WO2013190090A1 - Gene signatures for classifying and grading lung cancer - Google Patents

Gene signatures for classifying and grading lung cancer Download PDF

Info

Publication number
WO2013190090A1
WO2013190090A1 PCT/EP2013/062993 EP2013062993W WO2013190090A1 WO 2013190090 A1 WO2013190090 A1 WO 2013190090A1 EP 2013062993 W EP2013062993 W EP 2013062993W WO 2013190090 A1 WO2013190090 A1 WO 2013190090A1
Authority
WO
WIPO (PCT)
Prior art keywords
genes
lung cancer
stage
cel
gene
Prior art date
Application number
PCT/EP2013/062993
Other languages
French (fr)
Inventor
Stéphanie BOUE
Florian Martin
Marja TALIKKA
Yang Xiang
Original Assignee
Philip Morris Products S.A.
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Philip Morris Products S.A. filed Critical Philip Morris Products S.A.
Publication of WO2013190090A1 publication Critical patent/WO2013190090A1/en

Links

Classifications

    • C—CHEMISTRY; METALLURGY
    • C12—BIOCHEMISTRY; BEER; SPIRITS; WINE; VINEGAR; MICROBIOLOGY; ENZYMOLOGY; MUTATION OR GENETIC ENGINEERING
    • C12Q—MEASURING OR TESTING PROCESSES INVOLVING ENZYMES, NUCLEIC ACIDS OR MICROORGANISMS; COMPOSITIONS OR TEST PAPERS THEREFOR; PROCESSES OF PREPARING SUCH COMPOSITIONS; CONDITION-RESPONSIVE CONTROL IN MICROBIOLOGICAL OR ENZYMOLOGICAL PROCESSES
    • C12Q1/00—Measuring or testing processes involving enzymes, nucleic acids or microorganisms; Compositions therefor; Processes of preparing such compositions
    • C12Q1/68—Measuring or testing processes involving enzymes, nucleic acids or microorganisms; Compositions therefor; Processes of preparing such compositions involving nucleic acids
    • C12Q1/6876—Nucleic acid products used in the analysis of nucleic acids, e.g. primers or probes
    • C12Q1/6883—Nucleic acid products used in the analysis of nucleic acids, e.g. primers or probes for diseases caused by alterations of genetic material
    • C12Q1/6886—Nucleic acid products used in the analysis of nucleic acids, e.g. primers or probes for diseases caused by alterations of genetic material for cancer
    • C—CHEMISTRY; METALLURGY
    • C12—BIOCHEMISTRY; BEER; SPIRITS; WINE; VINEGAR; MICROBIOLOGY; ENZYMOLOGY; MUTATION OR GENETIC ENGINEERING
    • C12Q—MEASURING OR TESTING PROCESSES INVOLVING ENZYMES, NUCLEIC ACIDS OR MICROORGANISMS; COMPOSITIONS OR TEST PAPERS THEREFOR; PROCESSES OF PREPARING SUCH COMPOSITIONS; CONDITION-RESPONSIVE CONTROL IN MICROBIOLOGICAL OR ENZYMOLOGICAL PROCESSES
    • C12Q2600/00—Oligonucleotides characterized by their use
    • C12Q2600/112—Disease subtyping, staging or classification
    • C—CHEMISTRY; METALLURGY
    • C12—BIOCHEMISTRY; BEER; SPIRITS; WINE; VINEGAR; MICROBIOLOGY; ENZYMOLOGY; MUTATION OR GENETIC ENGINEERING
    • C12Q—MEASURING OR TESTING PROCESSES INVOLVING ENZYMES, NUCLEIC ACIDS OR MICROORGANISMS; COMPOSITIONS OR TEST PAPERS THEREFOR; PROCESSES OF PREPARING SUCH COMPOSITIONS; CONDITION-RESPONSIVE CONTROL IN MICROBIOLOGICAL OR ENZYMOLOGICAL PROCESSES
    • C12Q2600/00—Oligonucleotides characterized by their use
    • C12Q2600/158—Expression markers

Definitions

  • the present invention relates to gene signatures that are indicative for the class and grade of lung cancer.
  • the present invention also relates to methods of diagnosing classifying and grading a lung cancer tumor.
  • the invention further relates to arrays and computer readable media comprising such gene signatures.
  • Lung cancer is the most common cancer in the world, both in rates of incidence and in rates of mortality, and is most prevalent in Europe and North America. Lung cancer is often attributed to both genetic factors and exposure to environmental factors (e.g., of radon gas, asbestos, and air pollution).
  • Lung cancers are classified by histological type: non-small cell lung carcinoma (NSCLC), small cell lung carcinoma (SCLC), and carcinoid.
  • NSCLC non-small cell lung carcinoma
  • SCLC small cell lung carcinoma
  • carcinoid carcinoid
  • Non-small cell lung carcinomas are grouped together based on similar prognosis and management and comprise three main sub-types: squamous cell lung carcinoma (SCC), adenocarcinoma and large cell lung carcinoma (LCLC).
  • SCC squamous cell lung carcinoma
  • adenocarcinoma adenocarcinoma
  • large cell lung carcinoma LCLC
  • SCC account for approximately 25% of lung cancer incidents.
  • Adenocarcinoma of the lung accounts for approximately 40% of lung cancer incidents.
  • LCLC is a heterogeneous group of undifferentiated malignant neoplasms originating from transformed epithelial cells in the lung, which accounts for about 10% of lung cancer incidents.
  • NSCLC are staged according to the TNM system.
  • the T category defines the primary tumor by size and whether it has spread into the surrounding tissue.
  • the N category identifies any lymph node involvement in and around the lungs.
  • the M category indicates whether the cancer has metastasized.
  • Stage 1 the lung cancer is small and localized to a single area.
  • Stage 2 and stage 3 lung cancers are larger, may have grown into surrounding tissues and may involve lymph nodes in and around the lungs.
  • Stage 4 lung cancers have metastasized to another region of the body.
  • Lung cancer is currently diagnosed by X-ray or computed tomography (CT) screening.
  • CT computed tomography
  • X-ray analysis is typically performed if a patient reports symptoms that may suggest lung cancer and may reveal an obvious mass, widening of the mediastinum (suggestive of spread to lymph nodes there), atelectasis (collapse or closure of alveoli), consolidation (pneumonia), or pleural effusion.
  • CT imaging is subsequently performed to provide additional information about the type and extent of disease.
  • bronchoscopy or biopsy may be performed to analyze the suspicious tissue.
  • the diagnostic tests for lung cancer usually require the disease to have progressed to the point that lung function is moderately affected and tumor growth is visible.
  • a diagnostic test that can identify, classify and grade lung tumors in patients at early stages.
  • the present invention is directed to gene signatures for classifying, diagnosing or grading lung cancer in an individual.
  • a first aspect of the invention provides a method of classifying or grading a lung cancer tumor in an individual at risk for or having lung cancer.
  • the method comprises classifying a test sample as stage 1 lung adenocarcinoma, stage 2 lung
  • the method comprises measuring the expression levels of at least 2 genes listed in Table 1 in a test sample; and applying one or more network-based methods, one or more machine-learning based methods, or a combination of the foregoing methods to the expression levels to obtain a classification of the test sample as stage 1 lung adenocarcinoma, stage 2 lung adenocarcinoma, stage 1 squamous cell carcinoma or stage 2 squamous cell carcinoma.
  • a differential pattern of expression levels of said at least 2 genes in the test sample classifies the lung cancer tumor as one of stage 1 lung adenocarcinoma, stage 2 adenocarcinoma, stage 1 squamous cell carcinoma or stage 2 squamous cell carcinoma.
  • the differential pattern of expression levels is identified by a classifier based on a plurality of genes listed in Table 1 , including said at least two genes, said classifier having been trained by in silico analysis or one or more feature selection and classification algorithms.
  • the di fferential pattern of expression levels is identi fied by a classifier based on a plurality of genes listed m i able 1 , including said at least two genes, said classifier having been trained by one or more algorithms selected from the group consisting of dual ensemble, generalized simulated annealing, T- filter, CORG, CO G combined with support vector machine, dual bagging, single and pairs, forward learning, Laplacian based learning and learning method based on network perturbation amplitude.
  • the classifier may be trained with at least the data in the Gene Expression Omnibus datasets GSE21 09, GSE 10245, GSE 1 8842 riQP 1774 ;
  • the method comprises detecting the expression level of at least 2 of the genes listed in Table 1 in a test sample obtained from the individual; and comparing the expression level of the genes listed in Table 1 in the test sample to the expression level of the genes listed in Table 1 in a control sample. In some embodiments, the method further comprises detecting the expression level of the genes listed in Table 1 in the control sample.
  • the at least 2 genes are selected from the group consisting of: ZIC2, LOC I 0013 1262, CD83, EML1 , PAIP 1 , NIPBL, CREB3L 1 , SLC37A 1 , and SFMBT2.
  • the test sample or the control sample is selected from blood, scrum, plasma, sputum, saliva, tissue, bronchia brushings, exhaled breath, and urine.
  • the tissue is lung tissue, such as tissue obtained by biopsy from a tumor.
  • the control sample is lung tissue, such as tissue obtained by biopsy from healthy lung tissue.
  • the healthy lung tissue is obtained from the individual at risk for or having lung cancer.
  • the control sample is obtained from an individual that does not have lung cancer.
  • the expression level of the genes listed in Table 1 in the test sample and the expression level of the genes listed in Table 1 in the control sample arc detected by measuring mRNA levels.
  • the expression level of the genes listed in Table 1 the test sample are detected by using a human genome-wide array, a human lung tissue array or a custom array comprising polynucleotides of a plurality of genes in Table 1 and said at least 2 genes.
  • the expression level of the genes listed in Table 1 in the test sample and the expression level of the genes listed in Table 1 in the control sample are detected by measuring the level of proteins encoded by the genes.
  • the expression level of the genes listed in ' fable 1 in the test sample and the expression level of the genes listed in ' fable 1 in the control sample are detected by measuring both mRNA levels and the level of
  • the expression level of the genes listed in Table 1 in the test sample and the expression level of the genes listed in ' fable 1 in the control sample are compared by in silico analysis (e.g., network-based analysis or machine-learning methods).
  • a second aspect of the invention provides an array for use in classi fying or grading a lung cancer tumor.
  • the array comprises polynucleotides immobilized on a solid surface that can hybridize to at least 10 lung cancer signature genes, wherein the lung cancer signature genes are selected from the group consisting of the genes listed in Table 1.
  • the array comprises polynucleotides hybridizing to at least 2 lung cancer signature genes immobilized on a solid surface, wherein the lung cancer signature genes arc selected from the genes listed in Table 1.
  • the array is not a human genome- wide array.
  • a third aspect of the invention provides a panel for use in classifying or grading a lung cancer tumor.
  • the panel comprises antibodies immobilized on a solid surface that bind to proteins encoded by at least 2 lung cancer signature genes, wherein the lung cancer signature genes are selected from the group consisting of the genes listed in Table 1 .
  • the computer readable medium comprises a lung cancer gene signature, wherein the gene signature comprises at least 2 genes selected from the genes listed in Table 1 .
  • the computer readable medium or computer program product comprises a classifier based on at least two genes listed in Table 1 , said classifier having been trained by in silico analysis or one or more feature selection and classification algorithms.
  • the classifier is trained by one or more algorithms selected from the group consisting of dual ensemble, generalized simulated annealing, T-filter, CORG, CORG combined with support vector machine, dual bagging, single and pairs, forward learning, Laplacian based learning and learning method based on network perturbation amplitude.
  • the classifier may be trained with at least the data in the Gene Expression Omnibus datasets GSE2109, GSE10245, GSE 18842 and GSE37745.
  • the at least two genes are selected from the group consisting of ZIC2, lAJU i uu i J l zoz, iiiviL i , vir l , NirBL, L, CJL>J L I , 5 LU ,' A L , anu
  • a fifth aspect of the invention provides a device for classifying and grading a lung cancer tumor.
  • the device comprises means for detecting the expression level of the genes listed in Table 1 in a test sample; means for correlating the expression level with a grade or classification of the tumor as stage 1 or stage 2 lung adenocarcinoma or stage 1 or stage 2 squamous cell carcinoma; and means for outputting the lung cancer tumor grade or classification.
  • the device further comprises means for detecting the expression level of the genes listed in Table 1 in a control sample.
  • a sixth aspect of the invention provides a kit for classifying and grading a lung cancer tumor.
  • the kit comprises a set of reagents that detects expression levels of the genes listed in Table 1 in a test sample and instructions for using said kit for classifying and grading the lung cancer tumor.
  • the kit is for assessing the prognosis of a lung cancer tumor in an individual.
  • the kit comprises a set of reagents that detects expression levels of the genes listed in Table 1 in a test sample from the individual and instructions for using said kit for determining the prognosis of the lung cancer tumor in said individual.
  • the set of reagents that detects expression levels of the genes listed in Table 1 in the test sample may also be used to detect expression levels of the genes listed in Table 1 in a control sample.
  • An seventh aspect of the invention provides a method of diagnosing the stage of a lung cancer tumor in an individual or of assessing the prognosis of an individual with a lung cancer tumor.
  • the method comprises a) measuring the expression level of at least 10 genes/biomarkers selected from the group consisting of the genes listed in Table 1 in a biological sample obtained from the individual; b) calculating a numerical biomarker score for the individual based on the expression levels of the biomarkers measured in step a); wherein the numerical biomarker score is predictive of the stage of lung cancer in the individual.
  • the method comprises measuring the expression level of at least 10 genes/biomarkers selected from the group consisting of the genes listed in Table 1 in a biological sample obtained from the individual;
  • step a) calculating a numerical biomarker score for the individual based on the expression levels of the biomarkers measured in step a): wherein the numerical biomarker score is predictive of the prognosis of the lung cancer in the individual.
  • the biological sample is selected from blood, scrum, plasma, sputum, saliva, tissue, bronchia brushings, exhaled breath, and urine.
  • the tissue is lung tissue, such as tissue obtained by biopsy from a tumor.
  • the expression level of the genes listed in Table 1 in the biological sample is detected by measuring mRNA levels.
  • the expression level of the genes listed in Table 1 the test sample are detected by using a human genomc-wide array, a human lung tissue array or a custom array comprising polynucleotides of a plurality of genes in Table 1 and said at least 2 genes.
  • the expression level of the genes listed in Table 1 in the biological sample is detected by measuring the level of proteins encoded by the genes.
  • the numerical biomarker score is calculated by in silico analysis.
  • the in silica analysis may be network based analysis or a machine-learning method.
  • the biomarkcrs are proteins encoded by the genes selected from the group consisting of the genes listed in Table 1.
  • a method of classi fying or grading a lung cancer tumor in an individual at risk for or having lung cancer comprising
  • the lung cancer tumor can be classified and staged as stage 1 lung adenocarcinoma, stage 2 adenocarcinoma, stage 1 squamous cell carcinoma or stage 2 squamous cell carcinoma.
  • test sample is selected from blood, serum, plasma, sputum, saliva, tissue, bronchia brushintis, exhaled breath, and urine.
  • the tissue is lung tissue. 5. The method according to paragraph 4, wherein the lung tissue is obtained by biopsy from a tumor.
  • control sample is selected from blood, serum, plasma, sputum, saliva, tissue, bronchia brushings, exhaled breath, and urine.
  • control sample is obtained from an individual that does not have lung cancer.
  • hybridization is a dot blot, a slot blot, an RNase protection assay, microarray hybridization, or in SHU hybridization.
  • the antibody assay is selected from Western analysis, immunofluorescence, ELIS/v, and
  • An array comprising polynucleotides hybridizing to at least 2 lung cancer signature genes immobilized on a solid surface, wherein the lung cancer signature genes are selected from the group consisting of at least 2 genes listed in Table 1 .
  • ⁇ panel comprising antibodies immobilized on a solid surface that bind to proteins encoded by at least 2 lung cancer signature genes, wherein the lung cancer signature genes arc selected from the group consisting of the genes listed in Table 1 .
  • a computer readable medium comprising a lung cancer gene signature, wherein the gene signature comprises at least 2 genes selected from the genes listed in Table 1 .
  • a device for classifying and grading a lung cancer tumor comprising: means for detecting the expression level of at least 2 genes listed in Table 1 in a test sample; means for correlating the expression level with a classification of the tumor as stage 1 or stage 2 lung adenocarcinoma or stage 1 or stage 2 squamous cell carcinoma; and means for outputting the lung cancer tumor classification.
  • kits for classifying and grading a lung cancer tumor comprising one or more reagents that detects expression levels of at least 2 genes listed in Table 1 in a test sample and instructions for using said kit for classifying and grading a lung cancer tumor.
  • a kit for assessing the prognosis of lung cancer in an individual comprising a set of reagents that detects expression levels of at least 2 genes listed in Table 1 in a test sample from the individual and instructions for using said kit for determining the prognosis of lung cancer in said individual.
  • a method of diagnosing the stage of lung cancer in an individual comprising the steps of:
  • step b) calculating a numerical biomarkcr score for the individual based on the expression levels of the biomarkcrs measured in step a);
  • the numerical biomarkcr score is predictive of the stage of lung cancer in the individual.
  • a method of assessing the prognosis of an individual with lung cancer comprising the steps of:
  • step b) calculating a numerical biomarkcr score for the individual based on the expression levels of the biomarkcrs measured in step a); wherein the numerical biomarkcr score is predictive of the prognosis of the lung cancer in the individual.
  • a method of diagnosing, prognosing, classifying or grading lung cancer in a biological sample or an individual comprising measuring the expression levels of at least 2 genes listed in Table 1 in the biological sample or a test sample from the individual; and applying one or more network-based methods, one or more machine-learning based methods, or a combination of the foregoing methods to the expression levels to obtain a classification of the test sample as stage 1 lung adenocarcinoma, stage 2 lung adenocarcinoma, stage 1 squamous cell carcinoma or stage 2 squamous cell carcinoma.
  • the classifier is obtained by training with a network-based method or a machine-learning based method using datascts obtained from subjects with stage 1 lung adenocarcinoma, stage 2 lung adenocarcinoma, stage 1 squamous cell carcinoma or stage 2 squamous cell carcinoma and datascts from subjects without lung cancer.
  • ⁇ method of classifying or grading a lung cancer tumor in an individual at risk for or having lung cancer comprising detecting the expression level of at least 2 genes listed in Table 1 in a test sample obtained from the individual; wherein a differential pattern of expression levels of said at least 2 genes in the test sample classifies the lung cancer tumor as one of stage 1 lung adenocarcinoma, stage 2 adenocarcinoma, stage 1 squamous cell carcinoma or stage 2 squamous cell carcinoma.
  • the di fferential pattern of expression levels is identified by a classifier based on a plurality of genes listed in Table 1 , including said at least two genes, said classifier having been trained by one or more algorithms selected from the group consisting of dual ensemble, generalized simulated annealing, T-filtcr, CORG, CORG combined with support vector machine, dual bagging, single and pairs, forward learning,
  • An array comprising polynucleotides hybridizing to at least 2 lung cancer signature genes immobilized on a solid surface, wherein the lung cancer signature genes are selected from the genes listed in Table 1 and said array is not a human genome-wide array.
  • a device comprising antibodies immobilized on a solid surface that bind to proteins encoded by at least 2 lung cancer signature genes, wherein the lung cancer signature genes arc selected from the group consisting of the genes listed in Table 1.
  • a computer readable medium or computer program product comprising a classifier based on at least two genes listed in Table 1 , said classifier having been trained by in silico analysis or one or more feature selection and classification algorithms.
  • a kit for classifying and grading a lung cancer tumor or for assessing the prognosis of lung cancer in an individual comprising one or more reagents that detects expression levels of at least 2 genes listed in Table 1 in a test sample and instructions for using said kit for classifying and grading a lung cancer tumor or for determining the prognosis of lung cancer in said individual.
  • Figure 1 provides a features selection and classification algorithm(s) used for prediction of a gene signature.
  • antibody refers to an immunoglobulin molecule capable of specific binding to a target, such as a carbohydrate, polynucleotide, lipid, polypeptide, etc., through at least one antigen recognition site, located in the variable region of the immunoglobulin molecule.
  • the term is intended to encompass not only intact polyclonal or monoclonal antibodies, but also engineered antibodies (e.g., chimeric, humanized and/or derivatized to alter effector functions, stability and other biological activities) and fragments thereof (such as Fab, Fab' , F(ab')2, Fv), single chain (ScFv) and domain antibodies, including shark and camelid antibodies), and fusion proteins comprising an antibody portion, multivalent antibodies, rnultispecii c antibodies (e.g., bispecific antibodies so long as they exhibit the desired biological activity) and antibody fragments as described herein, and any other modified configuration of the immunoglobulin molecule that comprises an antigen recognition site.
  • engineered antibodies e.g., chimeric, humanized and/or derivatized to alter effector functions, stability and other biological activities
  • fragments thereof such as Fab, Fab' , F(ab')2, Fv), single chain (ScFv) and domain antibodies, including shark and camelid antibodies
  • An antibody includes an antibody of any class, such as IgG, IgA, or IgM (or sub class thereof), and the antibody need not be of any particular class.
  • immunoglobulins can be assigned to different classes. There are five major classes of immunoglobulins: IgA, IgD, IgE, IgG, and IgM, and several of these may be further divided into subclasses (isotypes), e.g., IgG l , IgG2, IgG3, IgG4, IgAl and IgA2 in humans.
  • the heavy chain constant domains that correspond to the different classes of immunoglobulins are called alpha, delta, epsilon, gamma, and mu, respectively.
  • the subunit structures and three dimensional configurations of different classes of immunoglobulins are well known.
  • array refers to the arrangement of biomarker detection molecules, such as nucleic acid probes or antibodies, on a solid support that allows for high-throughput screening of a sample to detect the presence and/or quantity of a biomarker.
  • biomarker detection molecules such as nucleic acid probes or antibodies
  • Such arrays may be used, e.g. , to evaluate the expression levels of several genes of interest in a single high-throughput reaction,
  • the array may be a nucleic acid array, such as a nucleic acid microarray; a protein array, such as a protein microarray; a peptide array, such as a peptide microarray; a tissue microarray, such as a tissue microarray or an antibody microarray, such as an antibody microarray.
  • the solid substrate may be a microscopic bead, a glass slide, a plastic chip or a silicon chip.
  • biomarkcr refers to a characteristic whose presence, absence or level indicates a biological state. Typically, the properties of biomarkers indicate a normal process, a pathogenic process or a response to a pharmaceutical or therapeutic intervention.
  • ⁇ biomarker can be a cell, a gene, a gene product, an enzyme, a hormone, a protein, a peptide, an antibody, a nucleic acid molecule, a metabolite, a lipid, a free fatty acid, cholesterol or some other chemical compound.
  • ⁇ biomarker can be a morphologic biomarkcr (for example, a histological change, DNA ploidy, malignancy-associated changes in the cell nucleus and premalignant lesions) or a genetic biomarker (for example, DNA mutations, DNA adducts and apoptotic index).
  • morphologic biomarkcr for example, a histological change, DNA ploidy, malignancy-associated changes in the cell nucleus and premalignant lesions
  • a genetic biomarker for example, DNA mutations, DNA adducts and apoptotic index
  • the term "classifying a lung cancer" refers to a method for determining the type of tumor from which a subject suffers.
  • a subject can suffer from several different types of lung cancer, including but not limited to, adenocarcinoma, squamous cell carcinoma, large cell lung carcinoma, other non-small cell lung carcinomas, small cell lung carcinoma, carcinoid and unspecified lung cancer.
  • a lung cancer tumor may be classified as one of these types o f lung cancer.
  • a tumor may be classified based on histology, genetics or the presence, absence, alteration or levels of biomarkers.
  • a lung cancer may be classified based on the lung cancer gene signature. The lung cancer may be classified as either adenocarcinoma or squamous cell carcinoma.
  • a computer program refers to a sequence of instructions, written to perform a specified task within a computer.
  • a computer program product is described, the product comprising computer-readable instructions that, when executed in a computerized system comprising at least one processor, cause the processor to carry out one or more steps of any of the methods described above.
  • a computerized system is described, the system comprising a processor configured with non-transitory computer-readable instructions that, when executed, cause the processor to carry out any of the methods described herein.
  • the computer program product and the computerized methods described herein may be implemented in a computerized system having one or more computing devices, each including one or more processors.
  • the computerized systems described herein may comprise one or more engines, which include a processor or devices, such as a computer, microprocessor, logic device or other device or processor that is configured with hardware, firmware, and software to carry out one or more of the computerized methods described herein. Any one or more of these engines may be physically separable from any one or more other engines, or may include multiple physically separable components, such as separate processors on common or different circuit boards.
  • the computer systems of the present invention comprises means for implementing the methods and its various embodiments as described herein.
  • the computerized system described herein may include a distributed computerized system having one or more processors and engines that communicate through a network interface. Such an implementation may be appropriate for distributed computing over multiple communication systems.
  • Non-volatile media include, for example, optical, magnetic, or opto- magnetic disks, or integrated circuit memory, such as flash memory.
  • Volatile media include, for example, dynamic random access memory (DRAM), which typically constitutes the main memory.
  • DRAM dynamic random access memory
  • Computer-readable media include, for example, a floppy disk, a flexible disk, hard disk, magnetic tape, magnetic cards, magnetic ink characters, magnetic drums, any other magnetic medium, a CD-ROM, DVD, any other optical medium, punch cards, paper tape, any other physical medium with patterns of holes, a RAM, a PROM, an EPROM or EEPROM (electronically erasable programmable read-only memory), a
  • control sample' refers to a sample against which a test sample is compared in order to diagnose, prognose, classify or grade the test sample
  • ⁇ control sample may be healthy tissue or may be a well-characterized tumor sample, including but not limited to, stage 1 adenocarcinoma, stage 2 adenocarcinoma, stage 1 squamous cell carcinoma, or stage 2 squamous cell carcinoma.
  • a control sample can be analyzed concurrently with or separately from the test sample, including before or after analyzing the test sample.
  • control sample may be developed as a medical standard for comparison.
  • analysis of control samples has developed medical standards for normal fed and fasted blood glucose levels; normal, at risk, and hypertensive blood pressures, and normal resting heart rates.
  • control sample includes samples that provided a medical standard. Accordingly, a test sample may be compared against a medical standard generated from control samples. For example, expression of a variant or mutated form of a gene may be indicative of a change medical condition.
  • a control sample may be lung tissue, such as tissue obtained by biopsy from a healthy portion of a lung (e.g., distant from a suspected tumor), or some other sample.
  • a control sample may be blood, blood cells, serum, plasma, sputum, saliva, tissue, bronchial washing, bronchial aspirates, bronchia brushings, exhaled breath, lymph fluid, and urine.
  • Tissue specimens, such as those obtained by biopsy may be fixed (e.g., formaldehyde- fixed paraffin- embedded (FFPE)).
  • the control sample may be obtained from a tissue bank.
  • the control sample may also be obtained from a cadaver or an organ donor.
  • a fold-change is a number describing how much a measurement at a node (or biological entity) changes from an initial value to a final value between control data and treatment data, or between two sets of data representing different treatment conditions.
  • the fold-change number may represent the logarithm of the fold-change of the activity of the biological entity between the two conditions. ⁇ confidence interval for the significance of the fold- change number may also be assessed.
  • a normal state e.g., a normal state, an at-risk state, a diseased state (e.g., stage 1 lung adenocarcinoma, stage 2 lung
  • ⁇ gene signature may be characterized by which genes are expressed and/or at what level each gene is expressed. Gene signatures are particularly useful in diagnosing, prognosing, classifying or grading complex diseases states, which result from the combination of several genetic and environmental factors.
  • the gene signatures disclosed herein may be used, e.g., for the diagnosis, prognosis, classification or grading of lung cancer tumors in an individual .
  • the gene signature may be unique to the class and grade of the tumor.
  • stage 0 refers to a carcinoma in situ
  • stage 1 refers to cancers that are localized to one part of the body
  • stage 2 refers to cancers that arc locally advanced
  • stage 3 refers to cancers that are further advanced locally (e.g., as evidenced by increased lymph node involvement)
  • stage 4 refers to cancers that have metastasized.
  • Lung cancers are typically staged according to the TNM system.
  • the T category defines the primary tumor by size and whether it has spread into the surrounding tissue.
  • the N category identifies any lymph node involvement in and around the lungs.
  • the M category indicates whether the cancer has metastasized.
  • ⁇ tumor may be graded based on histology, genetics or the presence, absence, or levels of biomarkers.
  • a lung cancer may be graded based on its gene signature. The lung cancer may be graded as stage 1
  • adenocarcinoma stage 2 adenocarcinoma, stage 1 squamous cell carcinoma or stage 2 squamous cell carcinoma.
  • silico analysis refers to analysis performed on a computer or via computer simulation.
  • Gene signature analysis involves detection of gene expression based on identity and expression level for a multitude of genes.
  • In silico analysis may apply one or more network-based methods, one or more machine-learning based methods, or a combination of the foregoing methods to the expression levels to obtain a classification of the test sample, e.g., as stage 1 lung adenocarcinoma, stage 2 lung adenocarcinoma, stage 1 squamous cell carcinoma or stage 2 squamous cell carcinoma.
  • Comparisons between expression levels from test samples and control samples may require computer analysis to determine the degree and significance of any changes observed. See, e.g. , U.S. Provisional Patent Application entitled “Systems and Methods relating to Network-based Biomarker Signatures,” filed concurrently with the instant application,
  • the term "individual” refers to a vertebrate, preferably a mammal.
  • the mammal can be, without limitation, a mouse, a rat, a cat, a dog, a horse, a pig, a cow, a non-human primate or a human.
  • the term "individual at risk for" lung cancer” refers to an individual who is predisposed to lung cancer. Predisposition to lung cancer may be due to one or more genetic or environmental factors. For example, an individual related to a lung cancer patient is 2.4 times more likely to get lung cancer than an individual who is not related to a lung cancer patient. Further, exposure to environmental factors such as radon gas, asbestos, tobacco smoke, and air pollution can increase the risk for lung cancer and predispose an individual to lung cancer.
  • the term "individual having lung cancer” or “individual suffering from lung cancer” refers to an individual experiencing uncontrolled cell growth in the tissues of the lung. Lung cancers typically form solid tumors, which can be observed on a chest X-ray or by a CT scan.
  • MALDI-TOF matrix-assisted laser
  • Time-of-fiight (TOF) mass spectrometry refers to a method in which an ion's mass-to-charge ratio is determined by measuring the time that it takes an ionized particle to reach a detector at a known distance.
  • network-based analysis refers to an approach to identify biomarkers that is based on the properties of groups of functionally interrelated genes that form a network in a biological system, instead of treating individual genes in the biological system a priori as completely independent and identical.
  • machine learning methods refers to methods that allow a machine, such as a programmable computer, to improve its performance at a certain predictive task that is based on the known properties of examples or training data.
  • Machine learning methods include, without limitation, support vector machines (SVMs), network-based SVMs, ensemble classifiers, neural network-based classifiers, logistic regression classifiers, decision tree-based classifiers, classifiers employing a linear discriminant analysis technique, a random-forest analysis technique, or both.
  • numbererical biomarker score refers to a number that is representative of the result(s) of one or more of the network-based analysis or machine learning methods.
  • polynucleotide hybridizing to refers to a polynucleotide molecule that binds to a target nucleic acid molecule through complementary base pair sequencing. Hybridization typically requires two nucleic acids that contain complementary sequences, although depending on the stringency of the
  • hybridization mismatches between bases are possible.
  • the appropriate stringency for hybridizing nucleic acids depends on the length of the nucleic acids and the degree o f complementation, variables well known in the art. Exemplary high stringent hybridization conditions are equivalent to about 20-27 °C below the melting temperature (T m ) of the DNA duplex formed in about 1 M salt.
  • T m melting temperature
  • Many equivalent procedures exist and several popular molecular cloning manuals describe suitable conditions for highly stringent hybridization and, furthermore, provide formulas for calculating the length of hybrids expected to be stable under these conditions (see, e.g., Current Protocols in Molecular Biology, John Wiley & Sons, N. Y.
  • Timc-of-flight (TOP) mass spectrometry refers to a method in which an ion's mass-to-charge ratio is determined by measuring the time that it takes an ionized particle to reach a detector at a known distance.
  • test sample refers to a sample obtained from an individual at risk for, having or suffering from lung cancer.
  • a test sample may be any sample suspected of containing or exhibiting a biomarker.
  • the test sample is analyzed and compared to a control sample, including medical standards developed from control samples, to diagnose, prognose, classify or grade lung cancer in the individual.
  • ⁇ test sample may be obtained from lung tissue, such as tissue obtained by biopsy from a tumor, or other biological tissue.
  • a test sample may be blood, blood cells, scrum, plasma, sputum, saliva, tissue, bronchial washing, bronchial aspirates, bronchia brushings, exhaled breath, lymph fluids, and urine.
  • Tissue specimens, such as those obtained by biopsy may be fixed (e.g., formaldehydc- fixed paraffin-embedded (FFPE)).
  • FFPE formaldehydc- fixed paraffin-embedded
  • a training data set includes a set of samples, and each sample may correspond to a measurement from a different patient.
  • a machine learning technique is applied to the training data set to generate a "classifier,” which corresponds to a way of assigning each sample in the training data set to a category (such as "disease positive” or “disease free”).
  • a training class set is known.
  • the training class set includes a known category assigned to each sample (or person).
  • the categories predicted by the classifier are compared to the known categories. I f the predicted categories mostly match the known categories, the classifier has performed well.
  • the parameters of the machine learning technique may be updated, and the updated machine learning technique is applied. These steps are repeated until the performance of a classifier exceeds a threshold, and the final classifier is provided.
  • the final classifier may then be applied to a test data set.
  • the test data set may correspond to measured samples from different patients, but the patients in the test data set may have unknown categories (disease states). Thus, applying the final classifier to the test data set thus allows for prediction of the disease states of the patients.
  • the gene signature comprises at least 2 genes selected from the genes listed in Tabic 1. In some embodiments, the gene signature comprises at least 2, at least 3, at least 4, at least 5, at least 10, at least 15, at least 20, at least 25, at least 30, at least 35, at least 40, at least 45, at least 50, at least 55, at least 60, at selected from the genes listed in Table 1. In some embodiments, the gene signature comprises each of the genes listed in Table 1.
  • said at least 2, at least 3, at least 4, or at least 5 genes are selected from the group consisting o : ZIC2, LOCI 00131262, CD83, EML1 , PAIP 1 , NIPBL, CREB3L 1 , SLC37A 1 , and SFMBT2, which are the genes that appear in 4 of the 5 lists generated in Example
  • At least 2, at least 3, at least 4, at least 5, at least 10, at least 1 5, at least 20, at least 25, at least 30, at least 35, at least 40, at least 45, at least 50, at least 55, at least 60, at least 65, or at least 70 of the genes selected from the genes listed in Table 1 have increased expression compared to a control sample.
  • at least 2, at least 3, at least 4, at least 5, at least 10, at least 15, at least 20, at least 25, at least 30, at least 35, at least 40, at least 45, at least 50, at least 55, at least 60, at least 65, or at least 70 of the genes selected from the genes listed in Table 1 have decreased expression compared to a control sample.
  • the gene signature includes a degree of up- regulation of a subset of genes in the gene signature compared to the control sample.
  • each up-regulated gene in the gene signature may, independently, be up-regulated at least 1.5-fold, at least 2-fold, at least 2.5-fold, at least 3-fold, at least 3.5-fold, at least 4-fold, at least 4.5-fold, at least 5-fold, at least 6-fold, at least 7-fold, at least 8-fold, at least 9-fold, at least 10-fold, at least 100- fold, at least 1 ,000-fold or more compared to the control sample.
  • the gene signature includes a degree of down-regulation of ; subset of genes in the gene signature compared to the control sample.
  • each down-regulated gene in the gene signature may, independently, be down-regulated at least 1 .5-fold, at least 2-fold, at least 2.5-fold, at least 3-fold, at
  • Vlll A, B, C, D, E, F, G, H, and 1;
  • xlvii. A, B, C, D, E, F, G, I I, I, J, K, L, M, N, O, P, Q, R, S, T, U, W, X, Y, Z.
  • Iviii A, B, C, D. E, F, G, I I, I, J, K, L, M, N, 0, P, Q, R, S, T. U, W, X, Y, Z, AA, AB, AC, AD, AE, AF, AG, AH, AI, AJ, AK, AL, AM, AN, AO, AP,
  • Ixi A, B, C, D, E, F, G, II, I, J, , L, M, N, O, P, Q, R, S, T, U, W, X, Y, Z, AA, AB, AC, AD, AE, AF, AG, AH, AI, AJ, AK, AL, AM, AN, AO, AP,
  • BG B1 I, BI, BJ, BK, BL, BM, BN, BO, BP, BQ, BR and BS;
  • BG BH, BI, BJ, BK, BL, BM, BN, BO, BP, BQ, BR, BS, BT, BU, BV, BW and BX;
  • AA AB. AC, AD, AE, AF, AG, AH, AI, AJ, AK, AL. AM, AN, AO, AP, AQ, AR. AS, AT, AU, AV, AW, AX, AY, AZ, BA, BB, BC, BD, BE, BF,
  • BG BI I, BI, BJ, BK, BL, BM, BN, BO, BP, BQ, BR, BS, BT, BU, BV, BW, BX and BY;
  • BG BI I. BI, BJ, BK, BL, BM, BN, BO, BP, BQ, BR, BS, BT, BU, BV,
  • BG BI I, BI, BJ, BK, BL, BM, BN, BO, BP, BQ, BR, BS, BT, BU, BV,
  • BG BI I. BI, BJ, BK, BL, BM. BN, BO, BP, BQ, BR, BS, BT, BU, BV,
  • BG BI I, BI, BJ, BK, BL, BM, BN, BO, BP, BQ, BR, BS, BT, BU, BV,
  • BW, BX, BY, BZ, CA, CB and CC ixxxi.
  • Ixxxvii. A, B, C, D, E, F, G, I I, I, J, K, L, M, N, 0, P, Q, R, S, T, U, W, X, Y, Z,
  • CG, CH, CI and CJ are independently selected from the genes listed in Table 1 and each of A, B, C, D, E, F, G, I I, I, J, K, L, M. N, O, P, Q, R, S, T, U, W, X, Y, Z, AA, AB, AC, AD, AE, AF, AG, Al l, AI, AJ, AK, AL, AM, AN, AO, AP, AQ, AR, AS, AT, AU, AV, AW, AX, AY, AZ, BA, BB, BC, BD, BE, BF, BG, BH, BI, BJ, BK, BL, BM, BN, BO, BP, BQ, BR, BS, BT, BU, BV, BW, BX, BY, BZ, CA, CB, CC, CD, CE, CF, CG, CH,
  • the biomarkers and gene signatures of the invention may be used in methods of diagnosing, prognosing, classifying or grading lung cancer in a biological sample or an individual.
  • the invention encompasses a method for clilSsi f v ' n f ' tpif camnl p a « cHi p p 1 ⁇ aH nnr!irf in ma 3 ⁇ 4ta f > ? l i m p
  • adenocarcinoma stage 1 squamous cell carcinoma or stage 2 squamous cell carcinoma
  • the method comprising: measuring the expression levels of at least 2 genes listed in Table 1 in a test sample; and applying one or more network-based methods, one or more machine-learning based methods, or a combination of the foregoing methods to the expression levels to obtain a classification of the test sample as stage 1 lung adenocarcinoma, stage 2 lung adenocarcinoma, stage 1 squamous cell carcinoma or stage 2 squamous cell carcinoma.
  • stage 1 lung adenocarcinoma stage 2 lung adenocarcinoma
  • stage 1 squamous cell carcinoma stage 2 squamous cell carcinoma
  • the expression levels of at least 2, at least 3, at least 4, at least 5, at least 10, at least 1 5, at least 20, at least 25, at least 30, at least 35, at least 40, at least 45, at least 50, at least 55, at least 60, at least 65, at least 70, at least 75, at least 80, least about 85 or at least 87 genes selected from the genes listed in Table 1 are measured.
  • a differential pattern of expression levels of said at least 2 genes in the test sample classifies the lung cancer tumor as one of stage 1 lung adenocarcinoma, stage 2 adenocarcinoma, stage 1 squamous cell carcinoma or stage 2 squamous cell carcinoma.
  • the methods of the invention can be used to identify a gene signature and a classifier (e.g., a gene-signaturc-based classifier) that can distinguish datasets obtained from various classes and stages of lung cancer (e.g. stage 1 lung adenocarcinoma, stage 2 lung adenocarcinoma, stage 1 squamous cell carcinoma or stage 2 squamous cell carcinoma).
  • a classifier or a previously established standard may be used to determine whether a test sample is a lung cancer sample or the class and stage of lung cancer from which the test sample is obtained.
  • a classifier that is obtained by training with network-based or machine-learning methods using datasets obtained from subjects with various classes and stages of lung cancer (e.g. stage 1 lung adenocarcinoma, stage 2 lung adenocarcinoma, stage 1 squamous cell carcinoma or stage 2 squamous cell carcinoma) and datasets from subjects without lung cancer, can be used.
  • lung cancer e.g. stage 1 lung adenocarcinoma, stage 2 lung adenocarcinoma, stage 1 squamous cell carcinoma or stage 2 squamous cell carcinoma
  • one or more numerical scores (e.g., average fold change or rank abs tval) generated by the algorithms described herein may be used as a previously established standard.
  • the levels of expression of one or more of the genes listed in Table 1 in a test sample may be compared to the previously established standard, and the comparison may be used to classify the test sample as a lung cancer sample or a normal sample. In some embodiments, the comparison may be used to classify the test sample as stage 1 lung adenocarcinoma, stage 2 lung
  • adenocarcinoma stage 1 squamous cell carcinoma or stage 2 squamous cell carcinoma.
  • the invention provides a method of diagnosing, prognosing, classifying or grading lung cancer in a biological sample, wherein the method comprises determining the properties (for example, absence, presence or expression level) of one or more genes listed in Table 1 in the biological sample; and applying in silica analysis with a classifier obtained from a network-based method, a machine-learning based method, or a combination of the foregoing methods.
  • the classifier can be obtained from the network-based method, a machine-learning based method, or a combination of the foregoing methods by training with datascts obtained from subjects with stage 1 lung adenocarcinoma, stage 2 lung adenocarcinoma, stage 1 squamous cell carcinoma stage 2 squamous cell carcinoma, healthy subjects, or a combination of two or more of the foregoing.
  • a classifier may be linked to a specific prognosis of the lung cancer in the individual who provided the biological sample.
  • the classifier may indicate that the lung cancer in the individual who provided the biological sample is stage 1 lung adenocarcinoma, stage 2 lung adenocarcinoma, stage 1 squamous ceil carcinoma or stage 2 squamous cell carcinoma.
  • the classifier may also indicate that a particular treatment regimen should be used to treat the individual who provided the biological sample.
  • the methods of the invention comprise obtaining a test sample (such as a lung biopsy) from an individual, determining the absence, presence or expression level of one or more of the genes listed in Table 1 in the test sample, comparing said absence, presence or expression level to the absence, presence or expression level of the same gene(s) in a control sample, and selecting a lung cancer treatment regimen based on the comparison.
  • a test sample such as a lung biopsy
  • the invention provides a method for monitoring the progress of a lung cancer treatment in an individual, said method comprising determining at suitable time intervals before, during, or after therapy (for example, at different lime points during the treatment) in a sample taken from said individual differential expression of a panel of at least 2 genes selected from the genes listed in Table 1.
  • the invention encompasses a method that comprises collecting data on the properties of one or more genes in the gene signature without generating a gene signature.
  • the method of the invention comprises obtaining a test sample from an individual, and detecting the absence, presence or the expression level of one or more of the genes listed in Table i in the sample.
  • the invention encompasses a method that comprises using data on the properties of one or more genes in a gene signature that are already collected as training data to generate an improved gene signature using one or more network-based methods, one or more machine learning methods, or a combination of the foregoing methods.
  • the invention encompasses a method that comprises collecting data on the properties of one or more genes in a biological system which is included in a gene signature, and using the data to predict a classification of the state of the biological system associated with the collected data.
  • the method comprises detecting the expression level of at least 2 of the genes listed in Table 1 in a test sample obtained from the individual; and comparing the expression level of the genes listed in Table 1 in the test sample to the expression level of the genes listed in Table 1 in a control sample.
  • the lung cancer subtype and stage may be determined.
  • the lung cancer is classified as stage 1 lung adenocarcinoma, stage 2 lung adenocarcinoma, stage 1 squamous cell carcinoma or stage 2 squamous cell carcinoma.
  • the method further comprises detecting the expression level of the genes listed in Table 1 in the control sample.
  • the expression levels of at least 2, at least 3, at least 4, at least 5, at least 10, at least 15, at least 20, at least 25, at least 30, at least 35, at least 40, at least 45, at least 50, at least 55, at least 60, at least 65, at least 70, at least 75, at least 80, at least 84 genes or all genes listed in Table 1 are detected.
  • the expression level of the genes listed in Tabic 1 in the test sample and the expression level of the genes listed in Table 1 in the control non-tumor biological sample are compared by in silico analysis.
  • the in silica analysis may be network based analysis or a machine-learning method.
  • the test sample is selected from blood, serum, plasma, sputum, saliva, tissue, bronchia brushings, exhaled breath, and urine.
  • the tissue is lung tissue, such as tissue obtained by biopsy from a tumor.
  • the control sample is selected from blood, serum, plasma, sputum, saliva, tissue, bronchia brushings, exhaled breath, and urine.
  • the tissue is lung tissue, such as tissue obtained by biopsy from healthy lung tissue.
  • the healthy lung tissue is obtained from the individual at risk for or having lung cancer.
  • the control sample is obtained from an individual that does not have lung cancer.
  • the expression level of the genes listed in Table 1 in the test sample and the expression level of the genes listed in Table 1 in the control sample are detected by measuring mRNA levels.
  • mRNA level is measured by amplification, hybridization, mass spectroscopy, serial analysis of gene expression, or massive parallel signature sequencing.
  • the amplification is reverse transcription PGR, real time quantitative PGR, dilTerential display or TaqMan PGR.
  • the hybridization is a dot blot, a slot blot, an RNasc protection assay, microarray hybridization, or in situ hybridization.
  • the mass spectroscopy may be MALDI-TOF mass spectroscopy.
  • the expression level of the genes listed in Table 1 in the test sample arc detected by using a human genome-wide array, a human lung tissue array or a custom array comprising polynucleotides of a plurality of genes in Table 1 .
  • the expression level of the genes listed in Table 1 in the test sample and the expression level of the genes listed in Table 1 in the control sample are detected by measuring the level of proteins encoded by the genes.
  • the protein level is measured using an antibody assay or by mass spectroscopy.
  • the antibody assay is selected from
  • the mass spectroscopy may be MALDI-TOF mass spectroscopy or SELDI-TOF mass
  • the expression level of the genes listed in Table 1 in the test sample and the expression level of the genes listed in Table 1 in the control sample are detected by measuring both mRNA levels and the level of proteins encoded by the genes. In some embodiments, expression levels are measured using the amplification, hybridization, mass spectroscopy, serial analysis of gene expression, massive parallel signature sequencing, and antibody assays discussed above. Mcthods of Biomarkcr Detection, Arrays and Panels
  • Detection of the nucleic acid and/or protein biomarkcrs described herein in a test sample or a control sample may be performed in a variety of ways.
  • the methods of the invention rely on the detection of the presence or absence of biomarkcr genes and/or biomarkcr gene expression, or the qualitative or quantitative assessment of either over- or under-cxpression of a biomarkcr gene in a population of cells in a test sample relative to a standard (for example, a control sample).
  • a standard for example, a control sample.
  • Such methods utilize reagents such as biomarkcr polynucleotides and biomarkcr antibodies.
  • the presence, absence or level of expression of a biomarkcr gene may be determined by measuring the amount of biomarkcr messenger RNA (mRNA), for example, by DNA-DNA hybridization, RNA-DNA hybridization, reverse transcription-polymerase chain reaction (PGR), real time quantitative PGR, differential display or TaqMan PGR; followed by comparing the results to a reference based on a control sample (for example, samples from clinically- characterized patients and/or cell lines of a known genotypc/phenotype).
  • mRNA biomarkcr messenger RNA
  • mRNA expression or turnover may be measured.
  • Hybridization, mass spectroscopy e.g., MALDI-TOF or SELDI-TOF mass spectroscopy
  • serial analysis of gene expression or massive parallel signature sequencing assays can also be performed.
  • hybridization assays include a singleplcx or a multiplexed aptamcr assay, a dot blot, a slot blot, an RNase protection assay, microarray hybridization, Southern or Northern hybridization analysis and in situ hybridization (e.g. , fluorescent in situ hybridization).
  • microarray-bascd assays that can be used to detect and quantify the amount of biomarkcr gene transcript using cDNA- or oligonucleotide-bascd arrays.
  • Microarray technology allows multiple biomarkcr gene transcripts and/or samples from different subjects to be analyzed in one reaction.
  • mRNA isolated from a sample is converted into labeled nucleic acids by reverse transcription and optionally in vitro transcription (cDNAs or cRNAs labelled with, for example, Cy3 or Cy5 dyes) and hybridized in parallel to probes present on an array. See, for example, Schulze et al, Nature Cell Biol.
  • the expression level of the genes listed in Table 1 in the test sample are detected by using a human genome-wide array, a human lung tissue array or a custom array comprising polynucleotides of a plurality of genes in Table 1 .
  • biomarkers are detected using reagents that specifically detect the biomarker.
  • reagents may bind to a target gene or a target gene product (e.g., mRNA or protein), such that levels of the gene product may be quantified.
  • reagents may be nucleic acid molecules that hybridize to the mRNA or cDNA of target gene products.
  • the reagents may be molecules that label mRNA or cDNA for later detection, e.g., by binding to an array.
  • the reagents may bind to proteins encoded by the genes of interest.
  • the reagent may be an antibody or a binding protein that specifically binds to a protein encoded by a target gene of interest.
  • the reagent may label proteins for later detection, e.g., by binding to an antibody on a panel.
  • reagents are used in histology to detect histological and/or genetic changes in a sample.
  • the present invention provides isolated biomarker polynucleotides or variants thereof, which can be used, for example, as hybridization probes or primers ("biomarker probes” or “biomarker primers”) to detect or amplify nucleic acids encoding a biomarker polypeptide, particularly a biomarker polypeptide encoded by a biomarker gene or polynucleotide selected from the group depicted in Table 1.
  • biomarker probes or “biomarker primers”
  • Nucleic acid molecules comprising nucleic acid sequences encoding the biomarker polypeptides or proteins of the invention, or genomic nucleic acid sequences from the biomarker genes (e.g., intron sequences, 5 ' and 3 ' untranslated sequences), or complements thereof (i .e. , anti sense polynucleotides), are collectively referred to as "biomarker genes,” “biomarker polynucleotides” or
  • biomarker nucleic acid sequences of the invention.
  • the present invention also provides isolated biomarker polynucleotides or variants thereof, which can be used, for example, as hybridization probes or primers ("biomarker probes” or “biomarker primers”) to detect or amplify nucleic acids encoding a biomarker polypeptide of the invention.
  • biomarker gene product thus
  • RNA encompasses both mRNA as well as translated polypeptide as a gene product of a biomarker.
  • the isolated biomarker polynucleotide according to the invention may comprise flanking sequences (i. e. , sequences located at the 5 ' or 3 ' ends of the nucleic acid), which naturally flank the nucleic acid sequence in the genomic DNA of the organism from which the nucleic acid is derived.
  • flanking sequences i. e. , sequences located at the 5 ' or 3 ' ends of the nucleic acid
  • the isolated biomarker polynucleotide can comprise less than about 5 kb, 4 kb, 3 kb, 2 kb, 1 kb, 0.5 kb or 0.1 kb of nucleotide sequences which naturally flank the coding sequence in genomic DNA of the cell from which the nucleic acid is derived.
  • polynucleotide is about 10-20, 21 -50, 51 - 100, 101 -200, 201 -400, 401 -750, 75 1 - 1000, or 1001 - 1500 bases in length.
  • the biomarker polynucleotides of the invention are used as molecular probes in hybridization reactions or as molecular primers in nucleic acid extension reactions as described herein.
  • the biomarker polynucleotides may be referred to as biomarker probes and biomarker primers, respectively, and the biomarker polynucleotides present in a sample which are to be detected and/or quantified are referred to as target biomarker
  • biomarker forward primer and biomarker reverse primer depending on their 5 ' to 3 ' orientation relative to the direction of transcription.
  • the invention encompasses methods of detecting genetic change in a biomarker gene (e.g., a mutation or a change in copy number). In another embodiment, the invention encompasses methods of detecting a change in the mcthylation of a biomarker gene.
  • a biomarker probe or a biomarker primer is typically an oligonucleotide which binds through complementary base pairing to a subsequence of a target biomarker polynucleotide.
  • the biomarker probe may be, for example, a DNA fragmcnt prepared by amplification methods such as by PGR or it may be chemically synthesized.
  • a double-stranded fragment may then be obtained, if desired, by annealing the chemically synthesized single strands together under appropriate conditions or by synthesizing the complementary strand using DNA polymerase with an appropriate primer.
  • a specific nucleic acid sequence it is understood that the complementary strand is also identified and included as the complementary strand will work equally well in situations where the target is a double stranded nucleic acid.
  • ⁇ nucleic acid probe is
  • biomarkers may be detected in the test sample or the control sample by gene expression profiling.
  • m.RNA is prepared from a sample and mRNA expression levels are measured by reverse transcription quantitative polymerase chain reaction (RT-PCR followed with qPCR).
  • RT-PCR is used to create a cDNA from the corresponding mRNA.
  • the cDNA may be used in a qPCR assay to produce fluorescence as the DNA amplification process progresses. By comparison to a standard curve, qPCR can produce an absolute measurement such as number of copies of mRNA per cell.
  • Northern blots, microarrays, Invader assays, and RT-PCR combined with capillary electrophoresis may be used to measure expression levels of mRNA in a sample. Further details are provided, for example, in “Gene Expression Profiling: Methods and Protocols,” Richard A. Shimkcts, editor, Humana Press, 2004 and US patent application 2010/0070191.
  • the invention encompasses an array comprising polynucleotides that hybridize to genes listed in Table 1.
  • the array may comprise polynucleotides that hybridize to at least 2, at least 3, at least 4, at least 5, at least 10, at least 15, at least 20, at least 25, at least 30, at least 35, at least 40, at least 45, at least 50, at least 55, at least 60, at least 65, at least 70, at least 75, at least 80, least about 85, at least 87 or all genes listed in Table 1 .
  • the polynucleotides are immobilized on a solid surface. Examples of solid surfaces include paper, filler, nylon or other type of membrane, slide including glass slide, and chip (e.g., silicon, microarray chip).
  • the polynucleotides may be single-stranded nucleic acid molecules (e.g., antisense oligonucleotides or fragments of cDNA).
  • the array is not a human genome-wide array. Examples of human genome-wide array include, but are not limited to, Exon 1 .0 ST, Gene 1.0 ST, U 95, U133 , U133A 2.0, and U 1 33 Plus 2.
  • detection of the biomarkcrs described herein may be accomplished by an immunoassay procedure.
  • the immunoassay typically includes contacting a test sample with an antibody that specifically binds to or otherwise recognizes a biomarkcr, and detecting the presence of the antibody/biomarker complex in the sample.
  • the immunoassay procedure may be selected from a wide variety of immunoassay procedures known to those skilled in the art such as, for example, competitive or non-competitive enzyme-based immunoassays, immunoprccipitation, enzyme-linked immunosorbent assays
  • ELISA radioimmunoassay
  • RIA radioimmunoassay
  • II IC immunohistochemistry
  • Western blots e.g., Western blots.
  • multiplex assays may be used, including antibody panels or arrays, wherein several desired antibodies are placed on a support, such as a glass bead or plate, and reacted or otherwise
  • Antibodies used in these assays may be monoclonal or polyclonal, and may be of any type such as IgG, IgM, IgA, IgD and IgE. Monoclonal antibodies may be used to bind to a specific epitope offered by the biomarkcr molecule, and therefore may provide a more specific and accurate result. Antibodies may be produced by immunizing animals such as rats, mice, rabbits and goats. The antigen used for immunization may be isolated from the samples or synthesized by recombinant protein technology. Methods of producing antibodies and of performing antibody-based assays are well-known to the skilled artisan and are described, for example, more thoroughly in Antibodies: A Laboratory Manual ( 1988) by Harlow & Lane; Immunoassays: A Practical Approach, Oxford
  • the present invention also provides "biomarker antibodies” including polyclonal, monoclonal, or recombinant antibodies, and fragments and variants thereof, that immunospeci fically bind the respective biomarker proteins or polypeptides encoded by the genes or cDNAs (including polypeptides encoded by mRNA splice variants) as listed in Table 1.
  • Various chemical or biochemical derivatives of the antibodies or antibody fragments of the present invention can be produced using known methods.
  • One type of derivative which is diagnostically useful as an immunoconjugate comprising an antibody molecule, or an antigen-binding fragment thereof, to which is conjugated a detectable label.
  • the biomarker antibody is not labeled but in the course of an assay, it becomes indirectly labeled by binding to or being bound by another molecule that is labeled.
  • the invention encompasses molecular complexes comprising a biomarker antibody and a label, as well as immunocomplexes comprising a biomarker polypeptide, a biomarker antibody, and immunocomplexes comprising a biomarker polypeptide, a biomarker antibody, and a label.
  • Examples of detectable substances or detectable labels include various enzymes, prosthetic groups, fluorescent materials, luminescent materials, bioluminesccnt materials, and radioactive materials.
  • suitable enzymes include horseradish peroxidase, alkaline phosphatase, bcta-galactosidase and acetylcholinesterase.
  • suitable prosthetic group complexes include streptavidin/biotin and avidin/biotin.
  • fluorescent materials include umbeliifcroncs, fluoresceins, fluorescein isothiocyanate, rhodamincs, dichlorotriazinylamine fluorescein, dansyl chloride, phycoerythrins, Alexa Fluor 647, Alexa Fluor 680, DilC, 9 (3), Rhodamine Red-X, Alexa Fluor 660, Alexa Fluor 546, Texas Red, YOYO- 1 + DNA, tetramcthyl hodamine, Alexa Fluor 594.
  • BODIPY FL Alexa Fluor 488, Fluorescein, BODIPY TR, BODIPY TMR, carboxy SNARF- 1 , FM 1 -43 , Fura-2, Indo- 1 , Cascade Blue, NBD. 13 API, Alexa Fluor 350, aminomethylcoumarin, Lucifer yellow. Propidium iodide, or dansylamide.
  • An example of a luminescent material is luminol.
  • bioluminesccnt materials include green fluorescent proteins, modified green fluorescent proteins, lucifcrase, iucifcrin, and acquorin. Examples of suitable
  • radioactive material include I, I, S or H.
  • Immunoassays for biomarker polypeptides will typically comprise incubating a sample, such as a biological fluid, a tissue extract, freshly harvested cells, or lysates of cells, in the presence of a detectably labeled antibody capable of identifying biomarker gene products or conserved variants or peptide fragments thereof, and detecting the bound antibody by any of a number of techniques well- known in the art.
  • a sample such as a biological fluid, a tissue extract, freshly harvested cells, or lysates of cells
  • a detectably labeled antibody capable of identifying biomarker gene products or conserved variants or peptide fragments thereof
  • ELISA enzyme-linked immunosorbent assay
  • the biological sample may be brought in contact with and immobilized onto a solid phase support or carrier such as nitrocellulose, or other solid support which is capable of immobilizing cells, cell particles or soluble proteins.
  • a solid phase support or carrier such as nitrocellulose, or other solid support which is capable of immobilizing cells, cell particles or soluble proteins.
  • the support may then be washed with suitable buffers followed by treatment with the detectably labeled biomarker antibody.
  • the solid phase support may then be washed with the buffer a second time to remove unbound antibody.
  • the amount of bound label on solid support may then be detected by conventional means.
  • a well known example of such a technique is Western blotting.
  • the present invention provides compositions comprising labelled biomarker polynucleotides, or labelled biomarker antibodies to the biomarker proteins or polypeptides, or labeled biomarker polynucleotides and labeled biomarker antibodies to the biomarker proteins or polypeptides according to the invention as described herein.
  • Antibodies and other reagents may also be used to detect post- translational modifications (e.g., mcfhylafion, acetylation, farnesylation, biotinylation, stcaroylation, formylation, myristoylation, palmitoylation, geranylgeranylation, pegylation, phosphorylation, sulphation, glycosylation, sugar modification, lipidation, lipid modification, ubiquitination, sumolation, disulphidc bonding, cystcinylation, oxidation, glutathionylation, carboxylation,
  • post- translational modifications e.g., mcfhylafion, acetylation, farnesylation, biotinylation, stcaroylation, formylation, myristoylation, palmitoylation, geranylgeranylation, pegylation, phosphorylation, sulphation, glycosylation, sugar modification, lipidation, lipid
  • biomarker proteins or biomarker
  • the invention encompasses a panel comprising antibodies that bind to proteins encoded by genes listed in Table 1 .
  • the panel may comprise antibodies that bind to proteins encoded by at least 2, at least 3, at least 4, at least 5, at least 10, at least 15, at least 20, at least 25, at least 30, at least 35, at least 40, at least 45, at least 50, at least 55, at least 60, at least 65, at least 70, at least 75, at least 80, least about 85, at least 87 or all genes listed in Table 1 .
  • the panel of antibodies is immobilized on a solid surface. Examples of solid surfaces include microspheres, plates, wells, slides, and beads (e.g., protein A or protein G agarose).
  • the biomarkers described herein may also be detected and quantified by mass spectrometry.
  • Mass spectrometry is a method that employs a mass spectrometer to detect ionized protein markers or ionized peptides as digested from the protein markers by measuring the mass-to- charge ratio (m/z). Labelling of biomarkers (along with other proteins) with stable heavy isotopes (deuterium, carbon-13, nitrogen- 15, and oxygen- 18) can be used in quantitative proteomics. These are either incorporated metabolically in sample cells cultured briefly in vitro, or directly in samples by chemical or enzymatic reactions. Light and heavy labelled biomarker peptide ions segregate and their intensity values are used for quantification.
  • analytes may be introduced into an inlet system of the mass spectrometer and ionized in an ionization source, such as a laser, fast atom bombardment, plasma or other suitable ionization sources known to the art.
  • an ionization source such as a laser, fast atom bombardment, plasma or other suitable ionization sources known to the art.
  • the generated ions are typically collected by an ion optic assembly and introduced into mass analyzers for mass separation before their masses are measured by a detector. The detector then translates information obtained from the detected ions into mass-to-charge ratios.
  • the invention also encompasses methods that involve measuring the activity of a biomarker (e.g., enzymatic activity).
  • a biomarker e.g., enzymatic activity
  • enzymatic activity include, without limitation, kinase, phosphatase, protease, ubiquitination, oxidase and reductase activity.
  • the invention also provides compositions comprising biomarker polynucleotides, biomarker polypeptides, or biomarker antibodies according to the invention as described herein in the various embodiments.
  • the invention further provides diagnostic or detection reagents for use in the methods of the invention, for example, reagents for flow cytometry and/or immunoassays that comprise fluorochrome-labeled antibodies that bind to one of the biomarker polypeptides of the invention.
  • the invention provides diagnostic or detection reagents that comprise one or more biomarker probes, or one or more biomarker primers.
  • ⁇ diagnostic reagent may comprise biomarker probes and/or biomarker primers from the same biomarker gene or from multiple biomarker genes.
  • the invention also provides diagnostic compositions that comprise one or more biomarker probes and target biomarker polynucleotides, or one or more biomarker primers and target polynucleotides, or biomarker primers, biomarker probes and biomarker target polynucleotides.
  • the diagnostic compositions comprise biomarker probes and/or biomarker primers and a sample suspected to comprise biomarker target polynucleotides.
  • Such diagnostic compositions comprise biomarker probes and/or biomarker primers and the nucleic acid molecules (including RNA, mRNA, cRNA, cDNA, and/or genomic DNA) of a subject in need of a diagnosis/prognosis of lung cancer.
  • Biomarkers and gene signatures of the invention may be predicted based on gene expression patterns in lung cancer, including stages 1 and 2
  • biomarker and gene signature prediction comprises gene expression patterns in control (e.g., non-tumor) biological samples.
  • a heterogeneous ensemble learning approach may be used to classify samples based on their gene expression profiles. Such an approach may combine predictions from different approaches that use genes, gene set-derived features and/or causal network-derived features in order to get a classification and a prediction confidence for each classificd sample.
  • Methods that may be used to generate biomarkcrs and gene signatures of the invention include shrunken centroids, factor rotation, logistic regression models, network-based approaches, disease module-based approaches, linkage methods, modularity or pathway-based methods and diffusion-based methods.
  • the biological data (such as training data and test data) used in these methods may be drawn from the literature, databases (including data from preclinical, clinical and post-clinical trials of pharmaceutical products or medical devices), genome databases (genomic sequences and expression data, e.g.. Gene Expression Omnibus by National Center for Biotechnology Information or
  • ArrayExpress by European Bioinformatics Institute (Parkinson et al. 2010, Nucl. Acids Res., doi: 10.1093/nar/gkql 040. Pubmed ID 21071405)), commercially available databases (e.g., Gene Logic, Gaithersburg, MD, USA) or experimental work.
  • the REACTOME, KEGG or BIOCARTA pathway gene set collections from the Broad Institute (Cambridge, MA) may be used.
  • the data may be related to nucleic acid ⁇ e.g. , absolute or relative quantities of specific DNA or RNA species, changes in DNA sequence, RNA sequence, changes in tertiary structure, or methyl ati on pattern as determined by sequencing,
  • hybridization particularly to nucleic acids on microarray, quantitative polymerase chain reaction, or other techniques known in the art
  • protein/peptide e.g. , absolute or relative quantities of protein, specific fragments of a protein, peptides, changes in secondary or tertiary structure, or posttranslational modifications as determined by methods known in the art
  • functional activities e.g., enzymatic activities, proteolytic activities, transcriptional regulatory activities, transport activities, binding affinities to certain binding partners
  • Modifications including posttranslational modifications of protein or peptide, can include, but are not limited to, methylation, acetylation, farncsylation, biotinylation, stearoylation, formylation, myristoylation, palmitoylation, geranylgeranylation, pegylation, phosphorylation, sulphation, glycosylation, sugar modification, lipidation, lipid modification, ubiquitination, sumolation, di sulphide bonding, cysteinylation, oxidation, glutathionylation, carboxylation,
  • a protein can be modified posttranslationally by a dispenser of a protein and a protein.
  • a dispenser such as Amadori reactions, Schiff base reactions, and Maillard reactions resulting in glycated protein products.
  • test data sets may be processed and have their quality controlled separately, together if they are obtained from the same technology platform (e.g., an Affymctrix platform).
  • raw data files may be read by the same technology platform (e.g., an Affymctrix platform).
  • the quality may be controlled by:
  • RNA degradation plots using the AffyRNAdeg function of the affy package (Gautier, 2004)
  • NUSE and RLE plots using the function affyPLM (Brettschneider et al., Technomelrics, 50(3):241 -264 (2008)), calculating the MA(RLE) values;
  • Arrays passing quality control checks may be normalized using the gcrma algorithm (Wu et al., Journal of the American Statistical Association, 99:909 (2004)). If the datascts were obtained from a database, the samples classification may be obtained from the series matrix file of the same database for each datasct. The output of this part of the method may consist of: a gene expression matrix on training samples and test samples, probesets, and the class information for the training samples.
  • Non-limiting examples of methods that may be used to generate predictions are: transformation invariant (Tranlnv) (U.S. Provisional Patent Application entitled “Systems and Methods for Generating Biomarker Signatures with Integrated Bias Correction and Class Prediction,” filed concurrently with the instant application and having the attorney docket no.
  • Generalized simulated annealing may be modified for binary functions.
  • a dual binary generalized simulated annealing based method may be used (DualGcnsemble) (U.S. Provisional Patent Application entitled “Systems and Methods for Generating Biomarker Signatures with Integrated Dual Ensemble and Simulated Annealing Techniques," filed concurrently with the instant application, incorporated herein by reference in its entirety and having the attorney docket no. 106500-003 1 -001 ).
  • ' ⁇ -filter is a method of filtering genes based on the t-test by setting P-valuc and fold-change thresholds. CORG may be modified by calculating activity scores by leveraging the F-test instead of the T- test.
  • CORG may also be combined with SVM.
  • Dual bagging is a combination of bagging (Breiman, The Elements of Statistical Learning: Data Mining, Inference, and Prediction. 2nd cd., ed. T. Hastie, R. Tibshirani, and J. Friedman, (2009)) and the random subspace method (Bryll, Pattern Recognition, 20(6): 1291 - 1302 (2003); Ho, IEEE Transactions on Pattern Analysis and Machine Intelligence, 20(8):832- 844 (1998); Skurichina, Pattern Analysis and Applications, 5(2): 121 - 135 (2002)).
  • the single and pairs method may include the following steps:
  • the forward learning method may include the following steps:
  • N for example, 20, 100 or 200.
  • the Laplacian based learning method may include the following steps:
  • KPC main kerne! principal components
  • Network-based analysis can be combined with machine learning methods to generate predictions, for example, combining any one of CORG, dual bagging or T-filter with a network-based analysis.
  • methods used to generate predictions are further combined with another classification method (e.g., a method that is used for cross- validation).
  • classification methods include PAMR (Tibshirani et ah, Proc Natl Acad Set USA, 99(10):6567-6572 (2002)),
  • LDA Discrimination Analysis
  • ELDA Eiigengene-based Linear Discrimination Analysis
  • PCA Principal Components Analysis
  • RPART Recursive Partitioning Tree
  • SVM Support Vector Machine
  • KNN K-Nearest Neighbors
  • PLS.DA Partial Least Squares Discriminant Analysis
  • NPA network-based analysis that uses NPA may be combined with SVM (U.S. Provisional Patent Application entitled “Systems and Methods relating to Network-based Biomarkcr Signatures,” filed concurrently with the instant application, incorporated herein by reference in its entirety and having the attorney docket no. 106500-0022-001 ).
  • these methods may further include a step of ovcrsampling to balance classes.
  • the methods may include a step of filtering genes based on a simple T-test between the categories to be classified. The filtering step may reduce the number of genes to less than 1 ,500 or less than 2,000.
  • the genes obtained by these methods may be mapped to gene symbols using any suitable platform, for example, the Confero platform (Hermida et al., Confero: an Integrated Contrast and Gene Set Platform for Computational Analysis and Biological Interpretation ofOmics Data, submitted, 2012).
  • the Confero platform Hermida et al., Confero: an Integrated Contrast and Gene Set Platform for Computational Analysis and Biological Interpretation ofOmics Data, submitted, 2012.
  • the numerical methods for generating the gene signatures of the invention may include a testing step and confidence statistics for the genes.
  • the testing step (or phase) is an exemplary use of the gene signature in carrying out the claimed method.
  • the invention encompasses a method for classifying a test sample as stage 1 lung adenocarcinoma, stage 2 lung adenocarcinoma, stage 1 squamous cell carcinoma or stage 2 squamous cell carcinoma, the method comprising: measuring the expression levels of at least 2 genes listed in Table 1 in a test sample; and applying one or more network-based methods, one or more machine-learning based methods, or a combination of the foregoing methods to the expression levels to obtain a classification of the test sample as stage 1 lung adenocarcinoma, stage 2 lung adenocarcinoma, stage 1 squamous cell carcinoma or stage 2 squamous cell carcinoma.
  • the expression levels of at least 2, at least 3, at least 4, at least 5, at least 10, at least 1 5, at least 20, at least 25, at least 30, at least 35, at least 40, at least 45, at least 50, at least 55, at least 60, at least 65, at least 70, at least 75, at least 80, least about 85, at least 87 or all genes listed in Table 1 are measured.
  • the classifier has been trained by in silico analysis or one or more feature selection and classification algorithms.
  • One aspect of the invention encompasses a list of one or more biomarkcrs or gene signatures of the invention stored on a computer readable medium.
  • the absence, presence, activity or expression level of a biomarkcr in a biological sample may also be stored on the computer readable medium.
  • the computer readable medium may also include information that identifies the sample.
  • the computer readable medium may also include a computer program product.
  • the computer program product may include a classifier based on at least two genes listed in Table 1 .
  • the classifier may be based on at least 2, at least 3, at least 4, at least 5, at least 10, at least 15, at least 20, at least 25, at least 30, at least 35, at least 40, at least 45, at least 50, at least 55, at least 60, at least 65, at least 70, at least 75, at least 80, least about 85, at least 87 or all genes listed in Table 1 .
  • the classifier is trained by in silica analysis or one or more feature selection and classification algorithms.
  • the classifier is trained by one or more algorithms selected from the group consisting of dual ensemble, generalized simulated annealing, T- filter, CORG, CORG combined with support vector machine, dual bagging, single and pairs, forward learning, Laplacian based learning and learning method based on network perturbation amplitude.
  • the classi bomb may be trained with at least the data in Gene Expression Omnibus datasets GSE2109, GS El 0245, GSE18842 and GSE37745.
  • One aspect of the invention encompasses devices useful for performing methods o f the invention.
  • the devices may be used for diagnosing, classifying and/or grading lung cancer.
  • the devices can comprise means for detecting the expression level of at least 2 of the genes listed in Table 1 or the level of at least 2 gene products of such genes in a test sample.
  • Such means may include components for performing one or more methods of nucleic acid extraction, nucleic acid amplification, nucleic acid detection, protein isolation and/or protein detection.
  • Such components may include one or more of an ampli fication chamber (for example a thermal cycler), a plate reader, robotic sample handling
  • the devices can comprise means for detecting at least 2, at icast 3, at least 4, at least 5, at least 10, at least 15, at least 20, at least 25, at least 30, at least 35, at least 40, at least 45, at least 50, at least 55, at least 60, at least 65, at least 70, at least 75, at least 80, at Icast 84 or all genes listed in Table 1.
  • the devices can comprise means for detecting the gene products of at least 2, at least 3, at least 4, at least 5, at least 1 0, at least 15, at least 20, at least 25, at least 30, at least 35, at least 40, at least 45, at least 50, at least 55, at least 60, at least 65, at least 70, at least 75, at least 80, at least 84 or all genes listed in Table 1.
  • the devices optionally comprise a means for identifying a given test sample, and of linking the results obtained to that sample.
  • Such means can include manual labels, barcodes, and other indicators which can be linked to a sample container or receptacle.
  • Identification means may optionally be included in the sample itself, for example where an encoded particle is added to the sample.
  • the results may be linked to the sample, for example in a computer memory that contains a sample designation and a record of expression levels obtained from the sample. Linkage of the results to the sample can also include a linkage to a particular sample container or receptacle in the device, which is also linked to the sample identity.
  • the devices may comprise an excitation and/or a detection means.
  • Any instrument that provides a wavelength that can activate a label (e.g., fluorophore, fluorochrome and fluorescent dye) used on a detection reagent and is shorter than the emission wavelcngth(s) to be detected can be used for excitation.
  • excitation sources include a broadband ultraviolet light source such as a deuterium lamp with an appropriate filter, the output of a white light source such as a xenon lamp or a deuterium lamp after passing through a monochromator to extract out the desired wavelength(s), a continuous wave (cw) gas laser, a solid state diode laser, or any pulsed lasers.
  • Emitted light can be detected through any suitable component or technique; many suitable approaches are known in the art.
  • a fluorimcter or spectrophotometer may be used to detect whether the test sample emits light of a wavelength characteristic of a label used in a method of the invention.
  • the devices may comprise a means for correlating the expression levels of the genes being analyzed with an lung cancer status, prognosis, grade and/or classification.
  • Such means may comprise one or more of a variety of correlative techniques, including lookup tables, algorithms, multivariate models, and linear or nonlinear combinations of expression models or algorithms, such as any of the in silico and machine learning methods described above.
  • the expression levels may ⁇ be converted to one or more biomarker scores, indicating that the individual providing the sample is not suffering from adenocarcinoma or squamous cell carcinoma or is suffering from stage 1 adenocarcinoma, stage 2 adenocarcinoma, stage 1 squamous cell carcinoma or stage 2 squamous cell carcinoma.
  • the models and/or algorithms can be provided in computer readable format.
  • the devices may also comprise output means for outputting the lung cancer status, prognosis, grade and/or classification.
  • output means can take any form which transmits the results to an individual and/or a healthcare provider, and may include a monitor, a display, and/or a printer.
  • Output means may record the results to a computer readable medium.
  • the device may use a computer system for performing one or more of the steps provided.
  • a device of the invention comprises means for detecting the expression level of at least 2 genes listed in Table 1 in a test sample; means for correlating the expression level with a classification of the lung cancer as stage 1 adenocarcinoma, stage 2 adenocarcinoma, stage 1 squamous cell carcinoma or stage 2 squamous cell carcinoma; and means for outputting the lung cancer classification.
  • the device comprises means for detecting the expression level of at least 2, at least 3, at least 4, at least 5, at least 1 0, at least 15, at least 20, at least 25, at least 30, at least 35, at least 40. at least 45, at least 50, at least 55, at least 60, at least 65, at least 70, at least 75, at least 80, at least 84 genes or ail genes listed in Table 1 .
  • kits for practicing the methods of the invention may be used for classifying and grading lung cancer or for assessing the prognosis of lung cancer in an individual.
  • the kits can be used for clinical diagnosis and/or laboratory research.
  • a kit comprises in one or more containers one or more reagents that detect expression levels of genes that serve as biomarkcrs of lung cancer in a test sample.
  • the kit also comprises instructions in any tangible medium (e.g., written, tape, CD- ROM, DVD) on the use of the detection reagent(s) in one or more methods of the invention.
  • a detection reagent in the kit may comprise at least one polynucleotide, probe, and/or primer specific for the stage 1 adenocarcinoma, stage 2 adenocarcinoma, stage 1 squamous cell carcinoma and/or stage 2 squamous cell carcinoma the genes listed in ' fable 1 .
  • the nucleic-acid based detection reagents may comprise sequences complementary to a portion of the signature genes or sequences that are portions of the signature genes.
  • Such a kit may optionally provide in separate containers enzymes and/or buffers for reverse transcription, in vitro transcription, and/or 1 ) ⁇ ⁇ polymerization, nucleotides, and/or labeled nucleotides.
  • a detection reagent in the kit may comprise a biomarkcr antibody, which may be labeled or labelable.
  • the antibodies may bind to proteins encoded by stage 1 adenocarcinoma, stage 2 adenocarcinoma, stage 1 squamous cell carcinoma and/or stage 2 squamous cell carcinoma the genes listed in Table 1 .
  • the detection reagents recognize a post-translational modification (e.g., mcthylation, accty!ation, farncsylation, biotinylation, stcaroylation, formylation, myristoylation,
  • the kit may include a substrate for the biomarkcr and a detection reagent that recognizes the products and/or byproducts of the activity being measured.
  • kit may optionally provide, in separate containers, buffers, secondary antibodies, signal generating accessory molecules, and/or labeled secondary antibodies, including fiuorochrome-labcled secondary antibodies.
  • the kit may also include unlabeled or labeled antibodies to various cell surface antigens which can used for identification or sorting of subpopulations of cells.
  • the detection reagents may be labeled or labelable. by one or more detectable labels.
  • detectable labels include, without limitation, radiolabcls (e.g. radioactive nuclides), dyes, fluorescent proteins or materials (e.g., fluorochromes, fiuorophores, fluorescein and rhodamine), luminescent proteins or materials, bioluminescent proteins or materials (e.g., luciferase, aequorin and luciferin), enzymes (e.g., beta-galactosidase, alkaline phosphatase, horseradish peroxidase and acetylcholinesterase) and prosthetic groups (e.g., biotin, streptavidin and avidin).
  • radiolabcls e.g. radioactive nuclides
  • dyes e.g., fluorescent proteins or materials (e.g., fluorochromes, fiuorophores, fluorescein and rhodamine),
  • the detection reagents in the kit may be immobilized on a solid surface or packaged separately with reagents to immobilize them on a solid surface.
  • kits may be positive and negative controls for the methods of the invention.
  • the positive and/or negative controls included in a kit can be nucleic acids, polypeptides, cell lysate, ceil extract, whole ceils from patients, or whole cells from cell lines.
  • RNA degradation plots with the AffyRNAdeg function of the affy package
  • NUSE and RLE plots with the function affyPLM
  • Training set sample classifications are obtained from the series matrix file of the GEO database for each datasct.
  • GSM926787_PA 1 17_1 30A_10 l 126_I-IG-U 133_Plus_2_.CEL,
  • GSM926801 PA 1 1 7 173 B_ 101 130J-IG-U 1 33_Plus_2_. CEL,
  • GSM926826 PA1 17 285A 101210 HG-U 133 Plus 2 .CEL, GSM926840 J 5 A 1 17_336 ⁇ _101202 J IG-U 133_Plus_2_.CEL,
  • GSM926843_PA1 17_353 A J 01208 JHG-U 133 _Plus_2_.CEL,
  • GSM926854 PA 1 1 7_58A_1 01 123_HG-U 133_Plus_2_.CEL,
  • GSM102451 .CEL GSM 102455.CEL, GSM138003.CEL, GSM203732.CEL, GSM23 1874.CEL, GSM467024.CEL, GSM76585.CEL, GSM76595.CEL, GSM88997.CEL, GSM926766_PA025_human_365A_070907.CEL, GSM926739_PA025_human_256A_070926.CEL,
  • GSM138002.CEL GSM 137945. GEL, GSM13793 1.CEL, GSM l 37916.CEL, GSM 137910.CEL, GSM l 17763. CEL, GSMl 17632.CEL, GSM 3 17610.CEL, GSM 102555.CEL, GSM 102553. CEL, GSMl 02548. CEL., GS l 025 12, CE L, GSM l 02507. CEL, GSM 102447.CEL, GSMl 17763. CEL, GSM 1 52757.CEL, GSM258570.CEL, GSM258591.CEL. GSM258594.CEL. GSM258597.CEL,
  • GSM926745_PA025_human_284A_070912.CEL GSM926750_ _PA025_ human_300A_070919.CEL
  • GSM926786_ _ ⁇ 1 17_ 128 A J 01 126J IG-U 1 33_Plus 2_ .CEL,
  • GSM926809 PA1 17 190B J 01202JIG-U133 _Plus 2___ .CEL,
  • GSM102512.CEL GSM138002.CEL
  • GSM467032.CEL GSM467032.CEL
  • GSM926851 PA 1 17_52A_101 123 I IG-U 133_Plus_2_.CEL,
  • GSM102548.CEL GSM 102553.CEL, GSM 1 17610.CEL, GSM38103.CEL, GSM46868.CEL, GSM46936.CEL, GSM46941.CEL, GSM76587.CEL, GSM76590.CEL, GSM88962.CEL were not used for further analysis.
  • the output at this point comprises a gene expression matrix X on 410 samples (260 training samples and 150 test samples) and 54675 probescts, and the class information for the training samples.
  • a set of feature selection and classification algorithms arc used to obtain a number of classifications for each test sample.
  • Each method has defined input and output:
  • INPUT gene expression matrix X nxp on n samples and p genes, training samples and test samples, and the class information for the training samples
  • OUTPUT Class prediction for each test sample and a list of genes involved.
  • oversampling is, optionally, used to balance classes in the training dataset; (2) mapping probe sets to gene symbols (Entrez gene ids) using Confero platform (Hcrmida, 2012); and (3) optionally filtering the genes in the matrix based on simple T-test between the categories to be classified so that less than 1500 genes (for Dual Ensemble or T-fi!ter methods) or less than 2000 genes (for the other methods) remain.
  • SVM Support Vector Machine
  • KNN K-Nearest Neighbors
  • This dual ensemble method builds ensemble of multiple classification algorithms applied in randomly perturbed data.
  • the diversity of the ensemble classifier is imposed by using different classification algorithms and is further enhanced by data-level perturbation. See, e.g., Yang, 2010.
  • a molecular profile of a training dataset, TO.train and its associated phenotypc cl. train (control and treatment) are used as input.
  • the molecular profile of the test set TO. test is used to predict the phenotypc ci.test.
  • Genes are filtered based on t-tcst to obtain a list of N genes, by setting P- value and fold-change thresholds. Thresholds are decreased (resp. increase) automatically if the list size is over N. Any M is trained on the resulting subspace. N is chosen by cross-validation.
  • CORG-modified method This method is modified from CORG method (Chuang, 2007) as activity- scores arc calculated by leveraging F-test instead of T-test. It uses the c2.cp gene sets collection from the Broad Institute (Cambridge, ⁇ ) (Reactome, KEGG and Biocarta pathways).
  • a threshold (a 0 ) is selected.
  • the obtained list is used to train M on the reduced feature space.
  • N typically 20, 100, 200.
  • NIPBL 25836 Nipped-B homolog (Drosophila) 4/5 cAMP responsive element binding protein
  • solute carrier family 37 (glycerol-3-
  • IL1R2 7850 interleukin 1 receptor, type 11 3/5 immediate early response 3 interacting
  • MLL5 55904 3/5 leukemia 5 (trithorax homolog, Drosophila)
  • FAP 21 91 fibroblast activation protein, alpha 3/5
  • ZNF521 25925 zinc finger protein 521 3/5

Abstract

The present invention relates to biomarkers and gene signatures that are useful for diagnosing, classifying and prognosing lung cancer. The invention also relates to diagnostic methods and kits using these biomarkers and gene signatures.

Description

Gene signatures for classifying and grading lung cancer
Field of the invention
[00011 The present invention relates to gene signatures that are indicative for the class and grade of lung cancer. The present invention also relates to methods of diagnosing classifying and grading a lung cancer tumor. The invention further relates to arrays and computer readable media comprising such gene signatures.
Background of the Invention
[0002] Lung cancer is the most common cancer in the world, both in rates of incidence and in rates of mortality, and is most prevalent in Europe and North America. Lung cancer is often attributed to both genetic factors and exposure to environmental factors (e.g., of radon gas, asbestos, and air pollution).
Accordingly, researchers have attempted to identify biomarkers for use in risk assessment, screening, diagnosis, prognosis, selection of therapy and monitoring therapy. See, e.g. , Yip et al. (WO 2004/061410), Semmcns, ct al. (WO
2005/098445), Sungwhan et al. (US 2007/0264659), Gold et al. (WO
201 1/03 1344), and Birse (US 7,892,760).
[0003] Lung cancers are classified by histological type: non-small cell lung carcinoma (NSCLC), small cell lung carcinoma (SCLC), and carcinoid.
Approximately 80% of all lung cancers are NCSLC, with SCLC and carcinoid accounting for approximately 17% and 1 %, respectively. Approximately 2% of lung cancers are unspecified.
[0004] Non-small cell lung carcinomas are grouped together based on similar prognosis and management and comprise three main sub-types: squamous cell lung carcinoma (SCC), adenocarcinoma and large cell lung carcinoma (LCLC). SCC account for approximately 25% of lung cancer incidents. Adenocarcinoma of the lung accounts for approximately 40% of lung cancer incidents. LCLC is a heterogeneous group of undifferentiated malignant neoplasms originating from transformed epithelial cells in the lung, which accounts for about 10% of lung cancer incidents.
[00051 NSCLC are staged according to the TNM system. The T category defines the primary tumor by size and whether it has spread into the surrounding tissue. The N category identifies any lymph node involvement in and around the lungs. Finally, the M category indicates whether the cancer has metastasized. Under stage 1 , the lung cancer is small and localized to a single area. Stage 2 and stage 3 lung cancers are larger, may have grown into surrounding tissues and may involve lymph nodes in and around the lungs. Stage 4 lung cancers have metastasized to another region of the body.
[0006] Lung cancer is currently diagnosed by X-ray or computed tomography (CT) screening. X-ray analysis is typically performed if a patient reports symptoms that may suggest lung cancer and may reveal an obvious mass, widening of the mediastinum (suggestive of spread to lymph nodes there), atelectasis (collapse or closure of alveoli), consolidation (pneumonia), or pleural effusion. CT imaging is subsequently performed to provide additional information about the type and extent of disease. After a suspicious mass is identified by X-ray and/or CT imaging, bronchoscopy or biopsy may be performed to analyze the suspicious tissue. Accordingly, the diagnostic tests for lung cancer usually require the disease to have progressed to the point that lung function is moderately affected and tumor growth is visible. Thus, there is a need for a diagnostic test that can identify, classify and grade lung tumors in patients at early stages.
Summary of the invention
1 007 ] The present invention is directed to gene signatures for classifying, diagnosing or grading lung cancer in an individual. A first aspect of the invention provides a method of classifying or grading a lung cancer tumor in an individual at risk for or having lung cancer. In some embodiments the method comprises classifying a test sample as stage 1 lung adenocarcinoma, stage 2 lung
adenocarcinoma, stage 1 squamous cell carcinoma or stage 2 squamous ceil carcinoma. In some embodiments, the method comprises measuring the expression levels of at least 2 genes listed in Table 1 in a test sample; and applying one or more network-based methods, one or more machine-learning based methods, or a combination of the foregoing methods to the expression levels to obtain a classification of the test sample as stage 1 lung adenocarcinoma, stage 2 lung adenocarcinoma, stage 1 squamous cell carcinoma or stage 2 squamous cell carcinoma. In some embodiments, a differential pattern of expression levels of said at least 2 genes in the test sample classifies the lung cancer tumor as one of stage 1 lung adenocarcinoma, stage 2 adenocarcinoma, stage 1 squamous cell carcinoma or stage 2 squamous cell carcinoma.
[ 0008 j In some embodiments, the differential pattern of expression levels is identified by a classifier based on a plurality of genes listed in Table 1 , including said at least two genes, said classifier having been trained by in silico analysis or one or more feature selection and classification algorithms. Optionally, the di fferential pattern of expression levels is identi fied by a classifier based on a plurality of genes listed m i able 1 , including said at least two genes, said classifier having been trained by one or more algorithms selected from the group consisting of dual ensemble, generalized simulated annealing, T- filter, CORG, CO G combined with support vector machine, dual bagging, single and pairs, forward learning, Laplacian based learning and learning method based on network perturbation amplitude. For example, the classifier may be trained with at least the data in the Gene Expression Omnibus datasets GSE21 09, GSE 10245, GSE 1 8842 riQP 1774 ;
10009] In some embodiments, the method comprises detecting the expression level of at least 2 of the genes listed in Table 1 in a test sample obtained from the individual; and comparing the expression level of the genes listed in Table 1 in the test sample to the expression level of the genes listed in Table 1 in a control sample. In some embodiments, the method further comprises detecting the expression level of the genes listed in Table 1 in the control sample.
[0010] In some embodiments, the at least 2 genes are selected from the group consisting of: ZIC2, LOC I 0013 1262, CD83, EML1 , PAIP 1 , NIPBL, CREB3L 1 , SLC37A 1 , and SFMBT2.
[0011 ] In some embodiments, the test sample or the control sample is selected from blood, scrum, plasma, sputum, saliva, tissue, bronchia brushings, exhaled breath, and urine. Optionally, the tissue is lung tissue, such as tissue obtained by biopsy from a tumor. In some embodiments, the control sample is lung tissue, such as tissue obtained by biopsy from healthy lung tissue. In some embodiments, the healthy lung tissue is obtained from the individual at risk for or having lung cancer. In other embodiments, the control sample is obtained from an individual that does not have lung cancer.
[0012] In some embodiments, the expression level of the genes listed in Table 1 in the test sample and the expression level of the genes listed in Table 1 in the control sample arc detected by measuring mRNA levels. Optionally, the expression level of the genes listed in Table 1 the test sample are detected by using a human genome-wide array, a human lung tissue array or a custom array comprising polynucleotides of a plurality of genes in Table 1 and said at least 2 genes.
[0013] In some embodiments, the expression level of the genes listed in Table 1 in the test sample and the expression level of the genes listed in Table 1 in the control sample are detected by measuring the level of proteins encoded by the genes.
[0014] In some embodiments, the expression level of the genes listed in 'fable 1 in the test sample and the expression level of the genes listed in 'fable 1 in the control sample are detected by measuring both mRNA levels and the level of
|0() 15| In some embodiments, the expression level of the genes listed in Table 1 in the test sample and the expression level of the genes listed in 'fable 1 in the control sample are compared by in silico analysis (e.g., network-based analysis or machine-learning methods).
[0016] A second aspect of the invention provides an array for use in classi fying or grading a lung cancer tumor. In some embodiments, the array comprises polynucleotides immobilized on a solid surface that can hybridize to at least 10 lung cancer signature genes, wherein the lung cancer signature genes are selected from the group consisting of the genes listed in Table 1. Optionally, the array comprises polynucleotides hybridizing to at least 2 lung cancer signature genes immobilized on a solid surface, wherein the lung cancer signature genes arc selected from the genes listed in Table 1. In some embodiments, the array is not a human genome- wide array.
[0017] A third aspect of the invention provides a panel for use in classifying or grading a lung cancer tumor. In some embodiments, the panel comprises antibodies immobilized on a solid surface that bind to proteins encoded by at least 2 lung cancer signature genes, wherein the lung cancer signature genes are selected from the group consisting of the genes listed in Table 1 .
[0018 j Λ fourth aspect of the invention provides a computer readable medium for use in classifying or grading a lung cancer tumor. In some embodiments, the computer readable medium comprises a lung cancer gene signature, wherein the gene signature comprises at least 2 genes selected from the genes listed in Table 1 .
[0019] In some embodiments, the computer readable medium or computer program product comprises a classifier based on at least two genes listed in Table 1 , said classifier having been trained by in silico analysis or one or more feature selection and classification algorithms. Optionally, the classifier is trained by one or more algorithms selected from the group consisting of dual ensemble, generalized simulated annealing, T-filter, CORG, CORG combined with support vector machine, dual bagging, single and pairs, forward learning, Laplacian based learning and learning method based on network perturbation amplitude. The classifier may be trained with at least the data in the Gene Expression Omnibus datasets GSE2109, GSE10245, GSE 18842 and GSE37745. In some embodiments, the at least two genes are selected from the group consisting of ZIC2, lAJU i uu i J l zoz, iiiviL i , vir l , NirBL, L, CJL>J L I , 5 LU ,' A L , anu
SFMBT2.
[0020] A fifth aspect of the invention provides a device for classifying and grading a lung cancer tumor. In some embodiments, the device comprises means for detecting the expression level of the genes listed in Table 1 in a test sample; means for correlating the expression level with a grade or classification of the tumor as stage 1 or stage 2 lung adenocarcinoma or stage 1 or stage 2 squamous cell carcinoma; and means for outputting the lung cancer tumor grade or classification. Optionally, the device further comprises means for detecting the expression level of the genes listed in Table 1 in a control sample. [0021 ] A sixth aspect of the invention provides a kit for classifying and grading a lung cancer tumor. In some embodiments, the kit comprises a set of reagents that detects expression levels of the genes listed in Table 1 in a test sample and instructions for using said kit for classifying and grading the lung cancer tumor. In other embodiments, the kit is for assessing the prognosis of a lung cancer tumor in an individual. In such embodiments, the kit comprises a set of reagents that detects expression levels of the genes listed in Table 1 in a test sample from the individual and instructions for using said kit for determining the prognosis of the lung cancer tumor in said individual. In some embodiments, the set of reagents that detects expression levels of the genes listed in Table 1 in the test sample may also be used to detect expression levels of the genes listed in Table 1 in a control sample.
[0022] An seventh aspect of the invention provides a method of diagnosing the stage of a lung cancer tumor in an individual or of assessing the prognosis of an individual with a lung cancer tumor. In some embodiments, the method comprises a) measuring the expression level of at least 10 genes/biomarkers selected from the group consisting of the genes listed in Table 1 in a biological sample obtained from the individual; b) calculating a numerical biomarker score for the individual based on the expression levels of the biomarkers measured in step a); wherein the numerical biomarker score is predictive of the stage of lung cancer in the individual. In some embodiments, the method comprises measuring the expression level of at least 10 genes/biomarkers selected from the group consisting of the genes listed in Table 1 in a biological sample obtained from the individual;
calculating a numerical biomarker score for the individual based on the expression levels of the biomarkers measured in step a): wherein the numerical biomarker score is predictive of the prognosis of the lung cancer in the individual.
[0023] In some embodiments, the biological sample is selected from blood, scrum, plasma, sputum, saliva, tissue, bronchia brushings, exhaled breath, and urine. Optionally, the tissue is lung tissue, such as tissue obtained by biopsy from a tumor.
[0024] In some embodiments, the expression level of the genes listed in Table 1 in the biological sample is detected by measuring mRNA levels. Optionally, the expression level of the genes listed in Table 1 the test sample are detected by using a human genomc-wide array, a human lung tissue array or a custom array comprising polynucleotides of a plurality of genes in Table 1 and said at least 2 genes.
[0025] In some embodiments, the expression level of the genes listed in Table 1 in the biological sample is detected by measuring the level of proteins encoded by the genes.
[0026] In some embodiments, the numerical biomarker score is calculated by in silico analysis. The in silica analysis may be network based analysis or a machine-learning method.
[0027] In some embodiments, the biomarkcrs are proteins encoded by the genes selected from the group consisting of the genes listed in Table 1.
[0028] Particular embodiments of the invention arc set forth in the following numbered paragraphs:
1 . A method of classi fying or grading a lung cancer tumor in an individual at risk for or having lung cancer comprising
( 1 ) detecting the expression level of at least 2 genes listed in Table 1 in a test sample obtained from the individual; and
(2) comparing the expression level of said at least 2 genes in the test sample to the expression level of said at least 2 genes in a control sample,
wherein,
i f the expression level of said at least 2 genes is different in the test sample than in the control sample, then the lung cancer tumor can be classified and staged as stage 1 lung adenocarcinoma, stage 2 adenocarcinoma, stage 1 squamous cell carcinoma or stage 2 squamous cell carcinoma.
2. The method according to paragraph 1 , wherein the method further comprises detecting the expression level of said at least 2 genes in the control sample.
3. The method according to paragraph 1 or 2, wherein the test sample is selected from blood, serum, plasma, sputum, saliva, tissue, bronchia brushintis, exhaled breath, and urine.
4. The method according to paragraph 3, wherein the tissue is lung tissue. 5. The method according to paragraph 4, wherein the lung tissue is obtained by biopsy from a tumor.
6. The method according to any one of paragraphs 2-5, wherein the control sample is selected from blood, serum, plasma, sputum, saliva, tissue, bronchia brushings, exhaled breath, and urine.
7. The method according to paragraph 6, wherein the tissue is lung tissue.
8. The method according to paragraph 7, wherein the lung tissue is obtained by biopsy from healthy lung tissue.
9. The method according to paragraph 8, wherein the healthy lung tissue is obtained from the individual at risk for or having lung cancer.
10. The method according to paragraph 6, wherein the control sample is obtained from an individual that does not have lung cancer.
1 1 . The method according to any one of paragraphs 1 - 10, wherein the expression level of said at least 2 genes in the test sample and the expression level of said at least 2 genes in the control sample are detected by measuring mRNA levels.
12. The method according to paragraph 1 1 , wherein the mRNA is measured by amplification, hybridization, mass spectroscopy, serial analysis of gene expression, or massive parallel signature sequencing.
13. The method according to paragraph 12, wherein the amplification is reverse transcription PGR, real time quantitative PGR, differential display or TaqMan PGR.
14. The method according to paragraph 12, wherein the hybridization is a dot blot, a slot blot, an RNase protection assay, microarray hybridization, or in SHU hybridization.
1 5. The method according to paragraph 12, wherein the mass spectroscopy is MALDI-TOF mass spectroscopy.
16. The method according to any one of paragraphs 1 - 10, wherein the expression level of said at least 2 genes in the test sample and the expression level of said at least 2 genes in the control sample are detected by measuring the level of proteins encoded by the genes. 17. The method according to paragraph 16, wherein the protein level is measured using an antibody assay or by mass spectroscopy.
18. The method according to paragraph 17, wherein the antibody assay is selected from Western analysis, immunofluorescence, ELIS/v, and
immunohistochemistry.
19. The method according to any one of paragraphs 1 - 18, wherein the expression level of said at least 2 genes in the test sample and, optionally, the expression level of said at least 2 genes in the control sample are compared by in silica analysis.
20. The method according to paragraph 19, wherein the in silica analysis comprises using a classifier generated by one or more network-based methods or machine-learning based methods.
21 . An array comprising polynucleotides hybridizing to at least 2 lung cancer signature genes immobilized on a solid surface, wherein the lung cancer signature genes are selected from the group consisting of at least 2 genes listed in Table 1 .
22. Λ panel comprising antibodies immobilized on a solid surface that bind to proteins encoded by at least 2 lung cancer signature genes, wherein the lung cancer signature genes arc selected from the group consisting of the genes listed in Table 1 .
23. A computer readable medium comprising a lung cancer gene signature, wherein the gene signature comprises at least 2 genes selected from the genes listed in Table 1 .
35. A device for classifying and grading a lung cancer tumor, the device comprising: means for detecting the expression level of at least 2 genes listed in Table 1 in a test sample; means for correlating the expression level with a classification of the tumor as stage 1 or stage 2 lung adenocarcinoma or stage 1 or stage 2 squamous cell carcinoma; and means for outputting the lung cancer tumor classification.
36. A kit for classifying and grading a lung cancer tumor, comprising one or more reagents that detects expression levels of at least 2 genes listed in Table 1 in a test sample and instructions for using said kit for classifying and grading a lung cancer tumor. 37. A kit for assessing the prognosis of lung cancer in an individual, comprising a set of reagents that detects expression levels of at least 2 genes listed in Table 1 in a test sample from the individual and instructions for using said kit for determining the prognosis of lung cancer in said individual.
38. A method of diagnosing the stage of lung cancer in an individual, said method comprising the steps of:
a) measuring the expression level of at least 2 genes/biomarkcrs selected from the group consisting of the genes listed in Table 1 in a biological sample obtained from the individual;
b) calculating a numerical biomarkcr score for the individual based on the expression levels of the biomarkcrs measured in step a);
wherein the numerical biomarkcr score is predictive of the stage of lung cancer in the individual.
39. A method of assessing the prognosis of an individual with lung cancer, said method comprising the steps of:
a) measuring the expression level of at least 2 genes/biomarkers selected from the group consisting of the genes listed in Table 1 in a biological sample obtained from the individual;
b) calculating a numerical biomarkcr score for the individual based on the expression levels of the biomarkcrs measured in step a); wherein the numerical biomarkcr score is predictive of the prognosis of the lung cancer in the individual.
40. A method of diagnosing, prognosing, classifying or grading lung cancer in a biological sample or an individual comprising measuring the expression levels of at least 2 genes listed in Table 1 in the biological sample or a test sample from the individual; and applying one or more network-based methods, one or more machine-learning based methods, or a combination of the foregoing methods to the expression levels to obtain a classification of the test sample as stage 1 lung adenocarcinoma, stage 2 lung adenocarcinoma, stage 1 squamous cell carcinoma or stage 2 squamous cell carcinoma.
41. The method according to paragraph 40, wherein a classifier or a previously established standard is used to classify a test sample as stage 1 lung adenocarcinoma, stage 2 lung adenocarcinoma, stage 1 squamous cell carcinoma or stage 2 squamous cell carcinoma.
42. The method according to paragraph 41 , wherein the classifier is obtained by training with a network-based method or a machine-learning based method using datascts obtained from subjects with stage 1 lung adenocarcinoma, stage 2 lung adenocarcinoma, stage 1 squamous cell carcinoma or stage 2 squamous cell carcinoma and datascts from subjects without lung cancer.
43. Λ method of classifying or grading a lung cancer tumor in an individual at risk for or having lung cancer comprising detecting the expression level of at least 2 genes listed in Table 1 in a test sample obtained from the individual; wherein a differential pattern of expression levels of said at least 2 genes in the test sample classifies the lung cancer tumor as one of stage 1 lung adenocarcinoma, stage 2 adenocarcinoma, stage 1 squamous cell carcinoma or stage 2 squamous cell carcinoma.
44. The method according to paragraph 43, wherein the differential pattern of expression levels is identi fied by a classifier based on a plurality of genes listed in Table I , including said at least two genes, said classifier having been trained by in silico analysis or one or more feature selection and classification algorithms.
45. The method according to paragraph 43 or 44, wherein the di fferential pattern of expression levels is identified by a classifier based on a plurality of genes listed in Table 1 , including said at least two genes, said classifier having been trained by one or more algorithms selected from the group consisting of dual ensemble, generalized simulated annealing, T-filtcr, CORG, CORG combined with support vector machine, dual bagging, single and pairs, forward learning,
Laplacian based learning and learning method based on network perturbation amplitude.
46. The method according to any one of paragraphs 43-45, wherein said classifier having been trained with at least the data in the Gene Expression Omnibus datascts GSE2109, GSE10245, GSE 1 8842 and GSE37745.
47. The method according to any one of paragraphs 43-46, wherein the method further comprises comparing the expression level of said at least 2 genes in the test sample and a control sample; or detecting the expression level of said at least 2 gcnes in the control sample and comparing the expression level of said at least 2 genes in the test sample and control sample, to identify the differential pattern.
48. The method according to any one of claims 43-47, wherein said at least 2 genes are selected from the group consisting of: ZIC2, LOC I 00131262, CD83, EMU , ΡΛΙΡ 1 , NIPBL, CREB3L1 , SLC37A1 , and SFMBT2.
49. The method according to any one of paragraphs 43-48, wherein the expression level of said at least 2 genes in the test sample are detected by using a human genome-wide array, a human lung tissue array or a custom array comprising polynucleotides of a plurality of genes in Table 1 and said at least 2 genes.
50. The method according to any one of paragraphs 43-48, wherein the expression level of said at least 2 genes in the test sample are detected by measuring the level of proteins encoded by the genes.
51. An array comprising polynucleotides hybridizing to at least 2 lung cancer signature genes immobilized on a solid surface, wherein the lung cancer signature genes are selected from the genes listed in Table 1 and said array is not a human genome-wide array.
52. A device comprising antibodies immobilized on a solid surface that bind to proteins encoded by at least 2 lung cancer signature genes, wherein the lung cancer signature genes arc selected from the group consisting of the genes listed in Table 1.
53. A computer readable medium or computer program product comprising a classifier based on at least two genes listed in Table 1 , said classifier having been trained by in silico analysis or one or more feature selection and classification algorithms.
54. The computer readable medium or computer program product according to paragraph 53, wherein said classifier is trained by one or more algorithms selected from the group consisting of dual ensemble, generalized simulated annealing, T- filter, CORG, CORG combined with support vector machine, dual bagging, single and pairs, forward learning, Lapiacian based learning and learning method based on network perturbation amplitude. 55. The computer readable medium or computer program product according to paragraph 53 or 54, wherein said classifier is trained with at least the data in the Gene Expression Omnibus datascts GSE2109, GSE10245, GSE18842 and
GSE37745.
56. The computer readable medium or computer program product according to any one of paragraphs 53-55, wherein said at least two genes are selected from the group consisting of ZIC2, LOC I 00131262, CD83, EML1 , PAIP 1 , NIPBL, CREB3L 1 , SLC37A1 , and SFMBT2.
57. A kit for classifying and grading a lung cancer tumor or for assessing the prognosis of lung cancer in an individual, comprising one or more reagents that detects expression levels of at least 2 genes listed in Table 1 in a test sample and instructions for using said kit for classifying and grading a lung cancer tumor or for determining the prognosis of lung cancer in said individual.
Brief Description of the Drawings
[0029] Figure 1 provides a features selection and classification algorithm(s) used for prediction of a gene signature.
Detailed Description of the invention
[0030] In order that the invention described herein may be fully understood, the following detailed description is set forth.
[0031 ] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as those commonly understood by one of skill in the art to which this invention belongs. In case of conflict, the present specification, including definitions, will control. Although methods and materials similar or equivalent to those described herein can be used in the practice or testing of the present invention, suitable methods and materials are described below. The materials, methods and examples are illustrative only, and are not intended to be limiting. Ail publications, patents and other documents mentioned herein are incorporated by reference in their entirety.
[0032] Throughout this specification, the word "comprise" or variations such as "comprises" or "comprising" will be understood to imply the inclusion of a stated integer or groups of integers but not the exclusion of any other integer or group of integers.
[0033] The term "antibody" refers to an immunoglobulin molecule capable of specific binding to a target, such as a carbohydrate, polynucleotide, lipid, polypeptide, etc., through at least one antigen recognition site, located in the variable region of the immunoglobulin molecule. As used herein, unless otherwise indicated by context, the term is intended to encompass not only intact polyclonal or monoclonal antibodies, but also engineered antibodies (e.g., chimeric, humanized and/or derivatized to alter effector functions, stability and other biological activities) and fragments thereof (such as Fab, Fab' , F(ab')2, Fv), single chain (ScFv) and domain antibodies, including shark and camelid antibodies), and fusion proteins comprising an antibody portion, multivalent antibodies, rnultispecii c antibodies (e.g., bispecific antibodies so long as they exhibit the desired biological activity) and antibody fragments as described herein, and any other modified configuration of the immunoglobulin molecule that comprises an antigen recognition site. An antibody includes an antibody of any class, such as IgG, IgA, or IgM (or sub class thereof), and the antibody need not be of any particular class. Depending on the antibody amino acid sequence of the constant domain of its heavy chains, immunoglobulins can be assigned to different classes. There are five major classes of immunoglobulins: IgA, IgD, IgE, IgG, and IgM, and several of these may be further divided into subclasses (isotypes), e.g., IgG l , IgG2, IgG3, IgG4, IgAl and IgA2 in humans. The heavy chain constant domains that correspond to the different classes of immunoglobulins are called alpha, delta, epsilon, gamma, and mu, respectively. The subunit structures and three dimensional configurations of different classes of immunoglobulins are well known.
[0034] The term "array" refers to the arrangement of biomarker detection molecules, such as nucleic acid probes or antibodies, on a solid support that allows for high-throughput screening of a sample to detect the presence and/or quantity of a biomarker. Such arrays may be used, e.g. , to evaluate the expression levels of several genes of interest in a single high-throughput reaction, lire array may be a nucleic acid array, such as a nucleic acid microarray; a protein array, such as a protein microarray; a peptide array, such as a peptide microarray; a tissue microarray, such as a tissue microarray or an antibody microarray, such as an antibody microarray. The solid substrate may be a microscopic bead, a glass slide, a plastic chip or a silicon chip.
[00351 The ter "biomarkcr" refers to a characteristic whose presence, absence or level indicates a biological state. Typically, the properties of biomarkers indicate a normal process, a pathogenic process or a response to a pharmaceutical or therapeutic intervention. Λ biomarker can be a cell, a gene, a gene product, an enzyme, a hormone, a protein, a peptide, an antibody, a nucleic acid molecule, a metabolite, a lipid, a free fatty acid, cholesterol or some other chemical compound. Λ biomarker can be a morphologic biomarkcr (for example, a histological change, DNA ploidy, malignancy-associated changes in the cell nucleus and premalignant lesions) or a genetic biomarker (for example, DNA mutations, DNA adducts and apoptotic index).
[0036] The term "classifying a lung cancer" refers to a method for determining the type of tumor from which a subject suffers. A subject can suffer from several different types of lung cancer, including but not limited to, adenocarcinoma, squamous cell carcinoma, large cell lung carcinoma, other non-small cell lung carcinomas, small cell lung carcinoma, carcinoid and unspecified lung cancer. Accordingly, a lung cancer tumor may be classified as one of these types o f lung cancer. A tumor may be classified based on histology, genetics or the presence, absence, alteration or levels of biomarkers. A lung cancer may be classified based on the lung cancer gene signature. The lung cancer may be classified as either adenocarcinoma or squamous cell carcinoma.
[0037] As used herein, the term "computer program" refers to a sequence of instructions, written to perform a specified task within a computer. For example, a computer program product is described, the product comprising computer-readable instructions that, when executed in a computerized system comprising at least one processor, cause the processor to carry out one or more steps of any of the methods described above. In another example, a computerized system is described, the system comprising a processor configured with non-transitory computer-readable instructions that, when executed, cause the processor to carry out any of the methods described herein. The computer program product and the computerized methods described herein may be implemented in a computerized system having one or more computing devices, each including one or more processors.
Generally, the computerized systems described herein may comprise one or more engines, which include a processor or devices, such as a computer, microprocessor, logic device or other device or processor that is configured with hardware, firmware, and software to carry out one or more of the computerized methods described herein. Any one or more of these engines may be physically separable from any one or more other engines, or may include multiple physically separable components, such as separate processors on common or different circuit boards. The computer systems of the present invention comprises means for implementing the methods and its various embodiments as described herein. The computerized system described herein may include a distributed computerized system having one or more processors and engines that communicate through a network interface. Such an implementation may be appropriate for distributed computing over multiple communication systems.
[0038] The term "computer readable medium" refers to a medium capable of storing data, such that the data may be accessed by a computer. Such a medium may take many forms, including but not limited to, non-volatile media and volatile media. Non-volatile media include, for example, optical, magnetic, or opto- magnetic disks, or integrated circuit memory, such as flash memory. Volatile media include, for example, dynamic random access memory (DRAM), which typically constitutes the main memory. Common forms of computer-readable media include, for example, a floppy disk, a flexible disk, hard disk, magnetic tape, magnetic cards, magnetic ink characters, magnetic drums, any other magnetic medium, a CD-ROM, DVD, any other optical medium, punch cards, paper tape, any other physical medium with patterns of holes, a RAM, a PROM, an EPROM or EEPROM (electronically erasable programmable read-only memory), a
FLASH-EEPROM, barcodes, semiconductors, microchips and any other memory chip or cartridge. [0039] The term "control sample'" refers to a sample against which a test sample is compared in order to diagnose, prognose, classify or grade the test sample, Λ control sample may be healthy tissue or may be a well-characterized tumor sample, including but not limited to, stage 1 adenocarcinoma, stage 2 adenocarcinoma, stage 1 squamous cell carcinoma, or stage 2 squamous cell carcinoma. A control sample can be analyzed concurrently with or separately from the test sample, including before or after analyzing the test sample. The data from the analysis of a control sample may be stored, e.g., in a computer readable medium or in a manual, for comparison against test samples analyzed in the future or as data for training network-based or machine-learning methods. A control sample may be developed as a medical standard for comparison. For example, analysis of control samples has developed medical standards for normal fed and fasted blood glucose levels; normal, at risk, and hypertensive blood pressures, and normal resting heart rates. As used herein, the term "control sample" includes samples that provided a medical standard. Accordingly, a test sample may be compared against a medical standard generated from control samples. For example, expression of a variant or mutated form of a gene may be indicative of a change medical condition.
Alternatively, a change in expression level of a gene may be indicative of a change in medical condition. A control sample may be lung tissue, such as tissue obtained by biopsy from a healthy portion of a lung (e.g., distant from a suspected tumor), or some other sample. For example, a control sample may be blood, blood cells, serum, plasma, sputum, saliva, tissue, bronchial washing, bronchial aspirates, bronchia brushings, exhaled breath, lymph fluid, and urine. Tissue specimens, such as those obtained by biopsy, may be fixed (e.g., formaldehyde- fixed paraffin- embedded (FFPE)). The control sample may be obtained from a tissue bank. The control sample may also be obtained from a cadaver or an organ donor.
[0040J The terms "differential pattern of expression" and "differential expression" arc used interchangeably herein and refer to a difference in an activity measurement (e.g., the variability or difference of genetic expression) of a biological entity under different conditions. For example, one condition may refer to an experimental treatment (such as exposure to a potentially carcinogenic agent), and another condition may refer to a control treatment (such as a null treatment). In an example, a fold-change is a number describing how much a measurement at a node (or biological entity) changes from an initial value to a final value between control data and treatment data, or between two sets of data representing different treatment conditions. The fold-change number may represent the logarithm of the fold-change of the activity of the biological entity between the two conditions. Λ confidence interval for the significance of the fold- change number may also be assessed.
[ 0041 ] The terms "gene signature" and "genetic signature" are used
interchangeably herein and refer to a group of genes expressed in a cell, whose combined expression pattern may be indicative of, e.g., a normal state, an at-risk state, a diseased state (e.g., stage 1 lung adenocarcinoma, stage 2 lung
adenocarcinoma, stage 1 squamous cell carcinoma or stage 2 squamous cell carcinoma), a treated state or a recovery state. Λ gene signature may be characterized by which genes are expressed and/or at what level each gene is expressed. Gene signatures are particularly useful in diagnosing, prognosing, classifying or grading complex diseases states, which result from the combination of several genetic and environmental factors. The gene signatures disclosed herein may be used, e.g., for the diagnosis, prognosis, classification or grading of lung cancer tumors in an individual . The gene signature may be unique to the class and grade of the tumor.
[0042 j The term "grading a lung cancer" refers to a method for determining the grade of tumor from which a subject suffers. A subject can suffer from several different grades of lung cancer, which reflect the seventy and invasiveness of the lung cancer. For example, stage 0 refers to a carcinoma in situ; stage 1 refers to cancers that are localized to one part of the body; stage 2 refers to cancers that arc locally advanced; stage 3 refers to cancers that are further advanced locally (e.g., as evidenced by increased lymph node involvement) and stage 4 refers to cancers that have metastasized. Lung cancers are typically staged according to the TNM system. The T category defines the primary tumor by size and whether it has spread into the surrounding tissue. The N category identifies any lymph node involvement in and around the lungs. Finally, the M category indicates whether the cancer has metastasized. Λ tumor may be graded based on histology, genetics or the presence, absence, or levels of biomarkers. A lung cancer may be graded based on its gene signature. The lung cancer may be graded as stage 1
adenocarcinoma, stage 2 adenocarcinoma, stage 1 squamous cell carcinoma or stage 2 squamous cell carcinoma.
[0043] The term "/« silico analysis" refers to analysis performed on a computer or via computer simulation. Gene signature analysis involves detection of gene expression based on identity and expression level for a multitude of genes. In silico analysis may apply one or more network-based methods, one or more machine-learning based methods, or a combination of the foregoing methods to the expression levels to obtain a classification of the test sample, e.g., as stage 1 lung adenocarcinoma, stage 2 lung adenocarcinoma, stage 1 squamous cell carcinoma or stage 2 squamous cell carcinoma. Comparisons between expression levels from test samples and control samples may require computer analysis to determine the degree and significance of any changes observed. See, e.g. , U.S. Provisional Patent Application entitled "Systems and Methods relating to Network-based Biomarker Signatures," filed concurrently with the instant application,
incorporated herein by reference in its entirety and having the attorney docket no. 106500-0022-001 ; U.S. Provisional Patent Application entitled "Systems and Methods for Generating Biomarker Signatures," filed concurrently with the instant application, incorporated herein by reference in its entirety and having the attorney docket no. 106500-0028-001 ; U.S. Provisional Patent Application entitled
"Systems and Methods for Generating Biomarker Signatures with Integrated Bias Correction and Class Prediction," filed concurrently with the instant application, incorporated herein by reference in its entirety and having the attorney docket no. 106500-0032-001 ; and U.S. Provisional Patent Application entitled "Systems and Methods for Generating Biomarker Signatures with Integrated Dual Ensemble and Simulated Annealing Techniques," filed concurrently with the instant application, incorporated herein by reference in its entirety and having the attorney docket no. 106500-0031 -001.
[0044] The term "individual" refers to a vertebrate, preferably a mammal. The mammal can be, without limitation, a mouse, a rat, a cat, a dog, a horse, a pig, a cow, a non-human primate or a human. [0045] The term "individual at risk for" lung cancer" refers to an individual who is predisposed to lung cancer. Predisposition to lung cancer may be due to one or more genetic or environmental factors. For example, an individual related to a lung cancer patient is 2.4 times more likely to get lung cancer than an individual who is not related to a lung cancer patient. Further, exposure to environmental factors such as radon gas, asbestos, tobacco smoke, and air pollution can increase the risk for lung cancer and predispose an individual to lung cancer.
[0046] The term "individual having lung cancer" or "individual suffering from lung cancer" refers to an individual experiencing uncontrolled cell growth in the tissues of the lung. Lung cancers typically form solid tumors, which can be observed on a chest X-ray or by a CT scan.
[0047] The term "MALDI-TOF" refers matrix-assisted laser
desorption/ionization time of flight mass spectroscopy. Matrix-assisted laser desorption/ionization (MALDI) is a two step process that uses laser-triggered desorption of protonated and deprotonatcd matrix materials to protonate or deprotonatc analyte molecules (e.g., DNA, RNA, and proteins). Time-of-fiight (TOF) mass spectrometry refers to a method in which an ion's mass-to-charge ratio is determined by measuring the time that it takes an ionized particle to reach a detector at a known distance.
[0048] The term " network-based analysis" refers to an approach to identify biomarkers that is based on the properties of groups of functionally interrelated genes that form a network in a biological system, instead of treating individual genes in the biological system a priori as completely independent and identical.
[0049] The term "machine learning methods" refers to methods that allow a machine, such as a programmable computer, to improve its performance at a certain predictive task that is based on the known properties of examples or training data. Machine learning methods include, without limitation, support vector machines (SVMs), network-based SVMs, ensemble classifiers, neural network-based classifiers, logistic regression classifiers, decision tree-based classifiers, classifiers employing a linear discriminant analysis technique, a random-forest analysis technique, or both. [0050] The term "numerical biomarker score" refers to a number that is representative of the result(s) of one or more of the network-based analysis or machine learning methods.
[0051 j The term "polynucleotide hybridizing to" refers to a polynucleotide molecule that binds to a target nucleic acid molecule through complementary base pair sequencing. Hybridization typically requires two nucleic acids that contain complementary sequences, although depending on the stringency of the
hybridization, mismatches between bases are possible. The appropriate stringency for hybridizing nucleic acids depends on the length of the nucleic acids and the degree o f complementation, variables well known in the art. Exemplary high stringent hybridization conditions are equivalent to about 20-27 °C below the melting temperature (Tm) of the DNA duplex formed in about 1 M salt. Many equivalent procedures exist and several popular molecular cloning manuals describe suitable conditions for highly stringent hybridization and, furthermore, provide formulas for calculating the length of hybrids expected to be stable under these conditions (see, e.g., Current Protocols in Molecular Biology, John Wiley & Sons, N. Y. ( 1989), 6.3.1 6 or 13.3.6; or pages 9.47-9.57 of Sambrook, et al. (1989) Molecular Cloning. 2nd cd., Cold Spring Harbor Press). "High stringency" refers to hybridization and/or washing conditions at 68 °C in 0.2 x SSC, at 42 °C in 50 % formamide, 4 x SSC, or under conditions that afford levels of hybridization equivalent to those observed under either of these two conditions. The greater the degree of similarity or homology between two nucleotide sequences, the greater the value of Tm, for hybrids of nucleic acids having those sequences. The relative- stability (corresponding to higher Tm) of nucleic acid hybridizations decreases in the following order: RNA:RNA, DNA:RNA, DNA:DNA.
[0052] The terms "protein," "polypeptide" and "peptide" are used
interchangeably and indicate at least one molecular chain of amino acids linked through covalcnt or non-covalent bonds. The terms do not refer to a specific length of the molecular chain. Peptides, oligopeptides and proteins are included within the definition of "polypeptide". The terms include post-translational modifications of the molecule, e.g., phosphorylation, glycosylation and acetylalion. The terms also include protein fragments, fusion proteins, mutant proteins and variant proteins.
[0053] The term "SELDI-TOF" refers surface-enhanced laser
desorption/ionization time of flight mass spectroscopy. Surface-enhanced laser desorption/ionization (SELDI) is a variant of MALDI that uses a target with a biochemical affinity for the analyte. Timc-of-flight (TOP) mass spectrometry refers to a method in which an ion's mass-to-charge ratio is determined by measuring the time that it takes an ionized particle to reach a detector at a known distance.
[0054] The term "test sample" refers to a sample obtained from an individual at risk for, having or suffering from lung cancer. A test sample may be any sample suspected of containing or exhibiting a biomarker. The test sample is analyzed and compared to a control sample, including medical standards developed from control samples, to diagnose, prognose, classify or grade lung cancer in the individual. Λ test sample may be obtained from lung tissue, such as tissue obtained by biopsy from a tumor, or other biological tissue. For example, a test sample may be blood, blood cells, scrum, plasma, sputum, saliva, tissue, bronchial washing, bronchial aspirates, bronchia brushings, exhaled breath, lymph fluids, and urine. Tissue specimens, such as those obtained by biopsy, may be fixed (e.g., formaldehydc- fixed paraffin-embedded (FFPE)).
[0055] As used herein, to "train" a data set means to generate a classifier that can accurately predict classifications of a set of test samples. For example, a training data set includes a set of samples, and each sample may correspond to a measurement from a different patient. A machine learning technique is applied to the training data set to generate a "classifier," which corresponds to a way of assigning each sample in the training data set to a category (such as "disease positive" or "disease free"). In addition to the training data set, a training class set is known. The training class set includes a known category assigned to each sample (or person). The categories predicted by the classifier are compared to the known categories. I f the predicted categories mostly match the known categories, the classifier has performed well. However, if there are substantial differences between the predicted categories and the known categories, the parameters of the machine learning technique may be updated, and the updated machine learning technique is applied. These steps are repeated until the performance of a classifier exceeds a threshold, and the final classifier is provided. The final classifier may then be applied to a test data set. The test data set may correspond to measured samples from different patients, but the patients in the test data set may have unknown categories (disease states). Thus, applying the final classifier to the test data set thus allows for prediction of the disease states of the patients.
Gene Signatures
[0056] One aspect of the invention provides gene signatures useful for diagnosing, prognosing, classifying or grading a lung cancer tumor. In some embodiments, the gene signature comprises at least 2 genes selected from the genes listed in Tabic 1. In some embodiments, the gene signature comprises at least 2, at least 3, at least 4, at least 5, at least 10, at least 15, at least 20, at least 25, at least 30, at least 35, at least 40, at least 45, at least 50, at least 55, at least 60, at selected from the genes listed in Table 1. In some embodiments, the gene signature comprises each of the genes listed in Table 1. Optionally, said at least 2, at least 3, at least 4, or at least 5 genes are selected from the group consisting o : ZIC2, LOCI 00131262, CD83, EML1 , PAIP 1 , NIPBL, CREB3L 1 , SLC37A 1 , and SFMBT2, which are the genes that appear in 4 of the 5 lists generated in Example
1 .
[0057] In some embodiments, at least 2, at least 3, at least 4, at least 5, at least 10, at least 1 5, at least 20, at least 25, at least 30, at least 35, at least 40, at least 45, at least 50, at least 55, at least 60, at least 65, or at least 70 of the genes selected from the genes listed in Table 1 have increased expression compared to a control sample. In some embodiments, at least 2, at least 3, at least 4, at least 5, at least 10, at least 15, at least 20, at least 25, at least 30, at least 35, at least 40, at least 45, at least 50, at least 55, at least 60, at least 65, or at least 70 of the genes selected from the genes listed in Table 1 have decreased expression compared to a control sample.
[0058] In some embodiments, the gene signature includes a degree of up- regulation of a subset of genes in the gene signature compared to the control sample. For example, each up-regulated gene in the gene signature may, independently, be up-regulated at least 1.5-fold, at least 2-fold, at least 2.5-fold, at least 3-fold, at least 3.5-fold, at least 4-fold, at least 4.5-fold, at least 5-fold, at least 6-fold, at least 7-fold, at least 8-fold, at least 9-fold, at least 10-fold, at least 100- fold, at least 1 ,000-fold or more compared to the control sample. Similarly, in embodiments, the gene signature includes a degree of down-regulation of ; subset of genes in the gene signature compared to the control sample. For example, each down-regulated gene in the gene signature may, independently, be down-regulated at least 1 .5-fold, at least 2-fold, at least 2.5-fold, at least 3-fold, at
10 least 3.5-fold, at least 4-fold, at least 4.5-fold, at least 5-fold, at least 6-fold, at least 7-fold, at least 8-fold, at least 9-fold, at least 10-fold, at least 100-fold, at least 1 ,000-fold or more compared to the control sample.
fhe present invention encompasses the following gene signati
1. Λ and B;
15 ii. A, B, and C;
iii. A, B, C, and D;
iv. A, B, C, D, and E;
v. A, B, C, D, E, and F;
vi. A, B, C, D, E, F, and G;
~>n A, B, C, D, E, F, G, and II;
Vlll. A, B, C, D, E, F, G, H, and 1;
ix. A, B, C, D, E, F, G, I I, I, and J;
v, i3, u, r., r , vjr, i i, i, j , anu is.
xi. A, B, C, D, E, F, G, I I, I, J, , and L
25 xii. A, B, C, D, E, F, G, 11, 1, J, K, L and M
xiii. A, B, C, D, E, F, G, I I, I, J, K, L, M, and N
xiv. A, B, C, D, E, F, G, H, I, J, , L, M, N, and O;
XV. A, B, C, D, E, F, G, II, I, J, K, L, M, N, O, and P;
xvi. A, B, C, D, E, F, G, I L I, J, K, L, M, N, O, P, and Q
XVil. A, B, C, D, E, F, G, 11, 1, J, , L, , N, O. P, Q, and R;
xviii. A. B, C, D, E, F, G, II, I, J, K, L, M, N, O, P, Q, R, and S;
xix. A, B. C, D, E, F, G, II, I, J, K, L. M, N, O, P, Q, R, S, and xx. A, B, C, D, E, F, G, I I, I, J, K, L, M, N, O, P, Q, R, S, T, and U;
xxi. A, B, C, D, E, F, G, I I, I, J, , L, M, N, O, P, Q, R, S, T, U, and V;
xxii. A, B, C, D, E, F, G, I I, I, J, K. L, M, N, O, P, Q, R, S, T, U, V, and W; xxiii. A, B, C, D, E, F, G, I I, I, J, , L, M, N, O, P, Q, R, S, T, U, V, W, and X; xxiv. A, B, C, D, E, F, G, I I, I, J, K, L, M, N, 0, P, Q, R, S, T, U, V, W, X, and
Y;
xxv. A, B, C, D, E, F, G, I I, I, J, K, L, M, N, 0, P, Q, R, S, T, U, V, W, X, Y, and Z;
xxvi. A, B, C, D, E, F, G, I I, I, J, K, L, M, N, O, P, 0, R, S, ! . U, W, X, Y, Z, and AA;
xxvii. A, B, C, D, E, F, G, I I, I, J, K, L, M, N, O, P, Q, R, S, T, U, W, X, Y, Z,
AA, and BB;
xxviii. Λ, B, C, D, E, F, G, I I, I, J, K, L, M, N, 0, P, Q. R, S, T, U, W, X, Y, Z,
AA, AB and AC;
xxix. A, B, C, D, E, F, G, I I, I, J, K, L, M, N, O, P, Q, R, S, T, U. W, X, Y, Z.
AA, AB, AC, and AD;
xxx. A. B, C, D, E, F, G, I I, I, J, K, L, M, N, O, P, Q, R, S, T, U, \Y. X, Y, Z, ΛΛ, AB, AC, AD and AE;
xxxi. A, B, C, D. E, F, G, I I, I, J, K, L, M, N, 0, P, Q, R, S, T, U, W, X, Y. Z, ΛΛ, AB, AC, AD. AE and AF;
xxxii. A, B, C, D, E, F, G, I I, I, J, K, L, M, N, O, P, Q, R, S, T, U, W, X, Y, Z,
AA, AB, AC, AD, AE, AF and AG;
xxxiii A B C D E F Γ, H J } K N 0 P Π R S T U W Y Y Z
AA, AB, AC, AD, AE. AF, AG and AH;
xxxiv. A, B, C, D, E, F, G, I I, I, J, K, L, M, N, O, P, Q, R, S, T, U, W, X, Y, Z,
AA, AB, AC, AD, AE, AF, AG. AH and ΛΙ;
xxxv. A, B, C, D, E, F, G, I I, I, J, K, L, M, N, O, P, Q, R, S, T, U, W, X, Y, Z,
AA, AB, AC, AD, AE, AF, AG, AH, AI and AJ;
xxxvi. A, B, C, D, E, F, G, I I, I, J, , L, , N, O, P, Q, R, S, T, U, W, X, Y, Z,
AA, AB, AC, AD, AE, AF, AG, AH, AI, AJ and ΛΚ;
xxxvii. A, B, C, D, E, F, G, I I, I, J, K, L, M, N, O, P, Q, R, S, T, U, W, X, Y, Z,
AA, AB, AC, AD, AE, AF, AG, AH, AI, AJ, AK and AL; xxxviii. Λ, B. C, D, E, F, G, II, I, J, K, L, M, N, O, P, Q. R, S, T, U, W, X, Y, Z, ΛΛ, ΛΒ, AC, AD, AE, AF, AG, AH, AI, AJ, AK, AL and AM; xxxix. A, B, C, D, E, F, G, I I, I, J, K, L, M, N, O, P, Q, R, S, T, U, W, X, Y, Z,
AA, AB, AC, AD, AE, AF, AG. AH, AI, AJ, AK, AL, AM and AN; xl. A, B, C, D, E, F, G, H, I, J, K, L, M, N, O, P, Q, R, S, T, U, W, X, Y, Z, AA, AB, AC, AD, AE, AF, AG, AH, AI, AJ, AK, AL, AM, AN and AO; xli. A, B, C, D, E, F, G, I I, I, J, K, L, M, N, O, P, Q, R, S, T, U, W, X, Y, Z, AA, AB, AC, AD, AE, AF, AG, AH, AI, AJ, AK, AL, AM, AN, AO and AP;
xlii. A, B, C, D, E, F, G, H, I, J, K, L, M, N, O, P, Q, R, S, T, U. W, X, Y, Z, AA, AB, AC, AD, AE, AF, AG. AH, AI, AJ, AK, AL, AM, AN. AO, AP and AQ;
xliii. A, B, ( '. D, E, F, G, I t, I, J, K, L, M, N, O, P, Q, R, S, T, U, W, X, Y, Z, AA. AB, AC, AD, AE, AF, AG, AH, AI, AJ, AK, AL, AM, AN, AO, AP, AQ and AR;
xliv. A, B, C, D, E, F, G, I I, I, J, K, L, M, N, O, P, Q, R, S, T, U, W, X, Y, Z, AA, AB, AC, AD, AE, AF, AG, AH, AI, AJ, AK, AL, AM, AN, AO, AP, AQ, AR and AS;
xlv. A, B, C, D, E, F, G, H, I, J, K, L, M, N, O, P, Q, R, S, T, U, W, X, Y, Z, AA, AB, AC, AD, AE, AF, AG, AH, AI, AJ, AK, AL, AM, AN, AO, AP,
AQ, AR, AS and AT;
xlvi. A, B, C, D, E. F, G, II, I, J, K. L, M, N, O, P, Q, R, S, T, U, W, X, Y, Z, AA, AB, AC, AD, AE, AF, AG, AH, AI, AJ, AK, AL, AM, AN, AO, AP, AQ. AR, AS, AT and AIJ;
xlvii. A, B, C, D, E, F, G, I I, I, J, K, L, M, N, O, P, Q, R, S, T, U, W, X, Y, Z.
AA, AB, AC, AD, AE, AF, AG, All, AI, AJ, AK, AL, AM, AN, AO, AP, AQ, AR, AS, AT, AU and AV;
xlviii. A, B, C, D, E, F, G, IT, I, J, K, L, M, N, O, P, Q, R, S, T, U, W, X, Y, Z„
AA, AB, AC, AD, AE, AF, AG, AH, AI, AJ, AK, AL, AM, AN, AO, AP, AQ, AR, AS, AT, AU, AV and AW; xlix. A, B, C, D, E, F, G, H, I, J, K, L, M, N, 0, P, Q, R, S. T, U, W, X, Y, Z, AA, AB. AC, AD, ΛΕ, AF, AG, AH, AI, AJ, AK, AL. AM, AN, AO, AP, AQ, AR, AS, AT, AU, AV, AW and AX;
I. A, B, C, D, E, F, G, I I, I, J. K, L, M, N, 0, P, Q, R, S, T, U, W, X, Y, Z, AA, AB, AC, AD, AE, AF, AG, AH, AI, AJ, AK, AL, AM, AN, AO, AP,
AQ, AR, AS, AT, AU, AV, AV/, AX and AY;
li. A, B, C, D, E, F, G, I I, I, J, K, L, M, N, 0, P, Q, R, S, T, U, W, X, Y, Z, AA, AB. AC, AD, AE, AF, AG. Al l, ΛΙ, AJ, AK, AL, AM. AN, AO, AP, AQ, AR, AS, AT, AU, AV, AW, AX, AY and AZ;
Hi. A, B, C, D, E, F, G, I I, I, J, K, L, M, N, O, P, Q, R, S, T, U, W, X, Y, Z, AA, ΛΒ, AC, AD, AE, AF, AG, AH, AI, AJ, AK, AL, AM, AN, AO, AP, AQ, AR, AS, AT, AU, AV, AW. AX, AY, AZ and BA;
liii. A, B, C, D. E, F, G, II, I, J, K, L, M, N, O, P, Q, R, S, T, U. W, X, Y, Z, ΛΛ, ΛΒ, AC, AD, AE, AF, AG, AH, AI, AJ, AK, AL, AM, AN, AO, ΛΡ, AQ, AR, AS, AT, AU, AV, AW, AX, AY, ΛΖ, BA and BB;
liv. A, B, C. D, E, F, G, II, I, J, K, L, M, N, O, P, Q, R, S, T, U, W, X, Y, Z, AA, AB. AC, AD, AE, AF, AG, AH, AI, AJ, AK, AL. AM, AN, AO, ΛΡ, AQ, AR. AS. AT, AU, AV, AW. AX, AY, AZ, BA, BB and BC;
Iv. A, B. C, D, E, F, G, Π, I, J, K, L, M, N, O, P, Q, R, S, T, U, W, X, Y. Z, AA, ΛΒ, AC, AD, AE, AF, AG, AH, AI, AJ, AK, AL, AM, AN, AO, AP,
AQ, AR, AS, AT, AU, AV, AW, AX, AY, AZ, BA, BB, BC and BD; Ivi. A, B, C, D, E, F, G, H, I, J, K, L, M, N, O, P, Q, R, S, T, U, W, X, Y, Z, AA, AB, AC, AD, AE, AF, AG, AH, AI, AJ, AK, AL,, AM, AN, AO, AP, AQ, AR, AS, AT, AU, AV, AW, AX, AY, AZ, BA, BB, BC, BD and BE; Ivii. A, B, C, D, E, F, G, I I , I, J. K. L, M, N, O, P, Q, R, S, T, U, W, X, Y, Z, AA, AB, AC, AD, AE, AF, AG, AH, AI, AJ, AK, AL, AM, AN, AO, AP, AQ, AR, AS, AT, AU, AV, AW, AX, AY, AZ, BA, BB, BC, BD, BE and BF;
Iviii. A, B, C, D. E, F, G, I I, I, J, K, L, M, N, 0, P, Q, R, S, T. U, W, X, Y, Z, AA, AB, AC, AD, AE, AF, AG, AH, AI, AJ, AK, AL, AM, AN, AO, AP,
AQ, AR, AS, AT, AU, AV, AW, AX, AY, AZ, BA, BB, BC, BD, BE, BF and BG; lix. Λ, B, C, D, E, F, G, I I, I, J, K, L, M, N, O, P, Q, R, S, T, U, W, X, Y, Z, AA, AB, AC, AD, AE, AF, AG, AH, AI, AJ, AK, AL, AM, AN. AO, AP, AQ, AR, AS, AT, AU, AV, AW, AX, AY, AZ, BA, BB, BC, BD, BE, BF, BG and BIT;
lx. A, B, C, D, E, F, G. I I, I, J, K. L, M, N, 0, P, Q, R, S, T, U, W, X. Y, Z, AA, AB, AC, AD, AE, AF, AG, AH, AI, AJ, AK, AL, AM, AN, AO, AP, AQ, AR, AS, AT, AU, AV, AW, AX, AY, AZ, BA, BB, BC, BD, BE, BF, BG, B1 I and Bl;
Ixi. A, B, C, D, E, F, G, II, I, J, , L, M, N, O, P, Q, R, S, T, U, W, X, Y, Z, AA, AB, AC, AD, AE, AF, AG, AH, AI, AJ, AK, AL, AM, AN, AO, AP,
AQ, AR, AS, AT, AU, AV, AW, AX, AY, AZ, BA, BB, BC, BD, BE, BF, BG, ΒΠ, Bl and BJ;
ixii. A, B, C, D, E, F, G, I I, I, J, K. L, M, N, O, P, Q, R, S. T, U, W, X, Y, Z, AA, AB, AC, AD, AE, AF, AG, AH, AI, AJ, AK, AL, AM, AN, AO, AP, AQ, AR, AS, AT, AU, AV, AW, AX, AY, AZ, BA, BB, BC, BD, BE, BF,
BG, BI L BI, BJ and BK;
Ixiii. A, B, C, D, E, F, G, H, I, J, K, L, M, N, O, P, Q, R, S, T, U, W, X, Y, Z, AA, AB, AC, AD, AE, AF, AG, AH, AI, AJ, AK, AL, AM, AN, AO, AP, AQ. AR. AS, AT, AU, AV, AW, ΛΧ, AY, AZ, BA, BB, BC, BD, BE, BF, BG, BH, Bl, BJ, BK and BL;
lxiv. A, B, C, D, E, F . G. II, I, J, K, L, M, N, O, P, Q, R, S, T, U, W, X, Y, Z, AA, AB, AC, AD, AE, AF, AG, AH, AI, AJ, AK, AL, AM, AN, AO, AP, AQ, AR, AS, AT, AU, AV, AW, AX, AY, AZ, BA, BB, BC, BD, BE, BF, BG, BI L Bl, BJ, BK, BL and BM;
Ixv. A, B, C, D, E, F, G, II, I, J, K, L, M, N, O, P, Q, R, S, T, U, W, X, Y, Z, AA, AB, AC, AD, AE, AF, AG, AH, AI, AJ, AK, AL, AM, AN, AO, AP, AQ, AR, AS, AT, AU, AV, AW, AX, AY, AZ, BA, BB, BC, BD, BE, BF, BG, BH, Bl, BJ, BK, BL, BM and BN;
lxvi. A, B, C, D, E, F, G, I I, I, J, K, L, M, N, O, P, Q, R, S, T, U, W, X. Y, Z, AA, AB, AC, AD, AE, AF, AG, AH, AL A J, AK, AL, AM, AN, AO, AP,
AQ, AR, AS, AT, AU, AV, AW, AX, AY, AZ, BA, BB, BC, BD, BE, BF, BG, BH. Bl, BJ, BK, BL, BM, BN and BO; lxvii. A, B, C, D, E, F, G, H, I, J, K, L, M, N, O, P, Q, R, S, T, U, W, X, Y, Z, AA, AB, AC, AD, AE, AF, AG, AH, ΛΪ, A J, AK, AL, AM, AN, AO, AP, AQ, AR, AS, AT, AU, AV, AW, AX, AY, AZ, BA, BB, BC, BD, BE, BF, BG, BH, BJ, BJ, BK, BL, BM, BN, BO and BP;
lxviii. A, B, C, D, E, F, G, H, I, J, K, L, M, N, O, P, Q, R, S, T, U, W, X, Y, Z,
AA, AB, AC, AD, AE, AF, AG, AH, ΛΙ, AJ, AK, AL, AM, AN, AO, AP, AQ, AR, AS, AT, AU, AV, AW, AX, AY, AZ, BA, BB, BC, BD, BE, BF, BG, BH, BI, BJ, BK, BL, BM, BN, BO, BP and BQ;
lxix. A, B, C, D, E, F, G, H, I, J, K, L5 M, N, O, P, Q, R, S, T, U, W, X, Y, Z, AA, AB, AC, AD, AE, AF, AG, AH, AI, AJ, AK, AL, AM, AN, AO, AP,
AQ, AR, AS, AT, AU, AV, AW, AX, AY, AZ, BA, BB, BC, BD, BE, BF, BG, BH, BI, BJ, BK, BL, BM, BN, BO, BP, BQ and BR;
lxx. A, B, C, D, E, F, G, H, I, J, K, L, M, N, O, P, Q, R, S, T, U, W, X, Y, Z.
AA, AB, AC, AD, AE, AF, AG, AH, AI, AJ, AK, AL, AM, AN, AO, AP, AQ, AR, AS, AT, AU, AV, AW, AX, AY, AZ, BA, BB, BC, BD, BE, BF,
BG, B1 I, BI, BJ, BK, BL, BM, BN, BO, BP, BQ, BR and BS;
Ixxi. A, B, C, D, E, F, G, I I, I, J, K, L, M, N, O, P, Q, R, S, T. U, W, X, Y, Z, AA, AB, AC, AD, AE, AF, AG, AH, AI, AJ, AK, AL, AM, AN, AO, AP, AQ, AR, AS, AT, AU, AV, AW, AX, AY, AZ, BA, BB, BC, BD, BE, BF, BG, BH. B l, BJ. BK, BL, BM. BN, BO, BP, BQ, BR, BS and BT;
lxxii. A, B, C, D. E, F, G, II, I, J, K, L, M, N, O, P, Q, R, S, T, U, W, X, Y, Z.
AA, AB, AC, AD, AE, AF, AG, AH, AI, AJ, AK, AL, AM, AN, AO, AP, AQ, AR, AS, AT, AU, AV, AW, AX, AY, AZ, BA, BB, BC, BD, BE, BF, BG, BH, BI, BJ, BK, BL, BM, BN, BO, BP, BQ, BR, BS, BT and BU; Ixxiii. A, B, C, D, E, F, G, H, I, J, K, L, M, N, O, P, Q, R, S, T, U, W, X, Y, Z,
AA, AB, AC, AD, AE, AF, AG, AH, AI, AJ, AK, AL, AM, AN, AO, AP, AQ, AR, AS, AT, AU, AV, AW, AX, AY, AZ, BA, BB, BC, BD, BE, BF, BG, BH, BI, BJ, BK, BL, BM, BN, BO, BP, BQ, BR, BS, BT, BU and BV; Ixxiv. A, B, O D, E, F, G, H, I, J, K, L, M, N, O, P, Q, R, S, T, U, W, X, Y, Z,
AA, AB, AC, AD, AE, AF, AG, AH, AI, AJ, AK, AL, AM, AN, AO, AP,
AQ, AR, AS, AT, AU, AV, AW, AX, AY, AZ, BA, BB, BC, BD, BE, BF, BG, BI I, BI, BJ, B , BL, BM, BN, BO, BP, BQ, BR, BS, BT, BU, BV and
BW;
lxxv. A, B, C, D, E, F, G, I I, I, J, K, L, M, N, O, P, Q, R, S, T, IJ, W, X, Y, Z, AA, ΛΒ, AC, AD, AE, AF, AG, AH, AI, AJ, AK, AL, AM, AN, AO, AP, AQ, AR, AS. AT, AU, AV, AW, AX, AY, AZ, BA, BB, BC, BD, BE, BF,
BG, BH, BI, BJ, BK, BL, BM, BN, BO, BP, BQ, BR, BS, BT, BU, BV, BW and BX;
Ixxvi. A, B, C, D, E, F, G, H, I, J, K, L, M, N, (), P, Q, R, S, T, U, W, X, Y, Z,
AA, AB. AC, AD, AE, AF, AG, AH, AI, AJ, AK, AL. AM, AN, AO, AP, AQ, AR. AS, AT, AU, AV, AW, AX, AY, AZ, BA, BB, BC, BD, BE, BF,
BG, BI I, BI, BJ, BK, BL, BM, BN, BO, BP, BQ, BR, BS, BT, BU, BV, BW, BX and BY;
lxxvii. A, B, C, D, E, F, G, H, I, J, K, L, M, N, O, P, Q, R, S, T, U, W, X, Y, Z,
AA, AB, AC, AD, AE, AF, AG, AH, AI, AJ, AK, AL, AM, AN, AO, AP, AQ, AR. AS, AT, AU, AV, AW, AX, AY, AZ, BA, BB, BC, BD, BE, BF,
BG, BI I. BI, BJ, BK, BL, BM, BN, BO, BP, BQ, BR, BS, BT, BU, BV,
BW, BX, B Y and BZ;
Ixxviii. A, B, C, D, E, F, G, H, I, J, K, L, M, N, O, P, Q, R, S, T, U, W, X, Y, Z,
AA, AB, AC, AD, AE, AF, AG, AH, AI, AJ, AK, AL, AM, AN, AO, AP, AQ, AR, AS, AT, AU, AV, AW, AX, AY, AZ, BA, BB, BC, BD, BE, BF,
BG, BI I, BI, BJ, BK, BL, BM, BN, BO, BP, BQ, BR, BS, BT, BU, BV,
BW, BX, BY, BZ and CA;
Ixxix. A, B, C, D, E, F, G, IT, I, J, K, L, M, N, O, P, Q, R, S, T, U, W, X, Y, Z,
AA, AB, AC, AD, AE, AF, AG, AH, AI, AJ, AK, AL, AM, AN, AO, AP, AQ, AR, AS, AT, AU, AV, AW, AX, AY, AZ, BA, BB, BC, BD, BE, BF,
BG, BI I. BI, BJ, BK, BL, BM. BN, BO, BP, BQ, BR, BS, BT, BU, BV,
BW, BX. BY, BZ, CA and CB;
Ixxx. A, B, C, D, E, F, G, II, I, J, K, L, M, N, O, P, Q, R, S, T, U, W, X, Y, Z,
AA, AB, AC, AD, AE, AF, AG, AH, AI, AJ, AK, AL, AM, AN, AO, AP, AQ, AR, AS, AT, AU, AV, AW, AX, AY, AZ, BA, BB, BC, BD, BE, BF,
BG, BI I, BI, BJ, BK, BL, BM, BN, BO, BP, BQ, BR, BS, BT, BU, BV,
BW, BX, BY, BZ, CA, CB and CC; ixxxi. A, B, C, D, E, F, G, H, I, J, K. L, M, N, O, P, Q, R, S, T, U, W, X, Y, Z, ΑΛ, AB, AC, AD, ΛΕ, AF, AG, AH, AI, AJ, AK, AL, AM, AN, AO, AP, AQ, AR, AS, AT, AU, AV, AW, AX, AY, AZ, BA, BB, BC, BD, BE, BF, BG, BH, BI, BJ, BK, BL, BM, BN, BO, BP, BQ, BR, BS, BT, BU, BV, BW, BX, BY, BZ, CA, CB, CC and CD;
lxxxii. A, B, C, D, E, F, G, H, I, J, K, L, M, N, O, P, Q, R, S, T, U, W, X, Y, Z,
AA, AB, AC, AD, AE, AF, AG, AH, AI, AJ, AK, AL, AM, AN, AC), AP, AQ, AR, AS, AT, AU, AV, AW, AX, AY, AZ, BA, BB, BC, BD, BE, BP, BG, BH, BI. BJ, BK, BL, BM, BN, BO, BP, BQ, BR, BS, BT, BU, BV, B W, BX, BY, BZ. CA, CB, CC, CD and CE;
Ixxxiii. A, B, C, D, E, F, G, I I, I, J, K, L, M, N, 0, P, Q, R, S, T, U, W, X, Y, Z,
AA, AB, AC, AD, AE, AF, AG, AH, AI. AJ, AK, AL, AM, AN, AO, AP, AQ, AR, AS, AT, AU, AV, AW, AX, AY, AZ, BA, BB, BC, BD, BE, BF, BG, BH, BI, BJ, BK, BL, BM, BN, BO, BP, BQ, BR, BS, BT, BU, BV, BW, BX, BY, BZ, CA, CB, CC, CD, CE and CF; and
Ixxxiv. A, B, C, D, E, F, G, II, I, J, K, L, M, N, 0, P, Q, R, S, T, U, W, X, Y, Z,
AA, AB, AC, AD, AE, AF, AG, AH, AI. AJ, AK, AL, AM, AN, AO, AP, AQ, AR, AS, AT, AU. AV, AW, AX, AY, AZ, BA, BB, BC, BD, BE, BF, BG, BH, BI, BJ, BK, BL, BM, BN, BC), BP, BQ, BR, BS, BT, BU, BV, BW, BX, BY, BZ, CA, CB, CC, CD, CE, CF and CG;
Ixxxv. A, B, C, D, E, F, G, II, I, J, K, L, M, N, 0, P, Q, R, S, T, U, W, X, Y, Z,
AA, AB, AC, AD, AE, AF, AG, AH, AI, AJ, AK, AL, AM, AN, AO, AP, AQ, AR, AS, AT, AU, AV, AV.7, AX, ΛΥ, AZ, BA, BB, BC, BD, BE, BF, BG, BH, BI, BJ, BK, BL, BM, BN, BO, BP, BQ, BR, BS, BT, BU, BV, BW, BX, BY, BZ. CA, CB, CC, CD, C E, CF, CG and CH;
Ixxxvi. A, B, C, D, E, F, G, I I, I, J, K, L, M, N, 0, P, O, R, S, T, U, W, X, Y, Z,
AA, AB, AC, AD, AE, AF, AG, AH, AI, AJ, AK, AL, AM, AN. AC), AP, AQ. AR, AS, AT, AU. AV, AW, AX, AY, AZ, BA, BB, BC, BD, BE, BF, BG, BH, BI, BJ, BK, BL, BM, BN. BO, BP, BQ, BR, BS, BT, BU, BV, BW, BX, BY, BZ, CA, CB, CC, CD, CE, CF, CG, CH and CI;
Ixxxvii. A, B, C, D, E, F, G, I I, I, J, K, L, M, N, 0, P, Q, R, S, T, U, W, X, Y, Z,
AA, AB, AC, AD, AE, AF, AG, AH, AI, AJ, AK, AL, AM, AN. AC), AP, AQ, AR, AS, AT, AU, AV, AW, AX, AY, AZ, BA, BB, BC, BD, BE, BF, BG, BH, BI, BJ, BK, BL, BM, BN, BO, BP, BQ, BR, BS, BT, BU, BV, BW, BX, BY, BZ, CA, CB, CC, CD, CE, CF, CG, CH, CI and CJ;
wherein each of A, B, C, D, E, F, G, II, I, J, K, L, M, N, O, P, Q, R, S, T, U, W, X, Y, Z, ΛΛ, AB, AC, AD, AE, AF, AG, AH, ΛΙ, AJ, AK, AL, AM, AN, ΛΟ, AP, AQ, AR, AS, AT, AU, AV, AW, AX, AY, AZ, BA, BB, BC, BD, BE, BF, BG, BH, BI, BJ, BK, BL, BM, BN, BO, BP, BQ, BR, BS, BT, BU, BV, BW, BX, BY, BZ, CA. CB, CC, CD, CE, CF. CG, CH, CI and CJ are independently selected from the genes listed in Table 1 and each of A, B, C, D, E, F, G, I I, I, J, K, L, M. N, O, P, Q, R, S, T, U, W, X, Y, Z, AA, AB, AC, AD, AE, AF, AG, Al l, AI, AJ, AK, AL, AM, AN, AO, AP, AQ, AR, AS, AT, AU, AV, AW, AX, AY, AZ, BA, BB, BC, BD, BE, BF, BG, BH, BI, BJ, BK, BL, BM, BN, BO, BP, BQ, BR, BS, BT, BU, BV, BW, BX, BY, BZ, CA, CB, CC, CD, CE, CF, CG, CH, CI and CJ are different. Methods of Using Biomarkcrs and Gene Signatures
[0060] The biomarkers and gene signatures of the invention may be used in methods of diagnosing, prognosing, classifying or grading lung cancer in a biological sample or an individual. The invention encompasses a method for clilSsi fv'n f' tpif camnlp a« cHi pp 1 Ιϊΐηο aH nnr!irf in ma ¾ta f> ? l i m p
adenocarcinoma, stage 1 squamous cell carcinoma or stage 2 squamous cell carcinoma, the method comprising: measuring the expression levels of at least 2 genes listed in Table 1 in a test sample; and applying one or more network-based methods, one or more machine-learning based methods, or a combination of the foregoing methods to the expression levels to obtain a classification of the test sample as stage 1 lung adenocarcinoma, stage 2 lung adenocarcinoma, stage 1 squamous cell carcinoma or stage 2 squamous cell carcinoma. In some
embodiments the expression levels of at least 2, at least 3, at least 4, at least 5, at least 10, at least 1 5, at least 20, at least 25, at least 30, at least 35, at least 40, at least 45, at least 50, at least 55, at least 60, at least 65, at least 70, at least 75, at least 80, least about 85 or at least 87 genes selected from the genes listed in Table 1 are measured. In some embodiments, a differential pattern of expression levels of said at least 2 genes in the test sample classifies the lung cancer tumor as one of stage 1 lung adenocarcinoma, stage 2 adenocarcinoma, stage 1 squamous cell carcinoma or stage 2 squamous cell carcinoma.
[0061] In one embodiment, the methods of the invention can be used to identify a gene signature and a classifier (e.g., a gene-signaturc-based classifier) that can distinguish datasets obtained from various classes and stages of lung cancer (e.g. stage 1 lung adenocarcinoma, stage 2 lung adenocarcinoma, stage 1 squamous cell carcinoma or stage 2 squamous cell carcinoma). In some methods of the invention, control data is not collected or used, instead, a classifier or a previously established standard may be used to determine whether a test sample is a lung cancer sample or the class and stage of lung cancer from which the test sample is obtained. For example, a classifier that is obtained by training with network-based or machine-learning methods using datasets obtained from subjects with various classes and stages of lung cancer (e.g. stage 1 lung adenocarcinoma, stage 2 lung adenocarcinoma, stage 1 squamous cell carcinoma or stage 2 squamous cell carcinoma) and datasets from subjects without lung cancer, can be used.
Alternatively, one or more numerical scores (e.g., average fold change or rank abs tval) generated by the algorithms described herein may be used as a previously established standard. The levels of expression of one or more of the genes listed in Table 1 in a test sample may be compared to the previously established standard, and the comparison may be used to classify the test sample as a lung cancer sample or a normal sample. In some embodiments, the comparison may be used to classify the test sample as stage 1 lung adenocarcinoma, stage 2 lung
adenocarcinoma, stage 1 squamous cell carcinoma or stage 2 squamous cell carcinoma.
[0062] In one embodiment, the invention provides a method of diagnosing, prognosing, classifying or grading lung cancer in a biological sample, wherein the method comprises determining the properties (for example, absence, presence or expression level) of one or more genes listed in Table 1 in the biological sample; and applying in silica analysis with a classifier obtained from a network-based method, a machine-learning based method, or a combination of the foregoing methods. The classifier can be obtained from the network-based method, a machine-learning based method, or a combination of the foregoing methods by training with datascts obtained from subjects with stage 1 lung adenocarcinoma, stage 2 lung adenocarcinoma, stage 1 squamous cell carcinoma stage 2 squamous cell carcinoma, healthy subjects, or a combination of two or more of the foregoing. In another embodiment, such a classifier may be linked to a specific prognosis of the lung cancer in the individual who provided the biological sample. In a further embodiment, the classifier may indicate that the lung cancer in the individual who provided the biological sample is stage 1 lung adenocarcinoma, stage 2 lung adenocarcinoma, stage 1 squamous ceil carcinoma or stage 2 squamous cell carcinoma. The classifier may also indicate that a particular treatment regimen should be used to treat the individual who provided the biological sample.
| ()()63 | In one embodiment, the methods of the invention comprise obtaining a test sample (such as a lung biopsy) from an individual, determining the absence, presence or expression level of one or more of the genes listed in Table 1 in the test sample, comparing said absence, presence or expression level to the absence, presence or expression level of the same gene(s) in a control sample, and selecting a lung cancer treatment regimen based on the comparison. In a further
embodiment, the invention provides a method for monitoring the progress of a lung cancer treatment in an individual, said method comprising determining at suitable time intervals before, during, or after therapy (for example, at different lime points during the treatment) in a sample taken from said individual differential expression of a panel of at least 2 genes selected from the genes listed in Table 1.
[0064] In one embodiment, the invention encompasses a method that comprises collecting data on the properties of one or more genes in the gene signature without generating a gene signature. For example, the method of the invention comprises obtaining a test sample from an individual, and detecting the absence, presence or the expression level of one or more of the genes listed in Table i in the sample. In one embodiment, the invention encompasses a method that comprises using data on the properties of one or more genes in a gene signature that are already collected as training data to generate an improved gene signature using one or more network-based methods, one or more machine learning methods, or a combination of the foregoing methods. In one embodiment, the invention encompasses a method that comprises collecting data on the properties of one or more genes in a biological system which is included in a gene signature, and using the data to predict a classification of the state of the biological system associated with the collected data.
[0065] In some embodiments, the method comprises detecting the expression level of at least 2 of the genes listed in Table 1 in a test sample obtained from the individual; and comparing the expression level of the genes listed in Table 1 in the test sample to the expression level of the genes listed in Table 1 in a control sample. In some embodiments, if the expression level of the genes listed in Table 1 is different in the test sample than in the control sample, then the individual suffers from lung cancer. In some embodiments, the lung cancer subtype and stage may be determined. In some embodiments, the lung cancer is classified as stage 1 lung adenocarcinoma, stage 2 lung adenocarcinoma, stage 1 squamous cell carcinoma or stage 2 squamous cell carcinoma. In some embodiments, the method further comprises detecting the expression level of the genes listed in Table 1 in the control sample. In some embodiments, the expression levels of at least 2, at least 3, at least 4, at least 5, at least 10, at least 15, at least 20, at least 25, at least 30, at least 35, at least 40, at least 45, at least 50, at least 55, at least 60, at least 65, at least 70, at least 75, at least 80, at least 84 genes or all genes listed in Table 1 are detected.
[0066] In some embodiments, the expression level of the genes listed in Tabic 1 in the test sample and the expression level of the genes listed in Table 1 in the control non-tumor biological sample are compared by in silico analysis. The in silica analysis may be network based analysis or a machine-learning method.
[0067] In some embodiments, the test sample is selected from blood, serum, plasma, sputum, saliva, tissue, bronchia brushings, exhaled breath, and urine. Optionally, the tissue is lung tissue, such as tissue obtained by biopsy from a tumor.
[0068] In some embodiments, the control sample is selected from blood, serum, plasma, sputum, saliva, tissue, bronchia brushings, exhaled breath, and urine. In some embodiments, the tissue is lung tissue, such as tissue obtained by biopsy from healthy lung tissue. In some embodiments, the healthy lung tissue is obtained from the individual at risk for or having lung cancer. In other embodiments, the control sample is obtained from an individual that does not have lung cancer.
[0069] In some embodiments, the expression level of the genes listed in Table 1 in the test sample and the expression level of the genes listed in Table 1 in the control sample are detected by measuring mRNA levels. For example, mRNA level is measured by amplification, hybridization, mass spectroscopy, serial analysis of gene expression, or massive parallel signature sequencing. Optionally, the amplification is reverse transcription PGR, real time quantitative PGR, dilTerential display or TaqMan PGR. In some embodiments, the hybridization is a dot blot, a slot blot, an RNasc protection assay, microarray hybridization, or in situ hybridization. The mass spectroscopy may be MALDI-TOF mass spectroscopy. In some embodiments, the expression level of the genes listed in Table 1 in the test sample arc detected by using a human genome-wide array, a human lung tissue array or a custom array comprising polynucleotides of a plurality of genes in Table 1 .
[0070] In some embodiments, the expression level of the genes listed in Table 1 in the test sample and the expression level of the genes listed in Table 1 in the control sample are detected by measuring the level of proteins encoded by the genes. Optionally, the protein level is measured using an antibody assay or by mass spectroscopy. In some embodiments, the antibody assay is selected from
Western analysis, immunofluorescence, ELISA, and immunohistochemistry. The mass spectroscopy may be MALDI-TOF mass spectroscopy or SELDI-TOF mass
[0071] In some embodiments, the expression level of the genes listed in Table 1 in the test sample and the expression level of the genes listed in Table 1 in the control sample are detected by measuring both mRNA levels and the level of proteins encoded by the genes. In some embodiments, expression levels are measured using the amplification, hybridization, mass spectroscopy, serial analysis of gene expression, massive parallel signature sequencing, and antibody assays discussed above. Mcthods of Biomarkcr Detection, Arrays and Panels
[0072] Detection of the nucleic acid and/or protein biomarkcrs described herein in a test sample or a control sample may be performed in a variety of ways.
[0073] In one aspect, the methods of the invention rely on the detection of the presence or absence of biomarkcr genes and/or biomarkcr gene expression, or the qualitative or quantitative assessment of either over- or under-cxpression of a biomarkcr gene in a population of cells in a test sample relative to a standard (for example, a control sample). Such methods utilize reagents such as biomarkcr polynucleotides and biomarkcr antibodies.
[0074] In particular, the presence, absence or level of expression of a biomarkcr gene may be determined by measuring the amount of biomarkcr messenger RNA (mRNA), for example, by DNA-DNA hybridization, RNA-DNA hybridization, reverse transcription-polymerase chain reaction (PGR), real time quantitative PGR, differential display or TaqMan PGR; followed by comparing the results to a reference based on a control sample (for example, samples from clinically- characterized patients and/or cell lines of a known genotypc/phenotype). In one embodiment, microRNA expression or turnover may be measured. Hybridization, mass spectroscopy (e.g., MALDI-TOF or SELDI-TOF mass spectroscopy), serial analysis of gene expression or massive parallel signature sequencing assays can also be performed. Non-limiting examples of hybridization assays include a singleplcx or a multiplexed aptamcr assay, a dot blot, a slot blot, an RNase protection assay, microarray hybridization, Southern or Northern hybridization analysis and in situ hybridization (e.g. , fluorescent in situ hybridization).
[0075] For example, these techniques find application in microarray-bascd assays that can be used to detect and quantify the amount of biomarkcr gene transcript using cDNA- or oligonucleotide-bascd arrays. Microarray technology allows multiple biomarkcr gene transcripts and/or samples from different subjects to be analyzed in one reaction. Typically, mRNA isolated from a sample is converted into labeled nucleic acids by reverse transcription and optionally in vitro transcription (cDNAs or cRNAs labelled with, for example, Cy3 or Cy5 dyes) and hybridized in parallel to probes present on an array. See, for example, Schulze et al, Nature Cell Biol. , 3 :E190 (2001 ); and Klein et al., J Exp Med, 194: 1625-1 638 (2001 ), which arc incorporated herein by reference in their entirety. Standard Northern analyses can be performed if a sufficient quantity of the test cells can be obtained. Utilizing such techniques, quantitative as well as size-related differences between biomarker transcripts can also be detected. In some embodiments, the expression level of the genes listed in Table 1 in the test sample are detected by using a human genome-wide array, a human lung tissue array or a custom array comprising polynucleotides of a plurality of genes in Table 1 .
[0076 j In some embodiments biomarkers are detected using reagents that specifically detect the biomarker. Such reagents may bind to a target gene or a target gene product (e.g., mRNA or protein), such that levels of the gene product may be quantified. Such reagents may be nucleic acid molecules that hybridize to the mRNA or cDNA of target gene products. Alternatively, the reagents may be molecules that label mRNA or cDNA for later detection, e.g., by binding to an array. The reagents may bind to proteins encoded by the genes of interest. For example, the reagent may be an antibody or a binding protein that specifically binds to a protein encoded by a target gene of interest. Alternatively, the reagent may label proteins for later detection, e.g., by binding to an antibody on a panel. In some embodiments, reagents are used in histology to detect histological and/or genetic changes in a sample.
[0077] The present invention provides isolated biomarker polynucleotides or variants thereof, which can be used, for example, as hybridization probes or primers ("biomarker probes" or "biomarker primers") to detect or amplify nucleic acids encoding a biomarker polypeptide, particularly a biomarker polypeptide encoded by a biomarker gene or polynucleotide selected from the group depicted in Table 1.
[0078] Nucleic acid molecules comprising nucleic acid sequences encoding the biomarker polypeptides or proteins of the invention, or genomic nucleic acid sequences from the biomarker genes (e.g., intron sequences, 5 ' and 3 ' untranslated sequences), or complements thereof (i .e. , anti sense polynucleotides), are collectively referred to as "biomarker genes," "biomarker polynucleotides" or
"biomarker nucleic acid sequences" of the invention. The present invention also provides isolated biomarker polynucleotides or variants thereof, which can be used, for example, as hybridization probes or primers ("biomarker probes" or "biomarker primers") to detect or amplify nucleic acids encoding a biomarker polypeptide of the invention. The term "biomarker gene product" thus
encompasses both mRNA as well as translated polypeptide as a gene product of a biomarker.
1 07 1 The isolated biomarker polynucleotide according to the invention may comprise flanking sequences (i. e. , sequences located at the 5 ' or 3 ' ends of the nucleic acid), which naturally flank the nucleic acid sequence in the genomic DNA of the organism from which the nucleic acid is derived. For example, in various embodiments, the isolated biomarker polynucleotide can comprise less than about 5 kb, 4 kb, 3 kb, 2 kb, 1 kb, 0.5 kb or 0.1 kb of nucleotide sequences which naturally flank the coding sequence in genomic DNA of the cell from which the nucleic acid is derived. In other embodiments, the isolated biomarker
polynucleotide is about 10-20, 21 -50, 51 - 100, 101 -200, 201 -400, 401 -750, 75 1 - 1000, or 1001 - 1500 bases in length.
[0080] In various embodiments, the biomarker polynucleotides of the invention are used as molecular probes in hybridization reactions or as molecular primers in nucleic acid extension reactions as described herein. In these instances, the biomarker polynucleotides may be referred to as biomarker probes and biomarker primers, respectively, and the biomarker polynucleotides present in a sample which are to be detected and/or quantified are referred to as target biomarker
polynucleotides. Two biomarker primers are commonly used in DNA
amplification reactions and they are referred to as biomarker forward primer and biomarker reverse primer depending on their 5 ' to 3 ' orientation relative to the direction of transcription.
[0081 1 In one embodiment, the invention encompasses methods of detecting genetic change in a biomarker gene (e.g., a mutation or a change in copy number). In another embodiment, the invention encompasses methods of detecting a change in the mcthylation of a biomarker gene.
[0082] A biomarker probe or a biomarker primer is typically an oligonucleotide which binds through complementary base pairing to a subsequence of a target biomarker polynucleotide. The biomarker probe may be, for example, a DNA fragmcnt prepared by amplification methods such as by PGR or it may be chemically synthesized. A double-stranded fragment may then be obtained, if desired, by annealing the chemically synthesized single strands together under appropriate conditions or by synthesizing the complementary strand using DNA polymerase with an appropriate primer. Where a specific nucleic acid sequence is given, it is understood that the complementary strand is also identified and included as the complementary strand will work equally well in situations where the target is a double stranded nucleic acid. Λ nucleic acid probe is
complementary to a target nucleic acid when it will anneal only to a single desired position on that target nucleic acid under proper annealing conditions which depend, for example, upon a probe's length, base composition, and the number of mismatches and their position on the probe, and must often be determined empirically. Such conditions can be determined by those of skill in the art.
[0083] In one aspect of the invention, biomarkers may be detected in the test sample or the control sample by gene expression profiling. In these methods, m.RNA is prepared from a sample and mRNA expression levels are measured by reverse transcription quantitative polymerase chain reaction (RT-PCR followed with qPCR). RT-PCR is used to create a cDNA from the corresponding mRNA. The cDNA may be used in a qPCR assay to produce fluorescence as the DNA amplification process progresses. By comparison to a standard curve, qPCR can produce an absolute measurement such as number of copies of mRNA per cell. Northern blots, microarrays, Invader assays, and RT-PCR combined with capillary electrophoresis may be used to measure expression levels of mRNA in a sample. Further details are provided, for example, in "Gene Expression Profiling: Methods and Protocols," Richard A. Shimkcts, editor, Humana Press, 2004 and US patent application 2010/0070191.
[0084] The invention encompasses an array comprising polynucleotides that hybridize to genes listed in Table 1. The array may comprise polynucleotides that hybridize to at least 2, at least 3, at least 4, at least 5, at least 10, at least 15, at least 20, at least 25, at least 30, at least 35, at least 40, at least 45, at least 50, at least 55, at least 60, at least 65, at least 70, at least 75, at least 80, least about 85, at least 87 or all genes listed in Table 1 . In one embodiment, the polynucleotides are immobilized on a solid surface. Examples of solid surfaces include paper, filler, nylon or other type of membrane, slide including glass slide, and chip (e.g., silicon, microarray chip). The polynucleotides may be single-stranded nucleic acid molecules (e.g., antisense oligonucleotides or fragments of cDNA). In some embodiments, the array is not a human genome-wide array. Examples of human genome-wide array include, but are not limited to, Exon 1 .0 ST, Gene 1.0 ST, U 95, U133 , U133A 2.0, and U 1 33 Plus 2.
[ 00851 In another aspect of the invention, detection of the biomarkcrs described herein may be accomplished by an immunoassay procedure. The immunoassay typically includes contacting a test sample with an antibody that specifically binds to or otherwise recognizes a biomarkcr, and detecting the presence of the antibody/biomarker complex in the sample. The immunoassay procedure may be selected from a wide variety of immunoassay procedures known to those skilled in the art such as, for example, competitive or non-competitive enzyme-based immunoassays, immunoprccipitation, enzyme-linked immunosorbent assays
(ELISA), radioimmunoassay (RIA), immunofluorescence, immunohistochemistry (II IC), cytological assays and Western blots. Further, multiplex assays may be used, including antibody panels or arrays, wherein several desired antibodies are placed on a support, such as a glass bead or plate, and reacted or otherwise
^υπια^ itu wi in liiC LuSl mple Or niC ujiu ui sam le.
[0086] Antibodies used in these assays may be monoclonal or polyclonal, and may be of any type such as IgG, IgM, IgA, IgD and IgE. Monoclonal antibodies may be used to bind to a specific epitope offered by the biomarkcr molecule, and therefore may provide a more specific and accurate result. Antibodies may be produced by immunizing animals such as rats, mice, rabbits and goats. The antigen used for immunization may be isolated from the samples or synthesized by recombinant protein technology. Methods of producing antibodies and of performing antibody-based assays are well-known to the skilled artisan and are described, for example, more thoroughly in Antibodies: A Laboratory Manual ( 1988) by Harlow & Lane; Immunoassays: A Practical Approach, Oxford
University Press, Gosling, J. P. (ed.) (2001 ) and/or Current Protocols in Molecular Biology (Ausubel et al.) which is regularly and periodically updated. [0087] In certain embodiments, the present invention also provides "biomarker antibodies" including polyclonal, monoclonal, or recombinant antibodies, and fragments and variants thereof, that immunospeci fically bind the respective biomarker proteins or polypeptides encoded by the genes or cDNAs (including polypeptides encoded by mRNA splice variants) as listed in Table 1.
[0088] Various chemical or biochemical derivatives of the antibodies or antibody fragments of the present invention can be produced using known methods. One type of derivative which is diagnostically useful as an immunoconjugate comprising an antibody molecule, or an antigen-binding fragment thereof, to which is conjugated a detectable label. However, in many embodiments, the biomarker antibody is not labeled but in the course of an assay, it becomes indirectly labeled by binding to or being bound by another molecule that is labeled. The invention encompasses molecular complexes comprising a biomarker antibody and a label, as well as immunocomplexes comprising a biomarker polypeptide, a biomarker antibody, and immunocomplexes comprising a biomarker polypeptide, a biomarker antibody, and a label.
[0089] Examples of detectable substances or detectable labels include various enzymes, prosthetic groups, fluorescent materials, luminescent materials, bioluminesccnt materials, and radioactive materials. Examples of suitable enzymes include horseradish peroxidase, alkaline phosphatase, bcta-galactosidase and acetylcholinesterase. Examples of suitable prosthetic group complexes include streptavidin/biotin and avidin/biotin. Examples of suitable fluorescent materials include umbeliifcroncs, fluoresceins, fluorescein isothiocyanate, rhodamincs, dichlorotriazinylamine fluorescein, dansyl chloride, phycoerythrins, Alexa Fluor 647, Alexa Fluor 680, DilC,9(3), Rhodamine Red-X, Alexa Fluor 660, Alexa Fluor 546, Texas Red, YOYO- 1 + DNA, tetramcthyl hodamine, Alexa Fluor 594.
BODIPY FL, Alexa Fluor 488, Fluorescein, BODIPY TR, BODIPY TMR, carboxy SNARF- 1 , FM 1 -43 , Fura-2, Indo- 1 , Cascade Blue, NBD. 13 API, Alexa Fluor 350, aminomethylcoumarin, Lucifer yellow. Propidium iodide, or dansylamide. An example of a luminescent material is luminol. Examples of bioluminesccnt materials include green fluorescent proteins, modified green fluorescent proteins, lucifcrase, iucifcrin, and acquorin. Examples of suitable
125 13 1 35 3
radioactive material include I, I, S or H.
[0090] Immunoassays for biomarker polypeptides will typically comprise incubating a sample, such as a biological fluid, a tissue extract, freshly harvested cells, or lysates of cells, in the presence of a detectably labeled antibody capable of identifying biomarker gene products or conserved variants or peptide fragments thereof, and detecting the bound antibody by any of a number of techniques well- known in the art. One way of measuring the level of biomarker polypeptide with a specific biomarker antibody of the present invention is by enzyme immunoassay (EIA) such as an enzyme-linked immunosorbent assay (ELISA) (Vollcr, A. et al , J, Clin. Pathol 57 :507-520 (1978); Butler, I.E., Melh. Enzymol. 75:482-523 (1981 ); Maggio, E. (ed.), Enzyme Immunoassay, CRC Press, Boca Raton, FL, 1980). The enzyme, either conjugated to the antibody or to a binding partner for the antibody, when later exposed to an appropriate substrate, will react with the substrate in such a manner as to produce a chemical moiety which can be detected, for example, by spectrophotometric, or fluorimetric means.
10091 ] The biological sample may be brought in contact with and immobilized onto a solid phase support or carrier such as nitrocellulose, or other solid support which is capable of immobilizing cells, cell particles or soluble proteins. The support may then be washed with suitable buffers followed by treatment with the detectably labeled biomarker antibody. The solid phase support may then be washed with the buffer a second time to remove unbound antibody. The amount of bound label on solid support may then be detected by conventional means. A well known example of such a technique is Western blotting.
[0092] In various embodiments, the present invention provides compositions comprising labelled biomarker polynucleotides, or labelled biomarker antibodies to the biomarker proteins or polypeptides, or labeled biomarker polynucleotides and labeled biomarker antibodies to the biomarker proteins or polypeptides according to the invention as described herein.
[0093] Antibodies and other reagents may also be used to detect post- translational modifications (e.g., mcfhylafion, acetylation, farnesylation, biotinylation, stcaroylation, formylation, myristoylation, palmitoylation, geranylgeranylation, pegylation, phosphorylation, sulphation, glycosylation, sugar modification, lipidation, lipid modification, ubiquitination, sumolation, disulphidc bonding, cystcinylation, oxidation, glutathionylation, carboxylation,
glucuronidation, and deamidation) of biomarker proteins or biomarker
polypeptides.
[0094] The invention encompasses a panel comprising antibodies that bind to proteins encoded by genes listed in Table 1 . The panel may comprise antibodies that bind to proteins encoded by at least 2, at least 3, at least 4, at least 5, at least 10, at least 15, at least 20, at least 25, at least 30, at least 35, at least 40, at least 45, at least 50, at least 55, at least 60, at least 65, at least 70, at least 75, at least 80, least about 85, at least 87 or all genes listed in Table 1 . In one embodiment, the panel of antibodies is immobilized on a solid surface. Examples of solid surfaces include microspheres, plates, wells, slides, and beads (e.g., protein A or protein G agarose).
[0095] In addition to antibody-based techniques, the biomarkers described herein may also be detected and quantified by mass spectrometry. Mass spectrometry is a method that employs a mass spectrometer to detect ionized protein markers or ionized peptides as digested from the protein markers by measuring the mass-to- charge ratio (m/z). Labelling of biomarkers (along with other proteins) with stable heavy isotopes (deuterium, carbon-13, nitrogen- 15, and oxygen- 18) can be used in quantitative proteomics. These are either incorporated metabolically in sample cells cultured briefly in vitro, or directly in samples by chemical or enzymatic reactions. Light and heavy labelled biomarker peptide ions segregate and their intensity values are used for quantification. For example, analytes may be introduced into an inlet system of the mass spectrometer and ionized in an ionization source, such as a laser, fast atom bombardment, plasma or other suitable ionization sources known to the art. The generated ions are typically collected by an ion optic assembly and introduced into mass analyzers for mass separation before their masses are measured by a detector. The detector then translates information obtained from the detected ions into mass-to-charge ratios.
[0096] The invention also encompasses methods that involve measuring the activity of a biomarker (e.g., enzymatic activity). Examples of enzymatic activity include, without limitation, kinase, phosphatase, protease, ubiquitination, oxidase and reductase activity.
[0097] The invention also provides compositions comprising biomarker polynucleotides, biomarker polypeptides, or biomarker antibodies according to the invention as described herein in the various embodiments. The invention further provides diagnostic or detection reagents for use in the methods of the invention, for example, reagents for flow cytometry and/or immunoassays that comprise fluorochrome-labeled antibodies that bind to one of the biomarker polypeptides of the invention.
[0098] In one embodiment, the invention provides diagnostic or detection reagents that comprise one or more biomarker probes, or one or more biomarker primers. Λ diagnostic reagent may comprise biomarker probes and/or biomarker primers from the same biomarker gene or from multiple biomarker genes. In another embodiment, the invention also provides diagnostic compositions that comprise one or more biomarker probes and target biomarker polynucleotides, or one or more biomarker primers and target polynucleotides, or biomarker primers, biomarker probes and biomarker target polynucleotides. In some embodiments, the diagnostic compositions comprise biomarker probes and/or biomarker primers and a sample suspected to comprise biomarker target polynucleotides. Such diagnostic compositions comprise biomarker probes and/or biomarker primers and the nucleic acid molecules (including RNA, mRNA, cRNA, cDNA, and/or genomic DNA) of a subject in need of a diagnosis/prognosis of lung cancer.
In silico Analysis and Computer Readable Media
1009 1 Biomarkers and gene signatures of the invention may be predicted based on gene expression patterns in lung cancer, including stages 1 and 2
adenocarcinoma and stages 1 and 2 squamous cell carcinoma. In some embodiments, biomarker and gene signature prediction comprises gene expression patterns in control (e.g., non-tumor) biological samples. A heterogeneous ensemble learning approach may be used to classify samples based on their gene expression profiles. Such an approach may combine predictions from different approaches that use genes, gene set-derived features and/or causal network-derived features in order to get a classification and a prediction confidence for each classificd sample. Methods that may be used to generate biomarkcrs and gene signatures of the invention include shrunken centroids, factor rotation, logistic regression models, network-based approaches, disease module-based approaches, linkage methods, modularity or pathway-based methods and diffusion-based methods.
[0100] The biological data (such as training data and test data) used in these methods may be drawn from the literature, databases (including data from preclinical, clinical and post-clinical trials of pharmaceutical products or medical devices), genome databases (genomic sequences and expression data, e.g.. Gene Expression Omnibus by National Center for Biotechnology Information or
ArrayExpress by European Bioinformatics Institute (Parkinson et al. 2010, Nucl. Acids Res., doi: 10.1093/nar/gkql 040. Pubmed ID 21071405)), commercially available databases (e.g., Gene Logic, Gaithersburg, MD, USA) or experimental work. In one embodiment, the REACTOME, KEGG or BIOCARTA pathway gene set collections from the Broad Institute (Cambridge, MA) may be used. The data may be related to nucleic acid {e.g. , absolute or relative quantities of specific DNA or RNA species, changes in DNA sequence, RNA sequence, changes in tertiary structure, or methyl ati on pattern as determined by sequencing,
hybridization - particularly to nucleic acids on microarray, quantitative polymerase chain reaction, or other techniques known in the art), protein/peptide {e.g. , absolute or relative quantities of protein, specific fragments of a protein, peptides, changes in secondary or tertiary structure, or posttranslational modifications as determined by methods known in the art) and functional activities (e.g., enzymatic activities, proteolytic activities, transcriptional regulatory activities, transport activities, binding affinities to certain binding partners) under certain conditions, among others. Modifications, including posttranslational modifications of protein or peptide, can include, but are not limited to, methylation, acetylation, farncsylation, biotinylation, stearoylation, formylation, myristoylation, palmitoylation, geranylgeranylation, pegylation, phosphorylation, sulphation, glycosylation, sugar modification, lipidation, lipid modification, ubiquitination, sumolation, di sulphide bonding, cysteinylation, oxidation, glutathionylation, carboxylation,
glucuronidation, and deamidation. In addition, a protein can be modified posttranslationally by a scries of reactions such as Amadori reactions, Schiff base reactions, and Maillard reactions resulting in glycated protein products.
[01 0 11 The test data sets may be processed and have their quality controlled separately, together if they are obtained from the same technology platform (e.g., an Affymctrix platform). For example, raw data files may be read by the
cadAffy function of the affy package (Gautier et al, Bioinformatics, 20:307-31 5 (2004)) belonging to Bioconductor (Gentleman et al., Genome Biol, 5(10):R80 (2004)) in R (R Development Core Team, R: A Language and Environment for Statistical Computing, 2007). The quality may be controlled by:
1 . Generating RNA degradation plots (using the AffyRNAdeg function of the affy package (Gautier, 2004)), NUSE and RLE plots (using the function affyPLM) (Brettschneider et al., Technomelrics, 50(3):241 -264 (2008)), calculating the MA(RLE) values;
2. Excluding arrays from the training datascts that fall below a set of thresholds on the quality control checks or that do not correspond to the test parameters;
or both.
Arrays passing quality control checks may be normalized using the gcrma algorithm (Wu et al., Journal of the American Statistical Association, 99:909 (2004)). If the datascts were obtained from a database, the samples classification may be obtained from the series matrix file of the same database for each datasct. The output of this part of the method may consist of: a gene expression matrix on training samples and test samples, probesets, and the class information for the training samples.
[0102] Non-limiting examples of methods that may be used to generate predictions are: transformation invariant (Tranlnv) (U.S. Provisional Patent Application entitled "Systems and Methods for Generating Biomarker Signatures with Integrated Bias Correction and Class Prediction," filed concurrently with the instant application and having the attorney docket no. 106500-0032-001 ), dual ensemble (Yang et al., Current Bioinformatics, 5(4):296-308 (2010)), generalized simulated annealing (Tsallis and Stariolo, Bhysica A: Statistical Mechanics and Its Applications, 233( l ):395-406 (1996); Xiang and Gong, Physical Review E, 62(3):4473 (2000); Xiang et al., Physics Letters A, 233(3):216-220 (1997); Xiang ct al, The Journal of Physical Chemistry A, 104(12):2746-275 1 (2000)), T-filter, CORG (Chuang et al., Mol Syst Biol, 3 : 140 (2007)), single and pairs, dual bagging, forward learning, NPA (network perturbation amplitude) (see, e.g., International Patent Application No. PCT/EP2012/061035, filed June 1 1 , 2012 and U.S.
Provisional Patent Application entitled "Systems and Methods Relating to
Network-Based Biomarker Signatures," filed concurrently with the instant application and having the attorney docket no. 106500-0022-001 ) and Laplacian based learning. Each of the foregoing patent applications and publications are incorporated herein by reference in their entirety.
[0103] Generalized simulated annealing may be modified for binary functions. In one embodiment, a dual binary generalized simulated annealing based method may be used (DualGcnsemble) (U.S. Provisional Patent Application entitled "Systems and Methods for Generating Biomarker Signatures with Integrated Dual Ensemble and Simulated Annealing Techniques," filed concurrently with the instant application, incorporated herein by reference in its entirety and having the attorney docket no. 106500-003 1 -001 ). 'Γ-filter is a method of filtering genes based on the t-test by setting P-valuc and fold-change thresholds. CORG may be modified by calculating activity scores by leveraging the F-test instead of the T- test. CORG may also be combined with SVM. Dual bagging is a combination of bagging (Breiman, The Elements of Statistical Learning: Data Mining, Inference, and Prediction. 2nd cd., ed. T. Hastie, R. Tibshirani, and J. Friedman, (2009)) and the random subspace method (Bryll, Pattern Recognition, 20(6): 1291 - 1302 (2003); Ho, IEEE Transactions on Pattern Analysis and Machine Intelligence, 20(8):832- 844 (1998); Skurichina, Pattern Analysis and Applications, 5(2): 121 - 135 (2002)).
[0104] The single and pairs method may include the following steps:
1 . Select a threshold ao,
2. For each pair of genes compute the leave-one-out cross-validation of a quadratic discriminant analysis and record the accuracy. Save the pair of genes if this accuracy is less than ao. 3. For each gene, compute the ao and 1 - ao quantile of the values in each class to discriminate the classes Λ and B, Qao(A), Q i -ao(A), Qao(B), Q \. a0(B). Select the gene if cither: Qa0(A) > Q i-a0(B) or Qi.a0(A) < Qa0(B).
4. Use the obtained list to train a classi fication algorithm on the reduced feature space.
5. Choose ao by cross-validation.
[0105] The forward learning method may include the following steps:
Set IN to the empty list, choose N (for example, 20, 100 or 200).
For n = 1 , . . . N do
a. For each gene, g, not in IN, compute a randomForcst (ntrce=500) on the subspacc corresponding to {IN,g} , record the out-of-the bag true positive rates (TPr) and true negative rates (TNr) and compute the g- pcrformance VTPr * TNr.
b. Select the gene, gmax> for which the g-performance is maximum and add it to the list IN:= {IN,gmax} .
c. Then train a classification algorithm on the sub space given by IN. N is chosen by cross-validation.
[01 06] The Laplacian based learning method may include the following steps:
1 . Compute Spearman or Pearson correlation between the samples based on their gene expression profiles lor both test and training data.
Normalize the distance matrix obtained from the correlation matrix
(Kij =Kij /sqrt(Kii * Kjj )) .
2. Compute the k-ncarest neighbors of each sample (k chosen by cross-validation, usually k=2,3,4,5)).
3. Define a graph with samples as nodes and put an edge between neighbors.
4. Create the (combinatorial) Laplacian of the graph and get its generalized inverse G, which is a positive definite kernel.
5. Extract main kerne! principal components (KPC) from G and train a Svm on it. The number of KPC's is chosen with rdetools package function rde (Braun et al., The Journal of Machine Learning Research, 9: 1875-1908 (2008)). 6. Train a SVM with the training samples and get performance by cross-validation.
7. Predict the test cases.
[0107] Network-based analysis can be combined with machine learning methods to generate predictions, for example, combining any one of CORG, dual bagging or T-filter with a network-based analysis.
[0108] In some embodiments, methods used to generate predictions are further combined with another classification method (e.g., a method that is used for cross- validation). Non-limiting examples of classification methods include PAMR (Tibshirani et ah, Proc Natl Acad Set USA, 99(10):6567-6572 (2002)),
RandomForest (Breiman, Machine Learning, 45( l ):5-32 (2001 )), Linear
Discrimination Analysis (LDA), Eiigengene-based Linear Discrimination Analysis (ELDA), Principal Components Analysis (PCA), Recursive Partitioning Tree (RPART), Support Vector Machine (SVM), K-Nearest Neighbors (KNN) (Bishop, Neural Networks for Pattern Recognition, ed. O.U. Press, 1995) and Partial Least Squares Discriminant Analysis (PLS.DA). In one embodiment, a network-based analysis that uses NPA may be combined with SVM (U.S. Provisional Patent Application entitled "Systems and Methods relating to Network-based Biomarkcr Signatures," filed concurrently with the instant application, incorporated herein by reference in its entirety and having the attorney docket no. 106500-0022-001 ).
[0109] In one embodiment, these methods may further include a step of ovcrsampling to balance classes. The methods may include a step of filtering genes based on a simple T-test between the categories to be classified. The filtering step may reduce the number of genes to less than 1 ,500 or less than 2,000.
[0110] After predictions are generated by several methods, a vote may be made to obtain the classi fication as well as the confidence for the prediction of each sample of the sample set. If a method provides cross-validation results far below the other methods, it may be excluded. Such additional steps are contemplated in the methods of the invention.
[ 01 1 1 ] The union of the gene signatures extracted by these methods may be considered as the larger gene signature. A weight may be given to genes to take into consideration the number of times they appear in a list. See, for example, in Table 1 the column "present/total lists" which shows the number of times each gene appears in one of the predicted gene signatures.
[0112] The genes obtained by these methods may be mapped to gene symbols using any suitable platform, for example, the Confero platform (Hermida et al., Confero: an Integrated Contrast and Gene Set Platform for Computational Analysis and Biological Interpretation ofOmics Data, submitted, 2012).
| 01 13] The numerical methods for generating the gene signatures of the invention may include a testing step and confidence statistics for the genes. The testing step (or phase) is an exemplary use of the gene signature in carrying out the claimed method.
[01 14 | The invention encompasses a method for classifying a test sample as stage 1 lung adenocarcinoma, stage 2 lung adenocarcinoma, stage 1 squamous cell carcinoma or stage 2 squamous cell carcinoma, the method comprising: measuring the expression levels of at least 2 genes listed in Table 1 in a test sample; and applying one or more network-based methods, one or more machine-learning based methods, or a combination of the foregoing methods to the expression levels to obtain a classification of the test sample as stage 1 lung adenocarcinoma, stage 2 lung adenocarcinoma, stage 1 squamous cell carcinoma or stage 2 squamous cell carcinoma. In some embodiments, the expression levels of at least 2, at least 3, at least 4, at least 5, at least 10, at least 1 5, at least 20, at least 25, at least 30, at least 35, at least 40, at least 45, at least 50, at least 55, at least 60, at least 65, at least 70, at least 75, at least 80, least about 85, at least 87 or all genes listed in Table 1 are measured. In some embodiments, the classifier has been trained by in silico analysis or one or more feature selection and classification algorithms.
[0115] One aspect of the invention encompasses a list of one or more biomarkcrs or gene signatures of the invention stored on a computer readable medium. The absence, presence, activity or expression level of a biomarkcr in a biological sample (such as a control sample or a test sample) may also be stored on the computer readable medium. The computer readable medium may also include information that identifies the sample. The computer readable medium may also include a computer program product. [0116] The computer program product may include a classifier based on at least two genes listed in Table 1 . The classifier may be based on at least 2, at least 3, at least 4, at least 5, at least 10, at least 15, at least 20, at least 25, at least 30, at least 35, at least 40, at least 45, at least 50, at least 55, at least 60, at least 65, at least 70, at least 75, at least 80, least about 85, at least 87 or all genes listed in Table 1 .
[0117] Optionally, the classifier is trained by in silica analysis or one or more feature selection and classification algorithms. In some embodiments, the classifier is trained by one or more algorithms selected from the group consisting of dual ensemble, generalized simulated annealing, T- filter, CORG, CORG combined with support vector machine, dual bagging, single and pairs, forward learning, Laplacian based learning and learning method based on network perturbation amplitude. The classi fier may be trained with at least the data in Gene Expression Omnibus datasets GSE2109, GS El 0245, GSE18842 and GSE37745.
Devices and Kits [0118] One aspect of the invention encompasses devices useful for performing methods o f the invention. For example, the devices may be used for diagnosing, classifying and/or grading lung cancer. The devices can comprise means for detecting the expression level of at least 2 of the genes listed in Table 1 or the level of at least 2 gene products of such genes in a test sample. Such means may include components for performing one or more methods of nucleic acid extraction, nucleic acid amplification, nucleic acid detection, protein isolation and/or protein detection. Such components may include one or more of an ampli fication chamber (for example a thermal cycler), a plate reader, robotic sample handling
components, a capillary electrophoresis apparatus, a spectrophotometer, a mass spectrometer and/or a chip reader. These components can obtain data that reflects the expression level of the genes being analyzed. In some embodiments, the devices can comprise means for detecting at least 2, at icast 3, at least 4, at least 5, at least 10, at least 15, at least 20, at least 25, at least 30, at least 35, at least 40, at least 45, at least 50, at least 55, at least 60, at least 65, at least 70, at least 75, at least 80, at Icast 84 or all genes listed in Table 1. In some embodiments, the devices can comprise means for detecting the gene products of at least 2, at least 3, at least 4, at least 5, at least 1 0, at least 15, at least 20, at least 25, at least 30, at least 35, at least 40, at least 45, at least 50, at least 55, at least 60, at least 65, at least 70, at least 75, at least 80, at least 84 or all genes listed in Table 1.
[0119] The devices optionally comprise a means for identifying a given test sample, and of linking the results obtained to that sample. Such means can include manual labels, barcodes, and other indicators which can be linked to a sample container or receptacle. Identification means may optionally be included in the sample itself, for example where an encoded particle is added to the sample. The results may be linked to the sample, for example in a computer memory that contains a sample designation and a record of expression levels obtained from the sample. Linkage of the results to the sample can also include a linkage to a particular sample container or receptacle in the device, which is also linked to the sample identity.
[0120] The devices may comprise an excitation and/or a detection means. Any instrument that provides a wavelength that can activate a label (e.g., fluorophore, fluorochrome and fluorescent dye) used on a detection reagent and is shorter than the emission wavelcngth(s) to be detected can be used for excitation. Examples of excitation sources include a broadband ultraviolet light source such as a deuterium lamp with an appropriate filter, the output of a white light source such as a xenon lamp or a deuterium lamp after passing through a monochromator to extract out the desired wavelength(s), a continuous wave (cw) gas laser, a solid state diode laser, or any pulsed lasers. Emitted light can be detected through any suitable component or technique; many suitable approaches are known in the art. For example, a fluorimcter or spectrophotometer may be used to detect whether the test sample emits light of a wavelength characteristic of a label used in a method of the invention.
[0121 ] The devices may comprise a means for correlating the expression levels of the genes being analyzed with an lung cancer status, prognosis, grade and/or classification. Such means may comprise one or more of a variety of correlative techniques, including lookup tables, algorithms, multivariate models, and linear or nonlinear combinations of expression models or algorithms, such as any of the in silico and machine learning methods described above. The expression levels may¬ be converted to one or more biomarker scores, indicating that the individual providing the sample is not suffering from adenocarcinoma or squamous cell carcinoma or is suffering from stage 1 adenocarcinoma, stage 2 adenocarcinoma, stage 1 squamous cell carcinoma or stage 2 squamous cell carcinoma. The models and/or algorithms can be provided in computer readable format.
[ 122] The devices may also comprise output means for outputting the lung cancer status, prognosis, grade and/or classification. Such output means can take any form which transmits the results to an individual and/or a healthcare provider, and may include a monitor, a display, and/or a printer. Output means may record the results to a computer readable medium. The device may use a computer system for performing one or more of the steps provided.
[0123] In one embodiment, a device of the invention comprises means for detecting the expression level of at least 2 genes listed in Table 1 in a test sample; means for correlating the expression level with a classification of the lung cancer as stage 1 adenocarcinoma, stage 2 adenocarcinoma, stage 1 squamous cell carcinoma or stage 2 squamous cell carcinoma; and means for outputting the lung cancer classification. In some embodiments, the device comprises means for detecting the expression level of at least 2, at least 3, at least 4, at least 5, at least 1 0, at least 15, at least 20, at least 25, at least 30, at least 35, at least 40. at least 45, at least 50, at least 55, at least 60, at least 65, at least 70, at least 75, at least 80, at least 84 genes or ail genes listed in Table 1 .
[0124] Another aspect of the invention encompasses kits for practicing the methods of the invention. Such kits may be used for classifying and grading lung cancer or for assessing the prognosis of lung cancer in an individual. The kits can be used for clinical diagnosis and/or laboratory research. In one embodiment, a kit comprises in one or more containers one or more reagents that detect expression levels of genes that serve as biomarkcrs of lung cancer in a test sample. Preferably, the kit also comprises instructions in any tangible medium (e.g., written, tape, CD- ROM, DVD) on the use of the detection reagent(s) in one or more methods of the invention.
[0125] For nucleic acid-based methods (for example, amplification assays, hybridization assays, sequencing or polymerase chain reactions), a detection reagent in the kit may comprise at least one polynucleotide, probe, and/or primer specific for the stage 1 adenocarcinoma, stage 2 adenocarcinoma, stage 1 squamous cell carcinoma and/or stage 2 squamous cell carcinoma the genes listed in 'fable 1 . The nucleic-acid based detection reagents may comprise sequences complementary to a portion of the signature genes or sequences that are portions of the signature genes. Such a kit may optionally provide in separate containers enzymes and/or buffers for reverse transcription, in vitro transcription, and/or 1 )\ \ polymerization, nucleotides, and/or labeled nucleotides.
[0126] For protein-based methods, such as immunoassays, a detection reagent in the kit may comprise a biomarkcr antibody, which may be labeled or labelable. The antibodies may bind to proteins encoded by stage 1 adenocarcinoma, stage 2 adenocarcinoma, stage 1 squamous cell carcinoma and/or stage 2 squamous cell carcinoma the genes listed in Table 1 . Tn one embodiment, the detection reagents recognize a post-translational modification (e.g., mcthylation, accty!ation, farncsylation, biotinylation, stcaroylation, formylation, myristoylation,
palmitoylation, geranylgeranylation, pegylation, phosphorylation, sulphation, glycosylation, sugar modi fication, lipidation, lipid modification, ubiquitination, sumolation, disulphide bonding, cysteinylation, oxidation, glutathionylation, carboxylation, glucuronidation, and deamidation) of a protein encoded by a gene selected from the genes listed in Table 1. For protein-based methods that involve measuring the activity of a biomarkcr (e.g., enzymatic activity), the kit may include a substrate for the biomarkcr and a detection reagent that recognizes the products and/or byproducts of the activity being measured. Such a kit may optionally provide, in separate containers, buffers, secondary antibodies, signal generating accessory molecules, and/or labeled secondary antibodies, including fiuorochrome-labcled secondary antibodies. The kit may also include unlabeled or labeled antibodies to various cell surface antigens which can used for identification or sorting of subpopulations of cells.
[0127] The detection reagents may be labeled or labelable. by one or more detectable labels. Examples of detectable labels include, without limitation, radiolabcls (e.g. radioactive nuclides), dyes, fluorescent proteins or materials (e.g., fluorochromes, fiuorophores, fluorescein and rhodamine), luminescent proteins or materials, bioluminescent proteins or materials (e.g., luciferase, aequorin and luciferin), enzymes (e.g., beta-galactosidase, alkaline phosphatase, horseradish peroxidase and acetylcholinesterase) and prosthetic groups (e.g., biotin, streptavidin and avidin).
[0128] The detection reagents in the kit may be immobilized on a solid surface or packaged separately with reagents to immobilize them on a solid surface.
[012 1 Also included in the kit may be positive and negative controls for the methods of the invention. The positive and/or negative controls included in a kit can be nucleic acids, polypeptides, cell lysate, ceil extract, whole ceils from patients, or whole cells from cell lines. Example 1 Generation of Lung Cancer Gene Signatures
[0130] This example is for the purpose of illustration only and is not to be construed as limiting the scope of the invention in any way. A heterogeneous ensemble learning approach aimed at classifying samples based on their gene expression profile is applied to extract genes whose expression levels allow determination of adenocarcinoma from squamous cell carcinoma lung biopsies and determination of stage 1 from stage 2 for each type of carcinoma. In summary, predictions from di fferent approaches that use genes and gene sets-derived features are combined to get the most accurate classifier possible. Gene lists are extracted from these methods and combined to generate the list presented in Table 1 .
[0131 ] A schematic overview of the data and strategy used to generate the list given in Table 1 is given below:
1. Public datasets and a dataset downloaded from the SBV diagnostic signature challenge (http://sbvimprovcr.com) are used as the source of gene expression data.
The following public datasets are downloaded from the Gene Expression Omnibus (GEO) (http://www.ncbi. nlm.nih.gov/geo/) repository:
•GSE2109 (www .ncbi .nl m.nih . gov/gco/ qucry/acc. cgi ?acc=G SE2109)
•GS 1 0245 (www.ncbi.nlm.nih. gov/geo/query/acc.cgi?acc=GSE 10245) •GSE1 8842 ( www.ncbi.nlm.nih. gov/geo/query/acc.cgi?acc=GSEl 8842) «GSE37745 (www.ncbi.nlm.nih. go v/geo/query/acc.cgi?acc=GSE37745) 2. As both training datasets are on the same Affymctrix platform as the test datasct (HGU-133 + 2), they are processed and have their quality controlled together. In summary, raw data files are read by the Read Affy function of the affy package (Gauticr, 2004) belonging to Bioconductor (Gentleman, 2004) in R (R Development Core Team, 2007), and the quality is controlled by:
a. generating RNA degradation plots (with the AffyRNAdeg function of the affy package), NUSE and RLE plots (with the function affyPLM
(Brettschncider, 2008)), and calculating the MA(RLE) values;
b. excluding arrays from the training datasets that fell below a set of thresholds on the quality control checks or that are duplicated in the above datasets; and
c. normalizing arrays that pass quality control checks using the gcrma algorithm (Wu, 2004). Training set sample classifications are obtained from the series matrix file of the GEO database for each datasct.
Arrays GSM926687_PA025_203B_081029 J IG-U 133 _Plus_2 _.CEL,
GSM926694J>A025_224AJ)81028JlG-U133_Plus_2_.CEL,
GSM926700_PA025_260A_081029 _HG-U 133_Plus_2_.CEL,
GSM926706_PA025_296A _08121 1_HG-U133_Plus_2_.CEL,
GSM926712 PA025 344A 081 106 HG-U 1 33_Plus_2_.CEL,
GSM926717 _P A025_98A_081029_HG-U 133_Plus_2_.CEL,
GSM926718_PA025_human_l 12B_070912.CEL,
GSM926721 _PA025 _human_l 31 A_070228.CEL,
GSM926728_PA025Jniman_175B_070228.CEL,
GS 926747_PA025_human_295B_070912.CEL,
GSM926758 _PA025_human_33 1 A_070914.CEL,
GSM926761 _ ΡΛ025 Jiuman _342A_070907.CEL,
GSM926762_PA025_human_347AJ)70907.CEL,
GSM926770_PA025_human_37A_070228.CEL,
GSM926776_PA025_human_88A_070228.CEL?
GSM926787_PA 1 17_1 30A_10 l 126_I-IG-U 133_Plus_2_.CEL,
GSM926801 PA 1 1 7 173 B_ 101 130J-IG-U 1 33_Plus_2_. CEL,
GSM926826 PA1 17 285A 101210 HG-U 133 Plus 2 .CEL, GSM926840 J5A 1 17_336Λ_101202 J IG-U 133_Plus_2_.CEL,
GSM926843_PA1 17_353 A J 01208 JHG-U 133 _Plus_2_.CEL,
GSM926854 PA 1 1 7_58A_1 01 123_HG-U 133_Plus_2_.CEL,
GSM926856_PA 1 17_5A_101 1 18_HG-U133_Plus_2_.CEL,
GSM926866 _PA1 17_7A_101 1 18_HG-U 133_Plus_2_.CEL,
GS 926872_PA 1 17_96A_101 126J IG-U 133_Plus_2_.CEL,
GSM102451 .CEL, GSM 102455.CEL, GSM138003.CEL, GSM203732.CEL, GSM23 1874.CEL, GSM467024.CEL, GSM76585.CEL, GSM76595.CEL, GSM88997.CEL, GSM926766_PA025_human_365A_070907.CEL, GSM926739_PA025_human_256A_070926.CEL,
GSM926738_PA025_human_255A_070926.CEL, GSM89060.CEL, GSM76590.CEL, GSM76587.CEL, GSM53 170.CEL, GSM53 167.CEL, GSM46817.CEL, GSM467030.CEL, GSM467029.CEL, GSM466975.CEL, GSM466952.CEL, GSM38104.CEL, GSM38103.CEE, GSM203641 .GEL, GSM 152757.CEL, GSM 152681.CEL, GSM 152670.CEL, GSM 152624.CEL,
GSM138002.CEL, GSM 137945. GEL, GSM13793 1.CEL, GSM l 37916.CEL, GSM 137910.CEL, GSM l 17763. CEL, GSMl 17632.CEL, GSM 3 17610.CEL, GSM 102555.CEL, GSM 102553. CEL, GSMl 02548. CEL., GS l 025 12, CE L, GSM l 02507. CEL, GSM 102447.CEL, GSMl 17763. CEL, GSM 1 52757.CEL, GSM258570.CEL, GSM258591.CEL. GSM258594.CEL. GSM258597.CEL,
GSM258601.CEL, GSM258603.CEL, GSM258606.CEL, GSM466973.CEL, GSM466982.CEL, GSM467029.CEL, GSM46833.CEL, GSM53 167.CEL, GSM53 170. CEL, GSM926683_PA025_1 9A_081028_HG- U 133_PIus_2_.CEL, GSM926692JPA025_217A_081029J IG- U 133_Plus_2_.CEL, GSM926703_PA025_282A_08 1029 J IG-
U 133_Plus _2_.CEL, GSM926707 _PA025_301 A_081 029 _HG- U 133_Plus_2_.CEL, GSM926708_PA025 _302A_08 1 1 06J 1G- U 133_Plus_2_.CEL, GSM926733_PA025_hiunan_207A_070912. CEL ,GSM926734_PA025_human_229A_070912. CEL,
GSM926735_PA025_human_234A_070912. CEL.
GSM926741J5A025Jiuman_265AJ)71005.CEL,
GSM926745_PA025_human_284A_070912.CEL, GSM926750_ _PA025_ human_300A_070919.CEL,
GS 926755_ _ΡΛ025_ human J 17A_070926.CEL,
GSM926760_ _ΡΛ025_ _human_33A_070509.CEL,
GS 926769_ _ΡΛ025_ _human_373B_070907.CEL,
GSM926772_ _PA025_ _human_48A_070509.CEL5
GSM926784_ _PA1 17_ J 16A 101 126_HG-U133_Plus 2_ Λ C i ..
GSM926786_ _ΡΛ1 17_ 128 A J 01 126J IG-U 1 33_Plus 2_ .CEL,
GSM926798 _ΡΛ1 17_ _164A_101210_HG-U 133_Plus .CEL,
GSM926803_ _ΡΛ1 17_ _ 179A_101 130_HG-U 133_Plus 2_ .CEL,
GSM926809 PA1 17 190B J 01202JIG-U133 _Plus 2__ .CEL,
GSM926818_ _ΡΛ1 17_ 241 B_l 01203 _HG-U133_Plus 2 .CEL,
GSM926836 _ΡΛ1 17_ _321A_101203 J IG-U 1 3 ^ Plus 2^ .CEL,
GSM926861 ΡΑ1 17_ 74 A 01 215J 1G-U 133_PlusJ 2 .( EL,
GSM926862_ _ΡΑ1 17_ 75A 101 126 J IG-U 133_Plus_: : :..ci L.
GSM102512.CEL, GSM138002.CEL, GSM467032.CEL,
GSM926693J5A025„223A_081028JIG-U133J¾s_2_.CEL,
GSM926756_PA025_human_319A^070914.CEL,
GSM926849_PA 1 17_41 A_ 1 01 123 I IG-U 133 _Plus_2_.CEL,
GSM926851 PA 1 17_52A_101 123 I IG-U 133_Plus_2_.CEL,
GSM102548.CEL, GSM 102553.CEL, GSM 1 17610.CEL, GSM38103.CEL, GSM46868.CEL, GSM46936.CEL, GSM46941.CEL, GSM76587.CEL, GSM76590.CEL, GSM88962.CEL were not used for further analysis.
The output at this point comprises a gene expression matrix X on 410 samples (260 training samples and 150 test samples) and 54675 probescts, and the class information for the training samples.
3. Features selection and classification algorithm(s) used for prediction of a gene signature follow the illustrated strategy of Figure 1 :
Briefly, a set of feature selection and classification algorithms arc used to obtain a number of classifications for each test sample. Each method has defined input and output:
INPUT: gene expression matrix Xnxp on n samples and p genes, training samples and test samples, and the class information for the training samples OUTPUT: Class prediction for each test sample and a list of genes involved.
[0132] Prior to applying features selection and classification methods, the following steps are performed: (1) oversampling is, optionally, used to balance classes in the training dataset; (2) mapping probe sets to gene symbols (Entrez gene ids) using Confero platform (Hcrmida, 2012); and (3) optionally filtering the genes in the matrix based on simple T-test between the categories to be classified so that less than 1500 genes (for Dual Ensemble or T-fi!ter methods) or less than 2000 genes (for the other methods) remain.
[0133] Cross validation of the results is performed using any of the following supervised methods:
PAMR (Tibshirani, 2002)
RandomForest (Breiman, 2001 )
Linear Discriminant Analysis (LDA)
Support Vector Machine (SVM)
K-Nearest Neighbors (KNN) (Bishop, 1995)
Partial Least Squares Discriminant Analysis (PLS.DA)
[0134] The following methods are used to generate predictions:
a. Dual Ensemble method.
This dual ensemble method builds ensemble of multiple classification algorithms applied in randomly perturbed data. The diversity of the ensemble classifier is imposed by using different classification algorithms and is further enhanced by data-level perturbation. See, e.g., Yang, 2010. A molecular profile of a training dataset, TO.train and its associated phenotypc cl. train (control and treatment) are used as input. The molecular profile of the test set TO. test is used to predict the phenotypc ci.test.
b. Ί -filter method
Genes are filtered based on t-tcst to obtain a list of N genes, by setting P- value and fold-change thresholds. Thresholds are decreased (resp. increase) automatically if the list size is over N. Any M is trained on the resulting subspace. N is chosen by cross-validation.
e. CORG-modified method This method is modified from CORG method (Chuang, 2007) as activity- scores arc calculated by leveraging F-test instead of T-test. It uses the c2.cp gene sets collection from the Broad Institute (Cambridge, ΜΛ) (Reactome, KEGG and Biocarta pathways).
d. Single and Pairs method
i m- u Ϊiwu,
1 . A threshold (a0) is selected.
2. For each pair of genes the leave-one-out cross-validation of a quadratic discriminant analysis is computed and the accuracy recorded. If this accuracy is less than ao, the pair of genes is saved.
3. For each gene, the ao and 1 - a0 quantile of the values in each class to discriminating the classes A and B, Qao(A), Qi-ao(A), Qa0(B), Q i-ao(B), is computed. If cither: Qa()(A)> Q i -a0(B) or Q l_a0(A)< Qa0(B), the gene is selected.
4. The obtained list is used to train M on the reduced feature space.
5. ao is chosen by cross-validation.
e. Forward Learning method
In this method,
1. Set IN to the empty list, choose N (typically 20, 100, 200).
i i. i i i , . . 1 1 v
a. For each gene, g, not in IN, compute a randomForcst (ntree=500) on the subspace corresponding to {IN,g} , record the out-of-the bag true positive rates (TPr) and true negative rates (TNr) and compute the g- pcrformancc VTPr * TNr.
b. Select the gene, gmax, for which the g-performance is maximum and add it to the list IN:={IN,gmax} .
c. Then train M on the subspace given by IN. N is chosen by cross- validation.
[01 35] The union of the gene signatures extracted from the results of the foregoing methods is considered as the larger gene signature. A weight is given to each of the genes in the union of the gene signature on the basis of the number of times it appears in a generated signature and the length of each generated gene signaturc. Genes in Table 1 are those that appear in at least 3 of 5 lists generated. The genes in Fable 1 that appear in at least 4 of the 5 lists generated are more predictive of lung cancer status than those appearing in 3 of the 5 lists generated.
[0136] While implementations of the invention have been particularly shown and described with reference to specific examples, it should be understood by those skilled in the art that various changes in form and detail may be made therein without departing from the spirit and scope of the disclosure.
Table 1 : present /
Gene Symbol ENTREZ G FN E J D Gene Title
total lists
Zic family member 2 (odd-paired homolog,
ZIC2 7546 4/5
Drosophila)
LOC100131262 100131262 hypothetical LOCI 00131262 4/5
CD83 9308 CD83 molecule 4/5 echinodcrm microtubule associated protein
EMLI 2009 4/5 like 1
poly(A) binding protein interacting protein
PAIP1 10605 4/5
1
NIPBL 25836 Nipped-B homolog (Drosophila) 4/5 cAMP responsive element binding protein
CREB3L1 90993 4/5
3-iikc 1
solute carrier family 37 (glycerol-3-
SLC37A1 54020 4/5 phosphate transporter), member 1
SFMBT2 57713 Scm-Iikc with four mbt domains 2 4/5 aldehyde dehydrogenase 3 family, member
ALDII3B1 221 3/5
Bl
wingless-type MMTV integration site
WNTS Λ 7474 3/5 family, member 5Λ
u Ccilmodulin-likc 3 3/5
SLC44A4 80736 solute carrier family 44, member 4 3/5
USPS 9101 u iquitin specific peptidase 8 3/5
SLC41A2 84102 solute carrier family 41, member 2 3/5
CTSH 1512 cathepsin II 3/5
CSTA 1475 cystatin A (stcfin A) 3/5
HNF!B 6928 HNF1 homeobox B 3/5
DSC3 1825 desmocollin 3 3/5
C180RF! 753 chromosome 18 open reading frame 1 3/5
C8ORF40 114926 chromosome 8 open reading frame 40 3/5
SLC25A40 55972 solute carrier family 25, member 40 3/5 protein tyrosine phosphatase-Iike (proline
PTPLB 201562
instead of catalytic arginine), member b
RORC 6097 R AR-related orphan receptor C 3/5
DENND2C 163259 DENN/MADD domain containing 2C 3/5
US PI 7398 ubiquitin specific peptidase 1 3/5
FMOD 2331 fibromodulin 3/5
PBJ murine osteosarcoma viral oncogene
FOSB 2354 3/5 homolog B
PRAF2 11230 PRAl domain family, member 2 3/5
IL1R2 7850 interleukin 1 receptor, type 11 3/5 immediate early response 3 interacting
IER3IP1 51124 3/5 protein 1
NSUN5 55695 NOP2/Sun domain family, member 5 3/5
DOK5 55816 docking protein 5 3/5 present /
Gene Symbol ENTREZ GENE ID Gene Title
total lists
ECHDC2 55268 enoyl CoA hydratase domain containing 2 3/5
FKBP 1 1 5 1303 FK506 binding protein 1 1 , 19 kDa 3/5 tRNA-histidine guanylyltransfcrase 1 -like
THG 1 L 54974 3/5
(S. ccrevisiae)
AVEN 57099 apoptosis, caspase activation inhibitor 3/5
8645 potassium channel, subfamily K, member 5 3/5
C90RF 167 54863 chromosome 9 open reading frame 1 67 3/5
TMEM 1 17 84216 transmembrane protein 1 17 3/5
LIFR 3977 leukemia inhibitory factor receptor alpha 3/5 myeloid/lymphoid or mixed-lineage
MLL5 55904 3/5 leukemia 5 (trithorax homolog, Drosophila)
membrane associated guanylate kinase,
MAGI3 260425 3/5
WW and PDZ domain containing 3
SI E 1 80143 suppressor of I BKE 1 3/5
LOC I 005071 53 100507153 hypothetical LOC I 005071 53 3/5 deleted in lymphocytic leukemia 2 (non¬
DLEU2 8847 3/5 protein coding)
postmeiotic segregation increased 2-like 5-
LOCI 001 32832 100132832 3/5 like
CCND2 894 cyclin D2 3/5
MYL9 10398 myosin, light chain 9, regulatory 3/5
Matrix metallopeptidase 2 (gelatinasc A,
MMP2 43 1 3 72kDa gelatinasc, 72kDa type IV 3/5 collagenase)
HTRA l 5654 HtrA serine peptidase 1 3/5 immunoglobulin (CD79A) binding protein
IGBP 1 3476 3/5
1
72 actm, gamma 2, smooth muscle, entenc 3/5 transforming, acidic coiled-coil containing
TACC2 1 0579 3/5 protein 2
MYO I E 4643 myosin IE 3/5
FBLN2 2199 fibulin 2 3/5
SRSF I O 1 0772 scrinc/arginine-rich splicing factor 1 0 3/5 protection of telomeres 1 homolog (S.
ΡΟΊΊ 259 13 3/5 pombe)
PSPI I 5723 phosphoserine phosphatase 3/5
PDGFRL 5 157 platelet-derived growth factor receptor-like 3/5
GOLGA8A 23015 golgin A8 family, member A 3/5
TST 7263 thiosul fate sulfurtransferase (rhodanese) 3/5
FAP 21 91 fibroblast activation protein, alpha 3/5
DEAD (Asp-Glu-Ala-Asp) box polypeptide
DDX39B 1 02 12 3/5
39
BTG3 1 0950 BTG family, member 3 3/5
DUSP7 1 849 dual specificity phosphatase 7 3/5
RABGAP 1 L 991 0 RAB GTPase activating protein 1 -like 3/5
ZYX 7791 zyxin 3/5 present /
Gene Symbol EN T REZ G EN E_ I D Gene Title
total lists solute carrier family 7 (amino acid
SLC7A8 23428 3/5 transporter, L-type), member 8
IDE 3416 insulin-degrading enzyme 3/5
WFDC1 58189 WAP four-disulfide core domain 1 3/5
KDELC1 79070 KDEL (Lys-Asp-Glu-Leu) containing 1 3/5
1 potassium voltage-gated channel, Isk-
KCNE4 23704 3/5 related family, member 4
HORMADi 84072 IIORMA domain containing 1 3/5
CDCA7L 55536 cell division cycle associated 7-1 ike 3/5
CTIIRC1 115908 collagen triple helix repeat containing 1 3/5
HIP1 3092 huntingtin interacting protein 1 3/5
ATPase, 11+ transporting, lysosomal 42kDa,
ATP6V1C1 528 3/5
VI subunit CI
ZNF521 25925 zinc finger protein 521 3/5
LOCI 00131564 100131564 hypothetical LOG 100 lj 1564 3/5
ERP27 121506 endoplasmic reticulum protein 27 3/5 immunoglobulin superfamily, DCC
IGDCC4 57722 3/5 subclass, member 4
SCAI 286205 suppressor of cancer cell invasion 3/5 family with sequence similarity 165,
FAM165I3 54065 3/5 member B
asparagine-linked glycosylation 10, alpha-
AEG 1 OB 144245 3/5
1,2-glucosyltransfcrasc hemolog B (yeast)
ZNF92 168374 zinc finger protein 92 3/5

Claims

We claim:
1 . A method of classifying or grading a lung cancer tumor in an individual at risk for or having lung cancer comprising detecting the expression level of at least 2 genes listed in Table 1 in a test sample obtained from the individual; wherein a differential pattern of expression levels of said at least 2 genes in the test sample classifies the lung cancer tumor as one of stage 1 lung adenocarcinoma, stage 2 adenocarcinoma, stage 1 squamous cell carcinoma or stage 2 squamous cell carcinoma,
2. The method according to claim 1 , wherein the differential pattern of expression levels is identified by a classifier based on a plurality of genes listed in Table 1 , including said at least two genes, said classifier having been trained by in silico analysis or one or more feature selection and classification algorithms.
3. The method according to claim 1 or 2, wherein the differential pattern of expression levels is identified by a classifier based on a plurality of genes listed in
Table 1 , including said at least two genes, said classifier having been trained by one or more algorithms selected from the group consisting of dual ensemble, generalized simulated annealing, T-filter, CORG, CORG combined with support vector machine, dual bagging, single and pairs, forward learning, Laplacian based learning and learning method based on network perturbation amplitude.
4. The method according to any one of claims 1 -3, wherein said classi fier having been trained with at least the data in the Gene Expression Omnibus datasets GSE2109, GSE10245, GSE18842 and GSE37745.
5. The method according to any one of claims 1 -4, wherein the method further comprises comparing the expression level of said at least 2 genes in the test sample and a control sample; or detecting the expression level of said at least 2 genes in thc control sample and comparing the expression level of said at least 2 genes in the test sample and control sample, to identify the differential pattern.
6. The method according to any one of claims 1 -5, wherein said at least 2 genes are selected from the group consisting of: ZIC2, LOC I OO I 3 1262, CD83, EML1 , ΡΛΙΡ 1 , NIPBL, CREB3L1 , SLC37A1 , and SFMBT2.
7. The method according to any one of claims 1 -6, wherein the expression level of said at least 2 genes in the test sample are detected by using a human genome-wide array, a human lung tissue array or a custom array comprising polynucleotides of a plurality of genes in Table 1 and said at least 2 genes.
8. The method according to any one of claims 1 -6, wherein the expression level of said at least 2 genes in the test sample are detected by measuring the level of proteins encoded by the genes.
9. An array comprising polynucleotides hybridizing to at least 2 lung cancer signature genes immobilized on a solid surface, wherein the lung cancer signature genes are selected from the genes listed in Table 1 and said array is not a human genorne-wide array.
10. A device comprising antibodies immobilized on a solid surface that bind to proteins encoded by at least 2 lung cancer signature genes, wherein the lung cancer signature genes arc selected from the group consisting of the genes listed in Table 1.
1 1. A computer readable medium or computer program product comprising a classifier based on at least two genes listed in Table 1 , said classifier having been trained by in silico analysis or one or more feature selection and classification algorithms.
12. The computer readable medium or computer program product according to claim 1 1 , wherein said classifier is trained by one or more algorithms selected from the group consisting of dual ensemble, generalized simulated annealing, T-filter, CORG, CORG combined with support vector machine, dual bagging, single and pairs, forward learning, Laplacian based learning and learning method based on network perturbation amplitude.
13. The computer readable medium or computer program product according to claim 1 1 or 12, wherein said classifier is trained with at least the data in the Gene Expression Omnibus datasets GSE2109, GSE 10245, GSE1 8842 and GSE37745.
14. The computer readable medium or computer program product according to any one of claims 1 1 - 13, wherein said at least two genes are selected from the group consisting of ZIC2, LOC 100131262, CD83, EML l , PAIP 1 , NIPBL, CREB3L1 , SLC37A1 , and SFMBT2.
1 5. A kit for classifying and grading a lung cancer tumor or for assessing the prognosis of lung cancer in an individual, comprising one or more reagents that detects expression levels of at least 2 genes listed in Table 1 in a test sample and instructions for using said kit for classifying and grading a lung cancer tumor or for determining the prognosis of lung cancer in said individual.
PCT/EP2013/062993 2012-06-21 2013-06-21 Gene signatures for classifying and grading lung cancer WO2013190090A1 (en)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
US201261662769P 2012-06-21 2012-06-21
US61/662,769 2012-06-21

Publications (1)

Publication Number Publication Date
WO2013190090A1 true WO2013190090A1 (en) 2013-12-27

Family

ID=48874248

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/EP2013/062993 WO2013190090A1 (en) 2012-06-21 2013-06-21 Gene signatures for classifying and grading lung cancer

Country Status (1)

Country Link
WO (1) WO2013190090A1 (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP3149209A4 (en) * 2014-05-30 2017-12-27 Genecentric Therapeutics, Inc. Methods for typing of lung cancer
EP4058484A4 (en) * 2019-11-15 2024-04-03 Gritstone Bio Inc Antigen-binding proteins targeting shared neoantigens

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2004061410A2 (en) 2002-12-18 2004-07-22 Ciphergen Biosystems, Inc. Serum biomarkers in lung cancer
WO2005098445A2 (en) 2004-03-30 2005-10-20 Eastern Virginia Medical School Lung cancer biomarkers
US20070264659A1 (en) 2006-05-11 2007-11-15 Sungwhan An Lung cancer biomarker discovery
WO2009067546A2 (en) * 2007-11-19 2009-05-28 Celera Corpration Lung cancer markers and uses thereof
WO2009146545A1 (en) * 2008-06-05 2009-12-10 University Health Network Compositions and methods for classifying lung cancer and prognosing lung cancer survival
US20100070191A1 (en) 2008-09-09 2010-03-18 Somalogic, Inc. Lung Cancer Biomarkers and Uses Thereof

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2004061410A2 (en) 2002-12-18 2004-07-22 Ciphergen Biosystems, Inc. Serum biomarkers in lung cancer
WO2005098445A2 (en) 2004-03-30 2005-10-20 Eastern Virginia Medical School Lung cancer biomarkers
US20070264659A1 (en) 2006-05-11 2007-11-15 Sungwhan An Lung cancer biomarker discovery
WO2009067546A2 (en) * 2007-11-19 2009-05-28 Celera Corpration Lung cancer markers and uses thereof
US7892760B2 (en) 2007-11-19 2011-02-22 Celera Corporation Lung cancer markers, and uses thereof
WO2009146545A1 (en) * 2008-06-05 2009-12-10 University Health Network Compositions and methods for classifying lung cancer and prognosing lung cancer survival
US20100070191A1 (en) 2008-09-09 2010-03-18 Somalogic, Inc. Lung Cancer Biomarkers and Uses Thereof
WO2011031344A1 (en) 2009-09-09 2011-03-17 Somalogic, Inc. Cancer biomarkers and uses thereof

Non-Patent Citations (33)

* Cited by examiner, † Cited by third party
Title
"Bishop, Neural Networks for Pattern Recognition", 1995, O.U. PRESS
"Current Protocols in Molecular Biology", 1989, JOHN WILCY & SONS, pages: 6.3.1 6 - 13.3.6
AUSUBEL, CURRENT PROTOCOLS IN MOLECULAR BIOLOGY
BRAUN ET AL., THE JOURNAL OF MACHINE LEARNING RESEARCH, vol. 9, 2008, pages 1875 - 1908
BREIMAN, MACHINE LEARNING, vol. 45, no. 1, 2001, pages 5 - 32
BREIMAN: "The Elements of Statistical Learning: Data Mining, Inference, and Prediction. 2nd ed.,", 2009
BRETTSCHNEIDER ET AL., TECHNOMETRICS, vol. 50, no. 3, 2008, pages 241 - 264
BRYLL, PATTERN RECOGNITION, vol. 20, no. 6, 2003, pages 1291 - 1302
BUTLER, J.E., METH. ENZYMOL., vol. 73, 1981, pages 482 - 523
CHUANG ET AL., MOL SYST BIOL, vol. 3, 2007, pages 140
GAUTICR ET AL., BIOINFORMATICS, vol. 20, 2004, pages 307 - 315
GENTLEZNAN ET AL., GENOME BIOL, vol. 5, no. 10, 2004, pages R80
GOSLING, J. P.: "Immunoassays: A Practical Approach", 2001, OXFORD UNIVERSITY PRESS
HARLOW; LANE: "Antibodies: A Laboratory Manual", 1988
HERMIDA ET AL., CONFERO: AN INTEGRATED CONTRAST AND GENE SET PLATFORM FOR COMPUTATIONAL ANALYSIS AND BIOLOGICAL INTERPRETATION OF OMICS DATA., 2012
HO, IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, vol. 20, no. 8, 1998, pages 832 - 844
KLEIN ET AL., J EXP MED, vol. 194, 2001, pages 1625 - 1638
L. GUO ET AL: "Constructing Molecular Classifiers for the Accurate Prognosis of Lung Adenocarcinoma", CLINICAL CANCER RESEARCH, vol. 12, no. 11, 1 June 2006 (2006-06-01), pages 3344 - 3354, XP055077626, ISSN: 1078-0432, DOI: 10.1158/1078-0432.CCR-05-2336 *
M. ROTUNNO ET AL: "A Gene Expression Signature from Peripheral Whole Blood for Stage I Lung Adenocarcinoma", CANCER PREVENTION RESEARCH, vol. 4, no. 10, 1 October 2011 (2011-10-01), pages 1599 - 1608, XP055077722, ISSN: 1940-6207, DOI: 10.1158/1940-6207.CAPR-10-0170 *
MAGGIO, E.: "Enzyme immunoassay", 1980, CRC PRESS
R DEVELOPMENT CORE TEAM, R: A LANGUAGE AND ENVIRONMENT FOR STATISTICAL COMPUTING., 2007
RICHARD A. SHIMKETS,: "Gene Expression Profiling: Methods and Protocols", 2004, ILUMANA PRESS
SAMBROOK: "Molecular Cloning, 2nd ed.,", 1989, COLD SPRING HARBOR PRESS, pages: 9.47 - 9.57
SCHULZE, NATURE CELL BIOL., vol. 3, 2001, pages E190
SKURICHINA, PATTERN ANALYSIS AND APPLICATIONS, vol. 5, no. 2, 2002, pages 121 - 135
TIBSHIRANI ET AL., PROC NATL ACAD SCI USA, vol. 99, no. 10, 2002, pages 6567 - 6572
TSALLIS; STARIOLO, PHYSICA A: STATISTICAL MECHANICS AND ITS APPLICATIONS, vol. 233, no. 1, 1996, pages 395 - 406
VOLLER, A. ET AL., J . PATHOL., vol. 31, 1978, pages 507 - 520
WU ET AL., JOURNAL OF THE AMERICAN STATISTICAL ASSOCIATION, vol. 99, 2004, pages 909
XIANG ET AL., PHYSICS LETTERS A, vol. 233, no. 3, 1997, pages 216 - 220
XIANG ET AL., THE JOURNAL OF PHYSICAL CHEMISTRY A, vol. 104, no. 12, 2000, pages 2746 - 2751
XIANG; GONG, PHYSICAL REVIEW E, vol. 62, no. 3, 2000, pages 4473
YANG ET AL., CURRENT BIOINFORMATICS, vol. 5, no. 4, 2010, pages 296 - 308

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP3149209A4 (en) * 2014-05-30 2017-12-27 Genecentric Therapeutics, Inc. Methods for typing of lung cancer
EP4058484A4 (en) * 2019-11-15 2024-04-03 Gritstone Bio Inc Antigen-binding proteins targeting shared neoantigens

Similar Documents

Publication Publication Date Title
US20190241969A1 (en) Compositions and methods for diagnosing thyroid tumors
JP5405110B2 (en) Methods and materials for identifying primary lesions of cancer of unknown primary
DK2158332T3 (en) PROGRAM FORECAST FOR MELANANCANCES
WO2013190092A1 (en) Gene signatures for copd diagnosis
IL243201A (en) Method for using gene expression to determine prognosis of prostate cancer
KR20130105764A (en) Prognosis prediction for colorectal cancer
Matsui Genomic biomarkers for personalized medicine: development and validation in clinical studies
US20100105564A1 (en) Stroma Derived Predictor of Breast Cancer
US20150344962A1 (en) Methods for evaluating breast cancer prognosis
WO2013190090A1 (en) Gene signatures for classifying and grading lung cancer
WO2015117210A1 (en) Process, apparatus or system and kit for classification of tumor samples of unknown and/or uncertain origin and use of genes of the group of biomarkers
EP2607494A1 (en) Biomarkers for lung cancer risk assessment
US20070275380A1 (en) Method for Distinguishing Aml Subtypes With Aberrant and Prognostically Intermediate Karyotypes
US20070122814A1 (en) Methods for distinguishing prognostically definable aml
WO2005043161A2 (en) Method for distinguishing leukemia subtypes

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 13740216

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 13740216

Country of ref document: EP

Kind code of ref document: A1