WO2006048270A2 - Methods of detecting leukemia and its subtypes - Google Patents

Abstract

The present invention relates to rapid and reliable approaches to detecting, diagnosing, and subtyping leukemia by gene expression profiling. In addition to these methods leukemia, the invention also provides related kits and systems.

Description

METHODS OF DETECTING LEUKEMIA

FIELD OF THE INVENTION

The present invention relates to the detection of leukemia and accordingly, provides diagnostic and/or prognostic information in certain embodiments.

BACKGROUND OF THE INVENTION Leukemias are generally classified into four different groups or types: acute myeloid (AML), acute lymphatic (ALL), chronic myeloid (CML) and chronic lymphatic leukemia (CLL). Within these groups, several subcategories or subtypes can be identified using various approaches. These different subcategories of leukemia are associated with varying clinical outcomes and therefore can serve as guides to the selection of different treatment strategies. The importance of highly specific classification may be illustrated for AML as a very heterogeneous group of diseases. Effort has been aimed at identifying biological entities and to distinguish and classify subgroups of AML that are associated with, e.g., favorable, intermediate or unfavorable prognoses. In 1976, for example, the FAB classification was proposed by the French- American-British co-operative group that utilizes cytomorphology and cytochemistry to separate AML subgroups according to the morphological appearance of blasts in the blood and bone marrow. In addition, genetic abnormalities occurring in leukemic blasts were recognized as having a major impact on the morphological picture and on prognosis. As a consequence, the karyotype of leukemic blasts is commonly used as an independent prognostic factor regarding response to therapy as well as survival.

A combination of methods is typically used to obtain the diagnostic information in leukemia. To illustrate, the analysis of the morphology and cytochemistry of bone marrow blasts and peripheral blood cells is commonly used to establish a diagnosis. In some cases, for example, immunophenotyping is also utilized to separate very undifferentiated AML from acute lymphoblastic leukemia and CLL. In certain instances, leukemia subtypes can be diagnosed by cytomorphology alone, but this typically requires that an expert review sample smears. However, genetic analysis based on, e.g., chromosome analysis, fluorescence in situ hybridization (FISH), or reverse transcription PCR (RT-PCR) and immunophenotyping is also generally used to accurately assign cases to the correct category. An aim of these techniques, aside from diagnosis, is to determine the prognosis of the leukemia under consideration. One disadvantage of these methods, however, is that viable cells are generally necessary, as the cells used for genetic analysis need to divide in vitro in order to obtain metaphases for the analysis. Another exemplary problem is the long lag period (e.g., 72 hours) that typically occurs between the receipt of the materials to be analyzed in the laboratory and the generation of results. Furthermore, great experience in preparing chromosomes and analyzing karyotypes is generally needed to obtain correct results in most cases. Using these techniques in combination, hematological malignancies in a first approach can be separated into CML, CLL, ALL, and AML. Within the latter three disease entities, several prognostically relevant subtypes have been identified. As a second approach this further sub-classification is based mainly on genetic abnormalities of the leukemic blasts and clearly is associated with different prognoses.

The sub-classification of leukemias is used increasingly as a guide to the selection of appropriate therapies. The development of new, specific drugs and treatment approaches often includes the identification of specific subtypes that may benefit from a distinct therapeutic protocol and thus, improve the outcomes of distinct subsets of leukemia. For example, the therapeutic drug 25 (STI571) inhibits the CML specific chimeric tyrosine kinase BCR-ABL generated from the genetic defect observed in CML, the BCR-ABL-rearrangement due to the translocation between chromosomes 3 and 22 (t(9;22) (q34; qll)). In patients treated with this new drug, the therapy response is dramatically higher as compared to all other drugs that had been previously been used. Another example is a subtype of acute myeloid leukemia, AML M3 and its 30 variant M3v, which both include the karyotype t[l 5;17)(q22; qll- 12). The introduction of all-trans retinoic acid (ATRA) has improved the outcome in this subgroup of patient from about 50% to 85% long-term survivors. Accordingly, the rapid and accurate identification of distinct leukemia subtypes is of consequence to further drug development in addition to diagnostics and prognostics. According to Golub et al. (Science, 1999, 286, 531-7, which is incorporated by reference), gene expression profiles can be used for class prediction and discriminating AML from ALL samples. However, for the analysis of acute leukemias the selection of the two different subgroups was performed using exclusively morphologic-phenotypical criteria. This was only descriptive and does not provide deeper insights into the pathogenesis or the underlying biology of the leukemia. The approach reproduces only very basic knowledge of cytomorphology and intends to differentiate classes. The data is not sufficient to predict prognostically relevant cytogenetic aberrations.

SUMMARY OF THE INVENTION

The present invention relates to rapid and reliable approach.es to detecting and subtyping leukemia. Aside from providing diagnostic information to patients, this information can also assist in selecting appropriate therapies and in prognostication. In some embodiments, these methods include profiling the expression of selected populations of genes using oligonucleotide arrays, such as DNA microarrays. In addition to methods, the invention also provides related kits and systems.

In one aspect, the invention provides a method of detecting leukemia. The method includes detecting an expression of at least one gene population of at least one cell, which gene population comprises at least one set of genes listed in one or more of Tables I-XII, thereby detecting leukemia. Typically, the cell is derived from at least one subject.

In some embodiments, the method also includes subtyping the leukemia. In certain embodiments, for example, the method includes correlating a detected expression of one or more sets of genes listed in at least one of Tables I- VI with the cell being an acute myeloid leukemia (AML) cell. In some embodiments, the method includes correlating a detected expression of a set of genes listed in Table I with the cell being an AML cell with a t(15;17). In some embodiments, the method includes correlating a detected expression of a set of genes listed in Table II with the cell being an AML cell with a t(8;21). Further, the method optionally includes correlating a detected expression of a set of genes listed in Table III with the cell being an AML cell with an inv(16). In certain embodiments, the method includes correlating a detected expression of a set of genes listed in Table IV with the cell being an AML cell with a normal karyotype or another cytogenetic abnormality. In some embodiments, the method includes correlating a detected expression of a set of genes listed in Table V with the cell being an AML cell with a 1 Iq23/MLL rearrangement. Optionally, the method includes correlating a detected expression of a set of genes listed in Table VI with the cell being an AML cell with a complex aberrant karyotype.

In some embodiments, the method includes correlating a detected expression of one or more sets of genes listed in at least one of Tables VII-X with the cell being an acute lymphoblastic leukemia (ALL) cell. In certain embodiments, the method includes correlating a detected expression of a set of genes listed in Table VII with the cell being a Pro-B-ALL/t(llq23) cell. In some embodiments, the method includes correlating a detected expression of a set of genes listed in Table VIII with the cell being a mature B-ALL/t(8;14) cell. Optionally, the method includes correlating a detected expression of a set of genes listed in Table IX with the cell being a c-ALL/Pre-B-ALL cell with or without t(9;22). In some embodiments, the method includes correlating a detected expression of a set of genes listed in Table X with the cell being a T-ALL cell.

In addition, the method optionally includes correlating a detected expression of a set of genes listed in Table XI with the cell being a chronic myeloid leukemia (CML) cell. In some embodiments, the method includes correlating a detected expression of a set of genes listed in Table XII with the cell being a chronic lymphatic leukemia (CLL) cell.

Expression levels are detected using essentially any gene expression profiling technique. In some embodiments, for example, the method includes measuring expression levels of genes of the cell on at least one probe array that comprises oligonucleotides with nucleotide sequences that correspond to at least subsequences of one or more sets of genes listed in at least one of Tables I-XII. In certain embodiments, the detection of the expression of the gene population comprises hybridizing transcribed polynucleotides or portions thereof to complementary polynucleotides or portions thereof. For example, the transcribed polynucleotides (e.g., mRNAs, cDNAs, etc.) or portions thereof are hybridized under stringent hybridization conditions in certain embodiments. In some embodiments, the expression of the gene population is detected by amplifying nucleic acid sequences associated with the genes to produce amplicons and detecting the amplicons. In these embodiments, the amplicons are typically detected using a process that comprises one or more of: hybridizing the amplicons to an oligonucleotide array, digesting the amplicons with a restriction enzyme, or real-time polymerase chain reaction (PCR) analysis. Optionally, the detection of the expression of the gene population comprises measuring quantities of transcribed polynucleotides (e.g., mRNAs, cDNAs, etc.) or portions thereof expressed or derived from the genes. In some embodiments, the detection of the expression of the gene population comprises contacting polynucleotides and/or polypeptides from the cell with compounds (e.g., aptamers, antibodies or fragments thereof) that specifically bind the polynucleotides and/or polypeptides. Typically, the detection of the expression of the gene population comprises using an array, a robotics system, and/or a microfluidic device.

In another aspect, the invention provides a method of differentiating between leukemia and non-leukemia cells. The method includes measuring expression levels of at least one gene population of at least one cell to produce expression data. The method also includes correlating the expression data with at least one set of genes listed in one or more of Tables I-XII, thereby detecting the leukemia cell, or with a set of genes listed in Tables XIII, thereby detecting the non-leukemia cell.

In still another aspect, the invention provides a kit that includes one or more probe biomolecules (e.g., polynucleotides, polypeptides, etc.) corresponding to one or more genes or portions thereof listed in one or more of Tables I-XIII. The kit also includes instructions for correlating detected expression levels of one or more polynucleotides or polypeptides in at least one target cell, which polynucleotides or polypeptides detectably bind to one or more of the probe biomolecules, with the target cell being a leukemia cell or a non-leukemia cell. In certain embodiments, at least one solid support comprises the probe biomolecules, e.g., in the form of an oligonucleotide array. In some embodiments, the kit includes one or more additional reagents to perform real-time PCR analyses. In another aspect, the invention provides a system that includes one or more probe biomolecules corresponding to one or more genes or portions thereof listed in one or more of Tables I-XIII. The system also includes at least one reference data bank for correlating detected expression levels of polynucleotides or polypeptides in target cells, which polynucleotides or polypeptides detectably bind to one or more of the probe biomolecules, with the target cell being a leukemia cell or a non- leukemia cell. In some embodiments, the system includes one or more additional reagents and/or components to perform real-time PCR analyses. Typically, the reference data bank is produced by: (a) compiling a gene expression profile of a patient sample by determining the expression level at least one of the markers, and

(b) classifying the gene expression profile using a machine learning algorithm. The machine learning algorithm is typically selected from, e.g., a weighted voting algorithm, a K-nearest neighbors algorithm, a decision tree induction algorithm, a support vector machine, a feed-forward neural network, or the like. In yet another aspect, the invention provides a method of producing a reference data bank for distinguishing leukemia and non-leukemia cells from one another. The method includes (a) compiling a gene expression profile of a patient sample by determining the expression level of genes listed in one or more of Tables I-XIII. The method also includes (b) classifying the gene expression profile using a machine learning algorithm.

BRIEF DESCRIPTION OF THE DRAWINGS Figure 1 shows a hierarchical cluster analysis of n=937 samples. The analysis of the n=937 samples (columns) used a set of 1,019 differentially expressed genes (rows). The normalized expression value for each gene is coded by color (standard deviation from mean). Red cells indicate high expression and green cells indicate low expression. The major leukemia types are separated by bars. For each of the 13 classes the top 100 differentially expressed genes, according to t-test statistic, were used. Of the 1,300 genes, 281 were repeatedly identified as important diagnostic markers and were overlapping between the respective top 100 gene lists. Thus, this results in a list of 1 ,019 non-overlapping genes. Figure 2 is a plot of a three-dimensional principal component analysis (PCA) that shows the distinction between precursor B-ALL and T-ALL. In the three- dimensional PCA n=l 14 ALL samples were projected into the feature space consisting of a combination of the top 100 differentially expressed genes when comparing precursor B-ALL vs. the other 12 classes or T-ALL vs. the other 12 classes. Data points with similar characteristics will cluster together. As shown, each patient's expression pattern is represented by a single color-coded sphere. The respective label, i.e. precursor B-ALL, or T-ALL was unknown to the algorithm. The labels and coloring of the classes were added after the analysis for means for better visualisation. The n=82 precursor B-ALL samples are colored blue and include 42 c-ALL/Pre-B-ALL with t(9;22), and 40 c-ALL/Pre-B-ALL without t(9;22). The n=32 T-ALL samples are colored turquoise.

Figure 3 is a plot of a three-dimensional principal component analysis (PCA) that shows the distinction between c-ALL/Pre-B-ALL with t(9;22) and CML. In the three-dimensional PCA n=l 17 samples were projected into the feature space consisting of a combination of the top 100 differentially expressed genes when comparing c-ALL/Pre-B-ALL with or without t(9;22) samples vs. the other 12 classes and CML vs. the other 12 classes. Data points with similar characteristics will cluster together. Here, each patient's expression pattern is represented by a single color-coded sphere. The respective label, i.e. precursor B-ALL, or T-ALL was unknown to the algorithm. The labels and coloring of the classes were added after the analysis for means for better visualisation. The n=42 c-ALL/Pre-B-ALL with t(9;22) samples are colored red, the n=75 CML samples are colored green, respectively. Figure 4 (supporting gene list for figure 4 is contained in table XIV) show the identification of c-ALL/Pre-B-ALL samples with or without t(9;22). Analysis of^" n=82 c-ALL/Pre-B-ALL samples based on a supervised identification of differentially expressed genes between 42 cases demonstrating a t(9;22), colored in red, and 40 cases without t(9;22), colored in blue, respectively. The labels and coloring of the classes were added after the analysis for means for better visualisation. (Panel A) In the hierarchical cluster analysis the normalized expression value for each gene (given in rows) is coded by color (standard deviation from mean). Red cells indicate high expression and green cells indicate low expression. (Panel B) In the three-dimensional principal component analysis (PCA) the c-ALL/Pre-B-ALL samples were projected into the feature space consisting of the top 100 differentially expressed genes when comparing t(9;22) positive vs. negative cases. Data points with similar characteristics will cluster together. Here, each patient's expression pattern is represented by a single color- coded sphere.

Figure 5 shows the distinction between immature and cortical T-ALL samples (supporting gene list for figure 5 is contained in table XV). Analysis of n=32 T-

ALL samples based on a supervised identification of differentially expressed genes between 12 immature T-ALL samples, colored in orange, and 20 cortical T-ALL samples, colored in purple, respectively. The labels and coloring of the classes were added after the analysis for means for better visualisation. (Panel A) In the hierarchical cluster analysis the normalized expression value for each gene (given in rows) is coded by color (standard deviation from mean). Red cells indicate high, expression and green cells indicate low expression. (Panel B) In the three- dimensional principal component analysis (PCA) the T-ALL samples were projected into the feature space consisting of the top 100 differentially expressed genes when comparing immature vs. cortical T-ALL cases. Data points with similar characteristics will cluster together. Here, each patient's expression pattern is represented by a single color-coded sphere.

Figure 6 shows plots of the event- free survival (left) and overall survival (right) in cytogenetically defined subgroups of AML. Patients analyzed in the present microarray study are shown. Patients with complex aberrant karyotypes and those with AML and t(l Iq23)/MLL have the worst prognosis, patients with AML and t(15;17), t(8;21), or inv(16) have a relatively good prognosis. Importantly, there is no difference with regard to prognosis between patients with a normal karyotype and those with other karyotype, i.e. those not in the before-mentioned subgroups. Figure 7 shows plots of comparisons of gene expression assessed by both microarray and Mirco Fluidic Cards (MFC). DETAILED DESCRIPTION

DEFINITIONS

Before describing the present invention in detail, it is to be understood that this invention is not limited to particular embodiments. It is also to be understood that the terminology used herein is for the purpose of describing particular embodiments only, and is not intended to be limiting. Units, prefixes, and symbols are denoted in the forms suggested by the International System of Units (SI), unless specified otherwise. Numeric ranges are inclusive of the numbers defining the range. As used in this specification and the appended claims, the singular ^■£< brms "a", "an" and "the" also include plural referents unless the context clearly dictates otherwise. Further, unless defined otherwise, all technical and scientific teαrms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the invention pertains. The terms defined below, and grammatical variants thereof, are more fully defined by reference to the specification in its entirety.

"Ilq23/MLL" refers to a 1 Iq23 rearrangement of the human MLL gene.

An "antibody" refers to a polypeptide substantially encoded by at least one immunoglobulin gene or fragments of at least one immunoglobulin gene, which can participate in specific binding with a ligand. The term "antibody" includes polyclonal and monoclonal antibodies and biologically active fragments trxereof including among other possibilities "univalent" antibodies (Glennie et al. Q 1982) Nature 295:712); Fab proteins including Fab' and F(ab')₂ fragments whether covalently or non-covalently aggregated; light or heavy chains alone, typically variable heavy and light chain regions (V_H and V_L regions), and more typically including the hypervariable regions (otherwise known as the complementa-rity determining regions (CDRs) of the V_H and V_L regions); F_c proteins; "hybrid" antibodies capable of binding more than one antigen; constant- variable region chimeras; "composite" immunoglobulins with heavy and light chains of different origins; "altered" antibodies with improved specificity and other characteristics as prepared by standard recombinant techniques, by mutagenic techniques, oar other directed evolutionary techniques known in the art. Derivatives of antibodies include scFvs, chimeric and humanized antibodies. See, e.g., Harlow and Lane, Antibodies, a laboratory manual, CSH Press (1988), which is incorporated by reference. For the detection of polypeptides using antibodies or fragments thereof, there are a variety of methods known to a person skilled in the art, which are optionally utilized. Examples include immunoprecipitations, Western blottings, Enzyme-linked immuno sorbent assays (ELISA), radioimmunoassays (RIA), dissociation-enhanced lanthanide fluoro immuno assays (DELFIA), scintillation proximity assays (SPA). To facilitate detection, an antibody is typically labeled by one or more of the labels described herein or otherwise known to persons skilled, in the art. In general, an "array" or "microarray" refers to a linear or two- or three dimensional arrangement of preferably discrete nucleic acid or polypeptide probes which comprises an intentionally created collection of nucleic acid or polypeptide probes of any length spotted onto a substrate/solid support. The person skilled in the art knows a collection of nucleic acids or polypeptide spotted onto a substrate/solid support also under the term "array". As also known to the person skilled in the art, a microarray usually refers to a miniaturized array arrangement, with the probes being attached to a density of at least about 10, 20, 50, 100 nucleic acid molecules referring to different or the same genes per cm . Furthermore, where appropriate an array can be referred to as "gene chip". The array itself can have different formats, e.g., libraries of soluble probes or libraries of probes tethered to resin beads, silica chips, or other solid supports.

A "biomolecule" refers to an organic molecule typically made by living organisms. This includes, for example, molecules comprising nucleotides, amino acids, sugars, fatty acids, steroids, nucleic acids, polypeptides, peptides, peptide fragments, carbohydrates, lipids, and combinations of these (e.g., glycoproteins, ribonucleoproteins, lipoproteins, or the like).

"Complementary" and "complementarity", respectively, can be described by the percentage, i.e., proportion, of nucleotides that can form base pairs between two polynucleotide strands or within a specific region or domain of the two strands. Generally, complementary nucleotides are, according to the base pairing rules, adenine and thymine (or adenine and uracil), and cytosine and guanine. Complementarity may be partial, in which only some of the nucleic acids' bases are matched according to the base pairing rules. Or, there may be a complete or total complementarity between the nucleic acids. The degree of complementarity between nucleic acid strands has effects on the efficiency and strength of hybridization between mαcleic acid strands.

Two nucleic acid strands are considered to be 100% complementary to each other over a defined length if in a defined region all adenines of a first strand can pair with a thymine (or an uracil) of a second strand, all guanines of a first strand can pair with a cytosine of a second strand, all thymine (or uracils) of a first strand can pair with an adenine of a second strand, and all cytosines of a first strand can pair with a guanine of a second strand, and vice versa. According to the present invention, the degree of^* complementarity is determined over a stretch of about 20 or 25 nucleotides, i.e., a 60% complementarity means that within a region of 20 nucleotides of two nucleic acid strands 12 nucleotides of the first strand can base pair with 12 nucleotides of the second strand according to the above base pairing rules, either as a stretch of 12 contiguous nucleotides or interspersed by non-pairing nucleotides, when the two strands are attached to each other over the region of 20 nucleotides. The degree of complementarity can range from at least about 50% to full, i.e., 100% complementarity. Two single nucleic acid strands are said to be "substantially complementary" when they are at least about 80% complementary, and more typically about 90% complementary or higher. For carrying out the methods of present invention substantial complementarity is generally utilized.

Two nucleic acids "correspond" when they have substantially identical or complementary sequences, when one nucleic acid is a subsequence of the other, or when one sequence is derived naturally or artificially from the other.

The term "detectably bind" refers to binding between at least two molecular species (e.g., a probe biomolecule acid and a target polynucleotide, an antibody and a target polynucleotide or polypeptide, etc.) that is detectable above a background signal (e.g., noise) using one or more methods of detection. The term "differential gene expression" refers to a gene or set of genes whose expression is activated to a higher or lower level in a subject suffering from a disease, (e.g., cancer) relative to its expression in a normal or control subject. Differential gene expression can also occur between different types or subtypes of diseased cells. The term also includes genes whose expression is activated to a higher or lower level at different stages of the same disease. It is also understood that a differentially expressed gene may be either activated or inhibited at the nucleic acid level or protein level, or may be subject to alternative splicing to result in a different polypeptide product. Such differences may be evidenced by a change in mRNA levels, surface expression, secretion or other partitioning of a polypeptide, for example. Differential gene expression may include a comparison of expression between two or more genes or their gene products, or a comparison of the ratios of the expression between two or more genes or their gene products, or even a comparison of two differently processed products of the same gene, which differ between, e.g., normal subjects and subjects suffering from a disease, various stages of the same disease, different types or subtypes of diseased cells, etc. Differential expression includes both quantitative, as well as qualitative, differences in the temporal or cellular expression pattern in a gene or its expression products among, for example, normal and diseased cells, or among cells which have undergone different disease events or disease stages. In certain embodiments, "differential gene expression" is considered to be present when there is at least an about two-fold, typically at least about four-fold, more typically at least about six¬ fold, most typically at least about ten-fold difference between, e.g., the expression of a given gene in normal and diseased subjects, in various stages of disease development in a diseased subject, different types or subtypes of diseased cells, etc.

The term "expression" refers to the process by which mRNA or a polypeptide is produced based on the nucleic acid sequence of a gene, i.e., "expression" also includes the formation of mRNA in the process of transcription. The term "determining the expression level" refers to the determination of the level of expression of one or more markers.

The term "gene" refers to a nucleic acid sequence encoding a gene product. The gene optionally comprises sequence information required for expression of the gene (e.g., promoters, enhancers, etc.). The term "'gene expression data" refers to one or more sets of data that contain information regarding different aspects of gene expression. The data set optionally includes information regarding: the presence of target-transcripts in cell or cell- derived samples; the relative and absolute abundance levels of target transcripts; the ability of various treatments to induce expression of specific genes; and the ability of various treatments to change expression of specific genes to different levels.

Nucleic acids "hybridize" when they associate, typically in solution. Nucleic acids hybridize due to a variety of well-characterized physico-chemical forces, such as hydrogen bonding, solvent exclusion, base stacking and the like. In certain embodiments, hybridization occurs under conventional hybridization conditions, such as under stringent conditions as described, for example, in Sambrook et al., in "Molecular Cloning: A Laboratory Manual" (1989), Eds. J. Sambrook, E. F. Fritsch and T. Maniatis, Cold Spring Harbour Laboratory Press, Cold Spring Harbour, NY, which is incorporated by reference. Such conditions are, for example, hybridization in 6x SSC, pH 7.0 / 0.1 % SDS at about 45⁰C for 18-23 hours, followed by a washing step with 2x SSC/1 % SDS at 50⁰C. In order to select the stringency, the salt concentration in the washing step can, for example, be chosen between 2x SSC/0.1 % SDS at room temperature for low stringency and 0.2x SSC/0.1 % SDS at 50°C for high stringency. In addition, the temperature of the washing step can be varied between room temperature (ca. 22 °C), for low stringency, and 65⁰C to 70°C for high stringency. Also contemplated are polynucleotides that hybridize at lower stringency hybridization conditions. Changes in the stringency of hybridization and signal detection are primarily accomplished through the manipulation of, e.g., formamide concentration (lower percentages of formamide result in lowered stringency), salt conditions, or temperature. For example, lower stringency conditions include an overnight incubation at 37⁰C in a solution comprising 6X SSPE (2OX SSPE = 3M NaCl; 0.2M NaH₂PO₄; 0.02M EDTA, pH 7.4), 0.5% SDS, 30% formamide, 100 mg/mL salmon sperm blocking DNA, followed by washes at 5O⁰C with 1 X SSPE, 0.1 %

SDS. hi addition, to achieve even lower stringency, washes performed following stringent hybridization can be done at higher salt concentrations (e.g., 5x SSC). Variations in the above conditions may be accomplished through the inclusion and/or substitution of alternate blocking reagents used to suppress background in hybridization experiments. The inclusion of specific blocking reagents may require modification of the hybridization conditions described herein, due to problems with compatibility. An extensive guide to the hybridization of nucleic acids is found in

Tijssen (1993) Laboratory Techniques in Biochemistry and Molecular Biology- Hybridization with Nucleic Acid Probes part I chapter 2, "Overview of principles of hybridization and the strategy of nucleic acid probe assays," (Elsevier, New York), as well as in Ausubel (Ed.) Current Protocols in Molecular Biology, Volumes I, II, and III, (1997), which are each incorporated by reference. Hames and Higgins (1995) Gene Probes 1 IRL Press at Oxford University Press, Oxford, England, (Hames and Higgins 1) and Hames and Higgins (1995) Gene Probes 2 IRL Press at Oxford University Press, Oxford, England (Hames and Higgins 2) provide details on the synthesis, labeling, detection and quantification of DNA and RNA, including oligonucleotides. Both Hames and Higgins 1 and 2 are incorporated by reference.

"inv(16)" refers to AML with inversion 16.

A "label" refers to a moiety attached (covalently or non-covalerifly), or capable of being attached, to a molecule (e.g., a polynucleotide, a polypeptide, etc.), which moiety provides or is capable of providing information about the molecule (e.g., descriptive, identifying, etc. information about the molecule) or another molecule with which the labeled molecule interacts (e.g., hybridizes, etc.) . Exemplary labels include fluorescent labels (including, e.g., quenchers or absorbers), non- fluorescent labels, colorimetric labels, chemiluminescent labels, bioluminescent labels, radioactive labels (such as ³H, ³⁵S, ³²P, ¹²⁵1, ⁵⁷Co or ¹⁴C), mass-modifying groups, antibodies, antigens, biotin, haptens, digoxigenin, enzymes (including, e.g., peroxidase, phosphatase, etc.), and the like. To further illustrate, fluorescent labels may include dyes that are negatively charged, such as dyes of the fluorescein family, or dyes that are neutral in charge, such as dyes of the rhodamine family, or dyes that are positively charged, such as dyes of the cyanine family. Dyes of the fluorescein family include, e.g., FAM, HEX, TET, JOE, NAN aαnd ZOE. Dyes of the rhodamine family include, e.g., Texas Red, ROX, RI lO, R6G, and TAMRA. FAM, HEX, TET, JOE, NAN, ZOE, ROX, Rl 10, R6G, and TAMRA are commercially available from, e.g., Perkin-Elmer, Inc. (Wellesley, MA, USA), and Texas Red is commercially available from, e.g., Molecular Probes, Inc. (Eugene, OR, USA). Dyes of the cyanine family include, e.g., Cy2, Cy3, Cy3.5, Cy5, Cy5.5, and Cy7, and are commercially available from, e.g., Amersham Biosciences Corp. (Piscataway, NJ, USA). Suitable methods include the direct labeling (incorporation) method, an amino-modified (amino-allyl) nucleotide method (available e.g. from Ambion, Inc. (Austin, TX, USA), and the primer tagging method (DNA dendrirner labeling, as kit available e.g. from Geni sphere, Inc.

(Hatfield, PA, USA)). In some embodiments, biotin or biotinylated nucleotides are used for labeling, with the latter generally being directly incorporated into, e.g., the cRNA polynucleotide by in vitro transcription. The term "lower expression" refers an expression level of one or more markers from a target that is less than a corresponding expression level of the markers in a reference. In certain embodiments, "lower expression" is assigned to all by numbers and Affymetrix Id. definable polynucleotides the t-values and fold change (fc) values of which are negative. Similarly, the term "higher expression" refers an expression level of one or more markers from a target that is more than a corresponding expression level of the markers in a reference. In some embodiments, "higher expression" is assigned to all by numbers and Affymetrix Id. definable polynucleotides the t-values and fold change (fc) values of which are positive.

A "machine learning algorithm" refers to a computational-based prediction methodology, also known to persons skilled in the art as a "classifier", employed for characterizing a gene expression profile. The signals corresponding to certain expression levels, which are obtained by, e.g., microarray-based [hybridization assays, are typically subjected to the algorithm in order to classify the expression profile. Supervised learning generally involves "training" a classifier to recognize the distinctions among classes and then "testing" the accuracy of the classifier on an independent test set. For new, unknown samples the classifier can be used to predict the class in which the samples belong. The term "marker" refers to a genetically controlled difference that can be used in the genetic analysis of a test or target versus a control or reference sample for the purpose of assigning the sample to a defined genotype or phenotype. In certain embodiments, for example, "markers" refer to genes, polynucleotides, polypeptides, or fragments or portions thereof that are differentially expressed in, e.g., different leukemia types and/or subtypes. The markers can be defined by their gene symbol name, their encoded protein name, their transcript identification number (cluster identification number), the data base accession number, public accession number and/or GenBank identifier. Markers can also be defined by their Affymetrix identification number, chromosomal location, UniGene accession number and cluster type, and/or LocusLink accession number. The Affymetrix identification number (affy id) is accessible for anyone and the person skilled in the art by entering the "gene expression omnibus" internet page of the National Center for Biotechnology Information (NCBI) on the world wide web at ncbi.nlm.nih.gov/geo/ as of 11/4/2004. In particular, the affy id's of the polynucleotides used for certain embodiments of the methods described herein are derived from the so-called human genome Ul 33 chip (Affymetrix, Inc., Santa Clara, CA, USA). The sequence data of each identification number can be viewed on the world wide web at, e.g., ncbi.nlm.nih.gov/projects/geo/ as of 11/4/2004 using the accession number GPL96 for Ul 33 A annotational data and accession number GPL97 for U133B annotational data. In some embodiments, the expression level of a marker is determined by the determining the expression of its corresponding polynucleotide.

The term "normal karyotype" refers to a state of those cells lacking any visible karyotype abnormality detectable with chromosome banding analysis.

The term "nucleic acid" refers to a polymer of monomers that can be corresponded to a ribose nucleic acid (RNA) or deoxyribose nucleic acid (DNA) polymer, or analog thereof. This includes polymers of nucleotides such as RNA and DNA, as well as modified forms thereof, peptide nucleic acids (PNAs), locked nucleic acids (LNA™s), and the like. In certain applications, the nucleic acid can be a polymer that includes multiple monomer types, e.g., both RNA and DNA subunits. A nucleic acid can be or include, e.g., a chromosome or chromosomal segment, a vector (e.g., an expression vector), an expression cassette, a naked DNA or RNA polymer, the product of a polymerase chain reaction (PCR) or other nucleic acid amplification reaction, an oligonucleotide, a probe, a primers, etc. /\ nucleic acid can be e.g., single-stranded or double-stranded. Unless otherwise indicated, a particular nucleic acid sequence optionally comprises or encodes complementary sequences, in addition to any sequence explicitly indicated.

Oligonucleotides (e.g., probes, primers, etc.) of a defined sequence maybe produced by techniques known to those of ordinary skill in the art, such as by chemical or biochemical synthesis, and by in vitro or in vivo expression from recombinant nucleic acid molecules, e.g., bacterial or retroviral vectors.

Oligonucleotides which are primer and/or probe sequences, as described below, may comprise DNA, RNA or nucleic acid analogs such as uncharged nucleic acid analogs including but not limited to peptide nucleic acids (PNAs) which are disclosed in International Patent Application WO 92/20702 or morpholino analogs which are described in U.S. Pat. Nos. 5,185,444, 5,034,506, and 5,1 42,047 all of which are incorporated by reference. Such sequences can routinely be synthesized using a variety of techniques currently available. For example, a sequence of DNA can be synthesized using conventional nucleotide phosphoramidite chemistry and the instruments available from Applied Biosystems, Inc, (Foster City, CA, USA);

DuPont, (Wilmington, DE, USA); or Milligen, (Bedford, MA, USA.). Similarly, and when desirable, the sequences can be labeled using methodologies well known in the art such as described in U.S. patent application numbers 5,46-4,746; 5,424,414; and 4,948,882 all of which are incorporated by reference. A nucleic acid, nucleotide, polynucleotide or oligonucleotide can comprise the five biologically occurring bases (adenine, guanine, thymine, cytosine and uracil) and/or bases other than the five biologically occurring bases. These bases may serve a number of purposes, e.g., to stabilize or destabilize hybridization; to promote or inhibit probe degradation; or as attachment points for detectable moieties or quencher moieties. For example, a polynucleotide of trie invention can contain one or more modified, non-standard, or derivatized base moieties, including, but not limited to, N -methyl-adenine, N⁶-tert-butyl-benzyl-adenine, imidazole, substituted imidazoles, 5-fluorouracil, 5-brornouracil, 5-chlorouracil, 5- iodouracil, hypoxanthine, xanthine, 4-acetylcytosine,

5-(carboxyhydroxymethyl)uracil, 5-carboxymethylamiriomethyl-2-thiouridine, 5-carboxymethylaminomethyluracil, dihydrouracil, beta-D-galactosylqueosine, inosine, N6-isopentenyladenine, 1-methylguanine, 1-methylinosine, 2,2- dimethyl guanine, 2-methyladenine, 2-methyl guanine, 3 -methyl cytosine, 5- methylcytosine, N6-methyladenine, 7-methylguanine, 5 -methylaminomethyluracil, 5-methoxyaminomethyl-2-thiouracil, beta-D mannosylqueosine, 5'- methoxycarboxymethyluracil, 5-methoxyuracil, 2-methylthio-N6- isopentenyladenine, uracil-5-oxyacetic acid (v), wybutoxosine, pseudouracil, queosine, 2-thiocytosine, 5-methyl-2-thiouracil, 2-thioιzracil, 4-thiouracil, 5- methyluracil, uracil-5- oxyacetic acidmethylester, 3-(3-amino-3-N-2- carboxypropyl) uracil, (acp3)w, 2,6- diaminopurine, and 5-propynyl pyrimidine. Other examples of modified, non-standard, or dervatized base moieties may be found in U.S. Patent Nos. 6,001,611, 5,955,589, 5,844,106, 5,789,562, 5,750,343, 5,728,525, and 5,679,785, each of which is incorporated by reference.

Furthermore, a nucleic acid, nucleotide, polynucleotide or oligonucleotide can comprise one or more modified sugar moieties including, but not limited to, arabinose, 2-fiuoroarabinose, xylulose, and hexose. A nucleic acid, nucleotide, polynucleotide or oligonucleotide can comprise phospkodi ester linkages or modified linkages including, but not limited to phosphotriester, phosphoramidate, siloxane, carbonate, carboxymethylester, acetamidate, carbamate, thioether, bridged phosphoramidate, bridged methylene phosphonate, phosphorothioate, methylphosphonate, phosphorodithioate, bridged phosphorothioate or sulfone linkages, and combinations of such linkages.

The term "polynucleotide" refers to a DNA, in particular cDNA, or RNA, in particular a cRNA, or a portion thereof. In the case of RNA (or cDNA), the polynucleotide is formed upon transcription of a nucleotide sequence that is capable of expression. "Polynucleotide fragments" refer to fragments of between at least 8, such as 10, 12, 15 or 18 nucleotides and at least 50, such as 60, 80, 100, 200 or 300 nucleotides in length, or a complementary sequence thereto, e.g., representing a consecutive stretch of nucleotides of a gene, cDNA. or mRNA. In some embodiments, polynucleotides also include any fragment (or complementary sequence thereto) of a sequence corresponding to or derived from any of the markers defined herein.

The term "polypeptide" refers to a polymer of amino acid residues. The term applies to amino acid polymers in which one or more amino acid residues are analogs, derivatives or mimetics of corresponding naturally occurring amino acids, as well as to naturally occurring amino acid polymers. For example, polypeptides can be modified or derivatized, e.g., by the addition of carbohydrate residues to form glycoproteins.

The term "primer" refers to an oligonucleotide having a hybridization specificity sufficient for the initiation of an enzymatic polymerization under predetermined conditions, for example in an amplification technique such as polymerase chain reaction (PCR), in a process of sequencing, in a method of reverse transcription and the like. The term "probe" refers to an oligonucleotide having a hybridization specificity sufficient for binding to a defined target sequence under predetermined conditions, for example in an amplification technique such as a 5' -nuclease reaction, in a hybridization-dependent detection method, such as a Southern or Northern blot, and the like. In certain embodiments, probes correspond at least in part to selected markers. Primers and probes may be used in a variety of ways and may be defined by the specific use. For example, a probe can be immobilized on a solid support by any appropriate means, including, but not limited, to: by covalent bonding, by adsorption, by hydrophobic and/or electrostatic interaction, or by direct synthesis on a solid support (see in particular patent application WO

92/10092). A probe may be labeled by means of a label chosen, for example, from radioactive isotopes, enzymes, in particular enzymes capable of acting on a chromogenic, fluorescent or luminescent substrate (in particular a peroxidase or an alkaline phosphatase), chromophoric chemical compounds, chrornogenic, fluorigenic or luminescent compounds, analogues of nucleotide bases, and ligands such as biotin. Illustrative fluorescent compounds include, for example, fluorescein, carboxyfluorescein, tetrachlorofluorescein, hexachlorofluorescein, Cy3, tetramethylrhodamine, Cy3.5, carboxy-x-rhodamine, Texas Red, Cy5, and Cy5.5. Illustrative luminescent compounds include, for example, luciferin and 2,3- dihydrophthalazinediones, such as luminol. Other suitable labels are described herein or are otherwise known to those of skill in the art.

Oligonucleotides (e.g., primers, probes, etc.), whether hybridization assay probes, amplification primers, or helper oligonucleotides, may be modified with chemical groups to enhance their performance or to facilitate the characterization of amplification products. For example, backbone-modified oligonucleotides such as those having phosphorothioate or methylphosphonate groups which render the oligonucleotides resistant to the nucleolytic activity of certain polymerases or to nuclease enzymes may allow the use of such enzymes in an amplification or other reaction. Another example of modification involves using non-nucleotide linkers (e.g., Arnold, et ah, "Non- Nucleotide Linking Reagents for Nucleotide Probes", EP 0 313 219, which is incorporated by reference) incorporated between nucleotides in the nucleic acid chain which do not interfere with hybridization or the elongation of the primer. Amplification oligonucleotides may also contain mixtures of the desired modified and natural nucleotides.

A "reference" in the context of gene expression profiling refers to a cell and/or genes in or derived from the cell (or data derived therefrom) relative to which a target is compared. In some embodiments, for example, the expression of one or more genes from a target cell is compared to a corresponding expression of the genes in or derived from a reference cell.

A "sample" refers to any biological material containing genetic information in the form of nucleic acids or proteins obtainable or obtained from one or more subjects or individuals. In some embodiments, samples are derived from subjects having leukemia, e.g., AML. Exemplary samples include tissue samples, cell samples, bone marrow, and/or bodily fluids such as blood, saliva, semen, urine, and the like. Methods of obtaining samples and of isolating nucleic acids and proteins from sample are generally known to persons of skill in the art. A "set" refers to a collection of one or more things. For example, a set may include 1, 2, 3, 4, 5, 10, 20, 50, 100, 1,000 or another number of genes or other types of molecules.

A "solid support" refers to a solid material that can be derivatized with, or otherwise attached to, a chemical moiety, such as an oligonucleotide probe or the like. Exemplary solid supports include plates (e.g., multi-well plates, etc.), beads, microbeads, tubes, fibers, whiskers, combs, hybridization chips (including microarray substrates, such as those used in GeneChip® probe arrays (Affymetrix, Inc., Santa Clara, CA, USA) and the like), membranes, single crystals, ceramic layers, self-assembling monolayers, and the like.

"Specifically binding" means that a compound is capable of discriminating between two or more polynucleotides or polypeptides. For example, the compound binds to the desired polynucleotide or polypeptide, but essentially does not bind to a non-target polynucleotide or polypeptide. The compound can be an antibody, or a fragment thereof, an enzyme, a so-called small molecule compound, a protein- scaffold (e.g., an anticalin).

A "subject" refers to an organism. Typically, the organism is a mammalian organism, particularly a human organism.

The term "substantially identical" in the context of gene expression refers to levels of expression of genes that are approximately equal to one another. In some embodiments, for example, the expression levels of genes being compared are substantially identical to one another when they differ by less than about 5% (e.g., about 4%, about 3%, about 2%, about 1%, etc.).

"t(8;16)" refers to AML with translocation (8;16). "t(15;17)" refers to AML with translocation (15;17).

"t(8;21)" refers to AML with translocation (8;21).

The term "target" refers to an object that is the subject of analysis. In some embodiments, for example, targets are specific nucleic acid sequences (e.g., mRNAs of expressed genes, etc.), the presence, absence or abundance of which are to be determined. In certain embodiments, targets include polypeptides (e.g., proteins, etc.) of expressed genes. Typically, the sequences subjected to analysis are in or derived from "target cells", such as a particular type of leukemia cell.

INTRODUCTION

The present invention provides methods, reagents, systems, and kits for detecting leukemia. In some embodiments, for example, the methods include detecting an expression of a gene population of a cell, which gene population comprises a set of genes listed in one or more of Tables I-XII to detect the leukemia. To further illustrate, the methods also include subtyping the leukemia in certain embodiments. For example, the methods optionally include correlating a detected expression of one or more sets of genes listed in at least one of Tables I- VI with the cell being an acute myeloid leukemia (AML) cell. In some embodiments, the methods include correlating a detected expression of one or more sets of genes listed in at least one of Tables VII-X with the cell being an acute lymphoblastic leukemia (ALL) cell. In certain embodiments, the methods include correlating a detected expression of a set of genes listed in Table XI with the cell being a chronic myeloid leukemia

(CML) cell. In some embodiments, the methods include correlating a detected expression of a set of genes listed in Table XII with the cell being a chronic lymphatic leukemia (CLL) cell.

The use of one or more of the markers described herein, e.g., utilizing a microarray technology or other gene expression profiling techniques, provides various advantages, including: (1) rapid and accurate diagnoses, (2) ease of use in laboratories without specialized knowledge, and (3) eliminates the need for analyzing viable cells for chromosome analysis, tfciereby eliminating cell sample transport issues. Aspects of the present invention are further illustrated in the examples provided below.

In practicing the present invention, many conventional techniques in, hematology, molecular biology and recombinant DNA are optionally used. These techniques are well known and are explained in, for example, Current Protocols in Molecular Biology, Volumes I, II, and III, 1997 (F. M. Ausubel ed.); Sambrook et al., Molecular Cloning: A Laboratory Manual. 3rd Ed., Cold Spring Harbor Laboratory

Press, Cold Spring Harbor, N. Y., 2001 ; Berger and Kimmel, Guide to Molecular Cloning Techniques. Methods in Enzymology volume 152 Academic Press, Inc., San Diego, CA (Berger), DNA Cloning: A Practical Approach, Volumes I and II, 1985 (D. N. Glover ed.); Oligonucleotide Synthesis, 1984 (M. L. Gait ed.); Nucleic Acid Hybridization. 1985, (Hames and Higgins); Transcription and Translation, 1984 (Hames and Higgins eds.); Animal Cell Culture, 1986 (Freshney ed.);

Immobilized Cells and Enzymes, 1986 (IRL Press); Perbal, 1984, A Practical Guide to Molecular Cloning; the series, Methods in Enzymology (Academic Press, Inc.); Gene Transfer Vectors for Mammalian Cells. 1987 (J. H. Miller and M. P. Calos eds., Cold Spring Harbor Laboratory); Greer et al. (Eds.), Wintrobe's Clinical Hematology, 11^th Ed., Lippincott Williams & Wilkins (2003); Shirlyn et al.,

Clinical Laboratory Hematology, Prentice Hall (2O02); Lichtman et al., Williams Manual of Hematology, 6^th Ed., McGraw-Hill Professional (2002); and Methods in Enzvmology Vol. 154 and Vol. 155 (Wu and Grossman, and Wu, eds., respectively), all of which are incorporated by reference. In addition to the methods of detecting leukemia, the related kits, systems, and other aspects of the invention are also described farther below.

SAMPLE COLLECTION AND PREPARATION

Samples are collected and prepared for analysis using essentially any technique known to those of skill in the art. In certain embodiments, for example, blood samples are obtained from subjects via venipuncture. Whole blood specimens are optionally collected in EDTA, Heparin or ACD vacutainer tubes. In other embodiments, the samples utilized for analysis comprise bone marrow aspirates, which are optionally processed, e.g., by erythrocyte lysis techniques, Ficoll density gradient centrifugations, or the like. Samples are typically either analyzed immediately following acquisition or stored frozen at, e.g., -80°C until being subjected to analysis. Sample collection and handling are also described in, e.g., Garland et al., Handbook of Phlebotomy and Patient Service Techniques, Lippincott Williams & Wilkins (1998), and Slockbower et al. (Eds.), Collection and Handling of Laboratory Specimens: A Practical Guide, Lippincott Williams & Wilkins (1983), which are both incorporated by reference. Treatment of Cells

The cells lines or sources containing the target nucleic acids and/or expression products thereof, are optionally subjected to one or more specific treatments that induce changes in gene expression, e.g., as part of processes to identify candidate modulators of gene expression. For example, a cell or cell line can be treated with or exposed to one or more chemical or biochemical constituents, e.g., pharmaceuticals, pollutants, DNA damaging agents, oxidative stress-inducing agents, pH-altering agents, membrane-disrupting agents, metabolic blocking agent, a chemical inhibitors, cell surface receptor ligands, antibodies, transcription promoters/enhancers/inhibitors, translation promoters/enhancers/inhibitors, protein- stabilizing or destabilizing agents, various toxins, carcinogens or teratogens, characterized or uncharacterized chemical libraries, proteins, lipids, or nucleic acids. Optionally, the treatment comprises an environmental stress, such as a change in one or more environmental parameters including, but not limited to, temperature (e.g. heat shock or cold shock), humidity, oxygen concentration (e.g., hypoxia), radiation exposure, culture medium composition, or growth saturation. Responses to these treatments may be followed temporally, and the treatment can be imposed for various times and at various concentrations. Target sequences can also be derived from cells exposed to multiple specific treatments as described above, either concurrently or in tandem (e.g., a cancerous cell or tissue sample may be further exposed to a DNA damaging agent while grown in an altered medium composition).

RNA Isolation

In some embodiments, total RNA is isolated from samples for use as target sequences. Cellular samples are lysed once culture with or without the treatment is complete by, for example, removing growth medium and adding a guanidinium- based lysis buffer containing several components to stabilize the RNA. In certain embodiments, the lysis buffer also contains purified RNAs as controls to monitor recovery and stability of RNA from cell cultures. Examples of such purified RNA templates include the Kanamycin Positive Control RNA from Promega (Madison,

WI, USA), and 7.5 kb Poly(A)-Tailed RNA from Life Technologies (Rockville, MD, USA). Lysates may be used immediately or stored frozen at, e.g., -80°C. Optionally, total RNA is purified from cell lysates (or other types of samples) using silica-based isolation in an automation- compatible, 96-well format, such as the Rneasy® purification platform (Qiagen, Inc. (Valencia, CA, USA)). Alternatively, RNA is isolated using solid-phase oligo-dT capture using oligo-dT bound to microbeads or cellulose columns. This method has the added advantage of isolating mRNA from genomic DNA and total RNA, and allowing transfer of the mRNA-capture medium directly into trie reverse transcriptase reaction. Other RNA isolation methods are contemplated, such as extraction with silica-coated beads or guanidinium. Further methods for RNA isolation and preparation can be devised by one skilled in the art.

Alternatively, the methods of the present invention are performed using crude cell lysates, eliminating the need to isolate RNA. RNAse inhibitors are optionally added to the crude samples. When using crude cellular lysates, genomic DNA could contribute one or more copies of target sequence, depending on the sample. In situations in which the target sequence is derived from one or more highly expressed genes, the signal arising from genomic DNA may not be significant. But for genes expressed at very low levels, the background can be eliminated by treating the samples with DNAse, or by using primers that target splice junctions. One skilled in the art can design a variety of specialized priming applications that would facilitate use of crude extracts as samples for the purposes of this invention.

GENE EXPRESSION PROFILING

The determination of gene expression levels may be effected at the transcriptional and/or translational level, i.e., at the level of mRNA or at the protein level. Essentially any method of gene expression profiling can be used or adapted for use in performing the methods described herein including, e.g., methods based on hybridization analysis ofpolynucleotid.es, and methods based on sequencing of polynucleotides. To illustrate, commonly used methods for the quantification of mRNA expression in a sample include northern blotting and in situ hybridization (Parker & Barnes, Methods in Molecular Biology 106:247-283 (1999)), RNAse protection assays (Hod, Biotechniques 13:852-854 (1992)), and reverse transcription polymerase chain reaction (RT-PCR) (Weis et al., Trends in Genetics 8:263-264 (1992)). Alternatively, antibodies may be employed that can recognize specific duplexes, including DNA duplexes, RJST A duplexes, and DNA-RNA hybrid duplexes or DNA-protein duplexes. Representative methods for sequencing-based gene expression analysis include Serial Analysis of Gene Expression (SAGE), and gene expression analysis by massively parallel signature sequencing (MPSS). Optionally, molecular species, such as antibodies, aptamers, etc. that can specifically bind to proteins or fragments thereof are used for analysis (see, e.g., Beilharz et al, Brief Funct Genomic Proteomic 3(2):103-l 11 (2004)). Some of these techniques, with a certain degree of overlap in some cases, are described further below. In certain embodiments, for example, the methods described herein include determining the expression levels of transcribed polynucleotides. In some of these embodiments, the transcribed polynucleotide is an mRNA, a cDNA and/or a cRNA. Transcribed polynucleotides are typically isolated from a sample, reverse transcribed and/or amplified, and labeled by techniques referred to above or otherwise known to persons skilled in the art. In order to determine the expression level of transcribed polynucleotides, the methods of the invention generally include hybridizing transcribed polynucleotides to a complementary polynucleotide, or a portion thereof, under a selected hybridization condition (e.g., a stringent hybridization condition), as described herein. In some embodiments, the detection and quantification of amounts of polynucleotides to determine the level of expression of a marker are performed according to those described by, e.g., Sambrook et al., supra, or real time methods known in the art as 5'-nuclease methods disclosed in, e.g., WO 92/02638, U.S. Pat. No. 5,210,015, U.S. Pat. No. 5,804,375, and U.S. Pat. No. 5,487,972, which are each incorporated by reference. In some embodiments, for example, 5'-nuclease methods utilize the exonuclease activity of certain polymerases to generate signals. In these approaches, target nucleic acids are detected in processes that include contacting a sample with an oligonucleotide containing a sequence complementary to a region of the target nucleic acid component and a labeled oligonucleotide containing a sequence complementary to a second region of the same target nucleic acid component sequence strand, but not including the nucleic acid sequence defined by the first oligonucleotide, to create a mixture of duplexes during hybridization conditions, wherein the duplexes comprise the target nucleic acid annealed to the first oligonucleotide and to the labeled oligonucleotide such that the 3 '-end of the first oligonucleotide is adjacent to the 5 '-end of the labeled oligonucleotide. Then this mixture is treated with a template-dependent nucleic acid polymerase having a 5' to 3' nuclease activity under conditions sufficient to permit the to 3' nuclease activity of the polymerase to cleave the annealed, labeled oligonucleotide and release labeled fragments. The signal generated by the hydrolysis of the labeled oligonucleotide is detected and/or measured. 5 '-nuclease technology eliminates the need for a solid phase bound reaction complex to be formed and made detectable. Other exemplary methods include, e.g., fluorescence resonance energy transfer between two adjacently hybridized probes as used in the LightCycler® format described in, e.g., U.S. Pat. No. 6,174,670, which is incorporated by reference. In one protocol, the marker, i.e., the polynucleotide, is in form of a transcribed nucleotide, where total RNA is isolated, cDNA and, subsequently, cRNA is synthesized and biotin is incorporated during the transcription reaction. The purified cRNA is applied to commercially available arrays that can be obtained from, e.g., Affymetrix, Inc. (Santa Clara, CA USA). The hybridized cRNA is optionally detected according to the methods described in the examples provided below. The arrays are produced by photolithography or other methods known to persons skilled in the art. Some of these techniques are also described in, e.g. U.S. Pat. No. 5,445,934, U.S. Pat. No. 5,744,305, U.S. Pat. No. 5,700,637, U.S. Pat. No. 5,945,334, EP 0 619 321, and EP 0 373 203, which are each incorporated by reference.

In another embodiment, the polynucleotide or at least one of the polynucleotides is in form of a polypeptide (e.g., expressed from the corresponding polynucleotide). The expression level of the polynucleotides or polypeptides is optionally detected using a compound that specifically binds to target polynucleotides or target polypeptides. These and other exemplary gene expression profiling techniques are described further below.

Blotting Techniques

Some of the earliest expression profiling methods are based on the detection of a label in RNA hybrids or protection of RNA from enzymatic degradation (see, e.g.,

Ausubel et al., supra). Methods based on detecting hybrids include northern blots and slot/dot blots. These two techniques differ in that the components of the sample being analyzed are resolved by size in a northern blot prior to detection, which enables identification of more than one species simultaneously. Slot blots are generally carried out using unresolved mixtures or sequences, but can be easily performed in serial dilution, enabling a more quantitative analysis.

In Situ Hybridization

In situ hybridization is a technique that monitors transcription by directly visualizing RNA hybrids in the context of a whole cell. This method provides information regarding subcellular localization of transcripts (see, e.g., Suzuki et al.,

Pigment Cell Res. 17(l):10-4 (2004)).

Assays Based on Protection from Enzymatic Degradation

Techniques to monitor RNA that make use of protection from enzymatic degradation include Sl analysis and RNAse protection assays (RPAs). Both of these assays employ a labeled nucleic acid probe, which is hybridized to the RNA species being analyzed, followed by enzymatic degradation of single-stranded regions of the probe. Analysis of the amount and length of probe protected from degradation is used to determine the quantity and endpoints of the transcripts being analyzed.

Reverse Transcriptase PCR (RT-PCR) and Real-Time Detection

RT-PCR can be used to compare, e.g., mRNA levels in different sample populations, in normal and tumor tissues, with or without drug treatment, to characterize patterns of gene expression, to discriminate between closely related mRNAs, and to analyze RNA structure. These assays are derivatives of PCR in which amplification is preceded by reverse transcription of mRNA into cDNA.

Accordingly, an initial step in these processes is generally the isolation of mRNA from a target sample (e.g., leukemia cells). The starting material is typically total RNA isolated from cancerous tissues or cells (e.g., bone marrow, peripheral blood aliquots, etc.), and in certain embodiments, from corresponding normal tissues or cells. General methods for mRNA extraction are well known in the art and are disclosed in standard textbooks of molecular biology, including Ausubel et al., supra. Methods for RNA extraction from paraffin embedded tissues are disclosed, for example, in Rupp and Locker, Lab Invest. 56:A67 (1987), and De Andres et al., BioTechniques 18:42044 (1995), which are each incorporated by reference. In particular, RNA isolation can be performed using purification kit, buffer set and protease from commercial manufacturers, such as Qiagen, according to the manufacturer's instructions. For example, total RNA from cells in culture can be isolated using Qiagen Rneasy® mini-columns (referred to above). Other commercially available RNA isolation kits include MasterPure™ Complete DNA and RNA Purification Kit (EPICENTRE™, Madison, Wis.), and Paraffin Block

RNA Isolation Kit (Ambion, Inc.). Total RNA from tissue samples can be isolated using RNA Stat-60 (Tel-Test). RNA prepared from tumor can be isolated, for example, by cesium chloride density gradient centrifugation.

Since RNA generally cannot serve as a template for PCR, the process of gene expression profiling by RT-PCR typically includes the reverse transcription of the

RNA template into cDNA, followed by its exponential amplification in a PCR reaction. Two commonly used reverse transcriptases are avilo myeloblastosis virus reverse transcriptase (AMV-RT) and Moloney murine leukemia virus reverse transcriptase (MMLV-RT). The reverse transcription step is typically primed using specific primers, random hexamers, or oligo-dT primers, depending on the particular circumstances of expression profiling analysis. For example, extracted RNA can be reverse-transcribed using a GeneAmp RNA PCR kit (Perkin Elmer, CA, USA), following the manufacturer's instructions. The derived cDNA can then be used as a template in the subsequent PCR reaction. Although the PCR step can use a variety of thermostable DNA-dependent DNA polymerases, it typically employs the Taq DNA polymerase, which has a 5 '-3' nuclease activity but lacks a 3'-5' proofreading endonuclease activity. Thus, TaqMan® PCR typically utilizes the 5'-nuclease activity of Taq or Tth polymerase to hydrolyze a hybridization probe bound to its target amplicon, but any enzyme with equivalent 5' nuclease activity can be used. Pairs of primers are generally used to generate amplicons in PCR reactions. A third oligonucleotide, or probe, is designed to bind to nucleotide sequence located between PCR primer pairs. Probe are generally non-extendible by Taq DNA polymerase enzyme, and are typically labeled with, e.g., a reporter fluorescent dye and a quencher fluorescent dye. Laser-induced emission from the reporter dye is quenched by the quenching dye when the two dyes are located close together, such as in an intact probe. During the amplification reaction, the Taq DNA polymerase enzyme cleaves the probe in a template-dependent manner. The resultant probe fragments disassociate in solution, and signal from the released reporter dye is free from the quenching effect of the second fluorophore. One molecule of reporter dye is typically liberated for each new molecule synthesized, and detection of the unquenched reporter dye provides the basis for quantitative interpretation of the data.

TaqMan® RT-PCR can be performed using commercially available equipment, such as, for example, a LightCycler® system (Roche Molecular Biochemicals, Mannheim, Germany) or an ABI PRISM 7700™ Sequence Detection System™ (Perkin-Elmer-Applied Biosystems, Foster City, CA, USA).

To minimize errors and the effect of sample-to-sample variation, RT-PCR is typically performed using an internal standard. An ideal internal standard is expressed at a relatively constant level among different cells or tissues, and is unaffected by the experimental treatment. Exemplary RNAs frequently used to normalize patterns of gene expression are mRNAs transcribed from for the housekeeping genes glyceraldehyde-3-phosphate-dehydrogenase (GAPDH) and β- actin.

Other exemplary methods for targeted mRNA analysis include differential display reverse transcriptase PCR (DDRT-PCR) and RNA arbitrarily primed PCR (RAP- PCR) (see, e.g., U.S. Patent No. 5,599,672; Liang and Pardee (1992) Science

257:967-971; Welsh et al. (1992) Nucleic Acids Res. 20:4965-4970, which are each incorporated by reference). Both methods use random priming to generate RT-PCR fingerprint profiles of transcripts in an unfractionated RNA preparation. The signal generated in these types of analyses is a pattern of bands separated on a sequencing gel. Differentially expressed genes appear as changes in the fingerprint profiles between two samples, which can be loaded in separate wells of the same gel. This type of readout allows identification of both up- and down-regulation of genes in the same reaction, appearing as either an increase or decrease in intensity of a band from one sample to another.

Molecular beacons are oligonucleotides designed for real time detection and quantification of target nucleic acids. The 5' and 3' termini of molecular beacons collectively comprise a pair of moieties, which confers the detectable properties of the molecular beacon. One of the termini is attached to a fluorophore and the other is attached to a quencher molecule capable of quenching a fluorescent emission of the fluorophore. To illustrate, one example fluorophore-quencher pair can use a fluorophore, such as EDA]STS or fluorescein, e.g., on the 5 '-end and a quencher, such as Dabcyl, e.g., on the 3'-end. When the molecular beacon is present free in solution, i.e., not hybridized, to a second nucleic acid, the stem of the molecular beacon is stabilized by complementary base pairing. This self-complementary pairing results in a "hairpin loop" structure for the molecular beacon in which the fluorophore and the quenching moieties are proximal to one another. In this confirmation, the fluorescent moiety is quenched by the quenching moiety. The loop of the molecular beacon typically comprises the oligonucleotide probe and is accordingly complementary to a sequence to be detected in the target microbial nucleic acid, such that hybridization of the loop to its complementary sequence in the target forces disassociation of the stem, thereby distancing the fluorophore and quencher from each other. This results in unquenching of the fluorophore, causing an increase in fluorescence of the molecular beacon.

Details regarding standard methods of making and using molecular beacons are well established in the literature and molecular beacons are available from a number of commercial reagent sources. Further details regarding methods of molecular beacon manufacture and use are found, e.g., in Leone et al. (1995) "Molecular beacon probes combined with amplification by NASBA enable homogenous real-time detection of RNA," Nucleic Acids Res. 26:2150-2155; Kostrikis et al. (1998) "Molecular beacons: spectral genotyping of human alleles" Science 279:1228-1229; Fang et al. (1999) "Designing a novel molecular beacon for surface-immobilized DNA hybridization studies" J. Am. Chem. Soc. 121 :2921 -

2922; and Marras et al. (1999) "Multiplex detection o>f single-nucleotide variation using molecular beacons" Genet. Anal. Biomol. Eng. 14:151-156, all of which are incorporated by reference. A variety of commercial suppliers produce standard and custom molecular beacons, including Oswel Research Products Ltd. (UK), Research Genetics (a division of Invitrogen, Huntsville, AL, USA), the Midland

Certified Reagent Company (Midland, TX, USA), and Gorilla Genomics, LLC (Alameda, CA, USA). A variety of kits which utilize molecular beacons are also commercially available, such as the Sentinel™ Molecular Beacon Allelic Discrimination Kits from Stratagene (La Jolla, CA, USA) and various kits from Eurogentec SA (Belgium) and Isogen Bioscience BV (Netherlands).

Nucleic Acid Array-Based Analysis

Differential gene expression can also be identified, or confirmed using arrayed oligonucleotides (e.g., microarrays), which have the benefit of assaying for sample hybridization to a large number of probes in a highly parallel fashion. In these approaches, polynucleotide sequences of interest (e.g., probes, such as cDNAs, mRNAs, oligonucleotides, etc.) are plated, synthesized, or otherwise disposed on a microchip substrate or other type of solid support (see, e.g., U.S. Patent Nos. 5,143,854 and 5,807,522; Fodor et al. (1991) Science 251:767-773; and Schena et al. (1995) Science 270:467-470, which are each incorporated by reference). Sequences of interest can be obtained, e.g., by creating a cDNA library from an mRNA source or by using publicly available databases, such as GenBank, to annotate the sequence information of custom cDNA libraries or to identify cDNA clones from previously prepared libraries. The arrayed sequences are then hybridized with target nucleic acids from cells or tissues of interest. As in the RT- PCR assays referred to above, the source of mRNA t;ypically is total RNA isolated from a sample. In certain embodiments, high-density oligonucleotide arrays are produced using a light-directed chemical synthesis process (i.e., photolithography). Unlike common cDNA arrays, oligonucleotide arrays (according, e.g., to the Affymetrix technology) typically use a single-dye technology. Given the sequence information of the probes or markers, the sequences are typically synthesized directly onto the array, thus, bypassing the need for physical intermediates, such as PCR products, commonly utilized in making cDNA arrays. For this purpose, selected markers, or partial sequences thereof, can be represented by, e.g., between about 14 to 20 features, typically by less then 14 features, more typically less then about 10 features, even more typically by about 6 features or less, with each feature generally being a short sequence of nucleotides (oligonucleotide), which is typically a perfect match (PM) to a segment of the respective gene. The PM oligonucleotides are paired with mismatch (MM) oligonucleotides, which have a single mismatch at the central base of the nucleotide and are used as "controls". The chip exposure sites are typically defined by masks and are de-protected by the use of light, followed by a chemical coupling step resulting in the synthesis of one nucleotide. The masking, light deprotection, and coupling process can then be repeated to synthesize the next nucleotide, until the nucleotide chain is of the specified length.

To illustrate other embodiments of microarray-based assays, PCR amplified inserts of cDNA clones are applied to a substrate in a dense array. In some embodiments., for example, at least 10,000 different cDNA probe sequences are applied to a given solid support. Fluorescently labeled cDNA targets may be generated through incorporation of fluorescent nucleotides by reverse transcription of RNA extracted, from the samples of interest. Labeled cDNA targets applied to the chip hybridize with corresponding probes on the array. After washing (e.g., under stringent conditions) to remove non-specifϊcally bound probes, the chip is typically scanned, by confocal laser microscopy or by another detection method, such as a CCD camera. Quantitation of hybridization of each arrayed element allows for assessment of corresponding mRNA abundance. With dual color fluorescence, for example, separately labeled cDNA probes generated from two sources of RNA can be hybridized concurrently to the arrayed probes. The relative abundance of the transcripts from the two sources corresponding to each specified gene can thus be determined simultaneously. The miniaturized scale of the hybridization affords a convenient and rapid evaluation of the expression pattern for large numbers of genes. Such methods have been shown to have the sensitivity required to detect rare transcripts, which are expressed at a few copies per cell, and to reproducibly detect at least approximately two-fold differences in the expression levels (Schena et al, Proc. Natl. Acad. Sci. USA 93(2):106-149 (1996), which is incorporated by reference). Other microarray-based assay formats are also optionally utilized. Microarray analysis can be performed using commercially available equipment, following manufacturer's protocols, such as by using the Affymetrix GeneChip® technology, or Agilent's microarray technology.

If the polynucleotide being detected is mRNA, cDNA may be prepared into which a detectable label, as exemplified herein, is incorporated. For example, labeled cDNA, in single-stranded form, may then be hybridized (e.g., under stringent or highly stringent conditions) to a panel of single-stranded oligonucleotides representing different genes and affixed to a solid support, such as a chip. Upon applying appropriate washing steps, those cDNAs that have a counterpart in the oligonucleotide panel or array will be detected (e.g., quantitatively detected). Various advantageous embodiments of this general method are feasible. For example, mRNA or cDNA may be amplified, e.g., by a polymerase chain reaction or another nucleic acid amplification technique. In some embodiments, where quantitative assessments are sought, it is generally desirable that the number of amplified copies corresponds to the number of mRNAs originally present in the cell. Optionally, cDNAs are transcribed into cRNAs prior to hybridization steps in a given assay. In these embodiments, labels can be attached or incorporated cRNAs during or after the transcription step.

To further illustrate, one exemplary embodiment of the methods of the invention includes, as follows (1) obtaining a sample, e.g. bone marrow or peripheral blood aliquots, from a patient; (2) extracting RNA, e.g., mRNA, from the sample; (3) reverse transcribing the RNA into cDNA; (4) in vitro transcribing the cDNA into cRNA; (5) fragmenting the cRNA; (6) hybridizing the fragmented cRNA on selected microarrays (e.g., the HG-Ul 33 niicroarray set available from Affymetrix, Inc. (Santa Clara, CA USA)); and (7) detecting hybridization.

Serical Analysis of Gene Expression (SAGE)

Serial analysis of gene expression (SAGE) is a method that allows the simultaneous and quantitative analysis of a large number of gene transcripts, without the need for providing an individual hybridization probe for each transcript. Initially, a short sequence tag (e.g., about 10-14 bp) is generated that contains sufficient information to uniquely identify a transcript, provided that the tag is obtained from a unique position within each transcript. Then, many transcripts are linked together to form long serial molecules, that can be sequenced, revealing the identity of the multiple tags simultaneously. The expression pattern of any population of transcripts can be quantitatively evaluated by determining the abundance of individual tags, and identifying the gene corresponding to each tag. SAGE-based assays are also described in, e.g. Velculescu et al., Science 270:484- 487 (1995) and Velculescu et al., Cell 88:243-51 (1997), which are both incorporated by reference.

Gene Expression Analysis by Massively Parallel Signature Sequencing (MPSS)

These sequencing approaches generally combine non-gel-based signature sequencing with in vitro cloning of millions of templates on separate 5 μm diameter microbeads. Typically, a microbead library of DNA templates is constructed by in vitro cloning. This is generally followed by the assembly of a planar array of the template-containing microbeads in a flow cell at a high density (typically greater than 3 x 10 microbeads/cm ). The free ends of the cloned templates on each microbead are analyzed simultaneously, using a fluorescence- based signature sequencing method that does not require DNA fragment separation. This method can be used to simultaneously and accurately provide, in a single operation, hundreds of thousands of gene signature sequences firom cDNA libraries. MPSS is also described in, e.g., Brenner et al., (2000) Nature Biotechnology 18:630-634, which is incorporated by reference. Immunoassays and proteomics

Essentially any available technique for the detection of proteins is optionally utilized in the methods of the invention. Exemplary protein analysis technologies include, e.g., one- and two-dimensional SDS-P AGE-based separation and detection, immunoassays (e.g., western blotting, etc.), aptamer-based detection, mass spectrometric detection, and the like. These and other techniques are generally well-known in the art.

To illustrate, immunohistochemical methods are optionally used for detecting the expression levels of the targets described herein. Thus, antibodies or antisera (e.g., polyclonal antisera) and in certain embodiments, monoclonal antibodies specific for particular targets are used to detect expression. In some of these embodiments, antibodies are directly labeled, e.g., with radioactive labels, fluorescent labels, haptens, chemiluminescent dyes, enzyme substrates or co-factors, enzyme inhibitors, free radicals, enzymes (e.g., horseradish peroxidase or alkaline phosphatase), or the like. Such labeled reagents may be used in a variety of well known assays, such as radioimmunoassays, enzyme immunoassays, e.g., ELISA., fluorescent immunoassays, and the like. See, e.g., U.S. Pat. Nos. 3,766,162; 3,791,932; 3,817,837; and 4,233,402, which are each incorporated by reference. Additional labels are described further herein. Alternatively, unlabeled primary antibodies are used in conjunction with labeled secondary antibodies, comprising antisera, polyclonal antisera or a monoclonal antibody specific for the primary antibody. Immunohistochemistry protocols and kits are well known in the art and are commercially available.

To further illustrate, proteins from a cell or tissue under investigation may be contacted with a panel or array of aptamers or of antibodies or fragments or derivatives thereof. These biomolecules may be affixed to a solid support, such as a chip. The binding of proteins indicative of a given leukemia type or subtype is optionally verified by binding to a detectably labeled secondary antibody or aptamer. The labeling of antibodies is also described in, e.g., Harlow and Lane, Antibodies, a laboratory manual, CSH Press (1988), which is incorporated by reference. To further illustrate, a minimum set of proteins necessary for detecting various leukemia types or subtypes may be selected for the creation of a protein array for use in making diagnoses with, e.g., protein lysates of bone marrow samples directly. Protein array systems for the detection of specific protein expression profiles are commercially available from various suppliers, including the Bio-Plex™ platform available from BIO-RAD Laboratories (Munich, Germany). In some embodiments of the invention, antibodies against the target proteins are produced and immobilized on a solid support, e.g., a glass slide or a well of a microtiter plate. The immobilized antibodies can be labeled with a reactant that is specific for the target proteins. These reactants can include, e.g., enzyme substrates, DNA, receptors, antigens or antibodies to create for example a capture sandwich immunoassay.

Target proteins can also be detected using aptamers including photoaptamers. Aptamers generally are single-stranded oligonucleotides (e.g., typically DNA for diagnostic applications) that assume a specific, sequence-dependent shape and binds to target proteins based on a "lock-and-key" fit between the two molecules. Aptamers can be identified using the SELEX process (Gold (1996) "The SELEX process: a surprising source of therapeutic and diagnostic compounds," Harvey Lect. 91 :47-57, which is incorporated by reference). Aptamer arrays are commercially available from various suppliers including, e.g., SomaLogic, Inc. (Boulder, CO, USA). The detection of proteins via mass includes various formats that can be adapted for use in the methods of the invention. Exemplary formats include matrix assisted laser desorption/ionization- (MALDI) and surface enhanced laser desorption/ionization-based (SELDI) detection. MALDI- and SELDI-based detection are also described in, e.g., Weinberger et al. (2000) "Recent trends in protein biochip technology," Pharmaco genomics 1(4):395-416, Forde et al. (20O2)

"Characterization of transcription factors by mass spectrometry and the role of SELDI-MS," Mass Spectrom. Rev. 21(6):419-439, and Leushner (2001) "MALEI TOF mass spectrometry: an emerging platform for genomics and diagnostics," Expert Rev. MoI. Diagn. 1(1): 1 1-18, which are each incorporated by reference. Protein chips and related instrumentation are available from commercial suppliers, such as Ciphergen Biosystems, Inc. (Fremont, CA, USA). OLIGONUCLEOTIDE PREPARATION

Various approaches can be utilized by one of skill in the art to design oligonucleotides for use as probes and/or primers. To illustrate, the DNAstar software package available from DNASTAR, Inc. (Madison, WI) can be used for sequence alignments. For example, target nucleic acid sequences and non-target nucleic acid sequences can be uploaded into DNAstar EditSeq program as individual files, e.g., as part of a process to identify regions in these sequences that have low sequence similarity. To further illustrate, pairs of sequence files can be opened in the DNAstar MegAlign sequence alignment program and the Clustal W method of alignment can be applied. The parameters used for Clustal W alignments are optionally the default settings in the software. MegAlign typi cally does not provide a summary of the percent identity between two sequences. This is generally calculated manually. From the alignments, regions having, e.g., less than 85% identity with one another are typically identified and oligonucleotide sequences in these regions can be selected. Many other sequence alignment algorithms and software packages are also optionally utilized. Sequence alignment algorithms are also described in, e.g., Mount, Bioinformatics: Sequence and Genome Analysis, Cold Spring Harbor Laboratory Press (2001), and Durbin et al., Biological Sequence Analysis: Probabilistic Models of Proteins and Nucleic Acids, Cambridge University Press (1998), which are both incorporated by reference.

To further illustrate, optimal alignment of sequences for comparison can be conducted, e.g., by the local homology algorithm of Smith & Waterman (1981) Adv. Appl. Math. 2:482, by the homology alignment algorithm of Needlemaxi & Wunsch (1970) J. MoI. Biol. 48:443, by the search for similarity method of Pearson & Lipman (1988) Proc. Nat'l. Acad. Sci. USA 85:2444, which are each incorporated by reference, and by computerized implementations of these algorithms (e.g., GAP, BESTFIT, FASTA, and TFASTA in the Wisconsin Genetics Software Package, Genetics Computer Group (Madison, WI)), or by even by visual inspection. Another example algorithm that is suitable for determining percent sequence identity is the BLAST algorithm, which is described in, e.g., Altschul et al. (1990) J. MoI. Biol. 215 :403-410, which is incorporated by reference. Software for performing versions of BLAST analyses is publicly available through the National Center for Biotechnology Information on the world wide web at ncbi.nlm.nih.gov/ as of 11/4/2004.

An additional example of a useful sequence alignment algorithm is PILEUP. PILEUP creates a multiple sequence alignment from a group of related sequences using progressive, pairwise alignments. It can also plot a tree showing the clustering relationships used to create the alignment. PILEUP uses a simplification of the progressive alignment method of Feng & Doolittle (1987) J. MoI. Evol. 35:351-360, which is incorporated by reference. Oligonucleotide probes and primers are optionally prepared using essentially any technique known in the art. In certain embodiments, for example, the oligonucleotide probes and primers are synthesized chemically using essentially any nucleic acid synthesis method, including, e.g., according to the solid phase phosphoramidite method described by Beaucage and Caruthers (1981) Tetrahedron Letts. 22(20): 1859-1862, which is incorporated by reference. To further illustrate, oligonucleotides can also be synthesized using a tri ester method (see, e.g., Capaldi et al. (2000) "Highly efficient solid phase synthesis of oligonucleotide analogs containing phosphorodithioate linkages" Nucleic Acids Res. 28(9):e40 and Eldrup et al. (1994) "Preparation of oligodeoxyribonucleoside phosphorodithioates by a triester method" Nucleic Acids Res. 22(10): 1797-1804, which are both incorporated by reference). Other synthesis techniques known in the art can also be utilized, including, e.g., using an automated synthesizer, as described in Needham-VanDevanter et al. (1984) Nucleic Acids Res. 12:6159-6168, which is incorporated by reference. A wide variety of equipment is commercially available for automated oligonucleotide synthesis. Multi-nucleotide synthesis approaches

(e.g., tri-nucleotide synthesis, etc.) are also optionally utilized. Moreover, the primer nucleic acids optionally include various modifications. In certain embodiments, for example, primers include restriction site linkers, e.g., to facilitate subsequent amplicon cloning or the like. To further illustrate, primers are also optionally modified to improve the specificity of amplification reactions as described in, e.g., U.S. Pat. No. 6,001,611, entitled "MODIFIED NUCLEIC ACID AMPLIFICATION PRIMERS," issued December 14, 1999 to Will, which is incorporated by reference. Primers and probes can also be synthesized with various other modifications as described herein or as otherwise known in the art.

Probes and/or primers utilized in the methods and other aspects of the invention are typically labeled to permit detection of probe- target hybridization duplexes. In general, a label can be any moiety that can be attached to a nucleic acid and provide a detectable signal (e.g., a quantifiable signal). Labels may be attached to oligonucleotides directly or indirectly by a variety of techniques known in the art. To illustrate, depending on the type of label used, the label can be attached to a terminal (5' or 3' end of an oligonucleotide primer and/or probe) ox a non-terminal nucleotide, and can be attached indirectly through linkers or spacer arms of various sizes and compositions. Using commercially available phosphoramidite reagents, one can produce oligonucleotides containing functional groups (e.g., thiols or primary amines) at either the 5' or 3' terminus via an appropriately protected phosphoramidite, and can label such oligonucleotides using protocols described in, e.g., Innis et al. (Eds.) PCR Protocols: A Guide to Methods and Applications, Elsevier Science & Technology Books (1990)(Innis), which is incorporated by reference.

Essentially any labeling moiety is optionally utilized to label a probe and/or primer by techniques well known in the art. In some embodiments, for ex: ample, labels comprise a fluorescent dye (e.g., a rhodamine dye (e.g., R6G, RI l O, TAMRA, ROX, etc.), a fluorescein dye (e.g., JOE, VIC, TET, HEX, FAM₅ etc.), a halofluorescein dye, a cyanine dye (e.g., CY3, CY3.5, CY5, CY5.5, etc.), a BODIPY® dye (e.g., FL, 530/550, TR, TMR, etc.), an ALEXA FLUOR® dye (e.g., 488, 532, 546, 568, 594, 555, 653, 647, 660, 680, etc.), a dictilororhodamine dye, an energy transfer dye (e.g., BIGD YE™ v 1 dyes, BIGDYE^TIM v 2 dyes, BIGDYE™ v 3 dyes, etc.), Lucifer dyes (e.g., Lucifer yellow, etc.), CASCADE BLUE®, Oregon Green, and the like. Additional examples of fluorescent dyes are provided in, e.g., Haugland, Molecular Probes Handbook of Fluorescent Probes and Research Products, Ninth Ed. (2003) and the updates thereto, which are each incorporated by reference. Fluorescent dyes are generally readily available from various commercial suppliers including, e.g., Molecular Probes, Inc. (Eugene, OR), Amersham Biosciences Corp. (Piscataway, NJ), Applied Biosystems (Foster City, CA), etc. Other labels include, e.g., biotin, weakly fluorescent labels (Yin et al. (2003) Appl Environ Microbiol. 69(7):3938, Babendure et al. (2003) Anal. BiocheriL 317(1):1, and Jankowiak et al. (2003) Chem Res Toxicol. 16(3):304), non-fluorescent labels, colorimetric labels, chemiluminescent labels (Wilson et al. (2003) Analyst. 128(5):480 and Roda et al. (2003) Luminescence 18(2):72), Raman labels, electrochemical labels, bioluminescent labels (Kitayama et al. (2003) Photochem Photobiol. 77(3):333, Arakawa et al. (2003) Anal. Biochiem. 314(2):206, and Maeda (2003) J. Pharm. Biomed. Anal. 30(6):1725), and an alpha- methyl-PEG labeling reagent as described in, e.g., U.S. Provisional Patent Application No. 60/428,484, filed on Nov. 22, 2002, which references are each incorporated by reference. Nucleic acid labeling is also described further below. In some embodiments, labeling is achieved using synthetic nucleotides (e.g., synthetic ribonucleotides, etc.) and/or recombinant phycoerythrin (PE).

In addition, whether a fluorescent dye is a label or a quencher is generally defined by its excitation and emission spectra, and the fluorescent dye with ^"which it is paired. Fluorescent molecules commonly used as quencher moieties in probes and primers include, e.g., fluorescein, FAM, JOE, rhodamine, R6G, TA]MRA, ROX, DABCYL, and EDANS. Many of these compounds are available from the commercial suppliers referred to above. Exemplary non-fluorescent or dark quenchers that dissipate energy absorbed from a fluorescent dye include the Black Hole Quenchers™ or BHQ™, which are commercially available from Biosearch Technologies, Inc. (Novato, CA, USA). To further illustrate, essentially any nucleic acid (and virtually any labeled nucleic acid, whether standard or non-standard) can be custom or standard ordered from any of a variety of commercial sources, such as The Midland Certified Reagent Company, The Great American Gene Company, ExpressGen Inc., Operon Technologies Inc., Proligo LLC, and many others. In certain embodiments, modified nucleotides are included in probes and primers.

To illustrate, the introduction of modified nucleotide substitutions into oligonucleotide sequences can, e.g., increase the melting temperature of the oligonucleotides. In some embodiments, this can yield greater sensitivity relative to corresponding unmodified oligonucleotides even in the presence of one or more mismatches in seqαience between the target nucleic acid and the particular oligonucleotide. Exemplary modified nucleotides that can be substituted or added in oligonucleotides include, e.g., C5-ethyl-dC, C5-methyl-dU, C5-ethyl-dU., 2,6- diaminopurines, C5-propynyl-dC, C7-propynyl-dA, C7-propynyl-dG, C5- propargylamino-dC, C5-propargylamino-dU, C7-propargylamino-dA, C7- propargylamino-dG, 7-deaza-2-deoxyxanthosine, pyrazolopyrimidine analogs, pseudo-dU, nitro pyrrole, nitro indole, 2'-0-m ethyl Ribo-U, 2'-0-methyl Ribo-C, an

8-aza-dA, an 8-aza-dG, a 7-deaza-dA, a 7-deaza-dG, N4-ethyl-dC, N6-metbιyl-dA, etc. To further illustrate, other examples of modified oligonucleotides include those having one or more LNA™ monomers. Nucleotide analogs such as these are also described in, e.g., U.S. Pat. No. 6,639,059, entitled "SYNTHESIS OF [2.2.I]BICYCLO NUCLEOSIDES," issued October 28, 2003 to Kochkine et al.,

U.S. Pat. No. 6,303,315, entitled "ONE STEP SAMPLE PREPARATION AND DETECTION OF NUCLEIC ACIDS IN COMPLEX BIOLOGICAL SAMPLES," issued October 16, 2001 to Skouv, and U.S. Pat. Application Pub. No. 2003/0092905, entitled "SYNTHESIS OF [2.2.I]BICYCLO NUCLEOSIDES," by Kochkine et al. that published May 15, 2003, which are each incorporated by reference. Oligorrucleotides comprising LNA™ monomers are commercially available through, e.g., Exiqon A/S (Vedbsek, DK). Additional oligonucleotide modifications are referred to herein, including in the definitions provided above.

ARRAY FORMATS In certain embodiments, oligonucleotide probes designed to hybridize with target nucleic acids are covalently or noncovalently attached to solid supports. In these embodiments, labeled amplicons derived from patient samples are typically contacted with these solid support-bound probes to effect hybridization and detection. In other embodiments, amplicons are attached to solid supports and contacted with labeled probes. Optionally, antibodies, aptamers, or other pxobe biomolecules utilized in a given assay are similarly attached to solid supports. Essentially any substrate material can be adapted for use as a solid support. In certain embodiments, for example, substrates are fabricated from silicon, glass, or polymeric materials (e.g., glass or polymeric microscope slides, silicon wafers, wells of microwell plates, etc.). Suitable glass or polymeric substrates, including microscope slides, are available from various commercial suppliers, such as Fisher

Scientific (Pittsburgh, PA, XJSA) or the like. In some embodiments, solid supports utilized in the invention are membranes. Suitable membrane materials are optionally selected from, e. g. polyaramide membranes, polycarbonate membranes, porous plastic matrix membranes (e.g., POREX® Porous Plastic, etc.), nylon membranes, ceramic membranes, polyester membranes, polytetrafluoroethylene

(TEFLON®) membranes, nitrocellulose membranes, or the like. Many of these membranous materials are widely available from various commercial suppliers, such as, PJ. Cobert Associates, Inc. (St. Louis, MO, USA), Millipore Corporation (Bedford, MA, USA), or trie like. Other exemplary solid supports that are optionally utilized include, e.g., ceramics, metals, resins, gels, plates, beads (e.g., magnetic microbeads, etc.), whiskers, fibers, combs, single crystals, self- assembling monolayers, and the like.

Nucleic acids are directly or indirectly (e.g., via linkers, such as bovine serum albumin (BSA) or the like) attached to the supports, e.g., by any available chemical or physical method. A wide variety of linking chemistries are available for linking molecules to a wide variety of solid supports. More specifically, nucleic acids may be attached to the solid support by covalent binding, such as by conjugation with a coupling agent or by non-covalent binding, such as electrostatic interactions, hydrogen bonds or antibody-antigen coupling, or by combinations thereof. Typical coupling agents include biotin/avidin, biotin/streptavidin, Staphylococcus aureus protein A/IgG antibody F_c fragment, and streptavidin/protein A chimeras (Sano et al. (1991) Bio/Technolo gy 9:1378, which is incorporated by reference), or derivatives or combinations of these agents. Nucleic acids may be attached to the solid support by a photocleavable bond, an electrostatic bond, a disulfide bond, a peptide bond, a diester bond or a combination of these bonds. Nucleic acids are also optionally attached to solid supports by a selectively releasable bond such as 4,4'-dimethoxytrityl or its derivative. Cleavable attachments can be created by attaching cleavable chemical moieties between the probes and the solid support including, e.g., an oligopeptide, oligonucleotide, oligopolyamide, oligoacrylamide, oligoethylene glycerol, alkyl chains of between about 6 to 20 carbon atoms, and combinations thereof. These moieties may be cleaved with, e.g., added chemical agents, electromagnetic radiation, or enzymes. Exemplary attachments cleavable by enzymes include peptide bonds, which can be cleaved by proteases, and phosphodiester bonds which can be cleaved by nucleases.

Chemical agents such as β-mercaptoethanol, dithiothreitol (DTT) and other reducing agents cleave disulfide bonds. Other agents which maybe useful include oxidizing agents, hydrating agents and other selectively active compounds. Electromagnetic radiation such as ultraviolet, infrared and visible light cleave photocleavable bonds. Attachments may also be reversible, e.g., using heat or enzymatic treatment, or reversible chemical or magnetic attachments. Release and reattachment can be performed using, e.g., magnetic or electrical fields.

A number of array systems have been described and can be adapted for use in the detection of target microbial nucleic acids. Aspects of array construction and use are also described in, e.g., Sapolsky et al. (1999) "High-throughput polymorphism screening and genotyping with high-density oligonucleotide arrays" Genetic Analysis: Biomolecular Engineering 14:187-192, Lockhart (1998) "Mutant yeast on drugs" Nature Medicine 4:1235-1236, Fodor (1997) "Genes, Chips and the Human Genome" FASEB Journal 11:A879, Fodor (1997) "Massively Parallel Genomics" Science 277: 393-395, and Chee et al. (1996) "Accessing Genetic Information with High-Density DNA Arrays" Science 274:610-614, all of which are incorporated by reference.

NUCLEIC ACID HYBRIDIZATION

The length of complementary region or sequence between primer or probes and their binding partners (e.g., target nucleic acids) should generally be sufficient to allow selective or specific hybridization of the primers or probes to the targets at the selected annealing temperatures used for a particular nucleic acid amplification protocol, expression profiling assay, etc. Although other lengths are optionally utilized, complementary regions of, for example, between about 10 and about 50 nucleotides (e.g., about 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, or 25 or more nucleotides) are typically used in a given application. "Stringent hybridization wash conditions" in the context of nucleic acid hybridization experiments, such as Southern and northern hybridizations, are sequence dependent, and are different under different environmental parameters. An extensive guide to the hybridization of nucleic acids is found in Tijssen (1993), supra, and in Hames and Higgins 1 and Hames and Higgins 2, supra. For purposes of the present invention, generally, "highly stringent" hybridization and wash conditions are selected to be about 5° C or less lower than the thermal melting point (T_m) for the specific sequence at a defined ionic strength and pH (as noted below, highly stringent conditions can also be referred to in comparative terms). The T_1n is the temperature (under defined ionic strength and pH) at which 50% of the test sequence hybridizes to a perfectly matched primer or probe. Very stringent conditions are selected to be equal to the T_m for a particular primer or probe.

The T_m is the temperature of the nucleic acid duplexes indicates the temperature at which the duplex is 50% denatured under the given conditions and its represents a direct measure of the stability of the nucleic acid hybrid. Thus, the T_m corresponds to the temperature corresponding to the midpoint in transition from helix to random coil; it depends on length, nucleotide composition, and ionic strength for long stretches of nucleotides.

After hybridization, unhybridized nucleic acid material can be removed by a series of washes, the stringency of which can be adjusted depending upon the desired results. Low stringency washing conditions (e.g., using higher salt and lower temperature) increase sensitivity, but can product nonspecific hybridization signals and high background signals. Higher stringency conditions (e.g., using lower salt and higher temperature that is closer to the hybridization temperature) lowers the background signal, typically with only the specific signal remaining. See, e.g., Rapley et al. (Eds.), Molecular Biomethods Handbook (Humana Press, Inc. 1998), which is incorporated by reference. Thus, one measure of stringent hybridization is the ability of the primer or probe to hybridize to one or more of the target nucleic acids (or complementary polynucleotide sequences thereof) under highly stringent conditions. Stringent hybridization and wash conditions can easily be determined empirically for any test nucleic acid.

For example, in determining highly stringent hybridization and wash conditions, the hybridization and wash conditions are gradually increased (e.g., by increasing temperature, decreasing salt concentration, increasing detergent concentration and/or increasing the concentration of organic solvents, such as formalin, in the hybridization or wash), until a selected set of criteria is met. For example, the hybridization and wash conditions are gradually increased until a target nucleic acid, and complementary polynucleotide sequences thereof, binds to a perfectly matched complementary nucleic acid. A target nucleic acid is said to specifically hybridize to a primer or probe nucleic acid when it hybridizes at least Vi as well to the primer or probe as to a perfectly matched complementary target, i.e., with a signal to noise ratio at least Vi as high as hybridization of the primer or probe to the target under conditions in which the perfectly matched primer or probe binds to the perfectly matched complementary target with a signal to noise ratio that is at least about 2.5x-10x, typically 5x-10x as high as that observed for hybridization to any of the unmatched target nucleic acids.

NUCLEIC ACID AMPLIFICATION

In some embodiments, RNA is converted to cDNA in a reverse-transcription (RT) reaction using, e.g., a target-specific primer complementary to the RNA for each gene target being monitored. Methods of reverse transcribing RNA. into cDNA are well known, and described in Sambrook, supra. Alternative methods for reverse transcription utilize thermostable DNA polymerases, as described in the art. As an exemplary embodiment, avian myeloblastosis virus reverse transcriptase (AMV- RT), or Maloney murine leukemia virus reverse transcriptase (MoMLV-RT) is used, although other enzymes are also optionally utilized. An advantage of using target-specific primers in the RT reaction is that only the desired sequences are converted into a PCR template. Superfluous primers or cDNA products are generally not carried into subsequent PCR amplifications. In another embodiment, RNA targets are reverse transcribed using non-specific primers, such as an anchored oligo-dT primer, or random sequence primers. An advantage of this embodiment is that the "unfractionated" quality of the mRNA sample is maintained because the sites of priming are non-specific, i.e., the products of this RT reaction will serve as template for any desired target in the subsequent PCR amplification. This allows samples to be archived in the form of DNA, which is more stable than RNA. In other embodiments, transcription-based amplification systems CTAS) are used, such as that first described by Kwoh et al. (Proc. Natl. Acad. Sci. (1989) 86(4):1173-7), or isothermal transcription-based systems such as 3SR (Self- Sustained Sequence Replication; Guatelli et al. (1990) Proc. Natl. Acad. Sci. 87:1874-1878) or NASBA (nucleic acid sequence based amplification; Kievits et al. (1991) J Virol Methods. 35(3):273-86), which are each incorporated by reference. In these methods, the mRNA target of interest is copied into cDNA by a reverse transcriptase. The primer for cDNA synthesis includes the promoter sequence of a designated DNA-dependent RNA polymerase 5' to the primer's region of homology with the template. The resulting cDNA products can then serve as templates for multiple rounds of transcription by the appropriate RNA polymerase. Transcription of the cDNA template rapidly amplifies the signal from the original target mRNA. The isothermal reactions bypass the need for denaturing cDNA strands from their RNA templates by including RNAse H to degrade RNA hybridized to DNA. In other exemplary embodiments, amplification is accomplished by used of the ligase chain reaction (LCR), disclosed in European Patent Application No. 320,308 (Backman and Wang), or by the ligase detection reaction (LDR), disclosed in U.S. Patent No. 4,883,750 (Whiteley et al.), which are each incorporated by reference. In LCR, two probe pairs are typically prepared, which are complimentary each other, and to adjacent sequences on both strands of the target. Each pair will bind to opposite strands of the target such that they abut. Each of the trwo probe pairs can then be linked to form a single unit, using a thermostable ligase. By temperature cycling, as in PCR, bound ligated units dissociate from the target, then both molecules can serve as "target sequences" for ligation of excess probe pairs, providing for an exponential amplification. The LDR is very similar to LCR. In this variation, oligonucleotides complimentary to only one strand of the target are used, resulting in a linear amplification of ligation products, since only the original target DNA can serve as a hybridization template. It is used following a PCR amplification of the target in order to increase signal.

In further embodiments, several methods generally known in the art would be suitable methods of amplification. Some additional examples include, but are not limited to, strand displacement amplification (Walker et al. (1992) ]Nucleic Acids

Res. 20:1691-1696), repair chain reaction (REF), cyclic probe reaction (REF), solid-phase amplification, including bridge amplification (Mehta and Singh (1999) BioTechniques 26(6): 1082-1086), rolling circle amplification (ICool, U.S. Patent No. 5,714,320), rapid amplification of cDNA ends (Frohman (1988) Proc. Natl. Acad. Sci. 85: 8998-9002), and the "invader assay" (Griffin et al. (1999) Proc.

Natl. Acad. Sci. 96: 6301-6306), which are each incorporated by reference. Amplicons are optionally recovered and purified from other reaction components by any of a number of methods well known in the art, including electrophoresis, chromatography, precipitation, dialysis, filtration, and/or centrifugation. Aspects of nucleic acid purification are described in, e.g., Douglas et al., DNA

Chromato graph y, Wiley, John & Sons, Inc. (2002), and Schott, Affinity Chromatography: Template Chromatography of Nucleic Acids and Proteins, Chromatographic Science Series, #27, Marcel Dekker (1984), both of which are incorporated by reference. In certain embodiments, amplicons are not purified prior to detection, such as when amplicons are detected simultaneoαis with amplification.

DATA COLLECTION

The number of species than can be detected within a mixture depends primarily on the resolution capabilities of the separation platform used, and the detection methodology employed. In some embodiments, separation steps are is based upon size-based separation technologies. Once separated, individual species are detected and quantitated by either inherent physical characteristics of the molecules themselves, or detection of an associated label.

Embodiments employing other separation methods are also described. For example, certain types of labels allow resolution of two species of the same mass through deconvolution of the data. Non-size based differentiation methods (such as deconvolution of data from overlapping signals generated by two different fluorophores) allow pooling of a plurality of multiplexed reactions to further increase throughput.

Separation Methods Certain embodiments of the invention incorporate a step of separating the products of a reaction based on their size differences. The PCR products generated during an amplification reaction typically range from about 50 to about 500 bases in length, which can be resolved from one another by size. Any one of several devices may be used for size separation, including mass spectrometry, any of several electrophoretic devices, including capillary, polyacrylamide gel, or agarose gel electrophoresis, or any of several chromatographic devices, including column chromatography, HPLC, or FPLC.

In some embodiments, sample analysis includes the use of mass spectrometry. Several modes of separation that determine mass are possible, including Time-of- Flight (TOF), Fourier Transform Mass Spectrometry (FTMS), and quadruple mass spectrometry. Possible methods of ionization include Matrix- Assisted Laser Desorption and Ionization (MALDI) or Electrospray Ionization (ESI). A preferred embodiment for the uses described in this invention is MALDI-TOF (Wu, et al. (1993) Rapid Communications in Mass Spectrometry 7:142-146, which is incorporated by reference). This method may be used to provide unfragmented mass spectra of mixed-base oligonucleotides containing between about 1 and about 1000 bases. In preparing the sample for analysis, the analyte is mixed into a matrix of molecules that resonantly absorb light at a specified wavelength. Pulsed laser light is then used to desorb oligonucleotide molecules out of the absorbing solid matrix, creating free, charged oligomers and minimizing fragmentation. An exemplary solid matrix material for this purpose is 3-hydroxypicolinic acid (Wu, supra), although others are also optionally used. In another embodiment, a microcapillary is used for analysis of nucleic acids obtained from the sample. Microcapillary electrophoresis generally involves the use of a thin capillary or channel, which may optionally be filled with a particular medium to improve separation, and employs an electric field to separate components of the mixture as the sample travels through the capillary. Samples composed of linear polymers of a fixed charge-to -mass ratio, such as DNA or RNA, will separate based on size. The high surface to volume ratio of these capillaries allows application of very high electric fields across the capillary without substantial thermal variation, consequently allowing very rapid separations. When combined with confocal imaging methods, these methods provide sensitivity in the range of attomoles, comparable to the sensitivity of radioactive sequencing methods. The use of microcapillary electrophoresis in size separation of nucleic acids has been reported in Woolley and Mathies (Proc. Natl. Acad. Sci. USA (1994) 91 :11348-11352), which is incorporated by reference. Capillaries are optionally fabricated from fused silica, or etched, machined, or molded into planar substrates. In many microcapillary electrophoresis methods, the capillaries are filled with an appropriate separation/sieving matrix. Several sieving matrices are known in the art that may be used for this application, including, e.g., hydroxyethyl cellulose, polyacrylamide, agarose, and the like. Generally, the specific gel matrix, running buffers and running conditions are selected to obtain the separation required for a particular application. Factors that are considered include, e.g., sizes of the nucleic acid fragments, level of resolution, or the presence of undenatured nucleic acid molecules. For example, running buffers may include agents such as urea to denature double-stranded nucleic acids in a sample.

Microfluidic systems for separating molecules such as DNA and RNA are commercially available and are optionally employed in the methods of the present invention. For example, the "Personal Laboratory System" and the "High Throughput System" have been developed by Caliper Lifesciences Corp. (Mountain View, CA). The Agilent 2100, which uses Caliper Lifesciences'

LabChip™ microfluidic systems, is available from Agilent Technologies (Palo Alto, CA, USA). Currently, specialized microfluidic devices, which provide for rapid separation and analysis of both DNA and RNA are available from Caliper Lifesciences for the Agilent 2100.

Other embodiments are generally known in the art for separating PCR amplification products by electrophoresis through gel matrices. Examples include polyacrylamide, agarose-acrylamide, or agarose gel electrophoresis, using standard methods (Sambrook, supra).

Alternatively, chromatographic techniques may be employed for resolving amplification products. Many types of physical or chemical characteristics may be used to effect chromatographic separation in the present invention, including adsorption, partitioning (such as reverse phase), ion-exchange, and size exclusion.

Many specialized techniques have been developed for their application including methods utilizing liquid chromatography or HPLC (Katz and Dong (1990) BioTechniques 8(5):546-55; Gaus et al. (1993) J. Immunol. Methods 158:229-236). In yet another embodiment, cDNA products are captured by their affinity for certain substrates, or other incorporated binding properties. For example, labeled cDNA products such as biotin or antigen can be captured with beads bearing avidin or antibody, respectively. Affinity capture is utilized on a solid support to enable physical separation. Many types of solid supports are known in the art that would be applicable to the present invention. Examples include beads (e.g. solid, porous, magnetic), surfaces (e.g. plates, dishes, wells, flasks, dipsticks, membranes), or chromatographic materials (e.g. fibers, gels, screens). Certain separation embodiments entail the use of microfluidic techniques. Technologies include separation on a microcapillary platform, such as designed by ACLARA BioSciences Inc. (Mountain View, CA), or the LabChip™ microfluidic devices made by Caliper Lifesciences Corp. Another technology developed by

Nanogen, Inc. (San Diego, CA), utilizes microelectronics to move and concentrate biological molecules on a semiconductor microchip. The microfluidics platforms developed at Orchid Biosciences, Inc. (Princeton, NJ), including the Chemtel™ Chip, which provides for parallel processing of hundreds of reactions, can also be used in certain embodiments. These microfluidic platforms require only nanoliter sample volumes, in contrast to the microliter volumes required b y other conventional separation technologies. Some of the processes usually involved in genetic analysis have been miniaturized using microfluidic devices. For example, PCT publication WO 94/05414 reports an integrated micro-PCR apparatus for collection and amplification of nucleic acids from a specimen. U.S. Patent Nos. 5,304,487 (Wilding et al.) and 5,296,375 (Kricka et al.) discuss devices for collection and analysis of cell-containing samples. U.S. Patent No. 5,856,174 (Lipshutz et al.) describes an apparatus that combines the various processing and analytical operations involved in nucleic acid analysis. Each of these references is incorporated by reference. Additional technologies are also contemplated. For example, Kasianowicz et al. (Proc. Natl. Acad. Sci. USA (1996) 93:13770-13773, which is incorporated by reference) describes the use of ion channel pores in a lipid bllayer membrane for determining the length of polynucleotides. In this system, an electric field is generated by the passage of ions through the pores. Polynucleotide lengths are measured as a transient decrease of ionic current due to bloclcage of ions passing through the pores by the nucleic acid. The duration of the current decrease was shown to be proportional to polymer length. Such a system can be applied as a size separation platform in certain embodiments of the present invention. Primers are useful both as reagents for hybridization in solution, such as priming PCR amplification, as well as for embodiments employing a solid phase, such as microarrays. With microarrays, sample nucleic acids such as mRNA or DNA are fixed on a selected matrix or surface. PCR products may be attached to the solid surface via one of the amplification primers, then denatured to provide single- stranded DNA. This spatially-partitioned, single-stranded nucleic acid is then subject to hybridization with selected probes under conditions that allow a quantitative determination of target abundance. In this embodiment, amplification products from each individual reaction are not physically separated, but are differentiated by hybridizing with a set of probes that are differentially labeled. Alternatively, unextended amplification primers may be physically immobilized at discreet positions on the solid support, then hybridized with the products of a nucleic acid amplification for quantitation of distinct species within the sample, hi this embodiment, amplification products are separated by way of hybridization with probes that are spatially separated on the solid support. Separation platforms may optionally be coupled to utilize two different separation methodologies, thereby increasing the multiplexing capacity of reactions beyond that which can be obtained by separation in a single dimension. For example, some of the RT-PCR primers of a multiplex reaction may be coupled with a moiety that allows affinity capture, while other primers remain unmodified. Samples are then passed through an affinity chromatography column to separate PCR products arising from these two classes of primers. Flow-through fractions are collected and the bound fraction eluted. Each fraction may then be further separated based on other criteria, such as size, to identify individual components. Detection Methods

Following separation of the different products of a multiplex amplification, one or more of the amplicons are detected and/or quantitated. Some embodiments of the methods of the present invention enable direct detection of products. Other embodiments detect reaction products via a label associated with one or more of the amplification primers. Many types of labels suitable for use in the present invention are known in the art, including chemiluminescent, isotopic, fluorescent, electrochemical, inferred, or mass labels, or enzyme tags. In further embodiments, separation and detection may be a multi-step process in which samples are fractionated according to more than one property of the products, and detected one or more stages during the separation process.

An exemplary embodiment of the invention that does not use labeling or modification of the molecules being analyzed is detection of trie mass-to-charge ratio of the molecule itself. This detection technique is optionally used when the separation platform is a mass spectrometer. An embodiment for increasing resolution and throughput with mass detection is in mass-modifying the amplification products. Nucleic acids can be mass-modified through either the amplification primer or the chain-elongating nucleoside triphosphates. Alternatively, the product mass can be shifted without modification of the individual nucleic acid components, by instead varying the number of bases in the primers. Several types of moieties have been shown to be coinpatible with analysis by mass spectrometry, including polyethylene glycol, halogens, alkyl, aryl, or aralkyl moieties, peptides (described in, for example, U.S. Patent No. 5,691,141, which is incorporated by reference). Isotopic variants of specified atoms, such as radioisotopes or stable, higher mass isotopes, are also used to vary the mass of the amplification product. Radioisotopes can be detected based on the energy released when they decay, and numerous applications of their use are generally known in the art. Stable (non-decaying) heavy isotopes can be detected based on the resulting shift in mass, and are useful for distinguishing between two amplification products that would otherwise have similar or equal masses. Other embodiments of detection that make use of inherent properties of the molecule being analyzed include ultraviolet light absorption (UV) or electrochemical detection. Electrochemical detection is based on oxidation or reduction of a chemical compound to which a voltage has been applied. Electrons are either donated (oxidation) or accepted (reduction), which can be monitored as current. For both UV absorption and electrochemical detection, sensitivity for each individual nucleotide varies depending on the component base, but with molecules of sufficient length this bias is insignificant, and detection levels can be taken as a direct reflection of overall nucleic acid content.

Some embodiments of the invention include identifying molecules indirectly by detection of an associated label. A number of labels may be employed that provide a fluorescent signal for detection. If a sufficient quantity of a given species is generated in a reaction, and the mode of detection has sufficient sensitivity, then some fluorescent molecules may be incorporated into one or more of the primers used for amplification, generating a signal strength proportional to the concentration of DNA molecules. Several fluorescent moieties, including Alexa 350, Alexa 430, AMCA, BODIPY 630/650, BODIPY 650/665, BODIPY-FL, BODIPY-R6G, BODIPY-TMR, BODIPY-TRX, carboxyfluorescein, Cascade

Blue, Cy3, Cy5, 6-FAM, Fluorescein, HEX, 6- JOE, Oregon Green 488, Oregon Green 500, Oregon Green 514, Pacific Blue, REG, Rhodamine Green, Rhodamine Red, ROX, TAMRA, TET, Tetramethylrhodamine, and Texas R.ed, are generally known in the art and routinely used for identification of discrete nucleic acid species, such as in sequencing reactions. Many of these dyes ha~ve emission spectra distinct from one another, enabling deconvolution of data from incompletely resolved samples into individual signals. This allows pooling of^" separate reactions that are each labeled with a different dye, increasing the throughput during analysis, as described in more detail below. Additional examples of suitable labels are described herein.

The signal strength obtained from fluorescent dyes can be enhanced through use of related compounds called energy transfer (ET) fluorescent dyes. After absorbing light, ET dyes have emission spectra that allow them to serve as "donors" to a secondary "acceptor" dye that will absorb the emitted light and emit a lower energy fluorescent signal. Use of these coupled-dye systems can significantly amplify fluorescent signal. Examples of ET dyes include the ABI PRISM BigDye terminators, recently commercialized by Perkin- Elmer Corporation (Foster City,

CA, USA) for applications in nucleic acid analysis. These chromophores incorporate the donor and acceptor dyes into a single molecule and an energy transfer linker couples a donor fluorescein to a dichlororhodamine acceptor dye, and the complex is attached, e.g., to a primer. Fluorescent signals can also be generated by non-covalent intercalation of fluorescent dyes into nucleic acids after their synthesis arid prior to separation. This type of signal will vary in intensity as a function of the length of the species being detected, and thus signal intensities must be normalized based on size. Several applicable dyes are known in the art, including, boat not limited to, ethidium bromide and Vistra Green. Some intercalating dyes, such, as YOYO or TOTO, bind so strongly that separate DNA molecules can each be bound with a different dye and then pooled, and the dyes will not exchange between DNA species. This enables mixing separately generated reactions in order to increase multiplexing during analysis. Alternatively, technologies such as the use of nanocrystals as a fluorescent DNA label (Alivisatos, et al. (1996) Nature 382:609-11, which is incorporated by reference) can be employed in the methods of the present invention. Another method, described by Mazumder, et al. (Nucleic Acids Res. (1998) 26:1996-2000, which is incorporated by reference), describes hybridization of a labeled oligonucleotide probe to its target without physical separation from unhybridized probe. In this method, the probe is labeled with a chemiluminescent molecule that in the unbound form is destroyed by sodium sulfite treatment, but is protected in probes that have hybridized to target sequence.

In other embodiments, both electrochemical and infrared methods of detection can be amplified over the levels inherent to nucleic acid molecules through attachment of EC or IR labels. Their characteristics and use as labels are described in, for example, PCT publication WO 97/27327, which is incorporated by reference. Some preferred compounds that can serve as an IR label include an aromatic nitrile, aromatic alkynes, or aromatic azides. Numerous compounds can serve as an EC label; many are listed in PCT publication WO 97/27327. Enzyme-linked reactions are also employed in the detecting step of the methods of the present invention. Enzyme-linked reactions theoretically yield an infinite signal, due to amplification of the signal by enzymatic activity. In this embodiment, an enzyme is linked to a secondary group that has a strong binding affinity to the molecule of interest. Following separation of the nucleic acid products, enzyme is bound via this affinity interaction. Nucleic acids are then detected by a chemical reaction catalyzed by the associated enzyme. Various coupling strategies are possible utilizing well-characterized interactions generally known in the art, such as those between biotin and avidin, an antibody and antigen, or a sugar and lectin. Various types of enzymes can be employed, generating colorimetric, fluorescent, chemiluminescent, phosphorescent, or other types of signals. As an illustration, a primer may be synthesized containing a biotin molecule. After amplification, amplicons are separated by size, and those made with the biotinylated primer are detected by binding with streptavidin that is covalently coupled to an enzyme, such as alkaline phosphatase. A subsequent chemical reaction is conducted, detecting bound enzyme by monitoring the reaction product. The secondary affinity group may also be coupled to an enzymatic substrate, which is detected by incubation with unbound enzyme. One of skill in the art can conceive of many possible variations on the different embodiments of detection methods described above. In some embodiments, it may be desirable prior to detection to separate a subset of amplification products from other components in the reaction, including other products. Exploitation of known high-affinity biological interactions can provide a mechanism for physical capture. Some examples of high-affinity interactions include those between a hormone with its receptor, a sugar with a lectin, avidin and biotin, or an antigen with its antibody. After affinity capture, molecules are retrieved by cleavage, denaturation, or eluting with a competitor for binding, and then detected as usual by monitoring an associated label. In some embodiments, the binding interaction providing for capture may also serve as the mechanism of detection.

Furthermore, the size of an amplification product or products are optionally changed, or "shifted," in order to better resolve trie amplification products from other products prior to detection. For example, chemically cleavable primers can be used in the amplification reaction. In this embodiment, one or more of the primers used in amplification contains a chemical linkage that can be broken, generating two separate fragments from the primer. Cleavage is performed after the amplification reaction, removing a fixed number of nucleotides from the 5' end of products made from that primer. Design and use of such primers is described in detail in, for example, PCT publication WO 96/37630, which is incorporated by reference.

DATA ANALYSIS

For reliably detecting leukemia it is generally desirable to determine the expression of more than one of the markers described herein. As an exemplary criterion for the choice of markers, the statistical significance of markers as expressed in q oxp values based on the concept of the false discovery rate is optionally determined. In doing so, a measure of statistical significance called the q value is associated with each tested feature. The q value is similar to the p value, except it is a measure of significance in terms of the false discovery rate rather than the false positive rate

(see, e.g., Storey et al. (2003) Proc.Natl.Acad.Sci. 100:9440-5, which is incorporated by reference).

In some embodiments, the markers described herein have ^-values of less than about 3E-06, typically less than about 1.5E-09, more typically less than about 1.5E- 11, even more typically less than about 1.5E-20, and still more typically less than about 1.5E-30. Of the markers described or referred to herein, the expression level of at least about two, typically of at least about ten, more typically of at least about 25, and even more typically of at least about 50 of these markers is determined as described herein or by another technique known to those of skill in the art. In some 5 embodiments, for example, expression levels of one or more markers listed in

Tables I-XIII are determined in a given sample. In certain embodiments, expression levels of each of these genes in a sample is determined and compared with expression levels detected in one or more reference cells (e.g., non-leukemic cells, leukemic cells, particular types or subtypes of leukemic cells, etc.). . 0 Furthermore, the International Publication No. WO O3/039443, which is incorporated by reference, discloses certain marker genes the expression levels of which are characteristic for certain leukemia. Certain of the markers and/or methods disclosed therein are optionally utilized in performing the methods described herein.

L 5 The level of the expression of a marker is typically indicative of the type of cell under consideration. The level of expression of a marker or group of markers is measured and is generally compared with the level of expression of the same marker or the same group of markers from other cells or samples. The comparison may be effected in an actual experiment or in silico. There is a meaningful

20 difference in these levels of expression, e.g., when these expression levels (also referred to as expression pattern, expression signature, or expression profile) are measurably different. In some embodiments, the difference is typically at least about 5%, 10% or 20%, more typically at least about 50% or may even be as high as 75% or 100%. To further illustrate, the difference in the level of expression is

25 optionally at least about 200%, i.e., two fold, at least about 500%, i.e., five fold, or at least about 1000%, i.e., 10 fold in some embodiments.

In certain embodiments, for example, the expression, level of markers expressed lower in a first subtype than in at least one second subtype, which differs from the first subtype, is at least about 5%, 10% or 20%, more typically at least about 50% 30 or may even be about 75% or about 100%, more typically at least about 10-fold, even more typically at least 50-fold, and still more typically at least about 100-fold lower in the first subtype. On the other hand, the expression level of markers expressed higher in a first subtype than in at least one second subtype, which differs from the first subtype, is at generally least about 5%, 10% or 20%, more generally at least about 50% or may even be about 75% or about 100%, more generally at least 10-fold, still more generally at least about 50-fold, and even more generally at least about 100-fold higher in the first subtype.

The classification accuracy of a given gene list for a set of microarray experiments is preferably estimated using Support Vector Machines (SVM), because there is evidence that SVM-based prediction slightly outperforms other classification techniques, such as k-Nearest Neighbors (k-NN). The LIBSVM software package version 2.36, for example, is optionally used (SVM-type: SVC, linear kernel (http://www.csie.ntu.edu.tw/-cjlin/libsvrn/)). Machine learning algorithms are also described in, e.g., Brown et al. (2000) Proc.Natl.Acad.Sci.. 97:262-267, Furey et al. (2000) Bioinformatics, 16:906-914, and Vapnik, Statistical Learning Theory. Wiley ( 1998), which are each incorporated by reference.

To further illustrate, the classification accuracy of a given gene list for a set of microarray experiments can be estimated using Support Vector Machines (SVM) as supervised learning techniques. Generally, SVMs are trained using differentially expressed genes, which were identified on a subset of the data and then this trained model is employed to assign new samples to those trained groups from a second and different data set. Differentially expressed genes are optionally identified, e.g., applying analysis of variance (ANOVA) and t-test-statistics (Welch t-test). Based on identified distinct gene expression signatures, respective training sets consisting of, e.g., 2/3 of cases and test sets with 1/3 of cases to assess classification accuracies can be designated. Assignment of cases to training and test sets is optionally randomized and balanced by diagnosis. Based on the training set, a Support Vector Machine (SVM) model can be built using this approach.

The apparent accuracy of prediction, i.e., the overall rate of correct predictions of the complete data set can be estimated by, e.g., lOfold cross validation. This process typically includes dividing the data set into 10 approximately equally sized subsets, training an S VM-model for 9 subsets, and generating predictions for the remaining subset. This training and prediction, process can be repeated 10 times to include predictions for each subset. Subsequently the data set can be split into a training set, consisting of two thirds of the samples, and a test set with the remaining one third. Apparent accuracy for trie training set can also be estimated

5 by lOfold cross validation (analogous to apparent accuracy for complete set). An

SVM-model of the training set is optionally biαilt to predict diagnosis in the independent test set, thereby estimating true accuracy of the prediction model. This prediction approach can be applied both for overall classification (multi-class) and binary classification (diagnosis X => yes or no). For the latter, sensitivity and

O specificity are optionally calculated, as follows:

Sensitivity = (number of positive samples predicted)/(number of true positive) Specificity = (number of negative samples predicted)/(number of true negatives).

SYSTEMS FOR GENE EXPRESSION ANALYSIS

The present invention also provides systems For analyzing gene expression. The

L 5 system includes one or more probes that correspond to at least portions of genes or expression products thereof. The genes are selected from, e.g., the markers listed in Tables I-XIII. In some embodiments, for example, the probes are nucleic acids (e.g., oligonucleotides, cDNAs, cRNAs, etc.), whereas in other embodiments, the probes are biomolecules (e.g., antibodies, aptrners, etc.) designed to detect

ZO expression products of the genes (e.g., proteins or fragments thereof). In certain embodiments, the probes are arrayed on a solid support, whereas in others, they are provided in one or more containers, e.g., for assays performed in solution. The system also includes at least one reference data bank or database for correlating detected expression levels of polynucleotides or polypeptides in target cells, which

25 polynucleotides or polypeptides detectably bind to one or more of the probe biomolecules, with the target cell being a leukemia cell or a non-leukemia cell. In some embodiments, the reference data bank is backed up on a computational data memory chip or other computer readable medium, which can be inserted in as well as removed from system of the present invention, e.g., like an interchangeable 0 module, in order to use another data memory chip containing a different reference data bank. In certain embodiments, the systems also include detectors (e.g., spectrometers, etc.) that detect binding between the probes and targets. Other detectors are described further below. In addition, the systems also generally include at least one controller operably connected to the reference data bank and/or to the detector. In some embodiments, for example, the controller is integral with the reference data bank.

The systems of the present invention that include a desired reference data bank can be used in a way such that an unknown sample is, first, subjected to gene expression profiling, e.g., by microarray analysis in a manner as described herein or otherwise known to person skilled in the art, and the expression level data obtained by the analysis are, second, fed into the system and compared with the data of the reference data bank obtainable by the above method. For this purpose, the apparatus suitably contains a device for entering the expression level of the data, for example, a control panel such as a keyboard. The results, whether and how the data of the unknown sample fit into the reference data bank can be made visible on a monitor or display screen and, if desired, printed out on an incorporated of connected printer. Computer components are described further below.

In some embodiments, a system optionally further includes a thermal modulator operably connected to containers to modulate temperature in the containers (e.g., to effect thermocycling when target nucleic acids are amplified in the containers), and/or fluid transfer components (e.g., automated pipettors, etc.) that transfer fluid to and/or from the containers. Optionally, these systems also include robotic components for translocating solid supports, containers, and the like, and/or separation components (e.g., microfluidic devices, chromatography columns, etc.) for separating the products of amplification, reactions from one another. The invention further provides a computer or computer readable medium that includes a data set that comprises a plurality of character strings that correspond to a plurality of sequences (or subsequences thereof) that correspond genes selected from, e.g., the markers listed in Tables I-XIII. Typically, the computer or computer readable medium further includes an automatic synthesizer coupled to an output of the computer or computer readable medium. The automatic synthesizer accepts instructions from the computer or computer readable medium, which instructions direct synthesis of, e.g., one or more probe nucleic acids that correspond to one or more character strings in the data set.

Detectors are structured to detect detectable signals produced, e.g., in or proximal to another component of the system (e.g., in container, on a solid support, etc.). Suitable signal detectors that are optionally utilized, or adapted for use, in these systems detect, e.g., fluorescence, phosphorescence, radioactivity, absorbance, refractive index, luminescence, or the like. Detectors optionally monitor one or a plurality of signals from upstream and/or downstream of the performance of, e.g., a given assay step. For example, the detector optionally monitors a plurality of optical signals, which correspond in position to "real time" results. Example detectors or sensors include photomultiplier tubes, CCD arrays, optical sensors, temperature sensors, pressure sensors, pH sensors, conductivity sensors, scanning detectors, or the like. Each of these as well as other types of sensors is optionally readily incorporated into the systems described herein. Optionally, the systems of the present invention include multiple detectors.

More specific exemplary detectors that are optionally utilized in these systems include, e.g., a resonance light scattering detector, an emission spectroscope, a fluorescence spectroscope, a phosphorescence spectroscope, a luminescence spectroscope, a spectrophotometer, a photometer, and the like. Various synthetic components are also utilized, or adapted for, use in the systems of the invention including, e.g., automated nucleic acid synthesizers, e.g., for synthesizing the oligonucleotides probes described herein. Detectors and synthetic components that are optionally included in the systems of the invention are described further in, e.g., Skoog et al., Principles of Instrumental Analysis, 5^th Ed., Harcourt Brace College Publishers (1998) and Currell, Analytical Instrumentation: Performance

Characteristics and Quality, John Wiley & Sons, Inc. (2000), both of which are incorporated by reference.

The systems of the invention also typically include controllers that are operably connected to one or more components (e.g., detectors, synthetic components, thermal modulator, fluid transfer components, etc.) of the system to control operation of the components. More specifically, controllers are generally included either as separate or integral system components that are utilized, e.g., to receive data from detectors, to effect and/or regulate temperature in the containers, to effect and/or regulate fluid flow to or from selected containers, or the like. Controllers and/or other system components is/are optionally coupled to an appropriately programmed processor, computer, digital device, or other information appliance

(e.g., including an analog to digital or digital to analog converter as needed), which functions to instruct the operation of these instruments in accordance with preprogrammed or user input instructions, receive data and information from these instruments, and interpret, manipulate and report this information to the user. Suitable controllers are generally known in the art and are available from various commercial sources.

Any controller or computer optionally includes a monitor which is often a cathode ray tube ("CRT") display, a flat panel display (e.g., active matrix liquid crystal display, liquid crystal display, etc.), or others. Computer circuitry is often placed in a box, which includes numerous integrated circuit chips, such as a microprocessor, memory, interface circuits, and others. The box also optionally includes a hard disk drive, a floppy disk drive, a high capacity removable drive such as a writeable CD-ROM, and other common peripheral elements. Inputting devices such as a keyboard or mouse optionally provide for input from a user. These components are illustrated further below.

The computer typically includes appropriate software for receiving user instructions, either in the form of user input into a set of^" parameter fields, e.g., in a GUI, or in the form of preprogrammed instructions, e.g., preprogrammed for a variety of different specific operations. The software then converts these instructions to appropriate language for instructing the operation of one or more controllers to carry out the desired operation. The computer then receives the data from, e.g., sensors/detectors included within the system, and interprets the data, either provides it in a user understood format, or uses that data to initiate further controller instructions, in accordance with the programming, e.g., such as controlling fluid flow regulators in response to fluid weight data received from weight scales or the like. The computer can be, e.g., a PC (Intel x86 or Pentium chip-compatible DOS™, OS2™, WINDOWS™, WINDOWS NT™, WINDOWS95™, WINDOWS98™, WINDOWS2000™, WINDOWS XP™, LINUX-based machine, a MACINTOSH™, Power PC, or a UNIX-based (e.g., SUN™ work station) machine) or other common commercially available computer which is known to one of skill. Standard desktop applications such as word processing software (e.g., Microsoft Word™ or Corel WordPerfect™) and database software (e.g.., spreadsheet software such as Microsoft Excel™, Corel Quattro Pro™, oar database programs such as Microsoft Access™ or Paradox™) can be adapted to tlie present invention. Software for performing, e.g., controlling temperature modulators and fluid flow regulators is optionally constructed by one of skill using a standard programming language such as Visual basic, Fortran, Basic, Java, or the like.

Reference data banks can be produced by, e.g., (a) compiling a gene expression profile of a patient sample by determining the expression level of genes listed in one or more of Tables I-XIII, and (b) classifying the gene expression profile using a machine learning algorithm. Exemplary machine learning algorithms are optionally selected from, e.g., Weighted Voting, K-Nearest Neighbors, Decision Tree Induction, Support Vector Machines (SVM), and Feed-Forward Neural Networks. In some embodiments, for example, the machine learning algorithm is an SVM, such as polynomial kernel, linear kernel, and Gaussian Radial Basis

Function-kernel SVM models.

KITS

The present invention also provides kits that include at least one probe a.s described herein for distinguishing between leukemic and non-leukemic cells. The kits also include instructions for correlating detected expression levels of one or more polynucleotides or polypeptides in a target cell, which polynucleotides or polypeptides detectably bind to one or more of the probe biomolecules, "with the target cell being a leukemia cell or a non-leukemia cell Typically, the kit includes suitable auxiliaries, such as buffers, enzymes, labeling compounds, andAor the like. In some embodiments, probes are attached to solid supports, e.g. the wells of microtiter plates, nitrocellulose membrane surfaces, glass surfaces, to particles in solution, etc. As another option, probes are provided free in solution in containers, e.g., for performing the methods of the invention in a solution phase. In certain embodiments, kits also contain at least one reference for a leukemia and/or non- leukemia cell. For example, the reference can be a sample, a database, or the like. In some embodiments, the kit includes primers and other reagents for amplifying target nucleic acids. Typically, kits also include at least one container for packaging the probes, the set of instructions, and any other included components.

EXAMPLES

It is understood that the examples and embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the claimed invention. It is also understood that various modifications or changes in light the examples and embodiments described herein will be suggested to persons skilled in the art and are to be included within the spirit and purview of this application and scope of the appended claims.

EXAMPLE 1: DIAGNOSIS OF LEUKEMIA USING GENE EXPRESSION

PROFILING

INTRODUCTION

The diagnosis and classification of leukemias rely on the simultaneous application of multiple techniques. Cytomorphology and histomorphology are combined with cytochemistry and multiparameter flow cytometry in order to assign the diagnostic sample to the correct entity. Furthermore, chromosomal analysis supplemented in many cases by fluorescence in situ hybridization (FISH) and molecular techniques such as polymerase chain reaction (PCR) are needed to definitely confirm the supposed diagnosis. A comprehensive and standardized algorithm for a diagnostic work-flow and an effective and carefully designed combination of methods is essential to guarantee that all of the required diagnostic information is gathered.

This huge amount of laboratory assessment is not only necessary to diagnose and classify leukemia samples correctly but also results in the discovery of major clinically relevant implications. Thus, the detailed leukemia classification according to FAB(I ;2), which has been used for nearly three decades, and its improvement by thoroughly defined genetic and other characteristics, resulted in the new WHO classification(3) and leads to new prognostic markers and even to disease-specific therapeutic approaches. The prime example for this strong link between a comprehensive diagnosis and a disease-specific treatment approach has been the use of all-trans retinoic acid (ATRA) in patients with acute promyelocyte leukemia — both the correct diagnosis and the efficacy of the specific treatment are based on the presence of the translocation (15; 17) and of the corresponding PML/RARA fusion gene(4-6). Although it did not result in development of a new targeted drug therapy the identification of acute myeloid leukemia (AML) with a complex aberrant karyotype, depending on the age of the patient, is nonetheless highly relevant for the management of the patient. Depending on the age of the patient this very dismal diagnosis is the basis for the decision to apply allogeneic stem cell transplantation very early or to even withhold any anti-leukemic therapy(7-10). The recent introduction of imatinib into the therapeutic management of patients with chronic myeloid leukemia (CML) has revolutionized the treatment strategies in this disease and may change therapeutic concepts also for BCR/ AB L- positive acute lymphoblastic leukemias (ALL) in the near future(l 1-16). Again, the basis for both the correct diagnosis and the specifically targeted therapy is the presence of the genetic alteration, the t(9;22) translocation. In addition, the BCR/ ABL fusion gene is increasingly used to sensitively assess response to therapy by monitoring minimal residual disease (MRD) levels(17). Also in acute leukemias and chronic lymphatic leukemias (CLL), monitoring of MRD is increasingly used to guide risk-adapted therapy( 18-22).

To achieve the correct and complete diagnosis in each analyzed sample it is necessary for a modem and state-of-the-art laboratory to provide significant resources with regard to laboratory equipment and working time as well as skilled and experienced personnel. The availability of a novel diagnostic tool which provides the opportunity to satisfy diagnostic needs and is at the same time an efficient use of resources is necessary.

Microarray analyses used to perform gene expression profiling may be the method of choice in this regard(23-25). Microarrays allow the simultaneous detection of the expression of nearly all human genes in one experimental approach and thereby provide a maximized insight into gene regulation and alterations present in the analyzed sample on the transcriptional level. While many studies exploited this technique to gain clues to the pathogenesis of a large variety of malignant diseases and to characterize not yet identified disease subentities, little focus has been put an the issue of applying gene expression profiling for diagnostic purposes(26-28).

This example describes the results of an extensive microarray study on leukemia samples which has been performed in parallel to all standard techniques and whicb. results in a global one-step diagnostic approach. In 937 samples from patients withi newly diagnosed leukemia and normal bone marrow from healthy donors the analysis focused on all leukemic subentities which are clinically relevant with respect to specific treatment approaches and prognostication. Using unsupervised and supervised biostatistical methods mostly based on support vector machines (SVM) the analysis confirmed and reproduced 12 predefined leukemia subtypes and separated all of these from each other and from healthy bone marrow samples with an accuracy of 95.1 %. Thus, the single method, gene expression profiling using microarrays, may effectively be applied as a complementary diagnostic method in leukemia or as a substitute for other methods.

METHODS

DESCRIPTION OF PATIENTS AND SAMPLES A total of 937 bone marrow or peripheral blood samples of patients with newly diagnosed leukemias were included into the analysis. In general, CLL samples comprised peripheral blood and samples from all other entities comprised bone marrow. The samples were sent to the Laboratory for Leukemia Diagnostics in Munich, Germany, between February 1998 and February 2004 for reference diagnosis from local and national hospitals. The median shipment time was 1 day

(range, 0 to 3 days). All samples underwent a standardized processing including sample central registration(29), preparation, and evaluation by cytomorphology(30), cytochemistry, multiparameter immunophenotyping(31), cytogenetics(32), fluorescence in situ hybridization (FISH), and molecular genetics(33). Stabilized cell lysates were stored at -8O⁰C for a median of 13 months (range, 0 to 67 months). The different types of diagnoses as well as other sample and patient characteristics are given in table 1. There were specific inclusion criteria besides the respective sample including one of the targeted subgroups (see below). After 800 samples had been identified for inclusion in the study an additional 137 samples were selected according to their diagnosis in order to achieve a distribution among the different disease subtypes to ensure an adequately powered study. In all cases with balanced translocations the corresponding fusion transcript was verified on the molecular level, i.e. PML/RARA for t(15;17), AML1/ETO for t(8;21), CBFB/MYH11 for inv(16)/t(16;16), MLL/various partner genes for t(l Iq23) in both AML and ALL, and BCR/ ABL for t(9;22) in both ALL and CML. In addition, in each of these cases as well as for MYC/IGH for t(8;14), cases FISH was applied using standard procedures(34).

TABLE 1 : SAMPLE AND PATIENT CHARACTERISTICS (TOTAL=937) median range

Shipment time (days) 1 0-3

Storage time at -80⁰C (months) 13 0-67

Patient age (years) 57 16-90

AML patients 61 18-90

ALL patients 46 16-86

CML patients 49 21-82

CLL patients 63 36-84 non-leukemia cases 45 18-83

Sex (male/female) 53%/47%

WBC count (G/l) 28.8 0.4-514

Bone marrow ^"blasts (acute leukemias only)* 85% 10%-100%

AML Total 620 (66%) t(15;17) 42 (4%) t(8;21) 38 (4%) inv(16)/t(16;16) 49 (5%) t(l l q23) 47 (5%)

Complex aberrant 75 (8%) other abnormalities 176 (19%) normal karyotype 193 (21%) ALL Total 152 (16%)

Pro-B-ALL/t(l lq23) 26 (3%) c-ALL/Pre-B-ALL with t(9;22) 42 (4%) c-ALL/Pre-B-ALL without t(9;22) 40 (4%) mature B-ALL/t(8;14) 12 (1%) cortical T-ALL 20 (2%) immature T-ALL 12 (1%)

CML, chronic phase 75 (8%)

CLL 45 (5%)

Non-leukemia 45 (5%)

Threshold for definition of AML according to WH0(3) is a bone marrow blast count of at least 20% which may be even lower by definition, however, if recurrent balanced translocations are present.

DESCRIPTION OF LEUKEMIA ENTITIES SELECTED TO BE IDENTIFIED BY GENE EXPRESSION PROFILING

A focus of this analysis was to identify all leukemia subgroups which are clinically relevant with regard to both specific treatment and prognostication. Besides the distinction between the four main categories of leukemia, i.e. AML, ALL, CML, and CLL, these relevant groups also comprise specifically defined subentities. In addition to the group designated "non-leukemia", which comprised healthy bone marrow, reactive bone marrow conditions, vitamin B 12 or iron deficiency, or idiopathic thrombocytopenic purpura, the following 12 clinically relevant subgroups were analyzed:

1. AML with t(l 5; 17), 2. AML with t(8;21),

3. AML with inv(16),

4. AML with normal karyotype or so-called "other" cytogenetic abnormalities,

5. AML with 11 q23/MLL rearrangement,

6. AML with complex aberrant karyotype, 7. Pro-B-ALL/t(l lq23), 8. mature B-ALL/t(8; 14),

9. c-ALL/Pre-B-ALL with or without t(9;22),

10. T-ALL,

11. CML, 12. CLL.

GENE EXPRESSION PROFILING

Microarray analyses were performed utilizing the GeneChip® System (Affyαmetrix, Santa Clara, USA) and the HG-Ul 33 microarray set. This two-array set provides comprehensive coverage of Λvell-substantiated genes in the human genome. It can be used to analyse the expression level of 39,000 transcripts and variants, including greater than 33,000 well-characterized human genes. The two arrays comprise more than 44,000 probe sets and 1,000,000 distinct oligonucleotide features. For gene expression profiling cell lysates of the leukemia samples were thawed, homogenized (QIAshredder, Qiagen), and total RNA was extracted (RNeasy Mini Kit, Qiagen). The subsequent target preparation steps as well as hybridization, washing and staining of the probe arrays were performed according to recommended protocols (Affymetrix Technical Manual). The Affymetrix software package (Microarray Suite 5.0.1) extracted fluorescence intensities from each element on the microarrays as detected by confocal laser scanning. Detection calls (present, marginal, or absent) were determined by default parameters(35). Signal intensity values were calculated by scaling the raw data intensities to a common target intensity (Ul 33 mask file; TGT value: 5000). Each human GeneChip expression array features 100 human maintenance genes that serve as a tool to normalize and scale the data before performing data comparisons. As recommended by the manufacturer, these 100 probe sets were used for normalization (http://www.affymetrix.com/support/technical/mask_files. affix).

DATA ANAJLYSIS

Gene expression data was preprocessed according to Affymetrix recommendations and analyzed as follows: For each disease entity differentially expressed genes were calculated by means oft-test-statistic (two-sample t-test, unequal variances) in a one-versus-all (OVA) approach. The software package R version 1.7.1 (http://www.r-project.org/) was applied. To address the multiple testing problem, false discovery rates (FDR) of genes were calculated according to Storey et al(36). Class prediction was performed using support vector machines (SVMC)(37), because there is evidence that SVM-based prediction slightly outperforms other classification techniques(38;39). Prediction accuracy was estimated by 10-fold cross validation and assessed for robustness in a resampling approach..

More specifically, a SVM is a supervised learning algorithm developed over the past decade by Vapnik et al(79) and has also recently been used for gene expression data analysis(80-83). The SVM algorithm operates by mapping the given training set of samples into a possibly high-dimensional feature space and attempting to locate in that space a plane that separates the positive from the negative examples. Having found such a plane, the SVM can then predict the classification of an unlabeled example by mapping it into the feature space and asking on which side of the separating plane the example lies.

In this analysis, multi-class SVM classifiers were built with linear kernels based on class-specific genes using library LIBSVM version 2.36 (http://www.csie.ntu.edu.tw/~cjlin/libsvm/). Apparent accuracy of the complete data set was estimated by 10 fold cross validation. This means that the data set was divided into 10 approximately equally sized subsets, an SVM-model was trained for 9 subsets and predictions were generated for the remaining subset. This training / prediction process was repeated 10 times to include predictions for each subset. Apparent accuracy is the overall rate of correct predictions. Sensitivity and specificity were calculated- as follows:

- Sensitivity = (number of positive samples predicted)/(number of tnxe positives)

- Specificity = (number of negative samples predicted)/(number of true negatives)

A resampling approach was applied to assess robustness of class prediction: The data set was randomly, but balanced by the leukemia subtype, split into a training set, consisting of two thirds of samples, and an independent test set with the remaining one third. Differential genes were identified in the training set (t-test- statistic, OVA), an SVM-model was built from the training set and predictions were made in the test set. This complete process was repeated 100 times. By this means, 95% confidence intervals were estimated for accuracy, sensitivity and specificity.

REAL TIME PCR

The expression of the top three to four differentially expressed genes from each subgroup (total, n=42) that had been analyzed on the microarray was reevaluated by quantitative real time PCR using Applied Biosystems 7900HT Micro Fluidic Cards Configuration 2 and the ABI PRISM 7900 HT Sequence Detection System

(Applied Biosystems, Foster City, CA, USA). A 1 μg aliquot of total RNA isolated for microarray analysis was reverse transcribed using hexanucleotide random priming and 300 U of Superscript II enzyme (Invitrogene, Karlsruhe, Germany). 100 ng of the cDNA were amplified using the TaqMan Universal Master Mix (Applera, Darmstadt, Germany) according to the manufacturer's recommendations.

A calibrator sample was prepared using a mix of mononucleated blood cells from 10 healthy volunteers. Results were given as fold changes compared to this calibrator.

Relative gene expression values were obtained using the comparative Cj method (ΔΔCT) for relative quantification(40). By use of the ΔΔCT method the expression quantity was expressed relative to the calibrator sample that was used as the basis for comparative results. Therefore, the calibrator is the Ix sample and all other quantities are expressed as an n-fold difference relative to the calibrator. Calculations were done by the SDS2.1 software (Applied Biosystems). In addition, to equalize for uneven cDNA loading as a maintainance gene GAPDH was assessed in parallel. RESULTS

TECHNICAL ASPECTS ON SAMPLE TARGET PREPARATION AND SCAN QUALITY

A total of 965 target preparations for gene expression profiling were performed yielding sufficient cRNA, i.e. > 20 μg, after in vitro transcription, for hybridization to microarrays. These were hybridized to Ul 33 A and B microarrays.

Each scan was visually inspected. In 28 (2.9%) cases samples were excluded that did not meet a combination of the following stringent criteria: (i) percentage of present called probe sets of the Ul 33 A array > 30%, (ii) low 375 'ratio of GAPDH probe sets (normally less than 3.0), and (iii) no scan artefacts detected, i.e. bubbles, scratches, or high background.

For the remaining 937 (97.1%) samples included in this analysis, the median value of the percentage of present called genes was 46.3% (U133A) and 31.3% (U133 B), respectively, the median 375 'ratio of GAPDH probe sets was 1.65 (Ul 33A) and 1.87 (Ul 33B), respectively.

PREDICTION OF 13 SUBGROUPS

Predicting the respective leukemia type or subtype based on differential gene expression signatures was approached using Support Vector Machines (SVM). The complete data set was randomly, but equally split into training and independent test cohorts for the 13 different subgroups. Then differentially expressed genes were identified in the training set, calculated by means oft-test-statistic, and a SVM model was built based on the genes that demonstrate differential expression between the respective subclasses in the training set.

This SVM model was used to predict samples in the test cohort. The application of the top 100 genes per group resulted in best prediction accuracies (superior to top

20, 50, 150, 200, 250, and 300 genes, respectively; data not shown) and were used for all subsequent analyses. Descriptions of these differentially expressed genes identified in one-versus-all (OVA) comparisons are provided in Tables I-XIII, in which the genes are listed according to the corresponding Affymetrix probe set description, gene symbol and title, q-value(78), and t-test statistic. The gene symbols and titles provided in Tables I-XIII were annotated using the NetAffx. analysis center (April 2004 release date)(84). Table 2 represents a confusion matrix of subgroup predictions based on their gene expression signature using a 10-fold cross validation approach (9/10 for training and 1/10 for testing, 10 iterations so that each sample is classified once). Overall, a 95.1% accuracy of subgroup prediction has been achieved analyzing 13 subgroups. Specifically, the highest accuracy was achieved for seven of the 13 subgroups, i.e. AML with t(15;17), 100% accurate predictions; AML with inv(16), 98.0%; CLL, 97.8%; CML, 97.3%; AML normal/other, 97.3%; Pro-B-ALL/t(l Iq23), 96.2%; AML with t(8;21), 94.7%. For the other six subgroups the percentage of accurate predictions ranged between 83.3% and 93.3%.

Most of the niisclassifications occurred in subgroups that either had relatively low sample numbers or which are characterized by a high intra-subgroup biologic heterogeneity. The first aspect clearly applies to mature B-ALL/t(8;14) with a sample number of 12 and 83.3% accurate predictions. The latter aspect is reflected in AML with t(l Iq23) (89.4% accurate predictions) with balanced translocations involving the MLL gene and six different fusion partner genes (AF4, AF6, AFQ, AFlO, ELL, ENL). Another example of biologic heterogeneity is AML with complex aberrant karyotype (88.0% accurate predictions) comprised of a wide range of three to 30 chromosomal abnormalities (median, 9). As anticipated, most of the misclassifϊcations of these groups (4 out of 5 for AML with t(l Iq23) and 8 out of 9 for AML with complex aberrant karyotype) were due to a prediction of the samples as AML normal/other. A third aspect to consider is the relative similarity of distinct subgroups with regard to specific characteristics for instance flow cytometrically detected expression of myeloid antigens on immature T-ALL cases(41). Probably due to these complexities, 4 out of 32 cases with T-ALL in the present series were classified as AML normal/other. TABLE 2: 13 GROUP PREDICTION CONFUSION MATRIX DETERMINED BY 10-FOLD CROSS VALIDATION

-4

Note, that the classification on the basis of routine diagnostics, including cytomorphology, cytochemistry, immunophenotyping, cytogenetics, and molecular genetics, is given in columns. The predicted classification using gene expression profiling is given in rows.

Thus, there are 891 out of 937 cases (95.1%) which were correctly classified.

In order to assess the robustness of class prediction a resampling approach was applied, i.e. the complete SVM classification procedure was repeated 100 times. For each of the 100 runs all samples were randomly divided into a training set (2/3 of all samples, n=625) and a test set (1/3 of all samples, n=312). Thus, the test set for each run consisted of n=28 c-ALL/Pre-B-ALL, n=9 Pro-B-ALL/t(l Iq23), n=4 mature B-ALL/t(8;14), n=10 T-ALL, n=14 AML with t(15;17), n=12 AML with t(8;21), n=16 AML with invlό, n=16 AML with t(l Iq23), n=25 AML witli complex karyotype, n=123 AML with normal karyotype or other aberrations, n=15 CLL, n=25 CML, and n=15 non-leukemia samples, respectively. The matrix in Table 3 gives the average number of class predictions as determined after 100 runs of SVM-based classifications. For example, n=9 Pro-B-ALL/t( 1 Iq23) samples were predicted by the algorthim n=900 times (each sample 100 times). Of the 900 predictions the class label Pro-B-ALL/t(l Iq23) was assigned correctly 854 times, i.e. on average 8.54 per run. In 2 individual predictions, Pro-B-ALL/t(l Iq23) samples were predicted as c-ALL/Pre-B-ALL, in 44 predictions as AML with normal karyotype or other aberrations, respectively.

Confirming the data obtained by 10-fold cross validation, the overall median accuracy amounts to 93.8% (95% confidence interval: 91.4% to 95.8%). Io particular and similar to the 10-fold cross validation approach, a very high degree of accurate predictions was achieved in seven of the 13 subgroups, i.e. AML with t(15;17), 100% median accuracy; AML with inv(16), 98.1%; CLL, 97.5%; CML, 95.3%; AML normal/other, 96.8%; Pro-B-ALL/t(l lq23), 94.9%; AML with t(8;21), 95.3%. For the other six subgroups the median prediction accuracies ranged between 64.3% and 92.3%. Thus, the results obtained for the subgroups by applying the resampling approach are highly consistent with those obtained by 10- fold cross validation and strongly confirm the capability of gene expression profiling to predict leukemia subtypes. The reasons for the misclassifications are most likely the same as those described above, in particular the relatively 1 ow sample number of cases with mature B-ALL/t(8;14). The sensitivities and specificities of the predictions for each of the 13 subclasses are given in Table 4. According to the accuracy data given above, the specificity overall is very high, more than 99% for all but one subgroup. Since most misclassified samples were classified as AML normal/other, the specificity of this subgroup was slightly lower than for other subgroups and amounted to 93.65%. The median sensitivity ranged between 75% and 100% for all subgroups due to the reasons outlined above for the results of the 10-fold cross validation.

TABLE 3: CONFUSION MATRIX FOR 13 SUBTYPES AS DETERMINED BY RESAMPLING USING SVM

-4

Note, that the matrix shows the predicted class as determined after 100 runs of SVM-based classifications. Average numbers of predictions per run are given. The total data set (n=937) was randomly separated into a training set (n=625) and a test set (n=312) for each of the 100 runs. Data given are the average numbers of respective classifications.

TABLE 4: SENSITIVITIES AND SPECIFICITIES FOR LEUKEMIA CLASSIFICATION USING SVM IN 13 SUBGROUPS

-4

Note, that the overall median accuracy is 93.8% (95% confidence interval: 91 4% to 95.8%)

CLUSTER ANALYSIS OF 13 SUBGROUPS

To further validate the findings described above cluster analyses (CA) and principal component analyses (PCA) have been performed for the 13 groups analyzed as well as for the paired comparison of selected groups. The CA of all of the analyzed samples reflects the clearly differing gene expression patterns of the 13 groups resulting in a highly accurate separation of this large and comprehensive series of 937 samples (Figure 1). Applying three-dimensional PCA (Figure 2), the power of the gene expression profile-based leukemia classification is demonstrated by the clear separation of T-precursor ALL from c-ALL/Pre-B-ALL (with or without t(9;22)/BCR/ABL). Similarly, three-dimensional PCA provides a cl ear distinction between both t(9;22)-positive entities, CML and c-ALL/Pre-B-AXL (Figure 3). Interestingly, the one sample of t(9;22)-positive c-ALL/Pre-B-AIX shown in the proximity of the CML samples is characterized by only 50% leukemic bone marrow infiltration; thus, the normal hematopoiesis present in this sample, which is largely myelomonocytic, and the forced assignment to either of the two groups are the likely reasons for this result.

IDENTIFICATION OF CORTICAL T-ALL AND B- PRECURSOR ALL WITH T(9;22)

The classification capabilities of this approach was further refined by identifying the clinically distinct entities, c-ALL/Pre-B-ALL with t(9;22) and cortical T-ALL, out of the groups classified as c-ALL/Pre-B-ALL and T-ALL, respectively.

The cluster analysis (Figure 4) shows that the majority of the cases fall either into the branch of c-ALL/Pre-B-ALL without t(9;22) or into the branch of c-ALL/Pre- B-ALL with t(9;22) (supporting gene list for figure 4 is contained in table KIV). The remaining 21 samples (26%) fall into a third branch characterized by a gene expression profile clearly differing from the other two groups. Accordingly_^ the 10- fold cross validation analysis (allowing the separation into two groups only) reveals an accuracy of^" 82.9%. Importantly, misclassifications occured in both directions, i.e. cases with t(9;22) were classified as without it and vice versa. Resampling of the training and test sets, respectively, applying 100 runs of SVM-based classification (median accuracy, 77.8%, range, 61.0% to 90.8%) indicated that these misclassifications are not limited to distinct samples, i.e. the percentages of misclassifi cation per sample range from 3.1% to 88.1%, probably reflecting a significant overlap of gene expression signatures between both groups or being due to the presence of a clinically not yet identified third group of c-ALL/Pre-B- ALL. The separation of cortical T-ALL samples from immature T-ALL samples is very clear as shown in the cluster analysis (Figure 5) (supporting gene list for figure 5 is contained in table XV). Two samples of immature T-ALL show a gene expression profile slightly different from the other immature T-ALL cases as visualized by cluster analysis (Figure 5). In fact, these two samples are the ones lying nearest to the cortical T-ALL samples in the PCA. According to the relative vicinity of these two samples to samples of cortical T-ALL, the accuracy of the 10-fold cross validation is 84.38% and resampling applying 100 runs of SVM-based classification results in a median accuracy of 80.00% (range, 60.00% to 100%).

The separation of samples with AML and normal karyotypes from those with AML and "other" cytogenetic aberrations was not approached in this example since the prognosis of both subgroups was identical. This is illustrated in Figure 6 showing that both event-free survival and overall survival are identical for these two subgroups when applying a standardized treatment approach(30;42;43). Thus, there was no clinical relevance or need for the distinction between these two groups in this analysis.

FALSE DISCOVERY RATE

The false discovery rate is an accepted methodology to calculate statistical significance in microarray studies(77,78). A measure of statistical significance called the q value is associated with each tested feature taking into account that thousands of genes are simultaneously tested. The q value of a particular feature in a microarray data set is the expected proportion of false positives incurred when calling that feature significant. VALIDATION OF SIGNIFICANT GENES BY REAL-TIME

PCR

A total of 42 genes was reevaluated by real time PCR with micro fluidic cards. For some genes more than one probe set was represented on the microarray. Probe sets generating only absent calls on the array were excluded for this quantitative comparison. Comparisons were performed by Spearman's rank correlation. As illustrated in Table 5, which includes results from the one group versus all comparison, highly significant correlations of both methods could be demonstated for 41 of the 42 analyzed genes for all probe sets analyzed. The lack of correlation for RAD21 most probably was due to a saturation of the probe sets on the mircoarray because all intensity signals assessed with this method were extremely high. For most of the genes real time PCR showed a greater dynamic range in comparison to microarray analysis. Some exemplary comparisons are plotted in Figure 7 demonstrating that both methods generally lead to concordant results.

TABLE 5: CORRELATION OF GENE EXPRESSION AS ASSESSED BY

MICROARRAY AND RT-PCR EXPERIMENTS

DISCUSSION

Diagnosing and classifying leukemias is a clinically highly relevant task wlhich requires a comprehensive and well-structured approach in the laboratory to guarantee the appropriateness of the results. Significant resources with regard to time, well-trained and skilled personnel and laboratory space and equipment are needed to cover this approach. Furthermore, the interlaboratory reproducibility of the currently applied diagnostic methods, i.e. cytomorphology, cytochemistry, immunophenotyping, cytogenetics, and molecular genetics, ranges between only 56% and 90% in experienced hands and therefore clearly needs improvement(9;44- 48). Gene expression profiling using microarray technology has the potential of optimizing leukemia diagnostics and overcoming the above mentioned short¬ comings of current methods.

This example focused on the identification of all clinically relevant subtypes of leukemia by gene expression profiling and resulted in finding significant differences for intra-group separation within the four main leukemia groups, i.e. AML, ALL, CML, and CLL. With regard to AML, more than 50 different recurrent cytogenetic abnormalities have been described. However, reliable data on their prognostic impact are available only for the most frequent ones. These include t(15;17), t(8;21), and inv(16), which are associated with a favorable outcome, and complex aberrant karyotypes and t(l Iq23) carrying an unfavorable prognosis(8- 10;32;49). The remaining cases, i.e. normal karyotypes and so-called other cytogenetic abnormalities, have an intermediate prognosis. Figure 6 demonstrates that the separation between these subgroups results in highly differing prognoses supporting the clinical relevance of the selection of AML subgroups in the present study. In addition, the relative distribution of the analyzed cases with regard to these subgroups reflects that of all patients diagnosed in our laboratory confirming that there is no selection bias. Also, the age distribution of the analyzed cohort is very similar to the true age distribution of patients with AML as well as with the other diseases analyzed. With regard to clinical relevance, similar characteristics apply to the different entities of ALL. Besides the separation of T-precursor ALL from B-precursor ALL, clinically relevant because of different treatment strategies, it is important to identify those with Pro-B-ALL and t(l Iq23), c-ALL or Pre-B- ALL and t(9;22), as well as mature B-ALL and t(8;14). These subentities differ highly with respect to prognostic impact and require substantially differing therapies which is true for mature B-ALL in particular(l 1 ;50). Because it was known that the gene expression profiles of ALL with and without t(9;22), respectively, are difficult to distinguish from other cases with t(9;22) it was not approached in the first step but in a second step of the analyses. The overall smaller numbers of cases with CLL, CML, and "non-leukemia" cases was chosen because these entities are biologically and clinically more homogeneous as compared to the acute leukemia cases discussed above.

This analysis demonstrates, inter alia, a very high degree of accuracy for the correct assignment of bone marrow samples to all clinically relevant subgroups of leukemia and to normal bone marrow, respectively. An important aspect for the achievement of this accuracy was the careful and comprehensive use of standard methods to characterize all of the samples before they underwent microarray analysis. Besides the use of cytomorphology and cytochemistry the samples were processed applying immunophenotyping, cytogenetics, and molecular genetics in order to allow the identification of subtype-specific gene expression patterns and to exclude any misclassifϊcation of samples or overlaps between the subcategories focused on in this microarray analyses. In AML, the detection of six different subgroups was approached in this analysis. For the classification of AML with t(15;17), t(8;21), and inv(16), respectively, the highest degree of accuracy was achieved with 42 out of 42, 36 out of 38, and 48 out of 49 correct assignments by 10-fold cross validation and an average number of correct predictions of 14 out of 14, 11.43 out of 12, and 15.7 out of 16, respectively, by resampling. Accordingly, all of the median sensitivities and specificities were 100%. This is in line with previous reports describing a unique biologic background for these subentities(51-53) which is reflected in distinct gene expression profiles(54-57). However, since the latter have not yet been assessed by microarray analysis in the context of the full spectrum of AML and the other leukemias, the present analysis adds important information by clearly demonstrating that based on their distinct features these subentities can be accurately predicted even in the context of the very heterogeneous background of other leukemias. The other three AML subgroups, AML with t(l Iq23)/MLL, AML with complex karyotype, and AML normal/other, have a more heterogeneous biology. This is reflected by different partner genes of the MLL gene and an overall heterogeneity with regard to cytogenetic and molecular genetic aberrations, respectively. With these complexities in mind, it was anticipated that misclassifications would occur. Importantly, however, out of the 24 misclassifications (total, 491 classifications) in these subgroups during 10-fold cross validation only four were misclassifications into the non-AML subgroups. As a consequence, while the median specificities for AML with t(l Iq23) and for AML with complex aberrant karyotype were very high (99.66% and 99.65%, respectively) the median specificity for AML normal/other of 93.65% points at the need for further improvements of the applied method or for the use of supplemental analyses in these cases. In particular, this is true since a small number of samples with ALL (n=l 1), CLL (n=l), CML (n=2), and non-leukemia (n=2) were classified into this subgroup by 10-fold cross validation. With the exception of the three samples described above there have been no misclassifi cations in CLL and CML by 10-fold cross validation. Accordingly, there were 14.62 out of 15 and 23.82 out of 25 correct assignments, respectively, by resampling in these entities. As a result, the median sensitivities (100% and 96.00%) and the clinically most important median specificities (100% and 99.65%) were very high for these distinct disease entities.

Each of the four subgroups of ALL analyzed could be classified with a high median accuracy (99.65% for c-ALL/Pre-B-ALL, 100% for the other subgroups). As discussed above, most of the misclassifi cation (11 out of 13) occured into the group AML normal/other. These cases did not feature the immunophenotype of an aberrant expression of myeloid antigens which is often observed in ALL cases.

Since previous studies reported on difficulties in separating c-ALL/Pre-B-ALL cases with t(9;22) from other B-precursor ALL cases resulting in prediction accuracies of 80%(26) the approach in this example was to include c-ALL/Pre-B- ALL cases combined as one subgroup irrespective of the presence of t(9;22) into the analysis and to separate cases positive for t(9;22) from those without in a second step. While the separation of c-ALL/Pre-B-ALL cases from the other entities was straight forward, difficulties were observed in separating t(9;22)- positive cases from negative ones with the accuracy being only 82.9%. The cluster analysis demonstrates that the majority of cases are accurately classified in one of the two categories, however, a third branch becomes evident revealing a gene expression pattern distinct from the further two groups. The hypothesis that a further and not yet identified genetic lesion could be responsible for this third branch has been discarded due to cluster analysis and SVM not revealing a reproducible gene expression pattern different from the other two groups (data not shown). Furthermore, the use of SVM with differentially expressed genes selected based on the comparison of only the first two more homogeneous groups did not result in a more accurate assignment of samples of the third group either (data not shown). Taken together, this supports the concept that BCR/ ABL represents a type 1 mutation(58) and down-stream pathways are shared by many other master genes. Thus, the gene expression profile of BCR/ABL-positive ALL cases is not highly reproducible and future microarray-based diagnostic tools should include oligonucleotides targeting the bcr/abl fusion transcript in order to accurately predict this entity with a higher accuracy which is anticipated to also increase the sensitivities and specificities for the classification of the other subgroups. Another clinically relevant subgroup has been approached also in a second step. After separation of T-ALL from all other entities immature T-ALL has been discriminated from cortical T-ALL, which in the clinical setting is characterized by a favorable prognosis(41). Again, the separation of both entities has been highly accurate with the exception of two samples that originally were classified by immunophenotyping as immature T-ALL. It is importaat to note that the definition of cortical T-ALL in this context is based only on the po>sitivity for CDla(59), while other T-cell markers like CD7, CD2, CD5, CD4, or CD8 may be positive in either subgroup. Intriguingly, while the use of CDIa is a diagnostic standard the present analysis suggests that in the two misclassified cases the overall gene expression profile is very similar to the cortical T-ALL signature. Thus, these two cases may rather be cortical T-ALL featuring an aberrant lack of CDIa expression than truely immature T-ALL. As one consequence of this analysis, it may be considered to base the classification of cortical T-ALL not only on the positivity for CDIa but to include other markers such as PAWR(6O). Further implications may be gained from analyzing the cellular function of differentially expressed genes. It is known that dexametliasone leads to a downregulation of CARD4(6\) which encodes a pro-apoptotically acting protein(62;63). Since CARD4 is highly expressed in cortical T-ALL corticoid therapy may be less effective in this entity as compared to immature T-ALL. However, clinical studies may be needed to prove this h/ypothesis.

One exemplary issue which has not yet been substantially addressed in other microarray studies(64-66) is the identification of non-lexαkemic bone marrow and its discrimination from all leukemia subtypes. In the present analysis, 42 out of 45 non-leukemia samples have been predicted accurately while one sample was classified as AML with t(l Iq23) and two as AML normal/other by 10-fold cross validation. Accordingly, the median accuracy applying resampling is 13.23 out of 15. Importantly, the median specificity for non-leukemia is 99.6% while the sensitivity is 90.0%. Thus, until improvements of the applied methods are achieved which better characterize the heterogeneous subgroup of AML normal/other it seems appropriate to add conventional methods, if the microarray analysis result assigns a sample to the latter subgroup. In contrast, due to its high specificity the result "non-leukemia" can be the basis to exclude the presence of leukemia in a given sample analyzed.

In general, there are two strategies to handle the occurrence of misclassifications obtained by microarray analysis. The first one is to identify the most frequent false positive result, i.e. the subgroup with the lowest specificity, and to add conventional diagnostics to confirm or revise a malignant diagnosis. Clearly, this applies for AML normal/other with a median specificity of 93.7% (range, 90.2% to 96.6%). By the use of cytochemistry, immunophenotyping, and cytogenetics the discrimination of this subgroup from c-ALL/Pre-B-ALL, AML with t(llq23), and AML with complex aberrant karyotype is straight forward although resource- consuming. Another possible application for additional methods is the use of PCR to identify or exclude the presence of the BCBJABL fusion gene once c-ALL/Pre-B- ALL is diagnosed.

The second and more promising strategy would be an improvement of the capabilities of the microarray technology by taking advantage of the additional representation of fusion gene-specific oligonucleotides. By this approach, many of the misclassifications should be avoidable, e.g. c-ALL/Pre-B-ALL with t(9;22) should TDe identifyable by the detection of BCBJABL as should AML with t(l Iq23) by the detection of fusion genes involving MLL and various partners(67). Following this approach would potentially result in even higher accuracies in the subgroups discussed above as well as improving accuracies in ottxer subgroups.

There axe even more subgroups, in particular in AML, which have been suggested to feature a homogeneous biologic background with potential impact on the clinical course of patients being affected by these entities(68). Examples are mutations of CEBPJ4(69), length mutations of FLT3(33), and partial tandem duplications of MLL(IO). However, since this evidence is still under evalution in clinical trials these sixb groups are not within the focus of the present study.

There is a growing body of published microarray studies addressing the identification of specific gene expression profile in distinct subentities of leukemia. Along this line, the respective groups of AML with recurrent balanced translocations as well as AML with trisomy 8 have been described to carry a typical genetic signature which in some cases is highly specific(28;55;71 ;72). The present analysis followed these important studies and in addition provided the opportunity, by focusing on all clinically relevant subtypes of chronic and acute leukemias in a single comprehensive approach, to build on these signatures for a highly accurate diagnostic tool capable of predicting leukemia subtypes. In addition, the separation of leukemia samples from samples with non-malignant diseases and healthy samples has been accomplished. In accordance to these analyses of leukemias, it is anticipated that similar approaches can also b>e taken to diagnose and classify myelodysplastic syndromes and lymphomas(66;73;74).

A future scenario which may result from this study includes the wide-spiead use of micorarray technology applying a carefully designed and comprehensive leukemia diagnostic microarray which allows a significant improvement of current standard diagnostics by strengthening the diagnostic accuracy and by more efficient use of resources. Furthermore, it is envisioned that this technology will provide significant insights into the specific genetic alterations of distinct entities and thereby will allow the detection of novel markers which can be targeted by PCR- based methods and multiparameter flow cytometry to quantify minimal xesidual disease during the course of anti-leukemic treatment(75). The identification of progno stic markers or marker constellations providing the opportunity to predict the response to anti-leukemic treatment is another clinically relevant topic which evolves and will be covered by future microarray trials(57;76). EXAMPLE 2: ADDITIONAL INFORMATION REGARDING GENERAL MATERIALS. METHODS AND DEFINITIONS OF FUNCTIONAL ANNOTATIONS

[0001] The methods section contains both information on statistical analyses used for identification of differentially expressed genes and detailed annotation data of identified microarray probe sets.

AFFYMETRIX PROBESET ANNOTATION

All annotation data of GeneChip® arrays are extracted from the NetAffx™ Analysis Center (internet website: www.affymetrix.com). Files for Ul 33 set arrays, including Ul 33 A and U133B microarrays are derived from the June 2003 release. The original publication refers to: Liu et al. (2003) "NetAffx: Affymetrix probe sets and annotations," Nucleic Acids Res. 3 l(l):82-6, which is incorporated by reference.

The sequence data are omitted due to their large size, and because they do not change, whereas the annotation data are updated periodically, for example new information on chromosomal location and functional annotation of the respective gene products. Sequence data are available to download in the NetAffx Download Center on the world wide web at affymetrix.com.

DATA FIELDS In the following section, the content of each field of the data files is described. Microarray probe sets, for example, found to be differentially expressed between different types of leukemia samples are further described by additional information. The fields are of the following types:

1. GeneChip Array Information 2. Probe Design Information

3. Public Domain and Genomic References

1. GeneChip Array Information HG-U133 ProbeSetJD:

HG-U133 ProbeSet_ID describes the probe set identifier. Examples are: 200007 _at, 20001 l_s_at,200012_x_at. GeneChip

The description of the GeneChip® probe array name where the respective probe set represented. Examples are: Afϊymetrix Human Genome U 133 A Array or Affymetrix Human Genome U133B Array. 2. Probe Design Information

Sequence Type The Sequence Type indicates whether the sequence is an Exemplar, Consensus or

Control sequence. An Exemplar is a single nucleotide sequence taken directly from a public database. This sequence could be an. mRNA or an expressed sequence tag (EST). A Consensus sequence is a nucleotide sequence assembled by Affymetrix, based on one or more sequence taken from a public database.

Transcript ID: The cluster identification number with a sub-cluster identifier appended.

Sequence Derived From: The accession number of the single sequence, or representative sequence on which the probe set is based. Refer to the "Sequence Source" field to determine the database used.

Sequence ID: For Exemplar sequences: Public accession number or GenBank identifier. For Consensus sequences: Affymetrix identification number or public accession number.

Sequence Source The database from which the sequence used to design this probe set was taken.

Examples are: GenBank®, RefSeq, UniGene, TIGR (annotations from The Institute for Genomic Research).

3. Public Domain and Genomic Rejferences Most of the data in this section is from the LocusLink and UniGene databases, and are annotations of the reference sequence on which the probe set is modeled.

Gene Symbol and Title: A gene symbol and a short title, when one is available. Such symbols are assigned by different organizations for different species. Affymetrix annotational data comes from the UniGene record. There is no indication which species-specific databank was used, but some of the possibilities include for example HUGO: The Human Genome Organization.

MapLocation: The map location describes the chromosomal location when one is available.

Unigene_Accession: UniGene accession number and cluster type. Cluster type can be "full length" or

"est", or "---" if unknown.

LocusLink: This information represents the LocusLink accession number.

Full Length Ref. Sequences Indicates the references to multiple sequences in RefSeq. The field contains the ID and description for each entry, and there can be multiple entries per probeSet.

EXAMPLE 3; ADDITIONAL INFORMATION REGARDING SAMPLE PREPARATION, PROCESSING AND DATA ANALYSIS

Method 1 : Microarray analyses were performed utilizing the GeneChip® System (Affymetrix,

Santa Clara, USA). Hybridization target preparations were performed according to recommended protocols (Affymetrix Technical Manual). More specifically, at time of diagnosis, mononuclear cells were purified by Ficoll-Hypaque density centrifugation. They had been lysed immediately in RLT buffer (Qiagen, Hilden, Germany), frozen, and stored at -8O⁰C from 1 week to 38 months. For gene expression profiling cell lysates of the leukemia samples were thawed, homogenized (QIAshredder, Qiagen), and total RNA was extracted (RNeasy Mini Kit, Qiagen). Subsequently, 5-10 μg total RNA isolated from 1 x 10⁷ cells was used as starting material for cDNA synthesis with oligo[(dT)₂₄T7promotor]₆₅ primer (cDNA Synthesis System, Roche Applied Science, Mannheim, Germany). cDNA products were purified by phenol/chloroform/IAA extraction (Ambion, Austin, TX, USA) and acetate/ethanol-precipitated overnight. For detection of the hybridized target nucleic acid biotin-labeled ribonucleotides were incorporated during the following in vitro transcription reaction (Enzo BioArray High Yield RNA Transcript Labeling Kit, Enzo Diagnostics). After quantification by spectrophotometric measurements and 260/280 absorbance values assessment for quality control of the purified cRNA (RNeasy Mini Kit, Qiagen), 15 μg cRNA was fragmented by alkaline treatment (200 mM Tris-acetate, pH 8.2/500 mM potassium acetate/150 mM magnesium acetate) and added to the hybridization cocktail sufficient for five hybridizations on standard GeneChip® microarrays (300 μL final volume). Washing and staining of the probe arrays was performed according to the recommended Fluidics Station protocol (EukGE-WS2v4). Affymetrix Microarray Suite software (version 5.0.1) extracted fluorescence signal intensities from each feature on the microarrays as detected by confocal laser scanning according to the manufacturer's recommendations.

Expression analysis quality assessment parameters included visual array inspection of the scanned image for the presence of image artifacts and correct grid alignment for the identification of distinct probe cells as well as both low 3V5' ratio of housekeeping controls (mean: 1.90 for GAPDH) and high percentage of detection calls (mean: 46.3% present called genes). The 3' to 5' ratio of GAPDH probesets can be used to assess RNA sample and assay quality. Signal values o f the 3' probe sets for GAPDH are compared to the Signal values of the corresponding 5' probe set. The ratio of the 3' probe set to the 5' probe set is generally no more than 3.0. A high 3' to 5' ratio may indicate degraded RNA or inefficient synthesis of ds cDNA or biotinylated cRNA (GeneChiplKJ Expression Analysis Teclinical Manual, www.affymetrix.com). Detection calls are used to determine whether the transcript of a gene is detected (present) or undetected (absent) and were calculated using default parameters of the Microarray Analysis Suite MAS 5.0 software package.

Method 2: Bone marrow (BM) aspirates are taken at the time of the initial diagnostic biopsy and remaining material is immediately lysed in RLT buffer (Qiagen), frozen and stored at -80°C until preparation for gene expression analysis. For microarray analysis the GeneChip® System (Affymetrix, Santa Clara, CA, USA) is used. The targets for GeneChip® analysis are prepared according to the current Expression Analysis. Briefly, frozen lysates of the leukemia samples are thawed, homogenized (QIAshredder, Qiagen) and total RNA extracted (RN easy Mini Kit, Qiagen). Normally 10 μg total RNA isolated from 1 x 10⁷ cells is used as starting material in the subsequent cDNA-Synthesis using Oligo-dT-T7- Promotor Primer (cDNA synthesis Kit, Roche Molecular Biochemicals). The cDNA is purified by plienol- chloroform extraction and precipitated with 100% Ethanol overnight. For detection of the hybridized target nucleic acid biotin-labeled ribonucleotides are incorporated during the in vitro transcription reaction (Enzo BioArray™ High Yield RNA Transcript Labeling Kit, ENZO). After quantification of the purified cRNΛ. (RNeasy Mini Kit, Qiagen), 15 μg are fragmented by alkaline treatment (2O0 mM Tris-acetate, pH 8.2, 500 mM potassium acetate, 150 mM magnesium acetate) and added to the hybridization cocktail sufficient for 5 hybridizations on standard GeneChip® microarrays. Before expression profiling Test3 Probe Arrays (Affymetrix) are chosen for monitoring of the integrity of the cRNA. Only labeled cRNA-cocktails which show a ratio of the measured intensity of the 3' to trie 5' end of the GAPDH gene less than 3.0 are selected for subsequent hybridization on HG- U133 probe arrays (Affymetrix). Washing and staining the Probe arrays is performed as described (see, Affymetrix-Original-Literature (LOCKHART und LIPSHUTZ). The Affymetrix software (Microarray Suite, Version 4.0.1) extracted fluorescence intensities from each element on the arrays as detected by con. focal laser scanning according to the manufacturers recommendations.

REFERENCES

(1) Bennett JM, Catovsky D, Daniel MT, Flandrin G, Galton DA, Gralnick HR et al. Proposed revised criteria for the classification of acute myeloid leukemia. A report of the French- American-British Cooperative Group. Ann Intern Med 1985; 103(4):620-625.

(2) Bennett JM, Catovsky D, Daniel MT, Flandrin G, Galton DA, Gralnick HR et al. Proposals for the classification of the acute leukaemias. French- American- British (FAB) co-operative group. Br J Haematol 1976; 33(4):451-458. (3) Jaffe ES, Harris NL, Stein H, Vardiman JW. World Health Organization Classification of Tumours. Pathology and Genetics of Tumours of Haematopoietic and Lymphoid Tissues. 2001. Lyon, IARC Press.

(4) Tallman MS, Andersen JW, Schiffer CA, Appelbaum FR, Feusner JH, Ogden A et al. AU-trans-retinoic acid in acute promyelocytic leukemia [see comments]

[published erratum appears in N Engl J Med 1997 Nov 27; 337(22):1639]. N Engl J Med 1997; 337(15):1021-1028.

(5) Warrell RP, Jr., de The H, Wang ZY, Degos L. Acute promyelocytic leukemia [see comments]. N Engl J Med 1993; 329(3): 177- 189. (6) Warrell RP, Jr., Frankel SR, Miller WH, Jr., Scheinberg DA, Itri LM, Hitt elman WN et al. Differentiation therapy of acute promyelocytic leukemia with tretinoin (all-trans-retinoic acid). N Engl J Med 1991; 324(20): 1385- 1393.

(7) Lowenberg B, Downing JR, Burnett A. Acute myeloid leukemia. N Engl J Med 1999; 341(14):1051-1062. (8) Schoch C, Haferlach T, Haase D, Fonatsch C, Loffler H, Schlegelberger B et al. Patients with de novo acute myeloid leukaemia and complex karyotype aberrations show a poor prognosis despite intensive treatment: a study of 90 patients. Br J Haematol 2001; 112(1):118-126.

(9) Grimwade D, Walker H, Oliver F, Wheatley K, Harrison C, Harrison G et al. The importance of diagnostic cytogenetics on outcome in AML: analysis of 1 ,612 patients entered into the MRC AML 10 trial. The Medical Research Council -Adult and Children's Leukaemia Working Parties. Blood 1998; 92(7):2322-2333.

(10) Grimwade D, Walker H, Harrison G, Oliver F, Chatters S, Harrison CJ et al. The predictive value of hierarchical cytogenetic classification in older adults -with acute myeloid leukemia (AML): analysis of 1065 patients entered into the United Kingdom Medical Research Council AMLl 1 trial. Blood 2001; 98(5):1312-1 320.

(11) Pui CH, Relling MV, Downing JR. Acute lymphoblastic leukemia. N Engl J Med 2004; 350(15):1535-1548. (12) Goldman JM, MeIo JV. Chronic myeloid leukemia— advances in biology and new approaches to treatment. N Engl J Med 2003; 349(15):1451 -1 464.

(13) Hughes TP, Kaeda J, Branford S, Rudzki Z, Hochhaus A, Hensley ML et al. Frequency of major molecular responses to imatinib or interferon alfa plus cytarabine in newly diagnosed chronic myeloid leukemia. N Engl J Med 2003; 349(15):1423-1432.

(14) O'Brien SG, Guilhot F, Larson RA, Gathmann I, Baccarani ML, Cervantes F et al. Imatinib compared with interferon and low-dose cytarabine for newly diagnosed chronic-phase chronic myeloid leukemia. N Engl J Med 2003; 348(1 1):994-1004. (15) Druker BJ, Sawyers CL, Kantarjian H, Resta DJ, Reese SF, Ford JM et al. Activity of a specific inhibitor of the BCR-ABL tyrosine kinase in the blast crisis of chronic myeloid leukemia and acute lymphoblastic leukemia with the Philadelphia chromosome. N Engl J Med 2001; 344(14): 1038-104-2.

(16) Kantarjian H, Sawyers C, Hochhaus A, Guilhot F, Schiffer C, Gambacorti- Passerini C et al. Hematologic and cytogenetic responses to imatinib mesylate in chronic myelogenous leukemia. N Engl J Med 2002; 346(9):645-652.

(17) Scheming UJ, Pfeifer H, Wassmann B, Brack P, Atta J, Petershofen EK et al. Early minimal residual disease (MRD) analysis during treatment of Philadelphia chromosome/Bcr-Abl-positive acute lymphoblastic leukemia with, the Abl-tyrosine kinase inhibitor imatinib (STI571). Blood 2003; 101(l):85-90.

(18) Kern W, Voskova D, Schoch C, Hiddemann W, Schnittger S, Haferlach T. Determination of relapse risk based on assessment of minimal residual disease during complete remission by multiparameter flow cytometry in uxiselected patients with acute myeloid leukemia. Blood 2004; in press. (19) Schnittger S, Weisser M, Schoch C, Hiddemann W, Haferlacli T, Kern W. New score predicting for prognosis in PML-RARA+, AML1-ETO+, or CBFBMYHl 1+ acute myeloid leukemia based on quantification of fusion transcripts. Blood 2003; 102(8):2746-2755. (20) Coustan-Smith E, Sancho J, Behm FG, Hancock ML, Razzouk BI, Ribeiro RC et al. Prognostic importance of measuring early clearance of leukemic cells by flow cytometry in childhood acute lymphoblastic leukemia. Blood 2002; lO0(l):52-58.

(21) Campana D, Behm FG. Immunophenotyping of leukemia. J Imπrunol Methods 2000; 243(l-2):59-75.

(22) Rawstron AC, Kennedy B, Moreton P, Dickinson AJ, Cullen MJ, Richards SJ et al. Early prediction of outcome and response to alemtuzumab therapy in chronic lymphocytic leukemia. Blood 2003.

(23) Grimwade D, Haferlach T. Gene- expression profiling in acute myeloid leukemia. N Engl J Med 2004; 350(16):1676-1678.

(24) Golub TR, Slonim DK, Tamayo P, Huard C, Gaasenbeek M, Mesirov JP et al. Molecular classification of cancer: class discovery and class prediction by gene expression monitoring. Science 1999; 286(5439):531-537.

(25) Dugas M, Merk S, Breit S, Dirschedl P. mdclust—exploratory microarray analysis by multidimensional clustering. Bioinformatics 2004; 20(6):931-936.

(26) Yeoh EJ, Ross ME, Shurtleff SA, Williams WK, Patel D, Mahfouz R et al. Classification, subtype discovery, and prediction of outcome in pediatric acute lymphoblastic leukemia by gene expression profiling. Cancer Cell 20O2; 1(2):133- 143. (27) Ross ME, Zhou X, Song G, Shurtleff SA, Girtman K, Williams WK et al. Classification of pediatric acute lymphoblastic leukemia by gene expression profiling. Blood 2003; 102(8):2951-2959.

(28) Ross ME, Mahfouz R, Onciu M, Liu HC, Zhou X, Song G et al. Gene Expression Profiling of Pediatric Acute Myelogenous Leukemia. Blood 2004. (29) Dugas M, Schoch C, Schnittger S, Haferlach T, Danhauser-Riedl S, Hiddemann W et al. A comprehensive leukemia database:integration of cytogenetics,molecular genetics and microarray data with clinical information,cytomorphology and immunophenotyping. Leukemia 20Ol ; 15(12):1805-1810. (30) Haferlach T, Schoch C, Loffler H, Gassmann W, Kern W, Scknittger S et al. Morphologic dysplasia in de novo acute myeloid leukemia (AML) is related to unfavorable cytogenetics but has no independent prognostic relevance under the conditions of intensive induction therapy: results of a multiparameter analysis from the German. AML Cooperative Group studies. J Clin Oncol 2003 ; 2 1 (2):256-265.

(31) Kern W, Voskova D, Schoch C, Schnittger S, Hiddemann W, Haferlach T. Prognostic impact of early response to induction therapy as assessed by multiparameter flow cytometry in acute myeloid leukemia. Haematologica 2004; 89(5):528-540. (32) Schocti C, Schnittger S, Klaus M, Kern W, Hiddemann W, Haferlach T. AML with 1 Iq23/MLL abnormalities as defined by the WHO classification: incidence, partner chromosomes, FAB subtype, age distribution, and prognostic impact in an unselected series of 1897 cytogenetically analyzed AML cases. Blood 2003; 102(7):2395-2402. (33) Schnittger S, Schoch C, Dugas M, Kern W, Staib P, Wuchter C et al. Analysis of FLT3 length mutations in 1003 patients with acute myeloid leukemia: correlation to cytogenetics, FAB subtype, and prognosis in the AMLCG study and usefulness as a marker for the detection of minimal residual disease. Blood 2002; 100(l):59-66. (34) Schocli C, Schnittger S, Bursch S, Gerstner D, Hochhaus A, Berger U et al. Comparison of chromosome banding analysis, interphase- and hypermetaphase- FISH, qualitative and quantitative PCR for diagnosis and for follow-up in chronic myeloid leukemia: a study on 350 cases. Leukemia 2002; 16(l):53-59.

(35) Liu WM, Mei R, Di X, Ryder TB, Hubbell E, Dee S et al. Analysis of high density expression microarrays with signed-rank call algorithms. Bioinformatics

2002; 18(12):1593-1599.

(36) Storey JD, Tibshirani R. Statistical significance for genomewide studies. Proc Natl Acad Sci U S A 2003; 100(16): 9440-9445.

(37) Vapnilc V. Statistical Learning Theory. 1998. New York, Wiley. (38) Furey TS, Cristianini N, Duffy N, Bednarski DW, Schummer M, Haussler D. Support vector machine classification and validation of cancer tissue samples using microarray expression data. Bioinformatics 2000; 16(10):906-914_

(39) Brown MP, Grundy WN, Lin D, Cristianini N, Sugnet CW, Fxirey TS et al. Knowledge-based analysis of microarray gene expression data by using support vector machines. Proc Natl Acad Sci U S A 2000; 97(l):262-267.

(40) Livak KJ, Schmittgen TD. Analysis of relative gene expression data using real-time quantitative PCR and the 2(-Delta Delta C(T)) Method. Methods 2001; 25(4):402-408. (41) Onciu M, Lai R, Vega F, Bueso-Ramos C, Medeiros LJ. Precursor T-cell acute lymphoblastic leukemia in adults: age-related immunophenotypic, cytogenetic, and molecular subsets. Am J Clin Pathol 2002; 117(2):252-258.

(42) Buchner T, Hiddemann W, Berdel WE, Wormann B, Schoch C, Fonatsch C et al. 6-Thioguanine, Cytarabine, and Daunorubicin (TAD) and Higri-Dose Cytarabine and Mitoxantrone (HAM) for Induction, TAD for Consolidation, and Either Prolonged Maintenance by Reduced Monthly TAD or TAD-HAM-TAD and One Course of Intensive Consolidation by Sequential HAM in Adxilt Patients at All Ages With De Novo Acute Myeloid Leukemia (AML): A Randomized Trial of the German AML Cooperative Group. J Clin Oncol 2003; 21(24):4496-4504. (43) Kern W, Haferlach T, Schoch C, Loffler H, Gassmann W, Heinecke A et al. Early blast clearance by remission induction therapy is a major independent prognostic factor for both achievement of complete remission and long-term outcome in acute myeloid leukemia: data from the German AML Cooperative Group (AMLCG) 1992 Trial. Blood 2003; 101(1):64-70. (44) Lucio P, Gaipa G, van Lochem EG, Van Wering ER, Porwit-MacDonald A, Faria T et al. BIOMED-I concerted action report: flow cytometric immunophenotyping of precursor B-ALL with standardized triple-stainings. BIOMED-I Concerted Action Investigation of Minimal Residual Disease in Acute Leukemia: International Standardization and Clinical Evaluation. Leukemia 2001; 15(8):1185-1192. (45) Gleissner B, Rieder H, Thiel E, Fonatsch C, Janssen LA, Heinze B et al. Prospective BCR-ABL analysis by polymerase chain reaction (RT-PCR) in adult acute B-lineage lymphoblastic leukemia: reliability of RT-nested-PCR and comparison to cytogenetic data. Leukemia 2001; 15(12): 1834-184O. (46) Argyle JC, Benjamin DR, Lampkin B, Hammond D. Acute nonlymphocytic leukemias of childhood. Inter-observer variability and problems in the use of the FAB classification. Cancer 1989; 63(2):295-301.

(47) Bennett JM, Begg CB. Eastern Cooperative Oncology Group study of the cytochemistry of adult acute myeloid leukemia by correlation of subtypes with response and survival. Cancer Res 1981; 41(11 Pt 2):4833-4837.

(48) Byrd JC, Mrozek K, Dodge RK, Carroll AJ, Edwards CG, Arthur DC et al. Pretreatment cytogenetic abnormalities are predictive of induction success, cumulative incidence of relapse, and overall survival in adult patients with de novo acute myeloid leukemia: results from Cancer and Leukemia Group B (CALGB 8461). Blood 2002; 100(13):4325-4336.

(49) Sctioch C, Kern W, Schnittger S, Hiddemann W, Haferlach T. Karyotype is an independent prognostic parameter in therapy-related acute myeloid leukemia (t- AML): an analysis of 93 patients with t-AML in comparison to 1091 patients with de novo AML. Leukemia 2004; 18(l):120-125. (50) Hoelzer D, Ludwig WD, Thiel E, Gassmann W, Loffler H, Fonatsch C et al. Improved outcome in adult B-cell acute lymphoblastic leukemia. Blood 1996; 87(2):495-508.

(51) Texien DG. Disruption of differentiation in human cancer: AML shows the way. Nat Rev Cancer 2003; 3(2):89-101. (52) Mecucci C, Rosati R, Starza RL. Genetic profile of acute myeloid leukemia. Rev Clin Exp Hematol 2002; 6(l):3-25.

(53) Alcalay M, Orleth A, Sebastiani C, Meani N, Chiaradonna F, Casciari C et al. Common themes in the pathogenesis of acute myeloid leukemia. Oncogene 2001 ; 20(40): 5680-5694. (54) Kohlmann A, Schoch C, Schnittger S, Dugas M, Hiddemann W, Kern W et al. Molecular characterization of acute leukemias by use of microarray technology. Genes Chromosomes Cancer 2003; 37:396-405.

(55) Schoch. C, Kohlmann A, Schnittger S, Brors B, Dugas M, Mergenthaler S et al. Acute myeloid leukemias with reciprocal rearrangements can be distinguished by specific gene expression profiles. Proc Natl Acad Sci U S A 2002; 99(15):100O8-10013.

(56) VaIk PJ, Verhaak RG, Beijen MA, Erpelinck CA, Barjesteh van Waalwijk van Doorn-Khosrovani, Boer JM et al. Prognostically useful gene-expression profiles in acute myeloid leukemia. N Engl J Med 2004; 350(16):1617-1628.

(57) Bullinger L, Dohner K, Bair E, Frohling S, Schlenk RF, Tibshirani R et al. Use of gene-expression profiling to identify prognostic subclasses in adult acute myeloid leukemia. N Engl J Med 2004; 350(16):1605-1616.

(58) Gilliland DG, Griffin JD. The roles of FLT3 in hematopoiesis and leukemia. Blood 2002; 100(5): 1532-1542.

(59) Bene MC, Castoldi G, Knapp W, Ludwig WD, Matutes E, Orfao A et al. Proposals for the immunological classification of acute leukemias. European Group for the Immunological Characterization of Leukemias (EGIL). Leukemia 1995; 9(10):1783-1786. (60) Johnstone RW, See RH, Sells SF, Wang J, Mufhukkumar S, Englert C et al. A novel repressor, par-4, modulates transcription and growth suppression functions of the Wilms' tumor suppressor WTl. MoI Cell Biol 1996; 16(12):6945-6956.

(61) Galon J, Franchimont D, Hiroi N, Frey G, Boettner A, Ehrhart-Bornstein M et al. Gene profiling reveals unknown enhancing and suppressive actions of glucocorticoids on immune cells. FASEB J 2002; 16(1):61-71.

(62) Inohara N, Koseki T, del Peso L, Hu Y, Yee C, Chen S et al. N^"odl, an Apaf-1- like activator of caspase-9 and nuclear factor-kappaB. J Biol Chem 1999; 274(21):14560-14567. (63) Bertin J, Nir WJ, Fischer CM, Tayber OV, Errada PR, Grant JR et al. Human CARD4 protein is a novel CED-4/Apaf-l cell death family member that activates NF-kappaB. J Biol Chem 1999; 274(19):12955-12958.

(64) Whitney AR, Diehn M, Popper SJ, Alizadeh AA, Boldrick JC, Relman DA et al. Individuality and variation in gene expression patterns in human blood. Proc

Natl Acad Sci U S A 2003; 100(4): 1896- 1901.

(65) Jelinek DF, Tschumper RC, Stolovitzky GA, Iturria SJ, Tu ^"Y, Lepre J et al. Identification of a global gene expression signature of B-chronic lymphocytic leukemia. MoI Cancer Res 2003; 1(5):346-361. (66) Hofmann WK, de Vos S, Komor M, Hoelzer D, Wachsman W, Koeffler HP. Characterization of gene expression of CD34+ cells from normal and myelodysplastic bone marrow. Blood 2002; 100(10):3553-3560.

(67) Repp R, Borkhardt A, Haupt E, Kreuder J, Brettreich S, Harnmermann J et al. Detection of four different 1 Iq23 chromosomal abnormalities by multiplex-PCR and fluorescence-based automatic DNA-fragment analysis. Leukemia 1995; 9(l):210-215.

(68) Tallman MS. Relevance of pathologic classifications and diagnosis of acute myeloid leukemia to clinical trials and clinical practice. Cancer Treat Res 2004; 121 :45-67. (69) Preudhomme C, Sagot C, Boissel N, Cayuela JM, Tigaud I, De Botton S et al. Favorable prognostic significance of CEBPA mutations in patients with de novo acute myeloid leukemia: a study from the Acute Leukemia French Association (ALFA). Blood 2002; 100(8):2717-2723.

(70) Schnittger S, Kinkelin U, Schoch C, Heinecke A, Haase D, Haferlach T et al. Screening for MLL tandem duplication in 387 unselected patients with AML identify a prognostically unfavorable subset of AML. Leukemia 2000; 14(5):796- 804.

(71) Virtaneva K, Wright FA, Tanner SM, Yuan B, Lemon WJ, Caligiuri MA et al. Expression profiling reveals fundamental biological differences in acute myeloid leukemia with isolated trisomy 8 and normal cytogenetics. Proc Natl Acad Sci U S A 2001 ; 98(3):1124-1129.

(72) Debernardi S, Lillington DM, Chaplin T, Tomlinson S, Amess J, Rohatiner A et al. Genome- wide analysis of acute myeloid leukemia with normal karyotype reveals a unique pattern of homeobox gene expression distinct from those with translocation-mediated fusion events. Genes Chromosomes Cancer 2003; 37(2):149-158.

(73) Alizadeh AA, Ross DT, Perou CM, van de RM. Towards a novel classification of human malignancies based on gene expression patterns. J Pathol 2001; 195(l):41-52.

(74) Staudt LM. Molecular diagnosis of the hematologic cancers. 3ST Engl J Med 2003; 348(18):1777-1785.

(75) Rosenwald A, Alizadeh AA, Widhopf G, Simon R, Davis RE, Yu X et al. Relation of gene expression phenotype to immunoglobulin mutation genotype in B cell chronic lymphocytic leukemia. J Exp Med 2001; 194(11):1639-1647.

(76) Greiner TC. mRNA microarray analysis in lymphoma and leukemia. Cancer Treat Res 2004; 121 :1-12.

(77) Tusher VG, Tibshirani R, Chu G. Significance analysis of microarrays applied to the ionizing radiation response. Proc Natl Acad Sci U S A 2001 ; 98(9):5116- 5121.

(78) Storey JD, Tibshirani R. Statistical significance for genomew^ride studies. Proc Natl Acad Sci U S A 2003; 100(16):9440-9445.

(79) Vapnik V. Statistical Learning Theory. 1998. New York, Wiley. Ref Type: Serial (Book,Monograph) (80) Brown MP, Grundy WN, Lin D, Cristianini N, Sugnet CW, Furey TS et al. Knowledge-based analysis of microarray gene expression data by using support vector machines. Proc Natl Acad Sci U S A 2000; 97(l):262-267. (81) Furey TS, Cristianini N, Duffy N, Bednarski DW, Schummer M, Haussler D. Support vector machine classification and validation of cancer tissue samples using microarray expression data. Bioinformatics 2000; 16(10):906-914.

(82) Ross ME, Zhou X, Song G, Shurtleff SA, Girtman K, Williams WK et al. Classification of pediatric acute lymphoblastic leukemia by gene expression profiling. Blood 2003; 102(8):2951-2959.

(83) Kohlmann A, Schoch C, Schnittger S, Dugas M, Hiddemann W, Kern W et al. Pediatric acute lymphoblastic leukemia (ALL) gene expression signatures classify an independent cohort of adult ALL patients. Leukemia 2004; 18(l):63-71. (84) Liu G, Loraine AE, Shigeta R, Cline M, Cheng J, Valmeekam V et al.

NetAffx: Affymetrix probesets and annotations. Nucleic Acids Res 2003; 31(1):82- 86.

While the foregoing invention has been described in some detail for purposes of clarity and understanding, it will be clear to one^Λ skilled in the art from a reading of this disclosure that various changes in form and detail can be made without departing from the true scope of the invention. For example, all the techniques and apparatus described above can be used in various combinations. All publications, patents, patent applications, and/or other documents cited in this application are incorporated by reference in their entirety for all purposes to the same extent as if each individual publication, patent, patent application, and/or other document were individually indicated to be incorporated by reference for all purposes.

Table 1

AML with t(15,17) versus rest

One-Versus-All (OVA) comparisons AML_t(15,17) versus rest list 1 AML with t 15,17 versus rest

Table I

AML with t(15;17) versus rest

Table I

AML with t(15;17) versus rest

Table I

AMLwitht(15;17) versus rest

Table I

AML with t(15;17) versus rest

Table I

AML with t(15,17) versus rest

Table I

AMLwitht(15;17) versus rest

Table I

AML with t(15;17) versus rest

Table I

AML with t(15;17) versus rest

Table I

AML with t(15;17) versus rest

1O

Tablβ 1

AML with t(15,17) versus rest

11

Table I

AML with t(15;17) versus rest

12

Table I

AML with t(15;17) versus rest

13

Table I

AML with t(15;17) versus rest

14

Table I

AMLwitht(15;17) versus rest

15

Table I

AMLwitht(15;17) versus rest

16

Table Il

AML with t(8;21) versus rest

One-Versus-AII (OVA) comparisons AML with t(8;21 ) versus rest

Table Il

AML with t(8,21) versus rest

Table Il

AML with t(8;21) versus rest

Table Il

AML with t(8;21) versus rest

Table Il

AML with t(8;21) versus rest

Table Il

AML with t(8,21) versus rest

Table Il

AML with t(8,21) versus rest

Table 11

AML with t(8;21) versus rest

Table Il

AML with t(8,21) versus rest

Table Il

AML with t(8;21) versus rest

10

Table Il

AML with t(8;21) versus rest

11

Table Il

AML with t(8;21) versus rest

12

Table Il

AML with t(8;21) versus rest

13

Table Il

AML with t(8;21) versus rest

14

Table III

AML with inv(16) versus rest

One-Versus-AII (OVA) comparisons AML with inv(16) versus rest

Table III

AMLwithinv(16) versus rest

Table III

AMLwithinv(16) versus rest

Table III

AML with inv(16) versus rest

Table III

AMLwithinv(16) versus rest

Table III

AML with inv(16) versus rest

Table III

AML with inv(16) versus rest

Table 111

AMLwithinv(16) versus rest

Table III

AML with ιnv(16) versus rest

Table III

AML withinv(16) versus rest

10

Table III

AML with inv(16) versus rest

11

Table III

AMLwithinv(16) versus rest

12

Table III

AMLwithinv(16) versus rest

13

Table III

AMLwithinv(16) versus rest

14

Table IV

AML with normal karyotype or so-called other cytogenetic abnormalities versus rest

One-Versus-AII (OVA) comparisons AML with normal karyotype or so-called other cytogenetic abnormalities versus rest

Table IV

AML with normal karyotype or so-called other c to enetic abnormalities versus rest

Table IV

AML with normal karyotype or so-called other c to enetic abnormalities versus rest

Table IV

AML with normal karyotype or so-called other cytogenetic abnormalities versus rest

Table IV

AML with normal karyotype or so-called other cytogenetic abnormalities versus rest

Table IV

AML with normal karyotype or so-called other c to enetic abnormalities versus rest

Table IV

AML with normal karyotype or so-called other c to enetic abnormalities versus rest

Table IV

AML with normal karyotype or so-called other c to enetic abnormalities versus rest

Table IV

AML with normal karyotype or so-called other cytogenetic abnormalities versus rest

Table IV

AML with normal karyotype or so-called other c to enetic abnormalities versus rest

10

Table IV

AML with normal karyotype or so-called other cytogenetic abnormalities versus rest

11

Table IV

AML with normal karyotype or so-called other cytogenetic abnormalities versus rest

12

Table IV

AML with normal karyotype or so-called other c to enetic abnormalities versus rest

13

Table IV

AML with normal karyotype or so-called other c to enetic abnormalities versus rest

14

Table IV

AML with normal karyotype or so-called other c to enetic abnormalities versus rest

15

Table IV

AML with normal karyotype or so-called other c to enetic abnormalities versus rest

16

Table V

AML with 11q23/MLL rearrangement versus rest

One-Versus-All (OVA) comparisons AML with 11q23/MLL rearrangement versus rest list 5

Table V

AML with 11q23/MLL rearran ement versus rest

Table V

AMLwith11q23/MLL rearran ement versus rest

Table V

AML with 11q23/MLL rearrangement versus rest

Table V

AMLwιth11q23/MLL rearrangement versus rest

Table V

AMLwith11q23/MLL rearran ement versus rest

Table V

AMLwith11q23/MLL rearran ement versus rest

Table V

AMLwith11q23/MLL rearrangement versus rest

Table V

AIVIL with 11q23/MLL rearran ement versus rest

Table V

AMLwιth11q23/MLL rearrangement versus rest

10

11

Table V

AMLwιth11q23/MLL rearrangement versus rest

12

Table V

AML with 11q23/MLL rearran ement versus rest

13

Table V

AML with 11q23/MLL rearran ement versus rest

14

Table Vl AML with complex aberrant karyotype versus rest

Table Vl AML with complex aberrant karyotype versus rest

Table Vl AML with complex aberrant karyotype versus rest

Table Vl AML with complex aberrant karyotype versus rest

Table Vl AML with complex aberrant karyotype versus rest

Table Vl AML with complex aberrant karyotype versus rest

Table Vl AML with complex aberrant karyotype versus rest

Table Vl AML with complex aberrant karyotype versus rest

Table Vl AML with complex aberrant karyotype versus rest

Table Vl AML with complex aberrant karyotype versus rest

10

Table Vl AML with complex aberrant karyotype versus rest

11

Table Vl AML with complex aberrant karyotype versus rest

12

Table Vl AML with complex aberrant karyotype versus rest

13

Table Vl AML with complex aberrant karyotype versus rest

14

Table Vl AML with complex aberrant karyotype versus rest

15

Table Vl AML with complex aberrant karyotype versus rest

16

TableVII Pro-B-ALL/t(11q23) versus rest

Table VII Pro-B-ALL/t(11q23) versus rest

Table VII Pro-B-ALL/t(11q23) versus rest

TableVII Pro-B-ALL/t(11q23) versus rest

Table VII Pro-B-ALL/t(11q23) versus rest

Table VII Pro-B-ALL/t(11q23) versus rest

Table VII Pro-B-ALL/t(11q23) versus rest

Table VII Pro-B-ALL/t(11q23) versus rest

Table VII Pro-B-ALL/t(11q23) versus rest

TableVII Pro-B-ALL/t(11q23) versus rest

10

Table VII Pro-B-ALL/t(11q23) versus rest

11

Tsble VII Pro-B-ALL/t(-l1q23) versus rest

12

TableVII Pro-B-ALL/t(11 q23) versus rest

13

Table VII Pro-B-ALL/t(11q23) versus rest

14

Table VIII mature B-ALL/t(8;14) versus rest

Table VlIl mature B-ALL/t(8;14) versus rest

Table VlIl mature B-ALL/t(8;14) versus rest

Table VIII mature B-ALL/t(8;14) versus rest

Table VIIl mature B-ALL/t(8;14) versus rest

Table VIII mature B-ALL/t(8;14) versus rest

TableVIII mature B-ALL/t(8;14) versus rest

Table VIII mature B-ALL/t(8;14) versus rest

TableVIII mature B-ALL/t(8;14) versus rest

Table VIII mature B-ALL/t(8;14) versus rest

10

TableVIII mature B-ALLΛ(8;14) versus rest

11

Table VIll mature B-ALL/t(8;14) versus rest

12

Table VIII mature B-ALL/t(8;1₄) versus rest

13

Table VIII mature B-ALL/t(8,14) versus rest

14

Tablβ VIII mature B-ALL/t(8, 14) versus rest

15

Table VIII mature B-ALL/t(8;14) versus rest

16

Table IX c-ALL/Pre-B-ALL with or without t(9;22) versus rest

Table IX c-ALL/Pre-B-ALL with or without t(9;22) versus rest

Table IX c-ALL/P re-B-ALL with or without t(9;22) versus rest

Table IX c-ALL/Pre-B-ALL with or without t(9;22) versus rest

Table IX c-ALL/Pre-B-ALL with or without t(9;22) versus rest

Table IX c-ALL/Pre-B-ALL with or without t(9;22) versus rest

Table IX c-ALL/Pre-B-ALL with or without t(9;22) versus rest

Table IX c-ALL/Pre-B-ALL with or without t(9;22) versus rest

Table IX c-ALL/Pre-B-ALL with or without t(9;22) versus rest

Table IX o-ALL/Pre-B-ALL with or without t(9;22) versus rest

10

Table IX c-ALL/Pre-B-ALL with or without t(9;22) versus rest

11

Table IX c-ALL/Pre-B-ALL with or without t(9;22) versus rest

12

Table IX c-ALL/Pre-B-ALL with or without t(9;22) versus rest

13

Table IX c-ALL/Pre-B-ALL with or without t(9;22) versus rest

14

Table X T-ALL versus rest

Oπe-Versus-All (OVA) comparisons

Table X T-ALL versus rest

Table X T-ALL versus rest

Table X T-ALL versus rest

Table X T-ALL versus rest

Table X T-ALL versus rest

Table X T-ALL versus rest

Table X T-ALL versus rest

Table X T-ALL versus rest

Table X T-ALL versus rest

10

Table X T-ALL versus rest

11

Table X T-ALL versus rest

12

Table X T-ALL versus rest

13

Table X T-ALL versus rest

14

Table Xl CML vs Rest

Oπe-Versus-AII (OVA) comparisons

Table Xl CML vs Rest

Table Xl CML vs Rest

Table Xl CML vs Rest

Table Xl GML vs Rest

Table Xl CML vs Rest

Table Xl CML vs Rest

Table Xl CML vs Rest

Table Xl CML vs Rest

Table Xl CML vs Rest

10

Table Xl CML vs Rest

11

Table Xl CML vs Rest

12

Table Xl CML vs Rest

13

Table Xl CML vs Rest

14

Table XII CLL versus rest

Table XlI CLL versus rest

Table XII CLL versus rest

Table XII CLL versus rest

Table XlI CLL versus rest

Table XII CLL versus rest

Table XII CLL versus rest

Table XII CLL versus rest

Table XII CLL versus rest

Table XII CLL versus rest

10

TableXII CLL versus rest

11

Table XII CLL versus rest

12

Table XII CLL versus rest

13

Table XII CLL versus rest

14

Table XlIl non-leukemia versus rest

One-Versus-AII (OVA) comparisons

Table XIII non-leukemia versus rest

TableXIII non-leukemia versus rest

Table XIII non-leukemia versus rest

Table XIII non-leukemia versus rest

Table XIII non-leukemia versus rest

TableXiπ non-leukemia versus rest

Table XIII non-leukemia versus rest

Table XlII non-leukemia versus rest

TableXIII non-leukemia versus rest

10

Table XIII non-leukemia versus rest

11

Table XIII non-leukemia versus rest

12

Table XIII non-leukemia versus rest

13

Table XIII non-leukemia versus rest

14

Table XIV ALL with t(9;22) vs ALL without t(9;22)

List

Table XlV ALL with t(9;22) vs ALL without t(9;22)

Table XIV ALL with t(9;22) vs ALL without t(9;22)

Table XlV ALL with t(9;22) vs ALL without t(9;22)

Table XIV ALL with t(9;22) vs ALL without t(9;22)

Table XIV ALL with t(9;22) vs ALL without t(9;22)

Table XIV ALL with t(9,22) vs ALL without t(9,22)

Table XIV ALL with t(9;22) vs ALL without t(9;22)

Table XIV ALL with t(9;22) vs ALL without t(9;22)

Table XIV ALL with t(9;22) vs ALL without t(9;22)

10

Table XIV ALL with t(9;22) vs ALL without t(9;22)

11

Table XIV ALL with t(9;22) vs ALL without t(9;22)

12

Table XV cortical T-ALL vs immature T-ALL

List

Table XV cortical T-ALL vs immature T-ALL

Table XV cortical T-ALL vs immature T-ALL

Table XV cortical T-ALL vs immature T-ALL

Table XV cortical T-ALL vs immature T-ALL

Table XV cortical T-ALL vs immature T-ALL

Table XV cortical T-ALL vs immature T-ALL

Table XV cortical T-ALL vs immature T-ALL

Table XV cortical T-ALL vs immature T-ALL

Table XV cortical T-ALL vs immature T-ALL

10

Claims

PATENT CLAIMS

1. A method of detecting leukemia, the method comprising detecting an expression of at least one gene population of at least one cell, which gene population comprises at least one set of genes listed in one or more of Tables I-XII, thereby detecting leukemia.

2. The method of claim 1, comprising measuring expression levels of genes of the cell on at least one probe array that comprises oligonucleotides with nucleotide sequences that correspond to at least subsequences of one or more sets of genes listed in at least one of Tables I-XII.

3. The method of claim 1, wherein the cell is derived from at least one subject.

4. The method of claim 1, comprising correlating a detected expression of a set of g $eenneess l liisstteedd i inn T Taabbllee X XII w wiitthh the cell being a chronic myeloid leukemia (CML) cell j. The method of claim 1, comprising correlating a detected expression of a set of ge :nneess l liisstteedd i inn T Taabbllee X XIIII with the cell being a chronic lymphatic leukemia (CLL) cell

6. The method of claim 1, comprising correlating a detected expression of one or more sets of genes listed in at least one of Tables I-VI with the cell being an ite myeloid leukemia (AML) cell.

7. The method of claim 6, comprising correlating a detected expression of a of genes listed in Table I with the cell being an AML cell with a t(15;17).

8. The method of claim 6, comprising correlating a detected expression of a set of genes listed in Table II with the cell being an AML cell with a t(8;21).

9. The method of claim 6, comprising correlating a detected expression of a set of genes listed in Table III with the cell being an AML cell with an inv(16).

10. The method of claim 6, comprising correlating a detected expression of a set of genes listed in Table IV with the cell being an AML cell with a normal karyotype or another cytogenetic abnormality. 11. The method of claim 6, comprising correlating a detected expression of a set of genes listed in Table V with the cell being an AML cell with, a 1 Iq23/MLL rearrangement.

12. The method of claim 6, comprising correlating a detected expression of a set of genes listed in Table VI with the cell being an AML cell with a complex aberrant karyotype.

13. The method of claim 1, comprising correlating a detected expression of one or more sets of genes listed in at least one of Tables VII-X with the cell being an acute lymphoblastic leukemia (ALL) cell. 14. The method of claim 13, comprising correlating a detected expression of a set of genes listed in Table VII with the cell being a Pro-B-ALLVt(llq23) cell.

15. The method of claim 13, comprising correlating a detected expression of a set of genes listed in Table VIII with the cell being a mature B-ALL/t(8;14) cell. 16. The method of claim 13, comprising correlating a detected expression of a set of genes listed in Table IX with the cell being a c-ALL/Pre-B-ALL cell with or without t(9;22).

17. The method of claim 13, comprising correlating a detected expression of a set of genes listed in Table X correlates with the cell being a T-ALL cell. 18. The method of claim I₅ wherein detecting the expression of the gene population comprises hybridizing transcribed polynucleotides or portions thereof to complementary polynucleotides or portions thereof.

19. The method of claim 18, wherein the transcribed polynucleotides or portions thereof are hybridized under stringent hybridization conditions. 20. The method of claim 1, wherein detecting the expression of the gene population comprises using an array, a robotics system, and/or a microfluidic device. 21. The method of claim 1, wherein the expression of the gene population is detected by amplifying nucleic acid sequences associated with the genes to produce amplicons and detecting the amplicons.

22. The method of claim 21, wherein the amplicons are detected using a process that comprises one or more of: hybridizing the amplicons to an oligonucleotide array, digesting the amplicons with a restriction enzyme, or real¬ time polymerase chain reaction (PCR) analysis.

23. The method of claim 1, wherein detecting the expression of the gene population comprises measuring quantities of transcribed polynucleotides or portions thereof expressed or derived from the genes.

24. The method of claim 23, wherein the transcribed polynucleotides are mRNAs or cDNAs.

25. The method of claim 1, wherein detecting the expression level comprises contacting polynucleotides or polypeptides expressed from the genes with compounds that specifically bind the polynucleotides or polypeptides.

26. The method of claim 25, wherein the compounds comprise aptamers, antibodies or fragments thereof.

27. A method of differentiating between leukemia and non-leukemia cells, the method comprising: measuring expression levels of at least one gene population of at least one cell to produce expression data; and correlating the expression data with at least one set of genes listed in one or more of Tables I-XII, thereby detecting the leukemia cell, or with a set of genes listed in Tables XIII, thereby detecting the non-leukemia cell. 28. A kit, comprising: one or more probe biomolecules corresponding to one or more genes or portions thereof listed in one or more of Tables I-XIII; and, instructions for correlating detected expression levels of one or more polynucleotides or polypeptides in at least one target cell, which polynucleotides or polypeptides detectably bind to one or more of the probe biomolecules, with the target cell being a leukemia cell or a non-leukemia cell,

29. The kit of claim 28, comprising one or more additional reagents to perform real-time PCR analyses. 30. The kit of claim 28, wherein at least one solid support comprises the probe biomolecules.

31. The kit of claim 28, wherien the probe biomolecules comprise polynucleotides.

32. A system, comprising: one or more probe biomolecules corresponding to one or more genes or portions thereof listed in one or more of Tables I-XIII; and, at least one reference data bank for correlating detected expression levels of polynucleotides or polypeptides in target cells, which polynucleotides or polypeptides detectably bind to one or more of the probe biomolecules, with the target cell being a leukemia cell or a non-leukemia cell.

33. The system of claim 32, comprising one or more additional reagents and/or components to perform real-time PCR analyses.

34. The system of claim 32, wherein the reference data bank is produced by: (a) compiling a gene expression profile of a patient sample by determining the expression level at least one of the markers, and

(b) classifying the gene expression profile using a machine learning algorithm.

35. The system of claim 34, wherein the machine learning algorithm is selected from the group consisting of: a weighted voting algorithm, a K-nearest neighbors algorithm, a decision tree induction algorithm, a support vector machine, and a feed- forward neural network.

36. A method of producing a reference data bank for distinguishing leukemia and non-leukemia cells from one another, the method comprising: (a) compiling a gene expression profile of a patient sample by determining the expression level of genes listed in one or more of Tables I-XIII, and;

(b) classifying the gene expression profile using a machine learning algorithm.

37. The reference data bank produced by the method of claim 36.