The invention provides for the identification and use of gene expression
profiles, or patterns, with clinical relevance to breast cancer.
In particular, the invention provides the identities of genes that
are correlated with patient survival and breast cancer recurrence.
The gene expression profiles may be embodied in nucleic acid expression,
protein expression, or other expression formats and used to predict
the survival of subjects afflicted with breast cancer and to predict
breast cancer recurrence and. The profiles may also be used in the
study and/or diagnosis of breast cancer cells and tissue, including
the grading of invasive breast cancer, as well as for the study
and/or determination of prognosis of a patient. When used for diagnosis
or prognosis, the profiles may be used to determine the treatment
of breast cancer based upon the likelihood of life expectancy and
5. A method to determine the survival outcome of a breast cancer
afflicted subject or determine prognosis of a subject having breast
cancer, said method comprising assaying a sample of breast cancer
cells of said subject for the expression level(s) of one or more
genes listed in Table 2, 3, 4, and/or 6.
7. A method of determining the prognosis of a subject having breast
cancer correlated with the over or under expression of one or more
genes in Table 2, 3, 4, and/or 6 said method comprising assaying
for the expression level(s) of said one or more genes in a breast
cancer cell from said subject.
8. The method of claim 5 wherein said assaying comprises preparing
RNA from said sample.
9. The method of claim 8 wherein said RNA is used for quantitative
10. The method of claim 5 wherein said assaying comprises using
11. The method of claim 5 wherein said sample is a ductal lavage
or fine needle aspiration sample.
12. The method of claim 11 wherein said sample is microdissected
to isolate one or more cells suspected of being breast cancer cells.
13. The method of claim 5 wherein said assaying comprises preparing
RNA from said sample and optionally using said RNA for quantitative
14. The method of claim 8 wherein said assaying comprises using
15. The method of claim 5 wherein said sample is a ductal lavage
or fine needle aspiration sample, which sample is optionally microdissected
to isolate one or more cells suspected of being breast cancer cells.
16. The method of claim 7 wherein said assaying comprises preparing
RNA from said cell and optionally using said RNA for quantitative
17. The method of claim 7 wherein said assaying comprises using
18. The method of claim 7 wherein said cell is present in a ductal
lavage or fine needle aspiration sample, which sample is optionally
microdissected to isolate one or more cells suspected of being breast
19. A method to determine the grade of breast cancer in a subject
comprising assaying a sample of breast cancer cells of said subject
for the expression level(s) of one or more genes listed in Table
20. A method to determine therapeutic treatment for a breast cancer
patient based upon said patient's expected survival, said method
comprising determining a survival outcome for said patient by assaying
a sample of breast cancer cells from said patient for the expression
level(s) of one or more one genes listed in Table 2, 3, 4, and/or
6; and selecting the appropriate treatment for a patient with such
a survival outcome.
 This application claims benefit of priority from U.S. Provisional
Patent Application 60/479,963, filed Jun. 18, 2003, and 60/545,810,
filed Feb. 18, 2004, both of which are incorporated by reference
as if fully set forth.
FIELD OF THE INVENTION
 The invention relates to the identification and use of gene
expression profiles, or patterns, with clinical relevance to breast
cancer. In particular, the invention provides the identities of
genes that are correlated with patient survival and breast cancer
recurrence. The gene expression profiles, whether embodied in nucleic
acid expression, protein expression, or other expression formats,
may be used to predict the survival of subjects afflicted with breast
cancer and to predict breast cancer recurrence and. The profiles
may also be used in the study and/or diagnosis of breast cancer
cells and tissue, including the grading of invasive breast cancer,
as well as for the study and/or determination of prognosis of a
patient. When used for diagnosis or prognosis, the profiles are
used to determine the treatment of breast cancer based upon the
likelihood of life expectancy and recurrence.
BACKGROUND OF THE INVENTION
 Breast cancer is by far the most common cancer among women.
Each year, more than 180,000 and 1 million women in the U.S. and
worldwide, respectively, are diagnosed with breast cancer. Breast
cancer is the leading cause of death for women between ages 50-55,
and is the most common non-preventable malignancy in women in the
Western Hemisphere. An estimated 2,167,000 women in the United States
are currently living with the disease (National Cancer Institute,
Surveillance Epidemiology and End Results (NCI SEER) program, Cancer
Statistics Review (CSR), www-seer.ims.nci.nih.gov/Publications/CSR1973
(1998)). Based on cancer rates from 1995 through 1997, a report
from the National Cancer Institute (NCI) estimates that about 1
in 8 women in the United States (approximately 12.8 percent) will
develop breast cancer during her lifetime (NCI's Surveillance, Epidemiology,
and End Results Program (SEER) publication SEER Cancer Statistics
Review 1973-1997). Breast cancer is the second most common form
of cancer, after skin cancer, among women in the United States.
An estimated 250,100 new cases of breast cancer are expected to
be diagnosed in the United States in 2001. Of these, 192,200 new
cases of more advanced (invasive) breast cancer are expected to
occur among women (an increase of 5% over last year), 46,400 new
cases of early stage (in situ) breast cancer are expected to occur
among women (up 9% from last year), and about 1,500 new cases of
breast cancer are expected to be diagnosed in men (Cancer Facts
& Figures 2001 American Cancer Society). An estimated 40,600
deaths (40,300 women, 400 men) from breast cancer are expected in
2001. Breast cancer ranks second only to lung cancer among causes
of cancer deaths in women. Nearly 86% of women who are diagnosed
with breast cancer are likely to still be alive five years later,
though 24% of them will die of breast cancer after 10 years, and
nearly half (47%) will die of breast cancer after 20 years.
 Every woman is at risk for breast cancer. Over 70 percent
of breast cancers occur in women who have no identifiable risk factors
other than age (U.S. General Accounting Office. Breast Cancer, 1971-1991:
Prevention, Treatment and Research. GAO/PEMD-92-12; 1991). Only
5 to 10% of breast cancers are linked to a family history of breast
cancer (Henderson I C, Breast Cancer. In: Murphy G P, Lawrence W
L, Lenhard R E (eds). Clinical Oncology. Atlanta, Ga.: American
Cancer Society; 1995:198-219).
 Each breast has 15 to 20 sections called lobes. Within each
lobe are many smaller lobules. Lobules end in dozens of tiny bulbs
that can produce milk. The lobes, lobules, and bulbs are all linked
by thin tubes called ducts. These ducts lead to the nipple in the
center of a dark area of skin called the areola. Fat surrounds the
lobules and ducts. There are no muscles in the breast, but muscles
lie under each breast and cover the ribs. Each breast also contains
blood vessels and lymph vessels. The lymph vessels carry colorless
fluid called lymph, and lead to the lymph nodes. Clusters of lymph
nodes are found near the breast in the axilla (under the arm), above
the collarbone, and in the chest.
 Breast tumors can be either benign or malignant. Benign
tumors are not cancerous, they do not spread to other parts of the
body, and are not a threat to life. They can usually be removed,
and in most cases, do not come back. Malignant tumors are cancerous,
and can invade and damage nearby tissues and organs. Malignant tumor
cells may metastasize, entering the bloodstream or lymphatic system.
When breast cancer cells metastasize outside the breast, they are
often found in the lymph nodes under the arm (axillary lymph nodes).
If the cancer has reached these nodes, it means that cancer cells
may have spread to other lymph nodes or other organs, such as bones,
liver, or lungs.
 Major and intensive research has been focused on early detection,
treatment and prevention. This has included an emphasis on determining
the presence of precancerous or cancerous ductal epithelial cells.
These cells are analyzed, for example, for cell morphology, for
protein markers, for nucleic acid markers, for chromosomal abnormalities,
for biochemical markers, and for other characteristic changes that
would signal the presence of cancerous or precancerous cells. This
has led to various molecular alterations that have been reported
in breast cancer, few of which have been well characterized in human
clinical breast specimens. Molecular alterations include presence/absence
of estrogen and progesterone steroid receptors, HER-2 expression/amplification
(Mark H F, et al. HER-2/neu gene amplification in stages I-IV breast
cancer detected by fluorescent in situ hybridization. Genet Med;
1(3):98-103 1999), Ki-67 (an antigen that is present in all stages
of the cell cycle except GO and used as a marker for tumor cell
proliferation, and prognostic markers (including oncogenes, tumor
suppressor genes, and angiogenesis markers) like p53, p27, Cathepsin
D, pS2, multi-drug resistance (MDR) gene, and CD31.
 van't Veer et al. (Nature 415:530-536, 2002) describe gene
expression profiling of clinical outcome in breast cancer. They
identified genes expressed in breast cancer tumors, the expression
levels of which correlated either with patients afflicted with distant
metastases within 5 years or with patients that remained metastasis-free
after at least 5 years.
 Ramaswamy et al. (Nature Genetics 33:49-54, 2003) describe
the identification of a molecular signature of metastasis in primary
solid tumors. The genes of the signature were identified based on
gene expression profiles of 12 metastatic adenocarcinoma nodules
of diverse origin (lung, breast, prostate, colorectal, uterus) compared
to expression profiles of 64 primary adenocarcinomas representing
the same spectrum of tumor types from different individuals. A 128
gene set was identified.
 Both of the above described approaches, however, utilize
heterogeneous populations of cells found in a tumor sample to obtain
information on gene expression patterns. The use of such populations
may result in the inclusion or exclusion of multiple genes that
are differentially expressed in cancer cells. The gene expression
patterns observed by the above described approaches may thus provide
little confidence that the differences in gene expression are meaningfully
associated with breast cancer recurrence or survival.
 Citation of documents herein is not intended as an admission
that any is pertinent prior art. All statements as to the date or
representation as to the contents of documents is based on the information
available to the applicant and does not constitute any admission
as to the correctness of the dates or contents of the documents.
BRIEF SUMMARY OF THE INVENTION
 The present invention relates to the identification and
use of gene expression patterns (or profiles or "signatures")
which are clinically relevant to breast cancer. In particular, the
identities of genes that are correlated with patient survival and
breast cancer recurrence are provided. The gene expression profiles,
whether embodied in nucleic acid expression, protein expression,
or other expression formats, may be used to predict survival of
subjects afflicted with breast cancer and the likelihood of breast
 The invention thus provides for the identification and use
of gene expression patterns (or profiles or "signatures")
which correlate with (and thus able to discriminate between) patients
with good or poor survival outcomes. In one embodiment, the invention
provides patterns that are able to distinguish patients with estrogen
receptor (ER) positive breast tumors into those with a survival
outcome poorer than that of patients with ER negative breast tumors
and those with a better survival outcome than that of patients with
ER positive breast tumors. These patterns are thus able to distinguish
patients with ER positive breast tumors into at least two subtypes.
 The invention also provides for the identification and use
of gene expression patterns which correlate with the recurrence
of breast cancer at the same location and/or in the form of metastases.
The pattern is able to distinguish patients with breast cancer into
at least those with good or poor survival outcomes.
 In another aspect of the invention, the ability to identify
the grade of invasive breast cancer by gene expression patterns
of the invention is provided. In particular, gene expression patterns
in a cell containing sample that distinguish "high-grade"
(or "grade 3") invasive breast tumors from "low-grade"
(or grades "1" and "2") invasive breast tumors
are provided. The invention thus permits the distinguishing (or
grading) of a subject's invasive tumors into two types which may
be differentially treated based on the expected outcome associated
with each type.
 The present invention provides a non-subjective means for
the identification of patients with breast cancer as likely to have
a good or poor survival outcome by assaying for the expression patterns
disclosed herein. Thus where subjective interpretation may have
been previously used to determine the prognosis and/or treatment
of breast cancer patients, the present invention provides objective
gene expression patterns, which may used alone or in combination
with subjective criteria to provide a more accurate assessment of
breast cancer patient outcomes, including survival and the recurrence
of cancer. The expression patterns of the invention thus provide
a means to determine breast cancer prognosis. Furthermore, the expression
patterns can also be used as a means to assay small, node negative
tumors that are not readily assayed by other means.
 The gene expression patterns comprise one or more than one
gene capable of discriminating between breast cancer outcomes with
significant accuracy. The gene(s) are identified as correlated with
various breast cancer outcomes such that the levels of their expression
are relevant to a determination of the preferred treatment protocols,
of a breast cancer patient. Thus in one aspect, the invention provides
a method to determine the outcome of a subject afflicted with, or
suspected of having, breast cancer by assaying a cell containing
sample from said subject for expression of one or more than one
gene disclosed herein as correlated with breast cancer outcomes.
 Gene expression patterns of the invention are identified
as described below. Generally, a large sampling of the gene expression
profile of a sample is obtained through quantifying the expression
levels of mRNA corresponding to many genes. This profile is then
analyzed to identify genes, the expression of which are positively,
or negatively, correlated, with a breast cancer outcome. An expression
profile of a subset of human genes may then be identified by the
methods of the present invention as correlated with a particular
breast cancer outcome. The use of multiple samples increases the
confidence which a gene may be believed to be correlated with a
particular survival outcome. Without sufficient confidence, it remains
unpredictable whether a particular gene is actually correlated with
a breast cancer outcome and also unpredictable whether a particular
gene may be successfully used to identify the outcome for a breast
 A profile of genes that are highly correlated with one outcome
relative to another may be used to assay an sample from a subject
afflicted with, or suspected of having, breast cancer to predict
the outcome of the subject from whom the sample was obtained. Such
an assay may be used as part of a method to determine the therapeutic
treatment for said subject based upon the breast cancer outcome
 The correlated genes may be used singly with significant
accuracy or in combination to increase the ability to accurately
correlating a molecular expression phenotype with a breast cancer
outcome. This correlation is a way to molecularly provide for the
determination of survival outcomes as disclosed herein. Additional
uses of the correlated gene(s) are in the classification of cells
and tissues; determination of diagnosis and/or prognosis; and determination
and/or alteration of therapy.
 The ability to discriminate is conferred by the identification
of expression of the individual genes as relevant and not by the
form of the assay used to determine the actual level of expression.
An assay may utilize any identifying feature of an identified individual
gene as disclosed herein as long as the assay reflects, quantitatively
or qualitatively, expression of the gene in the "transcriptome"
(the transcribed fraction of genes in a genome) or the "proteome"
(the translated fraction of expressed genes in a genome). Identifying
features include, but are not limited to, unique nucleic acid sequences
used to encode (DNA), or express (RNA), said gene or epitopes specific
to, or activities of, a protein encoded by said gene. All that is
required is the identity of the gene(s) necessary to discriminate
between breast cancer outcomes and an appropriate cell containing
sample for use in an expression assay.
 In one embodiment, the invention provides for the identification
of the gene expression patterns by analyzing global, or near global,
gene expression from single cells or homogenous cell populations
which have been dissected away from, or otherwise isolated or purified
from, contaminating cells beyond that possible by a simple biopsy.
Because the expression of numerous genes fluctuate between cells
from different patients as well as between cells from the same patient
sample, multiple data from expression of individual genes and gene
expression patterns are used as reference data to generate models
which in turn permit the identification of individual gene(s), the
expression of which are most highly correlated with particular breast
 In another aspect, the invention provides physical and methodological
means for detecting the expression of gene(s) identified by the
models generated by individual expression patterns. These means
may be directed to assaying one or more aspect of the DNA template(s)
underlying the expression of the gene(s), of the RNA used as an
intermediate to express the gene(s), or of the proteinaceous product
expressed by the gene(s).
 In a further aspect, the gene(s) identified by a model as
capable of discriminating between breast cancer outcomes may be
used to identify the cellular state of an unknown sample of cell(s)
from the breast. Preferably, the sample is isolated via non-invasive
means. The expression of said gene(s) in said unknown sample may
be determined and compared to the expression of said gene(s) in
reference data of gene expression patterns correlated with breast
cancer outcomes. Optionally, the comparison to reference samples
may be by comparison to the model(s) constructed based on the reference
 One advantage provided by the present invention is that
contaminating, non-breast cells (such as infiltrating lymphocytes
or other immune system cells) are not present to possibly affect
the genes identified or the subsequent analysis of gene expression
to identify the survival outcomes of patients with breast cancer.
Such contamination is present where a biopsy is used to generate
gene expression profiles.
 In another aspect, the invention provides the identification
and use of four gene sequences the expression of which are significantly
associated with tumor recurrence. Elevated expression of each one
of the four gene sequences is correlated with increased likelihood
of tumor recurrence and decreased patient survival. Therefore, the
expression of each of these gene sequences may be used in the same
manner as described herein for gene expression patterns.
 The first set of sequences is that of mitotic spindle associated
protein (also known as mitotic spindle coiled-coil related protein,
ASTRIN or DEEPEST). Human DEEPEST protein has been characterized
by Mack et al. (Proc Natl Acad Sci USA. 2001 98(25): 14434-9).
 The second set of sequences is that of the "Rac GTPase
activating protein 1" (RACGAP1).
 The third set of sequences is that of the "zinc finger
protein 145" or "PLZF" (Kruppel-like zinc finger
protein, expressed in promyelocytic leukemia) which is also referred
to as ZNF145.
 The fourth set of sequences is that of "MS4A7"
(membrane-spanning 4-domains, subfamily A, member 7).
 While the present invention is described mainly in the context
of human breast cancer, it may be practiced in the context of breast
cancer of any animal known to be potentially afflicted by breast
cancer. Preferred animals for the application of the present invention
are mammals, particularly those important to agricultural applications
(such as, but not limited to, cattle, sheep, horses, and other "farm
animals"), animal models of breast cancer, and animals for
human companionship (such as, but not limited to, dogs and cats).
DETAILED DESCRIPTION OF THE INVENTION
 Definitions of terms as used herein:
 A gene expression "pattern" or "profile"
or "signature" refers to the relative expression of a
gene between two or more breast cancer survival outcomes which is
correlated with being able to distinguish between said outcomes.
 A "gene" is a polynucleotide that encodes a discrete
product, whether RNA or proteinaceous in nature. It is appreciated
that more than one polynucleotide may be capable of encoding a discrete
product. The term includes alleles and polymorphisms of a gene that
encodes the same product, or a functionally associated (including
gain, loss, or modulation of function) analog thereof, based upon
chromosomal location and ability to recombine during normal mitosis.
 The terms "correlate" or "correlation"
or equivalents thereof refer to an association between expression
of one or more genes and a physiologic state of a breast cell to
the exclusion of one or more other state as identified by use of
the methods as described herein. A gene may be expressed at higher
or lower levels and still be correlated with one or more breast
cancer state or outcome.
 A "polynucleotide" is a polymeric form of nucleotides
of any length, either ribonucleotides or deoxyribonucleotides. This
term refers only to the primary structure of the molecule. Thus,
this term includes double- and single-stranded DNA and RNA. It also
includes known types of modifications including labels known in
the art, methylation, "caps", substitution of one or more
of the naturally occurring nucleotides with an analog, and internucleotide
modifications such as uncharged linkages (e.g., phosphorothioates,
phosphorodithioates, etc.), as well as unmodified forms of the polynucleotide.
 The term "amplify" is used in the broad sense
to mean creating an amplification product can be made enzymatically
with DNA or RNA polymerases. "Amplification," as used
herein, generally refers to the process of producing multiple copies
of a desired sequence, particularly those of a sample. "Multiple
copies" mean at least 2 copies. A "copy" does not
necessarily mean perfect sequence complementarity or identity to
the template sequence.
 By corresponding is meant that a nucleic acid molecule shares
a substantial amount of sequence identity with another nucleic acid
molecule. Substantial amount means at least 95%, usually at least
98% and more usually at least 99%, and sequence identity is determined
using the BLAST algorithm, as described in Altschul et al. (1990),
J. Mol. Biol. 215:403-410 (using the published default setting,
i.e. parameters w=4, t=17). Methods for amplifying mRNA are generally
known in the art, and include reverse transcription PCR (RT-PCR)
and those described in U.S. patent application Ser. No. 10/062,857
(filed on Oct. 25, 2001), as well as U.S. Provisional Patent Application
60/298,847 (filed Jun. 15, 2001) and 60/257,801 (filed Dec. 22,
2000), all of which are hereby incorporated by reference in their
entireties as if fully set forth. Another method which may be used
is quantitative PCR (or Q-PCR). Alternatively, RNA may be directly
labeled as the corresponding cDNA by methods known in the art.
 A "microarray" is a linear or two-dimensional
array of preferably discrete regions, each having a defined area,
formed on the surface of a solid support such as, but not limited
to, glass, plastic, or synthetic membrane. The density of the discrete
regions on a microarray is determined by the total numbers of immobilized
polynucleotides to be detected on the surface of a single solid
phase support, preferably at least about 50/cm2, more preferably
at least about 100/cm2, even more preferably at least about 500/cm2,
but preferably below about 1,000/cm2. Preferably, the arrays contain
less than about 500, about 1000, about 1500, about 2000, about 2500,
or about 3000 immobilized polynucleotides in total. As used herein,
a DNA microarray is an array of oligonucleotides or polynucleotides
placed on a chip or other surfaces used to hybridize to amplified
or cloned polynucleotides from a sample. Since the position of each
particular group of primers in the array is known, the identities
of a sample polynucleotides can be determined based on their binding
to a particular position in the microarray.
 Because the invention relies upon the identification of
genes that are over- or under-expressed, one embodiment of the invention
involves determining expression by hybridization of mRNA, or an
amplified or cloned version thereof, of a sample cell to a polynucleotide
that is unique to a particular gene sequence. Preferred polynucleotides
of this type contain at least about 20, at least about 22, at least
about 24, at least about 26, at least about 28, at least about 30,
or at least about 32 consecutive basepairs of a gene sequence that
is not found in other gene sequences. The term "about"
as used in the previous sentence refers to an increase or decrease
of 1 from the stated numerical value. Even more preferred are polynucleotides
of at least or about 50, at least or about 100, at least about or
150, at least or about 200, at least or about 250, at least or about
300, at least or about 350, or at least or about 400 basepairs of
a gene sequence that is not found in other gene sequences. The term
"about" as used in the preceding sentence refers to an
increase or decrease of 10% from the stated numerical value. Such
polynucleotides may also be referred to as polynucleotide probes
that are capable of hybridizing to sequences of the genes, or unique
portions thereof, described herein. Preferably, the sequences are
those of mRNA encoded by the genes, the corresponding cDNA to such
mRNAs, and/or amplified versions of such sequences. In preferred
embodiments of the invention, the polynucleotide probes are immobilized
on an array, other devices, or in individual spots that localize
 Alternatively, and in another embodiment of the invention,
gene expression may be determined by analysis of expressed protein
in a cell sample of interest by use of one or more antibodies specific
for one or more epitopes of individual gene products (proteins)
in said cell sample. Such antibodies are preferably labeled to permit
their easy detection after binding to the gene product.
 The term "label" refers to a composition capable
of producing a detectable signal indicative of the presence of the
labeled molecule. Suitable labels include radioisotopes, nucleotide
chromophores, enzymes, substrates, fluorescent molecules, chemiluminescent
moieties, magnetic particles, bioluminescent moieties, and the like.
As such, a label is any composition detectable by spectroscopic,
photochemical, biochemical, immunochemical, electrical, optical
or chemical means.
 The term "support" refers to conventional supports
such as beads, particles, dipsticks, fibers, filters, membranes
and silane or silicate supports such as glass slides.
 As used herein, a "breast tissue sample" or "breast
cell sample" refers to a sample of breast tissue or fluid isolated
from an individual suspected of being afflicted with, or at risk
of developing, breast cancer. Such samples are primary isolates
(in contrast to cultured cells) and may be collected by any non-invasive
means, including, but not limited to, ductal lavage, fine needle
aspiration, needle biopsy, the devices and methods described in
U.S. Pat. No. 6,328,709, or any other suitable means recognized
in the art. Alternatively, the "sample" may be collected
by an invasive method, including, but not limited to, surgical biopsy.
 "Expression" and "gene expression" include
transcription and/or translation of nucleic acid material.
 As used herein, the term "comprising" and its
cognates are used in their inclusive sense; that is, equivalent
to the term "including" and its corresponding cognates.
 Conditions that "allow" an event to occur or conditions
that are "suitable" for an event to occur, such as hybridization,
strand extension, and the like, or "suitable" conditions
are conditions that do not prevent such events from occurring. Thus,
these conditions permit, enhance, facilitate, and/or are conducive
to the event. Such conditions, known in the art and described herein,
depend upon, for example, the nature of the nucleotide sequence,
temperature, and buffer conditions. These conditions also depend
on what event is desired, such as hybridization, cleavage, strand
extension or transcription.
 Sequence "mutation," as used herein, refers to
any sequence alteration in the sequence of a gene disclosed herein
interest in comparison to a reference sequence. A sequence mutation
includes single nucleotide changes, or alterations of more than
one nucleotide in a sequence, due to mechanisms such as substitution,
deletion or insertion. Single nucleotide polymorphism (SNP) is also
a sequence mutation as used herein. Because the present invention
is based on the relative level of gene expression, mutations in
non-coding regions of genes as disclosed herein may also be assayed
in the practice of the invention.
 "Detection" includes any means of detecting, including
direct and indirect detection of gene expression and changes therein.
For example, "detectably less" products may be observed
directly or indirectly, and the term indicates any reduction (including
the absence of detectable signal). Similarly, "detectably more"
product means any increase, whether observed directly or indirectly.
 Unless defined otherwise all technical and scientific terms
used herein have the same meaning as commonly understood to one
of ordinary skill in the art to which this invention belongs.
 The present invention relates to the identification and
use of gene expression patterns (or profiles or "signatures")
which discriminate between (or are correlated with) breast cancer
survival and recurrence outcomes in a subject. Such patterns may
be determined by the methods of the invention by use of a number
of reference cell or tissue samples, such as those reviewed by a
pathologist of ordinary skill in the pathology of breast cancer,
which reflect breast cancer cells as opposed to normal or other
non-cancerous cells. The outcomes experienced by the subjects from
whom the samples may be correlated with expression data to identify
patterns that correlate with the outcomes. Because the overall gene
expression profile differs from person to person, cancer to cancer,
and cancer cell to cancer cell, correlations between certain cells
and genes expressed or underexpressed may be made as disclosed herein
to identify genes that are capable of discriminating between breast
 The present invention may be practiced with any number of
the genes believed, or likely to be, differentially expressed with
respect to breast cancer outcomes. The identification may be made
by using expression profiles of various homogenous breast cancer
cell populations, which were isolated by microdissection, such as,
but not limited to, laser capture microdissection (LCM) of 100-1000
cells. The expression level of each gene of the expression profile
may be correlated with a particular outcome. Alternatively, the
expression levels of multiple genes may be clustered to identify
correlations with particular outcomes.
 Genes with significant correlations to breast cancer survival
or recurrence outcomes may be used to generate models of gene expressions
that would maximally discriminate between outcomes. Alternatively,
genes with significant correlations may be used in combination with
genes with lower correlations without significant loss of ability
to discriminate between outcomes. Such models may be generated by
any appropriate means recognized in the art, including, but not
limited to, cluster analysis, supported vector machines, neural
networks or other algorithm known in the art. The models are capable
of predicting the classification of a unknown sample based upon
the expression of the genes used for discrimination in the models.
"Leave one out" cross-validation may be used to test the
performance of various models and to help identify weights (genes)
that are uninformative or detrimental to the predictive ability
of the models. Cross-validation may also be used to identify genes
that enhance the predictive ability of the models.
 The gene(s) identified as correlated with particular breast
cancer outcomes by the above models provide the ability to focus
gene expression analysis to only those genes that contribute to
the ability to identify a subject as likely to have a particular
outcome relative to another. The expression of other genes in a
breast cancer cell would be relatively unable to provide information
concerning, and thus assist in the discrimination of, a breast cancer
 As will be appreciated by those skilled in the art, the
models are highly useful with even a small set of reference gene
expression data and can become increasingly accurate with the inclusion
of more reference data although the incremental increase in accuracy
will likely diminish with each additional datum. The preparation
of additional reference gene expression data using genes identified
and disclosed herein for discriminating between different outcomes
in breast cancer is routine and may be readily performed by the
skilled artisan to permit the generation of models as described
above to predict the status of an unknown sample based upon the
expression levels of those genes.
 To determine the (increased or decreased) expression levels
of genes in the practice of the present invention, any method known
in the art may be utilized. In one preferred embodiment of the invention,
expression based on detection of RNA which hybridizes to the genes
identified and disclosed herein is used. This is readily performed
by any RNA detection or amplification+detection method known or
recognized as equivalent in the art such as, but not limited to,
reverse transcription-PCR, the methods disclosed in U.S. patent
application Ser. No. 10/062,857 (filed on Oct. 25, 2001) as well
as U.S. Provisional Patent Application 60/298,847 (filed Jun. 15,
2001) and 60/257,801 (filed Dec. 22, 2000), and methods to detect
the presence, or absence, of RNA stabilizing or destabilizing sequences.
 Alternatively, expression based on detection of DNA status
may be used. Detection of the DNA of an identified gene as methylated
or deleted may be used for genes that have decreased expression
in correlation with a particular breast cancer outcome. This may
be readily performed by PCR based methods known in the art, including,
but not limited to, Q-PCR. Conversely, detection of the DNA of an
identified gene as amplified may be used for genes that have increased
expression in correlation with a particular breast cancer outcome.
This may be readily performed by PCR based, fluorescent in situ
hybridization (FISH) and chromosome in situ hybridization (CISH)
methods known in the art.
 Expression based on detection of a presence, increase, or
decrease in protein levels or activity may also be used. Detection
may be performed by any immunohistochemistry (IHC) based, blood
based (especially for secreted proteins), antibody (including autoantibodies
against the protein) based, exfoliate cell (from the cancer) based,
mass spectroscopy based, and image (including used of labeled ligand)
based method known in the art and recognized as appropriate for
the detection of the protein. Antibody and image based methods are
additionally useful for the localization of tumors after determination
of cancer by use of cells obtained by a non-invasive procedure (such
as ductal lavage or fine needle aspiration), where the source of
the cancerous cells is not known. A labeled antibody or ligand may
be used to localize the carcinoma(s) within a patient.
 A preferred embodiment using a nucleic acid based assay
to determine expression is by immobilization of one or more sequences
of the genes identified herein on a solid support, including, but
not limited to, a solid substrate as an array or to beads or bead
based technology as known in the art. Alternatively, solution based
expression assays known in the art may also be used. The immobilized
gene(s) may be in the form of polynucleotides that are unique or
otherwise specific to the gene(s) such that the polynucleotide would
be capable of hybridizing to a DNA or RNA corresponding to the gene(s).
These polynucleotides may be the full length of the gene(s) or be
short sequences of the genes (up to one nucleotide shorter than
the full length sequence known in the art by deletion from the 5'
or 3' end of the sequence) that are optionally minimally interrupted
(such as by mismatches or inserted non-complementary basepairs)
such that hybridization with a DNA or RNA corresponding to the gene(s)
is not affected. Preferably, the polynucleotides used are from the
3' end of the gene. Polynucleotides containing mutations relative
to the sequences of the disclosed genes may also be used so long
as the presence of the mutations still allows hybridization to produce
a detectable signal.
 The immobilized gene(s) may be used to determine the state
of nucleic acid samples prepared from sample breast cell(s) for
which the outcome of the sample's subject (e.g. patient from whom
the sample is obtained) is not known or for confirmation of an outcome
that is already assigned to the sample's subject. Without limiting
the invention, such a cell may be from a patient with breast cancer
or alternatively suspected of being afflicted with, or at risk of
developing, breast cancer. The immobilized polynucleotide(s) need
only be sufficient to specifically hybridize to the corresponding
nucleic acid molecules derived from the sample under suitable conditions.
While even a single correlated gene sequence may to able to provide
adequate accuracy in discriminating between two breast cancer outcomes,
two or more, three or more, four or more, five or more, six or more,
seven or more, eight or more, nine or more, ten or more, or eleven
or more of the genes identified herein may be used as a subset capable
of discriminating may be used in combination to increase the accuracy
of the method. The invention specifically contemplates the selection
of more than one, two or more, three or more, four or more, five
or more, six or more, seven or more, eight or more, nine or more,
ten or more, or eleven or more of the genes disclosed in the tables
and figures herein for use as a subset in the identification of
breast cancer survival outcome.
 Of course 15 or more, 20 or more, 30 or more, 40 or more,
50 or more, 60 or more, 70 or more, 80 or more, 90 or more, 100
or more, 110 or more, 120 or more, 130 or more, 140 or more, or
all the genes provided in Tables 2, 3, and/or 4 below may be used.
"CloneID" as used in the context of Tables 2, 3, and 4
as well as the present invention refers to the IMAGE Consortium
clone ID number of each gene, the sequences of which are hereby
incorporated by reference in their entireties as they are available
from the Consortium at http://image.llnl.gov/ as accessed on the
filing date of the present application. Also provided in the tables
are GenBank accession numbers which are comprised of letters, numbers
and optionally underscores. P value refers to values assigned as
described in the Examples below. The indications of "E-xx"
where "xx" is a two digit number refers to alternative
notation for exponential figures where "E-xx" is "10-xx".
Thus in combination with the numbers to the left of "E-xx",
the value being represented is the numbers to the left times 10-xx.
Description provides a brief identifier of what the gene encodes.
 Genes with a correlation identified by a p value below or
about 0.02, below or about 0.01, below or about 0.005, or below
or about 0.001 are preferred for use in the practice of the invention.
The present invention includes the use of genes that identify different
ER positive subtypes and breast cancer recurrence and invasive tumor
grade to permit simultaneous identification of breast cancer survival
outcome of a patient based upon assaying a breast cancer sample
from said patient.
 In embodiments where only one or a few genes are to be analyzed,
the nucleic acid derived from the sample breast cancer cell(s) may
be preferentially amplified by use of appropriate primers such that
only the genes to be analyzed are amplified to reduce contaminating
background signals from other genes expressed in the breast cell.
Alternatively, and where multiple genes are to be analyzed or where
very few cells (or one cell) is used, the nucleic acid from the
sample may be globally amplified before hybridization to the immobilized
polynucleotides. Of course RNA, or the cDNA counterpart thereof
may be directly labeled and used, without amplification, by methods
known in the art.
 The above assay embodiments may be used in a number of different
ways to identify or detect the invasive breast cancer grade, if
any, of a breast cancer cell sample from a patient. In many cases,
this would reflect a secondary screen for the patient, who may have
already undergone mammography or physical exam as a primary screen.
If positive, the subsequent needle biopsy, ductal lavage, fine needle
aspiration, or other analogous methods may provide the sample for
use in the above assay embodiments. The present invention may be
used in combination with non-invasive protocols, such as ductal
lavage or fine needle aspiration, to prepare a breast cell sample.
 The present invention provides a more objective set of criteria,
in the form of gene expression profiles of a discrete set of genes,
to discriminate (or delineate) between breast cancer outcomes. In
particularly preferred embodiments of the invention, the assays
are used to discriminate between good and poor outcomes within 5,
or about 5, years after surgical intervention to remove breast cancer
tumors or within about 95 months after surgical intervention to
remove breast cancer tumors. Comparisons that discriminate between
outcomes after about 10, about 20, about 30, about 40, about 50,
about 60, about 70, about 80, about 90, or about 100 months may
also be performed.
 While good and poor survival outcomes may be defined relatively
in comparison to each other, a "good" outcome may be viewed
as a better than 50% survival rate after about 60 months post surgical
intervention to remove breast cancer tumor(s). A "good"
outcome may also be a better than about 60%, about 70%, about 80%
or about 90% survival rate after about 60 months post surgical intervention.
A "poor" outcome may be viewed as a 50% or less survival
rate after about 60 months post surgical intervention to remove
breast cancer tumor(s). A "poor" outcome may also be about
a 70% or less survival rate after about 40 months, or about a 80%
or less survival rate after about 20 months, post surgical intervention.
 In one embodiment of the invention, the isolation and analysis
of a breast cancer cell sample may be performed as follows:
 (1) Ductal lavage or other non-invasive procedure is performed
on a patient to obtain a sample.
 (2) Sample is prepared and coated onto a microscope slide.
Note that ductal lavage results in clusters of cells that are cytologically
examined as stated above.
 (3) Pathologist or image analysis software scans the sample
for the presence of non-normal and/or atypical breast cancer cells.
 (4) If such cells are observed, those cells are harvested
(e.g. by microdissection such as LCM).
 (5) RNA is extracted from the harvested cells.
 (6) RNA is purified, amplified, and labeled.
 (7) Labeled nucleic acid is contacted with a microarray
containing polynucleotides of the genes identified herein as correlated
to discriminations between breast cancer outcomes under suitable
hybridization conditions, then processed and scanned to obtain a
pattern of intensities of each spot (relative to a control for general
gene expression in cells) which determine the level of expression
of the gene(s) in the cells.
 (8) The pattern of intensities is analyzed by comparison
to the expression patterns of the genes in known samples of breast
cancer cells correlated with outcomes (relative to the same control).
 A specific example of the above method would be performing
ductal lavage following a primary screen, observing and collecting
non-normal and/or atypical cells for analysis. The comparison to
known expression patterns, such as that made possible by a model
generated by an algorithm (such as, but not limited to nearest neighbor
type analysis, SVM, or neural networks) with reference gene expression
data for the different breast cancer survival outcomes, identifies
the cells as being correlated with subjects with good or poor outcomes.
Another example would be taking a breast tumor removed from a subject
after surgical intervention, isolation and preparation of breast
cancer cells from the tumor for determination/identification of
atypical, non-normal, or cancer cells, and isolation of said cells
followed by steps 5 through 8 above.
 Alternatively, the sample may permit the collection of both
normal as well as cancer cells for analysis. The gene expression
patterns for each of these two samples will be compared to each
other as well as the model and the normal versus individual comparisons
therein based upon the reference data set. This approach can be
significantly more powerful that the cancer cells only approach
because it utilizes significantly more information from the normal
cells and the differences between normal and cancer cells (in both
the sample and reference data sets) to determine the breast cancer
outcome of the patient based on gene expression in the cancer cells
from the sample.
 With use of the present invention, skilled physicians may
prescribe treatments based on prognosis determined via non-invasive
samples that they would have prescribed for a patient which had
previously received a diagnosis via a solid tissue biopsy.
 The above discussion is also applicable where a palpable
lesion is detected followed by fine needle aspiration or needle
biopsy of cells from the breast. The cells are plated and reviewed
by a pathologist or automated imaging system which selects cells
for analysis as described above.
 The present invention may also be used, however, with solid
tissue biopsies. For example, a solid biopsy may be collected and
prepared for visualization followed by determination of expression
of one or more genes identified herein to determine the breast cancer
outcome. One preferred means is by use of in situ hybridization
with polynucleotide or protein identifying probe(s) for assaying
expression of said gene(s).
 In an alternative method, the solid tissue biopsy may be
used to extract molecules followed by analysis for expression of
one or more gene(s). This provides the possibility of leaving out
the need for visualization and collection of only cancer cells or
cells suspected of being cancerous. This method may of course be
modified such that only cells that have been positively selected
are collected and used to extract molecules for analysis. This would
require visualization and selection as a prerequisite to gene expression
 In a further modification of the above, both normal cells
and cancer cells are collected and used to extract molecules for
analysis of gene expression. The approach, benefits and results
are as described above using non-invasive sampling.
 The genes identified herein may be used to generate a model
capable of predicting the breast cancer survival and recurrence
outcomes of an unknown breast cell sample based on the expression
of the identified genes in the sample. Such a model may be generated
by any of the algorithms described herein or otherwise known in
the art as well as those recognized as equivalent in the art using
gene(s) (and subsets thereof) disclosed herein for the identification
of breast cancer outcomes. The model provides a means for comparing
expression profiles of gene(s) of the subset from the sample against
the profiles of reference data used to build the model. The model
can compare the sample profile against each of the reference profiles
or against a model defining delineations made based upon the reference
profiles. Additionally, relative values from the sample profile
may be used in comparison with the model or reference profiles.
 In a preferred embodiment of the invention, breast cell
samples identified as normal and cancerous from the same subject
may be analyzed for their expression profiles of the genes used
to generate the model. This provides an advantageous means of identifying
survival and recurrence outcomes based on relative differences from
the expression profile of the normal sample. These differences can
then be used in comparison to differences between normal and individual
cancerous reference data which was also used to generate the model.
 The detection of gene expression from the samples may be
by use of a single microarray able to assay gene expression from
some or all genes disclosed herein for convenience and accuracy.
 Other uses of the present invention include providing the
ability to identify breast cancer cell samples as correlated with
particular breast cancer survival or recurrence outcomes for further
research or study. This provides a particular advantage in many
contexts requiring the identification of cells based on objective
genetic or molecular criteria.
 The materials for use in the methods of the present invention
are ideally suited for preparation of kits produced in accordance
with well known procedures. The invention thus provides kits comprising
agents for the detection of expression of the disclosed genes for
identifying breast cancer outcomes. Such kits optionally comprising
the agent with an identifying description or label or instructions
relating to their use in the methods of the present invention, is
provided. Such a kit may comprise containers, each with one or more
of the various reagents (typically in concentrated form) utilized
in the methods, including, for example, pre-fabricated microarrays,
buffers, the appropriate nucleotide triphosphates (e.g., dATP, dCTP,
dGTP and dTTP; or rATP, rCTP, rGTP and UTP), reverse transcriptase,
DNA polymerase, RNA polymerase, and one or more primer complexes
of the present invention (e.g., appropriate length poly(T) or random
primers linked to a promoter reactive with the RNA polymerase).
A set of instructions will also typically be included.
 The methods provided by the present invention may also be
automated in whole or in part. All aspects of the present invention
may also be practiced such that they consist essentially of a subset
of the disclosed genes to the exclusion of material irrelevant to
the identification of breast cancer survival outcomes via a cell
 Having now generally described the invention, the same will
be more readily understood through reference to the following examples
which are provided by way of illustration, and are not intended
to be limiting of the present invention, unless specified.
 Clinical specimen collection and clinicopathological parameters.
Laser capture microdissected invasive cancer cells from a total
of 124 breast cancer biopsies were used to discover two sets of
genes, the expression levels of which correlate with clinical breast
cancer outcomes. These genes could thus be used either individually
or in combination as prognostic factors for breast cancer management.
The characteristics of the 124 patient profiles in the study are
shown in Table 1.
 Relative expression levels of .about.22000 genes were measured
from the invasive cancer cells for each of the 124 patients. Genes
varying by at least 3-fold from the median expression level across
the 124 patients in at least 10 patients were selected, resulting
in 7090 genes.
 In particular, 4 genes (DEEPEST, RACGAP1, ZNF145, MS4A7)
were shown to be strong prognostic factors individually for predicting
tumor recurrence after surgery and adjuvant therapies.
1 TABLE 1 Group N % Age <=45 30 24.2 45-55 27 21.8 >55 67
54 ER positive 66 53.2 negative 58 46.7 Node positive 58 55.2 negative
47 44.8 Not avail. 19 15 Grade 1 8 9.6 2 29 34.9 3 46 55.4 Not avail.
Identification of ER Positive Subtypes with Different Survival
 Hierarchical clustering, based on the 7090 genes described
in Example 1, of the resulting gene expression matrix (7090.times.124)
revealed a cluster of 67-genes (the Ki67 set) the expressions of
which differentiates estrogen receptor positive patients into two
subgroups with distinct clinical outcomes based on overall survival
 As shown in FIG. 1, left panel, a Kaplan-Meier curve on
the left compares the disease-free survival of patients based on
ER status, which shows slightly better survival for ER positive
patients but with an insignificant p value (log-rank test). In contrast,
and as shown in the right panel, when the ER positive patients are
subdivided into two subgroups (A and B) based on the expression
levels of the Ki67 signature genes, which are all expressed at levels
above the median to define subgroup A and below the median to define
 The three-group (ER+, subgroup A; ER+, subgroup B; and ER-)
comparison shows significant differences in survival such that subgroup
B subjects had significantly better survival outcomes than those
of subgroup A. The ER- curve remains unchanged. This indicates that
the Ki67 signature, and individual or groups of genes therein, can
be used to subdivide ER positive patients into two clinically distinct
subgroups based upon survival outcomes.
 The identities of the genes in the Ki67 signature are shown
in Table 2.
2TABLE 2 Genes, the expressions of which define two ER+ subgroups
CloneID Gene Description 2967734 BC007491 EXO1 .vertline. exonuclease
1 NM_000057 BLM .vertline. Bloom syndrome 2849551 AW512559 CDC25C
.vertline. cell division cycle 25C 3634656 BC010044 CDC20 .vertline.
CDC20 cell division cycle 20 homolog (S. cerevisiae) 2961114 BC008718
BIRC5 .vertline. baculoviral IAP repeat-containing 5 (survivin)
NM_012112 C20orf1 .vertline. chromosome 20 open reading frame 1
AF399910 DEEPEST .vertline. mitotic spindle coiled-coil related
protein NM_032997 ZWINT .vertline. ZW10 interactor AF331796 HCAP-G
.vertline. chromosome condensation protein G 2175265 AI524385 ANLN
.vertline. "anillin, actin binding protein (scraps homolog,
Drosophila)" 3873367 BC010658 KIAA0008 .vertline. KIAA0008
gene product 1338423 AA810180 FLJ10517 .vertline. hypothetical protein
FLJ10517 AL136794 RACGAP1 .vertline. Rac GTPase activating protein
1 AF334184 FKSG42 .vertline. FKSG42 4420248 BC017705 KNSL5 .vertline.
kinesin-like 5 (mitotic kinesin-like protein 1) AB035898 KNSL7 .vertline.
kinesin-like 7 1240937 AA714213 ESTs, Highly similar to T47163 hypothetical
protein DKFZp762E1312.1 [H. sapiens] 2820741 BC001940 DKFZp762E1312
.vertline. hypothetical protein DKFZp762E1312 AF017790 HEC .vertline.
"highly expressed in cancer, rich in leucine heptad repeats"
4048625 BC013919 TYMS .vertline. thymidylate synthetase 1699365
AI049877 KIAA0186 .vertline. KIAA0186 gene product 3139011 BC001459
RAD51 .vertline. "RAD51 homolog (RecA homolog, E. coli) (S.
cerevisiae)" AF053306 BUB1B .vertline. BUB1 budding uninhibited
by benzimidazoles 1 homolog beta (yeast) 3908972 BC015050 OIP5 .vertline.
Opa-interacting protein 5 1911633 AI268609 ESPL1 .vertline. extra
spindle poles like 1 (S. cerevisiae) NM_002417 MKI67 .vertline.
antigen identified by monoclonal antibody Ki-67 2988318 BC013966
FLJ10156 .vertline. hypothetical protein 2964488 BC013300 STK12
.vertline. serine/threonine kinase 12 NM_016343 CENPF .vertline.
"centromere protein F (350/400 kD, mitosin)" AF161499
HSPC150 .vertline. HSPC150 protein similar to ubiquitin-conjugating
enzyme AL136840 MCM10 .vertline. MCM10 minichromosome maintenance
deficient 10 (S. cerevisiae) 3028566 BC008947 FLJ10540 .vertline.
hypothetical protein FLJ10540 1986322 AI273114 ESTs, Weakly similar
to I78885 serine/threonine-specific protein kinase [H. sapiens]
3909951 BC015706 Homo sapiens, Similar to RIKEN cDNA 2810433K01
gene, clone MGC: 10200 IMAGE: 3909951, mRNA, complete cds AF095289
PTTG3 .vertline. pituitary tumor-transforming 3 2262695 AI811894
PTTG2 .vertline. pituitary tumor-transforming 2 1186167 AA648922
CDC25A .vertline. cell division cycle 25A 3138951 BC002551 MGC2577
.vertline. hypothetical protein MGC2577 U74612 FOXM1 .vertline.
forkhead box M1 3347875 BC000703 FLJ10468 .vertline. hypothetical
protein FLJ10468 3347571 BC008764 KNSL6 .vertline. kinesin-like
6 (mitotic centromere-associated kinesin) 2822981 BC000404 TRIP13
.vertline. thyroid hormone receptor interactor 13 1678629 AI082049
ESTs NM_003504 CDC45L .vertline. CDC45 cell division cycle 45-like
(S. cerevisiae) 3345575 BC007656 UBE2C .vertline. ubiquitin-conjugating
enzyme E2C D84212 STK6 .vertline. serine/threonine kinase 6 AF011468
STK15 .vertline. serine/threonine kinase 15 3938081 BC011000 MGC16386
.vertline. similar to RIKEN cDNA2610036L13 AF277375 KIF4A .vertline.
kinesin family member 4A 3461992 BC000881 CENPA .vertline. centromere
protein A (17 kD) AF108138 PIF1 .vertline. DMA helicase homolog
PIF1 AF155827 HELLS .vertline. "helicase, lymphoid-specific"
NM_018492 TOPK .vertline. T-LAK cell-originated protein kinase 1686560
AI088843 ESTs 1241465 AA715810 ESTs, Weakly similar to YK61_YEAST
HYPOTHETICAL 39.6 KDA PROTEIN IN MTD1-NUP133 INTERGENIC REGION [S.
cerevisiae] 2823731 BC001068 C20orf129 .vertline. chromosome 20
open reading frame 129 AK026964 FLJ23311 .vertline. hypothetical
protein FLJ23311 3996265 BC005389 LOC51053 .vertline. geminin 3901250
BC010858 EZH2 .vertline. enhancer of zeste homolog 2 (Drosophila)
4547136 BC014039 KIAA0175 .vertline. likely ortholog of maternal
embryonic leucine zipper kinase 4091997 BC017575 CHEK1 .vertline.
CHK1 checkpoint homolog (S. pombe) 669114 AA232651 SUV39H2 .vertline.
suppressor of variegation 3-9 (Drosophila) homolog 2; hypothetical
protein FLJ23414 NM_002497 NEK2 .vertline. NIMA (never in mitosis
gene a)-related kinase 2 4107592 BC016330 PIR51 .vertline. RAD51-interacting
protein AF025840 POLE2 .vertline. "polymerase (DMA directed),
epsilon 2" 3510656 BC007633 EIF2C2 .vertline. "eukaryotic
translation initiation factor 2C, 2" AL050151 Homo sapiens
mRNA; cDNA DKFZp586J0720 (from clone DKFZp586J0720)
Molecular Signature that Correlates with the Recurrence of Breast
 A molecular signature that correlates with recurrence of
breast cancer after removal of cancer by surgery was identified
as follows. Each of the 7090 genes from Example 1 was used to fit
a univariate Cox proportional hazard regression model using the
survival information available for the patients in the study. A
total of 143 genes with significant p values (p<0.01) in these
univariate models were selected. Hierarchical clustering of patient
samples by the 143 recurrence-associated genes identified them as
having expression levels that correlated with the absence or presence
of breast cancer recurrence.
 These 143 genes are shown in Table 3. The sign of the coefficient
values in Table 3 correspond to whether a gene is positively or
negatively correlated with breast cancer recurrence. A positive
coefficient means that the gene is positively correlated (overexpressed)
in patients with a poor (shorter) survival outcome due to recurrence
and negatively correlated (underexpressed) in patients with a good
or better (longer) survival outcome due to the relative absence
of recurrence. A negative coefficient means that the gene is positively
correlated (overexpressed) in patients with a good or better (longer)
survival outcome (due to the relative absence of cancer recurrence)
and negatively correlated (underexpressed) in patients with a poor
(shorter) survival outcome (due to cancer recurrence).
 To validate this gene set, 22 of the top 27 genes from Table
3 (with the smallest p values) were mapped onto the microarray used
by van't Veer et al. (Supra) via the Unigene database. The top 27
genes are provided in Table 4 while the mapping of genes are shown
in Table 5 (showing identities of the genes via their GenBank ID,
van't Veer et al. reference, and Unigene ID numbers). Thirteen of
the 22 genes were filtered out due to low variance across the sample
set, reducing the number of genes for cluster analysis to 9. The
27 gene set was used with the data from the patients of Example
1 to classify them as being in either the good prognosis or the
poor prognosis group by hierarchical clustering based on disease-free
survival. The results are shown in FIG. 2, left panel (Kaplan-Meier
curves of patients stratified by the top 27 recurrence-associated
 The 9 genes not filtered out from the van't Veer et al.
data were used to with the patient data therein to classify them
as being in either the good prognosis or the poor prognosis group
by hierarchical clustering based on disease-free survival. The results
are shown in FIG. 2, right panel (Kaplan-Meier curves of patients
stratified by 9 of the top 27 recurrence-associated genes).
 Like FIG. 1, the horizontal axis of FIG. 2 is in time (months
or years) and the vertical axis is in survival probability (where
1.0 is survival of 100% of the subjects and 0.5 is survival of 50%
of the subjects). As shown in FIG. 2, differences in disease-free
survival between the two groups in both datasets were highly significant.
3TABLE 3 Genes, the expressions of which correlate with breast
cancer recurrence Clone ID gene p coef desc 1184567 AA648777 7.58E-06
-2.3582882 MS4A7 .vertline. membrane-spanning 4-domains, subfamily
A, member 7 2961112 BC005850 1.05E-04 -1.845548 CBFA2T1 .vertline.
core-binding factor, runt domain, alpha subunit 2; translocated
to, 1; cyclin D-related 3565773 BF432813 1.65E-03 -1.2898777 KLRB1
.vertline. killer cell lectin-like receptor subfamily B, member
1 1352935 AA830131 7.60E-03 -1.2502516 ZNF80 .vertline. zinc finger
protein 80 (pT17) 3915193 BC017022 1.80E-03 -0.9385143 Homo sapiens,
clone MGC: 8979 IMAGE: 3915193, mRNA, complete cds 2630949 AW150267
2.87E-04 -0.9172496 C21orf9 .vertline. chromosome 21 open reading
frame 9 2714519 AW137991 1.22E-03 -0.9027559 RELB .vertline. v-rel
reticuloendotheliosis viral oncogene homolog B, nuclear factor of
kappa light polypeptide gene enhancer in B-cells 3 (avian) 2365891
AI741785 7.56E-03 -0.8965429 SLIT3 .vertline. slit homolog 3 (Drosophila)
NM_006006 8.59E-04 -0.8880113 ZNF145 .vertline. zinc finger protein
145 (Kruppel-like, expressed in promyelocytic leukemia) 3645909
BF436656 3.08E-03 -0.8506381 MFAP4 .vertline. microfibrillar-associated
protein 4 2349778 AI806109 3.89E-03 -0.8414599 KIAA1580 .vertline.
KIAA1580 protein 3504259 BC000723 7.95E-03 -0.8413126 CRAT .vertline.
camitine acetyltransferase 4342203 BC018538 2.92E-04 -0.8358585
ALOX5AP .vertline. arachidonate 5-lipoxygenase-activating protein
AL122052 4.91E-04 -0.8340079 KIAA0793 .vertline. KIAA0793 gene product
AK025091 6.06E-03 -0.8263796 FLJ21438 .vertline. hypothetical protein
FLJ21438 2612878 AW130888 2.14E-03 -0.8099805 PTK2B .vertline. protein
tyrosine kinase 2 beta AF244129 8.92E-03 -0.807578 LY9 .vertline.
lymphocyte antigen 9 AK027120 8.04E-03 -0.8071951 FLJ23467 .vertline.
hypothetical protein FLJ23467 AF367473 1.85E-04 -0.8035463 NYD-SP21
.vertline. testes development-related NYD-SP21 4214447 BC009032
4.52E-03 -0.7976367 PR48 .vertline. protein phosphatase 2A 48 kDa
regulatory subunit AB045832 5.30E-03 -0.767161 P53AIP1 .vertline.
p53-regulated apoptosis-inducing protein 1 NM_000598 2.64E-03 -0.7567474
IGFBP3 .vertline. insulin-like growth factor binding protein 3 AI952055
5.06E-03 -0.7474092 ESTs NM_003734 4.42E-03 -0.7298537 AOC3 .vertline.
amine oxidase, copper containing 3 (vascular adhesion protein 1)
4291158 BC008392 6.23E-03 -0.7260119 UCP3 .vertline. uncoupling
protein 3 (mitochondrial, proton earner) 3840457 BC012990 2.07E-03
-0.7130147 Homo sapiens, clone IMAGE: 3840457, mRNA AB037886 2.20E-03
-0.6839819 NESH .vertline. NESH protein 3622951 BC004300 6.00E-03
-0.6820265 VILL .vertline. villin-like NM_015385 6.90E-03 -0.6555434
SH3D5 .vertline. SH3-domain protein 5 (ponsin) 289749 N59284 4.93E-03
-0.6497849 ESTs 3677098 BC004864 2.47E-03 -0.6450665 PPP3CC .vertline.
protein phosphatase 3 (formerly 2B), catalytic subunit, gamma isoform
(calcineurin A gamma) 2254324 AI620965 1.08E-03 -0.6415661 ESTs
1848897 AI247901 8.54E-03 -0.6392025 ESTs, Weakly similar to S23650
retrovirus-related hypothetical protein II [H. sapiens] 4699374
BC017839 1.24E-03 -0.6348455 CASP4 .vertline. caspase 4, apoptosis-related
cysteine protease U90878 3.45E-03 -0.6303001 PDLIM1 .vertline. PDZ
and LIM domain 1 (elfin) 2729801 AW293849 3.26E-03 -0.6188886 ESTs,
Moderately similar to I54374 gene NF2 protein [H. sapiens] AL137694
9.86E-03 -0.6154861 FLJ11286 .vertline. hypothetical protein FLJ11286
1884362 AI215902 8.92E-04 -0.6139098 ESTs, Highly similar to T50835
hypothetical protein [H. sapiens] 3543310 BC001609 7.61E-03 -0.6113849
WBSCR5 .vertline. Williams-Beuren syndrome chromosome region 5 1869453
AI264644 5.93E-03 -0.6105146 KIAA0775 .vertline. KIAA0775 gene product
3010091 BC006107 1.60E-03 -0.6088258 ARHGAP9 .vertline. Rho GTPase
activating protein 9 NM_002405 5.53E-03 -0.6041611 MFNG .vertline.
manic fringe homolog (Drosophila) AK026343 1.86E-04 -0.6018579 FLJ22690
.vertline. hypothetical protein FLJ22690 2227051 AI583109 2.88E-03
-0.5977938 STAT5A .vertline. signal transducer and activator of
transcription 5A 1144648 AA613560 6.79E-03 -0.5968032 ALOX5 .vertline.
arachidonate 5-lipoxygenase 206683 H59559 8.44E-03 -0.5967191 ESTs
AK021674 8.80E-03 -0.5841158 Homo sapiens cDNA FLJ11612 fis, clone
HEMBA1004011 2364492 AI741086 4.47E-03 -0.5804208 ESTs BF725007
8.40E-05 -0.5750824 ADRA2A .vertline. adrenergic, alpha-2A-, receptor
AL050391 3.94E-03 -0.5724837 Homo sapiens mRNA; cDNA DKFZp586A181
(from clone DKFZp586A181); partial cds AF367470 7.17E-03 -0.5715338
NYD-SP18 .vertline. testes development-related NYD-SP18 293605 AK026747
5.13E-03 -0.5606859 LOC54103 .vertline. hypothetical protein AK025732
6.04E-03 -0.5567897 ASAH .vertline. N-acylsphingosine amidohydrolase
(acid ceramidase) 1645681 AI026838 4.77E-03 -0.5536825 ESTs, Weakly
similar to NUCL_HUMAN NUCLEOLIN [H. sapiens] 1670862 AI081235 8.40E-03
-0.5523028 CD53 .vertline. CD53 antigen 3703127 BF433686 5.19E-03
-0.5433968 Homo sapiens cDNA FLJ32651 fis, clone SYNOV2001581 1837189
AF339781 4.94E-03 -0.5432118 GPR18 .vertline. G protein-coupled
receptor 18 4263201 BG236645 1.63E-03 -0.5426042 ESTs 4309471 BC009956
3.25E-03 -0.5390891 HLA-DPA1 .vertline. major histocompatibility
complex, class II, DP alpha 1 31047 R42463 2.11E-03 -0.5357526 ENTPD1
.vertline. ectonucleoside triphosphate diphosphohydrolase 1 L02785
6.69E-03 -0.5355299 SLC26A3 .vertline. solute carrier family 26,
member 3 NM_001337 2.41E-03 -0.5279489 CX3CR1 .vertline. chemokine
(C-X3-C) receptor 1 BC016758 2.54E-03 -0.5181311 HCLS1 .vertline.
hematopoietic cell-specific Lyn substrate 1 2214761 AI565489 5.26E-03
-0.5077388 PDE4A .vertline. phosphodiesterase 4A, cAMP-specific
(phosphodiesterase E2 dunce homolog, Drosophila) 2483676 BI492073
4.82E-03 -0.5068155 ITM2A .vertline. integral membrane protein 2A
128753 R16838 9.27E-03 -0.5059294 ESTs 4548935 BC014117 6.62E-03
-0.5027082 TBXAS1 .vertline. thromboxane A synthase 1 (platelet,
cytochrome P450, subfamily V) 1734062 AI191620 6.36E-03 -0.5006683
CDO1 .vertline. cysteine dioxygenase, type I NM_003820 8.71E-03
-0.5005279 TNFRSF14 .vertline. tumor necrosis factor receptor superfamily,
member 14 (herpesvirus entry mediator) AF305428 3.48E-03 -0.4933213
APOL1 .vertline. apolipoprotein L, 1 3163446 BC008734 6.56E-03 -0.4928143
FCGRT .vertline. Fc fragment of IgG, receptor, transporter, alpha
1964662 AJ420585 4.18E-04 -0.4903945 Homo sapiens mRNA full length
insert cDNA clone EUROIMAGE 1964662 1351991 AA807346 5.64E-03 -0.4893204
Homo sapiens cDNA FLJ14296 fis, clone PLACE1008455 NM_005211 8.18E-03
-0.485557 CSF1R .vertline. colony stimulating factor 1 receptor,
formerly McDonough feline sarcoma viral (v-fms) oncogene homolog
2161081 AI580271 9.09E-03 -0.4836746 AFP .vertline. alpha-fetoprotein
4862198 BC014456 7.46E-03 -0.4809022 CHRNA6 .vertline. cholinergic
receptor, nicotinic, alpha polypeptide 6 2728733 AW295170 8.18E-03
-0.4691647 ESTs 2163996 AI479461 5.00E-03 -0.4666083 CSR1 .vertline.
CSR1 protein 3086130 BF509235 4.52E-03 -0.4567332 KIAA1658 .vertline.
KIAA1658 protein 2364383 AI740671 2.10E-03 -0.4565665 Homo sapiens
cDNA FLJ32430 fis, clone SKMUS2001129, weakly similar to NAD-DEPENDENT
METHANOL DEHYDROGENASE (EC 184.108.40.206) 1272059 AA743283 6.97E-03
-0.4459972 GZMK .vertline. granzyme K (serine protease, granzyme
3; tryptase II) 3902651 BC016841 9.58E-03 -0.440234 RAB34 .vertline.
RAB34, member RAS oncogene family AB058708 8.44E-03 -0.4205688 KIAA1805
.vertline. KIAA1805 protein 40879 R56053 3.88E-03 -0.4205535 ME3
.vertline. malic enzyme 3, NADP(+)-dependent, mitochondrial 2222621
AI572605 7.12E-03 -0.3955994 HLA-DRA .vertline. major histocompatibility
complex, class II, DR alpha 2423726 AI860360 7.08E-03 -0.3889526
ESTs 2586524 AW080831 2.88E-03 -0.3552714 SEC14L2 .vertline. SEC14-like
2 (S. cerevisiae) 1056761 AA574174 8.03E-03 -0.3505373 CYP2A7 .vertline.
cytochrome P450, subfamily IIA (phenobarbital-inducible), polypeptide
7 NM_033380 8.90E-03 -0.3281259 COL4A5 .vertline. collagen, type
IV, alpha 5 (Alport syndrome) 2148123 AI467846 8.94E-03 -0.3273129
IAN4L1 .vertline. immune associated nucleotide 4 like 1 (mouse)
3026606 BE046325 6.30E-03 0.3235462 IGFBP5 .vertline. insulin-like
growth factor binding protein 5 3939513 BC013882 7.56E-03 0.3511008
EYA2 .vertline. eyes absent homolog 2 (Drosophila) AK057339 2.22E-03
0.359611 LOC81569 .vertline. actin like protein AF007194 9.59E-03
0.3752148 MUC3A .vertline. mucin 3A, intestinal AF288395 2.13E-03
0.3941496 C1orf15 .vertline. chromosome 1 open reading frame 15
3846346 BC017033 7.82E-03 0.4089227 SQLE .vertline. squalene epoxidase
3463613 BC003684 5.11E-03 0.42608 CXADR .vertline. coxsackie virus
and adenovirus receptor 2190016 AI538226 8.41E-04 0.4327835 GNG4
.vertline. guanine nucleotide binding protein 4 2138200 AI522215
8.62E-03 0.4352044 KIAA1804 .vertline. KIAA1804 protein 3087716
BF510979 7.81E-03 0.4409617 DHDH .vertline. dihydrodiol dehydrogenase
(dimeric) 5677199 BM129393 7.66E-03 0.4526335 GDF1 .vertline. growth
differentiation factor 1 2144913 AI452634 4.23E-03 0.4561948 GPR64
.vertline. G protein-coupled receptor 64 M95585 3.62E-03 0.5164108
HLF .vertline. hepatic leukemia factor 3932186 BC005345 6.63E-03
0.5470575 GTF2H2 .vertline. general transcription factor IIH, polypeptide
2 (44 kD subunit) 2968940 AW613854 7.48E-03 0.5487125 ESTs, Moderately
similar to S02826 nonhistone chromosomal protein HMG-1 [H. sapiens]
AF017790 5.02E-03 0.5639359 HEC .vertline. highly expressed in cancer,
rich in leucine heptad repeats AY049737 6.54E-03 0.5703928 NPM3
.vertline. nucleophosmin/nucleoplasmin, 3 U87791 8.91E-03 0.5755378
HBS1L .vertline. HBS1-like (S. cerevisiae) 3915484 BC017053 9.97E-03
0.5772075 ACOX3 .vertline. acyl-Coenzyme A oxidase 3, pristanoyl
AF100751 9.93E-03 0.5901307 LOC51661 .vertline. FK506-binding protein
AF206673 9.98E-03 0.6026123 BRF2 .vertline. BRF2, subunit of RNA
polymerase III transcription initiation factor, BRF1-like 1251833
AA731207 6.11E-03 0.6062738 FLJ10858 .vertline. hypothetical protein
FLJ10858 2507739 AI961369 2.77E-03 0.6190266 INSIG1 .vertline. insulin
induced gene 1 3504930 BC005141 5.35E-03 0.621928 GALK2 .vertline.
galactokinase 2 AL136570 6.98E-03 0.6225449 LHX6 .vertline. LIM
homeobox protein 6 2735278 AW450731 7.74E-03 0.6254488 FLJ14642
.vertline. hypothetical protein FLJ14642 AK025820 7.09E-03 0.6282847
FLJ22167 .vertline. hypothetical protein
FLJ22167 2975886 AW629176 9.47E-04 0.6387345 ESTs, Weakly similar
to I38022 hypothetical protein [H. sapiens] AF116670 4.34E-03 0.6509022
NP .vertline. nucleoside phosphorylase 1630968 AI018605 3.48E-03
0.6689059 ESTs 3996449 BC015107 7.48E-03 0.6910834 FLJ13433 .vertline.
hypothetical protein FLJ13433 AB049113 9.76E-03 0.7010916 DUT .vertline.
dUTP pyrophosphatase AK025543 8.56E-03 0.707967 KIAA1345 .vertline.
KIAA1345 protein AF053306 6.31E-03 0.7227021 BUB1B .vertline. BUB1
budding uninhibited by benzimidazoles 1 homolog beta (yeast) 3010092
BC008954 9.78E-03 0.7272841 SLC29A1 .vertline. solute carrier family
29 (nucleoside transporters), member 1 NM_001685 6.48E-03 0.7348505
ATP5J .vertline. ATP synthase, H+ transporting, mitochondrial F0
complex, subunit F6 4838878 BC016751 5.11E-03 0.7473835 PCDHB3 .vertline.
protocadherin beta 3 2175265 AI524385 2.93E-03 0.7701449 ANLN .vertline.
anillin, actin binding protein (scraps homolog, Drosophila) 3010727
BE206076 1.25E-04 0.7737357 ALK .vertline. anaplastic lymphoma kinase
(Ki-1) X62534 4.09E-03 0.7848284 HMG2 .vertline. high-mobility group
(nonhistone chromosomal) protein 2 NM_018669 3.59E-03 0.8006609
WDR4 .vertline. WD repeat domain 4 AB035898 3.72E-03 0.8063743 KNSL7
.vertline. kinesin-like 7 NM_006734 8.82E-03 0.8275957 HIVEP2 .vertline.
human immunodeficiency virus type I enhancer binding protein 2 AF399910
1.43E-03 0.8370408 DEEPEST .vertline. mitotic spindle coiled-coil
related protein AF331796 2.69E-03 0.8636081 HCAP-G .vertline. chromosome
condensation protein G AF073518 6.93E-03 0.8902775 SERF1A .vertline.
small EDRK-rich factor 1A (telomeric) 1870184 AI245807 1.20E-03
0.8964932 MGC14798 .vertline. similar to RIKEN cDNA 5730421E18 gene
4509200 BC012919 1.19E-03 0.9209921 KLF7 .vertline. Kruppel-like
factor 7 (ubiquitous) 3926227 BC009855 1.51E-03 0.9417395 FLJ14909
.vertline. hypothetical protein FLJ14909 1337864 AA811376 2.04E-03
1.0013265 FLJ10545 .vertline. hypothetical protein FLJ10545 4109322
BC016782 2.22E-03 1.0106451 KIAA0101 .vertline. KIAA0101 gene product
AF334184 1.66E-04 1.0332632 FKSG42 .vertline. FKSG42 AL136794 1.98E-04
1.1631056 RACGAP1 .vertline. Rac GTPase activating protein 1
4TABLE 4 Top 27 genes gene p coef desc BE206076 7.27446E-05 1.2270555
ALK .vertline. anaplastic lymphoma kinase (Ki-1) AF040628 0.000116979
0.9344455 ED1 .vertline. ectodermal dysplasia 1, anhidrotic BF725007
0.000331932 -0.5801744 ADRA2A .vertline. adrenergic, alpha-2A-,
receptor AF367473 0.00068377 -0.8560356 NYD-SP21 .vertline. testes
development-related NYD-SP21 AI245807 0.000789859 1.0889127 MGC14798
.vertline. similar to RIKEN cDNA 5730421E18 gene AI215902 0.00086535
-0.6485096 ESTs, Highly similar to T50835 hypothetical protein [H.
sapiens] AF334184 0.000883849 1.0194484 FKSG42 .vertline. FKSG42
AW137991 0.000969904 -1.1497247 RELB .vertline. v-rel reticuloendotheliosis
viral oncogene homolog B, nuclear factor of kappa light polypeptide
gene enhancer in B-cells 3 (avian) AA648777 0.001013976 -1.6577293
MS4A7 .vertline. membrane-spanning 4-domains, subfamily A, member
7 BC017053 0.001247656 0.8780177 ACOX3 .vertline. acyl-Coenzyme
A oxidase 3, pristanoyl AL136570 0.00133504 0.822904 LHX6 .vertline.
LIM homeobox protein 6 AL136794 0.001396755 1.1506902 RACGAP1 .vertline.
Rac GTPase activating protein 1 AF331796 0.001490968 1.2361318 HCAP-G
.vertline. chromosome condensation protein G BC005850 0.001567107
-1.5082099 CBFA2T1 .vertline. core-binding factor, runt domain,
alpha subunit 2; translocated to, 1; cyclin D-related AF399910 0.00160607
0.8764045 DEEPEST .vertline. mitotic spindle coiled-coil related
protein AK057339 0.001660726 0.4461839 LOC81569 .vertline. actin
like protein NM_003265 0.001664793 -0.7178243 TLR3 .vertline. toll-like
receptor 3 AK026343 0.001697437 -0.5688145 FLJ22690 .vertline. hypothetical
protein FLJ22690 BC018538 0.001728603 -0.7262933 ALOX5AP .vertline.
arachidonate 5-lipoxygenase-activating protein AI806109 0.001789313
-1.0762434 KIAA1580 .vertline. KIAA1580 protein AL122052 0.001810196
-0.9558644 KIAA0793 .vertline. KIAA0793 gene product BC012919 0.002008228
0.9606709 KLF7 .vertline. Kruppel-like factor 7 (ubiquitous) BC008392
0.002082612 -1.0314954 UCP3 .vertline. uncoupling protein 3 (mitochondrial,
proton carrier) BF432813 0.002110755 -1.2468785 KLRB1 .vertline.
killer cell lectin-like receptor subfamily B, member 1 AI741086
0.002281532 -0.675948 ESTs AK022729 0.002327927 -0.9643334 KIAA1681
.vertline. KIAA1681 protein NM_006006 0.002390747 -1.0628839 ZNF145
.vertline. zinc finger protein 145 (Kruppel-like, expressed in promyelocytic
5 TABLE 5 GenBank van't Veer et al. UniGene AA648777 AF201951 Hs.11090
AF367473 AL137391 Hs.28514 AK022729 Contig30485_RC Hs.42656 AI741086
Contig39054_RC Hs.115122 AK022729 Contig47136_RC Hs.42656 AI215902
Contig52342_RC Hs.88845 BF725007 Contig53357_RC Hs.249159 BF725007
NM_000681 Hs.249159 AF040628 NM_001399 Hs.105407 BC018538 NM_001629
HS.100194 BF432813 NM_002258 Hs.169824 NM_003265 NM_003265 Hs.29499
BC008392 NM_003356 Hs.101337 BC017053 NM_003501 HS.12773 BC012919
NM_003709 Hs.21599 BE206076 NM_004304 Hs.278572 BC005850 NM_004349
Hs.31551 NM_006006 NM_006006 Hs.37096 AF399910 NM_006461 Hs.16244
AW137991 NM_006509 Hs.858 AL136794 NM_013277 Hs.23900 AL136570 NM_014368
Hs.103137 AL122052 NM_014808 Hs.301283 AK057339 U20582 Hs.2149
Individual Genes that are Expressed at Higher than Median Levels
and Correlated with the Recurrence of Breast Cancer
 DEEPEST, RACGAP1, ZNF145 and MS4A7 were found to each be
significantly associated with tumor recurrence. In both the datasets
used in FIG. 2, patients were divided into high and low expression
groups relative to the overall median for each gene across all patients,
and their survival curves were compared (see FIG. 3, which shows
Kaplan-Meier disease-free survival curves). The first six graphs
in FIG. 3 display the results using the dataset from the 124 patients
of Example 1; the X-axis is in months. The second six graphs in
FIG. 3 display the results using the dataset from van't Veer et
al. with the X-axis in years. The Y-axis for both are "survival
probability" as described above. As control, MKI67 and CCNE1,
two genes known to be associated with aggressive cancers were analyzed
in the same manner.