US20090076734A1 - Gene Signature for the Prediction of Radiation Therapy Response - Google Patents

Abstract

Described are mathematical models and method, e.g., computer-implemented methods, for predicting tumor sensitivity to radiation therapy, which can be used, e.g., for selecting a treatment for a subject who has a tumor.

Description

CLAIM OF PRIORITY
• This application claims the benefit of U.S. Provisional Patent Application Ser. No. 60/972,544, filed on Sep. 14, 2007, and is a continuation in part of U.S. patent application Ser. No. 12/053,796, filed on Mar. 24, 2008, which claims the benefit of U.S. Provisional Patent Application Ser. No. 60/896,550, filed on Mar. 23, 2007, and U.S. Provisional Patent Application Ser. No. 60/896,350, filed on Mar. 22, 2007. The entire contents of the foregoing are hereby incorporated by reference.
FEDERALLY SPONSORED RESEARCH OR DEVELOPMENT
• This invention was made with Government support under Grant No. 5 K08 CA108926-03, NCI Grant R21CA101355, Department of Defense, National Functional Genomics Center—DAMD 17-02-2-0051. The Government has certain rights in the invention.
TECHNICAL FIELD
• This invention relates to mathematical models and methods for predicting tumor sensitivity to radiation therapy using biological assay data, which can be used, e.g., for selecting a treatment for a subject who has a tumor.
BACKGROUND
• Personalized medicine holds the promise that the diagnosis, prevention and treatment of cancer will be based on individual assessment of risk (Dalton and Friend, Science 2006; 312(5777):1165-8). The delivery of this promise in radiation oncology is dependent on the ability to define the variables that define response to clinical radiotherapy. Although most strategies in personalized medicine have focused on specific disease sites and/or drug therapies (van't Veer et al., Nature 2002; 415(6871):530-6; Beer et al., Nat Med 2002; 8(8):816-24; Chung et al., Cancer Cell 2004; 5(5):489-500; Eschrich et al., J Clin Oncol 2005; 23(15):3526-35; Giles et al., Semin Oncol 2008; 35(1 Suppl 1):S1-17), the impact of individualizing radiation therapy is significant. Approximately 60% of cancer patients are treated with radiation therapy during their diagnosis (Perez, Principles and Management of Radiation Therapy. Philadelphia-New York: Lippincott-Raven; 1998). Thus, radiation therapy provides a common denominator in cancer therapeutics.
• Significant advances towards personalized radiation therapy have been largely achieved by physical advances in radiotherapy treatment planning and delivery (Bucci et al., CA Cancer J Clin 2005; 55(2):117-34). In contrast, the efforts in understanding the biological parameters that define intrinsic radiosensitivity have not met the same success. Thus, radiotherapy is prescribed without considering the potential individual differences in tumor and patient radiosensitivity. However there is evidence to suggest that differences in intrinsic radiosensitivity exist (Zelefsky et al., J. Urology 2001; 166(3):876-81) and understanding their biological basis could significantly impact clinical practice. Thus, a successful radiosensitivity predictive assay would be central to the development of biologically-guided personalized treatment strategies in radiation oncology. However, although a number of promising approaches have been developed in the past (e.g., determination of ex-vivo tumor SF2, (Bjork-Eriksson et al., Int J Radiat Oncol Biol Phys 2000; 46(1):13-9; Buffa et al., Int J Radiat Oncol Biol Phys 2001; 50(5):1113-22; Eschwege et al., Int J Radiat Oncol Biol Phys 1997; 39(4):849-53; Taghian et al., Int J Radiat Oncol Biol Phys 1993; 25(2):243-9; West et al., British Journal of Cancer 1997; 76(9):1184-90; West et al., Br J Cancer 1993; 68(4):819-23); the use of electrodes to measure tumor hypoxia (Fyles et al., J Clin Oncol 2002; 20(3):680-7; Movsas et al., Urology 2002; 60(4):634-9); and determination of tumor proliferative potential (Tpot) (Begg et al., Radiother Oncol 1999; 50(1):13-23; Bourhis et al., Int J Radiat Oncol Biol Phys 1996; 35(3):471-6; Corvo et al., J Clin Oncol 1995; 13(8):1843-50), none has become routine in the clinic.
SUMMARY
• At least in part, the inventions described herein are based on the development of methods and models that predict intrinsic sensitivity of a tumor to radiation therapy based on a gene expression profile.
• In one aspect, the invention provides methods, e.g., computer-implemented methods, for predicting the sensitivity of a cell, i.e., a living cell, e.g., a tumor cell or a normal (non-tumor) cell, or a cultured cell, to a selected dose of radiation therapy. The methods include assigning a radiation sensitivity index to the cell based on expression levels of two or more signature genes in the cell, wherein the radiation sensitivity index indicates whether the cell is sensitive to radiation therapy.
• In an additional aspect, the invention provides methods for predicting the effect of radiation therapy on a tumor. The methods include assigning a radiation sensitivity index to the tumor based on expression levels of two or more signature genes in a cell from the tumor, wherein the radiation sensitivity index indicates whether the radiation therapy is likely to be effective.
• In yet another aspect, the invention provides methods for assessing a tumor in a subject for a radiation therapy regimen. The methods include assigning a radiation sensitivity index to the tumor based on expression levels of two or more signature genes in a cell from the tumor, wherein the radiation sensitivity index indicates whether the tumor in the subject should be treated with radiation therapy.
• In a further aspect, the invention provides methods for selecting a treatment regimen for a subject having a tumor. The methods include assigning a radiation sensitivity index to the tumor based on expression levels of two or more signature genes in a cell from the tumor, and selecting a treatment regimen for the subject based on the radiation sensitivity index. In general, a radiation sensitivity index below a threshold indicates that radiation therapy is likely to be effecting in treating the tumor, and the method includes selecting a treatment regimen including radiation therapy. Conversely, a radiation sensitivity index above a threshold indicates that radiation therapy is not likely to be effecting in treating the tumor, and the method includes selecting a treatment regimen excluding radiation therapy, or a treatment regime including a high dose of radiation therapy.
• In another aspect, the invention provides methods for selecting a dose of radiation to be administered to a subject having a tumor. The methods include assigning to the tumor a radiation sensitivity index for a preselected dose of radiation based on expression levels of two or more signature genes in a cell from the tumor, and selecting a dose of radiation for the subject based on the radiation sensitivity index at the preselected dose of radiation. In some embodiments, the methods include selecting a dose of radiation that is the same as or less than the preselected dose of radiation, if the radiation sensitivity index is below a threshold. In some embodiments, the methods include selecting a dose of radiation that is the greater than the preselected dose of radiation, if the radiation sensitivity index is above a threshold.
• In some embodiments of the methods described herein, assigning a radiation sensitivity index comprises applying a linear regression model to the gene expression levels, e.g., a rank-based linear regression model. In some embodiments, the expression levels of the two or more signature genes are weighted.
• In some embodiments, the linear regression model is represented by the following algorithm:
• 
  RSI=k₁*AR+k₂*c-jun+k₃*STAT1+k₄*PKC+k₅*RelA+k₆*cAbl+k₇*SUMO1+k₈*CDK1+k₉*HDAC+k₁₀*IRF1.  I
• In some embodiments of the methods described herein, assigning a radiation sensitivity index can include determining a level of expression of two or more signature genes in a cell.
• In some embodiments of the methods described herein, the signature genes are selected from the group consisting of Androgen receptor (AR); Jun oncogene (c-Jun); Signal transducer and activator of transcription 1 (STAT 1); Protein kinase C, beta (PRKCB or PKC); V-rel reticuloendotheliosis viral oncogene homolog A (avian) (RELA or p65); c-Abl oncogene 1, receptor tyrosine kinase (ABL 1 or c-Abl); SMT3 suppressor of mif two 3 homolog 1 (S. cerevisiae) (SUMO1); CDK1 (p34); Histone deacetylase 1 (HDAC1); and Interferon regulatory factor 1 (IRF1). In some embodiments, the signature genes comprise a subset of the genes as listed in Table 10, 11, or 12. Optionally, assigning a radiation sensitivity index can also include using gene expression levels of one or more genes listed inFIG. 6.
• In another aspect, the invention provides computer-implemented method of identifying genes associated with radiation sensitivity. The methods include assigning a radiation sensitivity value to one or more populations of cells, wherein the radiation sensitivity value represents the sensitivity of the cells to a selected dose of radiation; determining a level of gene expression in the one or more populations of cells for each of a plurality of genes; and identifying a subset comprising two or more genes, the expression of which is correlated with the radiation sensitivity value. In some embodiments, the subset of the genes is identified by a method including applying a model representing gene expression and radiosensitivity, e.g., a multivariate linear regression model. The model can include, e.g., at least one coefficient representing one, two, or all three of: tissue of origin, ras status, or p53 status. In addition, the model can include information regarding the dose of radiation administered and the presence of any additional treatment or factors relevant to the cell.
• In some embodiments, the methods further include associating a classifier representing biological importance with each gene in the subset of genes, expression of which is correlated with the radiation sensitivity value, and selecting a second subset comprising two or more genes wherein the classifier representing biological importance is above a preselected threshold, thereby selecting a subset of biologically important genes. In some embodiments, the classifier representing biological importance is based on a review of relevant scientific literature. In some embodiments, the model also includes a variable representing an effect of administration of a treatment on expression of each gene in the subset of biologically important genes.
• In some embodiments, the methods further include selecting a third subset comprising one or more genes based on the effect of administration of the treatment, thereby identifying a subset of therapeutic target genes, e.g., genes the expression of which may be usefully manipulated to alter (i.e., increase or decrease) sensitivity to radiation. In some embodiments, the effect of the treatment is an increase or decrease in radiosensitivity. In some embodiments, the methods further include selecting a treatment that has an effect on radiosensitivity in the model.
• In yet another aspect, the invention provides databases including a plurality of records, wherein each record includes data on the expression of at least two signature genes in a cell, and a value representing sensitivity of the cell to a selected dose of radiation. In some embodiments, the database also includes data regarding the administration of a treatment, e.g., chemotherapy, to the cell. In some embodiments, the database is in computer readable form.
• Also provided herein are microarrays including a substrate and a plurality of individually addressable hybridisable array elements arranged thereon, wherein the individually addressable hybridisable array elements are selective for at least two signature genes, and optionally at least one hybridisable array element selective for an internal normalization control gene. In some embodiments, the plurality of hybridisable array elements consists of at least one element selective for each of AR; c-Jun; STAT1; PKC; RelA (p65); c-Abl; SUMO-1; CDK1 (p34); HDAC1; and IRF1.
• In an additional aspect, the invention provides microfluidic devices including a substrate and a plurality of reaction chambers with reagents for selective quantification of at least two signature genes; and optionally at least one reaction chamber comprising reagents for selective quantification of an internal normalization control gene. In some embodiments, the devices include duplicate sets of the reaction chambers, e.g., to allow processing of multiple samples simultaneously.
• In a further aspect, the invention provides a medium, e.g., computer-readable medium, bearing instructions to cause a computer to perform a method described herein. For example, the medium can bear instructions to cause a computer to assign a radiation sensitivity index to a cell based on expression levels of two or more signature genes in the cell. In some embodiments, assigning a radiation sensitivity index comprises applying a linear regression model to the gene expression levels, e.g., a rank-based linear regression model, e.g., wherein the two or more signature genes are weighted.
• In an additional aspect, the invention provides a medium, e.g., computer-readable medium, bearing instructions to cause a computer to assign a radiation sensitivity value to one or more populations of cells, wherein the radiation sensitivity value represents the sensitivity of the cells to a selected dose of radiation; assign a level of gene expression in the one or more populations of cells for each of a plurality of genes; and identify a subset comprising two or more genes, the expression of which is correlated with the radiation sensitivity value. In some embodiments, identifying a subset of the genes comprises applying a model representing gene expression and radiosensitivity, e.g., a multivariate linear regression model. In some embodiments, the model includes at least one coefficient representing one, two, or all three of: tissue of origin, ras status, or p53 status.
• In some embodiments, the medium further comprises instructions to cause a computer to associate a classifier representing biological importance with each gene in the subset of genes, expression of which is correlated with the radiation sensitivity value, and selecting a second subset comprising two or more genes wherein the classifier representing biological importance is above a preselected threshold, thereby selecting a subset of biologically important genes.
• In some embodiments, the model further comprises a variable representing an effect of administration of a treatment on expression of each gene in the subset of biologically important genes. In some embodiments, the medium further includes instructions to cause a computer to select a third subset comprising two or more genes based on the effect of administration of the treatment, thereby identifying a subset of therapeutic target genes.
• Also provided by the present invention are kits including reagents for the specific quantification of gene expression levels of two or more signature genes in a cell, and instructions for carrying out a method as described herein. In some embodiments, the kits include a medium as described herein.
• The present invention has a number of advantages. The models and methods described herein provide an opportunity to individualize radiation dose parameters based on intrinsic radiosensitivity. Since higher doses of radiation therapy are associated with higher toxicity rate (Peeters et al., Int J Radiat Oncol Biol Phys 2005; 61(4):1019-34), dose personalization would result in a therapeutic ratio benefit. In addition the model may provide a unique framework to understand the differences between responders and non-responders that share a predicted radioresistant phenotype. This may allow the accurate identification of patients that benefit from the addition of concurrent chemotherapy.
• Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. Methods and materials are described herein for use in the present invention; other, suitable methods and materials known in the art can also be used. The materials, methods, and examples are illustrative only and not intended to be limiting. All publications, patent applications, patents, sequences, database entries, and other references mentioned herein are incorporated by reference in their entirety. In case of conflict, the present specification, including definitions, will control.
• Other features and advantages of the invention will be apparent from the following detailed description and figures, and from the claims.
DESCRIPTION OF DRAWINGS
• FIGS. 1A and B are bar graphs showing that the predicted tumor radiosensitivity is correlated with clinical response to concurrent radiochemotherapy in rectal and esophageal cancer patients. Predicted radiosensitivity indices (RSI) for each patient were generated using a ten-gene, rank-based linear regression model built from the cell line data as described herein. Statistical significance was determined using a one-sided Mann-Whitney test for differences. (1A) The mean predicted RSI of responders is significantly lower than in non-responders in both clinical cohorts (esophageal: p=0.05, rectal: p=0.03). (1B) Predicted RSI of each individual patient in both cohorts is significantly different relative to response (combined: p=0.001511).
• FIG. 2 is a ROC curve that was generated using the predicted RSI values to determine the sensitivity and specificity of the radiosensitivity predictor. Using a threshold RSI value of 0.4619592, the predictor has an 80% sensitivity and 82% specificity, with a positive predictive value (PPV) of 86%. The estimated area under the curve (AUC) is 0.84.
• FIG. 3 is a line graph demonstrating that predicted radiosensitivity distinguishes clinical populations with different disease related outcomes in head and neck cancer. Radiosensitivity predictions were generated with the gene expression model as described in 92 patients treated with definitive concurrent radiochemotherapy at the Netherlands Cancer Institute. Using the 25th percentile as cutpoint (RSI<0.023), there is a superior 2 year Recurrence Free Survival (RFS) in the predicted radiosensitive group (86% vs. 62%, p=0.06).
• FIG. 4 is a list setting forth gene combinations (profiles) that were evaluated and demonstrated significant association with radiosensitivity in the Rectal+Esophagus cohorts described herein. The gene symbols are joined by ‘_’ and the p-values from tests of significance between responders and non-responders are also given.
• FIG. 5 is a block diagram of computing devices and systems.
• FIG. 6 is a list setting forth 500 genes identified as being correlated with radiosensitivity using a systems network model as described herein.
• FIG. 7 is a list of exemplary control genes useful in the methods described herein.
DETAILED DESCRIPTION
• The advent of high dimensional and high-throughput technologies has provided an opportunity to address the development of biomarkers from a different perspective. For example, gene expression signatures have been shown to be prognostic in breast (van't Veer et al., Nature 2002; 415(6871):530-6), lung (Beer et al., Nat Med 2002; 8(8):816-24), head and neck (Chung et al., Cancer Cell 2004; 5(5):489-500) and colon cancer (Eschrich et al., J Clin Oncol 2005; 23(15):3526-35). Recent studies have identified biomarkers predictive of patient response to drug treatment, including response to Gleevec in Chronic Myelogenous Leukemia (CML) (Giles et al., Semin Oncol 2008; 35(1 Suppl 1):S1-17). In addition, gene expression can predict cellular intrinsic radiosensitivity (Torres-Roca et al., Cancer Res. 65(16):7169-76 (2005)). The present inventors have developed a gene expression model to predict radiosensitivity in patients.
• Described herein is a novel multi-gene expression model of intrinsic tumor radiosensitivity in a database of 48 human cancer cell lines. The model is based on the expression of sets of signature genes, which predicts a radiosensitivity index (RSI) that is directly proportional to tumor radioresistance. The model was clinically validated as a predictive factor of pathological response in two independent cohorts of esophageal (n=12) and rectal (n=14) cancer patients treated with preoperative concurrent chemoradiation in prospective clinical trials at Moffitt Cancer Center. In addition, RSI calculated by a method described herein was of prognostic value in a third external dataset of head and neck cancer patients (n=92) treated with definitive concurrent chemoradiation withinPhase 2 and 3 clinical trials at the Netherlands Cancer Institute. Thus this model can be used to individualize therapy in clinical radiation oncology. For example, the model provides an opportunity to individualize radiation dose parameters based on intrinsic radiosensitivity. Since higher doses of radiation therapy are associated with higher toxicity rate, dose personalization would result in a therapeutic ratio benefit. In addition the model provides a unique framework to understand the differences between responders and non-responders that share a predicted radioresistant phenotype. This allows more accurate identification of patients that benefit from the addition of concurrent chemotherapy.
• In the molecular medicine era, high-throughput technologies (e.g., microarrays and proteomics) have led to the identification of numerous molecular signatures of prognostic and/or predictive significance (van't Veer et al., Nature 2002; 415(6871):530-6; Beer et al., Nat Med 2002; 8(8):816-24; Chung et al., Cancer Cell 2004; 5(5):489-500; Eschrich et al., J Clin Oncol 2005; 23(15):3526-35; Alizadeh et al., Nature 2000; 403(6769):503-11; Bild et al., Nature 2006; 439(7074):353-7; van de Vijver et al., New Eng. J. Med. 2002; 347(25):1999-2009; Shedden et al., Nat Med 2008; 14(8):822-7). However the initial enthusiasm that these signatures would lead to personalized medicine has been dampened by their frequent lack of robustness (Simon et al., J Natl Cancer Inst 2003; 95(1): 14-8).
• The robustness of the radiosensitivity model described herein is supported by several lines of evidence. First, the algorithm was validated in three independent prospectively collected datasets in three different diseases. Second, the model was valid across different gene expression platforms. Gene expression data in the esophageal and rectal cancer cohorts were derived from Affymetrix U-133 Plus microarrays. However, gene expression in the head and neck dataset was derived from NKI arrays, which is a two channel based cDNA microarray platform. The observation that the algorithm is transferable across platforms is important as it demonstrates transferability to other clinical platforms (e.g., using RT-PCR/Formalin-fixed tissue). Third, all patients in the validating clinical cohorts were treated with concurrent chemoradiation, since we were unable to obtain a dataset of patients treated with radiation alone. However the algorithm was based on cellular radiosensitivity. Thus, in spite of this potential source of inaccuracy, the model was still validated. Finally, the model showed both predictive and prognostic value.
• The model described herein is designed to predict tumor radiosensitivity. RSI was prognostic in the head and neck cancer dataset, suggesting that the biological factors that determine radiosensitivity are related to disease prognosis after treatment. This is consistent with the observation that complete pathological response in both esophageal and rectal cancer has strong prognostic significance in several studies (Janjan et al., Am J Clin Oncol 2001; 24(2):107-12; Chirieac et al., Cancer 2005; 103(7):1347-55; Gavioli et al., Dis Colon Rectum 2005; 48(10):1851-7; Capirci et al., Int J Radiat Oncol Biol Phys. 2008 Sep. 1; 72(1):99-107. Epub 2008 Apr. 11). Thus a model that can identify complete responders would indeed be desirable. Using the current model, 6/8 complete responders fall below the threshold suggested by the ROC analysis (SF2=0.46) suggesting that this model successfully identifies this population.
• In addition, there is a role for identifying patients that are likely to be downstaged, particularly in rectal cancer. For example, this knowledge might lead to better counseling of patients with low-lying rectal tumors where sphincter-sparing surgery is being considered. Patients that have low lying rectal cancer are generally patients that have a tumor within 5 cm of the anal sphincter. Classically, these patients when operated were treated with an abdomino-perineal resection (APR), which removes the anal sphincter and thus requires the patient to have a permanent colostomy bag, which generally adversely affects the patient's quality of life. In order to address this, protocols were developed about 10-15 years ago to test whether using a course of preoperative radiation or chemoradiation would improve the ability of the surgeon to spare the sphincter; shrinking the tumor improves the chances that the surgeon can remove the cancer and reconnect the rectum and keep the normal sphincter mechanism intact. This approach has been successful, but the likelihood of sphincter-sparing surgery is related to the amount of downstaging achieved by chemoradiation. Thus, the model described herein can improve patient counseling before a treatment decision is made. For example, if the patient is likely to respond to preoperative treatment because the model determines that the rectal tumor is radiosensitive, then the chances of success are high. However, if the tumor is radioresistant, success is unlikely, and the patient can be counseled to go directly to surgery and thereby be spared the side effects of a treatment that is unlikely to be successful, or if preoperative treatment is still pursued then higher doses of radiation therapy could be prescribed to increase the chances of success.
• Determination of Radiosensitivity Index (RSI) of a Tumor
• The methods described herein use a rank-based linear algorithm to assign a radiosensitivity index (RSI) to a cell, i.e., a living cell, e.g., a tumor cell from a patient, a normal cell from a patient, or a cultured cell. In general, the methods are applicable to any mammal, particularly humans. The methods include determining expression levels of signature genes in a cell or cells of the tumor, and determining a RSI based on the expression levels. In some embodiments, the methods include the use of two or more, e.g., three, four, five, six, seven, eight, nine, or all ten signature genes, as follows: Androgen receptor (AR); Jun oncogene (c-Jun); Signal transducer and activator of transcription 1 (STAT1); Protein kinase C, beta (PRKCB or PKC); V-rel reticuloendotheliosis viral oncogene homolog A (avian) (RELA or p65); c-Abl oncogene 1, receptor tyrosine kinase (ABL1 or c-Abl); SMT3 suppressor of mif two 3 homolog 1 (S. cerevisiae) (SUMO1); CDK1 (p34); Histone deacetylase 1 (HDAC1); and Interferon regulatory factor 1 (IRF1).

• TABLE A
  Exemplary Sequences of Signature Genes - Human

  			Probe
  	Gene	Refseq	Design
  Gene Name	Symbol	Identifiers	Identifier
  Androgen Receptor	AR	NM_000044.2	M23263.1
  		NM_001011645.1
  Jun oncogene		NM_002228.3	J04111.1
  Signal transducer and	STAT1	NM_007315.2	M97935.1
  activator of		NM_139266.1
  transcription 1
  Protein kinase C, beta	PRKCB	NM_002738.5	X06318.1
  	PKC	NM_212535.1
  v-rel	RELA	NM_021975.2	U33838.1
  reticuloendotheliosis	P65
  viral oncogene
  homolog A (avian)
  c-Abl oncogene 1,	ABL1	NM_007313.2	X16416.1
  receptor tyrosine kinase	c-Abl	NM_005157.3
  SMT3 suppressor of	SUMO1	NM_001005781.1	U83117.1
  mif two 3homolog 1		NM_001005782.1
  (S. cerevisiae)		NM_003352.4
  P21 protein	PAK2	NM_002577.3	U24153.1
  (Cdc42/Rac)-activated	CDK1(p34)
  kinase 2
  Histone deacetylase 1	HDAC1	NM_004964.2	D50405.1
  	HDAC
  Interferon regulatory	IRF1	NM_002198.2	L05072.1
  factor 1

  
  Although the exemplary gene sequences set forth above are for the human genes, and thus are best suited for use in human cells, one of skill in the art could readily identify mammalian homologs using database searches (for known sequences) or routine molecular biological techniques (to identify additional sequences). In general, genes are considered homologs if they show at least 80%, e.g., 90%, 95%, or more, identity in conserved regions (e.g., biologically important regions).
• In some embodiments, the profile includes the signature genes listed in a profile shown in Table 10, Table 11, Table 12, orFIG. 4. In some embodiments, the profile includes at least c-jun, STAT1, cAbl, and IRF1. In some embodiments, the profile includes at least IRF1.
• A linear regression model useful in the methods described herein includes gene expression levels and coefficients, or weights, for combining expression levels. The coefficients can be calculated using a least-squares fit of the proposed model to a measure of cellular radiation sensitivity. One example described herein used the survival fraction at 2 Gy (SF2) although other measures at other dose levels (e.g., SF8) can be considered with different coefficients being determined from each. The functional form of the algorithm is given below, wherein each of the k_icoefficients will be determined by fitting expression levels to a particular RSI measure.
• 
  RSI=k₁*AR+k₂*c-jun+k₃*STAT1+k₄*PKC+k₅*RelA+k₆*cAbl+k₇*SUMO1+k₈*CDK1+k₉*HDAC+k₁₀*IRF1  I
• In some embodiments, the methods include applying an algorithm to expression level data determined in a cell; e.g., a rank-based linear regression algorithm as described herein. In some embodiments, the algorithm includes weighting coefficients for each of the genes.
• Methods of Use
• The methods described herein can be used to identify a radiation sensitivity index to a selected dose of radiation for any solid tumor in a subject. A solid tumor is an abnormal mass of hyperproliferative or neoplastic cells from a tissue other than blood, bone marrow, or the lymphatic system, which may be benign or cancerous. In general, the tumors treated by the methods described herein are cancerous. As used herein, the terms “hyperproliferative” and “neoplastic” refer to cells having the capacity for autonomous growth, i.e., an abnormal state or condition characterized by rapidly proliferating cell growth. Hyperproliferative and neoplastic disease states may be categorized as pathologic, i.e., characterizing or constituting a disease state, or may be categorized as non-pathologic, i.e., a deviation from normal but not associated with a disease state. The term is meant to include all types of solid cancerous growths, metastatic tissues or malignantly transformed cells, tissues, or organs, irrespective of histopathologic type or stage of invasiveness. “Pathologic hyperproliferative” cells occur in disease states characterized by malignant tumor growth. Examples of non-pathologic hyperproliferative cells include proliferation of cells associated with wound repair. Examples of solid tumors are sarcomas, carcinomas, and lymphomas. Leukemias (cancers of the blood) generally do not form solid tumors.
• The term “carcinoma” is art recognized and refers to malignancies of epithelial or endocrine tissues including respiratory system carcinomas, gastrointestinal system carcinomas, genitourinary system carcinomas, testicular carcinomas, breast carcinomas, prostatic carcinomas, endocrine system carcinomas, and melanomas. In some embodiments, the disease is lung carcinoma, rectal carcinoma, colon carcinoma, esophageal carcinoma, prostate carcinoma, head and neck carcinoma, or melanoma. Exemplary carcinomas include those forming from tissue of the cervix, lung, prostate, breast, head and neck, colon and ovary. The term also includes carcinosarcomas, e.g., which include malignant tumors composed of carcinomatous and sarcomatous tissues. An “adenocarcinoma” refers to a carcinoma derived from glandular tissue or in which the tumor cells form recognizable glandular structures.
• The term “sarcoma” is art recognized and refers to malignant tumors of mesenchymal derivation.
• In some embodiments, the tumors treated by a method described herein are of epithelial cell origin. In some embodiments, the tumors originate from lung, colon, rectal, esophageal, prostate, or head/neck tissues (e.g., originating from the upper aerodigestive tract, including the lip, oral cavity, nasal cavity, paranasal sinuses, pharynx, and larynx, e.g., squamous cell carcinomas originating from the mucosal lining (epithelium)). In some embodiments, the tumors are metastatic, and originate from an epithelial tissue (and are thus epithelial in origin) but have spread to another tissue, e.g., epithelial-origin prostate cancer that has spread to the bones of the pelvis, spine and/or ribs, or lung carcinoma that has metastasized to the adrenal glands, liver, brain, or bones.
• The methods described herein can identify tumors that are sensitive to radiation therapy, and thereby identify subjects who would benefit from administration of radiation therapy having, or who would benefit from concurrent administration of radiation therapy and radiation sensitizing chemotherapy. For example, once a RSI has been determined for a tumor, if the RSI is low and thus indicates that the tumor is sensitive to radiation (and thus is likely to be effectively treated with radiation), then a course of radiation alone can be prescribed for the patient, or radiation and possibly less invasive surgical removal methods, e.g., laparoscopic methods. Alternatively, if the RSI is high and thus indicates that the tumor is less sensitive or is not sensitive to radiation therapy, then a course of chemotherapy, e.g., radiation sensitizing chemotherapy, can be prescribed in combination with radiation therapy, and optionally more invasive or radical surgical resection. Thus the methods can be used to predict a subject's response to radiation therapy. In some embodiments, the threshold for sensitivity using an RSI as defined in esophageal and rectal cancer as 0.46.
• As one example, the methods can be used for identifying patients that are likely to be downstaged, particularly in rectal cancer. For example, this knowledge might lead to better counseling of patients with low-lying rectal tumors where sphincter-sparing surgery is being considered, as described above.
• In some embodiments, a subject having a tumor is identified (methods for diagnosing the presence of a tumor are well known in the art and need not be repeated herein). A test sample is obtained from the tumor and the level of signature protein or nucleic acid (e.g., mRNA) is evaluated, wherein the level of signature protein or nucleic acid is indicative of the sensitivity of the tumor to radiation therapy. As used herein, a “test sample” refers to a biological sample obtained from a subject of interest including a cell or cells, e.g., tissue, from the tumor.
• The assays described herein can also be used to determine whether a subject should be administered one or more of radiation therapy, chemotherapy, or surgical resection to treat a solid tumor, e.g., to select a therapy or therapeutic regime for a subject. For example, such methods can be used to determine whether a subject can be effectively treated with radiation therapy alone or radiation therapy with a second, non-radiation treatment modality, e.g., surgery or chemotherapy, or will need radiation, surgery, and chemotherapy.
• In addition, the methods described herein can be used on normal cells, i.e., non-tumor cells, to determine their sensitivity to radiation therapy. This allows the use of the model to predict the likelihood of radiation therapy-related toxicity or other side effects.
• The network system models described herein can also be used to select genes to target for agents, e.g., radiosensitizing or radioprotective agents. In these methods, the network models are modified to model the effects of modulating various genes. One such approach is to simulate the effects of biological targeting of one or more of the identified network hubs. This type of in silico perturbation of the developed model can provide additional information on the hubs most likely to effect radiation phenotype. The model can be perturbed by systematically reducing (using computer simulations) the rank or weight of each hub gene to the lowest possible value, in effect “knocking” the gene out. The altered expression pattern will be used to predict patient radiosensitivity using the same model previously constructed.
• Differences from the unaltered SF2 predictions will be recorded. These changes in SF2 will be examined and averaged over the entire patient cohort to estimate the impact of individual gene knockout.
• Radiation therapies, chemotherapies, surgical resection techniques, and methods that can be used to select specific therapies appropriate for a given tumor, are known in the art, see, e.g., “Practical Radiotherapy Planning,” Dobbs, Barrett, and Ash (1999) Arnold; “Walter & Miller's Textbook of Radiotherapy,” Bomford and Kunkler (2002) Churchill Livingstone; “Cancer Chemotherapy and Biotherapy: Principles and Practice,” Chabner and Longo (2005) Lippincott Williams & Wilkins; “Regional Chemotherapy: Theory and Practice,” Kerr and McArdle (2000) Informa Healthcare; and “Textbook of Surgery,” Tjandra et al. (2006) Wiley-Blackwell.
• Assays for Determining Expression Levels
• Any method known in the art for obtaining a sample comprising at least one living cell (preferably a plurality of cells), e.g., a cell from a tumor (e.g., from a biopsy), or a normal cell, or a cultured cell, can be used. Commonly used methods to obtain tumor cells include surgical (the use of tissue taken from the tumor after removal of all or part of the tumor) and needle biopsies. The samples should be treated in any way that preserves intact the gene expression levels of the living cells as much as possible, e.g., flash freezing or chemical fixation, e.g., formalin fixation.
• Any method known in the art can be used to extract material, e.g., protein or nucleic acid (e.g., mRNA) from the sample. For example, mechanical or enzymatic cell disruption can be used, followed by a solid phase method (e.g., using a column) or phenol-chloroform extraction, e.g., guanidinium thiocyanate-phenol-chloroform extraction of the RNA. A number of kits are commercially available for use in isolation of mRNA. Purification can also be used if desired. See, e.g., Peirson and Butler, Methods Mol. Biol. 2007; 362:315-27. A number of methods are also known in the art to obtain proteins from cells, see, e.g., “Protein Methods,” 2nd Edition by Bollag et al., Wiley Pub. (1996). Optionally, cDNA can be transcribed from the mRNA.
• Gene expression levels can be determined in many different ways, including the quantification of fluorescence of hybridized mRNA on glass slides, Northern blot analysis, real-time reverse transcription PCR(RT-PCR) or other measures of gene expression abundance. Each of these ways provides a different scale, however each approach is proportional to the abundance of a particular mRNA transcript.
• A number of assays suitable for the determination of expression levels of the signature genes in a biological sample are known in the art. For example, expression levels can be evaluated by obtaining a biological sample from tumor of a test subject and contacting the biological sample with a compound or an agent capable of detecting mRNA for the signature genes, or protein encoded by the signature genes, such that the level of the protein or nucleic acid is detected in the biological sample. The term “biological sample” includes tissues, cells and fluids comprising cells or tissues isolated from tumor of a subject, as well as tissues and cells and fluids present within a subject. A preferred biological sample is a biopsy sample taken from the tumor. The level of expression of the signature genes can be measured in a number of ways, including, but not limited to: measuring the mRNA encoded by the signature genes; measuring the amount of protein encoded by the signature genes; or measuring the activity of the protein encoded by the signature genes.
• The level of mRNA corresponding to the signature gene in a cell can be determined both by in situ and by in vitro formats.
• The isolated mRNA can be used in hybridization or amplification assays that include, but are not limited to, Southern or Northern analyses, polymerase chain reaction analyses and probe arrays. One exemplary diagnostic method for the detection of mRNA levels involves contacting the isolated mRNA with a nucleic acid molecule (probe) that can hybridize to the mRNA encoded by the gene being detected. The nucleic acid probe can be, for example, a full-length nucleic acid or an oligonucleotide of at least 7, 15, 30, 50, 100, 250 or 500 nucleotides in length and sufficient to specifically hybridize under stringent conditions to mRNA for a signature gene. Other suitable probes for use in the diagnostic assays are known in the art.
• In one format, mRNA (or cDNA) from the sample is immobilized on a surface and contacted with the probes, for example by running the isolated mRNA on an agarose gel and transferring the mRNA from the gel to a membrane, such as nitrocellulose. In an alternative format, the probes are immobilized on a surface and the mRNA (or cDNA) from the sample is contacted with the probes, for example, in a two-dimensional gene chip array. A skilled artisan can adapt known mRNA detection methods for use in detecting the level of mRNA encoded by the signature genes.
• The level of mRNA in a sample that is encoded by one of signature can be evaluated with nucleic acid amplification, e.g., by rtPCR (Mullis (1987) U.S. Pat. No. 4,683,202), ligase chain reaction (Barany (1991) Proc. Natl. Acad. Sci. USA 88:189-193), self sustained sequence replication (Guatelli et al., (1990) Proc. Natl. Acad. Sci. USA 87:1874-1878), transcriptional amplification system (Kwoh et al., (1989), Proc. Natl. Acad. Sci. USA 86:1173-1177), Q-Beta Replicase (Lizardi et al., (1988) Bio/Technology 6:1197), rolling circle replication (Lizardi et al., U.S. Pat. No. 5,854,033) or any other nucleic acid amplification method, followed by the detection of the amplified molecules using techniques known in the art. As used herein, amplification primers are defined as being a pair of nucleic acid molecules that can anneal to 5′ or 3′ regions of a gene (plus and minus strands, respectively, or vice-versa) and contain a short region in between. In general, amplification primers are from about 10 to 30 nucleotides in length and flank a region from about 50 to 200 nucleotides in length. Under appropriate conditions and with appropriate reagents, such primers permit the amplification of a nucleic acid molecule comprising the nucleotide sequence flanked by the primers.
• A preferred method is the use of microfluidic devices, e.g., for high-throughput real time-polymerase chain reaction (RT-PCR), e.g., as described herein.
• For in situ methods, a cell or tissue sample can be prepared/processed and immobilized on a support, typically a glass slide, and then contacted with a probe that can hybridize to mRNA that encodes the signature gene being analyzed.
• In another embodiment, the methods further contacting a control sample with a compound or agent capable of detecting signature mRNA, and comparing the presence of signature mRNA in the control sample with the presence of signature mRNA in the test sample.
• A variety of methods can be used to determine the levels of proteins encoded by the selected signature genes. In general, these methods include contacting an agent that selectively binds to the protein, such as an antibody, with a sample, and evaluating the level of protein in the sample. In a preferred embodiment, the antibody bears a detectable label. Antibodies can be polyclonal, or more preferably, monoclonal. An intact antibody, or a fragment thereof (e.g., Fab or F(ab′)2) can be used. The term “labeled,” with regard to the probe or antibody, is intended to encompass direct labeling of the probe or antibody by coupling (i.e., physically linking) a detectable substance to the probe or antibody, as well as indirect labeling of the probe or antibody by reactivity with a detectable substance. Examples of detectable substances are known in the art, as are methods of quantifying levels of proteins detected thereby.
• The detection methods can be used to detect signature protein in a biological sample in vitro as well as in vivo. In vitro techniques for detection of signature protein include enzyme linked immunosorbent assays (ELISAs), immunoprecipitations, immunofluorescence, enzyme immunoassay (EIA), radioimmunoassay (RIA), and Western blot analysis. In vivo techniques for detection of signature protein include introducing into a subject a labeled anti-signature antibody. For example, the antibody can be labeled with a radioactive marker whose presence and location in a subject can be detected by standard imaging techniques.
• In another embodiment, the methods further include contacting a control sample with a compound or agent capable of detecting signature protein, quantifying the level of signature protein, and comparing the level of signature protein in the control sample with the level of signature protein in the test sample.
• In some embodiments, the sensitivity of a tumor to radiation therapy can be predicted by determining a gene expression profile including expression levels for two or more of the signature genes described herein, and comparing that expression profile to a reference profile, e.g., a reference profile representing a tumor that is sensitive to radiation; in that case, substantial similarity between the reference profile and the profile of expression from the tumor would indicate that the tumor was sensitive to radiation. Methods for performing such methods are known in the art, e.g., as described in U.S. Pat. No. 7,148,008.
• Kits
• The invention also includes kits for detecting and quantifying the selected signature genes (e.g., mRNA or protein corresponding to the signature genes) in a biological sample. For example, the kit can include a compound or agent capable of detecting mRNA or protein corresponding to the signature genes in a biological sample; and a standard; and optionally one or more reagents necessary for performing detection, quantification, or amplification. The compounds, agents, and/or reagents can be packaged in a suitable container. The kit can further comprise instructions for using the kit to detect and quantify signature protein or nucleic acid.
• For antibody-based kits, the kit can include: (1) a first antibody (e.g., attached to a solid support) which binds to a polypeptide corresponding to a signature gene; and, optionally, (2) a second, different antibody which binds to either the polypeptide or the first antibody and is conjugated to a detectable agent.
• For oligonucleotide-based kits, the kit can include: (1) an oligonucleotide, e.g., a detectably labeled oligonucleotide, which hybridizes to a nucleic acid sequence corresponding to a signature gene or (2) a pair of primers useful for amplifying a nucleic acid molecule corresponding to a signature gene. The kit can also includes a buffering agent, a preservative, and/or a protein stabilizing agent. The kit can also include components necessary for detecting the detectable agent (e.g., an enzyme or a substrate). The kit can also contain a control sample or a series of control samples which can be assayed and compared to the test sample contained. Each component of the kit can be enclosed within an individual container and all of the various containers can be within a single package, along with instructions for interpreting the results of the assays performed using the kit.
• In some embodiments, the kits include reagents specific for the quantification of the signature genes listed in a profile shown in Table 10, Table 11, Table 12, orFIG. 4. In some embodiments, the kits include primers or antibodies selective for at least c-jun, STAT1, cAbl, and IRF1. In some embodiments, the kits include primers or antibodies selective for at least IRF1 and one additional signature gene. microfluidic devices for RT-PCR. In some embodiments, the kits also include primers or antibodies selective for a housekeeping or control gene, e.g., as listed inFIG. 7.
• Microarrays/Microfluidic Devices
• Also described herein are microarrays useful for detecting and quantifying levels of mRNA or protein corresponding to the signature genes. The microarray comprises a substrate and hybridisable array elements. For the detection of mRNA, the microarray will include a plurality of individually addressable areas including hybridizable array elements selective for the selected signature genes. For the detection of protein, the microarray will include a plurality of individually addressable areas including reagents for the detection of one or more proteins encoded by the signature genes, e.g., antibodies.
• In some embodiments, the microarrays include hybridisable array elements selective for the signature genes listed in a profile shown in Table 10, Table 11, Table 12, orFIG. 4. In some embodiments, the microarrays include hybridisable array elements selective for at least c-jun, STAT1, cAbl, and IRF1. In some embodiments, the microarrays include hybridisable array elements selective for at least IRF1 and one additional signature gene.
• The term “microarray” refers to a substrate having an ordered arrangement of hybridisable array elements arranged thereon. In some embodiments, the array elements are arranged so that there are preferably at least about 10 different array elements, on a 1 cm.sup.2 substrate surface. The maximum number of array elements is unlimited, but can be upwards of at least 100,000 array elements. Furthermore, a hybridization signal from each of the array elements is individually distinguishable. In a preferred embodiment, the array elements comprise polynucleotide probes.
• Hybridization causes a denatured polynucleotide probe and a denatured complementary target to form a stable duplex through base pairing. Hybridization methods are well known to those skilled in the art (See, e.g., Laboratory Techniques in Biochemistry and Molecular Biology, Vol. 24: Hybridization With Nucleic Acid Probes, P. Tijssen, ed. Elsevier Science, New York, N.Y. (1993)). Conditions can be selected for hybridization where exactly complementary target and polynucleotide probe can hybridize, i.e., each base pair must interact with its complementary base pair. Alternatively, conditions can be selected where target and polynucleotide probes have mismatches but are still able to hybridize. Suitable conditions can be selected, for example, by varying the concentrations of salt or formamide in the prehybridization, hybridization and wash solutions, or by varying the hybridization and wash temperatures.
• Hybridization can be performed at low stringency with buffers, such as 6×SSPE with 0.005% Triton X-100 at 37° C., which permits hybridization between target and polynucleotide probes that contain some mismatches to form target polynucleotide/probe complexes. Subsequent washes are performed at higher stringency with buffers, such as 0.5×SSPE with 0.005% Triton X-100 at 50° C., to retain hybridization of only those target/probe complexes that contain exactly complementary sequences. Alternatively, hybridization can be performed with buffers, such as 5×SSC/0.2% SDS at 60° C. and washes are performed in 2×SSC/0.2% SDS and then in 0.1×SSC. Stringency can also be increased by adding agents such as formamide. Background signals can be reduced by the use of detergent, such as sodium dodecyl sulfate, Sarcosyl or Triton X-100, or a blocking agent, such as sperm DNA.
• Hybridization specificity can be evaluated by comparing the hybridization of specificity-control polynucleotide probes to specificity-control target polynucleotides that are added to a sample in a known amount. The specificity-control target polynucleotides may have one or more sequence mismatches compared with the corresponding polynucleotide probes. In this manner, whether only complementary target polynucleotides are hybridizing to the polynucleotide probes or whether mismatched hybrid duplexes are forming is determined.
• After hybridization, the microarray is washed to remove non-hybridized nucleic acids and complex formation between the hybridisable array elements and the target polynucleotides is detected.
• Methods for detecting complex formation are known in the art. In some embodiments, the target polynucleotides are labeled with a fluorescent label and measurement of levels and patterns of fluorescence indicative of complex formation is accomplished by fluorescence microscopy, preferably confocal fluorescence microscopy. An argon ion laser excites the fluorescent label, emissions are directed to a photomultiplier and the amount of emitted light detected and quantitated. The detected signal should be proportional to the amount of probe/target polynucleotide complex at each position of the microarray. The fluorescence microscope can be associated with a computer-driven scanner device to generate a quantitative two-dimensional image of hybridization intensity. The scanned image is examined to determine the abundance/expression level of each hybridized target polynucleotide.
• Typically, microarray fluorescence intensities can be normalized to take into account variations in hybridization intensities when more than one microarray is used under similar test conditions. In a preferred embodiment, individual polynucleotide probe/target complex hybridization intensities are normalized using the intensities derived from internal normalization controls contained on each microarray, e.g., control genes, e.g., ubiquitin C; hydroxymethylbilane synthase; tyrosine 3-monooxygenase/tryptophan 5-monooxygenase activation protein, zeta polypeptide; polymerase (RNA) II (DNA directed) polypeptide A, 220 kDa;importin 8; hypoxanthine phosphoribosyltransferase 1 (Lesch-Nyhan syndrome); ribosomal protein, large P0 (RLPO); peptidylprolyl isomerase A (cyclophilin A); beta actin; beta glucuronidase; beta-2-microglobulin;phosphoglycerate kinase 1; glyceraldehyde-3-phosphate dehydrogenase; transferrin receptor (p90; CD71); TATA box binding protein; subunit A of the succinate dehydrogenase complex; and the 18s ribosomal RNA (seeFIG. 7).
• The microarrays described herein include (or consist of) individually addressable hybridisable array elements selective for the signature genes as described herein, or a subset thereof listed in any of Tables 10-12 orFIG. 4. In some embodiments, the microarrays also include one or more hybridisable array elements selective for an internal normalization control, e.g., as described herein. In some embodiments, the microarrays do not include hybridisable array elements selective for other genes.
• In some embodiments, a microfluidic RT-PCR/paraffin-preserved tissue platform can be used. There are several advantages to the use of this platform. First, this platform is practical to use for routine diagnostic application. For example, OncotypeDX, a multi-gene model for risk assessment in breast cancer is in an RT-PCR/paraffin-preserved tissue platform. Further, the RT-PCR approach is cost-efficient. In addition, a standardized and optimized test could readily be tested in banked tissue (paraffin-preserved), e.g., from cooperative group trials (e.g., the Radiation Therapy Oncology Group (RTOG)), for development into routine clinical use.
• Custom-design microfluidics cards can be obtained, e.g., from Applied Biosystems (ABI), that include all 10 genes in the hub model, or a subset thereof (e.g., as described herein), along with the standard 16 housekeeping genes recommended by ABI, or a subset thereof. Additional genes, e.g., 20, 30, 40, 50, 60, 70, 80, 90, or 100 genes, e.g., genes selected from the larger network model consisting of 500 genes (listed inFIG. 6), can also be included in the cards. In some embodiments, all genes will be present at least in duplicate, e.g., in duplicate or triplicate, enabling the analysis of two or more samples per card. As one example, if an ABI card is used, the microarray probeset targets identified using a method described herein will be sequence-aligned with the ABI probeset to determine the ABI probes closest to the microarray target.
• As one example, the microfluidics assays can be performed substantially as follows. Briefly, THE ABI TAQMAN LOW DENSITY ARRAY (TLDA) cards profile gene expression using the Comparative CT Method of relative quantification. Each card consists of a series of 384 interconnected wells divided into eight sets of assays. Each well contains dried Applied Biosystems TaqMan primers and probes for one mRNA target. Each of the 8 ports of the card is loaded with 100 μL of sample-specific PCR mix (Each 100-μL PCR mix should contain 1 ng to 100 ng of total RNA converted to cDNA. Once the card is loaded with samples, it is centrifuged to distribute the PCR mix throughout the 48 wells of each port. The TLDA card is run in the AB 7900HT using relative quantification analysis.
• General methods for making and using microfluidic devices are known in the art, see, e.g., U.S. Pat. Nos. 6,960,437 and 7,250,260.
• Databases
• In another aspect, the invention features a database comprising a plurality of records. Each record includes data on the expression of at least two signature genes in a tumor cell, and at least one, two or preferably all of the following: data on tissue of origin of the cell; data on ras status of the cell; and data on p53 status of the cell, and optionally data on a preselected factor relating to a subject who has the tumor. In some embodiments, the preselected factor can be one or more of: the presence of a treatment (e.g., the administration of a compound, e.g., a drug (e.g., chemotherapy), vitamin, food or dietary supplement); the presence of an environmental factor (e.g., the presence of a substance in the environment); the presence of a genetic factor or physical factor such as age.
• In some embodiments, the database includes at least two records, and the preselected factor in each of the records differs form the other record. For example, in one embodiment, the preselected factor can be administration of a compound and in one record the preselected factor includes administration of the compound and in the other record the compound is not administered, is administered at a different dose and/or a different compound is administered. In another embodiment, the preselected factor can be an environmental factor and in one record the factor is present and in the other record the environmental factor is not present or is present at a different level. In yet another embodiment, the preselected factor can be a physical factor such as age and the age in one record varies from the age in the other record, e.g., a difference in age of at least 5, 10, 15, 20 years or more.
• In some embodiments, each record of the database includes data on at least two preselected factors relating to the subject. In one embodiment, the database includes at least two records, and at least one preselected factor in each of the records differs from the other record. Preferably, the database includes at least two records and at least one preselected factor in the records differ and at least one of the other preselected factors is the same. In other embodiments, the database can include at least two records and each record includes at least one preselected factor and at least one preselected condition.
• In some embodiments, the database includes at least two records, wherein each record includes information regarding a cell including expression levels of a subset or all of the 10 signature genes as described herein, dose of radiation administered, and survival fraction in response to that dose of radiation (e.g., for a dose of 2 Gy, the survival fraction is referred to as SF2).
• The database can be any kind of storage system capable of storing the various data for each of the records as described herein. For example, the database may be a flat file, a relational database, a table in a database, an object in a computer readable volatile or non-volatile memory, data accessible by computer program, such as data stored in a resource fork of an application program file on a computer readable storage medium. Preferably, the database is in a computer readable medium (e.g., a computer memory or storage device).
• In some embodiments, each record can further include data on the expression of at least one internal control gene, e.g., as listed inFIG. 7.
• The information obtained by evaluating the efficacy of radiation therapy in treating a tumor can also be used to evaluate the effects that various factors and conditions, e.g., environmental conditions, can have on tumor treatment. In some embodiments, the information can be stored in a database as described herein.
• In another aspect, the invention features a method of evaluating the likelihood that radiation therapy will be effective in treating a tumor, using a database as described herein.
• The database can be any kind of storage system capable of storing various data for each of the records as described herein. In preferred embodiments, the database is a computer medium having a plurality of digitally encoded data records. The data record can be structured as a table, e.g., a table that is part of a database such as a relational database (e.g., a SQL database of the Oracle or Sybase database environments).
• As used herein, “machine-readable media” refers to any medium that can be read and accessed directly by a machine, e.g., a digital computer or analogue computer. Non-limiting examples of a computer include a desktop PC, laptop, mainframe, server (e.g., a web server, network server, or server farm), handheld digital assistant, pager, mobile telephone, and the like. The computer can be stand-alone or connected to a communications network, e.g., a local area network (such as a VPN or intranet), a wide area network (e.g., an Extranet or the Internet), or a telephone network (e.g., a wireless, DSL, or ISDN network). Machine-readable media include, but are not limited to: magnetic storage media, such as floppy discs, hard disc storage medium, and magnetic tape; optical storage media such as CD-ROM; electrical storage media such as RAM, ROM, EPROM, EEPROM, flash memory, and the like; and hybrids of these categories such as magnetic/optical storage media.
• A variety of data storage structures are available to a skilled artisan for creating a machine-readable medium having recorded thereon the data described herein. The choice of the data storage structure will generally be based on the means chosen to access the stored information. In addition, a variety of data processor programs and formats can be used to store the information of the present invention on computer readable medium.
• Computer Software/Hardware
• FIG. 5 is a block diagram of computing devices andsystems700,750 that may be used and implemented to perform operations associated with the audio file toolbox404.Computing device700 is intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers.Computing device750 is intended to represent various forms of mobile devices, such as personal digital assistants, cellular telephones, smartphones, and other similar computing devices. The components shown here, their connections and relationships, and their functions, are meant to be exemplary only, and are not meant to limit implementations of the inventions described and/or claimed in this document.
• Computing device700 includes aprocessor702,memory704, astorage device706, a high-speed interface708 connecting tomemory704 and high-speed expansion ports710, and alow speed interface712 connecting tolow speed bus714 andstorage device706. Each of thecomponents702,704,706,708,710, and712, are interconnected using various busses, and can be mounted on a common motherboard or in other manners as appropriate. Theprocessor702 can process instructions for execution within thecomputing device700, including instructions stored in thememory704 or on thestorage device706 to display graphical information for a GUI on an external input/output device, such asdisplay716 coupled tohigh speed interface708. In other implementations, multiple processors and/or multiple buses can be used, as appropriate, along with multiple memories and types of memory. Also,multiple computing devices700 can be connected, with each device providing portions of the operations (e.g., as a server bank, a group of blade servers, or a multi-processor system).
• Thememory704 stores information within thecomputing device700. In one implementation, thememory704 is a computer-readable medium. In one implementation, thememory704 is a volatile memory unit or units. In another implementation, thememory704 is a non-volatile memory unit or units.
• Thestorage device706 is capable of providing mass storage for thecomputing device700. In one implementation, thestorage device706 is a computer-readable medium. In various different implementations, thestorage device706 can be a floppy disk device, a hard disk device, an optical disk device, or a tape device, a flash memory or other similar solid state memory device, or an array of devices, including devices in a storage area network or other configurations. In one implementation, a computer program product is tangibly embodied in an information carrier. The computer program product contains instructions that, when executed, perform one or more methods, such as those described above. The information carrier is a computer- or machine-readable medium, such as thememory704, thestorage device706, memory onprocessor702, or a propagated signal.
• Thehigh speed controller708 manages bandwidth-intensive operations for thecomputing device700, while thelow speed controller712 manages lower bandwidth-intensive operations. Such allocation of duties is exemplary only. In one implementation, the high-speed controller708 is coupled to memory707, display716 (e.g., through a graphics processor or accelerator), and to high-speed expansion ports710, which can accept various expansion cards (not shown). In the implementation, low-speed controller712 is coupled tostorage device706 and low-speed expansion port714.
• The low-speed expansion port, which can include various communication ports (e.g., USB, Bluetooth, Ethernet, wireless Ethernet) can be coupled to one or more input/output devices, such as a keyboard, a pointing device, a scanner, or a networking device such as a switch or router, e.g., through a network adapter.
• Thecomputing device700 can be implemented in a number of different forms, as shown in the figure. For example, it can be implemented as astandard server720, or multiple times in a group of such servers. It can also be implemented as part of arack server system724. In addition, it can be implemented in a personal computer such as alaptop computer722. Alternatively, components fromcomputing device700 can be combined with other components in a mobile device (not shown), such asdevice750.
• Each of such devices can contain one or more ofcomputing device700,750, and an entire system can be made up ofmultiple computing devices700,750 communicating with each other.
• Computing device750 includes aprocessor752,memory764, an input/output device such as adisplay754, acommunication interface766, and atransceiver768, among other components. Thedevice750 can also be provided with a storage device, such as a microdrive or other device, to provide additional storage. Each of thecomponents750,752,764,754,766, and768, are interconnected using various buses, and several of the components can be mounted on a common motherboard or in other manners as appropriate.
• Theprocessor752 can process instructions for execution within thecomputing device750, including instructions stored in thememory764. The processor can also include separate analog and digital processors. The processor can provide, for example, for coordination of the other components of thedevice750, such as control of user interfaces, applications run bydevice750, and wireless communication bydevice750.
• Processor752 can communicate with a user throughcontrol interface758 anddisplay interface756 coupled to adisplay754. Thedisplay754 can be, for example, a TFT LCD display or an OLED display, or other appropriate display technology. Thedisplay interface756 can comprise appropriate circuitry for driving thedisplay754 to present graphical and other information to a user. Thecontrol interface758 can receive commands from a user and convert them for submission to theprocessor752. In addition, anexternal interface762 can be provide in communication withprocessor752, so as to enable near area communication ofdevice750 with other devices.External interface762 can provide, for example, for wired communication (e.g., via a docking procedure) or for wireless communication (e.g., via Bluetooth or other such technologies).
• Thememory764 stores information within thecomputing device750. In one implementation, thememory764 is a computer-readable medium. In one implementation, thememory764 is a volatile memory unit or units. In another implementation, thememory764 is a non-volatile memory unit or units.Expansion memory774 can also be provided and connected todevice750 throughexpansion interface772, which can include, for example, a SIMM card interface.Such expansion memory774 can provide extra storage space fordevice750, or can also store applications or other information fordevice750. Specifically,expansion memory774 can include instructions to carry out or supplement the processes described above, and can include secure information also. Thus, for example,expansion memory774 can be provide as a security module fordevice750, and can be programmed with instructions that permit secure use ofdevice750. In addition, secure applications can be provided via the SIMM cards, along with additional information, such as placing identifying information on the SIMM card in a non-hackable manner.
• The memory can include for example, flash memory and/or MRAM memory, as discussed below. In one implementation, a computer program product is tangibly embodied in an information carrier. The computer program product contains instructions that, when executed, perform one or more methods, such as those described above. The information carrier is a computer- or machine-readable medium, such as thememory764,expansion memory774, memory onprocessor752, or a propagated signal.
• Device750 can communicate wirelessly throughcommunication interface766, which can include digital signal processing circuitry where necessary.Communication interface766 can provide for communications under various modes or protocols, such as GSM voice calls, SMS, EMS, or MMS messaging, CDMA, TDMA, PDC, WCDMA, CDMA2000, or GPRS, among others. Such communication can occur, for example, through radio-frequency transceiver768. In addition, short-range communication can occur, such as using a Bluetooth, WiFi, or other such transceiver (not shown). In addition,GPS receiver module770 can provide additional wireless data todevice750, which can be used as appropriate by applications running ondevice750.
• Device750 can also communication audibly usingaudio codec760, which can receive spoken information from a user and convert it to usable digital information.Audio codex760 can likewise generate audible sound for a user, such as through a speaker, e.g., in a handset ofdevice750. Such sound can include sound from voice telephone calls, can include recorded sound (e.g., voice messages, music files, etc.) and can also include sound generated by applications operating ondevice750.
• Thecomputing device750 can be implemented in a number of different forms, as shown in the figure. For example, it can be implemented as acellular telephone780. It can also be implemented as part of asmartphone782, personal digital assistant, or other similar mobile device.
• Where appropriate, the systems and the functional operations described in this specification can be implemented in digital electronic circuitry, or in computer software, firmware, or hardware, including the structural means disclosed in this specification and structural equivalents thereof, or in combinations of them. The techniques can be implemented as one or more computer program products, i.e., one or more computer programs tangibly embodied in an information carrier, e.g., in a machine readable storage device or in a propagated signal, for execution by, or to control the operation of, data processing apparatus, e.g., a programmable processor, a computer, or multiple computers. A computer program (also known as a program, software, software application, or code) can be written in any form of programming language, including compiled or interpreted languages, and it can be deployed in any form, including as a stand alone program or as a module, component, subroutine, or other unit suitable for use in a computing environment. A computer program does not necessarily correspond to a file. A program can be stored in a portion of a file that holds other programs or data, in a single file dedicated to the program in question, or in multiple coordinated files (e.g., files that store one or more modules, sub programs, or portions of code). A computer program can be deployed to be executed on one computer or on multiple computers at one site or distributed across multiple sites and interconnected by a communication network.
• The processes and logic flows described in this specification can be performed by one or more programmable processors executing one or more computer programs to perform the described functions by operating on input data and generating output. The processes and logic flows can also be performed by, and apparatus can be implemented as, special purpose logic circuitry, e.g., an FPGA (field programmable gate array) or an ASIC (application specific integrated circuit).
• Processors suitable for the execution of a computer program include, by way of example, both general and special purpose microprocessors, and any one or more processors of any kind of digital computer. Generally, the processor will receive instructions and data from a read only memory or a random access memory or both. The essential elements of a computer are a processor for executing instructions and one or more memory devices for storing instructions and data. Generally, a computer will also include, or be operatively coupled to receive data from or transfer data to, or both, one or more mass storage devices for storing data, e.g., magnetic, magneto optical disks, or optical disks. Information carriers suitable for embodying computer program instructions and data include all forms of non volatile memory, including by way of example semiconductor memory devices, e.g., EPROM, EEPROM, and flash memory devices; magnetic disks, e.g., internal hard disks or removable disks; magneto optical disks; and CD ROM and DVD-ROM disks. The processor and the memory can be supplemented by, or incorporated in, special purpose logic circuitry.
• To provide for interaction with a user, aspects of the described techniques can be implemented on a computer having a display device, e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor, for displaying information to the user and a keyboard and a pointing device, e.g., a mouse or a trackball, by which the user can provide input to the computer. Other kinds of devices can be used to provide for interaction with a user as well; for example, feedback provided to the user can be any form of sensory feedback, e.g., visual feedback, auditory feedback, or tactile feedback; and input from the user can be received in any form, including acoustic, speech, or tactile input.
• The techniques can be implemented in a computing system that includes a back-end component, e.g., as a data server, or that includes a middleware component, e.g., an application server, or that includes a front-end component, e.g., a client computer having a graphical user interface or a Web browser through which a user can interact with an implementation, or any combination of such back-end, middleware, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication, e.g., a communication network. Examples of communication networks include a local area network (“LAN”) and a wide area network (“WAN”), e.g., the Internet.
• The computing system can include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other.
• One computer-implemented modeling algorithm is described herein (namely, the linear and quadratic analysis), although such algorithms themselves are generally outside the scope of the present invention. Other software-based modeling algorithms can also be utilized, alone or in combination, such as the classification or decision trees, support vector machines or neural networks.
EXAMPLES
• The invention is further described in the following examples, which do not limit the scope of the invention described in the claims.
Example 1A Radiosensitivity Systems Model Captures Central Regulatory Pathways in Radiation Response
• The model used in the methods described herein was developed in 48 cancer cell lines from the NCI panel of 60 (listed in Table 1). Radiosensitivity measurements (as determined by clonogenic survival at 2 Gy, SF2) were either determined using known methods (Gupta et al., Cancer Res 2001; 61:4278-82; Torres-Roca et al., Cancer Res 2005; 65(16):7169-76) (25 cell lines) or obtained from the literature (23 cell lines). SF2 results for each cell line are presented in Table 1.

• TABLE 1
  48 cell lines and measured SF2 values.

  		Recorded
  	Cell Line	SF2

  	BREAST_HS578T	0.79
  	BREAST_MDAMB231	0.82
  	COLON_HCT116	0.38
  	COLON_HCT15	0.4
  	COLON_SW620	0.62
  	LEUK_CCRFCEM	0.185
  	LEUK_HL60	0.315
  	LEUK_MOLT4	0.05
  	MELAN_SKMEL2	0.66
  	NSCLC_A549ATCC	0.61
  	NSCLC_H460	0.84
  	NSCLC_HOP62	0.164
  	NSCLC_NCIH23	0.086
  	OVAR_OVCAR5	0.408
  	RENAL_SN12C	0.62
  	BREAST_BT549	0.632
  	BREAST_MCF7	0.576
  	BREAST_MDAMB435	0.1795
  	BREAST_T47D	0.52
  	CNS_SF268	0.45
  	CNS_SF539	0.82
  	CNS_SNB19	0.43
  	CNS_SNB75	0.55
  	CNS_U251	0.57
  	COLON_COLO205	0.69
  	COLON_HCC-2998	0.44
  	COLON_HT29	0.79
  	COLON_KM12	0.42
  	MELAN_LOXIMVI	0.68
  	MELAN_M14	0.42
  	MELAN_MALME3M	0.8
  	MELAN_SKMEL28	0.74
  	MELAN_SKMEL5	0.72
  	MELAN_UACC257	0.48
  	MELAN_UACC62	0.52
  	NSCLC_EKVX	0.7
  	NSCLC_HOP92	0.43
  	OVAR_OVCAR3	0.55
  	OVAR_OVCAR4	0.29
  	OVAR_OVCAR8	0.6
  	OVAR_SKOV3	0.9
  	PROSTATE_DU145	0.52
  	PROSTATE_PC3	0.484
  	RENAL_7860	0.66
  	RENAL_A498	0.61
  	RENAL_ACHN	0.72
  	RENAL_CAKI1	0.37
  	RENAL_UO31	0.62

• Gene expression profiles for all cell lines at baseline were from Affymetrix HU6800 chips (7,129 genes) from a previously published study (Staunton et al., Proc Natl Acad Sci USA 2001; 98(19):10787-92). These are publicly available as supplemental data to the published study (Staunton et al., 2001). The gene expression data had been previously preprocessed using the Affymetrix MAS 4.0 algorithm in average difference units. Negative expression values were set to zero and the chips were normalized to the same mean intensity.
• From the total of 7,129, a subset of genes of interest was selected by a linear regression algorithm where radiosensitivity was modeled based on survival fraction at 2 Gy (SF2) in the 48 cancer cell line database. Gene expression profiles and SF2 for all cell lines in the database had been previously determined, as described above.
• A general linear model was created for each gene in the cell line dataset to model the SF2 values. Independent variables used within the linear model were gene expression, p53 mutation status (17 lines were wt, 31 were mutant), ras mutation status (33 wt, 15 mutant) and tissue of origin (TO), see Tables 2 and 3 for additional details on the numbers of cell lines within each category. Tissue of origin, p53 mutation and ras mutation were coded using “dummy” variables (0/1).

• TABLE 2
  Cell line characteristics for TO (Tissue of Origin)

  		Number of
  	Tissue of Origin	Cell lines

  Melanoma

  	8
  	Colon	7
  	Breast	6
  	Renal	6
  	Non-Small Cell Carcinoma (NSCLC)	6
  	CNS	5
  	Ovarian	5
  	Leukemia	3
  	Prostate	2
  	TOTAL	48

• TABLE 3
  Cell line characteristics by each biological variable.

  	ras wt	ras mut

  p53 wt

  	Tissue ofOrigin
  	Renal
  	5	0
  	Breast	0	0
  	CNS	0	0
  	Colon	0	0
  	Leukemia	0	0
  	Melanoma	4	0
  	NSCLC	2	0
  	Ovarian	4	0
  	Prostate	2	0

  P53 mut

  	Tissue oforigin
  	Renal
  	0	1
  	Breast	4	2
  	CNS	5	0
  	Colon	4	3
  	Leukemia	0	3
  	Melanoma	3	1
  	NSCLC	0	4
  	Ovarian	0	1
  	Prostate	0	0

• The linear model format initially considered all terms (9 TO, ras wt/mut, p53 wt/mut) and 2-, 3- and 4-way interactions among these terms. Without accounting for linearly dependent terms, there are 180 terms total, far more than the number of observations (i.e., 48). These include an intercept, 14 terms involving a single variable (gene expression, 9 TO, 2 p53, 2 ras), 53 paired terms, 76 triples and 36 terms with four variables interacting. However, the number of non-singular terms was far less due to the sample size (Table 3) and linearly dependent variables (typically interactions with no effect) are dropped from the model. Interactions of larger numbers of variables were dropped in favor of fewer in the case of linearly dependent variables. Thus there are only 29 terms in the linear model (an intercept, gene expression, 9 TO, p53, ras, 15 two-way interactions and 2 three-way interactions). When considering biological states, the intercept was not used thus producing 28 biological states. This model is expected to overfit the data significantly; however the model was used to describe the relationships in the data in an exploratory fashion as opposed to statistically determining a significant relationship.

• TABLE 4
  Terms used in linear modeling. The term (y) represents gene expression.
  The operator x represents an interaction term between
  two or more variables.
  Terms

  	y (Gene expression)
  	TissueTypeBREAST
  	TissueTypeCNS
  	TissueTypeCOLON
  	TissueTypeLEUK
  	TissueTypeMELAN
  	TissueTypeNSCLC
  	TissueTypeOVAR
  	TissueTypePROSTATE
  	RASmut
  	P53mut
  	y x TissueTypeBREAST
  	y x TissueTypeCNS
  	y x TissueTypeCOLON
  	y x TissueTypeLEUK
  	y x TissueTypeMELAN
  	y x TissueTypeNSCLC
  	y x TissueTypeOVAR
  	y x TissueTypePROSTATE
  	y x RASmut
  	y x P53mut
  	TissueTypeBREAST x RASmut
  	TissueTypeCOLON x RASmut
  	TissueTypeMELAN x RASmut
  	TissueTypeNSCLC x RASmut
  	TissueTypeOVAR x RASmut
  	y x TissueTypeBREAST x RASmut
  	y x TissueTypeCOLON x RASmut

• A model based on the description above was constructed for each gene in the dataset using a least-squares fit. The best fitting genes were selected, as measured by the sum of squares of residuals. The gene-based linear models were compared to the fit expected simply from the biological characteristics (tissue of origin, ras status (mut vs. wild-type) and p53 status (mut vs. wild-type)). This simpler model used 28 terms and resulted in a sum of squared error of residuals of 1.208211.
• The resulting model:
• 
  SF2=k₀+k₁(y_x)+k₂(TO)+k₃(rasstatus)+k₄(p53 status)+k₅(y_x)(TO)+k₆(y_x)(rasstatus)+k₇(TO)(rasstatus)+k₈(y_x)(p53 status)+k₉(TO)(p53)+k₁₀(rasstatus)(p53 status)+k₁₁(y_x)(TO)(rasstatus)+k₁₂(y_x)(rasstatus)(p53 status)+k₁₃(TO)(rasstatus)(p53 status)+k₁₄(y_x)(TO)(rasstatus)(p53 status)
• 500 gene-based models were chosen (threshold ssq=0.5416959) corresponding to at most 45% of the sum squared error from the biological characteristics model. The 500 genes, which are listed inFIG. 6, represent 7% of the total number of probesets on the chip.
• Next, pathway analysis was performed to examine the biological significance of the genes identified. 500 probesets representing these genes were loaded into GeneGO MetaCore software (GeneGO, Encinitas, Calif.) and analyzed for significant over-representation in various pathways; the primary edges (interconnections) were plotted using literature-based annotations and the model was reduced by identifying all genes (network hubs) with more than 5 edges and less than 50% of edges hidden within the network. 485 probesets were recognized in GeneGO.
• Hubs within a gene network were defined using the GeneGO™ software as nodes consisting of more than 5 connections and less than 50% of the edges hidden within the network. Table 5 details all defined hubs within the radiation response network together with the number of edges and the number of hidden edges (along with the probesets used on each platform for each of the hubs).

• TABLE 5
  Radiation network hub genes.

  		Number	Number of
  	Gene	of	Hidden	HU6800	U133Plus
  Gene Name	Number	Edges	Edges	Probeset	Probeset

  Androgen

  	1.	19	0	M23263_at	211110_s_at
  receptor
  c-Jun	2.	19	4	J04111_at	201466_s_at
  STAT1
  	3.	15	1	AFFX-	AFFX-
  				HUMISGF3A/	HUMISGF3A/
  				M97935_MA_at	M97935_MA_at
  PKC
  	4.	14	4	X06318_at	207957_s_at
  RelA (p65)	5.	14	2	U33838_at	201783_s_at
  c-Abl	6.	13	0	X16416_at	202123_s_at
  SUMO-1	7.	13	0	U83117_at	208762_at
  CDK1 (p34)	8.	11	3	U24153_at	205962_at
  HDAC
  	9.	10	0	D50405_at	201209_at
  Integrin
  	10.	7	4
  IRF1	11.	7	0	L05072_s_at	202531_at
  PKC-beta	12.	6	5
  Caspase-8	13.	5	0
  CDC25C	14.	5	4
  Cyclin D1	15.	5	0
  FasR (CD95)	16.	5	0
  Galpha(g)-specific	17.	5	5
  peptide GPCRs
  HES1
  	18.	5	0

• The Gather program (Gene Annotation Tool to Help Explain Relationships (Chang and Nevins, Bioinformatics 22(23):2926-2933, 2006) tool was used to identify significant relationships of terms from the 10 genes. A threshold of p<0.005 was used as a cutoff.
• All hubs with more than 5 connections and less than 50% of edges hidden within the network were selected as the major hubs for classification purposes. Genes listed in Table 6 were selected. The probes used on each platform (Affymetrix HU6800, HG U133Plus 2.0 and NKI cDNA arrays) are also listed in Table 3. Matches were identified via sequence similarity to the original HU6800 platform.
• Table 6 shows the ten “hub” genes on whose expression the radiosensitivity model is built. These genes are also referred to herein as “signature” genes.

• TABLE 6
  Radiation network hub (signature) genes.

  	HU6800	U133Plus	NKI
  Gene Name	Probeset	Probeset	Reporter
  Androgen receptor	M23263_at	211110_s_at	324293
  c-Jun	J04111_at	201466_s_at	329987
  STAT1	AFFX-	AFFX-	308421
  	HUMISGF3A/	HUMISGF3A/
  	M97935_MA_at	M97935_MA_at
  PKC	X06318_at	207957_s_at	322907
  RelA (p65)	U33838_at	201783_s_at	326475
  c-Abl	X16416_at	202123_s_at	304192
  SUMO-1	U83117_at	208762_at	308596
  CDK1 (p34)	U24153_at	205962_at	332859
  HDAC1	D50405_at	201209_at	308690
  IRF1	L05072_s_at	202531_at	310653

• The selected genes are biologically important, as they have been reported to be involved in regulating radiation signaling (Deng et al., Nat Genet. 2004; 36(8):906-12; Hallahan et al., International Journal of Radiation Oncology*Biology*Physics 1996; 36(2):355-60; Kao et al., J Biol Chem 1999; 274(49):34779-84; Li and Karin, PNAS 1998; 95(22):13012-7; Liu et al., Molecular Cell 2006; 21(4):467-80; Mårten Fryknäs et al., International Journal of Cancer 2007; 120(1):189-95; Nakajima et al., Radiat Res 2004; 161(5):528-34; Pamment et al., Oncogene 2002; 21(51):7776-85; Terzoudi et al., Int Radiat Biol 2000; 76(5):607-15; Wang et al., Nucleic Acids Res 2005; 33(13):4023-34). In addition, 7/10 (HDAC1, PKC-beta, NFKB, c-Abl, STAT1, AR, CDK1) have been studied as targets for radiosensitizer development (Wang et al., Nucleic Acids Res 2005; 33(13):4023-34; Russell et al., Cancer Res 2003; 63(21):7377-83; Ma et al., J Clin Oncol 2003; 21(14):2760-76; Cerna et al., Current topics in developmental biology 2006; 73:173-204; Milas et al., Head & Neck 2003; 25(2):152-67, Kaminski et al., Int J Radiat Oncol Biol Phys 2003; 57(1):24-8). Furthermore, the Gene Ontology (GO) terms captured by the 10 gene systems model, include DNA damage response, histone deacetylation, regulation of cell cycle, apoptosis and proliferation, all of which play an important role in radiation response (Marples et al., Int J Radiat Oncol Biol Phys 2008; 70(5):1310-8; Chinnaiyan et al., Int J Radiat Oncol Biol Phys 2005; 62(1):223-9) Lindsay et al., Br J Radiol. 2007 September; 80 Spec No 1:S2-6; Ma et al., 2003, supra). One notable exception includes hypoxia (Moeller et al., Cancer Metastasis Rev 2007; 26(2):241-8). However, since the analysis is based on data generated in normoxic conditions, we would not expect the model to capture hypoxia-related genes. In summary, the systems model captures central pathways and genes involved in regulating radiosensitivity.
• Eight different cell lines were used to validate the importance of c-jun in the systems model. In each experiment, a pooled siRNA and c-jun siRNA experiment were performed several times (i.e. replicates). One estimate of the impact, per cell line, of c-Jun knockdown is a Wilcoxon signed-rank test between the experiment and control that were run together. Table 7 represents the characteristics of the experiments performed, including the tissue of origin, number of times the experiment was performed, mean values (with standard deviation) and a p-value testing differences between mean values (Wilcoxon signed-rank test for cell line replicate experiments).

• TABLE 7
  Individual cell line siRNA experiments.

  			SF2 (Mean ± standard
  Tissue			deviation)
  of Origin	Cell Line	n	siRNA pool vs. c-Jun siRNA	p-value

  Lung

  A549

  	5	0.52 ± 0.13; 0.71 ± 0.11	0.062
  Cancer	H460		9	0.50 ± 0.06; 0.60 ± 0.08	0.004
  	Hop62	8	0.41 ± 0.16; 0.50 ± 0.18	0.039
  Colon	HCT116		7	0.23 ± 0.05; 0.30 ± 0.06	0.016
  Cancer	HCT15		7	0.59 ± 0.09; 0.66 ± 0.13	0.156
  	HT29	5	0.85 ± 0.21; 0.96 ± 0.31	0.312
  Breast	Hs578		10	0.62 ± 0.09; 0.61 ± 0.07	0.770
  Cancer	MDA231		6	0.61 ± 0.09; 0.67 ± 0.11	0.156

• These results demonstrate that at least one of the genes in the model, c-jun, is mechanistically involved in the cellular response to radiation.
Example 3Development of a Radiosensitivity Predictive Model Based on the Systems Model
• A linear regression algorithm to predict radiosensitivity was developed and optimized using gene expression of the 10 genes in the systems model.
• Translation of the model to other datasets was an important requirement, therefore the hubs were assigned ranks by expression (gene expression data for the 10 identified genes were rank ordered, so that lowest expression among the ten genes was ranked 1; HDAC gene expression was generally the highest of the ten genes and therefore was often ranked tenth) and the linear regression model was built from ranks (instead of absolute expression) (Xu et al., BMC Bioinformatics 2008; 9:125) based on expression levels in the 48 cell lines using the R statistical software package (available on the world wide web at r-project.org). The model predicts a continuous radiosensitivity index (RSI) that is based on the survival fraction at a given dose, measured for the cell lines in the database. Thus, the radiosensitivity index is directly proportional to radioresistance (high index=radioresistance). Since the 10 hubs were selected from the cell line data, cross-validation of this linear regression model would yield optimistically-biased estimates of accuracy. As a result, additional datasets were used for validation.
• A general model based on expression of all 10 hub genes is as follows, including the weighting coefficient for each term representing expression levels of the recited gene:
• 
  RSI=k₁*AR+k₂*c-jun+k₃*STAT1+k₄*PKC+k₅*RelA+k₆*cAbl+k₇*SUMO1+k₈*CDK1+k₉*HDAC+k₁₀*IRF1  I
• The rank-based linear regression equation for determining RSI at a dose of 2Gy identified using the present methods was the following:
• 
  RSI=−0.0098009*AR+0.0128283*c-jun+0.0254552*STAT1−0.0017589*PKC−0.0038171*RelA+0.1070213*cABL−0.0002509*SUMO1−0.0092431*CDK1−0.0204469*HDAC−0.0441683*IRF1  II
Example 3The Radiosensitivity Model Predicts Pathological Response to Chemoradiation in Rectal and Esophageal Cancer
• The regression model developed as described in Examples 1-2 was then applied to similarly rank-ordered patient data to generate a Radiation Sensitivity Index (RSI).
• The model was applied to the prediction of clinical response to concurrent radiochemotherapy in two independent prospectively-collected pilot cohorts of patients with rectal (n=14) and esophageal cancer (n=12). Pathological response was defined by T stage criteria (see methods).
• The Rectal Cancer Cohort consisted of 14 patients enrolled in an IRB-approvedprospective Phase 1 trial evaluating escalating doses of oral topotecan as a radiosensitizing agent in patients with rectal cancer. Informed consent was obtained for all patients prior to enrollment. The eligibility criteria included patients with histologically-confirmed rectal cancer with a primary tumor at least 3 cm in size and a clinical stage of T₂-T₄. An ECOG performance status of 2 or less was required as well as a life expectancy of more than 3 months. The diagnosis could not be more than 90 days from initial clinic visit or from the start of therapy. All study subjects were treated at the H Lee Moffitt Cancer Center and Research Institute. Subjects were clinically-staged by endoscopic ultrasound (EUS). Tumor and adjacent normal mucosa biopsies (a minimum of 5 core biopsies) were obtained for microarray analysis before initiation of therapy and betweenday 10 and 14 of preoperative radiochemotherapy. For the purposes of this study only the pretreatment tumor tissue microarray was utilized. Biopsies were snap frozen in liquid nitrogen. No macro or microdissection of the biopsies was performed.
• All study subjects were treated with preoperative concurrent radiochemotherapy and underwent surgical resection (APR or LAR in 13/14) within 8 weeks of completion of preoperative treatment. The starting dose of oral Topotecan was 0.25 mg/m²and it was administered at least 3 hours before XRT on a daily basis. Patients were treated with 45 Gy per day (prescribed to the isocenter) to a standard pelvic field with either a three field or four field 3-D conformal technique. Table 8 shows a summary of the clinical characteristics of this cohort.

• TABLE 8
  Clinical Characteristics for
  Rectal CancerTrial

  	Sex
  	Male

  	10
  	Female	4
  	Age (y)
  	Mean	69.4
  	Median (range)	72	(50-90)
  	Chemotherapy Dose
  	0.25 mg/m2/day	3	(21)
  	0.4 mg/m2/day	5	(36)
  	0.55 mg/m2/day	6	(43)
  	Ultrasound Tumor Stage
  	T3	14	(100)
  	Pathological Tumor Stage
  	T0	2	(14.3)
  	Tis	1	(7)
  	T1	2	(14.3)
  	T2	3	(21.4)
  	T3	5	(36)
  	T4	1	(7)
  	Downstaging
  	Yes	8	(57)
  	No	6	(43)
  	Values are number (percentage) unless otherwise noted.

• Pathological response in the rectal cancer cohort was defined by at least a decrease of one T stage in the primary tumor between the pretreatment EUS and the pathological evaluation of the specimen (Janjan et al., Int. J. Rad. Oncol. Biol. Phys. 1999; 44(5):1027-38; Janjan et al., Am J Clin Oncol 2001; 24(2):107-12). Pathological complete response was defined as no evidence of tumor in the surgical specimen (primary and nodes). Based on this definition, 57% (8/14) of the patients were considered responders.
• The Esophageal Cancer Cohort consisted of 12 patients enrolled in an IRB-approved prospective tissue collection trial, aimed at defining molecular signatures of prognostic value in esophageal cancer. Clinical management was not dictated by the protocol and left to the clinical judgment of the treating physicians. Treatment details are presented in table 9. Eligibility criteria included a histological diagnosis of esophageal cancer, deemed a reasonable candidate for preoperative radiochemotherapy and/or esophagectomy by the evaluating physician. An ECOG performance below 2 was required. In addition, patients were required to be chemotherapy näive. Study subjects were clinically staged by EUS. Biopsies of the tumor and normal mucosa were snap frozen in liquid nitrogen for microarray analysis.
• All subjects in this cohort were treated with concurrent radiochemotherapy to be followed by planned esophagectomy. 9/12 underwent planned esophagectomy. Three patients completed concurrent radiochemotherapy but were not operated because of patient or physician preference (2 patients) or progressive disease (one patient). The clinical characteristics of this cohort is summarized in Table 9.

• TABLE 9
  Clinical Characteristics for
  Esophageal Trial

  	Sex
  	Male	7	(58.3)
  	Female	5	(41.7)
  	Age (y)
  	Mean	67.08
  	Median (range)	66	(51-80)
  	Chemotherapy Regimen
  	CDDP + 5-FU	3	(25)
  	5-FU	2	(16.7)
  	Carbo/Tax + 5-FU	1	(8.3)
  	NA	6	(50)
  	Radiation Dose
  	45	1	(8.3)
  	50.4	4	(33.3)
  	54	2	(16.7)
  	61.2	1	(8.3)
  	NA	4	(33.3)
  	Clinical Tumor Stage
  	T2N1	1	(8.3)
  	T3N0	1	(8.3)
  	T3N1	7	(58.4)
  	T4N1	3	(25)
  	Pathological Tumor Stage
  	T0N0	4	(33.3)
  	T0N1	1	(8.3)
  	T1aN0	1	(8.3)
  	T1N1	2	(16.7)
  	T2bN1	1	(8.3)
  	T2N1	1	(8.3)
  	Progressive Dx	2	(16.7)
  	Downstaging
  	Yes	7	(58.3)
  	No	5	(41.7)
  	Values are number (percentage) unless otherwise noted.

• Clinical response in the esophageal cancer cohort was defined as a decrease of at least two T stages between the pretreatment EUS evaluation and the pathological evaluation of the specimen (Chirieac et al., Cancer 2005; 103(7):1347-55). Three patients in our cohort did not undergo esophagectomy. One had progressive disease during preoperative therapy; the other two experienced clinical complete responses (documented by PET and/or EUS and biopsy) and had no evidence of disease at least one year after completion of treatment. Based on thisdefinition 50% (6/12) of the patients were considered responders.
• The specimen sampled was flash frozen within 15 minutes of resection and the RNA was extracted. Total RNA from the excised tissue was isolated using the TRIZOL™ Reagant (Invitrogen, Carlsbad, Calif.) and the manufacturer's protocol. The aqueous phase containing the RNA separated from the TRIZOL™ reagent was further purified using the RNeasy cleanup procedure (Qiagen Inc., Valencia, Calif.). The quality of total RNA was then assessed by agarose gel electrophoresis and A_260/280ratio or by analysis on the Agilent 2100 Bioanalyzer. Five micrograms of total RNA from each sample was processed for microarray analysis. The poly(A) RNA was specifically converted to cDNA and then amplified and labeled with biotin following the procedure initially described by Van Gelder et al. (Proc Natl Acad Sci USA 1990; 87(5):1663-7). Hybridization with the biotin labeled RNA, staining, and scanning of the chips followed the prescribed procedure outlined in the Affymetrix technical manual and has been previously described (Dobbin et al., Clin Cancer Res 2005; 11(2 Pt 1):565-72).
• The oligonucleotide probe arrays used were the Affymetrix U133A 2.0 plus chips. Since the original cell line data was created on the HU6800 GeneChip, while the newer patient expression data was generated on HG-U133Plus chips, it was necessary to translate the hub probesets in Table 6. This was done using the blast program to find the best U133+ probeset match to the consensus sequence from which the 6800 probeset was designed. The Affymetrix NetAffx software was also used for this translation. Scanned output files were visually inspected for hybridization artifacts and then analyzed using the robust multi-array analysis method (RMA) (Irizarry et al., Nucleic Acids Res 2003; 31(4):e15(27)). Statistical testing for patient cohorts was determined from predicted RSI values using a one-sided Mann-Whitney test. The test was used to determine if the predicted RSI was significantly higher for non-responders. Bar-charts of patient response were graphed using mean and standard error values for each response group in both the rectal cancer and esophageal cancer data. Relapse-free survival differences between low and high RSI values were calculated using a log-rank test of censored survival times.
• As shown inFIG. 1, the model significantly separated responders (R) from non-responders (NR) in the pilot clinical cohort (all patients, mean predicted radiosensitivity index, R vs. NR 0.34 vs. 0.48, p=0.002). Importantly, the model was accurate in both disease cohorts in spite of the small number of patients (rectal cancer patients, mean predicted radiosensitivity index, R vs. NR 0.32 vs. 0.46, p=0.03) (esophageal cancer patients, mean predicted radiosensitivity index, R vs. NR 0.37 vs. 0.50, p=0.05).
• To further describe the model, an ROC curve (FIG. 2) was generated using the predicted radiosensitivity index values to determine the sensitivity and specificity of the predictor. Using a threshold RSI value of 0.46, the model has a sensitivity and specificity of 80 and 82% respectively, with a positive predictive value (PPV) of 86%. In addition, there were 8 patients that experienced a complete pathological response in the two cohorts. 6/8 complete responders had a predicted radiosensitivity index below the threshold. These numbers are encouraging since the predictor was not developed to account for the radiosensitizing effect of chemotherapy and the inclusion of chemotherapy was expected to account for prediction inaccuracies.
• These results show that RSI when analyzed as a continuous variable is correlated with pathological response in rectal and esophageal cancer patients treated with preoperative concurrent chemoradiation.
• It is important to note that false negatives (predicted radioresistant that responded) were the main inaccuracy when the model was dichotomized in the esophageal and rectal datasets. This population represented 60% of the misclassified cases in the esophageal and rectal cancer cohorts. It is possible that this inaccuracy is due to the radiosensitization effect of chemotherapy. The proportion of individuals within the rectal and esophageal dataset that are classified in this group (11.5%) is consistent with the observed improvement in clinical responses with concurrent chemotherapy over radiotherapy alone (Herskovic et al., The New England Journal of Medicine 1992; 326(24):1593-8; Al-Sarraf et al., J Clin Oncol 1997; 15(1):277-84 (published erratum appears in J Clin Oncol 1997 February; 15(2):866); Bosset et al., The New England journal of medicine 2006; 355(11): 1114-23). Therefore, it is possible that this effect can be addressed by analyzing differences between responders and non-responders that share a predicted radioresistant phenotype.
Example 4The Radiosensitivity Predictive Model is of Prognostic Value in Head and Neck Cancer
• The model was further tested as a prognostic marker in locally-advanced head and neck cancer patients treated with definitive concurrent radiochemotherapy.
• The Head and Neck Cancer Cohort consisted of 92 patients with head and neck cancer treated within prospective randomized Phase II-III trials at The Netherlands Cancer Institute. The majority of tumors were locally-advanced advanced (94% T3 and above, 74% N1 and above). The full clinical details of this cohort were previously published (Pramana et al., Int J Radiat Oncol Biol Phys 2007; 69(5):1544-52). All patients were treated with concurrent radiochemotherapy with cisplatin-based chemotherapy. Total radiation dose was 70Gy in 2Gy daily fractions in all cases. Two different schedules of cisplatin were given: 1. (high dose) 100 mg/m²IV three times during radiotherapy or 150 mg/m²given intra-arterially four times during radiotherapy; or 2. (low dose) 20×6 mg/m²daily. No disease outcome differences were found between chemotherapy schedules.
• Gene expression profiles for all patients were generated using the NKI array. These methods were previously published, see Pramana et al., Int J Radiat Oncol Biol Phys 2007; 69(5):1544-52. Probes were mapped from the HU6800 platform to the HG-U133 Plus 2.0 platform and NKI array format by mapping the probe sequences onto a corresponding NCBI refseq ID or genomic region, then identifying the closest probe match on the new microarray platform.
• Using the same algorithm developed in cell lines and tested in the rectal and esophageal cohorts, radiosensitivity predictions were generated for this dataset. The average radiosensitivity index prediction was lower in this disease site when compared with rectal and esophagus (predicted radiosensitivity index, head and neck vs. esophagus vs. rectal 0.06 vs. 0.43 vs. 0.39). Although this could be partly a function of radiosensitivity differences between these diseases, it could also be due to platform differences (Affymetrix U133 Plus vs. NKI array). In spite of these differences, the radiosensitivity index was still of prognostic value within the head and neck dataset. The predicted radiosensitive group had an improved 2 year Relapse-Free survival (2 yr RFS 86% vs. 62%, p=0.06), thus arguing that the model is capturing biological commonalties that determine tumor radiosensitivity across disease sites (FIG. 3).
• These results show that RSI is of prognostic significance in a cohort of 92 patients with locally-advanced head and neck cancer. The applicability of the model in three different disease sites strongly suggests that the model captures commonalties that define radiosensitivity across disease sites. Therefore the model should be generally applicable to other disease sites (e.g., lung, prostate, or cervical cancer).
• In addition, as noted above the gene expression in the head and neck dataset was derived from NKI arrays, which is a two channel based cDNA microarray platform, while the gene expression data in the esophageal and rectal cancer cohorts were derived from Affymetrix U-133 Plus microarrays. This indicates that the algorithm is transferable across platforms.
Example 5Identification of Subsets of Genes Significantly Associated with Radiation Sensitivity
• To determine whether all 10 of the above-described signature genes were necessary for a robust prediction, subset analysis was performed using the methods described herein.
• Considering the 10 signature genes, subsets of these genes were selected and tested for statistical significance in the patient cohorts described earlier. For each subset, the gene expression data was rank ordered and a linear regression model was built. The coefficients and ranks of these models differ from the 10-gene model. Each new model was evaluated by generating RSI predictions on the esophageal and rectal cancer patient cohort and using a one-sided Wilcox test for significant difference in RSI between responders and non-responders. In addition, a one-sided Student's t test was also used to assess statistical significance.
• Likewise, the RSI predictions were generated for the head and neck cancer patient set. Here, the 25th percentile of predicted RSI was used as described above as a threshold for calling a patient's tumor radiosensitive or radioresistant (for the purposes of time to local recurrence). A log-rank statistical test was performed on recurrence free survival times between the predicted radiosensitive and radioresistant groups to assess statistical significance. In addition, the mean predicted RSI was also used a threshold and evaluated.
• The rank-based approach to prediction does not allow single genes to be used. In addition, some subsets of two genes lead to identical ranking for all cell lines, thereby limiting the number of possible subsets to be evaluated.
• All of the potential gene combinations were evaluated in each of the patient cohorts described above. Table 10 and 11 show results for statistical significant subsets of genes (gene symbols joined by ‘_’) and p-values from tests of significance between responders and non-responders (in the manner described earlier). Many significant subsets were identified, ranging from subsets of 2 genes to 10 genes.
• In the head & neck trial there were 12 significant subsets of genes when considering the difference in recurrence free survival split at the 25th percentile of predicted RSI, which are listed in Table 10. Using the median of predicted RSI identified the gene subsets shown in Table 11.

• TABLE 10
  Subsets Significant in Head & Neck Cancer Cohort by 25thPercentile

  Number in Subset	Combination	25th Percentile

  4	STAT1_SUMO1_HDAC_IRF1	0.053506572
  5	AR_STAT1_SUMO1_HDAC_IRF1	0.050918175
  5	c-jun_STAT1_RelA_SUMO1_IRF1	0.059557472
  6	AR_c-jun_STAT1_RelA_SUMO1_IRF1	0.042062687
  6	AR_c-jun_STAT1_cAbl_SUMO1_IRF1	0.046890608
  6	AR_c-jun_STAT1_SUMO1_HDAC_IRF1	0.053117262
  6	c-jun_STAT1_RelA_cAbl_SUMO1_IRF1	0.059557472
  6	c-jun_STAT1_RelA_SUMO1_CDK1_IRF1	0.059557472
  7	AR_c-jun_STAT1_RelA_SUMO1_CDK1_IRF1	0.042062687
  7	c-jun_STAT1_PKC_RelA_cAbl_HDAC_IRF1	0.046900514
  8	AR_c-jun_STAT1_PKC_RelA_cAbl_HDAC_IRF1	0.032529876
  10	AR_c-jun_STAT1_PKC_RelA_cAbl_SUMO1_CDK1_HDAC_IRF1	0.062801635

• TABLE 11
  Subsets Significant in Head & Neck Cancer Cohort by Median

  Number inSubset	Combination	Median
  5	AR_PKC_cAbl_SUMO1_IRF1	0.039215827
  5	c-jun_STAT1_RelA_SUMO1_IRF1	0.059557472
  6	AR_c-jun_STAT1_RelA_SUMO1_IRF1	0.046890608
  6	AR_c-jun_STAT1_cAbl_SUMO1_IRF1	0.046890608
  6	c-jun_STAT1_RelA_cAbl_SUMO1_IRF1	0.059557472
  6	c-jun_STAT1_RelA_SUMO1_CDK1_IRF1	0.059557472
  7	AR_c-jun_STAT1_RelA_SUMO1_CDK1_IRF1	0.046890608
  7	AR_c-jun_STAT1_RelA_cAbl_SUMO1_IRF1	0.046890608
  8	AR_c-jun_STAT1_PKC_RelA_cAbl_SUMO1_IRF1	0.043331767

• The Rectal & Esophageal trial had 259 significant subsets, ranging from two hubs to all ten;FIG. 4 presents a list of those subsets.
• There were five gene subsets that generated RSI predictions that were of statistical significance in both the head/neck and rectal and esophageal cancer patient cohorts. All included c-jun, STAT1, cAbl, IRF1, and are listed in Table 12:

• TABLE 12
  Subsets Significant in Head & Neck, Esophageal, and Rectal
  Cancer Cohorts

  Number of Genes in
  Profile	Genes inProfile

  6	AR + c-jun + STAT1 + cAbl + SUMO1 + IRF1
  6	c-jun + STAT1 + RelA + cAbl + SUMO1 + IRF1
  7	c-jun + STAT1 + PKC + RelA + cAbl + HDAC +
  	IRF1
  8	AR + c-jun + STAT1 + PKC + RelA + cAbl +
  	HDAC +IRF1
  10	AR + c-jun + STAT1 + PKC + RelA + cAbl +
  	SUMO1 + CDK1 + HDAC + IRF1

• This indicates that these subsets of genes can be used in the present methods in place of the ten signature gene profile.
OTHER EMBODIMENTS
• It is to be understood that while the invention has been described in conjunction with the detailed description thereof, the foregoing description is intended to illustrate and not limit the scope of the invention, which is defined by the scope of the appended claims. Other aspects, advantages, and modifications are within the scope of the following claims.

Claims (52)

1. A method of predicting the sensitivity of a cell to a selected dose of radiation therapy, the method comprising:

assigning a radiation sensitivity index to the cell based on expression levels of two or more signature genes in the cell, wherein the two or more signature genes are selected from the group consisting of Androgen receptor (AR); Jun oncogene (c-Jun); Signal transducer and activator of transcription 1 (STAT1); Protein kinase C, beta (PRKCB or PKC); V-rel reticuloendotheliosis viral oncogene homolog A (avian) (RELA or p65); c-Abl oncogene 1, receptor tyrosine kinase (ABL1 or c-Abl); SMT3 suppressor of mif two 3 homolog 1 (S. cerevisiae) (SUMO1); CDK1 (p34); Histone deacetylase 1 (HDAC1); and Interferon regulatory factor 1 (IRF1);

wherein the radiation sensitivity index indicates whether the cell is sensitive to radiation therapy.

2. The method ofclaim 1, wherein assigning a radiation sensitivity index comprises applying a linear regression model to the gene expression levels.

3. The method ofclaim 2, wherein the model is a rank-based linear regression model.

4. The method ofclaim 1, wherein the cell is a tumor cell.

5. The method ofclaim 1, wherein the two or more signature genes are weighted.

6. The method ofclaim 5, wherein the linear regression model is represented by the following algorithm:

RSI=k₁*AR+k₂*c-jun+k₃*STAT1+k₄*PKC+k₅*RelA+k₆*cAbl+k₇*SUMO1+k₈*CDK1+k₉*HDAC+k₁₀*IRF1.  I

7. The method ofclaim 1, wherein the method is computer-implemented.

8. The method ofclaim 1, wherein assigning a radiation sensitivity index comprises:

determining a level of expression of the two or more genes in the cell.

9. The method ofclaim 1, wherein the signature genes comprise AR, c-jun, STAT1, cAbl, SUMO1, and IRF1.

10. The method ofclaim 1, wherein the signature genes comprise c-jun, STAT1, RelA, cAbl, SUMO1, and IRF1.

11. The method ofclaim 1, wherein the signature genes comprise c-jun, STAT1, PKC, RelA, cAbl, HDAC, and IRF1.

12. The method ofclaim 1, wherein the signature genes comprise AR, c-jun, STAT1, PKC, RelA, cAbl, HDAC, and IRF1.

13. The method ofclaim 1, wherein the signature genes comprise AR, c-jun, STAT1, PKC, RelA, cAbl, SUMO1, CDK1, HDAC, and IRF1.

14. A computer-implemented method of identifying genes associated with radiation sensitivity, the method comprising:

assigning a radiation sensitivity value to one or more populations of cells, wherein the radiation sensitivity value represents the sensitivity of the cells to a selected dose of radiation;

determining a level of gene expression in the one or more populations of cells for each of a plurality of genes; and

identifying a subset comprising two or more genes, the expression of which is correlated with the radiation sensitivity value.

15. The method ofclaim 14, wherein identifying a subset of the genes comprises applying a model representing gene expression and radiosensitivity.

16. The method ofclaim 15, wherein the model is a multivariate linear regression model.

17. The method ofclaim 15, wherein the model includes at least one coefficient representing one, two, or all three of: tissue of origin, ras status, or p53 status.

18. The method ofclaim 15, further comprising associating a classifier representing biological importance with each gene in the subset of genes, expression of which is correlated with the radiation sensitivity value, and selecting a second subset comprising two or more genes wherein the classifier representing biological importance is above a preselected threshold, thereby selecting a subset of biologically important genes.

19. The method ofclaim 18, wherein the model further comprises a variable representing an effect of administration of a treatment on expression of each gene in the subset of biologically important genes.

20. The method ofclaim 19, further comprising selecting a third subset comprising one or more genes based on the effect of administration of the treatment, thereby identifying a subset of therapeutic target genes.

21. The method ofclaim 20, wherein the effect of the treatment is an increase in radiosensitivity.

22. The method ofclaim 20, wherein the effect of the treatment is a decrease in radiosensitivity.

23. The method ofclaim 19, further comprising selecting a treatment that has an effect on radiosensitivity in the model.

24. A method of predicting the effect of radiation therapy on a tumor, the method comprising:

assigning a radiation sensitivity index to the tumor based on expression levels of two or more signature genes in a cell from the tumor, wherein the two or more signature genes are selected from the group consisting of Androgen receptor (AR); Jun oncogene (c-Jun); Signal transducer and activator of transcription 1 (STAT1); Protein kinase C, beta (PRKCB or PKC); V-rel reticuloendotheliosis viral oncogene homolog A (avian) (RELA or p65); c-Abl oncogene 1, receptor tyrosine kinase (ABL1 or c-Abl); SMT3 suppressor of mif two 3 homolog 1 (S. cerevisiae) (SUMO1); CDK1 (p34); Histone deacetylase 1 (HDAC1); and Interferon regulatory factor 1 (IRF1);

wherein the radiation sensitivity index indicates whether the radiation therapy is likely to be effective.

25. A method of assessing a tumor in a subject for a radiation therapy regimen, the method comprising:

assigning a radiation sensitivity index to the tumor based on expression levels of two or more signature genes in a cell from the tumor, wherein the two or more signature genes are selected from the group consisting of Androgen receptor (AR); Jun oncogene (c-Jun); Signal transducer and activator of transcription 1 (STAT1); Protein kinase C, beta (PRKCB or PKC); V-rel reticuloendotheliosis viral oncogene homolog A (avian) (RELA or p65); c-Abl oncogene 1, receptor tyrosine kinase (ABL1 or c-Abl); SMT3 suppressor of mif two 3 homolog 1 (S. cerevisiae) (SUMO1); CDK1 (p34); Histone deacetylase 1 (HDAC1); and Interferon regulatory factor 1 (IRF1);

wherein the radiation sensitivity index indicates whether the tumor in the subject should be treated with radiation therapy.

26. A method of selecting a treatment regimen for a subject having a tumor, the method comprising:

assigning a radiation sensitivity index to the tumor based on expression levels of two or more signature genes in a cell from the tumor, wherein the two or more signature genes are selected from the group consisting of Androgen receptor (AR); Jun oncogene (c-Jun); Signal transducer and activator of transcription 1 (STAT1); Protein kinase C, beta (PRKCB or PKC); V-rel reticuloendotheliosis viral oncogene homolog A (avian) (RELA or p65); c-Abl oncogene 1, receptor tyrosine kinase (ABL1 or c-Abl); SMT3 suppressor of mif two 3 homolog 1 (S. cerevisiae) (SUMO1); CDK1 (p34); Histone deacetylase 1 (HDAC1); and Interferon regulatory factor 1 (IRF1); and

selecting a treatment regimen for the subject based on the radiation sensitivity index.

27. The method ofclaim 26, wherein a radiation sensitivity index below a threshold indicates that radiation therapy is likely to be effecting in treating the tumor, and the method includes selecting a treatment regimen including radiation therapy.

28. The method ofclaim 26, wherein a radiation sensitivity index above a threshold indicates that radiation therapy is not likely to be effecting in treating the tumor, and the method includes selecting a treatment regimen excluding radiation therapy, or a treatment regime including a high dose of radiation therapy.

29. A method of selecting a dose of radiation to be administered to a subject having a tumor, the method comprising:

assigning to the tumor a radiation sensitivity index for a preselected dose of radiation based on expression levels of two or more signature genes in a cell from the tumor, wherein the two or more signature genes are selected from the group consisting of Androgen receptor (AR); Jun oncogene (c-Jun); Signal transducer and activator of transcription 1 (STAT1); Protein kinase C, beta (PRKCB or PKC); V-rel reticuloendotheliosis viral oncogene homolog A (avian) (RELA or p65); c-Abl oncogene 1, receptor tyrosine kinase (ABL1 or c-Abl); SMT3 suppressor of mif two 3 homolog 1 (S. cerevisiae) (SUMO1); CDK1 (p34); Histone deacetylase 1 (HDAC1); and Interferon regulatory factor 1 (IRF1); and

selecting a dose of radiation for the subject based on the radiation sensitivity index at the preselected dose of radiation.

30. The method ofclaim 29, comprising selecting a dose of radiation that is the same as or less than the preselected dose of radiation, if the radiation sensitivity index is below a threshold.

31. The method ofclaim 26, comprising selecting a dose of radiation that is the greater than the preselected dose of radiation, if the radiation sensitivity index is above a threshold.

32. A database comprising a plurality of records wherein each record includes data on the expression of at least two signature genes in a cell, wherein the two or more signature genes are selected from the group consisting of Androgen receptor (AR); c-Jun; STAT1; PKC; RelA (p65); c-Abl; SUMO-1; CDK1 (p34); HDAC1; and IRF1, and a value representing sensitivity of the cell to a selected dose of radiation.

33. The database ofclaim 29, further comprising data regarding the administration of a treatment to the cell.

34. The database ofclaim 29, wherein the treatment is the administration of chemotherapy.

35. The database ofclaim 29, wherein the database is in computer readable form.

36. A microarray comprising a substrate and a plurality of individually addressable hybridisable array elements arranged thereon, wherein the individually addressable hybridisable array elements are selective for at least two signature genes selected from the group consisting of Androgen Receptor (AR); c-Jun; STAT1; PKC; RelA (p65); c-Abl; SUMO-1; CDK1 (p34); HDAC1; and IRF1; and optionally at least one hybridisable array element selective for an internal normalization control gene.

37. The microarray ofclaim 33, wherein the plurality of hybridisable array elements consists of at least one element selective for each of the Androgen Receptor (AR); c-Jun; STAT1; PKC; RelA (p65); c-Abl; SUMO-1; CDK1 (p34); HDAC1; and IRF1.

38. A microfluidic device comprising a substrate and a plurality of reaction chambers, wherein the reaction chambers comprise reagents for selective quantification of at least two signature genes selected from the group consisting of Androgen Receptor (AR); c-Jun; STAT1; PKC; RelA (p65); c-Abl; SUMO-1; CDK1 (p34); HDAC1; and IRF1; and optionally at least one reaction chamber comprising reagents for selective quantification of an internal normalization control gene.

39. The microfluidic device ofclaim 35, wherein the plurality of reaction chambers comprise reagents for selective quantification of each of the Androgen Receptor (AR); c-Jun; STAT1; PKC; RelA (p65); c-Abl; SUMO-1; CDK1 (p34); HDAC1; IRF1; and at least one reaction chamber comprising reagents for selective quantification of an internal normalization control gene.

40. A medium bearing instructions to cause a computer to assign a radiation sensitivity index to a cell based on expression levels of two or more signature genes in the cell, wherein the two or more signature genes are selected from the group consisting of Androgen receptor (AR); Jun oncogene (c-Jun); Signal transducer and activator of transcription 1 (STAT1); Protein kinase C, beta (PRKCB or PKC); V-rel reticuloendotheliosis viral oncogene homolog A (avian) (RELA or p65); c-Abl oncogene 1, receptor tyrosine kinase (ABL1 or c-Abl); SMT3 suppressor of mif two 3 homolog 1 (S. cerevisiae) (SUMO1); CDK1 (p34); Histone deacetylase 1 (HDAC1); and Interferon regulatory factor 1 (IRF1).

41. The medium ofclaim 33, in which assigning a radiation sensitivity index comprises applying a linear regression model to the gene expression levels.

42. The medium ofclaim 33, wherein the model is a rank-based linear regression model.

43. The medium ofclaim 33, wherein the signature genes consist of the Androgen Receptor (AR); c-Jun; STAT1; PKC; RelA (p65); c-Abl; SUMO-1; CDK1 (p34); HDAC 1; and IRF1.

44. The medium ofclaim 33, wherein the two or more signature genes are weighted.

45. A medium bearing instructions to cause a computer to:

assign a radiation sensitivity value to one or more populations of cells, wherein the radiation sensitivity value represents the sensitivity of the cells to a selected dose of radiation;

assign a level of gene expression in the one or more populations of cells for each of a plurality of genes; and

identify a subset comprising two or more genes, the expression of which is correlated with the radiation sensitivity value.

46. The medium ofclaim 42, wherein identifying a subset of the genes comprises applying a model representing gene expression and radiosensitivity.

47. The medium ofclaim 43, wherein the model is a multivariate linear regression model.

48. The medium ofclaim 43, wherein the model includes at least one coefficient representing one, two, or all three of: tissue of origin, ras status, or p53 status.

49. The medium ofclaim 43, wherein the medium further comprises instructions to cause a computer to associate a classifier representing biological importance with each gene in the subset of genes, expression of which is correlated with the radiation sensitivity value, and to select a second subset comprising two or more genes wherein the classifier representing biological importance is above a preselected threshold, thereby selecting a subset of biologically important genes.

50. The medium ofclaim 46, wherein the model further comprises a variable representing an effect of administration of a treatment on expression of each gene in the subset of biologically important genes.

51. The medium ofclaim 47, further comprising selecting a third subset comprising two or more genes based on the effect of administration of the treatment, thereby identifying a subset of therapeutic target genes.

52. A kit comprising reagents for the specific quantification of gene expression levels of two or more signature genes in a cell, wherein the two or more signature genes are selected from the group consisting of Androgen receptor (AR); Jun oncogene (c-Jun); Signal transducer and activator of transcription 1 (STAT1); Protein kinase C, beta (PRKCB or PKC); V-rel reticuloendotheliosis viral oncogene homolog A (avian) (RELA or p65); c-Abl oncogene 1, receptor tyrosine kinase (ABL1 or c-Abl); SMT3 suppressor of mif two 3 homolog 1 (S. cerevisiae) (SUMO1); CDK1 (p34); Histone deacetylase 1 (HDAC1); and Interferon regulatory factor 1 (IRF1).

Images