and BCSS (0.31 and
0.287, respectively). In this case, the predictive accuracy is
reduced from 0.287 to 0.254 for BCSS (DMFS, from 0.310 to
0.282). In terms of explained variation, the St Gallen index is
again the worst, explaining only 1% to 2% of the variation in
this dataset. With 6% to 8% of the explained variation being
attributable to the Nottingham Prognostic Index dataset, it is
the best of the clinical staging systems. In contrast, the RNA
splicing/Immune signature explains a larger amount of the var-
iation (10% to 13%) on its own, and including it with one of
the three staging systems improves on this only marginally. For
this dataset, there is a gain in the predictive accuracy when
adding the RNA splicing/Immune classifier to existing predic-
tors. Additional data file 7 shows the prognostic capacity of
the signature conditional on the levels of the different clinical
staging systems.
Figure 5
Enrichment analysis for the seven enlarged signatures and Kaplan–Meier analysis of distant metastasis-free survivalEnrichment analysis for the seven enlarged signatures and Kaplan–Meier analysis of distant metastasis-free survival. (a) Enrichment analy-
sis for the seven enlarged signatures on a collection of 1,889 gene sets from the Reactome, Kyoto Encyclopedia of Genes and Genomes (KEGG),
Molecular Signatures Database, and Gene Ontology databases. Significance of the P values was computed with the hypergeometric test. All P val-
ues were adjusted for multiple testing using the Benjamini–Hochberg method. Each cell in the matrix represents the adjusted P value for a given
enlarged signature and a gene set. The ontology modules to which a particular gene set belongs are indicated at the top. See (b) for the link
between the color coding and the module identity. (b) Kaplan–Meier analysis of distant metastasis-free survival (DMFS) on the Chin–Loi training set
for each of the 11 ontology modules. The Kaplan–Meier analysis is based on the output of a nearest mean classifier trained on the genes in each of
the ontology modules. (c) Kaplan–Meier analysis of DMFS on van Vijver and colleagues' breast cancer series as stratified by the Immune and RNA
splicing module classifier, for both endpoints. Blue curve, low-risk group where both classifiers assign a patient to the good outcome group; grey
curve, intermediate-risk category where the classifiers are discordant; red curve, high-risk category where both classifiers assign a patient to the
poor outcome group. For comparison, the same dataset as stratified by the Netherlands Cancer Institute (NKI) 70-gene classifier for both endpoints
is also presented.
Page 10
Available online http://breast-cancer-research.com/content/10/6/R93
Page 11 of 15
(page number not for citation purposes)
Discussion
The analyses performed here demonstrate that each of the
nine gene expression signatures have similar classification
performance based on the sensitivity, specificity, negative pre-
dictive value, positive predictive value and predictive accuracy.
All gene expression signatures added independent informa-
tion to a multivariate model including standard pathological
and clinical criteria. Although the gene expression classifiers
were mostly defined to determine the risk of distant metastasis
events and not the risk of death from breast cancer, we found
similar results for each gene classifier when using either
DMFS or BCCS as the endpoint.
The Gene expression Grade Index [25] and the molecular
prognostic index signatures (T17 and T52) [20] were devel-
oped for ER-positive breast cancer, while the 76-gene expres-
sion classifier signature [8] was defined for LN-negative
tumors. Despite these prerequisites, we applied the nine gene
expression classifiers to the same dataset without any consid-
eration for the heterogeneity of these samples as our first goal
was to compare these different classifiers when applied to the
same dataset. More importantly, the generalization of sub-
group-specific classifiers (for example, LN0 classifiers) across
a complete cohort of breast cancer samples (both LN0 and
LN1) hints at the existence of common biological processes
determining the outcome. We were interested in revealing
these processes. Furthermore, when evaluating the signatures
in specific subgroups of patients, we showed similar behavior
for each of these nine gene signatures. No signature showed
strong association with survival when applied to LN-positive,
ER-negative or high-grade subgroups. These results are
potentially explained by the fact that these factors identify a set
of intrinsically poor outcome cases; that is, they contain no
good outcome cases. This emphasizes the fact that gene
expression classifiers should, in our opinion, not be regarded
as a tool to replace standard pathological and clinical criteria,
but should instead be integrated with clinical parameters.
Gene expression classifiers can be employed to improve strat-
ification in subgroups of breast cancer patients with good
prognosis, where the groups are defined based on standard
pathological and clinical criteria.
Fan and colleagues [24] showed similar performance and sig-
nificant concordance between the 70-gene signature [6,7],
the Core Serum Response signature [14], the Genomic
Health signature [10] and the Intrinsic/UNC gene set [15]
when applied to the dataset of van de Vijver and colleagues
[7]. In contrast, we show here that agreement between gene
expression signatures is low, with >50% of the samples hav-
ing at least one discordant class assignment. We showed that
the predictive accuracy dramatically decreases with the
number of poor prognostic assignments a sample receives.
The best classification performance was obtained for the
group of patients with only good outcome assignments. These
results immediately reveal the dilemma faced by a patient diag-
nosed with breast cancer, and determines consultation of a
collection of signatures to predict disease outcome. The result
obtained is uncertain in almost 50% of the cases. As our
results were less optimistic than those of Fan and colleagues
[24], we repeated our analyses as described above but this
Table 2
Explained variation and predictive inaccuracy for distant metastasis-free survival and breast cancer-specific survival in the 295
Netherlands Cancer Institute dataset
Model Distant metastasis-free survival Breast cancer-specific survival
Predictive inaccuracy Explained variation (%) Predictive inaccuracy Explained variation (%)
No predictors 0.314 ± 0.02 0.292 ± 0.02
Adjuvant online 0.303 ± 0.02 3.5 ± 1.8 0.274 ± 0.02 6.1 ± 2.0
Nottingham Prognostic Index 0.294 ± 0.02 6.2 ± 2.4 0.270 ± 0.02 7.8 ± 2.6
St Gallen 0.310 ± 0.02 1.2 ± 0.9 0.287 ± 0.02 1.9 ± 1.0
RNA splicing/Immune module classifier 0.282 ± 0.02 9.9 ± 3.0 0.254 ± 0.02 13.0 ± 3.8
Adjuvant online and module classifier 0.281 ± 0.02 10.3 ± 2.9 0.250 ± 0.02 14.5 ± 3.8
Nottingham Prognostic Index and module
classifier
0.277 ± 0.02 11.6 ± 3.2 0.248 ± 0.02 15.2 ± 3.8
St Gallen and module classifier 0.282 ± 0.02 9.9 ± 3.0 0.254 ± 0.02 13.1 ± 3.7
Gain by adding module classifier to Adjuvant
online
0.022 6.8 0.024 8.4
Gain by adding module classifier to
Nottingham Prognostic Index
0.017 5.4 0.022 7.4
Gain by adding module classifier to St Gallen 0.028 8.7 0.033 11.2
Data presented as the mean ± standard error.
Page 11
Breast Cancer Research Vol 10 No 6 Reyal et al.
Page 12 of 15
(page number not for citation purposes)
time used the dataset of van de Vijver and colleagues [7] and
the following signatures: the 70-gene signature [6,7] (employ-
ing Fan and colleagues' labeling [24]), the Core Serum
Response signature [14], the Genomic Health signature [10],
the Intrinsic/UNC gene set [15], and the Gene expression
Grade Index [25]. The results recapitulated our earlier results.
In particular, only 42% of the samples received a concordant
class assignment, while the ER status, HER2 status, patholog-
ical grading and tumor size were all correlated with the number
of times a sample was classified as poor outcome by the sig-
natures. As was demonstrated earlier, the predictive accuracy
decreased with the number of poor outcome assignments.
Larger datasets (such as those acquired in the TAILORx and
MINDACT trials [11,12] are required to shed more light on the
cases where the signatures give discordant class
assignments.
To gain more insight into the small degree of overlap between
the genes comprising the different classifiers, we generated
an enlarged signature for each signature. The intersection of
the enlarged signatures identified a core of 72 genes signifi-
cantly enriched in DNA replication, cell cycle and mitosis
ontology annotations, which we consider the common back-
ground of these gene signatures. Proliferation genes are a
major component of many prognostic signatures in breast car-
cinoma and other tumor types [44]. Among the 72 genes we
found AURKA, BIRC5, CCNB1, MKI67 and MYBL2, which
define the complete set of proliferation genes from the
Genomic Health signature [10]. The proliferation modules also
contain genes frequently described as markers of proliferation
in different types of cancer [45]: PLK1, BUB1, CCNA2,
CCNB1, CCNB2, CCNE2, FOXM1, and TOP2A. These
genes are derived from the functional intersection of the
enlarged gene expression signatures, indicating that prolifera-
tion is a major driver of the prognosis gene signatures.
The enrichment analysis of the enlarged signatures revealed
11 gene ontology modules. Identification of distinct biological
processes correlated with survival or other clinicopathological
features is a major step towards improving our understanding
of tumor development and to providing accurate information to
develop new targeted therapies. Yu and colleagues generated
500 gene signatures of ER-positive and ER-negative tumors
[46], and found the following pathways to be overrepresented
in the signatures: apoptosis, proliferation, focal adhesion, RNA
splicing and immunity. They emphasized that similar pathways
are common to different gene signatures, whereas the individ-
ual genes defining these pathways can still have varying
degrees of association with outcome.
We showed that the combination of the Immune and RNA
splicing modules define a classifier that is highly accurate in
predicting both DMFS and BCSS. In addition, the classifier
showed an improvement in predictive accuracy when com-
bined with commonly used clinical staging systems. This indi-
cates that not only proliferation but also other functional
processes have prognostic power. Teschendorff and col-
leagues recently showed that the overexpression of a seven-
gene immunity module is associated with good outcome in
186 ER-negative breast cancers [21]. No significant correla-
tion between lymphocyte infiltration and this immunity module
was found. Two of these seven genes (XCL2 and HLA-F) are
also in our classifier. Recent clinical and experimental studies
have revealed that not only cancer cell intrinsic processes, but
also cancer cell extrinsic processes – including angiogenesis,
remodeling of the extracellular matrix, and inflammation – are
critical in determining malignant outcome. The role of the
immune system during cancer progression has recently
gained much attention [47]. The reciprocal interaction
between the immune system and cancer can be regarded as
a double-edged sword: whereas certain interactions inhibit or
prevent cancer growth, other interactions actually contribute
to tumor progression. For example, in situ analysis of tumor-
infiltrating lymphocytes in human colorectal cancer samples
revealed that the influx of T lymphocytes is associated with
improved survival, and the immunological data were found to
be a better predictor of patient survival than the histopatholog-
ical methods currently used to stage colorectal cancer [48].
The RNA splicing process is a key molecular event in the gen-
eration of protein biodiversity. Alteration of the normal process
results in the production of altered mRNA or in an off-balance
production of tissue-specific mRNA isoforms [49-51]. The
main consequences of this abnormal RNA splicing process
are a reduction of the normal protein level or the production of
abnormal proteins. Aberrant mRNA splicing variants are found
in many cancers and can interfere with major biological events
such as apoptosis, cell-cycle control, adhesion, differentiation
or angiogenesis. Mutations in splicing cis-acting sequences
have been associated with the BRCA1 gene in breast cancer
[52] and the KIT oncogene in gastrointestinal stromal tumor
[53]. The RNA splicing module we identified contains several
genes that are individually strongly associated with survival.
More specifically, SFRS10 is significantly overexpressed in
breast cancer and might be responsible for splicing of CD44
isoforms associated with tumor progression and metastasis
[54]. SRPK1 is upregulated in breast cancer and its expres-
sion level is proportional to the tumor grade. Inhibition of
SRPK1 results in reduced phosphorylation of MAPK3,
MAPK1 and AKT [55]. Targeted SRPK1 treatment seems to
be a promising way to increase apoptosis, to decrease prolif-
eration and to enhance the sensitivity to chemotherapeutics
drugs [56]. LSM1 is located at 8p11-12 loci and is amplified
in almost 20% of breast cancer cases [57]. Streicher and col-
leagues showed that overexpression of hLsm1 transforms
mammary epithelial cells, and inhibition of its expression in
breast cancer cells reduces anchorage-independent prolifera-
tion [58]. Yang and colleagues similarly showed the same abil-
ity of LSM1 to transform human mammary epithelial cells in
vitro [57].
Page 12
Available online http://breast-cancer-research.com/content/10/6/R93
Page 13 of 15
(page number not for citation purposes)
Conclusion
In the present article we set out to address three questions
regarding the signatures considered in the study. First, we can
conclude that the nine gene expression signatures had similar
performance. This was observed for both the accuracy with
which samples were assigned to the dichotomous poor/good
outcome groups as well as the level of association with sur-
vival found. Nevertheless, the concordance of outcome
assignment between gene expression signatures is low, with
50% of the samples receiving at least one outcome
assignment that is discordant with the assignments of the
other signatures. This relatively high level of gene expression
classification instability is associated with a dramatic decrease
in predictive accuracy with an increase in the number of poor
outcome assignments. The heterogeneity in the outcome
assignments of the different classifiers can most probably be
attributed to the different approaches that were followed to
construct the classifiers, the heterogeneity in the sample pop-
ulations employed to construct the classifiers, and sample size
issues [59].
In the present study we showed that the common background
of the nine gene signatures investigated is a 72-gene cluster.
Eleven gene ontology modules were overrepresented in the
enlarged signatures. These modules revealed a wide array of
functional groups that were overrepresented in gene sets
highly correlated with the probes contained in the original sig-
natures. Finally, we demonstrated that the combination of the
Immune and RNA splicing modules defined an efficient classi-
fier for breast cancer. This result shows that pathway-level
analysis of microarrays is able to provide a functionally coher-
ent and highly efficient prognosis classifier, which will most
probably be more stable than the classifiers from which it
originates.
Competing interests
LvV is employed by Agendia BV, has an ownership interest in
MammaPrint, and is the named inventor on a patent to use
microarray technology to ascertain breast cancer prognosis
and holds equity interests in Agendia BV. The other authors
declare that they have no competing interests.
Authors' contributions
FR and MHvV contributed equally to this work, carrying out the
study design, data mining, data analysis and manuscript writ-
ing. NJA carried out the study design, data mining, data analy-
sis and manuscript writing. HMH carried out the study design
and data mining. KEdV carried out the data analysis and man-
uscript writing. MK carried out the study design and manu-
script writing. AET carried out the Molecular Prognosis Index
calculation. SM carried out the study design and manuscript
advising. LvV carried out the manuscript advising. CC carried
out the Molecular Prognosis Index calculation. RJS carried out
the fund raising and manuscript advising. MJvdV carried out
the study design, data analysis and manuscript writing. LFAW
carried out the study design, data mining, data analysis and
manuscript writing.
Additional files
The following Additional files are available online:
Additional file 1
A Word file containing the supplementary Materials and
methods section and four supplementary tables. Table
S1 lists data for multivariate and univariate Cox
regression analyses with DMFS and BCSS as the
endpoints. Table S2 lists data for multivariate Cox
regression analysis with selected clinical parameters –
ER status based on immunohistochemistry (DMFS only),
LN status (positive versus negative), histological grading
(Elston Ellis I, II and III) – and the output of the nine gene
expression classifiers as input and either DMFS and
BCSS as clinical endpoints. Table S3 lists a
performance analysis of the signatures on the complete
set of 1,127 patients with dichotomous outcome labels
of poor outcome and good outcome derived from DMFS
and BCSS. Table S4 lists data for multivariate Cox
regression analysis with selected clinical parameters –
ER status based on immunohistochemistry, LN status
(positive versus negative), histological grading (Elston
Ellis I, II and III) – tumor size and the output of the Immune
and RNA splicing modules gene signature (IR) or the 70-
gene signature (NKI) as the input and DMFS and BCSS
analysis as the clinical endpoint.
See http://www.biomedcentral.com/content/
supplementary/bcr2192-S1.doc
Additional file 2
An Adobe file containing a figure of the RNA splicing,
Immune and 72 Proliferation gene annotations
[Probe_ID, EntrezID, OMIM, Ensembl, UnigeneID,
Representative Public ID, RefSeq Transcript ID, Gene
Symbol, k-means metastasis, k-means no metastasis].
See http://www.biomedcentral.com/content/
supplementary/bcr2192-S2.png
Additional file 3
An Adobe file containing a figure showing the nine
signature Kaplan–Meier curves with BCSS as endpoints
(S1a), showing the performance of the signatures on
subgroups of the patient population (S1b (top panel)),
and showing) the time-censoring performance analysis
of the signatures (1b (bottom panel).
See http://www.biomedcentral.com/content/
supplementary/bcr2192-S3.pdf
Page 13
Breast Cancer Research Vol 10 No 6 Reyal et al.
Page 14 of 15
(page number not for citation purposes)
Acknowledgements
FR was supported by grants from the French 'Association pour la
Recherche contre le Cancer' and the French Academy of Medecine. The
present article is dedicated to JH Petersen.
References
1. Goldhirsch A, Wood WC, Gelber RD, Coates AS, Thürlimann B,
Senn H-J: Progress and promise: highlights of the international
expert consensus on the primary therapy of early breast can-
cer 2007. Ann Oncol 2007, 18:1133-1144.
2. Eifel P, Axelson JA, Costa J, Crowley J, Curran WJ, Deshler A, Ful-
ton S, Hendricks CB, Kemeny M, Kornblith AB, Louis TA, Markman
M, Mayer R, Roter D: National Institutes of Health Consensus
Development Conference Statement: adjuvant therapy for
breast cancer, November 1–3, 2000. J Natl Cancer Inst 2001,
93:979-989.
3. Blamey RW, Ellis IO, Pinder SE, Lee AHS, Macmillan RD, Morgan
DAL, Robertson JFR, Mitchell MJ, Ball GR, Haybittle JL, Elston
CW: Survival of invasive breast cancer according to the Not-
tingham Prognostic Index in cases diagnosed in 1990–1999.
Eur J Cancer 2007, 43:1548-1555.
4. Ravdin PM, Siminoff LA, Davis GJ, Mercer MB, Hewlett J, Gerson
N, Parker HL: Computer program to assist in making decisions
about adjuvant therapy for women with early breast cancer. J
Clin Oncol 2001, 19:980-991.
5. Early Breast Cancer Trialists' Collaborative Group: Effects of
chemotherapy and hormonal therapy for early breast cancer
on recurrence and 15-year survival: an overview of the ran-
domised trials. Lancet 2005, 365:1687-1717.
6. van 't Veer LJ, Dai H, Vijver MJ van de, He YD, Hart AAM, Mao M,
Peterse HL, Kooy K van der, Marton MJ, Witteveen AT, Schreiber
GJ, Kerhoven RM, Roberts C, Linsley PS, Bernards R, Friend SH:
Gene expression profiling predicts clinical outcome of breast
cancer. Nature 2002, 415:530-536.
7. Vijver MJ van de, He YD, van 't Veer LJ, Dai H, Hart AAM, Voskuil
DW, Schreiber GJ, Peterse JL, Roberts C, Marton MJ, Parrish M,
Atsma D, Witteveen A, Glas A, Delahaye L, Velde T van der, Bar-
telink H, Rodenhuis S, Rutgers ET, Friend SH, Bernards R: A
gene-expression signature as a predictor of survival in breast
cancer. N Engl J Med 2002, 347:1999-2009.
8. Wang Y, Klijn JG, Zhang Y, Sieuwerts AM, Look MP, Yang F,
Talantov D, Timmermans M, Meijer-van Gelder ME, Yu J, Jatkoe T,
Berns EM, Atkins D, Foekens JA: Gene-expression profiles to
predict distant metastasis of lymph-node-negative primary
breast cancer. The Lancet 2005, 365:671-679.
9. Desmedt C, Piette F, Loi S, Wang Y, Lallemand F, Haibe-Kains B,
Viale G, Delorenzi M, Zhang Y, d'Assignies MS, Bergh J, Lidereau
R, Ellis P, Harris AL, Klijn JGM, Foekens JA, Cardoso F, Piccart MJ,
Buyse M, Sotiriou C: Strong time dependence of the 76-gene
prognostic signature for node-negative breast cancer patients
in the TRANSBIG multicenter independent validation series.
Clin Cancer Res 2007, 13:3207-3214.
10. Paik S, Shak S, Tang G, Kim C, Baker J, Cronin M, Baehner FL,
Walker MG, Watson D, Park T, Hiller W, Fisher ER, Wickerham
DL, Bryant J, Wolmark N: A multigene assay to predict recur-
rence of tamoxifen-treated, node-negative breast cancer. N
Engl J Med 2004, 351:2817-2826.
11. Bogaerts J, Cardoso F, Buyse M, Braga S, Loi S, Harrison JA,
Bines J, Mook S, Decker N, Ravdin P, Therasse P, Rutgers E, van
't Veer LJ, Piccart M, the TRANSBIG consortium: Gene signature
evaluation as a prognostic tool: challenges in the design of the
MINDACT trial. Nat Clin Pract Oncol 2006, 3:540-551.
12. Paik S: Development and clinical utility of a 21-gene recur-
rence score prognostic assay in patients with early breast can-
cer treated with tamoxifen. Oncologist 2007, 12:631-635.
13. Carter SL, Eklund AC, Kohane IS, Harris LN, Szallasi Z: A signa-
ture of chromosomal instability inferred from gene expression
profiles predicts clinical outcome in multiple human cancers.
Nat Genet 2006, 38:1043-1048.
14. Chang HY, Nuyten DS, Sneddon JB, Hastie T, Tibshirani R, Sorlie
T, Dai H, He YD, van 't Veer LJ, Bartelink H, van Rijn M, Brown PO,
Vijver MJ van de: Robustness, scalability, and integration of a
wound-response gene expression signature in predicting
breast cancer survival. Proc Natl Acad Sci USA 2005,
102:3738-3743.
15. Hu Z, Cheng F, Oh DS, Marron JS, He X, Qaqish BF, Livasy C,
Carey LA, Reynolds E, Dressler L, Nobel A, Parker J, Ewend MG,
Sawyer LR, Wu J, Liu Y, Nanda R, Tretiakova M, Orrico AR, Dreher
D, Palazzo JP, Perreard L, Nelson E, Mone M, Hansen H, Mullins
M, Quackenbush JF, Ellis MJ, Olopade OI, Bernard PS, Perou CM:
The molecular portraits of breast tumors are conserved
across microarray platforms. BMC Genomics 2006, 7:96.
16. Liu R, Wang X, Grace YC, Dalerba P, Gurney A, Hoey T, Sherlock
G, Lewicki J, Shedden K, Clarke MF: The prognostic role of a
gene signature from tumorigenic breast-cancer cells. N Engl
J Med 2007, 356:217-226.
17. Loi S, Haibe-Kains B, Desmedt C, Lallemand F, Tutt AM, Gillet C,
Ellis P, Harris A, Bergh J, Foekens JA, Klijn JGM, Larsimont D,
Buyse M, Bontempi G, Delorenzi M, Piccart MJ, Sotiriou C: Defi-
nition of clinically distinct molecular subtypes in estrogen
receptor-positive breast carcinomas through genomic grade.
J Clin Oncol 2007, 25:1239-1246.
18. Miller LD, Smeds J, George J, Vega VB, Vergara L, Ploner A, Paw-
itan Y, Hall P, Klaar S, Liu ET, Bergh J: An expression signature
for p53 status in human breast cancer predicts mutation sta-
tus, transcriptional effects, and patient survival. Proc Natl Acad
Sci USA 2005, 102:13550-13555.
19. Minn AJ, Gupta GP, Siegel PM, Bos PD, Shu WP, Giri DD, Viale
A, Olshen AB, Gerald WL, Massague J: Genes that mediate
breast cancer metastasis to lung. Nature 2005, 436:518-524.
20. Teschendorff AE, Naderi A, Barbosa-Morais NL, Pinder SE, Ellis
IO, Aparicio S, Brenton JD, Caldas C: A consensus prognostic
Additional file 4
An image file containing a figure showing heatmaps of
the concordance of the nine classifiers across clinical
subgroups among the 1,127 human breast tumor
samples.
See http://www.biomedcentral.com/content/
supplementary/bcr2192-S4.png
Additional file 5
An image file containing a figure showing the overlap and
performance analysis of 403 samples with BCSS as the
endpoint.
See http://www.biomedcentral.com/content/
supplementary/bcr2192-S5.png
Additional file 6
An Excel file containing a figure showing the log-rank test
P value on the Chin–Loi training set for each pairwise
combination of the module classifiers.
See http://www.biomedcentral.com/content/
supplementary/bcr2192-S6.xls
Additional file 7
An Adobe file containing a figure showing the Kaplan–
Meier plots on the dataset of van de Vijver and
colleagues for the subgroups defined by the Nottingham
Prognostic Index, St Gallen, and AdjuvantOnline! clinical
staging systems – plots for the Immune/RNA splicing
module classifier within each of the clinical subgroups for
each staging system.
See http://www.biomedcentral.com/content/
supplementary/bcr2192-S7.pdf
Page 14
Available online http://breast-cancer-research.com/content/10/6/R93
Page 15 of 15
(page number not for citation purposes)
gene expression classifier for ER positive breast cancer.
Genome Biol 2006, 7:R101.
21. Teschendorff AE, Miremadi A, Pinder S, Ellis I, Caldas C: An
immune response gene expression module identifies a good
prognosis subtype in estrogen receptor negative breast
cancer. Genome Biol 2007, 8:R157.
22. Sotiriou C, Piccart MJ: Taking gene-expression profiling to the
clinic: when will molecular signatures become relevant to
patient care? Nat Rev Cancer 2007, 7:545-553.
23. Ein-Dor L, Zuk O, Domany E: Thousands of samples are needed
to generate a robust gene list for predicting outcome in
cancer. Proc Natl Acad Sci USA 2006, 103:5923-5928.
24. Fan C, Oh DS, Wessels L, Weigelt B, Nuyten DS, Nobel AB, van't
Veer LJ, Perou CM: Concordance among gene-expression-
based predictors for breast cancer. N Engl J Med 2006,
355:560-569.
25. Sotiriou C, Wirapati P, Loi S, Harris A, Fox S, Smeds J, Nordgren
H, Farmer P, Praz V, Haibe-Kains B, Desmedt C, Larsimont D, Car-
doso F, Peterse H, Nuyten D, Buyse M, Vijver MJ van de, Bergh J,
Piccart M, Delorenzi M: Gene expression profiling in breast can-
cer: understanding the molecular basis of histologic grade to
improve prognosis. J Natl Cancer Inst 2006, 98:262-272.
26. Mook S, Schmidt MK, Viale G, Pruneri G, Eekhout I, Floore A, Glas
AM, Bogaerts J, Cardoso F, Piccart-Gebhart MJ, Rutgers ET, Van't
Veer LJ, On behalf of the TRANSBIG consortium: The 70-gene
prognosis-signature predicts disease outcome in breast can-
cer patients with 1–3 positive lymph nodes in an independent
validation study. Breast Cancer Res Treat 2008 in press.
27. Goldstein LJ, Gray R, Badve S, Childs BH, Yoshizawa C, Rowley
S, Shak S, Baehner FL, Ravdin PM, Davidson NE, Sledge GW Jr,
Perez EA, Shulman LN, Martino S, Sparano JA: Prognostic utility
of the 21-gene assay in hormone receptor-positive operable
breast cancer compared with classical clinicopathologic
features. J Clin Oncol 2008, 26:4063-4071.
28. Pawitan Y, Bjöhle J, Amler L, Borg A-L, Egyhazi S, Hall P, Han X,
Holmberg L, Huang F, Klaar S, Liu ET, Miller L, Nordgren H, Ploner
A, Sandelin K, Shaw PM, Smeds J, Skoog L, Wedrén S, Bergh J:
Gene expression profiling spares early breast cancer patients
from adjuvant therapy: derived and validated in two popula-
tion-based cohorts. Breast Cancer Res 2005, 7:R953-R964.
29. Chin K, DeVries S, Fridlyand J, Spellman PT, Roydasgupta R, Kuo
W-L, Lapuk A, Neve RM, Qian Z, Ryder T: Genomic and tran-
scriptional aberrations linked to breast cancer
pathophysiologies.
Cancer Cell 2006, 10:529-541.
30. Gene Expression Omnibus [http://www.ncbi.nlm.nih.gov/geo/
]
31. ArrayExpress repository [http://www.ebi.ac.uk/microarray-as/
ae/]
32. R [http://cran.r-project.org/
]
33. Bioconductor [http://www.bioconductor.org
]
34. Matlab [http://www.mathworks.com/
]
35. Gene Ontology [http://www.ncbi.nlm.nih.gov/
]
36. Kyoto Encyclopedia of Genes and Genomes [http://
www.genome.ad.jp]
37. Reactome [http://www.reactome.org
]
38. Molecular Signatures Database [http://www.broad.mit.edu/
gsea/]
39. Benjamini Y, Hochberg Y: Controlling the false discovery rate: a
practical and powerful approach to multiple testing. J R Stat
Soc Ser B 1995, 57:289-300.
40. Schemper M, Henderson R: Predictive accuracy and explained
variation in Cox regression. Biometrics 2000, 56:249-255.
41. Lusa L, Miceli R, Mariani L: Estimation of predictive accuracy in
survival analysis using R and S-PLUS. Comput Methods Pro-
grams Biomed 2007, 87:132-137.
42. Dunkler D, Michiels S, Schemper M: Gene expression profiling:
does it add predictive accuracy to clinical characteristics in
cancer prognosis? Eur J Cancer 2007, 43:745-751.
43. Buyse M, Loi S, van't Veer L, Viale G, Delorenzi M, Glas AM,
d'Assignies MS, Bergh J, Lidereau R, Ellis P, Harris A, Bogaerts J,
Therasse P, Floore A, Amakrane M, Piette F, Rutgers E, Sotiriou C,
Cardoso F, Piccart MJ, the TRANSBIG Consortium: Validation
and clinical utility of a 70-gene prognostic signature for
women with node-negative breast cancer. J Natl Cancer Inst
2006, 98:1183-1192.
44. Whitfield ML, George LK, Grant GD, Perou CM: Common mark-
ers of proliferation. Nat Rev Cancer 2006, 6:99-106.
45. Chung CH, Bernard PS, Perou CM: Molecular portraits and the
family tree of cancer. Nat Genet 2002, 32(Suppl):533-540.
46. Yu J, Sieuwerts A, Zhang Y, Martens J, Smid M, Klijn J, Wang Y,
Foekens J: Pathway analysis of gene signatures predicting
metastasis of node-negative primary breast cancer. BMC
Cancer 2007, 7:182.
47. de Visser KE, Coussens LM: The inflammatory tumor microen-
vironment and its impact on cancer development. Contrib
Microbiol 2006, 13:118-137.
48. Galon J, Costes A, Sanchez-Cabo F, Kirilovsky A, Mlecnik B,
Lagorce-Pagès C, Tosolini M, Camus M, Berger A, Wind P, Zinzin-
dohoué F, Bruneval P, Cugnenc P-H, Trajanoski Z, Fridman W-H,
Pagès F: Type, density, and location of immune cells within
human colorectal tumors predict clinical outcome. Science
2006, 313:1960-1964.
49. Venables JP:
Aberrant and alternative splicing in cancer. Can-
cer Res 2004, 64:7647-7654.
50. Venables JP: Unbalanced alternative splicing and its signifi-
cance in cancer. Bioessays 2006, 28:378-386.
51. Pajares MJ, Ezponda T, Catena R, Calvo A, Pio R, Montuenga LM:
Alternative splicing: an emerging topic in molecular and clini-
cal oncology. Lancet Oncol 2007, 8:349-357.
52. Mazoyer S, Puget N, Perrin-Vidoz L, Lynch HT, Serova-Sinilnikova
OM, Lenoir GM: A BRCA1 nonsense mutation causes exon
skipping. Am J Hum Genet 1998, 62:713-715.
53. Chen LL, Sabripour M, Wu EF, Prieto VG, Fuller GN, Frazier ML: A
mutation-created novel intra-exonic pre-mRNA splice site
causes constitutive activation of KIT in human gastrointestinal
stromal tumors. Oncogene 2005, 24:4271-4280.
54. Watermann DO, Tang Y, Hausen AZ, Jäger M, Stamm S, Stickeler
E: Splicing factor Tra2-beta1 is specifically induced in breast
cancer and regulates alternative splicing of the CD44 gene.
Cancer Res 2006, 66:4774-4780.
55. Hayes GM, Carrigan PE, Miller LJ: Serine-arginine protein
kinase 1 overexpression is associated with tumorigenic imbal-
ance in mitogen-activated protein kinase pathways in breast,
colonic, and pancreatic carcinomas. Cancer Res 2007,
67:2072-2080.
56. Hayes GM, Carrigan PE, Beck AM, Miller LJ: Targeting the RNA
splicing machinery as a novel treatment strategy for pancre-
atic carcinoma. Cancer Res 2006, 66:3819-3827.
57. Yang ZQS, Katie L, Ray ME, Abrams J, Ethier SP: Multiple inter-
acting oncogenes on the 8p11-p12 amplicon in human breast
cancer. Cancer Res 2006, 66:11632-11643.
58. Streicher KL, Yang ZQ, Draghici S, Ethier SP: Transforming func-
tion of the LSM1 oncogene in human breast cancers with the
8p11-12 amplicon. Oncogene 2007, 26:2104-2114.
59. van Vliet MH, Reyal F, Horlings HM, Vijver MJ van de, Reinders MJ,
Wessels LF: Pooling breast cancer datasets has a synergetic
effect on classification performance and improves signature
stability. BMC Genomics 2008, 9:375.
Page 15
  • Source
    • "Another way of improving the stability of a gene signature suggested by Zhao et al was to combine the information obtained from published gene-sets prognosis signatures to build a higher prognostic performance signature using an independent gene expression dataset [41]. In a previous study, Reyal et al showed that the combination of the signatures could define an efficient classifier for breast cancer, which will most probably be more stable than the classifiers from which it originates [20]. It is therefore not surprising that the risks predicted by these models are in the end little used in clinical setting for adjuvant therapies prescription. "
    [Show abstract] [Hide abstract] ABSTRACT: Background: Numerous prognostic gene expression signatures have been recently described. Among the signatures there is variation in the constituent genes that are utilized. We aim to evaluate prognostic concordance among eight gene expression signatures, on a large dataset of ER positive HER2 negative breast cancers. Methods: We analysed the performance of eight gene expression signatures on six different datasets of ER+ HER2- breast cancers. Survival analyses were performed using the Kaplan-Meier estimate of survival function. We assessed discrimination and concordance between the 8 signatures on survival and recurrence rates The Nottingham Prognostic Index (NPI) was used to to stratify the risk of recurrence/death. Results: The discrimination ability of the whole signatures, showed fair discrimination performances, with AUC ranging from 0.64 (95%CI 0.55-0.73 for the 76-genes signatures, to 0.72 (95%CI 0.64-0.8) for the Molecular Prognosis Index T17. Low concordance was found in predicting events in the intermediate and high-risk group, as defined by the NPI. Low risk group was the only subgroup with a good signatures concordance. Conclusion: Genomic signatures may be a good option to predict prognosis as most of them perform well at the population level. They exhibit, however, a high degree of discordance in the intermediate and high-risk groups. The major benefit that we could expect from gene expression signatures is the standardization of proliferation assessment.
    Full-text · Article · Feb 2016 · PLoS ONE
  • Source
    • "Most studies evaluating various signatures [14-18] have been carried out on relatively small scales. Compatibility between the signatures and the targeted cohorts with respect to biological and pathological characteristics (Additional file 1: Table S1) is often ignored [16]. Use of validation sets not completely independent of the original training sets may have influenced the results leading to biased interpretation [14]. "
    [Show abstract] [Hide abstract] ABSTRACT: The aim was to assess and compare prognostic power of nine breast cancer gene signatures (Intrinsic, PAM50, 70-gene, 76-gene, Genomic-Grade-Index, 21-gene-Recurrence-Score, EndoPredict, Wound-Response and Hypoxia) in relation to ER status and follow-up time. A gene expression dataset from 947 breast tumors was used to evaluate the signatures for prediction of Distant Metastasis Free Survival (DMFS). A total of 912 patients had available DMFS status. The recently published METABRIC cohort was used as an additional validation set. Survival predictions were fairly concordant across most signatures. Prognostic power declined with follow-up time. During the first 5 years of followup, all signatures except for Hypoxia were predictive for DMFS in ER-positive disease, and 76-gene, Hypoxia and Wound-Response were prognostic in ER-negative disease. After 5 years, the signatures had little prognostic power. Gene signatures provide significant prognostic information beyond tumor size, node status and histological grade. Generally, these signatures performed better for ER-positive disease, indicating that risk within each ER stratum is driven by distinct underlying biology. Most of the signatures were strong risk predictors for DMFS during the first 5 years of follow-up. Combining gene signatures with histological grade or tumor size, could improve the prognostic power, perhaps also of long-term survival.
    Full-text · Article · Mar 2014 · BMC Cancer
  • Source
    • "Analyses of the prognostic information that lies in the 70-gene signature and other multigene signatures have shown that a large portion of the prognostic information lies in proliferation-related genes [42]. In fact, reanalyses of these signatures showed that the signature with proliferation-related genes had greater prognostic value than the original signature [43]–[45], and that the proliferation signature was correlated with the MAI (correlation coefficient, 0.968) [44]. One study showed that the non-proliferative genes had no prognostic power [43]. "
    [Show abstract] [Hide abstract] ABSTRACT: The overall survival rate is good for lymph-node-negative breast cancer patients, but they still suffer from serious over- and some undertreatments. Prognostic and predictive gene signatures for node-negative breast cancer have a high number of genes related to proliferation. The prognostic value of gene sets from commercial gene-expression assays were compared with proliferation markers. Illumina WG6 mRNA microarray analysis was used to examine 94 fresh-frozen tumour samples from node-negative breast cancer patients. The patients were divided into low- and high-risk groups for distant metastasis based on the MammaPrint-related genes, and into low-, intermediate- and high-risk groups based on the recurrence score algorithm with genes included in Oncotype DX. These data were then compared to proliferation status, as measured by the mitotic activity index, the expressions of phosphohistone H3 (PPH3), and Ki67. Kaplan-Meier survival analysis for distant-metastasis-free survival revealed that patients with weak and strong PPH3 expressions had 14-year survival rates of 87% (n = 45), and 65% (n = 49, p = 0.014), respectively. Analysis of the MammaPrint classification resulted in 14-year survival rates of 80% (n = 45) and 71% (n = 49, p = 0.287) for patients with low and high risks of recurrence, respectively. The Oncotype DX categorization yielded 14-year survival rates of 83% (n = 18), 79% (n = 42) and 68% (n = 34) for those in the low-, intermediate- and high-risk groups, respectively (p = 0.52). Supervised hierarchical cluster analysis for distant-metastasis-free survival in the subgroup of patients with strong PPH3 expression revealed that the genes involved in Notch signalling and cell adhesion were expressed at higher levels in those patients with distant metastasis. This pilot study indicates that proliferation has greater prognostic value than the expressions of either MammaPrint- or Oncotype-DX-related genes. Furthermore, in the subgroup of patients with high proliferation, Notch signalling pathway genes appear to be expressed at higher levels in patients who develop distant metastasis.
    Full-text · Article · Mar 2014 · PLoS ONE
Show more