BANERJI, Shantanu (203 Grenfell Boulevard, Winnipeg Manitoba, R3P 0B8, 0B8, CA)
GETZ, Gad (70 Hoitt Road, Belmont, MA, 02478, US)
CIBULSKIS, Kristian (40 Burnside Ave, Somerville, MA, 02144, US)
LAWRENCE, Michael (11 Norris Street, Cambridge, MA, 02140, US)
MIRANDA, Alfredo, Hidalgo (Periferico Sur 4809, Arenal TepepanDelegacion Tlalpa, Mexico D.F., 14610, MX)
ESCARENO, Claudia, Rangel (Periferico Sur 4809, Arenal TepepanDelegacion Tlalpa, Mexico D.F., 14610, MX)
TOKER, Alex (68 Jamaica Street, Apt. 3Boston, MA, 02130, US)
BROWN, Kristin, K. (120 Peterborough Street, Apt. 9Boston, MA, 02215, US)
BETH ISRAEL DEACONESS MEDICAL CENTER (Technology Deaconess Medical Center, Technology Venture Office330 Brookline Avenue, Mail Stop FN-, Boston MA, 02215, US)
DANA-FARBER CANCER INSTITUTE, INC. (450 Brookline Avenue, Boston, MA, 02215, US)
NATIONAL INSTITUTE OF GENOMIC MEDICINE, MEXICO (Periferico Sur 4809, Arenal TepepanDelegacion Tlalpa, Mexico D.F, 14610, MX)
MEYERSON, Matthew (307 Independence Road, Concord, MA, 01742, US)
BANERJI, Shantanu (203 Grenfell Boulevard, Winnipeg Manitoba, R3P 0B8, 0B8, CA)
GETZ, Gad (70 Hoitt Road, Belmont, MA, 02478, US)
CIBULSKIS, Kristian (40 Burnside Ave, Somerville, MA, 02144, US)
LAWRENCE, Michael (11 Norris Street, Cambridge, MA, 02140, US)
MIRANDA, Alfredo, Hidalgo (Periferico Sur 4809, Arenal TepepanDelegacion Tlalpa, Mexico D.F., 14610, MX)
ESCARENO, Claudia, Rangel (Periferico Sur 4809, Arenal TepepanDelegacion Tlalpa, Mexico D.F., 14610, MX)
TOKER, Alex (68 Jamaica Street, Apt. 3Boston, MA, 02130, US)
BROWN, Kristin, K. (120 Peterborough Street, Apt. 9Boston, MA, 02215, US)
1. A method of treating a subject having cancer comprising identifying a subject having a tumor expressing a MAGI3-AKT3 fusion gene and administering to said subject an AKT inhibitor
2. The method of claim 1, wherein said cancer is an epithelial cell cancer.
3. The method of claim 2, wherein the epithelial cell cancer is breast cancer.
4. The method of claim 3, wherein said breast cancer is triple negative breast cancer.
5. The method of claim 1, wherein the AKT inhibitor is an ATP-competitive AKT inhibitor.
6. The method of claim 5, wherein the ATP-competitive AKT inhibitor is GSK690693, A-
443654, CCT-128930, GSK-2141795, AZD-5363, GDC-0068, A-674565 or AT7867.
7. A method of decreasing tumor growth or inducing tumor cell apoptosis wherein said tumor expresses a MAGI3-AKT3 fusion gene comprising contacting said tumor with an AKT inhibitor.
8. The method of claim 7, wherein said tumor is an epithelial cell tumor.
9. The method of claim 8, wherein the epithelial cell tumor is breast cancer.
10. The method of claim 9, wherein said breast cancer is triple negative breast cancer.
11. The method of claim 7, wherein the AKT inhibitor is an ATP-competitive AKT inhibitor.
12. The method of claim 11, wherein the ATP-competitive AKT inhibitor is GSK690693, A-
443654, CCT-128930, GSK-2141795, AZD-5363, GDC-0068, A-674565 or AT7867.
13. A method of determining responsiveness of a subject having cancer to an AKT inhibitor comprising detecting the presence or absence of a MAGI3-AKT3 fusion gene in a cancer cell obtained from said subject.
14. The method of claim 13, wherein the presence of said fusion gene indicates that the subject will be responsive to an ATP-competitive AKT inhibitor.
15. The method of claim 13, wherein the absence of said fusion gene indicates that the subject will be not be responsive to an ATP-competitive AKT inhibitor.
16. The method of claim 14 or 15, wherein the ATP-competitive AKT inhibitor is GSK690693,
A-443654, CCT-128930, GSK-2141795, AZD-5363, GDC-0068, A-674565 or AT7867.
17. The method of claim 13 wherein the presence of said fusion gene indicates that the subject will be not be responsive to an allosteric AKT inhibitor.
18. The method of claim 17, wherein the allosteric AKT inhibitor is MK2206 or PHT-427. A method of selecting an AKT inhibitor for a subject having cancer comprising identifying the presence or absence of a MAGI3-AKT3 fusion gene in a cancer cell obtained from said subject and selecting an ATP-competitive AKT inhibitor when said subject has said fusion gene.
EXPRESSING A MAGI3 - AKT3 FUSION GENE
 This application claims the benefit of U.S. Provisional Application Nos. 61/499,458, filed June 21, 2011 and 61/609,815, filed March 12, 2012 the contents of each are incorporated herein by reference in their entireties.
FIELD OF THE INVENTION
 The present invention relates generally to the treatment of cancer using an AKT inhibitor. More specifically the invention is related to treating cancers that express a MAGI3- AKT3 fusion gene.
 This invention was made with government support under C A 122099 and awarded by the National Institutes of Health. The United States government has certain rights in the invention.
BACKGROUND OF THE INVENTION
 Breast carcinoma is the leading cause of cancer-related mortality in women worldwide with an estimated 1.38 million new cases and 458,000 deaths in 2008 alone (Jemal et al., CA Cancer J. Clin. (2011)). Clinically, breast cancer remains the benchmark for the success of targeted cancer therapies with approved drugs targeting estrogen receptor activity including tamoxifen and aromatase inhibitors, or targeting amplified ERBB2 with trastuzumab or lapatinib, while other promising agents such as poly-ADP-ribose polymerase (PARP) inhibitors are under investigation for treatment of BRCA1/BRCA2 mutant tumors (Lancet. (1998) 351, 1451-1467; Baum, M. et al., Lancet (359, 2131-2139; Piccart-Gebhart, M.J. et al., N. Engl. J. Med. (2005) 353, 1659-1672; Geyer et al., N. Engl. J. Med. (2006) 355, 2733-2743; Fong et al., N. Engl. J. Med. (2009) 361, 123-134).
 Breast cancer represents a heterogeneous group of tumors with characteristic molecular features, prognosis, and responses to available therapy. Some of this heterogeneity arises from genetic alterations found in breast cancer including germline (TP53, BRCA1/BRCA , PTEN, MLHl, LKBl, and APC) and somatic (TP53, PIK3CA, PTEN, and AKTl) gene mutations and somatic amplifications (ERBB2, CCND1, CCNE1, and MFC) (Malkin, D. et al, Science (1990) 250, 1233-1238; Futreal, P.A. et al, Science (1994) 266, 120-122; Wooster R. et al, Nature (1995) 378, 789-792; Polyak, K. et al. Expert Rev. Mol. Med. (2002) 4, 1-4; Sjoblom, T. et al., Science (2006) 314, 268-274; Samuels, Y. et al., Science (2004) 304, 554; Li, J. et al., Science (1997) 275, 1943-1947; Carpten, J.D. et al., Nature (2007) 448, 439-444; King, C.R. et al., (1985) 229, 974-976; Schuuring, E. et al., Oncogene (1992) 7, 355-361; Callagy, G. et al., J. Pathol. (2005) 205, 388-396; Escot, C. et al., Proc. Natl. Acad. Sci. USA (1986) 83, 4834-4838). However, known genetic alterations explain only a fraction of the heterogeneity. Additional genetic events are likely to contribute to breast cancer initiation and the development of therapy resistance. Breast tumors have been classified into five major subtypes, based on gene expression signatures: Luminal A, Luminal B, Her2, and Basal-like and other (Perou, CM. et al., Nature (2000) 406, 747-752) Both luminal subtypes are associated with expression of estrogen (ER + ) and progesterone (PR + ) receptors and differential luminal epithelial cell markers. The subtypes differ in genomic complexity, key genetic alterations, clinical prognosis, and predictive response to available therapies (S0lie, T. et al., Proc. Natl. Acad. Sci. USA (2001) 98, 10869- 10874; Stephans, P. J. et al., Nature (2009) 462, 1005-1010; Chin, K. et al., Cancer Cell (2006) 10, 529-541; Gatza, M.L. et al., Proc. Natl. Acad. Sci. USA (2010) 6994-6999). An improved understanding of the genomic alterations in each subtype may lead to further improvements in therapy.
 Triple negative breast cancer, those that are estrogen receptor (ER) negative, progesterone receptor (PR) negative, and Her-2 negative comprise approximately 15% of all breast cancers and have an aggressive clinical course with high rates of local and systemic relapse. The clinical course reflects the biology of the tumor as well as the absence of conventional targets for treatment such as hormonal therapy for ER or PR positive patients and trastuzumab for Her-2 over-expressing tumors. In addition, these cancers may have different sensitivity to chemotherapeutic agents. As such, there is a great deal of interest in determining novel therapeutic regimens for this aggressive disease.
SUMMARY OF THE INVENTION
 The invention features methods of treating a subject having cancer by identifying a subject having a tumor expressing a MAGI3-AKT3 fusion gene and administering the subject an AKT inhibitor. The AKT inhibitor is an ATP-competitive AKT inhibitor.  In another aspect the invention features methods of decreasing tumor growth or inducing tumor cell apoptosis when the tumor expresses a MAGI3-AKT3 fusion gene by contacting the tumor with an AKT inhibitor.
 In yet another aspect the invention features methods of determining the
responsiveness of a subject having cancer to an AKT inhibitor by detecting the presence or absence of a MAGI3-AKT3 fusion gene. The presence of the fusion gene indicates that the subject would be responsive to an ATP-competitive AKT inhibitor and/or not be responsive to an allosteric AKT inhibitor. The absence of the fusion gene indicates that the subject would not be responsive to an ATP-competitive AKT inhibitor.
 In a further aspect the invention provides a method of selecting an AKT inhibitor for a subject having cancer by identifying the presence or absence of a MAGI3-AKT3 fusion gene in a cancer cell obtained from the subject and selecting an ATP-competitive AKT inhibitor when the subject expressed the fusion gene.
 The cancer/tumor is an epithelial cell cancer such as breast cancer. Preferably, the cancer is triple negative breast cancer.
 The AKT inhibitor is an ATP-competitive AKT inhibitor such as GSK690693, A- 443654, CCT-128930, GSK-2141795, AZD-5363, GDC-0068, A-674565 or AT7867.
 Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. Although methods and materials similar or equivalent to those described herein can be used in the practice of the present invention, suitable methods and materials are described below. All publications, patent applications, patents, and other references mentioned herein are expressly incorporated by reference in their entirety. In cases of conflict, the present
specification, including definitions, will control. In addition, the materials, methods, and examples described herein are illustrative only and are not intended to be limiting.
 Other features and advantages of the invention will be apparent from and
encompassed by the following detailed description and claims.
 BRIEF DESCRIPTION OF THE DRAWINGS
 Figure 1: Most significantly-mutated genes in breast cancer as determined by whole exome sequencing. Upper histogram: samples ordered by overall mutation rate (point mutations and indels) with synonymous rate in green, non synonymous rate in blue. Left histogram: number of total mutations per gene observed and percentage of samples affected (color coding as in upper panel). Central heatmap: Distribution of significant mutations across sequenced samples (Other non synonymous mutations = nonsense, indel, splice- site). Right histogram: -loglO score of MutSig determined q value. Red line at q = 0.1. Lower chart: top - rates of non- silent mutations within categories indicated by legend; bottom - key molecular features of samples in each column (Duct. = Infiltrating ductal carcinoma, DCIS = Ductal carcinoma in situ, Lob. = Infiltrating lobular carcinoma).
 Figure 2: CBFB mutations and RUNXl deletions.
A. CBFB with RUNX binding domain in green. Mutations identified in this study (red bullets), previously identified mutations (Irizarry et al. Biostatistics (2003) 4, 249-264; Parker et al. Nature (2009) 27, 1160-1167) (black bullets), and known leukaemia CBFB-MYH11 fusion is indicated.
B. Detail of the 3 Mb region surrounding the RUNXl homozygous deletions in samples BR-M- 045 and BR-M- 174. Bi-allelic copy ratios from Affymetrix SNP 6.0 microarray data were analysed using HAPSEG. The copy-ratios of the (germline) heterozygous markers (colored points) clustered into two values (red and blue), corresponding to the copy-ratio of each homologous chromosome in the cancer sample (horizontal colored lines). Equal homologous copy-ratios are indicated by purple lines.
C. Analysis of genome-wide homologous copy-ratios using ABSOLUTE. The histogram of homologous copy-ratios (as in B) is shown for each sample, with the indicated absolute copy- numbers superimposed (dotted lines). The genomic region corresponding to RUNXl in (B) is present at an allelic copy-ratio corresponding to 0 copies per cancer cell.
 Figure 3: MAGI3-AKT3 fusion gene.
A. Diagram depicting balanced translocation involving MAGI3 and AKT3.
B. (top) Genomic DNA PCR analysis for AKT3, MAGI3, and both products of the balanced translocation in tumor (T) and normal (N). (bottom) cDNA PCR validation of fusion in tumor.
C. (above) MAGI3 and AKT3 protein structure with indicated domains (proteins not drawn to scale); (below) putative fusion protein.
D. Western blots of ZR-75 cells in low-serum media overexpressing vector, MAGI3-AKT3 fusion, or AKT1 E17K mutant, and comparing levels of phosphorylated Akt and GSK3P; (left) infected cells with and without insulin growth factor 1 (IGF-1) stimulation; (right) treatment of vector or MAGI3-AKT3 overexpressing cells with allosteric pan-Akt inhibitor (MK-2206) or ATP-competitive pan-Akt inhibitor (GSK-690693).
E. Focus formation assays with Rat-1 cells infected with vector or pLX overexpressing MAGI3- AKT3, and stained with crystal violet.
 Figure 4: Frequent genetic alterations across pathways and potentially druggable targets. A. Distribution of common genomic alterations in exome-sequenced samples, with genes in rows and samples in columns grouped by expression subtype. Druggable genes are indicated by name of drug or status of inhibitor development (Piccart-Gebhart, M.J. et al., N. Engl. J. Med. (2005) 353, 1659-1672; Geyer et al., N. Engl. J. Med. (2006) 355, 2733- 2743; Fong et al., N. Engl. J. Med. (2009) 361, 123-134; Pal, S.K. et al., Expert Opin. Investig. Drugs (2010) 19, 1355-1366; Wallin, J. J. et al., Sci. Transl. Med. (2010) 2, 48ra66; Kozicizak, M. et al., Oncogene (2004) 23, 3501-3508; Weiss, J. et al., Sci. Transl. Med. (2010) 2, 62ra93; Greulich, H. et al., PLoS Med. (2005) 2, e313; Chou, J. et al., J. Cell Physiol. (2010) 222, 42-49). Heatmap: purple - mutation, red - focal amplification, blue - focal deletion. B. Diagram of interconnected alterations observed in sequenced samples. Genes are depicted in relationship to locations of common cellular activity (outer boundary - plasma membrane; inner boundary - nuclear membrane). Color coding as shown in legend reflects major alteration type observed for each gene. Frequency of gene alteration across samples is shown as numerical percent adjacent to each gene box.
 Figure 5: Sample processing pipeline. Vietnamese samples for whole-genome sequencing were selected based on manual review showing at least one region of homozygous deletion while Mexican samples were selected based on tumor purity >50 and allelic fraction >20 .
 Figure 6: Sample analysis pipeline for determining somatic mutations and rearrangements.
 Figure 7: Ancestry analysis. A. Individual ancestry proportions. Four parental groups K=4: Asia (CHB+JPT), Europe (CEU), Africa (YRI) and Native Mexican - Zapoteca, Maya, and Tepehuano (ZAP, MAY, TEPEH) and Breast Cancer cases. B. Populations used for ancestry assessment (Silva-Zolezzi et al. Proc Natl Acad Sci USA (2009) 106, 8611-8616).  Figure 8: Mutation rate by expression subtype. Plot: Overall mutation rate of samples plotted according to breast expression subtype as determined by PAM50 classification. Histogram: Breakdown of mutation spectra across expression subtypes and samples.
 Figure 9: Comparison of somatic mutations identified via whole-exome and whole-genome sequencing. A. Plot of somatic mutations in genomes and exomes according to allelic fraction of event by each method. Mutations found by both methods shown in green. B. Concordance of somatic mutations called by both methods binned by allelic fraction. Whole- exome sequencing overall is able to identify mutations at a lower allelic fraction. Mutations found only in whole-genomes likely represent false-positive somatic mutations due to lower depth of sequence coverage.
 Figure 10: Representative ABSOLUTE plots of samples harboring RUNXl deletions. A. For each sample, plot on left demonstrates genome- wide view of copy ratios for both homologous chromosomes. The copy ratios are shown for each genomic segment with locally constant copy number. Color-axis indicates distance between low (blue) and high (red) homologue concentration; segments where these are similar (homologous-allele balance) are purple. Inset zoomed on Chromosome 21 with homozygous RUNXl deletion shown by arrow. Plot on right shows homologous copy-ratio histogram obtained by binning at 0.04 resolution (y- axis); the length of each block corresponds to the (haploid) genomic fraction (jc-axis) of each corresponding segment. B. Plot of relative copy number of RUNXl and AKT3 in tumor and normal DNA from samples suspected to harbor RUNXl homozygous deletions. Relative quantities are normalized to CDH7, used as a diploid internal control. Findings are consistent with homozygous RUNXl deletion in both tumor samples, with lower purity of BR-M-174 tumor DNA.
 Figure 11: ERBB2 copy number and mutation status across samples. DNA amplification shown in red. Samples with indicated ERBB2 somatic mutations shown on left.
 Figure 12: Somatic rearrangements observed in 22 whole-genomes. Genome- wide Circos plots organized by breast expression subtype. Chromosomal position shown in outer ring, copy number shown in inner ring. Inter-chromosomal rearrangements - red lines; intra-chromosomal rearrangements - green lines.
 Figure 13: MAGI3 copy number status and MAGI3-AKT3 expression. A. Copy number of MAGI3 as determined by SNP array. BR-M-045 sample with fusion gene shown at top. B. Expression levels of MAGI3 and AKT3 in tumor and normal control samples as determined by exon arrays. Red line indicates exon expression profile of MAGI3 and AKT3 in BR-M-045 sample.
 Figure 14: Recurrence of MAGI3-AKT3 fusion across breast cancer samples.
Sample identity indicated for lanes with positive band. BR-M-045 used as positive control and appears multiple times. Green asterisk indicates repeated sample.
 Figure 15: Significant GISTIC amplification and deletion peaks in our collection. Amplification in red and deletions in blue. Green line indicates FDR q-value=0.25.
Chromosomal position indicated to right of plot with focus of amplification and deletion as labeled.
 Figure 16: Power to detect mutations in significantly mutated genes as determined by ABSOLUTE. Samples along x-axis arranged from least to greatest tumor cell purity. Genes represented along y-axis with each row representing individual exons within gene. Dark green squares represent exons with zero power while salmon squares represent exons with power = 1. Grey squares represent regions of homozygous deletion. Mutations shown with cyan "x".
DETAILED DESCRIPTION OF THE INVENTION
 This invention is based upon the discovery of a recurrent translocation in breast cancer patients leading to an in-frame MAGI3-AKT3 fusion gene. This finding indicates that treatment with AKT inhibitors would provide therapeutic benefits to cancers harboring this translocation, in particular triple-negative breast cancers, a subtype where limited therapeutic options exist beyond systemic cytotoxic chemotherapy.
 Accordingly, the invention method of treating cancer by identifying in a tumor sample from a subject a MAGI3-AKT3 fusion gene and administering an AKT inhibitor. The cancer is any cancer in which the tumor has a translocation resulting in an MAGI3-AKT3 fusion gene. For example the cancer is an epithelial cell cancer such as breast cancer. In particular, the cancer is a triple negative breast cancer.
 An AKT inhibitor is a compound that decreases the expression or activity of AKT. AKT is a serine/threonine protein kinase that plays a key role in multiple cellular processes such as glucose metabolism, cell proliferation, apoptosis, transcription and cell migration. In humans, there are three genes in the "AKT family": Aktl, Akt2, and Akt3.
 A decrease in AKT3 expression or activity is defined by a reduction of a biological function of the serine/threonine kinase. A serine/threonine kinase biological function includes for example, catalyzing the phosphorylation of serine or threonine.
 An AKT inhibitor acts for example by, blocking kinase- substrate interaction, inhibiting the enzyme's adenosine triphosphate (ATP) binding site or blocking extracellular tyrosine kinase receptors on cells. Preferably, the inhibitor is an ATP-competitive AKT inhibitor.
 AKT kinase activity is measured by detecting phosphorylation of a protein. AKT inhibitors are known in the art or are identified using methods described herein. For example, an AKT inhibitor is identified by detecting a decrease the serine/threonine kinase mediated transfer phosphate from ATP to protein serine or threonine residues.
 ATP-competitive AKT inhibitor such as GSK690693, A-443654, CCT- 128930, GSK-2141795, AZD-5363, GDC-0068, A-674565 or AT7867.
 Other AKT inhibitors allosteric AKT inhibitors such as MK2206 or PHT-427.
 AKT inhibitors also include for example
 Palomid 529, Perifosine, PHT-427, KP372-1, 2-pyrimidyl-5-amidothiophenes, or isothiocyanates.
 Other AKT inhibitors include those described in U.S. Pat. Nos. 7,943,732; 7,625,890; 7,414,063; 7,919,504; 7,776,589; 6,809,194 and US Application No. 20110129455,
20110071182 each of which is hereby incorporated by reference in their entireties.
 Therapeutic Methods
 The growth of cells is inhibited, e.g. reduced or apoptosis is induced by contacting a cell with a composition containing an AKT inhibitor. By inhibition of cell growth is meant the cell proliferates at a lower rate or has decreased viability compared to a cell not exposed to the composition. Cell growth is measured by methods know in the art such as, the MTT cell proliferation assay. By inducing apoptosis is meant an increase of oxidative stress induced cell death. The process of apoptosis is characterized by, but not limited to, several events. Cells lose their cell junctions and microvilli, the cytoplasm condenses and nuclear chromatin marginates into a number of discrete masses. As the nucleus fragments, the cytoplasm contracts and mitochondria and ribosomes become densely compacted. After dilation of the endoplasmic reticulum and its fusion with the plasma membrane, the cell breaks up into several membrane- bound vesicles, apoptotic bodies, which are usually phagocytosed by adjacent bodies. As fragmentation of chromatin into oligonucleotides fragments is characteristic of the final stages of apoptosis, DNA cleavage patterns can be used as and in vitro assay for its occurrence (Cory, Nature 367: 317-18, 1994). Many methods for measuring apoptosis, including those described herein, are known to the skilled artisan including, but not limited to, the classic methods of DNA ladder formation by gel electrophoresis and of morphologic examination by electron microscopy. The more recent and readily used method for measuring apoptosis is flow cytometry.
 Cells are directly contacted with an inhibitor. Alternatively, the inhibitor is administered systemically. Inhibitors are administered in an amount sufficient to decrease (e.g., inhibit) cell proliferation or induce apoptosis.
 The cell is a tumor cell such as a carcinoma, adenocarcinoma, blastoma, leukemia, myeloma, or sarcoma. The cell is an epithelial cell cancer such as breast cancer. In particular, the cancer is a triple negative breast cancer.
 In various aspects the cell has translocation leading to an in-frame MAGI3-AKT3 fusion gene. A MAGI3-AKT3 fusion gene is identified by methods known in the art. Optionally, the cell is resistant to tamoxifen, aromatase inhibitors or trastuzumab.
 The methods are useful to alleviate the symptoms of a variety of cancers. Any cancer containing a MAGI3-AKT3 fusion gene is amendable to treatment by the methods of the invention. In some aspects he subject is suffering from triple negative breast cancer. The subject is resistant to tamoxifen, aromatase inhibitors or trastuzumab.
 Treatment is efficacious if the treatment leads to clinical benefit such as, a decrease in size, prevalence, or metastatic potential of the tumor in the subject. When treatment is applied prophylactically, "efficacious" means that the treatment retards or prevents tumors from forming or prevents or alleviates a symptom of clinical symptom of the tumor. Efficaciousness is determined in association with any known method for diagnosing or treating the particular tumor type
 Therapeutic Administration
 The invention includes administering to a subject a composition comprising an AKT inhibitor.  An effective amount of a therapeutic compound is preferably from about 0.1 mg/kg to about 150 mg/kg. Effective doses vary, as recognized by those skilled in the art, depending on route of administration, excipient usage, and coadministration with other therapeutic treatments including use of other anti-proliferative agents or therapeutic agents for treating, preventing or alleviating a symptom of a cancer. A therapeutic regimen is carried out by identifying a mammal, e.g., a human patient suffering from a cancer that has a MAGI3-AKT3 fusion gene using standard methods.
 The pharmaceutical compound is administered to such an individual using methods known in the art. Preferably, the compound is administered orally, rectally, nasally, topically or parenterally, e.g., subcutaneously, intraperitoneally, intramuscularly, and intravenously. The inhibitors are optionally formulated as a component of a cocktail of therapeutic drugs to treat cancers. Examples of formulations suitable for parenteral administration include aqueous solutions of the active agent in an isotonic saline solution, a 5% glucose solution, or another standard pharmaceutically acceptable excipient. Standard solubilizing agents such as PVP or cyclodextrins are also utilized as pharmaceutical excipients for delivery of the therapeutic compounds.
 The therapeutic compounds described herein are formulated into compositions for other routes of administration utilizing conventional methods. For example, the therapeutic compounds are formulated in a capsule or a tablet for oral administration. Capsules may contain any standard pharmaceutically acceptable materials such as gelatin or cellulose. Tablets may be formulated in accordance with conventional procedures by compressing mixtures of a therapeutic compound with a solid carrier and a lubricant. Examples of solid carriers include starch and sugar bentonite. The compound is administered in the form of a hard shell tablet or a capsule containing a binder, e.g., lactose or mannitol, conventional filler, and a tableting agent. Other formulations include an ointment, suppository, paste, spray, patch, cream, gel, resorbable sponge, or foam. Such formulations are produced using methods well known in the art.
 Therapeutic compounds are effective upon direct contact of the compound with the affected tissue. Accordingly, the compound is administered topically. Alternatively, the therapeutic compounds are administered systemically. For example, the compounds are administered by inhalation. The compounds are delivered in the form of an aerosol spray from pressured container or dispenser which contains a suitable propellant, e.g., a gas such as carbon dioxide, or a nebulizer.
 Additionally, compounds are administered by implanting (either directly into an organ or subcutaneously) a solid or resorbable matrix which slowly releases the compound into adjacent and surrounding tissues of the subject.
 EXAMPLE 1: GENERAL METHODS
 A. Sample and Collection Attributes, DNA RNA Collection, and Quality Control
 Clinical cohorts
 Mexican samples were collected under an IRB approved protocol during the
2008-2010 period at the Instituto de Enfermedades de la Mama - FUCAM A. C. hospital.
Tumor, normal adjacent tissue, and peripheral blood were obtained from each patient after informed consent during surgery by S.R.C. for tumor resection. After macroscopic inspection by a pathologist, sections of tumor and normal tissues were immediately frozen in liquid nitrogen and stored at -80°C until further processing. A section of these tissues were formalin fixed, paraffin embedded (FFPE) and 5-micron sections were stained using hematoxylin and eosin (H&E) for confirmation of diagnosis, assessing grade, and tumor cell content evaluation.
Estrogen and progesterone receptors as well as HER2 expression were evaluated using the ER/PR pharmDx and HercepTest, respectively (Dako, Denmark).
 Fresh frozen Vietnamese samples were acquired from the BioServe commercial tissue repository (www.bioserve.com) following careful review of IRB and informed consent documents applicable to each sample. According to their guidelines, a board certified pathologist reviewed all samples to confirm diagnosis, assess grade, and evaluate tumor cell content. H&E slides were provided for each sample to confirm of diagnosis. ER/PR/HER2 expression status by immunohistochemistry was available only if provided by the original hospital responsible for specimen collection.
 DNA/RNA extraction
 Mexican Samples: After tumor cell content confirmation, DNA and RNA were extracted from the frozen tissues and peripheral blood lymphocytes using the AUPrep DNA/RNA mini kit (Qiagen, Valencia, CA) according to manufacturer's instructions. DNA integrity was evaluated by 1% agarose gel electrophoresis and RNA integrity by capillary electrophoresis using the Bioanalyzer system (Agilent, Santa Clara, CA). Only samples with RNA integrity number (RIN) greater than 6.0 were used for expression microarray analysis.
 Vietnamese Samples: DNA extraction was performed on fresh frozen tumor and adjacent normal tissue using DNAQuik reagents developed by BioServe. DNA was run on 1% agarose gels to assess structural integrity. RNA extraction was performed using Trizol (Qiagen) and the quality was determined using the Bioanalyzer system. RNA with a RIN score >6.0 was stored at -80°C until use.
 DNA quality control
 We used standard Broad Institute protocols as recently described (Berger, F. et al.,
Nature (2011) 470, 214-220; Chapman, M.A. et al., Nature (2011) 471, 467; Stransky, N. et al., Science (2011) 333, 1157-1160). Tumor and normal DNA concentration was measured using PicoGreen® dsDNA Quantitation Reagent (Invitrogen, Carlsbad, CA). A minimum DNA concentration of 60 ng/μΐ was required for sequencing. In select cases where concentration was <60 ng/μΐ, ethanol precipitation and re-suspension was required to increase concentration. Gel electrophoresis confirmed that the large majority of DNA was high molecular weight. We prepared reserve stocks of each sample using whole genome amplification (WGA) for use in subsequent validation efforts. All Illumina sequencing libraries were created with the native DNA. The identities of all tumor and normal DNA samples (native and WGA product) were confirmed by mass spectrometric fingerprint genotyping of 24 common SNPs (Sequenom, San Diego, CA).
 B. cDNA Microarrays and Expression Subtype Determination
 Expression Microarrays
 cDNA generated from RNA was hybridized on human whole-transcript microarrays (Human Gene ST 1.0, Affymetrix, Santa Clara CA), according to manufacturer's instructions. Samples classified as 141 Mexican samples included 35 normal and 106 tumors that were processed at the Affymetrix Unit of the Instituto Nacional de Medicina Genomica
(INMEGEN) in Mexico City. cDNA from tumor for all Vietnamese samples was processed at the Genetic Analysis Platform (GAP) at the Broad Institute in Cambridge, MA.
 Breast expression subtyping
 Raw gene expression profiles from all 201 samples were obtained after low-level analysis and quality assessment for the two sets separately since no comparison between the two populations was planned. Probe level data on each set were log2 transformed, background corrected using RMA4 and normalized using quantile normalizations. These algorithms are coded in the "oligo" package in Bioconductor. Gene expression data was further processed to determine breast cancer molecular subtypes according to the expression profiles classification of PAM506. The PAM50 gene expression test aims at classifying breast cancer tumors into 5 known intrinsic subtypes: Luminal A, Luminal B, Her2, Basal-like, and Normal-like and also provides a continuous risk of recurrence (ROR) score based on the similarity of an individual sample to the prototypic subtypes.
 C. Single-Nucleotide Polymorphism (SNP) Array Based Analysis
 Single-nucleotide polymorphism arrays
 Non-WGA genomic DNA from tumor and paired normal samples was processed using Affymetrix Genome- Wide Human SNP Array 6.0 (Affymetrix, Inc.) according to manufacturer's protocols. DNA was digested with Nspl and Styl enzymes (New England Biolabs), ligated to the respective Affymetrix adapters using T4 DNA ligase (New England Biolabs), amplified (Clontech), purified using magnetic beads (Agencourt), labeled, fragmented, and hybridized to the arrays. Following hybridization, the arrays were washed and stained with streptavidin-phycoerythrin (Invitrogen). Array preparation and scanning was performed at the genotyping core laboratory of INMEGEN and GAP at the Broad Institute for the Mexican and Vietnamese samples respectively.
 Copy-number assessment
 Data preprocessing was performed using Affymetrix Power Tools. Copy number data was evaluated after segmenting the log 2 ratios between tumor and paired normal levels on a sample basis. Quality control, data integrity, segmentation and copy number analysis were performed as previously described7. Segmented copy number data was visualized with the Integrative Genomics Viewer (IGV) (Robinson, J.T. et al., Nat. Biotechnol. (2011) 29, 24-26). The Genomic Identification of Significant Targets in Cancer (GISTIC) algorithm was used to identify broad and focal regions of copy number alterations in individual samples as described (Beroukhim, R. et al., Nature (2010) 463, 899-905; Mermel, C.H. et al., Genome Biology (2011) 12, R41).
 Purity, ploidy, and allele -specific copy number analysis  SNP Array data was analyzed using the HAPSEG11 and ABSOLUTE algorithms to infer the tumor purity, average ploidy, and allele- specific copy number levels. Allelic fraction for each tumor was calculated, indicative of the fraction of sequence reads expected to harbor the non-reference allele at a locus with a somatic mutation existing at a single copy per nucleus.
 Ancestry Analysis
 A total of 140 sample SNP arrays were used for the ancestry analysis: 100 samples from Mexico and 40 from Vietnam. Genotypes of 301,219 common SNPs in three genotyping platforms (SNP 6.0, Affymetrix 500K, and Illumina 1M) were used. 196 HapMap samples (CEU: northern European ancestry; YRI, Africans from Nigeria and CHB+JPT, east Asian population) and 71 native Mexican samples were included as parental populations for the analysis, while 161 Mexican mestizo samples from the Mexican Genome Diversity Project (MGDP) were included to evaluate ancestry proportions in the general Mexican population (Silva-Zolezzi, I. et al., Proc. Natl. Acad. Sci. (2009) 106, 8611-8616). Four major quality control tests were performed: 1) Missing rate per person excluding individuals with more than 5% missing genotypes, 2) Missing rate per SNP: only SNPs with a 95% genotyping rate were included, 3) Exclusion of markers that failed the Hardy- Weinberg Equilibrium test at 0.00001 significance threshold, 4) Identity-by-descent (IBD) test to assess quality on the full set of samples. Quality of all samples was good and no familial relationships were found. Principal components analysis (PCA) was used to detect population substructure using genome- wide data using EIGENSOFT 3.013. Individual average ancestral proportions were determined using ADMIXTURE 1.1 software (Price, A.L. et al., Nat. Rev. Genet. (2010) 11, 459-463). Based on the origin of the samples, the value of K was chosen to be 4 meaning that four parental groups were considered to quantify ancestral contribution: CEU, YRI, CHB+JPT and NATMEX and to explain the major substructure in this set of 140 individuals (Figure 7).
 D. Sequence Data Generation
 A total of 125 samples were initially sequenced with 120 successfully completed
(Figure 5). Ninety-seven underwent whole exome sequencing only and 5 samples whole genome sequencing only. Additional 18 samples were sequenced with both methods. Tumor and normal samples were sequenced according to the manufacturer's protocols (Illumina, San Diego, CA) as previously described with a brief summary provided below (Berger, F. et al., Nature (2011) 470, 214-220; Chapman, M.A. et al., Nature (2011) 471, 467; Stransky, N. et al., Science (2011) 333, 1157-1160).
 Whole Genome Sequencing Library Construction
 We followed established protocols at the Broad Institute as previously described
(Stransky, N. et al., Science (2011) 333, 1157-1160). A total of 1 μg of genomic DNA was sheared to a range of 100-700 bp. Each of the resulting WGS libraries was sequenced on an average of 13 flow cells lanes of the Illumina GA-II or HiSeq sequencers. Using 101 bp paired- end reads, we aimed to reach 30X average genomic coverage for each of the tumor and normal genomes. The mean coverage achieved was 36x in tumors and 38x in normals.
 Whole Exome Sequencing Library Construction
 We follow the procedure described by Gnirke et al. adapted for production- scale exome capture libraries (Gnirke, A. et al., Nat. Biotechnol. (2009) 27, 182-189; Stransky, N. et al., Science (2011) 333, 1157-1160). Resulting exome sequencing libraries were sequenced on 3 lanes of an Illumina GA-II sequencer, using 76 bp paired-end reads. The mean coverage achieved was 141x in the tumors and 133x in the normals.
 Illumina sequencing
 Libraries were quantified using a SYBR Green qPCR protocol with specific probes for the ends of the adapters3. Libraries were normalized to 2nM and then denatured using 0.1 N NaOH. Cluster amplification of denatured templates occurred according to manufacturer's protocol (Illumina) using V2 Chemistry and V2 Flowcells (1.4mm channel width). SYBR Green dye was added to all flowcell lanes to provide a quality control checkpoint after cluster amplification to ensure optimal cluster densities on the flowcells. Flowcells were paired-end sequenced on Genome Analyzer II or HiSeq machines, using V3 Sequencing-by-Synthesis kits and analyzed with the Illumina vl.3.4 pipeline. Standard quality control metrics including error rates, % passing filter reads, and total Gb produced were used to characterize process performance prior to downstream analysis. The Illumina pipeline generates data files that contain the reads and qualities.
 E. Sequence Data Processing
 Sequencing data were processed using two consecutive pipelines (Berger, F. et al., Nature (2011) 470, 214-220; Chapman, M.A. et al., Nature (2011) 471, 467; Stransky, N. et al., Science (2011) 333, 1157-1160):  (1) Sequencing data-processing pipeline - "Picard" - uses the reads and qualities produced by the Illumina software for all lanes and libraries generated for a single sample (either tumor or normal) and produces a single BAM file (http://samtools.sourceforge.net/SAMl.pdf) representing the sample. The final BAM file stores all reads and calibrated qualities along with their alignments to the genome.
 (2) Cancer genome analysis pipeline - "Firehose" - takes the BAM files for the tumor and patient-matched normal samples and performs analyses including quality control, local- realignment, mutation calling, small insertion and deletion identification, rearrangement detection, coverage calculations and others as described briefly below and more extensively in Stransky et al (Stransky, N. et al., Science (2011) 333, 1157-1160).
 The Cancer Genome Analysis Pipeline ( "Firehose ")
 The pipeline represents a set of tools for analyzing massively parallel sequencing data for both tumor DNA samples and their patient-matched normal DNA samples. Firehose uses GenePatternl6 as its execution engine for pipelines and modules based on input files specified by Firehose. The pipeline contains the following steps (described in detail in Chapman, M.A. et al., Nature (2011) 471, 467 and Stransky, N. et al., Science (2011) 333, 1157-1160) (Figure 6):
 Quality control - confirms identity if individual tumor and normal to avoid mix-ups between tumor and normal data for the same individual.
 Local realignment of reads - realigns sites potentially harboring small insertions or deletions in either the tumor or the matched normal to decrease the number of false positive single nucleotide variations caused by misaligned reads.
 Identification of somatic single nucleotide variations (SSNVs) - MuTect algorithm - candidate SSNVs were detected using a statistical analysis of the bases and qualities in the tumor and normal BAMs.
 Identification of somatic small insertions and deletions - Indelocator algorithm - putative somatic events were first identified within the tumor BAM file and then filtered out using the corresponding normal data.
 Identification of inter-chromosomal and large intra-chromosomal structural rearrangements - dRanger algorithm - candidate rearrangements were identified as groups of paired-end reads which connected genomic regions with an unexpected orientation and/or distance. When possible, breakpoints are mapped to basepair resolution using BreakPointerl.  Mutation rate calculation - we calculated base mutation rates using the detected mutations (SSNVs and indels) and the coverage statistics.
 Identification of significantly mutated genes - MutSig algorithm - genes that harbored a greater number of mutations than expected by chance were detected by comparing the observed number of mutations across the samples to the expected number based on the background mutation rates and the covered bases in all samples. Genes list in Figure 1 of main manuscript were selected after filtering that included: eliminating gene with q value of >0.1 after correction for multiple hypothesis testing, manual review of reads, and fewer than 2 mutations per sample. Subsequent input of samples that failed orthogonal validation were used to correct the background mutation rate.
 Identification of significantly mutated genesets - using canonical pathways used in Gene Set Enrichment Analysis (GSEA) that contained at least one gene targeted in whole-exome sequencing— yielding 616 gene sets.
 Mutation annotation - Detected point mutation and indels calls were annotated with the annotation pipeline Oncotator.
 Mutation calling in non-coding regions with regulatory potential - regions were defined as previously described (Chapman, M. A. et al. Nature (2011) 471, 467). For each of the 2.38 million regions with regulatory-potential, we calculated the number of mutations observed across our panel of 22 cases with whole-genome sequencing, and the number of covered bases. We performed a p-value calculation and FDR correction for 2.38 million hypotheses. We manually removed all sites with mutations in only one patient or a single genomic site. This left 20 regions with a q-value <0.25 of which only 3 did not appear to be sequencing artifacts after manual review. To identify significant mutations in estrogen receptor (ER) binding sites, we first downloaded a list of ChlP-Seq annotated ER binding sites from Grober et all7: contained 6024 ER alpha binding sites, and 9702 ER beta binding sites comprising 1.5mb and 2.3mb of genomic territory respectively. Since these regions are overlapping, they were analyzed separately using the same method for identification of significantly mutated genes, using the ER-binding regions for the genomic territory instead of genes. A post-processing step was performed as above: manual review and regions significant with only one sample.
 F. High-Throughput Experimental Validation of Point Mutations, Indels, and Rearrangements  Somatic Mutations
 We obtained independent validation for 494 candidate mutations using mass spectrometric genotyping (Sequenom) of the tumor and normal DNA, or alternative next- generation sequencing of tumor DNA using 454 pyrosequencing, Pacific Biosciences SMRT cell targeted sequencing (for PIK3CA and TP53 mutations), and Illumina exome sequencing of frozen tumor-matched formalin-fixed paraffin embedded tissues. Whole-genome amplified DNA was used for all validation experiments except Illumina sequencing of FFPE tissue where non- amplified DNA was used. Mutations selected for validation included all candidate protein-coding mutations in significantly-mutated genes in the Illumina whole-exome data with q- value < 0.2, and all genes in significantly- mutated genesets with q- value < 0.1.
 Somatic Rearrangements
 We attempted PCR validation of all dRanger identified rearrangements across the 22 genomes that met the criteria: 1. dRanger score >= 4; 2. event results in duplication/deletion of entire exons, or results in frame protein/transcript fusion. Two unique primer pairs were designed for each event.
 G. Other Methods
 Germline mutation calling
 Mutation calling was performed using Unified Genotyper as previously described (Depristo, M.A. et al., Nat. Genet. (2011) 43, 491-498). Called germline variations were compared to a list of functionally annotated variations to assess for pathogenic significance (Osborne, R.H. et al., Med. J. Aust. (2000) 172, 463-464).
 qPCR to confirm absolute copy number
 To verify whether two samples, MEX-BR-45 and MEX-BR 174, had homozygous deleted RUNXl, qPCR was run following the manufacture's protocol for the Brilliant II SYBR® Green qPCR Master Mix with a gDNA input of 15ng per reaction in triplicates. The following primers were used.
 A  TCCTGTCCCTGCT  CGCTCCTCAGAGAA
KT3 GTTTACCCTGC CACCCGC
 CDH7 was used as the diploid endogenous control, and AKT3 as a representative amplified region in these two tumors. Tumor normal pairs were run on an ABI Prism 7900HT with an annealing temperature of 61°C at one minute and with the extension time at 45 seconds. Ct values, AACt values, and relative quantification for each target gene was determined by RQ Manager 1.2 software (Applied Biosystems).
 Independent validation of balanced translocation involving MAGI3 andAKT3 in tumor and normal genomic DNAfrom case BR-M-045
 Aliquots of tumor and blood normal DNA was obtained for case BR-M-045. PCR amplification was performed using AccuPrime Taq DNA Polymerase (Invitrogen) with the following primer pairs.
 PCR amplification of MAGI 3 -AKT3 fusion gene from patient cDNA
 Total RNA was obtained from tumor for case BR-M-045. Double stranded cDNA was made from 200 ng of RNA using the Superscript III cDNA Synthesis kit (Invitrogen) with and without the inclusion of RNA polymerase for first strand synthesis. PCR amplification was performed using the AccuPrime Taq DNA Polymerase on double strand cDNA using forward primer (5'-AAGCCCCTGAAGACTGTGAA-3') in MAGI3 and reverse primer (5'- ACTTGCCTTCTCTCGA ACC A- 3 ') in AKT3. 35 PCR cycles were performed as follows: 95°C for 2 min, 95°C for 30 sec, 57.2°C for 30 sec, 72°C for 100 sec, 72°C for 5 min, 4°C hold.
 Assessment of expressed MAGI3-AKT3 fusion prevalence across breast cancer samples
 Tumor RNA was obtained from the Massachusetts General Hospital via D.C.S., the Susan F. Smith Women's Cancers Tissue Repository at the Brigham and Women's Hospital via A.L.R., and from the Instituto de Enfermedades de la Mama FUCAM via S.R.C and A.H.M. First strand cDNA was made from 50 ng total RNA using the Superscript III First Strand cDNA Synthesis System (Invitrogen) and purified using the MinElute PCR Purification Kit
(Qiagen). PCR amplification was performed on the purified first strand cDNA using primers spanning the MAGI3-AKT3 breakpoint. The forward primer is located on exon six of MAGI3 (5'-AAGCCCCTGAAGACTGTGAA-3') and the reverse primer is located on exon five of AKT3 (5'- ACTTGCCTTCTCTCGAACCA-3'). Forty PCR cycles were performed as follows: 95°C for 2 min, 95°C for 30 sec, 57.2°C for 30 sec, 72°C for 100 sec, 72°C for 5 min, 4°C hold. An 833- bp PCR product was expected for the fusion. Bands were excised and purified using the
QIAquick Gel Extraction Kit (Qiagen). TOPO TA Cloning (Invitrogen) was performed prior to Sanger sequencing.
 Gateway Cloning of MAGI3-AKT3 fusion for validation experiments
 Double stranded cDNA was generated from total RNA from case BR-M-045 as described above, and purified using the MinElute PCR Purification Kit (Qiagen). Two PCR products were generated with overlapping sequence using Gateway® cloning compatible primers and Taq DNA polymerase HiFi (Invitrogen). Fragment 1: attbl (5'-
GGGGACAAGTTTGTAC AAAAAAGCAGGCTTAATGTCGAAGACGCTGAAG-3 ' ) and AKT3 reverse primer (5'- ACTTGCCTTCTCTCGAACCA-3'), Fragment 2: attb2 (5'- GGGGACC ACTTTGTACAAGAAAGCTGGGTCTTATTCTCGTCCACTTGC AGA-3 ' ) and the MAGI3 forward primer (5'-AAGCCCCTGAAGACTGTGAA-3').
 The 5' and 3' overlapping PCR products were added to the BP reaction with pDONR221 plasmid. In vitro recombination resulted in insertion of the full-length fusion as tested by restriction digest and PCR across the overlapping region. The fusion was subcloned into the pBabe-Puro and pLX304-Blast destination vectors.  Cell culture
 ZR75 cells were maintained in RPMI-1640 media (Cellgro; Manassas, VA) supplemented with 10% fetal bovine serum (FBS) (GIBCO; Carlsbad, CA). Rat-1 fibroblasts and HEK-293T cells were maintained in DMEM (Cellgro) supplemented with 10% FBS (GIBCO).
 Plasmids and lentivirus production
 Plasmids (pLX304 and pLX304-MAGI3-Akt3): pLX304 plasmid constructs were co- transfected in HEK-293T cells with the packaging vectors pCMV-VSVG and psPAX2 using polyethylenimine (PEI). Lenti viral supernatents were harvested 48 hr after transfection, passed through a 0.45 μιη filter and used to infect target cells.
 HA-Aktl Glul7Lys (E17K) was constructed by site-directed mutagenesis. Cells were transfected with pcDNA3-Aktl-E17K using Lipofectamine 2000 (Invitrogen; Carlsbad, CA) according to the manufacturer's protocol.
 Growth factors and inhibitors
 Serum-starved cells were stimulated with recombinant human IGF-1 (R&D Systems; Minneapolis, MN) at a final concentration of 100 ng/mL for 20 min. Serum-starved cells were exposed to the pan-Akt inhibitors MK-2206 (Active Biochem; Wanchai, China) and GSK- 690693 (SynKinase; Shanghai, China) at final concentrations of 1 μΜ for 20 min.
 ZR75 cells were infected with viral supernatent and 5 μg/mL polybrene (Millipore; Billerica, MA) or transfected with Aktl-E17K for 48 hr prior to serum- starvation for an additional 16 hr. Cells were exposed to IGF-1 or inhibitors, washed with ice-cold PBS and lysed in ice-cold lysis buffer (1% NP-40, 150 mM NaCl, 10 mM KCl, 20 mM Tris-HCl [pH 7.5], 0.1% SDC, 0.1% SDS, protease inhibitor cocktail [Sigma- Aldrich; St. Louis, MO], 50 nM calyculin A [Sigma- Aldrich], 1 mM sodium pyrophosphate, 20 mM sodium fluoride) for 20 min on ice. Cell extracts were cleared by centrifugation at 13,000 rpm for 10 min at 4°C and protein
concentration was measured with the Bio-Rad protein assay reagent (Bio-Rad; Hercules, CA). Lysates were resolved by SDS-PAGE and transferred to nitrocellulose membrane (Bio-Rad). Membranes were blocked in TBST buffer (10 mM Tris-HCl [pH 8], 150 mM NaCl, 0.2% Tween 20) containing 5% (w/v) non-fat dry milk and then incubated with primary antibodies diluted in TBST buffer containing 2% (w/v) non-fat dry milk at 4°C overnight. Membranes were washed in TBST and incubated with horseradish-peroxidase-conjugated secondary antibodies for 1 hr at room temperature. Membranes were washed in TBST and developed using a chemiluminescent substrate (Millipore).
 Anti-Akt, anti-phospho Akt (Ser473), anti-GSK3 and anti-phospho GSK3 (Ser9) antibodies were obtained from Cell Signaling Technology (Danvers, MA). Anti- -actin antibody was purchased from Sigma- Aldrich. Horseradish peroxidase-conjugated anti-mouse and anti- rabbit immunoglobulin (IgG) antibodies were purchased from Millipore.
 Focus formation assay
 Rat-1 cells were infected with viral supernatent and 5 μg/mL polybrene (Millipore). 48 hours after infection, cells were split into 100 mm dishes for focus formation. 8 days later, cells were fixed with ice-cold methanol and stained with crystal violet (0.5% crystal violet, 25% methanol). Images of cells and foci were acquired using an inverted microscope (Eclipse Ti; Nikon, Melville, NY).
 EXAMPLE 2: IDENTIFICATION OF SOMATIC MUTATIONS IN BREAST CANCER
 Whole-exome sequencing was performed on 103 tumor-normal pairs, 54 from Mexico and 49 from Vietnam, targeting 189,980 exons comprising 33 megabases of the genome and with a median of 85.1% of targeted bases covered at least 30-fold across the sample set. This analysis revealed a total of 4,985 candidate somatic substitutions in the target protein-coding regions and the adjacent splice sites, ranging from 14 to 307 putative events in individual samples. These mutations represented 3,153 missense, 1,157 silent, 242 nonsense, 97 splice site, 194 deletions, 110 insertions and 32 other mutations. The total mutation rate was 1.66 per Mb (range 0.47-10.5) with a non-silent mutation rate of 1.27 per Mb (range 0.31-8.05), similar to previous reports in breast carcinoma (Sjoblom, T. et al., Science (2006) 314, 268-274; Wood, L. D. et al., Science (2007) 318, 1108-1113; Shah, S. P. et al., Nature (2009) 809,813; Ding, L. et al., Nature (2010) 464, 999-1005). The mutation rate in breast cancer exceeds that of
hematologic malignancies and prostate cancer but is significantly lower than in lung cancer and melanoma (Kan, Z. et al, Nature (2010) 466, 869-873; Berger, M. F. et al., Nature (2011) 470, 214-220; Chapman, M. A. et al., Nature (2011) 471, 467; Pleasance, E.D. et al., Nature (2010) 463, 191-196; Pleasance , E.D. et al., Nature (2010) 463, 184-190). The most common mutation events observed are C to T transition events in CpG dinucleotides (Figure 1, Figure 8).  We performed validation experiments on 494 candidate mutations (representing all significantly mutated genes and genes in significantly mutated genesets) using a combination of mass-spectrometric genotyping, 454 pyrosequencing, Pacific Biosciences sequencing, and Illumina sequencing of matched formalin-fixed paraffin embedded tissue, and confirmed the presence of 94% of protein- altering point mutations; this validation rate is consistent with previous results that 95% of point mutations can be validated with orthogonal methods (Berger, M. F. et al, Nature (2011) 470, 214-220; Chapman, M. A. et al, Nature (2011) 471, 467). Only 18 of 39 (46%) indels among significantly mutated genes were confirmed.
 EXAMPLE 3: IDENTIFICATION OF SIGNIFICANTLY MUTATED GENES IN THE BREAST CANCER EXOME
 Six genes were found to be mutated with significant recurrence in the 103 whole exome sequenced samples, by analysis with the MutSig algorithm at a False Discovery Rate (FDR) <0.1 after correction for multiple hypothesis testing, manual review of reads, and subsequent orthogonal confirmation of somatic events (Figure 1) (Berger, M. F. et al., Nature (2011) 470, 214-220; Chapman, M. A. et al., Nature (2011) 471, 467). One gene, CBFB is identified for the first time as a significantly mutated gene in breast cancer or any other epithelial cancer, to our knowledge, while the other 5 genes (TP '53, PIK3CA, AKT1, GATA3, and
MAP3K1) have previously been reported as mutated in breast cancer (Wood, L. D. et al., Science (2007) 318, 1108-1113; Kan, Z. et al, Nature (2010) 466, 869-873; Usary, J. et al., Oncogene (2004) 23, 7669-7678). This significantly mutated genes list, as any list produced by a statistical method, is likely incomplete and reflects the statistical power of our cohort size— larger sample sets will provide further statistical power.
 Somatic mutations in TP53 and PIK3CA were each present in 27% of samples, consistent with published frequencies (Figure 1) (Kan, Z. et al, Nature (2010) 466, 869-873; Bachman, K.E. et al., Cancer Biol. Ther. (2004) 3, 772-775). TP53 mutations occur in samples with a higher mutation rate (T-test p = 0.0079 comparing samples with mutation rates greater than or less than the median 1.66 mutations/Mb) and were distributed across the gene in sites reported in COSMIC (http://ww , w.sanger,ac.ukgenetics/CGP/cosmic/). Also, using the
ABSOLUTE algorithm for determining allele- pecific copy number we observed that 21 of 31 TP53 mutations were homozygous. PIK3CA mutations were clustered in the helical (amino acids 542/545; 40%) and kinase domains (amino acid 1047; 47%) (Bachman, K.E. et al., Cancer Biol. Ther. (2004) 3, 772-775). Six samples harboured the AKT1 E17K mutation that alters the pleckstrin-homology (PH) domain and leads to activation of the kinase (Carpten, J.D. et al., Nature (2007) 448, 439-444). AKT1 and PIK3CA mutations, which activate the
phosphatidylinositol- 3 -kinase (PI3K) pathway, were mutually exclusive in our dataset.
MAP3K1, recently reported as mutated in ER + breast cancers, harboured 5 mutations in 3 patients with ER + disease, and followed a pattern consistent with positive selection for recessive inactivation of the gene (Kan, Z. et al, Nature (2010) 466, 869-873). In total, two frameshift, two nonsense, and one missense mutation combined with a homozygous deletion spanning the coding region were observed. Although the point mutations appeared to be heterozygous by copy-number analysis, two patients harboured dual mutations, consistent with compound heterozygous inactivation, although confirmatory phasing data were not available. The GATA3 transcription factor gene harboured mutations in 4 patients with luminal tumors, including 3 novel frameshift mutations near the 3'-end of the coding sequence. We also identified one previously described splice-site mutation that disrupts zinc-finger domains in Gata3 required for DNA binding (Usary, J. et al., Oncogene (2004) 23, 7669-7678).
 CBFB, encoding the core-binding-factor beta subunit, was mutated in 4 ER + samples, with one nonsense mutation and three truncating frameshift mutations (Figure 2A). CBFB somatic mutations have been noted in isolated cases of breast cancer (Sjoblom, T. et al., Science (2006) 314, 268-274; Kan, Z. et al., Nature (2010) 466, 869-873). This is the first report of these mutations recurring at a significant rate above background; the sample size is not sufficient to determine whether these mutations are specific for ER + subtypes. CBFB encodes the non-DNA binding component of a heterodimeric protein complex, together with the DNA-binding RUNX proteins encoded by RUNX1, RUNX2, and RUNX3. Copy-number analysis, using the
ABSOLUTE algorithm, provides further evidence for loss of function of the Runxl/Cbfb complex in breast cancer: the cases with CBFB mutations appear to have hemizygous deletions of one parental allele while two additional cases harbour homozygous deletions of RUNX1 (Figure 2B, C, Figure 10). Oncogenic rearrangements of RUNX1 or CBFB are common in acute myeloid leukemia (including the CBFB-MYH11 translocation believed to have dominant negative function) (Cameron, E.R. et al., Oncogene (2004) 23, 4308-4314; Shigesada, K. et al., Oncogene (2004) 23, 4297-4307). This is to our knowledge the first report of inactivation of this transcription factor complex in epithelial cancers.  Significance analysis restricted to somatic mutations in genes reported in COSMIC revealed 3 significantly mutated genes, including PIK3CA, TP53, and ERBB2, the latter below the significance threshold in the complete analysis. ERBB2 contained somatic mutations in three samples, with two being identical S310F mutations (these two samples are distinct based on their germline and somatic genotypes.) The S310F mutation can activate ERBB2 and is transforming in vitro. Neither sample with the S310F activating mutation has ERBB2 amplification (Figure 1 1 ). The two samples belong to the Her2-enriched and Luminal B subtypes, which typically have ERBB2 amplification; this supports the notion that the observed mutations have a driving role in these tumors (Kan, Z. et al., Nature (2010) 466, 869-873; Stephans, P. et al., Nature (2004) 431, 525-526).
 EXAMPLE 4: IDENTIFICATION AND CHARACTERIZATION OF A AKT3
 To identify candidate genomic rearrangements, we applied the dRanger algorithm to the 22 cases with paired tumor/normal whole-genome sequencing data. The rate of
rearrangements ranged from a median of 30 rearrangements per sample in the Luminal A subtype (range 0-218) to the basal-like and Her2-enriched subtypes with a median of 237 and 246 rearrangements, respectively (Figure 12) (Berger, M.R. et al., Nature (2011) 470, 214-220; Chapman, M. A. et al., Nature (2011) 471, 467); the rates are similar to a recent report (Stephans, P.J. et al., Nature (2009) 462, 1005-1010). We performed PCR amplification on a subset of the candidate rearrangements (confirmed 89 of 165 events (54%). No rearrangement was seen in more than one sample. In addition, we did not identify rearrangements previously observed by DNA sequencing nor by cDNA-sequencing, including MAST and Notch family-gene fusions (Stephans, P.J. et al., Nature (2009) 462, 1005-1010; Robinson, D.R. et al., Nature Medicine (2011) 17, 1646-1651).
 The discovery of recurrent driver rearrangements in other epithelial cancers led to a closer examination of the list of confirmed rearrangements (Soda, M. et al., Nature (2007) 448, 561-566; Tomlins, S.A. et al., Science (2005) 310, 644-648). In a triple-negative, basal-like subtype tumor, we observed a rearrangement between the genes MAGI3 (membrane associated guanylate kinase, WW and PDZ domain containing 3) on chromosome lp and AKT3 (v-akt murine thymoma viral oncogene homolog 3) on chromosome lq, resulting in a balanced translocation from intron 9 in MAGI3 to intron 1 of AKT3 (Figure 3A). The novel fusion genes were confirmed in tumor DNA by sequencing the product of PCR amplification (Figure 3B). The MAGI3 disruption is complemented by a hemizygous deletion of the other allele (Figure 13A). The expression levels of individual exons of MAGI3 and AKT3 correspond to the predicted 5'- MAGI3-AKT3-3 ' fusion (Figure 13B), with this sample having the highest AKT3 expression in the dataset. Expression of the fusion gene was confirmed in the tumor sample by PCR amplification of the cDNA (Figure 3B).
 The rearrangement produces an in-frame fusion gene with a predicted Magi3-Akt3 fusion protein that combines Magi3 lacking the second PDZ domain, reported to bind to Pten and be required for Pten's inhibitory effect on the PI3K pathway, together with an Akt3 region that retains an intact kinase domain but has a disruption of the pleckstrin homology domain prior to the glutamate at position 17 (Figure 3C) (Wu, Y. et al., J. of Biol. Chem. (2000) 275, 21477- 21485). AKT3 shares significant homology to AKT1 and is reported to be the dominant AKT family member expressed in hormone receptor negative breast cancers (Nakatani, K. et al., J. Biol. Chem. (1999) 274, 21528-21532) . Together, the MAGI3-AKT3 translocation and deletion of MAGI3 could result in the combined loss of function of a tumor suppressor gene (PTEN) and activation of an oncogene (AKT3).
 To evaluate oncogenic activity of the MAGI3-AKT3 fusion, we expressed the fusion gene ectopically in ZR-75 cells. The Magi3-Akt3 fusion protein is constitutively phosphorylated at serine 473 in the Akt3 kinase domain (numbered according to the wild-type protein) in the absence of growth factors (Figure 3D); ectopically expressed Aktl with an engineered E17K mutation is likewise constitutively phosphorylated (Figure 3D), as previously reported (Carpten, J.D. et al., Nature (2007) 448, 439-444). Constitutive activation of the Magi3-Akt3 kinase in turn activates downstream pathways as demonstrated by phosphorylation of GSK3 , an Akt substrate (Figure 3D). Phosphorylation of GSK3 by the MAGI3-AKT3 fusion can be inhibited with an ATP-competitive small molecule Akt inhibitor, GSK-690693, but not with an allosteric Akt inhibitor, MK-2206, that interacts with the PH domain of Akt (Figure 3D). Over-expression of the MAGI3-AKT3 fusion gene in Rat-1 fibroblast cell lines led to loss of contact inhibition and focus formation (Figure 3E).
 We screened 235 additional breast cancer samples for the presence of the 5'- AG/3- AKT3-3 ' fusion event by RT-PCR of cDNA followed by Sanger sequencing of breakpoints. The fusion was present in 8 of the 235 samples, including 5 out of 72 triple negative (ER7PR7Her2 ~ ) samples, (Figure 14).
 The power provided by whole-genome and whole-exome sequencing of a relatively large and diverse breast cancer sample set has enabled several significant discoveries including the identification of recurrent inactivating mutations in CBFB and of a recurrent translocation of MAGI3-AKT3. The mutations in CBFB, RUNX1 and GAT A3 suggest the importance of understanding epithelial cell differentiation and its regulatory transcription factors in breast cancer pathogenesis. The recurrent genomic fusion involving AKT3 suggests that the use of ATP-competitive Akt inhibitors should be evaluated in clinical trials for the treatment of fusion- positive triple-negative breast cancers, a subtype where limited therapeutic options exist beyond systemic cytotoxic chemotherapy.
 EXAMPLE 5: INTEGRATIVE MUTATION AND COPY ANALYSIS DEFINES THERAPEUTIC TARGETS WITHIN DISRUPTED PATHWAYS
 Summing the frequency of all 8 significantly- mutated genes across the 103 exome- sequenced breast cancers reveals a total of 76 mutations in 62 cases (Figure 4A). Among these are PIK3CA and AKT1, encoding components of the insulin signaling pathway, mutated in 33 mutually exclusive cases. These kinases along with the tumor suppressor PTEN, are central to the PI3K pathway and its role in cell proliferation, survival and metabolism (Figure 4B). Small- molecule inhibitors of PI3K and AKT have shown success in pre-clinical models of breast cancer and are currently the focus of early stage clinical trials (Pal, S.K. et al., Expert Opin. Investig. Drugs (2010) 19, 1355-1366; Wallin, J. J. et al., Sci. Transl. Med. (2010) 2, 48ra66). Tumors with the MAGFAKT3 fusion represent additional cases for evaluation with these inhibitors.
 Upstream of the PI3K pathway are many potentially draggable membrane- associated receptor tyrosine kinases (RTK) in which cancer-driving genetic alterations are reported frequently in epithelial tumors (Figure 4B). Expression of the RTK ERBB2 driven by gene amplification is a hallmark of breast cancer. Almost 1/3 of tumors in our study harbor ERBB2 amplifications and represent candidates to benefit from the monoclonal antibody trastuzumab shown to dramatically improve patient outcomes (Piccart-Gebhart et al. N Engl J Med (2005) 353,1659-1672). Additionally, cases with activating ERBB2 mutations, as identified in our study, should be evaluated for response to inhibitors like lapatinib (Figure 4A).  A closer look at the mutation and copy number data reveals additional RTKs with therapeutic potential. Here we highlight some RTKs affecting more than one individual case in our study. FGFR1 is significantly- amplified in our study occurring in 14 cases (Figure 4B). FGFR expression has been shown to accelerate tumorigenesis in a mouse model of breast cancer (Pond, A.C. et al., Cancer Research (2010) 70, 4868-4879). Inhibition of FGFR1 amplification- induced proliferation has been demonstrated in breast cancer and squamous cell lung cancer cell lines using the pan-FGFR inhibitor PD 173074 (Kozicizak, M. et al., Oncogene (2004) 23, 3501- 3508; Weiss, J. et al., Sci. Transl. Med. (2010) 2, 62ra93). While neither significantly mutated nor amplified in our study, a germline polymorphism in FGFR2 has been shown previously to be associated with breast cancer risk (Easton, D.F. et al., Nature (2007) 447, 1087-1093). We have also identified 3 somatic mutations in EGFR (Figure 4A). Lung adenocarcinomas with EGFR kinase domain mutations respond to the small molecule inhibitors erlotinib and gefitinib
(Greulich, H. et al., PLoS Med. (2005) 2, e313). Overall, 64% of cases in our study have alterations in one or more of these potentially draggable tyrosine kinases.
 A search for mutated genesets in the MSigDB Canonical Pathways database using all mutations in the sequenced samples revealed 147 candidate pathways out of 616 KEGG pathways. Many of these genesets are overlapping and heavily driven by mutations in TP53 and PIK3CA (Chapman, M.A. et al., Nature (2011) 471, 467). Beyond PI3K signaling described above, genesets frequently altered in breast cancer include genes involved in DNA repair and cell cycle control (Figure 4B). TP 53 is a central player in these pathways and is mutated or deleted in 30% of cases. Genes like BRCA1 , BRCA2, and ATM have important roles in DNA repair. Modifying these defects in DNA repair underlie the recent clinical success of PARP inhibitors (Fong, P.C. et al., N. Engl. J. Med. (2009) 361, 123-134). DNA repair is closely linked to cell cycle regulation, another target of genetic alteration in breast cancer. CCND1 promoting cell proliferation is the second most significant amplification event in breast cancer after ERBB2. CCND1 amplification and focal deletions involving the cyclin dependant kinase (CDK) inhibitor CDKN2A, lead to activation of the CDKs and subsequent progression through the cell cycle. CDK inhibitors may have a role in the treatment of tumors harboring these alterations (Swanton, C. et al., Lancet Oncol. (2004) 5, 27-36).
 Finally, other significant alterations involving the CBFB-RUNX1 axis and MLL genes, while not yet annotated as part of gene relationships depicted in Figure 4B, are likely to form the scaffold for other breast cancer pathways that have yet to be identified. The MLL genes via their effect on Hox expression can influence many aspects of the cell from receptor signaling to control of apoptosis and proliferation (Shah, N. et al., Nature Publishing Group (2010) 10, 361-371). Histone modification or targeting specific Hox genes are potential candidates for alternative targeted therapies. In total 81 (79%) cases in our study are potential candidates for existing or putative targeted therapies.
 EXAMPLE 6: COMPARISON OF THE RELATIVE UTILITY OF MUTATION DETECTION USING WHOLE-GENOME AND WHOLE-EXOME SEQUENCING APPROACHES
 There was high concordance between whole-genome and whole-exome sequencing for mutation detection at higher allelic fraction (Figure 9A). Whole-exome sequencing is more sensitive at detecting mutations at lower allelic fraction (Figure 9B). Mutations at a very low allelic fraction were detected only by whole-genome sequencing and likely represent mutation calling artifacts in regions of minimal sequence coverage.
 We looked for evidence of germline mutations in four breast-cancer susceptibility genes (BRCA1, BRCA2, TP53, and PTEN) and found a number of protein- altering variants, which we compared to an online curated list of breast cancer inherited variants and their functional significance^. Two Vietnamese cases (one Luminal B, one basal subtype) carried identical germline BRCA1 R1772* mutations. Structure-altering germline BRCA2 mutations were seen in 4 cases, 3 Vietnamese cases with distinct frameshift mutations and a Mexican case with a K3326* nonsense mutation. Two somatic mutations were seen in each BRCA1 and BRCA2.
 We also used the whole-genome sequence from the 22 samples to look for significantly mutated non-coding regions with regulatory potential. We defined regions as previously described in detail in Chapman, M.A. et al., Nature (2011) and subjected them to the same permutation analysis used for exonic regions. We found only 3 such regions, altered in a maximum of 5 cases (FDR <0.1, manual review of mutations to eliminate artifacts,); the functional significance of these regions requires further evaluation in a larger sample set. We also looked for recurrent mutations at binding sites for ER-alpha (6024 sites) and ER-beta (9702 sites), annotated based on ChlP-Seq datal7. Only one site was mutated twice. The site is flanked by the kallikrein genes KLK13 and KLK14 on Chromosome 19, and it is annotated as binding both ER alpha and ER beta. Kallikreins are serine proteases whose expression is regulated by steroid hormones. KLK13 expression has been shown to be a favorable prognostic marker for breast carcinoma.
 The power to detect a variant depends on the allelic fraction and local depth of coverage. For each ex on of the significantly mutated genes in each sample, we calculated the allelic fraction assuming a single mutated copy taking into account the local copy number of the exon and the purity of the sample. The average local depth of coverage was computed directly for each sample-ex on. Using this allelic fraction and average local depth, we calculated the power to have observed a clonal mutation in a single copy (Figure 16). Power was not uniform across samples and genomic regions. Some genomics regions have suboptimal coverage often due to failed hybrid-capture, GC-bias in sequencing, or lack of unique alignment to the genome. These regions are usually located at the 5'- and 3'- ends of genes. In our 6 significantly mutated genes, the power to detect mutations was not affected by the tumor purity in regions with adequate sequencing coverage (Figure 16). In regions with intermediate coverage, power is reduced in samples with lower purity. Therefore our observed frequency of mutations represents a lower-bound of the true mutation frequency.
Table 1 : Sample collections successfully completed sequencing and analysis
Patients N = 56 N = 52
Median Age (Range) 54 (37-92) 48 (31 -81 )
Source of Normal DNA Blood Adjacent Tissue
Pathology Subtype (Percent)
Ductal 46 (82%) 41 (79%)
Lobular 4 (7%) 0 (0%)
DCIS 0 (0%) 9 (17%)
Other 6 (1 1 %) * 2 (4%)§
0 0 (0%) 9 (17%)
I 8 (14%) 3 (6%)
II 36 (64%) 31 (60%)
III 12 (21 %) 9 (17%)
Expression Subtype (Percent)^
Luminal A 24 (43%) 14 (27%)
Luminal B 13 (23%) 9 (17%)
Her2 9 (16%) 12 (23%)
Basal 5 (9%) 8 (15%)
Unknown 2 (4%) 3 (6%)
Normal Like 3 (5%) 6 (1 1 %)
Includes tubular carcinoma, medullary carcinoma, mucinous carcinoma, and mixed carcinoma (3)
§ Includes mucinous carcinoma (2)
1f Based on PAM-50 classification
DCIS = Ductal carcinoma in situ