|Home | About | Journals | Submit | Contact Us | Français|
This review summarizes advances in our understanding of the genomic and epigenomic abnormalities in breast cancers that are being revealed by the increasingly powerful suite of genomic analysis technologies. It summarizes the remarkable genomic heterogeneity that characterizes the disease, describes mechanisms that shape cancer genomes as they evolve toward metastasis, summarizes important recurrent aberrations that exist in spite of the genomic chaos and that contribute to breast cancer pathophysiology, and describes the use of preclinical models to identify drugs that will be effective against subsets of breast cancers carrying specific genomic and epigenomic abnormalities.
Cancer is now recognized as disease in which abnormalities in the genome and epigenome accumulate as a result of exposure to endogenous and exogenous damaging agents thereby enabling cells to escape normal regulatory controls. The set of accumulated abnormalities in the tumor and in the microenvironment in which the tumors reside determine the course of the disease including propensity to invade and metastasize and to respond to treatment. A central goal of modern genomic projects like The Cancer Genome Atlas (TCGA; http://cancergenome.nih.gov/) Project  and the International Cancer Genome Consortium (ICGC; http://www.icgc.org/) is to determine the spectrum of genomic aberrations that make up important human cancer types and to understand how these aberrations contribute to the pathophysiology of these cancers. The technologies being employed by these and related projects are providing an increasingly detailed view of breast cancer genomes. Today’s comparative genomic hybridization (CGH) platforms provide genome-wide maps of genome copy number abnormalities with subgene resolution . In some cases the measurements are allele specific. Fluorescence in situ hybridization (FISH) provides information about copy number and structural heterogeneity including variation between cells and in histological context . Moreover, probes are available to interrogate essentially any part of the genome (http://bacpac.chori.org/). Genome sequencing is now providing detailed information about mutations and structural aberrations. Massively parallel sequencing platforms [4•] have now made practical, analysis of hundreds to thousands of genes in individual tumor genomes and comprehensive analyses of entire breast cancer genomes are beginning to emerge [5••]. Analysis of promoter methylation and of DNA sequences immunoprecipitated with antibodies against specific chromatin marks provides information on epigenomic events that regulate gene expression [6,7].
The picture of breast cancer genomes that emerges from large-scale studies is one of remarkable complexity and variability. Individual tumors often carry aberrations that deregulate hundreds or even thousands of genes. Genome copy number changes and epigenomic modifications seem to deregulate the largest number of genes. Figure 1 shows genome copy number abnormalities for three primary breast tumors. These show ‘strongly selected’ events such as high-level gene amplification and homozygous deletion as well as abnormalities that change genome copy number by only one or a few copies. In general, the strongly selected regions are narrow in genomic extent and involved only a few genes while low level genome copy number abnormalities (CNAs) that span larger regions of the genome up to and including whole chromosome losses or gains may affect thousands of genes. These data also show the remarkable differences in copy number abnormality structure that occur between clinically similar tumors. Although CGH profiles of breast cancers appear complicated, analyses of cancer genomes using DNA sequence-based methods such as end sequence profiling  or paired end sequencing [9••] show that the genome copy number abnormality maps yield an overly simple view of the events that conspire to change copy number. Figure 2, for example, shows results from a DNA sequence analysis of a ~150 kbp sequence from a region of amplification nominally from chromosome 20 . Remarkably, sequences from several normally separate parts of the genome have been ‘stitched together’ in the cancer genome and amplified together as a cassette — perhaps as a result of fusions that occur during a bridge breakage fusion process or during extrachromosomal fusion and amplification. Genome sequencing efforts are revealing large numbers of structural rearrangements in breast cancers. So far, however, important recurrent structural aberrations like the Ets-family fusions found in prostate cancer [10•] have not been discovered in breast cancer. Breast cancer genomes are further deregulated through mutational processes. In general, the number of genes deregulated via mutations is considerably less than through other mechanisms — typically tens of mutations per tumor .
Epigenomic events including promoter methylation, histone acetylation, methylation, and phosphorylation and chromatin remodeling further contribute to cancer pathophysiology [7,12]. Genome-wide analyses of promoter regions show that promoter CpG island hypermethylation is a mechanism by which tumor suppressor genes are inactivated [13,14] and is an especially important early event in breast cancer development [15–17]. Indeed, this appears to be a more common mechanism of tumor suppressor gene inactivation than mutation or homozygous deletion. The importance of epigenomic events as early events in breast cancer is noteworthy considering that the epigenome is subject to environmental-related and treatment-related modification [18•] and may even reflect environmental influences that are passed between generations .
Although the genomes of clinically similar breast cancers are remarkably different, some regions of the genome are recurrently aberrant. Figure 3, for example, shows the locations of significant genome copy number gains and losses across the genome measured using array CGH . The number of genes mapping to regions of significant abnormality is remarkable. In fact, correlative analyses of genome copy number and gene expression indicate that 10–15% of the entire genome is deregulated by recurrent genome copy number abnormalities [20,21]. Analysis of CGH data suggests the existence of at least three general classes — those with ‘simple’ genomes displaying relatively few genome aberrations, those with ‘complex’ genomes in displaying many aberrations and those displaying high-level amplification [20,22]. It is likely that these subtypes may be caused by differences in DNA repair defects. For example, complex genomes typically carry p53 mutations while simple genomes do not. CGH profiles for tumors from BRCA1 carriers also display high genome complexity although the spectrum of recurrent abnormalities differs between sporadic and heritable breast cancers. In particular, tumors from BRCA1 mutation carriers typically display recurrent aberrations involving chromosomes 3 and 5 that sporadic tumors do not have . Relatively few recurrent mutations have been found in breast cancers. High frequency somatic mutations (i.e. present in >3% of reported cases) include PIK3CA, TP53, CDH1, CDKN2A, and AKT1 (http://www.sanger.ac.uk/genetics/CGP/cosmic). However, almost two hundred of others have been reported as recurrently mutated but at lower frequency . Prevalent germ line mutations that contribute to breast cancer genesis involve BRCA1, BRCA2, CHEK1, and TP53.
The mechanisms by which low level copy number abnormalities contribute to cancer pathophysiology have not yet been determined definitively. However, we have suggested that these deregulate transcription of large numbers of genes that collectively increase metabolic activity . The contributions of regions of amplification at 8p11–12, 8q24, 11q13, 12q13, 17q11–12, 17q21–24 and 20q13 as somewhat better understood since each encodes a known oncogene (e.g. FGFR1, MYC, CCND1, MDM2, ERBB2, PS6K and ZNF217, respectively). However, substantial evidence is now emerging that demonstrates that multiple genes in each region of amplification contribute to the pathophysiology of the disease. Functionally important co-amplified genes demonstrated so far include LSM1, BAG4, and C8orf4 at 8p11–12 ; MYC and PVT1 at 8q24 ; CCND1 and EMSY at 11q13 ; and ERBB2 and GRB7 at 17q21 . Recent studies also suggest that amplifications of regions that are well separated in the normal genome are not independent events. For example, coamplification of regions at 8p11–8p12 and 11q12–11q14  or 8q24 and 17q21  have been reported to contribute collectively to breast cancer pathophysiology. These interactions may be just the tip of the iceberg since dozens of genes have been identified as over expressed as a result of recurrent, high-level amplification and most have not been functionally assessed . Coamplification of genes that are normally separate along the genome is plausible considering the fact that genes normally separate in the genome may be joined and contiguous in cancer genomes as a result of structural rearrangements . However, this remains to be investigated. Recurrent mutations, although relatively infrequent in breast cancer, provide particularly strong evidence of functional importance. Bringing all of these observations together, recent integrative analyses of homozygous deletions, high-level amplification and recurrent mutations suggest that genomic deregulation of pathways involving ERBB2, EGFR, and PI3K signaling and DNA topology is particularly important in breast cancer genesis and progression .
Epigenomic alterations also contribute to breast cancer pathophysiology [19,31]. Targeted and genome wide analyses of promoter region methylation in breast cancer have shown that genes with known or suspected tumor suppressor function including BRCA1, CCND2, CDKN2A, RARβ, and RASSF1A [32,33] and members of the WNT signaling pathway  are recurrently down-regulated by hypermethylation. Consistent with this, genes found to be recurrently mutated in large-scale genome analysis studies are frequently targets of hypermethylation and hypermethylation is usually mutually exclusive from genetic changes . At the chromatin level, modification via acetylation and/or phosphorylation directly modifies estrogen receptor alpha and other steroid hormone receptors superfamily in breast cancer thereby modifying ligand sensitivity and hormone antagonist responses . Chromatin remodeling also has been implicated in growth and metastasis. In particular, SATB1 has been reported to be a genome organizer that tethers multiple genomic loci and recruits chromatin-remodeling enzymes to regulate chromatin structure and gene expression. Overexpression of this ‘master chromatin regulator’ appears to drive breast cancer progression by upregulating metastasis-associated genes and down-regulating tumor-suppressor genes .
Breast cancers accumulate a remarkable number of genomic aberrations as they develop the cancer hallmarks that eventually enable metastatic spread . The wide variation in number and type of abnormalities that occurs between clinically similar tumors is understandable considering the evolutionary forces that each tumor experiences. It is useful to consider the shaping of cancer genomes as the result of overcoming checks and balances that have evolved over time to protect against cancer . Several possible barriers are illustrated schematically in Figure 4 and include DNA repair, differentiation, senescence, immune surveillance, limited angiogenesis, apoptotic surveillance, cell cycle regulation, and anticancer treatments. These are depicted separately but of course they interact considerably — for example, DNA repair is tightly linked to cell cycle regulation. Nonetheless, considering them as separate barriers is a convenient way of beginning to parse studies of cancer genome evolution. Since the ‘efficiency’ of each of these barriers is likely influenced by the individual genotype and by the microenvironment of the evolving cancer it is easy to see how individual tumors evolve such different genomic and epigenomic features.
Recent mouse genetic experiments comparing cancer incidence in crosses between cancer prone and cancer resistant mouse strains are revealing some of the important barriers to cancer progression that directly influence individual genotypes. Such experiments already have identified transcriptional networks including mitotic checkpoints, inflammation, apoptosis, and epidermal barrier function whose activity likely determines barrier efficiency [39••]. Genetically engineered mouse models of cancer also clearly demonstrate that the genetic background significantly influences the genome aberration spectrum of the resulting tumors . Combined genetic and genomic studies of mouse cancer models and human tumors show that tumors selectively amplify cancer susceptibility alleles thereby establishing a direct link between tumor genome aberration spectrum and individual genotype . Finally, it is telling that the recurrent genomic aberrations that arise in tumors in BRCA1 carriers are significantly different than those that arise in noncarriers. In short, the individual genetic background plays a major role in determining cancer susceptibility and in shaping cancer genomes .
A wealth of evidence now indicates that the extracellular matrix composition and organization strongly influences breast tumorigenesis [43,44] and that the extracellular environment and the tumor genome coevolve during progression. Key microenvironmental factors that contribute to cancer progression include microenvironmental remodeling via matrix metalloproteinases [45–47], macrophage-mediated metastasis , senescence-associated cytokine production , changes in tissue architecture , estrogen-related stromal–epithelial cell crosstalk , and TGF-beta signaling . Most evidence suggests that the microenvironment coevolves with breast tumors via epigenomic modifications  although some evidence supports the possibility that clonal genomic aberrations in breast cancer-associated stromal cells also may play a role [53•,54]. The latter possibility remains controversial .
Breast cancers progress through multiple genomic and epigenomic steps. However, the evolutionary process is unlikely to be the result of a linear and more or less constant rate of successive genomic and epigenomic aberration accumulation. It seems more likely that cancer progression varies between individuals and over time within an individual. The breast progenitor cell in which the tumor arises likely determines the spectrum of aberrations needed to enable progression. Evidence for this comes from studies showing that the spectrum of genomic aberrations accumulated is strongly influenced by the normal progenitor cell type in which the tumor arises [20,21]. In general, the events associated with early aspects of breast cancer appear to be epigenomic in nature, wherein stepwise DNA methylation changes enable escape of telomerase negative epithelial cells from proliferation barriers [56••]. This leads to proliferation in the absence of telomerase and culminates in a period of high genome instability owing to checkpoint deregulation [57,58] and/or entry into telomere crisis when telomeres become critically short [3,59•]. This barrier is highly effective but not perfectly so. As a consequence, most cells become genomically unstable and die. The increase in genome instability during breast cancer progression is illustrated in Figure 5. Rarely, a single cell accumulates genomic or epigenomic alterations that reactivate telomerase and confer a proliferative advantage. This cell might be considered the tumor initiation cell and will have multiple characteristics that appear ‘stem cell like’. The extent to which this is related to normal stem cells remains unclear. However, it is likely that the genomic characteristics of this cell — both transcriptional and genomic — will be reflected in subsequent progeny. This may explain why tumor genomes appear to evolve relatively slowly after telomere crisis and why metastases that develop years after immortalization usually retain the genomic characteristics of the primary tumor from which they were derived . That said, not all breast tumors progress this way. A recent analysis of tumor progression termed Sector-Ploidy-Profiling (SPP) demonstrates that breast tumor evolution may evolve as a single major clonal subpopulation or as multiple clonal subpopulations . The latter model may explain why a small percentage of metastatic cancers do not resemble the primary tumor from which they were derived.
Figure 5 also shows that the cells that survive telomere crisis remain genomic unstable. Thus, tumors are likely to be consist of a large number of cells that are not faithful genomic representations of the tumor-initiating cell and are likely to be biologically compromised. These cells are likely to be much more sensitive to treatment than the tumor initiating cells from which they were derived. This may partially explain why breast tumors initially respond well to treatment but fail to exhibit a durable response.
Molecular profiling technology promises to have a profound effect on the classification of breast cancer. Pioneering molecular profiling studies on invasive ductal carcinoma of the breast have recognized the presence of distinct breast cancer classes, based on expression patterns of the ‘intrinsic gene set’, a large panel of genes that shows heterogenous expression across entire tumor cohorts [62–64]. The tumors within these distinct clusters are defined by expression patterns that are similar to cells within the normal breast architecture (i.e. basal versus luminal clusters), or have known genomic aberrations (ERBB2/HER2 positive), or most closely resemble normal breast (‘normal-like’ breast cancer). Some of these clusters can be further subdivided on the basis of additional expression features (e.g. luminal A and luminal B). Patients within the distinct subclasses have significantly different clinical outcomes; basal and ERBB2/ HER2 patients have a poorer prognosis than their luminal A and normal subclass counterparts [63,64]. The most common histopathologic types of breast cancer, namely invasive ductal carcinoma and invasive lobular carcinoma, can be classified as any of these five subtypes . In contrast, most rare histologic types of breast cancer belong exclusively to single molecular subclasses, such as medullary and metaplastic (basal) and mucinous and neuroendocrine (lobular) . The classification of tumors into basal, lobular, and HER2 subgroups reflects the molecular characteristics of the disease, an advantage over those studies that simply develop predictive gene sets for outcome prediction. Since these initial studies, larger cohorts of tumors have been profiled on more comprehensive array platforms, leading to the refinement of the classification and the identification of specific genes that are definitive of each of the classes.
In keeping with their relatively good prognosis, the luminal class of tumors tends to be ER positive and express genes that are consistent with this hormone receptor status (ESR1, PGR, and GATA3) [63,66]. At the genomic level, luminal tumors have gain of 1q and 16p and loss of 16q and frequent high-level amplification of 8p11–12, 11q13–14, 12q13–14, 17q11–12, 17q21–24, and 20q13 (luminal A), while luminal B is characterized by gains of chromosomes 1q, 8q, 17q, and 20q and losses involving portions of 1p, 8p, 13q, 16q, 17p, and 22q, with frequent high-level amplification of 8p11–12, two regions of 8q, and 11q13–14 [20,21].
Similarly, the HER2 tumors are defined by the high expression of ERBB2 and GRB7, both of which lie within the 17q amplicon that is almost universally amplified in this type of breast cancer. ERBB2 tumors are relatively enriched for increased copy number at 1q, 7p, 8q, 16p, and 20q and reduced copy number at 1p, 8p, 13q, and 18q, with high levels of amplification of 17q as expected [20,21].
Basal-like tumors are ER negative, and are defined by the expression of genes associated with both basal epithelium and myoepithelium (e.g. CK5 and CK17) [62–64]. Basal tumors are relatively enriched for gains involving 3q, 8q, and 10p and losses involving 3p, 4p, 4q, 5q, 12q, 13q, 14q, and 15q [20,21]. Interestingly, high-level amplification at any locus is infrequent in these tumors. Basal tumors are aggressive, highly proliferative, and usually present as high-grade tumors, but the basis of this aggressive phenotype is not fully understood. Furthermore, it has become clear that basal tumors share many characteristics with breast tumors derived from patients with BRCA1 mutations [67,68]. In keeping with this observation, basal tumors often lose expression of BRCA1 through mechanisms such as promoter hypermethylation. Finally, many but not all of the basal tumors are exemplified by the lack of expression of hormone receptors (ER and PR) and HER2, forming the ‘triple negative’ type of breast cancer [67,69–71]. Although once thought to be exclusively basal in origin, ‘triple negative’ breast tumors are now recognized to include a subset of normal-like breast tumors [67,69]. Triple negative tumors are more common in African-American than Caucasian women, and are associated with shorter duration survival . Recent studies suggest that subsets of aggressive tumors are also defined by stem cell like (CD44 high, CD24 low) and claudin low genotypes as well that may overlap but be distinct from traditional basal subtypes .
Given the heterogeneity of expression and genomic composition within the distinct subclasses, it might also be expected that there will be distinct clinical responses to chemotherapeutic and targeted agents within the different subclasses. If agents can be identified that show enhanced efficacy against specific subclasses, this might represent the first step toward more personalized medicine for the treatment of breast cancer. Such class distinctions have already been identified, in particular for HER2 positive tumors. Targeted therapeutic agents such as the monoclonal antibody trastuzumab or the small molecule tyrosine kinase inhibitor lapatinib have both been approved for use in treating patients with HER2 positive breast cancer, leading to improved survival in these patients [73–75]. Given that HER2-directed therapies were developed before the identification of breast cancer subclasses, it is now of interest to see if additional subclass specific agents can be identified. Recent studies have demonstrated that inhibitors of poly (ADP-ribose) polymerase (PARP), an important transducer of BRCA-mediated DNA damage response, selectively target BRCA1 and BRCA2 mutant breast cancer . This is a form of synthetic lethality, since only cells with aberrant BRCA function will be affected. Indeed, the side effects in patients treated with PARP inhibitors have been much milder than most traditional chemotherapeutic agents. However, when patients with BRCA mutant tumors have been treated with PARP inhibitors, the result is prolonged patient survival . Given the similarity between BRCA1 mutant and the basal subclass of tumors, it is hoped that basal tumors will also demonstrate increased sensitivity to PARP inhibitors. One caveat to this approach has been the observation that BRCA may be dispensable for the growth of established breast cancers. This is evidenced by the findings that BRCA2 mutant tumors develop secondary resistance to DNA damaging agents through mutations that restore normal BRCA function and DNA damage response [77,78]. With the restoration of the DNA damage response, these tumors are less sensitive to the actions of DNA damaging agents such as cisplatin and carboplatin. Basal subtype tumors have also been shown to be exquisitely sensitive to dose dense regimens of epirubicin–cyclophosphamide combinations . Interestingly, of the 10 basal type tumors within the cohort, 9 of them showed complete response. Thus, these results demonstrate that both targeted and conventional chemotherapies are likely to demonstrate subclass specific efficacy.
The successes of genome marker targeted therapies described above have motivated efforts to identify other response-genome aberration associations that might guide the development of additional breast cancer treatments. This is complicated by the complexity of cancer genomes and the large and growing numbers of candidate anticancer agents that should be considered for use in the treatment. In fact, a 2009 survey by the pharmaceutical industry indicates that more than 800 experimental therapies are now being evaluated clinically including 106 in breast cancer (http://www.phrma.org/files/09-046PhRMACancer09_0331.pdf). Identification of genomic and epigenomic features that predict response to these agents is challenging in the context of clinical trials because of the difficulty of obtaining tumor samples during trials, the cost of comprehensive molecular profiling and the relatively large size of clinical trials needed to establish solid predictive markers. An alternative is to use preclinical models to identify associations between genome aberrations and response. The challenge in this approach is to develop models that reflect the range of genome and epigenome aberrations that influence response. Collections of breast cancer cell lines grown in vitro and as xenografts are now being used for this purpose as are collections of genetically engineered mouse models.
The use of collections of cell lines to identify molecular features associated with therapeutic response began with an effort by the NCI Developmental Therapeutics Program (http://dtp.nci.nih.gov/index.html) to assess responses to potential anticancer agents in a collection of 60 diverse cancer cell lines (the NCI-60). Over the past several decades, the concentrations of drug needed to inhibit growth by a predetermined amount were determined for more than 40 000 agents. Recently, the cell lines have been characterized in substantial molecular depth  so that the cell responses can be compared with molecular features to identify response-associated features that might be developed into predictive molecular markers that can be tested clinically. Several studies have illustrated the promise of this approach [80–82].
One problem with the NCI-60 cell line panel for the assessment of specific responses in breast cancer is the lack of representation of several of the major breast cancer subtypes in the cell line collection. To overcome that limitation, we and others have focused on testing drugs intended for breast cancer treatment in large panels of breast cancer cell lines selected to capture the molecular features that determines clinical responses. Analysis of genome copy number and mutations in commonly available breast cancer cell lines showed that the recurrent aberrations in the cell lines were similar to those found in collections of primary breast tumors [83,84]. However, analysis of the transcriptional features showed that the collection represented only three of the transcriptional subtypes seen clinically. These were designated basal B (now redefined as claudin low ) and basal A (now redefined as basal) and a single luminal subtype [83,84].
The relative ease of analysis in cell line panels allows efficient identification of candidate molecular predictors of response to new agents and combinations thereof that warrant clinical evaluation. This approach is supported by studies of breast cancer cell line panels that clearly show the association between ERBB2 amplification and response to trastuzumab  and lapatinib  that has been observed clinically. Other predictive markers suggested by cell line panel studies include:
Of course, cell line collections are necessarily imperfect models since they model only aspects of response that are determined by factors that are intrinsic to the tumors from which they were derived. Clearly, they cannot accurately identify predictors of response to agents that inhibit extrinsic cancer hallmarks such as angiogenesis, immune surveillance or metastatic spread. Their utility also is affected by the failure to accurately model interactions with the microenvironment . Culturing cells in different microenvironments can overcome this limitation to some extent but that can never be accomplished perfectly. Indeed, the microenvironments in which tumor cells reside vary within individuals — from in situ disease to diverse metastatic sites. Thus, drugs that are effective in a broad range of molecular environments and molecular markers that predict response independent of microenvironment may be most likely to predict durable clinical response. Studies of the role of the microenvironment on drug response may be facilitated by the recent development of microenvironment microarrays that allow the effects of growing cells in many different microenvironments to be assessed [93•].
Genetically engineered mouse (GEM) tumor models are important preclinical adjuncts to cell line models. Historically, the utility of GEM models in predictive marker development has been limited because they did not model the genetic diversity found in human tumors. However, this is changing thanks to the substantial increase in GEM tumor model diversity because of work in the NCI Mouse Models of Human Cancer Consortium (MMHCC) and throughout the mouse model community (http://cancermodels.nci.nih.gov/). A recent comparison of transcriptional characteristics of 13 GEM mouse models with human tumors showed that ‘many of the defining characteristics of human subtypes were conserved among the mouse models’ . Several interesting mammary models have been reported since that study. A first example is a GEM with MMTV driven HER2 overexpression combined with cre-recombinase mediated PTEN deletion  to model the luminal type of HER2 positive breast cancer. Tumors that arise in these mice are multifocal and metastatic, with hyperactivation of the PI3K pathway. Given that PTEN deficiency in HER2 positive tumors is associated with trastuzumab resistance, these mice are likely to represent a valuable resource to identify therapeutic combinations that are likely to overcome this resistance. A second example is a GEM with mammary specific expression of mutant Kras to model basal-like breast cancer. These mice are characterized by the overexpression of Igf1r . Treatment of tumor bearing mice with an IGF1R inhibitor has been shown to result in shrinkage of the tumor mass. Xenografts of the KRAS mutant cell line MDAMB231 into mice has shown that MDAMB231 is also highly sensitive to the IGF1R inhibitor, further reinforcing the basal-like phenotype of this mouse model.
The power of modern genome analysis tools — especially massively parallel sequencing — is increasing at a remarkable rate. It seems clear that the ‘thousand dollar genome’ is nearly at hand. Thus it seems reasonable to expect that comprehensive tumor scans of genomic and epigenomic aberrations will become an important and routine part of cancer management. The challenges will be to manage the information so that it is generally accessible to the scientific and medical communities and to interpret it in ways that lead to improved cancer management. This will require a substantial investment in functional studies since the roles of most genomic and epigenomic aberrations in cancer pathophysiology are not understood. It will also require full development and deployment of large-scale information management and interpretation systems. Success in these areas will allow the potential of cancer genome and epigenome in treatment personalization to be fully realized.
This work was supported by the Director, Office of Science, Office of Biological & Environmental Research, of the U.S. Department of Energy under contract No. DE-AC02-05CH11231, by the National Institutes of Health, National Cancer Institute grants U54 112970, and P50 CA 58207.
Papers of particular interest, published within the period of review, have been highlighted as:
• of special interest
• • of outstanding interest