Patent Application
20090105167
Cancer therapeutics are often effective only in a subset of patients. In addition, chemotherapeutic drugs often have toxic side effects. To address this problem, it will be useful to predict which cancer therapeutics will be effective for a given patient. This invention relates to a gene predictor set wherein altered expression of certain genes is correlated with high or low responsiveness to chemotherapeutic drugs. A tumor sample is collected from a patient and its gene expression profile is determined. This profile is then compared to a gene predictor set. This comparison allows one to select the therapy that is most likely to be effective for the individual patient.
Numerous advances in the development, selection, and application of chemotherapy agents, sometimes with remarkable successes as seen in the case of treatment for lymphomas or platinum-based therapy for testicular cancers (Herbst, R. S. et al. Clinical Cancer Advances 2005; major research advances in cancer treatment, prevention, and screening—a report from the American Society of Clinical Oncology. J. Clin. Oncol. 24, 190-205 (2006)). In addition, in several instances, combination chemotherapy in the adjuvant setting has been found to be curative. However, most patients with clinically or pathologically advanced solid tumors will relapse and die of their disease. Moreover, administration of ineffective chemotherapy increases the probability of side-effects, particularly from cytotoxic agents, and consequently a decrease in quality of life (Herbst, R. S. et al. Clinical Cancer Advances 2005; major research advances in cancer treatment, prevention, and screening—a report from the American Society of Clinical Oncology. J. Clin. Oncol. 24, 190-205 (2006), Breathnach, O. S. et al. Twenty-two years of phase III trials for patients with advanced non-small-cell lung cancer: sobering results. J. Clin. Oncol. 19, 1734-1742 (2001).).
Recent work has demonstrated the value in the use of biomarkers to select patients for various targeted therapeutics including tamoxifen, trastuzumab, and imatinib mesylate. In contrast, equivalent tools to select those patients most likely to respond to the commonly used chemotherapeutic drugs are lacking. A thorough understanding of drug resistance mechanisms should provide insight into how best to overcome resistance and, more importantly, the development of a strategy to match patients with drugs to which they are most likely to be sensitive and/or identify appropriate drug combinations for individual patient/patient groups is critical.
Throughout this specification, reference numbering is sometimes used to refer to the full citation for the references, which can be found in the “Reference Bibliography” after the Examples section. The disclosure of all patents, patent applications, and publications cited herein are hereby incorporated by reference in their entirety for all purposes.
In one aspect, the invention provides a method of identifying an effective cancer therapy agent for an individual with a platinum-resistant tumor, comprising: (a) obtaining a cellular sample from the individual; (b) analyzing said sample to obtain a first gene expression profile; (c) comparing said first gene expression profile to a platinum chemotherapy responsivity predictor set of gene expression profiles to identify whether said individual will be responsive to a platinum-based therapy; (d) if said individual is an incomplete responder to platinum based therapy, then comparing the first gene expression profile to a set of gene expression profiles comprising at least 5 genes from Table 1 that is capable of predicting responsiveness to other cancer therapy agents; thereby identifying whether said individual would benefit from the administration of one or more cancer therapy agents wherein said cancer therapy agents are not platinum-based.
In another aspect, the invention provides a method of treating an individual with ovarian cancer comprising: (a) obtaining a cellular sample from the individual; (b) analyzing said sample to obtain a first gene expression profile; (c) comparing said first gene expression profile to a platinum chemotherapy responsivity predictor set of gene expression profiles to identify whether said individual will be responsive to a platinum-based therapy; (d) if said individual is a complete responder or incomplete responder, then administering an effective amount of platinum-based therapy to the individual; (e) if said individual is predicted to be an incomplete responder to platinum based therapy, then comparing the first gene expression profile to a set of gene expression profiles comprising at least 5 genes from Table 1 that is predictive of responsivity to additional cancer therapeutics to identify to which additional cancer therapeutic the individual would be responsive; and (f) administering to said individual an effective amount of one or more of the additional cancer therapeutic that was identified in step (e); thereby treating the individual with ovarian cancer.
In certain embodiments, the cellular sample is taken from a tumor sample or ascites. In certain embodiments the set of gene expression profiles that is capable of predicting responsiveness to salvage therapy agents comprises at least 10 or 15 genes from Table 1. The cancer therapy agent may be a salvage therapy agent. In addition, the salvage therapy agent may be selected from the group consisting of topotecan, adriamycin, doxorubicin, cytoxan, cyclophosphamide, gemcitabine, etoposide, ifosfamide, paclitaxel, docetaxel, and taxol. Furthermore, the cancer therapy agent may target a signal transduction pathway that is deregulated. The cancer therapy agent may be selected from the group consisting of inhibitors of the Src pathway, inhibitors of the E2F3 pathway, inhibitors of the Myc pathway, and inhibitors of the beta-catenin pathway. In one embodiment, the platinum-based therapy is administered first, followed by the administration of one or more salvage therapy agent. The platinum-based therapy may also be administered concurrently with one or more salvage therapy agent. One or more salvage therapy agent may be administered by itself. Alternatively, the salvage therapy agent may be administered first, followed by the administration of one or more platinum-based therapy.
In yet another aspect, the invention provides for a gene chip for predicting an individual's responsivity to a salvage therapy agent comprising the gene expression profile of at least 5 genes selected from Table 1.
In yet another aspect, the invention provides for a gene chip for predicting an individual's responsivity to a salvage therapy agent comprising the gene expression profile of at least 10 genes selected from Table 1.
In yet another aspect, the invention provides for a gene chip for predicting an individual's responsivity to a salvage therapy agent comprising the gene expression profile of at least 20 genes selected from Table 1.
In yet another aspect, the invention provides for a kit comprising a gene chip for predicting an individual's responsivity to a salvage therapy agent and a set of instructions for determining an individual's responsivity to salvage chemotherapy agents.
In yet another aspect, the invention provides for a computer readable medium comprising gene expression profiles comprising at least 5 genes from any of Table 1.
In yet another aspect, the invention provides for a computer readable medium comprising gene expression profiles comprising at least 15 genes from Table 5.
In yet another aspect, the invention provides for a computer readable medium comprising gene expression profiles comprising at least 25 genes from Table 5.
In yet another aspect, the invention provides a method for estimating or predicting the efficacy of a therapeutic agent in treating an individual afflicted with cancer. In one aspect, the method comprises: (a) determining the expression level of multiple genes in a tumor biopsy sample from the subject; (b) defining the value of one or more metagenes from the expression levels of step (a), wherein each metagene is defined by extracting a single dominant value using singular value decomposition (SVD) from a cluster of genes associated tumor sensitivity to the therapeutic agent; and (c) averaging the predictions of one or more statistical tree models applied to the values of the metagenes, wherein each model includes one or more nodes, each node representing a metagene, each node including a statistical predictive probability of tumor sensitivity to the therapeutic agent, wherein at least one of the metagenes comprises at least 3 genes in metagenes 1, 2, 3, 4, 5, 6, or 7, thereby estimating the efficacy of a therapeutic agent in an individual afflicted with cancer. In certain embodiments, step (a) comprises extracting a nucleic acid sample from the sample from the subject. In certain embodiments, the method further comprising: (d) detecting the presence of pathway deregulation by comparing the expression levels of the genes to one or more reference profiles indicative of pathway deregulation, and (e) selecting an agent that is predicted to be effective and regulates a pathway deregulated in the tumor. In certain embodiments said pathway is selected from RAS, SRC, MYC, E2F, and β-catenin pathways.
In yet another aspect, the invention provides a method for estimating the efficacy of a therapeutic agent in treating an individual afflicted with cancer. In one aspect, the method comprises (a) determining the expression level of multiple genes in a tumor biopsy sample from the subject; (b) defining the value of one or more metagenes from the expression levels of step (a), wherein each metagene is defined by extracting a single dominant value using singular value decomposition (SVD) from a cluster of genes associated tumor sensitivity to the therapeutic agent; and (c) averaging the predictions of one or more binary regression models applied to the values of the metagenes, wherein each model includes a statistical predictive probability of tumor sensitivity to the therapeutic agent, wherein at least one of the metagenes comprises at least 3 genes in metagene 1, 2, 3, 4, 5, 6, or 7, thereby estimating the efficacy of a therapeutic agent in an individual afflicted with cancer.
In yet another aspect, the invention provides a method of treating an individual afflicted with cancer, said method comprising: (a) estimating the efficacy of a plurality of therapeutic agents in treating an individual afflicted with cancer according to the methods if the invention; (b) selecting a therapeutic agent having the high estimated efficacy; and (c) administering to the subject an effective amount of the selected therapeutic agent, thereby treating the subject afflicted with cancer. The method of estimating the efficacy may comprise (i) determining the expression level of multiple genes in a tumor biopsy sample from the subject and (ii) averaging the predictions of one or more statistical tree models applied to the values of one or more of metagenes 1, 2, 3, 4, 5, 6, and 7, wherein each model includes one or more nodes, each node representing a metagene, each node including a statistical predictive probability of tumor sensitivity to the therapeutic agent.
In yet another aspect, the invention provides a therapeutic agent having the high estimated efficacy is one having an estimated efficacy in treating the subject of at least 50%. In certain embodiments, the invention provides a therapeutic agent having the high estimated efficacy is one having an estimated efficacy in treating the subject of at least 80%.
In certain embodiments, the tumor is selected from a breast tumor, an ovarian tumor, and a lung tumor. In certain embodiments, the therapeutic agent is selected from docetaxel, paclitaxel, topotecan, adriamycin, etoposide, fluorouracil (5-FU), and cyclophosphamide, or any combination thereof.
In certain embodiments, the therapeutic agent is docetaxel and wherein the cluster of genes comprises at least 10 genes from metagene 1. In certain embodiments, the therapeutic agent is paclitaxel, and wherein the cluster of genes comprises at least 10 genes from metagene 2. In certain embodiments, wherein the therapeutic agent is topotecan, and wherein the cluster of genes comprises at least 10 genes from metagene 3. In certain embodiments, wherein the therapeutic agent is adriamycin, and wherein the cluster of genes comprises at least 10 genes from metagene 4. In certain embodiments, wherein the therapeutic agent is etoposide, and wherein the cluster of genes comprises at least 10 genes from metagene 5. In certain embodiments, wherein the therapeutic agent is fluorouracil (5-FU), and wherein the cluster of genes comprises at least 10 genes from metagene 6. In certain embodiments, wherein the therapeutic agent is cyclophosphamide and wherein the cluster of genes comprises at least 10 genes from metagene 7.
In certain embodiments, at least one of the metagenes is metagene 1, 2, 3, 4, 5, 6, or 7. In certain embodiments, the cluster of genes corresponding to at least one of the metagenes comprises 3 or more genes in common to metagene 1, 2, 3, 4, 5, 6, or 7. In certain embodiments, the cluster of genes corresponding to at least one metagene comprises 5 or more genes in common to metagene 1, 2, 3, 4, 5, 6, or 7. In certain embodiments, the cluster of genes corresponding to at least one metagene comprises at least 10 genes, wherein half or more of the genes are common to metagene 1, 2, 3, 4, 5, 6, or 7.
In certain embodiments, each cluster of genes comprises at least 3 genes. In certain embodiments, each cluster of genes comprises at least 5 genes. In certain embodiments, each cluster of genes comprises at least 7 genes. In certain embodiments, each cluster of genes comprises at least 10 genes. In certain embodiments, each cluster of genes comprises at least 12 genes. In certain embodiments, each cluster of genes comprises at least 15 genes. In certain embodiments, each cluster of genes comprises at least 20 genes.
In certain embodiments, a nucleic acid sample is extracted from a subject. In certain embodiments, the expression level of multiple genes in the tumor biopsy sample is determined by quantitating nucleic acids levels of the multiple genes using a DNA microarray.
In certain embodiments, at least one of the metagenes shares at least 3 of its defining genes in common with metagene 1, 2, 3, 4, 5, 6, or 7. In certain embodiments, at least one of the metagenes shares at least 50% of its defining genes in common with metagene 1, 2, 3, 4, 5, 6, or 7. In certain embodiments, at least one of the metagenes shares at least 75% of its defining genes in common with metagene 1, 2, 3, 4, 5, 6, or 7. In certain embodiments, at least one of the metagenes shares at least 90% of its defining genes in common with metagene 1, 2, 3, 4, 5, 6, or 7. In certain embodiments, at least one of the metagenes shares at least 95% of its defining genes in common with metagene 1, 2, 3, 4, 5, 6, or 7. In certain embodiments, at least one of the metagenes shares at least 98% of its defining genes in common with metagene 1, 2, 3, 4, 5, 6, or 7.
In certain embodiments, the cluster of genes for at least two of the metagenes share at least 50% of their genes in common with one of metagenes 1, 2, 3, 4, 5, 6, or 7. In certain embodiments, the cluster of genes for at least two of the metagenes share at least 75% of their genes in common with one of metagenes 1, 2, 3, 4, 5, 6, or 7. In certain embodiments, the cluster of genes for at least two of the metagenes share at least 90% of their genes in common with one of metagenes 1, 2, 3, 4, 5, 6, or 7. In certain embodiments, the cluster of genes for at least two of the metagenes share at least 95% of their genes in common with one of metagenes 1, 2, 3, 4, 5, 6, or 7. In certain embodiments, the cluster of genes for at least two of the metagenes share at least 98% of their genes in common with one of metagenes 1, 2, 3, 4, 5, 6, or 7.
In certain embodiments, the cluster of genes comprises at least 3 genes. In certain embodiments, the cluster of genes comprises at least 5 genes. In certain embodiments, the cluster of genes comprises at least 10 genes. In certain embodiments, the cluster of genes comprises at least 15 genes. In certain embodiments, the correlation-based clustering is Markov chain correlation-based clustering or K-means clustering.
The patent or application file contains at least one drawing executed in color. Copies of this patent or patent application publication with color drawing(s) will be provided by the Office upon request and payment of the necessary fee.
Table 1 lists the predictor set for commonly used chemotherapeutics.
Table 2 is a summary of the chemotherapy response predictors—validations in cell line and patient data sets.
Table 3 shows an enrichment analysis shows that a genomic-guided response prediction increases the probability of a clinical response in the different data sets studied.
Table 4 shows the accuracy of genomic-based chemotherapy response predictors is compared to previously reported predictors of response.
Table 5 lists the genes that constitute the predictor of PI3 kinase activation.
An individual who has cancer frequently has progressed to an advanced stage before any symptoms appear. The difficulty with administering one or more chemotherapeutic agents is that not all individuals with cancer will respond favorably to the chemotherapeutic agent selected by the physician. Frequently, the administration of one or more chemotherapeutic agents results in the individual becoming even more ill from the toxicity of the agent and the cancer still persists. Due to the cytotoxic nature of chemotherapeutic agents, the individual is physically weakened and his/her immunologically compromised system cannot generally tolerate multiple rounds of “trial and error” type of therapy. Hence a treatment plan that is personalized for the individual is highly desirable.
The inventors have described gene expression profiles associated with determining whether an individual afflicted with cancer will respond to a therapy, and in particular to a therapeutic agents such as salvage agents. This analysis has been coupled with gene expression signatures that reflect the deregulation of various oncogenic signaling pathways to identify unique characteristics of chemotherapeutic resistant cancers that can guide the use of these drugs in patients with chemotherapeutic resistant disease. The invention thus provides integrating gene expression profiles that predict chemotherapeutic response and oncogenic pathway status as a strategy for developing personalized treatment plans for individual patients.
“Platinum-based therapy” and “platinum-based chemotherapy” are used interchangeably herein and refers to agents or compounds that are associated with platinum.
As used herein, “array” and “microarray” are interchangeable and refer to an arrangement of a collection of nucleotide sequences in a centralized location. Arrays can be on a solid substrate, such as a glass slide, or on a semi-solid substrate, such as nitrocellulose membrane. The nucleotide sequences can be DNA, RNA, or any permutations thereof. The nucleotide sequences can also be partial sequences from a gene, primers, whole gene sequences, non-coding sequences, coding sequences, published sequences, known sequences, or novel sequences.
A “complete response” (CR) is defined as a complete disappearance of all measurable and assessable disease or, in the absence of measurable lesions, a normalization of the CA-125 level following adjuvant therapy. An individual who exhibits a complete response is known as a “complete responder.”
An “incomplete response” (IR) includes those who exhibited a “partial response” (PR), had “stable disease” (SD), or demonstrated “progressive disease” (PD) during primary therapy.
A “partial response” refers to a response that displays 50% or greater reduction in the product obtained from measurement of each bi-dimensional lesion for at least 4 weeks or a drop in the CA-125 by at least 50% for at least 4 weeks.
“Progressive disease” refers to response that is a 50% or greater increase in the product from any lesion documented within 8 weeks of initiation of therapy, the appearance of any new lesion within 8 weeks of initiation of therapy, or any increase in the CA-125 from baseline at initiation of therapy.
“Stable disease” was defined as disease not meeting any of the above criteria.
“Effective amount” refers to an amount of a chemotherapeutic agent that is sufficient to exert a biological effect in the individual. In most cases, an effective amount has been established by several rounds of testing for submission to the FDA. It is desirable for an effective amount to be an amount sufficient to exert cytotoxic effects on cancerous cells.
“Predicting” and “prediction” as used herein does not mean that the event will happen with 100% certainty. Instead it is intended to mean the event will more likely than not happen.
As used herein, “individual” and “subject” are interchangeable. A “patient” refers to an “individual” who is under the care of a treating physician. In one embodiment, the subject is a male. In one embodiment, the subject is a female.
The practice of the present invention will employ, unless otherwise indicated, conventional techniques of molecular biology (including recombinant techniques), microbiology, cell biology, biochemistry, nucleic acid chemistry, and immunology, which are well known to those skilled in the art. Such techniques are explained fully in the literature, such as, Molecular Cloning: A Laboratory Manual, second edition (Sambrook et al., 1989) and Molecular Cloning. A Laboratory Manual, third edition (Sambrook and Russel, 2001), (jointly referred to herein as “Sambrook”); Current Protocols in Molecular Biology (F. M. Ausubel et al., eds., 1987, including supplements through 2001); PCR: The Polymerase Chain Reaction, (Mullis et al., eds., 1994); Harlow and Lane (1988) Antibodies, A Laboratory Manual, Cold Spring Harbor Publications, New York; Harlow and Lane (1999) Using Antibodies. A Laboratory Manual Cold Spring Harbor Laboratory Press, Cold Spring Harbor, N.Y. (jointly referred to herein as “Harlow and Lane”), Beaucage et al. eds., Current Protocols in Nucleic Acid Chemistry John Wiley & Sons, Inc., New York, 2000) and Casarett and Doull's Toxicology The Basic Science of Poisons, C. Klaassen, ed., 6th edition (2001).
Gene expression profiles may be obtained from tumor samples taken during surgery to debulk individuals with ovarian cancer. It is also possible to generate a predictor set for predicting responsivity to common chemotherapy agents by using publicly available data. Numerous websites exist that share data obtained from microarray analysis. In one embodiment, gene expression profiling data obtained from analysis of 60 cancerous cells lines, known herein as NCI-60, can be used to generate a training set for predicting responsivity to cancer therapy agents. The NCI-60 training set can be validated by the same type of “Leave-one-out” cross-validation as described earlier.
The predictor sets for the other salvage therapy agents are shown in Table 1. The genes listed in Table 1 represent, to the best of Applicants' knowledge, a novel gene predictor set. The genes in the predictor set would not have been obvious to one of ordinary skill in the art. These predictor sets are used as a reference set to compare the first gene expression profile from an individual with ovarian cancer to determine if she will be responsive to a particular salvage agent. In certain embodiments, the methods of the application are performed outside of the human body.
Method of Treating Individuals with Ovarian Cancer
This methods described herein also include treating an individual afflicted with ovarian cancer. In the instance where the individual is predicted to be a non-responder to platinum-based therapy, a physician may decide to administer salvage therapy agent alone. In most instances, the treatment will comprise a combination of a platinum-based therapy and a salvage agent. In one embodiment, the treatment will comprise a combination of a platinum-based therapy and an inhibitor of a signal transduction pathway that is deregulated in the individual with ovarian cancer.
In one embodiment, the platinum-based therapy and a salvage agent are administered in an effective amount concurrently. In another embodiment, the platinum-based therapy and a salvage agent are administered in an effective amount in a sequential manner. In yet another embodiment, the salvage therapy agent is administered in an effective amount by itself. In yet another embodiment, the salvage therapy agent is administered in an effective amount first and then followed concurrently or step-wise by a platinum-based therapy.
Methods of Predicting/Estimating the Efficacy of a Therapeutic Agent in Treating a Individual Afflicted with Cancer
One aspect of the invention provides a method for predicting, estimating, aiding in the prediction of, or aiding in the estimation of, the efficacy of a therapeutic agent in treating a subject afflicted with cancer. In certain embodiments, the methods of the application are performed outside of the human body.
One method comprises (a) determining the expression level of multiple genes in a tumor biopsy sample from the subject; (b) defining the value of one or more metagenes from the expression levels of step (a), wherein each metagene is defined by extracting a single dominant value using singular value decomposition (SVD) from a cluster of genes associated tumor sensitivity to the therapeutic agent; and (c) averaging the predictions of one or more statistical tree models applied to the values of the metagenes, wherein each model includes one or more nodes, each node representing a metagene, each node including a statistical predictive probability of tumor sensitivity to the therapeutic agent, wherein at least one of the metagenes comprises at least 3 genes in metagenes 1, 2, 3, 4, 5, 6, or 7, thereby estimating the efficacy of a therapeutic agent in a subject afflicted with cancer. Another method comprises (a) determining the expression level of multiple genes in a tumor biopsy sample from the subject; (b) defining the value of one or more metagenes from the expression levels of step (a), wherein each metagene is defined by extracting a single dominant value using singular value decomposition (SVD) from a cluster of genes associated tumor sensitivity to the therapeutic agent; and (c) averaging the predictions of one or more binary regression models applied to the values of the metagenes, wherein each model includes a statistical predictive probability of tumor sensitivity to the therapeutic agent, wherein at least one of the metagenes comprises at least 3 genes in metagenes 1, 2, 3, 4, 5, 6, or 7, thereby estimating the efficacy of a therapeutic agent in a subject afflicted with cancer.
In one embodiment, the predictive methods of the invention predict the efficacy of a therapeutic agent in treating a subject afflicted with cancer with at least 70% accuracy. In another embodiment, the methods predict the efficacy of a therapeutic agent in treating a subject afflicted with cancer with at least 80% accuracy. In another embodiment, the methods predict the efficacy of a therapeutic agent in treating a subject afflicted with cancer with at least 85% accuracy. In another embodiment, the methods predict the efficacy of a therapeutic agent in treating a subject afflicted with cancer with at least 90% accuracy. In another embodiment, the methods predict the efficacy of a therapeutic agent in treating a subject afflicted with cancer with at least 70%, 80%, 85% or 90% accuracy when tested against a validation sample. In another embodiment, the methods predict the efficacy of a therapeutic agent in treating a subject afflicted with cancer with at least 70%, 80%, 85% or 90% accuracy when tested against a set of training samples. In another embodiment, the methods predict the efficacy of a therapeutic agent in treating a subject afflicted with cancer with at least 70%, 80%, 85% or 90% accuracy when tested on human primary tumors ex vivo or in vivo.
In one embodiment, the predictive methods of the invention comprise determining the expression level of genes in a tumor sample from the subject, preferably a breast tumor, an ovarian tumor, and a lung tumor. In one embodiment, the tumor is not a breast tumor. In one embodiment, the tumor is not an ovarian tumor. In one embodiment, the tumor is not a lung tumor. In one embodiment of the methods described herein, the methods comprise the step of surgically removing a tumor sample from the subject, obtaining a tumor sample from the subject, or providing a tumor sample from the subject. In one embodiment, the sample contains at least 40%, 50%, 60%, 70%, 80% or 90% tumor cells. In preferred embodiments, samples having greater than 50% tumor cell content are used. In one embodiment, the tumor sample is a live tumor sample. In another embodiment, the tumor sample is a frozen sample. In one embodiment, the sample is one that was frozen within less than 5, 4, 3, 2, 1, 0.75, 0.5, 0.25, 0.1, 0.05 or less hours after extraction from the patient. Preferred frozen sample include those stored in liquid nitrogen or at a temperature of about −80 C or below.
The expression of the genes may be determined using any methods known in the art for assaying gene expression. Gene expression may be determined by measuring mRNA or protein levels for the genes. In a preferred embodiment, an mRNA transcript of a gene may be detected for determining the expression level of the gene. Based on the sequence information provided by the GenBank™ database entries, the genes can be detected and expression levels measured using techniques well known to one of ordinary skill in the art. For example, sequences within the sequence database entries corresponding to polynucleotides of the genes can be used to construct probes for detecting mRNAs by, e.g., Northern blot hybridization analyses. The hybridization of the probe to a gene transcript in a subject biological sample can be also carried out on a DNA array. The use of an array is preferable for detecting the expression level of a plurality of the genes. As another example, the sequences can be used to construct primers for specifically amplifying the polynucleotides in, e.g., amplification-based detection methods such as reverse-transcription based polymerase chain reaction (RT-PCR). Furthermore, the expression level of the genes can be analyzed based on the biological activity or quantity of proteins encoded by the genes.
Methods for determining the quantity of the protein includes immunoassay methods. Paragraphs 98-123 of U.S. Patent Pub No. 2006-0110753 provide exemplary methods for determining gene expression. Additional technology is described in U.S. Pat. Nos. 5,143,854; 5,288,644; 5,324,633; 5,432,049; 5,470,710; 5,492,806; 5,503,980; 5,510,270; 5,525,464; 5,547,839; 5,580,732; 5,661,028; 5,800,992; as well as WO 95/21265; WO 96/31622; WO 97/10365; WO 97/27317; EP 373 203; and EP 785 280.
In one exemplary embodiment, about 1-50 mg of cancer tissue is added to a chilled tissue pulverizer, such as to a BioPulverizer H tube (Bio101 Systems, Carlsbad, Calif.). Lysis buffer, such as from the Qiagen Rneasy Mini kit, is added to the tissue and homogenized. Devices such as a Mini-Beadbeater (Biospec Products, Bartlesville, Okla.) may be used. Tubes may be spun briefly as needed to pellet the garnet mixture and reduce foam. The resulting lysate may be passed through syringes, such as a 21 gauge needle, to shear DNA. Total RNA may be extracted using commercially available kits, such as the Qiagen RNeasy Mini kit. The samples may be prepared and arrayed using Affymetrix U133 plus 2.0 GeneChips or Affymetrix U133A GeneChips.
In one embodiment, determining the expression level of multiple genes in a tumor sample from the subject comprises extracting a nucleic acid sample from the sample from the subject, preferably an mRNA sample. In one embodiment, the expression level of the nucleic acid is determined by hybridizing the nucleic acid, or amplification products thereof, to a DNA microarray. Amplification products may be generated, for example, with reverse transcription, optionally followed by PCR amplification of the products.
In one embodiment, the predictive methods of the invention comprise determining the expression level of all the genes in the cluster that define at least one therapeutic sensitivity/resistance determinative metagene. In one embodiment, the predictive methods of the invention comprise determining the expression level of at least 50%, 60%, 70%, 80%, 90%, 95%, 98%, 99% of the genes in each of the clusters that defines 1, 2, 3, 4 or 5 or more of metagenes 1, 2, 3, 4, 5, 6 and 7.
In one embodiment, at least 50%, 60%, 70%, 80%, 90%, 95%, 98%, 99% of the genes whose expression levels are determined to predict 5-FU sensitivity (or the genes in the cluster that define a metagene having said predictivity) are genes represented by the following symbols: LOC92755 (TUBB, LOC648765), CDKN2A, TRA@, GABRA3, COL1A2, ACTB, PDLIM4, ACTA2, FTSJ1, NBR1 (LOC727732), CFL1, ATP1A2, APOC4, KIAA1509, ZNF516, GRIK5, PDE5A, ARSF, ZC3H7B, WBP4, CSTB, TSPY1 (TSPY2, LOC653174, LOC728132, LOC728137, LOC728395, LOC728403, LOC728412), HTR2B, KBTBD11, SLC25A17, HMGN3, FIBP, IFT140, FAM63B, ZNF337, KIAA0100, FAM13C1, STK25, CPNE1, PEX19, EIF5B, EEF1A1 (APOLD1, LOC440595), SRR, THEM2, ID4, GGT1 (GGTL4), IFNA10, TUBB2A (TUBB4, TUBB2B), and TUBB3.
In one embodiment, at least 50%, 60%, 70%, 80%, 90%, 95%, 98%, 99% of the genes whose expression levels are determined to predict adriamycin sensitivity are genes represented by the following symbols: MLANA, CSPG4, DDR2, ETS2, EGFR, BIK, CD24, ZNF185, DSCR1, GSN, TPST1, LCN2, FAIM3, NCK2, PDZRN3, FKBP2, KRT8, NRP2, PKP2, CLDN3, CAPN1, STXBP1, LY96, WWC1, C10orf56, SPINT2, MAGED2, SYNGR2, SGCD, LAMC2, C19orf21, ZFHX1B, KRT18, CYBA, DSP, ID1, ID1, PSAP, ZNF629, ARHGAP29, ARHGAP8 (LOC553158), GPM6B, EGFR, CALU, KCNK1, RNF144, FEZ1, MEST, KLF5, CSPG4, FLNB, GYPC, SLC23A2, MITF, PITPNM1, GPNMB, PMP22, PLXNB3 (SRPK3), MIA, RAB40C, MAD2L1BP, PLOD3, VIL2, KLF9, PODXL, ATP6V1B2, SLC6A8, PLP1, KRT7, PKP3, DLG3, ZHX2, LAMA5, SASH1, GAS1, TACSTD1, GAS1, and CYP27A1.
In one embodiment, at least 50%, 60%, 70%, 80%, 90%, 95%, 98%, 99% of the genes whose expression levels are determined to predict cytoxan sensitivity (or the genes in the cluster that define a metagene having said predictivity) are genes represented by the following symbols: DAP3, RPS9, TTR, ACTB, MARCKS, GGT1 (GGT2), GGTL4, GGTLA4, LOC643171, LOC653590, LOC728226, LOC728441, LOC729838, LOC731629), FANCA, CDC42EP3, TSPAN4, C6orf145, ARNT2, KIF22 (LOC728037), NBEAL2, CAV1, SCRN1, SCHIP1, PHLDB1, AKAP12, ST5, SNAI2, ESD, ANP32B, CD59, ACTN1, CD59, PEG10, SMARCA1, GGCX, SAMD4A, CNN3, LPP, SNRPF, SGCE, CALD1, and C22orf5.
In one embodiment, at least 50%, 60%, 70%, 80%, 90%, 95%, 98%, 99% of the genes whose expression levels are determined to predict docetaxel sensitivity (or the genes in the cluster that define a metagene having said predictivity) are genes represented by the following symbols: BLR1, EIF4A2, FLT1, BAD, PIP5K3, BIN1, YBX1, BCKDK, DOHH, FOXD1, TEX261, NBR1 (LOC727732), APOA4, DDX5, TBCA, USP52, SLC25A36, CHP, ANKRD28, PDXK, ATP6AP1, SETD2, CCS, BRD2, ASPHD1, B4GALT6, ASL, CAPZA2, STARD3, LIMK2 (PPP1R14BP1), BANF1, GNB2, ENSA, SH3GL1, ACVR1B, SLC6A1, PPP2R1A, PCGF1, LOC643641, INPP5A, TLE1, PLLP, ZKSCAN1, TIAL1, TK1, PPP2R1A, and PSMB6.
In one embodiment, at least 50%, 60%, 70%, 80%, 90%, 95%, 98%, 99% of the genes whose expression levels are determined to predict etoposide sensitivity are genes represented by the following symbols: LIMK1, LIG3, AXL, IFI16, MMP14, GRB7, VAV2, FLT1, JUP, FN1, FN1, PKM2, LYPLA3, RFTN1, LAD1, SPINT1, CLDN3, PTRF, SPINT2, MMP14, FAAH, CLDN4, ST14, C19orf21, KIAA0506, LLGL2 (MADD), COBL, ZFHX1B, GBP1, IER2, PPL, TMEM30B, CNKSR1, CLDN7, BTN3A2, BTN3A2, TUBB2A, MAP7, HNRNPG-T, UGCG, GAK, PKP3, DFNA5, DAB2, TACSTD1, SPARC, and PPP2R5A.
In one embodiment, at least 50%, 60%, 70%, 80%, 90%, 95%, 98%, 99% of the genes whose expression levels are determined to predict taxol sensitivity (or the genes in the cluster that define a metagene having said predictivity) are genes represented by the following symbols: NR2F6, TOP2B, RARG, PCNA, PTPN11, ATM, NFATC4, CACNG1, C22orf31, PIK3R2, PRSS12, MYH8, SCCPDH, PHTF2, IQSEC2, TRPC3, TRAFD1, HEPH, SOX30, GATM, LMNA, HD, YIPF3, DNPEP, PCDH9, KLHDC3, SLC10A3, LHX2, CKS2, SECTM1, SF1, RPS6KA4, DYRK2, GDI2, and IFI30.
In one embodiment, at least 50%, 60%, 70%, 80%, 90%, 95%, 98%, 99% of the genes whose expression levels are determined to predict topotecan sensitivity (or the genes in the cluster that define a metagene having said predictivity) are genes represented by the following symbols: DUSP1, THBS1, AXL, RAP1GAP, QSCN6, IL1R1, TGFBI, PTX3, BLM, TNFRSF1A, FGF2, VEGFC, ACO2, FARSLA, RIN2, FGF2, RRAS, FIGF, MYB, CDH2, FGFR1, FGFR1, LAMC1, HIST1H4K (HIST1H4J), COL6A2, TMC6, PEA15, MARCKS, CKAP4, GJA1, FBN1, BASP1, BASP1, BTN2A1, ITGB1, DKFZP686A01247, MYLK, LOXL2, HEG1, DEGS1, CAP2, CAP2, PTGER4, BAI2, NUAK1, DLEU1 (SPANXC), RAB11FIP5, FSTL3, MYL6, VIM, GNA12, PRAF2, PTRF, CCL2, PLOD2, COL6A2, ATP5G3, GSR, NDUFS3, ST14, NID1, MYO1D, SDHB, CAV1, DPYSL3, PTRF, FBXL2, RIN2, PLEKHC1, CTGF, COL4A2, TPM1, TPM1, TPM1, FZD2, LOXL1, SYK, HADHA, TNFAIP1, NNMT, HPGD, MRC2, MEIS3P1, AOX1, SEMA3C, SEMA3C, SYNE1, SERPINE1, IL6, RRAS, GPD1L, AXL, WDR23, CLDN7, IL15, TNFAIP2, CYR61, LRP1, AMOTL2, PDE1B, SPOCK1, RA114, PXDN, COL4A1, CIR, KIAA0802 (C21orf57), C5orf13, TUFM, EDIL3, BDNF, PRSS23, ATP5A1, FRAT2, C16orf51, TUSC4, NUP50, TUBA3, NFIB, TLE4, AKT3, CRIM1, RAD23A, COX5A, SMCR7L, MXRA7, STARD7, STC1, TTC28, PLK2, TGDS, CALD1, OPTN, IFITM3, DFNA5, FGFR1, HTATIP, SYK, LAMB1, FZD2, SERPINE1, THBS1, CCL2, ITGA3, ITGA3, and UBE2A.
Table 1 shows the genes in the cluster that define metagenes 1-7 and indicates the therapeutic agent whose sensitivity it predicts. In one embodiment, at least 3, 5, 7, 9, 10, 12, 14, 16, 18, 20, 25, 30, 40 or 50 genes in the cluster of genes defining a metagene used in the methods described herein are common to metagene 1, 2, 3, 4, 5, 6 or 7, or to combinations thereof.
In one embodiment, the predictive methods of the invention comprise defining the value of one or more metagenes from the expression levels of the genes. A metagene value is defined by extracting a single dominant value from a cluster of genes associated with sensitivity to an anti-cancer agent, preferably an anti-cancer agent such as docetaxel, paclitaxel, topotecan, adriamycin, etoposide, fluorouracil (5-FU), and cyclophosphamide. In one embodiment, the agent is selected from alkylating agents (e.g., nitrogen mustards), antimetabolites (e.g., pyrimidine analogs), radioactive isotopes (e.g., phosphorous and iodine), miscellaneous agents (e.g., substituted ureas) and natural products (e.g., vinca alkyloids and antibiotics). In another embodiment, the therapeutic agent is selected from the group consisting of allopurinol sodium, dolasetron mesylate, pamidronate disodium, etidronate, fluconazole, epoetin alfa, levamisole HCL, amifostine, granisetron HCL, leucovorin calcium, sargramostim, dronabinol, mesna, filgrastim, pilocarpine HCL, octreotide acetate, dexrazoxane, ondansetron HCL, ondansetron, busulfan, carboplatin, cisplatin, thiotepa, melphalan HCL, melphalan, cyclophosphamide, ifosfamide, chlorambucil, mechlorethamine HCL, carmustine, lomustine, polifeprosan 20 with carmustine implant, streptozocin, doxorubicin HCL, bleomycin sulfate, daunirubicin HCL, dactinomycin, daunorucbicin citrate, idarubicin HCL, plimycin, mitomycin, pentostatin, mitoxantrone, valrubicin, cytarabine, fludarabine phosphate, floxuridine, cladribine, methotrexate, mercaptipurine, thioguanine, capecitabine, methyltestosterone, nilutamide, testolactone, bicalutamide, flutamide, anastrozole, toremifene citrate, estramustine phosphate sodium, ethinyl estradiol, estradiol, esterified estrogens, conjugated estrogens, leuprolide acetate, goserelin acetate, medroxyprogesterone acetate, megestrol acetate, levamisole HCL, aldesleukin, irinotecan HCL, dacarbazine, asparaginase, etoposide phosphate, gemcitabine HCL, altretamine, topotecan HCL, hydroxyurea, interferon alpha-2b, mitotane, procarbazine HCL, vinorelbine tartrate, E. coli L-asparaginase, Erwinia L-asparaginase, vincristine sulfate, denileukin diftitox, aldesleukin, rituximab, interferon alpha-2a, paclitaxel, docetaxel, BCG live (intravesical), vinblastine sulfate, etoposide, tretinoin, teniposide, porfimer sodium, fluorouracil, betamethasone sodium phosphate and betamethasone acetate, letrozole, etoposide citrororum factor, folinic acid, calcium leucouorin, 5-fluorouricil, adriamycin, cytoxan, and diamino-dichloro-platinum.
In a preferred embodiment, the dominant single value is obtained using single value decomposition (SVD). In one embodiment, the cluster of genes of each metagene or at least of one metagene comprises at least 3, 4, 5, 6, 7, 8, 9, 10, 12, 15, 18, 20 or 25 genes. In one embodiment, the predictive methods of the invention comprise defining the value of 2, 3, 4, 5, 6, 7, 8, 9 or 10 or more metagenes from the expression levels of the genes.
In preferred embodiments of the methods described herein, at least 1, 2, 3, 4, 5, 6, 7, 8 or 9 of the metagenes is metagene 1, 2, 3, 4, 5, 6, or 7. In one embodiment, at least one of the metagenes comprises 3, 4, 5, 6, 7, 8, 9 or 10 or more genes in common with any one of metagenes 1, 2, 3, 4, 5, 6, or 7. In one embodiment, a metagene shares at least 50%, 60%, 70%, 80%, 90%, 95%, 98%, 99% of the genes in its cluster in common with a metagene selected from 1, 2, 3, 4, 5, 6, or 7.
In one embodiment, the predictive methods of the invention comprise defining the value of 2, 3, 4, 5, 6, 7, 8 or more metagenes from the expression levels of the genes. In one embodiment, the cluster of genes from which any one metagene is defined comprises at least 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 22 or 25 genes.
In one embodiment, the predictive methods of the invention comprise defining the value of at least one metagene wherein the genes in the cluster of genes from which the metagene is defined, shares at least 50%, 60%, 70%, 80%, 90%, 95% or 98% of genes in common to any one of metagenes 1, 2, 3, 4, 5, 6, or 7. In one embodiment, the predictive methods of the invention comprise defining the value of at least two metagenes, wherein the genes in the cluster of genes from which each metagene is defined share at least 50%, 60%, 70%, 80%, 90%, 95% or 98% of genes in common to anyone of metagenes 1, 2, 3, 4, 5, 6, or 7. In one embodiment, the predictive methods of the invention comprise defining the value of at least three metagenes, wherein the genes in the cluster of genes from which each metagene is defined shares at least 50%, 60%, 70%, 80%, 90%, 95% or 98% of genes in common to anyone of metagenes 1, 2, 3, 4, 5, 6, or 7. In one embodiment, the predictive methods of the invention comprise defining the value of at least four metagenes, wherein the genes in the cluster of genes from which each metagene is defined shares at least 50%, 60%, 70%, 80%, 90%, 95% or 98% of genes in common to anyone of metagenes 1, 2, 3, 4, 5, 6, or 7. In one embodiment, the predictive methods of the invention comprise defining the value of at least five metagenes, wherein the genes in the cluster of genes from which each metagene is defined shares at least 50%, 60%, 70%, 80%, 90%, 95% or 98% of genes in common to anyone of metagenes 1, 2, 3, 4, 5, 6, or 7. In one embodiment, the predictive methods of the invention comprise defining the value of a metagene from a cluster of genes, wherein at least 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19 or 20 genes in the cluster are selected from the genes listed in Table 1.
In one embodiment, at least one of the metagenes is metagene 1, 2, 3, 4, 5, 6, or 7. In one embodiment, at least two of the metagenes are selected from metagenes 1, 2, 3, 4, 5, 6, or 7. In one embodiment, at least three of the metagenes are selected from metagenes 1, 2, 3, 4, 5, 6, or 7. In one embodiment, at least three of the metagenes are selected from metagenes 1, 2, 3, 4, 5, 6, or 7. In one embodiment, at least four of the metagenes are selected from metagenes 1, 2, 3, 4, 5, 6, or 7. In one embodiment, at least five or more of the metagenes are selected from metagenes 1, 2, 3, 4, 5, 6, or 7. In one embodiment of the methods described herein, one of the metagenes whose value is defined (i) is metagene 1 or (ii) shares at least 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12 or 13 genes in common with metagene 1. In one embodiment of the methods described herein, one of the metagenes whose value is defined (i) is metagene 2 or (ii) shares at least 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 or 12 genes in common with metagene 2. In one embodiment of the methods described herein, one of the metagenes whose value is defined (i) is metagene 3 or (ii) shares at least 2, 3 or 4 genes in common with metagene 3. In one embodiment of the methods described herein, one of the metagenes whose value is defined (i) is metagene 4 or (ii) shares at least 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24 or 25 genes in common with metagene 4. In one embodiment of the methods described herein, one of the metagenes whose value is defined (i) is metagene 5 or (ii) shares at least 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14 or 15 genes in common with metagene 5. In one embodiment of the methods described herein, one of the metagenes whose value is defined (i) is metagene 6 or (ii) shares at least 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12 or 13 genes in common with metagene 6. In one embodiment of the methods described herein, one of the metagenes whose value is defined (i) is metagene 7 or (ii) shares at least 2, 3, 4, 5, 6, 7, 8, 9 or 10 genes in common with metagene 7.
(E) Predictions from Tree Models
In one embodiment, the predictive methods of the invention comprise averaging the predictions of one or more statistical tree models applied to the metagenes values, wherein each model includes one or more nodes, each node representing a metagene, each node including a statistical predictive probability of sensitivity to an anti-cancer agent. The statistical tree models may be generated using the methods described herein for the generation of tree models. General methods of generating tree models may also be found in the art (See for example Pitman et al., Biostatistics 2004; 5:587-601; Denison et al. Biometrika 1999; 85:363-77; Nevins et al. Hum Mol Genet. 2003; 12:R153-7; Huang et al. Lancet 2003; 361:1590-6; West et al. Proc Natl Acad Sci USA 2001; 98:11462-7; U.S. Patent Pub. Nos. 2003-0224383; 2004-0083084; 2005-0170528; 2004-0106113; and U.S. application Ser. No. 11/198,782).
In one embodiment, the predictive methods of the invention comprise deriving a prediction from a single statistical tree model, wherein the model includes one or more nodes, each node representing a metagene, each node including a statistical predictive probability of sensitivity to an anti-cancer agent. In a preferred embodiment, the tree comprises at least 2 nodes. In a preferred embodiment, the tree comprises at least 3 nodes. In a preferred embodiment, the tree comprises at least 3 nodes. In a preferred embodiment, the tree comprises at least 4 nodes. In a preferred embodiment, the tree comprises at least 5 nodes.
In one embodiment, the predictive methods of the invention comprise averaging the predictions of one or more statistical tree models applied to the metagenes values, wherein each model includes one or more nodes, each node representing a metagene, each node including a statistical predictive probability of sensitivity to an anti-cancer agent. Accordingly, the invention provides methods that use mixed trees, where a tree may contain at least two nodes, where each node represents a metagene representative to the sensitivity/resistance to a particular agent.
In one embodiment, the statistical predictive probability is derived from a Bayesian analysis. In another embodiment, the Bayesian analysis includes a sequence of Bayes factor based tests of association to rank and select predictors that define a node binary split, the binary split including a predictor/threshold pair. Bayesian analysis is an approach to statistical analysis that is based on the Bayes law, which states that the posterior probability of a parameter p is proportional to the prior probability of parameter p multiplied by the likelihood of p derived from the data collected. This methodology represents an alternative to the traditional (or frequentist probability) approach: whereas the latter attempts to establish confidence intervals around parameters, and/or falsify a-priori null-hypotheses, the Bayesian approach attempts to keep track of how apriori expectations about some phenomenon of interest can be refined, and how observed data can be integrated with such a-priori beliefs, to arrive at updated posterior expectations about the phenomenon. Bayesian analysis have been applied to numerous statistical models to predict outcomes of events based on available data. These include standard regression models, e.g. binary regression models, as well as to more complex models that are applicable to multi-variate and essentially non-linear data.
Another such model is commonly known as the tree model which is essentially based on a decision tree. Decision trees can be used in clarification, prediction and regression. A decision tree model is built starting with a root mode, and training data partitioned to what are essentially the “children” nodes using a splitting rule. For instance, for clarification, training data contains sample vectors that have one or more measurement variables and one variable that determines that class of the sample. Various splitting rules may be used; however, the success of the predictive ability varies considerably as data sets become larger. Furthermore, past attempts at determining the best splitting for each mode is often based on a “purity” function calculated from the data, where the data is considered pure when it contains data samples only from one clan. Most frequently, used purity functions are entropy, gini-index, and towing rule. A statistical predictive tree model to which Bayesian analysis is applied may consistently deliver accurate results with high predictive capabilities.
Gene expression signatures that reflect the activity of a given pathway may be identified using supervised classification methods of analysis previously described (e.g., West, M. et al. Proc Natl Acad Sci USA 98, 11462-11467, 2001). The analysis selects a set of genes whose expression levels are most highly correlated with the classification of tumor samples into sensitivity to an anti-cancer agent versus no sensitivity to an anti-cancer agent. The dominant principal components from such a set of genes then defines a relevant phenotype-related metagene, and regression models assign the relative probability of sensitivity to an anti-cancer agent.
In one embodiment, the methods for defining one or more statistical tree models predictive of cancer sensitivity to an anti-cancer agent comprise identifying clusters of genes associated with metastasis by applying correlation-based clustering to the expression level of the genes. In one embodiment, the clusters of genes that define each metagene are identified using supervised classification methods of analysis previously described. See, for example, West, M. et al. Proc Natl Acad Sci USA 98, 11462-11467 (2001). The analysis selects a set of genes whose expression levels are most highly correlated with the classification of tumor samples into sensitivity to an anti-cancer agent versus no sensitivity to an anti-cancer agent. The dominant principal components from such a set of genes then defines a relevant phenotype-related metagene, and regression models assign the relative probability of sensitivity to an anti-cancer agent.
In one embodiment, identification of the clusters comprises screening genes to reduce the number by eliminating genes that show limited variation across samples or that are evidently expressed at low levels that are not detectable at the resolution of the gene expression technology used to measure levels. This removes noise and reduces the dimension of the predictor variable. In one embodiment, identification of the clusters comprises clustering the genes using k-means, correlated-based clustering. Any standard statistical package may be used, such as the xcluster software created by Gavin Sherlock (http://genetics.stanford.edu/˜sherlock/cluster.html). A large number of clusters may be targeted so as to capture multiple, correlated patterns of variation across samples, and generally small numbers of genes within clusters. In one embodiment, identification of the clusters comprises extracting the dominant singular factor (principal component) from each of the resulting clusters. Again, any standard statistical or numerical software package may be used for this; this analysis uses the efficient, reduced singular value decomposition function. In one embodiment, the foregoing methods comprise defining one or more metagenes, wherein each metagene is defined by extracting a single dominant value using single value decomposition (SVD) from a cluster of genes associated with estimating the efficacy of a therapeutic agent in treating a subject afflicted with cancer.
In one embodiment, the methods for defining one or more statistical tree models predictive of cancer sensitivity to an anti-cancer agent comprise defining a statistical tree model, wherein the model includes one or more nodes, each node representing a metagene, each node including a statistical predictive probability of the efficacy of a therapeutic agent in treating a subject afflicted with cancer. This generates multiple recursive partitions of the sample into subgroups (the “leaves” of the classification tree), and associates Bayesian predictive probabilities of outcomes with each subgroup. Overall predictions for an individual sample are then generated by averaging predictions, with appropriate weights, across many such tree models. Iterative out-of-sample, cross-validation predictions are then performed leaving each tumor out of the data set one at a time, refitting the model from the remaining tumors and using it to predict the hold-out case. This rigorously tests the predictive value of a model and mirrors the real-world prognostic context where prediction of new cases as they arise is the major goal.
In one embodiment, a formal Bayes' factor measure of association may be used in the generation of trees in a forward-selection process as implemented in traditional classification tree approaches. Consider a single tree and the data in a node that is a candidate for a binary split. Given the data in this node, one may construct a binary split based on a chosen (predictor, threshold) pair (χ, τ) by (a) finding the (predictor, threshold) combination that maximizes the Bayes' factor for a split, and (b) splitting if the resulting Bayes' factor is sufficiently large. By reference to a posterior probability scale with respect to a notional 50:50 prior, Bayes' factors of 2.2, 2.9, 3.7 and 5.3 correspond, approximately, to probabilities of 0.9, 0.95, 0.99 and 0.995, respectively. This guides the choice of threshold, which may be specified as a single value for each level of the tree. Bayes' factor thresholds of around 3 in a range of analyses may be used. Higher thresholds limit the growth of trees by ensuring a more stringent test for splits.
In one non-limiting exemplary embodiment of generating statistical tree models, prior to statistical modeling, gene expression data is filtered to exclude probe sets with signals present at background noise levels, and for probe sets that do not vary significantly across tumor samples. A metagene represents a group of genes that together exhibit a consistent pattern of expression in relation to an observable phenotype. Each signature summarizes its constituent genes as a single expression profile, and is here derived as the first principal component of that set of genes (the factor corresponding to the largest singular value) as determined by a singular value decomposition. Given a training set of expression vectors (of values across metagenes) representing two biological states, a binary probit regression model may be estimated using Bayesian methods. Applied to a separate validation data set, this leads to evaluations of predictive probabilities of each of the two states for each case in the validation set. When predicting sensitivity to an anti-cancer agent from an Tumor sample, gene selection and identification is based on the training data, and then metagene values are computed using the principal components of the training data and additional expression data. Bayesian fitting of binary probit regression models to the training data then permits an assessment of the relevance of the metagene signatures in within-sample classification, and estimation and uncertainty assessments for the binary regression weights mapping metagenes to probabilities of relative pathway status. Predictions of sensitivity to an anti-cancer agent are then evaluated, producing estimated relative probabilities—and associated measures of uncertainty—of sensitivity to an anti-cancer agent across the validation samples. Hierarchical clustering of sensitivity to anti-cancer agent predictions may be performed using Gene Cluster 3.0 testing the null hypothesis, which is that the survival curves are identical in the overall population.
In one embodiment, the each statistical tree model generated by the methods described herein comprises 2, 3, 4, 5, 6 or more nodes. In one embodiment of the methods described herein for defining a statistical tree model predictive of sensitivity/resistance to a therapeutic, the resulting model predicts cancer sensitivity to an anti-cancer agent with at least 70%, 80%, 85%, or 90% or higher accuracy. In another embodiment, the model predicts sensitivity to an anti-cancer agent with greater accuracy than clinical variables. In one embodiment, the clinical variables are selected from age of the subject, gender of the subject, tumor size of the sample, stage of cancer disease, histological subtype of the sample and smoking history of the subject. In one embodiment, the cluster of genes that define each metagene comprise at least 3, 4, 5, 6, 7, 8, 9, 10, 12 or 15 genes. In one embodiment, the correlation-based clustering is Markov chain correlation-based clustering or K-means clustering.
One aspect of the invention provides methods of conducting a diagnostic business, including a business that provides a health care practitioner with diagnostic information for the treatment of a subject afflicted with cancer. One such method comprises one, more than one, or all of the following steps: (i) obtaining an tumor sample from the subject; (ii) determining the expression level of multiple genes in the sample; (iii) defining the value of one or more metagenes from the expression levels of step (ii), wherein each metagene is defined by extracting a single dominant value using single value decomposition (SVD) from a cluster of genes associated with sensitivity to an anti-cancer agent; (iv) averaging the predictions of one or more statistical tree models applied to the values, wherein each model includes one or more nodes, each node representing a metagene, each node including a statistical predictive probability of sensitivity to an anti-cancer agent, wherein at least one metagene is one of metagenes 1-7; and (v) providing the health care practitioner with the prediction from step (iv).
In one embodiment, obtaining a tumor sample from the subject is effected by having an agent of the business (or a subsidiary of the business) remove a tumor sample from the subject, such as by a surgical procedure. In another embodiment, obtaining a tumor sample from the subject comprises receiving a sample from a health care practitioner, such as by shipping the sample, preferably frozen. In one embodiment, the sample is a cellular sample, such as a mass of tissue. In one embodiment, the sample comprises a nucleic acid sample, such as a DNA, cDNA, mRNA sample, or combinations thereof, which was derived from a cellular tumor sample from the subject. In one embodiment, the prediction from step (iv) is provided to a health care practitioner, to the patient, or to any other business entity that has contracted with the subject.
In one embodiment, the method comprises billing the subject, the subject's insurance carrier, the health care practitioner, or an employer of the health care practitioner. A government agency, whether local, state or federal, may also be billed for the services. Multiple parties may also be billed for the service.
In some embodiments, all the steps in the method are carried out in the same general location. In certain embodiments, one or more steps of the methods for conducting a diagnostic business are performed in different locations. In one embodiment, step (ii) is performed in a first location, and step (iv) is performed in a second location, wherein the first location is remote to the second location. The other steps may be performed at either the first or second location, or in other locations. In one embodiment, the first location is remote to the second location. A remote location could be another location (e.g. office, lab, etc.) in the same city, another location in a different city, another location in a different state, another location in a different country, etc. As such, when one item is indicated as being “remote” from another, what is meant is that the two items are at least in different buildings, and may be at least one mile, ten miles, or at least one hundred miles apart. In one embodiment, two locations that are remote relative to each other are at least 1, 2, 3, 4, 5, 10, 20, 50, 100, 200, 500, 1000, 2000 or 5000 km apart. In another embodiment, the two locations are in different countries, where one of the two countries is the United States.
Some specific embodiments of the methods described herein where steps are performed in two or more locations comprise one or more steps of communicating information between the two locations. “Communicating” information means transmitting the data representing that information as electrical signals over a suitable communication channel (for example, a private or public network). “Forwarding” an item refers to any means of getting that item from one location to the next, whether by physically transporting that item or otherwise (where that is possible) and includes, at least in the case of data, physically transporting a medium carrying the data or communicating the data. The data may be transmitted to the remote location for further evaluation and/or use. Any convenient telecommunications means may be employed for transmitting the data, e.g., facsimile, modem, internet, etc.
In one specific embodiment, the method comprises one or more data transmission steps between the locations. In one embodiment, the data transmission step occurs via an electronic communication link, such as the internet. In one embodiment, the data transmission step from the first to the second location comprises experimental parameter data, such as the level of gene expression of multiple genes. In some embodiments, the data transmission step from the second location to the first location comprises data transmission to intermediate locations. In one specific embodiment, the method comprises one or more data transmission substeps from the second location to one or more intermediate locations and one or more data transmission substeps from one or more intermediate locations to the first location, wherein the intermediate locations are remote to both the first and second locations. In another embodiment, the method comprises a data transmission step in which a result from gene expression is transmitted from the second location to the first location.
In one embodiment, the methods of conducting a diagnostic business comprise the step of determining if the subject carries an allelic form of a gene whose presence correlates to sensitivity or resistance to a chemotherapeutic agent. This may be achieved by analyzing a nucleic acid sample from the patient and determining the DNA sequence of the allele. Any technique known in the art for determining the presence of mutations or polymorphisms may be used. The method is not limited to any particular mutation or to any particular allele or gene. For example, mutations in the epidermal growth factor receptor (EGFR) gene are found in human lung adenocarcinomas and are associated with sensitivity to the tyrosine kinase inhibitors gefitinib and erlotinib. (See, e.g., Yi et al. Proc Natl Acad Sci USA. 2006 May 16; 103(20):7817-22; Shimato et al. Neuro-oncol. 2006 April; 8(2):137-44). Similarly, mutations in breast cancer resistance protein (BCRP) modulate the resistance of cancer cells to BCRP-substrate anticancer agents (Yanase et al., Cancer Lett. 2006 Mar. 8; 234(1):73-80).
Arrays and microarrays which contain the gene expression profiles for determining responsivity to platinum-based therapy and/or responsivity to salvage agents are also encompassed within the scope of this invention. Methods of making arrays are well-known in the art and as such, do not need to be described in detail here.
Such arrays can contain the profiles of at least 5, 10, 15, 25, 50, 75, 100, 150, or 200 genes as disclosed in Table 1. Accordingly, arrays for detection of responsivity to particular therapeutic agents can be customized for diagnosis or treatment of ovarian cancer. The array can be packaged as part of kit comprising the customized array itself and a set of instructions for how to use the array to determine an individual's responsivity to a specific cancer therapeutic agent.
Also provided are reagents and kits thereof for practicing one or more of the above described methods. The subject reagents and kits thereof may vary greatly. Reagents of interest include reagents specifically designed for use in production of the above described metagene values.
One type of such reagent is an array probe of nucleic acids, such as a DNA chip, in which the genes defining the metagenes in the therapeutic efficacy predictive tree models are represented. A variety of different array formats are known in the art, with a wide variety of different probe structures, substrate compositions and attachment technologies. Representative array structures of interest include those described in U.S. Pat. Nos. 5,143,854; 5,288,644; 5,324,633; 5,432,049; 5,470,710; 5,492,806; 5,503,980; 5,510,270; 5,525,464; 5,547,839; 5,580,732; 5,661,028; 5,800,992; the disclosures of which are herein incorporated by reference; as well as WO 95/21265; WO 96/31622; WO 97/10365; WO 97/27317; EP 373 203; and EP 785 280.
The DNA chip is convenient to compare the expression levels of a number of genes at the same time. DNA chip-based expression profiling can be carried out, for example, by the method as disclosed in “Microarray Biochip Technology” (Mark Schena, Eaton Publishing, 2000). A DNA chip comprises immobilized high-density probes to detect a number of genes. Thus, the expression levels of many genes can be estimated at the same time by a single-round analysis. Namely, the expression profile of a specimen can be determined with a DNA chip. A DNA chip may comprise probes, which have been spotted thereon, to detect the expression level of the metagene-defining genes of the present invention. A probe may be designed for each marker gene selected, and spotted on a DNA chip. Such a probe may be, for example, an oligonucleotide comprising 5-50 nucleotide residues. A method for synthesizing such oligonucleotides on a DNA chip is known to those skilled in the art. Longer DNAs can be synthesized by PCR or chemically. A method for spotting long DNA, which is synthesized by PCR or the like, onto a glass slide is also known to those skilled in the art. A DNA chip that is obtained by the method as described above can be used estimating the efficacy of a therapeutic agent in treating a subject afflicted with cancer according to the present invention.
DNA microarray and methods of analyzing data from microarrays are well-described in the art, including in DNA Microarrays: A Molecular Cloning Manual, Ed. by Bowtel and Sambrook (Cold Spring Harbor Laboratory Press, 2002); Microarrays for an Integrative Genomics by Kohana (MIT Press, 2002); A Biologist's Guide to Analysis of DNA Microarray Data, by Knudsen (Wiley, John & Sons, Incorporated, 2002); DNA Microarrays: A Practical Approach, Vol. 205 by Schema (Oxford University Press, 1999); and Methods of Microarray Data Analysis II, ed. by Lin et al. (Kluwer Academic Publishers, 2002).
One aspect of the invention provides a gene chip having a plurality of different oligonucleotides attached to a first surface of the solid support and having specificity for a plurality of genes, wherein at least 50% of the genes are common to those of metagenes 1, 2, 3, 4, 5, 6 and/or 7. In one embodiment, at least 70%, 80%, 90% or 95% of the genes in the gene chip are common to those of metagenes 1, 2, 3, 4, 5, 6 and/or 7.
One aspect of the invention provides a kit comprising: (a) any of the gene chips described herein; and (b) one of the computer-readable mediums described herein.
In some embodiments, the arrays include probes for at least 2, 3, 4, 5, 6, 7, 8, 9, 10, 15, 20, 25, 30, 40, or 50 of the genes listed in Table 1. In certain embodiments, the number of genes that are from Table 1 that are represented on the array is at least 5, at least 10, at least 25, at least 50, at least 75 or more, including all of the genes listed in the table. Where the subject arrays include probes for additional genes not listed in the tables, in certain embodiments the number % of additional genes that are represented does not exceed about 50%, 40%, 30%, 20%, 15%, 10%, 8%, 6%, 5%, 4%, 3%, 2% or 1%. In some embodiments, a great majority of genes in the collection are genes that define the metagenes of the invention, where by great majority is meant at least about 75%, usually at least about 80% and sometimes at least about 85, 90, 95% or higher, including embodiments where 100% of the genes in the collection are metagene-defining genes.
The kits of the subject invention may include the above described arrays. The kits may further include one or more additional reagents employed in the various methods, such as primers for generating target nucleic acids, dNTPs and/or rNTPs, which may be either premixed or separate, one or more uniquely labeled dNTPs and/or rNTPs, such as biotinylated or Cy3 or Cy5 tagged dNTPs, gold or silver particles with different scattering spectra, or other post synthesis labeling reagent, such as chemically active derivatives of fluorescent dyes, enzymes, such as reverse transcriptases, DNA polymerases, RNA polymerases, and the like, various buffer mediums, e.g. hybridization and washing buffers, prefabricated probe arrays, labeled probe purification reagents and components, like spin columns, etc., signal generation and detection reagents, e.g. streptavidin-alkaline phosphatase conjugate, chemifluorescent or chemiluminescent substrate, and the like.
In addition to the above components, the subject kits will further include instructions for practicing the subject methods. These instructions may be present in the subject kits in a variety of forms, one or more of which may be present in the kit. One form in which these instructions may be present is as printed information on a suitable medium or substrate, e.g., a piece or pieces of paper on which the information is printed, in the packaging of the kit, in a package insert, etc. Yet another means would be a computer readable medium, e.g., diskette, CD, etc., on which the information has been recorded. Yet another means that may be present is a website address which may be used via the internet to access the information at a removed site. Any convenient means may be present in the kits.
The kits also include packaging material such as, but not limited to, ice, dry ice, styrofoam, foam, plastic, cellophane, shrink wrap, bubble wrap, paper, cardboard, starch peanuts, twist ties, metal clips, metal cans, drierite, glass, and rubber (see products available from www.papermart.com. for examples of packaging material).
The invention also contemplates computer readable media that comprises gene expression profiles. Such media can contain all of part of the gene expression profiles of the genes listed in Table 1. The media can be a list of the genes or contain the raw data for running a user's own statistical calculation, such as the methods disclosed herein.
Another aspect of the invention provides a program product (i.e., software product) for use in a computer device that executes program instructions recorded in a computer-readable medium to perform one or more steps of the methods described herein, such for estimating the efficacy of a therapeutic agent in treating a subject afflicted with cancer.
One aspect of the invention provides a computer readable medium having computer readable program codes embodied therein, the computer readable medium program codes performing one or more of the following functions: defining the value of one or more metagenes from the expression levels genes; defining a metagene value by extracting a single dominant value using singular value decomposition (SVD) from a cluster of genes associated tumor sensitivity to a therapeutic agent; averaging the predictions of one or more statistical tree models applied to the values of the metagenes; or averaging the predictions of one or more binary regression models applied to the values of the metagenes, wherein each model includes a statistical predictive probability of tumor sensitivity to a therapeutic agent.
Another related aspect of the invention provides kits comprising the program product or the computer readable medium, optionally with a computer system. One aspect of the invention provides a system, the system comprising: a computer; a computer readable medium, operatively coupled to the computer, the computer readable medium program codes performing one or more of the following functions: defining the value of one or more metagenes from the expression levels genes; defining a metagene value by extracting a single dominant value using singular value decomposition (SVD) from a cluster of genes associated tumor sensitivity to a therapeutic agent; averaging the predictions of one or more statistical tree models applied to the values of the metagenes; or averaging the predictions of one or more binary regression models applied to the values of the metagenes, wherein each model includes a statistical predictive probability of tumor sensitivity to a therapeutic agent.
In one embodiment, the program product comprises: a recordable medium; and a plurality of computer-readable instructions executable by the computer device to analyze data from the array hybridization steps, to transmit array hybridization from one location to another, or to evaluate genome-wide location data between two or more genomes. Computer readable media include, but are not limited to, CD-ROM disks (CD-R, CD-RW), DVD-RAM disks, DVD-RW disks, floppy disks and magnetic tape.
A related aspect of the invention provides kits comprising the program products described herein. The kits may also optionally contain paper and/or computer-readable format instructions and/or information, such as, but not limited to, information on DNA microarrays, on tutorials, on experimental procedures, on reagents, on related products, on available experimental data, on using kits, on chemotherapeutic agents including there toxicity, and on other information. The kits optionally also contain in paper and/or computer-readable format information on minimum hardware requirements and instructions for running and/or installing the software. The kits optionally also include, in a paper and/or computer readable format, information on the manufacturers, warranty information, availability of additional software, technical services information, and purchasing information. The kits optionally include a video or other viewable medium or a link to a viewable format on the internet or a network that depicts the use of the use of the software, and/or use of the kits. The kits also include packaging material such as, but not limited to, styrofoam, foam, plastic, cellophane, shrink wrap, bubble wrap, paper, cardboard, starch peanuts, twist ties, metal clips, metal cans, drierite, glass, and rubber.
The analysis of data, as well as the transmission of data steps, can be implemented by the use of one or more computer systems. Computer systems are readily available. The processing that provides the displaying and analysis of image data for example, can be performed on multiple computers or can be performed by a single, integrated computer or any variation thereof. For example, each computer operates under control of a central processor unit (CPU), such as a “Pentium” microprocessor and associated integrated circuit chips, available from Intel Corporation of Santa Clara, Calif., USA. A computer user can input commands and data from a keyboard and display mouse and can view inputs and computer output at a display. The display is typically a video monitor or flat panel display device. The computer also includes a direct access storage device (DASD), such as a fixed hard disk drive. The memory typically includes volatile semiconductor random access memory (RAM).
Each computer typically includes a program product reader that accepts a program product storage device from which the program product reader can read data (and to which it can optionally write data). The program product reader can include, for example, a disk drive, and the program product storage device can include a removable storage medium such as, for example, a magnetic floppy disk, an optical CD-ROM disc, a CD-R disc, a CD-RW disc and a DVD data disc. If desired, computers can be connected so they can communicate with each other, and with other connected computers, over a network. Each computer can communicate with the other connected computers over the network through a network interface that permits communication over a connection between the network and the computer.
The computer operates under control of programming steps that are temporarily stored in the memory in accordance with conventional computer construction. When the programming steps are executed by the CPU, the pertinent system components perform their respective functions. Thus, the programming steps implement the functionality of the system as described above. The programming steps can be received from the DASD, through the program product reader or through the network connection. The storage drive can receive a program product, read programming steps recorded thereon, and transfer the programming steps into the memory for execution by the CPU. As noted above, the program product storage device can include any one of multiple removable media having recorded computer-readable instructions, including magnetic floppy disks and CD-ROM storage discs. Other suitable program product storage devices can include magnetic tape and semiconductor memory chips. In this way, the processing steps necessary for operation can be embodied on a program product.
Alternatively, the program steps can be received into the operating memory over the network. In the network method, the computer receives data including program steps into the memory through the network interface after network communication has been established over the network connection by well known methods understood by those skilled in the art. The computer that implements the client side processing, and the computer that implements the server side processing or any other computer device of the system, can include any conventional computer suitable for implementing the functionality described herein.
The mass storage 1508 may include one or more magnetic disk or tape drives or optical disk drives, for storing data and instructions for use by the CPU 1502. At least one component of the mass storage system 1508, preferably in the form of a disk drive or tape drive, stores one or more databases, such as databases containing of transcriptional start sites, genomic sequence, promoter regions, or other information.
The mass storage system 1508 may also include one or more drives for various portable media, such as a floppy disk, a compact disc read only memory (CD-ROM), or an integrated circuit non-volatile memory adapter (i.e., PC-MCIA adapter) to input and output data and code to and from the computer system 1500.
The computer system 1500 may also include one or more input/output interfaces for communications, shown by way of example, as interface 1510 for data communications via a network. The data interface 1510 may be a modem, an Ethernet card or any other suitable data communications device. To provide the functions of a computer system according to
The computer system 1500 also includes suitable input/output ports or use the interconnect bus 1506 for interconnection with a local display 1512 and keyboard 1514 or the like serving as a local user interface for programming and/or data retrieval purposes. Alternatively, server operations personnel may interact with the system 1500 for controlling and/or programming the system from remote terminal devices via the network.
The computer system 1500 may run a variety of application programs and stores associated data in a database of mass storage system 1508. One or more such applications may enable the receipt and delivery of messages to enable operation as a server, for implementing server functions relating to obtaining a set of nucleotide array probes tiling the promoter region of a gene or set of genes.
The components contained in the computer system 1500 are those typically found in general purpose computer systems used as servers, workstations, personal computers, network terminals, and the like. In fact, these components are intended to represent a broad category of such computer components that are well known in the art.
It will be apparent to those of ordinary skill in the art that methods involved in the present invention may be embodied in a computer program product that includes a computer usable and/or readable medium. For example, such a computer usable medium may consist of a read only memory device, such as a CD ROM disk or conventional ROM devices, or a random access memory, such as a hard drive device or a computer diskette, having a computer readable program code stored thereon.
The following examples are provided to illustrate aspects of the invention but are not intended to limit the invention in any manner.
To develop predictors of cytotoxic chemotherapeutic drug response, we used an approach similar to previous work analyzing the NCI-60 panel,49 first identifying cell lines that were most resistant or sensitive to docetaxel (
In addition to leave-one-out cross validation, we utilized an independent dataset derived from docetaxel sensitivity assays in a series of 30 lung and ovarian cancer cell lines for further validation. As shown in
The development of a gene expression signature capable of predicting in vitro docetaxel sensitivity provides a tool that might be useful in predicting response to the drug in patients. We have made use of published studies with clinical and genomic data that linked gene expression data with clinical response to docetaxel in a breast cancer neoadjuvant study50 (
We also performed a complementary analysis using the patient response data to generate a predictor and found that the in vivo generated signature of response predicted sensitivity of NCI-60 cell lines to docetaxel (
Given the development of a docetaxel response predictor, we have examined the NCI-60 dataset for other opportunities to develop predictors of chemotherapy response. Shown in
In addition to the capacity of each signature to distinguish cells that are sensitive or resistant to a particular drug, we also evaluated the extent to which a signature was also specific for an individual chemotherapeutic agent. From the example shown in
Given the ability of the in vitro developed gene expression profiles to predict response to docetaxel in the clinical samples, we extended this approach to test the ability of additional signatures to predict response to commonly used salvage therapies for ovarian cancer and an independent dataset of samples from adriamycin treated patients (Evans W, GSE650, GSE651). As shown in
Many therapeutic regimens make use of combinations of chemotherapeutic drugs raising the question as to the extent to which the signatures of individual therapeutic response will also predict response to a combination of agents. To address this question, we have made use of data from a breast neoadjuvant treatment that involved the use of paclitaxel, 5-fluorouracil, adriamycin, and cyclophosphamide (TFAC)55,56 (
As a further validation of the capacity to predict response to combination therapy, we have made use of gene expression data generated from a collection of breast cancer (n=45) samples from patients who received 5-fluorouracil, adriamycin and cyclophosphamide (FAC) in the adjuvant chemotherapy set. As shown in
As a further measure of the relevance of the predictions, we examined the prognostic significance of the ability to predict response to FAC. As shown in
When comparing individual genes that constitute the predictors, it was interesting to observe that the gene coding for MAP-Tau, described previously as a determinant of paclitaxel sensitivity,56 was also identified as a discriminator gene in the paclitaxel predictor generated using the NCI-60 data. Although, similar to the docetaxel example described earlier, a predictor for TFAC chemotherapy developed using the NCI-60 data was superior to the ability of the MAP-Tau based predictor described by Pusztai et al (Table 4). Similarly, p53, methyltetrahydrofolate reductase gene and DNA repair genes constitute the 5-fluorouracil predictor, and excision repair mechanism genes (e.g., ERCC4), retinoblastoma pathway genes, and bcl-2 constitute the adriamycin predictor, consistent with previous reports (Table 1).
The availability of genomic-based predictors of chemotherapy response could potentially provide an opportunity for a rational approach to selection of drugs and combination of drugs. With this in mind, we have utilized the panel of chemotherapy response predictors described in
As shown in
Most patients who are resistant to chemotherapeutic agents are then recruited into a second or third line therapy or enrolled to a clinical trial.38,59 Moreover, even those patients who initially respond to a given agent are likely to eventually suffer a relapse and in either case, additional therapeutic options are needed. As one approach to identifying such options, we have taken advantage of our recent work that describes the development of gene expression signatures that reflect the activation of several oncogenic pathways.36 To illustrate the approach, we first stratified the NCI cell lines based on predicted docetaxel response and then examined the patterns of pathway deregulation associated with docetaxel sensitivity or resistance (
The results linking docetaxel resistance with deregulation of the PI3 kinase pathway, suggests an opportunity to employ a PI3 kinase inhibitor in this subgroup, given our recent observations that have demonstrated a linear positive correlation between the probability of pathway deregulation and targeted drug sensitivity.36 To address this directly, we predicted docetaxel sensitivity and probability of oncogenic pathway deregulation using DNA microarray data from 17 NSCLC cell lines (
An analysis of a panel of ovarian cancer cell lines provided a second example. Ovarian cell lines that are predicted to be topotecan resistant (
Taken together, these data demonstrate an approach to the identification of therapeutic options for chemotherapy resistant patients, as well as the identification of novel combinations for chemotherapy sensitive patients, and thus represents a potential strategy to a more effective treatment plan for cancer patients, after future prospective validations trials (
NCI-60 data. The (−log 10(M)) GI50/IC50, TGI (Total Growth Inhibition dose) and LC50 (50% cytotoxic dose) data was used to populate a matrix with MATLAB software, with the relevant expression data for the individual cell lines. Where multiple entries for a drug screen existed (by NCS number), the entry with the largest number of replicates was included. Incomplete data were assigned as Nan (not a number) for statistical purposes. To develop an in vitro gene expression based predictor of sensitivity/resistance from the pharmacologic data used in the NCI-60 drug screen studies, we chose cell lines within the NCI-60 panel that would represent the extremes of sensitivity to a given chemotherapeutic agent (mean GI50+/−1SD). Relevant expression data (updated data available on the Affymetrix U95A2 GeneChip) for the solid tumor cell lines and the respective pharmacological data for the chemotherapeutics was downloaded from the NCI website (http://dtp.nci.nih.gov/docs/cancer/cancer_data.html). The individual drug sensitivity and resistance data from the selected solid tumor NCI-60 cell lines was then used in a supervised analysis using binary regression methodologies, as described previously,60 to develop models predictive of chemotherapeutic response.
Human ovarian cancer samples. We measured expression of 22,283 genes in 13 ovarian cancer cell lines and 119 advanced (FIGO stage III/IV) serous epithelial ovarian carcinomas using Affymetrix U133A GeneChips. All ovarian cancers were obtained at initial cytoreductive surgery from patients. All tissues were collected under the auspices of respective institutional (Duke University Medical Center and H. Lee Moffitt Cancer Center) IRB approved protocols involving written informed consent.
Full details of the methods used for RNA extraction and development of gene expression signatures representing deregulation of oncogenic pathways in the tumor samples are recently described.36 Response to therapy was evaluated using standard criteria for patients with measurable disease, based upon WHO guidelines.28
Lung and ovarian cancer cell culture. Total RNA was extracted and oncogenic pathway predictions was performed similar to the methods described previously.36
Cross-platform Affymetrix Gene Chip comparison. To map the probe sets across various generations of Affymetrix GeneChip arrays, we utilized an in-house program, Chip Comparer (http://tenero.duhs.duke.edu/genearray/perl/chip/chipcomparer.pl) as described previously.36
Cell proliferation assays. Growth curves for cells were produced by plating 500-10,000 cells per well in 96-well plates. The growth of cells at 12 hr time points (from t=12 hrs) was determined using the CellTiter 96 Aqueous One 23 Solution Cell Proliferation Assay Kit by Promega, which is a colorimetric method for determining the number of growing cells.36 The growth curves plot the growth rate of cells vs. each concentration of drug tested against individual cell lines. Cumulatively, these experiments determined the concentration of cells to use for each cell line, as well as the dosing range of the inhibitors. The final dose-response curves in our experiments plot the percent of cell population responding to the chemotherapy vs. the concentration of the drug for each cell line. Sensitivity to docetaxel and a phosphatidylinositol 3-kinase (PI3 kinase) inhibitor (LY-294002)36 in 17 lung cell lines, and topotecan and a Src inhibitor (SU6656) in 13 ovarian cell lines was determined by quantifying the percent reduction in growth (versus DMSO controls) at 96 hrs using a standard MTT colorimetric assay.36 Concentrations used ranged from 1-10 nM for docetaxel, 300 nM-10 μM (SU6656), and 300 nM-10M for LY-294002. All experiments were repeated at least three times.
Statistical analysis methods. Analysis of expression data are as previously described.36, 60-62 Briefly, prior to statistical modeling, gene expression data is filtered to exclude probesets with signals present at background noise levels, and for probesets that do not vary significantly across samples. Each signature summarizes its constituent genes as a single expression profile, and is here derived as the top principal components of that set of genes. When predicting the chemosensitivity patterns or pathway activation of cancer cell lines or tumor samples, gene selection and identification is based on the training data, and then metagene values are computed using the principal components of the training data and additional cell line or tumor expression data. Bayesian fitting of binary probit regression models to the training data then permits an assessment of the relevance of the metagene signatures in within-sample classification,60 and estimation and uncertainty assessments for the binary regression weights mapping metagenes to probabilities. To guard against over-fitting given the disproportionate number of variables to samples, we also performed leave-one-out cross validation analysis to test the stability and predictive capability of our model. Each sample was left out of the data set one at a time, the model was refitted (both the metagene factors and the partitions used) using the remaining samples, and the phenotype of the held out case was then predicted and the certainty of the classification was calculated. Given a training set of expression vectors (of values across metagenes) representing two biological states, a binary probit regression model, of predictive probabilities for each of the two states (resistant vs. sensitive) for each case is estimated using Bayesian methods. Predictions of the relative oncogenic pathway status and chemosensitivity of the validation cell lines or tumor samples are then evaluated using methods previously described36,60 producing estimated relative probabilities—and associated measures of uncertainty—of chemosensitivity/oncogenic pathway deregulation across the validation samples. In instances where a combined probability of sensitivity to a combination chemotherapeutic regimen was required based on the individual drug sensitivity patterns, we employed the theorem for combined probabilities as described by Feller: [Probability (Pr) of (A), (B), (C) . . . (N)=Pr (A)+Pr (B)+Pr (C) . . . +Pr (N)−[Pr(A)×Pr(B)×Pr(C) . . . ×Pr (N)]. Hierarchical clustering of tumor predictions was performed using Gene Cluster 3.0.63 Genes and tumors were clustered using average linkage with the uncentered correlation similarity metric. Standard linear regression analyses and their significance (log rank test) were generated for the drug response data and correlation between drug response and probability of chemosensitivity/pathway deregulation using GraphPad® software.
Drosophila)”
Drosophila)”
This invention was made with government support under NCI-U54 CA1112952-02 and R01-CA106520 awarded by the National Cancer Institute. The government has certain rights in the invention.
