WO2016118670A1 - Multigene expression assay for patient stratification in resected colorectal liver metastases - Google Patents

Abstract

Disclosed are methods, kits and systems for stratifying a patient according to a molecular risk score and employing such information to making a therapy decision involving a colorectal cancer patient diagnosed with one or more liver metastases. The method comprises (a) comparing expression levels within tumor cells obtained from the patient of at least 10 of a panel of 20 genes:, to expression levels of one or more reference genes or to one or more predetermined values correlating to expression levels of said one or more reference genes; (b) calculating a molecular risk score (MRS) for the patient based on differential expression of said genes compared to the one or more reference genes or value;(c) making a therapy decision with respect to said patient taking into account whether the MRS is high or low compared to a predetermined median score.

Description

MULTIGENE EXPRESSION ASSAY FOR PATIENT STRATIFICATION IN RESECTED COLORECTAL LIVER METASTASES

CROSS-REFERENCE TO RELATED APPLICATIONS

This application claims priority to U.S. Provisional Patent Application No.

62/105,616 filed January 20, 2015 the contents of which are incorporated by reference herein.

GOVERNMENT SPONSORED RESEARCH OR DEVELOPMENT

The work described in this disclosure has been indirectly funded in part by a grant from the National Institutes of Health (National Cancer Institute) NIH/NCI P30 CA008748 (Cancer Center Support Grant) to the Core Facility of the applicant institution. The U.S. government may have certain rights in this invention.

BACKGROUND OF THE DISCLOSURE

Field of the Disclosure

The present disclosure is in the field of developing molecular or genetic methods of predicting clinical outcome of a proposed therapeutic modality. More specifically the present disclosure pertains to the field of predicting clinical outcome of colorectal cancer patients who have liver metastases (such metastases referred to herein as "CRLM"), should such patients be subjected to surgery to resect the liver metastases. Description of the Related Art

Colorectal cancer is the third most common cancer and a leading cause of cancer death worldwide (Jemal et al. CA Cancer J Clin, 61(2): 69-90, 2011). Approximately 150,000 new cases per year are diagnosed in the United States. From those, 60,000-110,000 patients with colorectal cancer develop CRLM. Of the patients with metastatic disease, 15-25% have been reported to develop synchronous liver metastases at the time of the initial colorectal cancer diagnosis and 10-25% have been reported to develop metachronous CRLM following resection of the primary lesion (Luu et al., J Gastromiest Oncol , 4(3):328-36, 2013). Currently, the standard treatment for patients with CRLM comprises resection of metastatic lesions and chemotherapy. However, due to such factors as tumor size, location, multifocality or general condition of the patient, most patients are not considered surgical candidates, although in recent years the percentage of patients considered suitable for surgery has increased. Currently, -25% of patients with liver metastases are considered surgical candidates (Leonard et al, J Clin Oncol , 23(9):2038-48, 2005). Among those undergoing surgery, 5-year overall survival is approximately 50% with surgery alone, while only 15-25% of patients are alive, disease-free, and considered cured at 10 years. Outcomes after surgery are heterogeneous, with up to 70% of patients recurring and 30% dying from cancer within 2 years. These results underscore the need for accurate preoperative risk stratification, which would identify the likelihood of a patient's favorable clinical outcome following surgical resection of CRLM. Successful development of a risk score with strong prognostic ability would constitute a major advance in patient stratification and in the clinical decision making process.

Specific clinical endpoints, such as overall survival (OS), disease specific survival (DSS), and recurrence free survival (RFS) after resection of CRLM also reflect this heterogeneity. Various prognostic scoring systems have been developed with the aim of predicting OS, DSS, and RFS following a resection CRLM. For example, Fong, Nordlinger, Iwatsuki and others have developed clinical risk scores (CRSs) based on clinical, pathological, and treatment outcome data, with the Fong score being the most commonly utilized score (Fong et al. Ann Surg. 230 (3) 1999; Nordlinger et al. Cancer, 77(7), 1996; Iwatsuki et al. J Am Coll Surg. 189(3), 1999). For example, the Fong CRS is based on factors including lymph node status of the primary tumor (negative or positive), disease-free interval (<12 or >= 12 months), serum carcinoembryonic antigen (CEA) level prior to liver resection (>200 or <= 200 ng/mL), number of hepatic tumors (1 or >1), and tumor size (<= 5 or > 5 cm). In the Fong CRS calculation, patients receive 1 point for each adverse factor, and the CRS represents the sum, where a high risk CRS is CRS >= 3, and a low risk CRS is CRS < 3. However, these scoring systems have insufficient discriminatory power and have not consistently predicted outcomes. As a result, knowledge of a clinical risk score alone will rarely impact clinical decisions.

The inadequacy of the foregoing predictive methods has been revealed in clinical studies. For example, it was reported that none of the Fong, Nordlinger, or Iwatsuki scores stratified 662 patients with metastatic colorectal cancer treated surgically at the Mayo clinic (Zakaria et al. Ann Surg., 246(2), 2007). Furthermore, when eight different scoring systems were applied to 286 patients with 10-year follow-up, none showed sufficient predictive accuracy of recurrence or cancer specific death to be used in clinical practice (Roberts et al., Br J Surg., 101 (7), 2014).

While the above mentioned scoring systems have been developed using clinical, pathological, and treatment outcome data, more recent studies have focused on identification of molecular parameters statistically associated with clinical outcomes. Gene expression profiling has proven successful in predicting treatment- related clinical outcomes in various other types of cancer, such as lung cancer (Beer et al, Nat Med, 8 (8):816-24, 2002), breast cancer (Van't Veer et al. Nature, 415(6871):530-6, 2002; U.S. Published Patent Application 20140308202), prostate cancer (U.S. Published Patent Application 20140308202) and most recently melanoma (Delman, K. and Lawson, D., Emory University, unpublished data). However, since tumor-associated gene expression changes are different depending on the specific type of cancer, gene expression data obtained from lung cancer, breast cancer, or prostate cancer do not correlate with gene expression alterations associated with colorectal cancer or its liver metastasis. Recently, gene expression profiling was used in an attempt to identify gene scores capable of predicting which patients with colorectal liver metastases would benefit from surgery. To date, however, the results have not been satisfactory even though in certain cases some correlation was established (Ito et al. PLoS One, 8 (12), 2013). For example, a scoring method involving an 19- gene signature was developed by the present inventors and co-workers based on the same patient data as the derivation group discussed below which stratified patients based on DSS and LRFS (Ito et al. PLoS One, 8 (12), 2013), but this 19-gene signature did not constitute a prognostic tool for OS (unpublished information). The Ito et al. paper also reports a 115-gene signature which stratified patients only according to liver recurrence free survival (LFRS) and not according to DSS or RFS. (Interestingly, there is very little overlap in the gene signature between the work described in the present disclosure and these prior studies.) A major flaw of currently available scoring systems applied to resected

CRLM, whether molecularly or clinically based, is that they were derived based on a single-institution cohorts and data, reflecting the practice patterns and potential bias of a specific institution, such as referral patterns, patient selection, and treatment methodologies. Accordingly, validation of a risk score across different institutions is a critical step in establishing a consistent prognostic signature.

Certain subgroups of patients, but not all patients, may experience a benefit from perioperative chemotherapy. The success of Oncotype Dx, a multigene assay to risk stratify patients with resected primary breast cancer has shown the validity of a molecular approach to such prognostication (Paik et al. N Engl J Med 2004;351 :2817-26), and has been applied to other primary tumors (Hoshida et al. N Engl J Med 2008;359: 1995-2004). However, prior to the work described herein, a prognostic molecular signature has yet to be developed for resected CRLM, adjuvant chemotherapy or for metastasectomy of any solid tumor.

SUMMARY OF THE DISCLOSURE The present disclosure provides a method for making treatment decisions for colorectal cancer patients by determining prognostic outcome for a patient diagnosed with colorectal cancer liver metastases (CRLM) if the patient were subjected to surgical resection of such metastases or subjected to such surgery and administered adjuvant chemotherapy. The present disclosure provides a tumor gene set or panel, the differential expression levels of which can be used to determine a likelihood of beneficial outcome after resection of liver metastases in colorectal cancer patients or resection with adjuvant chemotherapy (low MRS score). Conversely, if the thus determined likelihood of beneficial clinical outcome is low (high MRS score), a decision of no further treatment can be made or the patient can be referred for example to a clinical trial or administered an alternative therapeutic regimen if one is or should become available for colorectal cancer, or more specifically, CRLM and resectable CRLM.

The gene set or panel disclosed herein has been identified using an iterative process a great number of times over and has been validated against patient data from a separate institution.

In one embodiment, the present disclosure provides a method for making a therapy decision for a colorectal cancer patient diagnosed with one or more liver metastases or referring the patient to a particular course of treatment, the method comprising: a) comparing expression levels within tumor tissue or cells of at least 10 of the following 20 genes: RBBP8, DKK1, LRRC42, REG4, RAD23B, FGFBP1, NUP62CL, RNF135, HOXC6, DNAJC12, SMIM24, LRP8, RPS24, TYMS, PLA2G2A, CES2, ODC1, SERPINB1, PLCB4, and STEAP1, to a median expression of each of said at least 10 genes across a cohort of patients, or to expression levels of one or more reference genes (such as for example the same gene from a non-cancerous cell) or to one or more predetermined values correlating to expression levels of said one or more reference genes.

b) calculating a molecular risk score (MRS) for the patient based on differential expression of said genes compared to the median expression of said genes across a studied cohort of patients, or the one or more reference genes or value, and;

c) making a therapy decision with respect to said patient taking into account whether the MRS is high or low compared to a predetermined median or other averaged score.

In some embodiments, the method further comprises treating the patient with surgery or surgery plus adjuvant chemotherapy, if the MRS is low or refraining from subjecting the patient to surgery and referring the patient to other available therapy options if the MRS is high. In some embodiments the tumor cells are contained in or obtained from a biological sample obtained from the patient. In some embodiments the comparison of expression levels is for at least 10 of the foregoing genes; In some embodiments, the comparison of expression levels is for at least 11, 12, 13, 14, 15, 16, 17, 18, 19, or all 20 of the foregoing genes.

In some embodiments, the MRS is high and the therapy decision is to forgo surgery; in some embodiments the MRS is low and the therapy decision is to proceed with surgery.

In another aspect, the present disclosure provides a kit that can be used for establishing gene expression levels, making the foregoing comparison, calculating an MRS score and comparing it to a predetermined median MRS value. The kit comprises a. reagents sufficient for measuring the expression levels of said at least 10 genes or at least 10 genes or at least 11, 12, 13, 14, 15, 16, 17, 18, 19, or all 20 of the foregoing genes from cancer cells of said patient and optionally for measuring expression levels of the reference genes such as the same genes from noncancerous liver cells from the patient if the comparison will be to the same genes from noncancerous cells; b. instructions for effecting the comparison; c. instructions for deriving an MRS score for the patient and comparing it to a predetermined median or other averaged score. If the comparison is to be made to a predetermined reference value that correlates to expression levels of said genes, then the kit may but need not contain reagents sufficient for measuring expression levels of reference genes but the comparison is made to predetermined reference values for such genes and instructions would be provided to that effect. BRIEF DESCRIPTION OF THE DRAWINGS

Figure 1 is a plot of percent survival over years after surgery. The graph shows overall survival (OS), disease specific survival (DSS), and recurrence free survival (RFS) in the MSKCC derivation cohort. Note the close correlation of OS and DSS, which is expected given the high mortality rate of colorectal cancer and the fact that the vast majority of newly diagnosed patients are 50 years old or older.

Figure 2 is a flow chart of investigational design used to identify prognostic genes and MRS. It begins with defining a derivation cohort, performing gene expression analysis, randomly assigning patients to a training set and a test set, performing supervised principal component analysis iteratively; identifying genes that are frequently dysregulated and using them to derive the MRS. The gene signature is then validated against a validation cohort.

Figures 3A-3C are bar graphs depicting preliminary data of expression of each of the 20 genes in tumor (mCRC which stands for "metastatic colorectal cancer") and published data for normal liver, normalized to the housekeeping gene GAPDH.

Figures 4A, 4B and 4C are plots of percent survival versus years after surgery. Figure 4C shows disease specific survival, Fig. 4B overall survival, and Fig. 4A recurrence-free survival in MSKCC cohort stratified by median MRS of the cohort. Circles and squares are censored events. Hazard ratios (HR), 95% confidence intervals (CI), and P values were determined by log rank method.

Figures 5A and 5B are plots of percent survival over years after surgery. Figure 5A shows overall survival in both MSKCC derivation and Dutch validation cohorts. Figure 5B shows recurrence-free survival MSKCC derivation and Dutch validation cohorts.

Figures 6 A and 6B are plots of percent survival over years after surgery. Figure 6 a shows overall (6A) and recurrence-free survival (6B) in Dutch cohort stratified by MSKCC molecular risk score (MRS). NR= Not reached. Hazard ratios (HR), 95% confidence intervals (CI), and P values were determined by log rank method. Figure 7 is a flowchart of the method for ascertaining a patient's MRS. Briefly, reagents enabling detection of genes included in the 20-gene signature are added to a patient biological sample or to a sample derived therefrom and containing cancer cells from liver metastases of colorectal cancer. Measured expression levels of genes are compared to measured expression levels of reference genes (which can be determined in parallel or which are provided in a predetermined reference). If a gene is determined to be highly differentially expressed, it is included in the panel used to calculate the MRS as described below. Calculated MRS is then compared to a predetermined median MRS, and the likelihood of positive clinical outcome following surgical resection of colorectal liver metastasis is determined based on whether the calculated MRS is higher or lower than the median MRS of the cohort.

Figure 8 is a schematic representation of exemplary architecture of apparatuses and systems that can be used to implement computing dependent aspects of the present disclosure.

DETAILED DESCRIPTION

Definitions

The following terms including the plural form and cognates thereof shall have the meanings ascribed to them below unless the context clearly requires otherwise.

As used herein, the term "differentially expressed gene" refers to a gene that is expressed at a higher or lower level in a cancer cell or tissue compared to median (or other averaged) expression of said gene across the studied group. Thus, the comparison can be made to a predetermined reference value such as the foregoing median or other averaged expression of the same gene across a studied group of patients or even individuals not diagnosed with cancer in the liver (metastatic or not). Alternatively, "differentially expressed gene" refers to a gene that is expressed at a higher or lower level in a cancer cell or tissue compared to a reference gene, such as one in a cell or tissue of the same type that is non-cancerous. For example, mRNA of a cancer cell or tissue is at levels at least about 25%, at least about 50%, at least about 75%, at least about 90%, at least about 1.5-fold, at least about 2-fold, at least about 5-fold, at least about 10-fold, or more, higher or lower than the corresponding non-cancerous cell or tissue (or, in the case of generation of the data that led to the present application, the median expression of the same gene in the study cohort).

As used herein, the term "Molecular Risk Score" or "MRS" refers to an outcome risk score calculated on the basis of differential expression of a panel of genes identified in this disclosure to statistically correlate with risk of clinical outcome. Molecular Risk Score (MRS) was calculated as

MRS ^:=: Sum [(log univariate Cox proportional hazards coefficients) * SGE j wherein SGE (standardized gene expression) is calculated as (absolute gene expression - mean gene expression/standard deviation. It should be noted that in the generation of the data mea gene expression was used as a reference. In practice, however, of the present methods, the reference can be expression of the same gene in a noncancerous cell or other reference expression level . As used herein, the term "correlates" or "correlating" refers to a statistical association between instances of two events, where events may include numbers, data sets, and the like. For example, when the events involve numbers, a positive correlation means that as one increases, the other increases as well. An inverse correlation means that as one increases, the other decreases. The present disclosure provides RNA transcripts, or expression products thereof, the levels of which are correlated with a particular outcome measure, such as between the level of particular RNA transcript or expression product thereof and the likelihood of beneficial response to resection of colorectal liver metastases. Such a correlation may be demonstrated statistically in various ways, e.g. by a low hazard ratio. In another example, levels of particular RNA transcripts or expression products thereof may correlate with a decreased likelihood of beneficial response to resection of colorectal liver metastases. Such correlation may be demonstrated statistically in various ways, e.g., by a high hazard ratio.

As used herein, the term "clinical outcome" refers to the resulting progression (or nonprogression) of disease and can be characterized for example by recurrence, period of time until recurrence, period of time until metastasis, number of metastases, and/or death due to disease. For example, a positive clinical outcome includes nonrecurrence over a given period of time, cure (nonrecurrence over a longer period of time, usually 10 years) and/or survival within a given period of time (especially without recurrence). Poor clinical outcome includes disease progression, recurrence and/or death within a given period of time.

As used herein, the term "prognosis" refers to a general forecast of the future course of a disease, such as the general probability of disease recurrence, metastatic spread, or cancer-attributable death.

As used herein, the term "prediction" refers to the likelihood that a particular patient will respond positively to a treatment or regimen, such as resection of liver metastases of colorectal cancer. The predictive methods of the present disclosure can be used clinically to make treatment decisions by choosing the most suitable treatment option for a particular patient.

As used herein, the term "overall survival" (OS) refers to time from liver resection to death from any cause or last follow-up.

As used herein, the term "disease-specific survival" (DSS) refers to time from liver resection to cancer-related death. As used herein, the term "recurrence-free survival" (RFS) refers to time from liver resection to recurrence.

As used herein, the term "responder" refers to patients in which the cancer/tumor(s) is eradicated, reduced, or stabilized such that the disease is not progressing. As used herein, the term "non-responder" refers to patients characterized by progressive disease.

As used herein, the term "normal" as used in the context of "normal cell or tissue," refers to a cell or tissue of an untransformed phenotype or exhibiting a morphology of a non-transformed cell or tissue type in question. As used herein, the term "polynucleotide" refers to any polyribonucleotide or polydeoxyribonucleotide, which may be RNA or DNA. For example, polynucleotides as used herein refers to, among others, single- and double-stranded DNA, DNA that is a mixture of single- and double-stranded regions, single- and double-stranded RNA, and RNA that is mixture of single- and double-stranded regions, hybrid molecules comprising DNA and RNA that may be single-stranded or, more typically, double-stranded or a mixture of single- and double-stranded regions. The term encompasses genes.

As used herein, the terms "peptide," "protein" and "polypeptide" refer to any polymer comprising any of the 20 protein amino acids, regardless of its size. Although "protein" is often used in reference to relatively large polypeptides, and "peptide" is often used in reference to small polypeptides, usage of these terms in the art overlaps and varies. The term "protein" as used herein refers to peptides, polypeptides and proteins, unless otherwise noted. As used herein, the terms "protein", "polypeptide" and "peptide" are used interchangeably herein when referring to a gene expression product.

As used herein, the term "expression" generally refers to the cellular processes by which an RNA is produced from a DNA template by RNA polymerase (RNA expression) or a polypeptide is produced from RNA (protein expression). Thus the term "expression" describes levels of either RNA or protein in a cell that can be quantified by methods described in the disclosure.

As used herein, the term "diagnosis" refers to a determination that has been made that the cancer is, for example, a colorectal cancer with liver metastases and in some embodiments with resectable liver metastases. A diagnosis may be made prior to (on a different sample) performing the present methods for determining the likelihood of beneficial clinical outcome to resection of liver metastases in colorectal cancer patients and/or to adjuvant chemotherapy. Alternatively, diagnosis may be made in conjunction (i.e., either concurrently or sequentially) with the present methods for determining the likelihood of beneficial clinical outcome to resection of liver metastases in colorectal cancer patients and/or to adjuvant chemotherapy.

As used herein, the term "microarray" refers to an ordered arrangement of hybridizable array elements, preferably polynucleotide probes, on a substrate.

As used herein, the term "primer" refers to a short segment of DNA or DNA- containing nucleic acid molecule, which (i) anneals under amplification conditions to a suitable portion of a DNA or RNA sequence to be amplified (e.g. a target sequence), and (ii) initiates extension, and is itself physically extended, via polymerase-mediated synthesis.

As used herein, the term "colorectal cancer" relates to cancer of the large intestine and/or rectum, and includes adenocarcinoma. In particular, in the present disclosure, the term refers to colorectal cancer with CRLM before or after resection of the latter.

As used herein, the term "metastasis" refers to the growth of a cancerous tumor in an organ or body part, which is not directly connected to the organ of the original cancerous tumor. Metastases will be understood to cancerous cells in an organ or body part which is different and in particular distant from the organ of the primary tumor.

As used herein, the term "liver" refers to the whole organ liver or parts thereof, liver tissue and/or liver cells.

As used herein, the term "liver resection" refers to partial removal of liver, or one or more of its vascular segments. As used herein, terms "liver resection" and "hepatic resection" are used interchangeably.

The term "chemotherapy" as used herein refers to the treatment of cancer using specific chemical agents or drugs that are destructive of malignant cells and tissues. In the particular case of colorectal cancer, chemical agents that are commonly used include, without limitation, platinum drugs, pyrimidine

antimetabolite drugs, leucovorin and combinations thereof. Different combinations of therapy have been developed and may be recommended for initial treatment. Some of the combination chemotherapy regimens include: oxaliplatin plus FU and leucovorin (referred to as FOLFOX), irinotecan plus FU and leucovorin (referred to as FOLFIRI), and oxaliplatin plus capecitabine (referred to as XELOX or CAPOX). Additionally, adding bevacizumab to FOLFOX, FOLFIRI, or XELOX increases the likelihood that the tumor will respond and prolongs survival compared with treatment without bevacizumab (Kabbinavar et al. J Clin Oncol,2\ :60-65, 2003; Saltz et al. J Clin Oncol. ;27(4):653, 2008). Lastly, cetuximab alone or in

combination with FOLFIRI has been approved for certain patients with colorectal cancer. All of these treatments are contemplated for use herein. As used herein, the term "neoadjuvant chemotherapy" relates to a preoperative therapy regimen consisting of one or more chemotherapeutic and/or antibody agents, which are used to reduce tumor burden, in an effort to y render local therapy (such as surgery) less extensive or more effective. As used herein, the term "adjuvant chemotherapy" relates to a postoperative therapy regimen consisting of one or more chemotherapeutic and/or biologic agents, which are used with a purpose of improving treatment (such as surgery) outcome.

As used herein, the term "sample" refers to any biological sample or specimen, such as a tumor biopsy sample, which can be obtained from the patient. The present method can be applied to any type of biological sample from a patient, such as a biopsy sample, core biopsy, fine needle aspiration biopsy, a tissue, cell, blood or a bodily fluid containing cancers cells. In a particular embodiment, the sample is a tumor tissue sample or portion thereof such as tumor cells. In a more particular embodiment, the tumor tissue sample is a liver metastasis tissue sample from a patient suffering from colorectal cancer. The sample can be obtained by any method, e.g., biopsy, by using methods well known to those of ordinary skill in the related medical arts. Additionally, samples can be frozen, or paraffin-embedded.

As used herein, in the context of determining differential expression in the tumor of a subject of one of the genes listed in Table 2, the term "reference gene" or "reference polynucleotide" refers to expression of either the same gene as that found in a cancer cell but in a non-cancerous cell or another reference expression level such as the mean expression level of the same gene across a collection of tumor samples from different patients as was done in the study that gave rise to the present disclosure. For example, expression of a gene from Table 2 can be compared to expression of the same gene in a non-cancerous cell or to another predetermined reference expression level. To minimize expression measurement variations due to non-biological variations in samples, the data are preferably normalized for both differences in the amount of RNA assayed and variability in the quality of the RNA used. The level of RNA or its expression product may be normalized relative to the mean levels obtained for one or more reference RNA transcripts or their expression products. Housekeeping genes can be chosen based on the relative invariability of their expression in the study samples and their lack of correlation with clinical outcome. Commonly used housekeeping genes include, but are not limited to, glyceraldehyde-3-phosphate-dehydrogenase (GAPDH), β-actin, RPLPO, GUS, TFRC, and 18S rRNA.

In the context of the present disclosure, reference to "at least 10", at least "fifteen", etc. of the genes listed in Table 2, means any and all combinations of 5 or more, 10 or more, 11 or more, 12 or more, 13 or more, 14 or more, 15 or more etc. of the 20 genes listed in Table 2.

Description of Specific Embodiments

Standard treatment for patients diagnosed with colorectal cancer liver metastases includes surgical resection, provided the remnant liver has sufficient volume and vascular supply to ensure adequate future liver function. While surgical resection of the liver provides significant benefit in OS, RFS, and DSS, there is significant variability in outcome within the group of colorectal cancer patients that undergo surgery. Stratification of patients according to likely clinical outcome using the Molecular Risk Score (MRS) disclosed herein provides a novel tool to improve the treatment decision-making process. The methods disclosed herein comprise calculating MRS using the selected panel of differentially expressed genes optionally in combination with one or more clinical variables, such as comorbidities, performance status and any other indication of the patient's physiologic ability to withstand major surgery. The MRS can thus be used to inform treatment decisions. Thus, a subject having a low risk score may benefit from surgical resection of colorectal liver metastases, whereas a subject having a high risk score may not be indicated for such surgery.

In the present disclosure, iterative supervised principle component analysis, which had not been used in previous studies of CRLM, was utilized in a significant number of iterations to identify an MRS capable of stratifying patients with CRLM thereby enabling a prediction of clinical outcome following resection of CRLM and/or adjuvant chemotherapy.

The present disclosure provides a panel of 20 genes all or a subset of which (at least 10 or at least 11, 12, 13, 14, 15, 16, 17, 18, or at least 19 or all 20 of them) can be used to predict OS and one or more of DSS, and RFS. This is significant as, unlike other multigene assays that only assess recurrence, the present disclosure provides an MRS that significantly stratifies OS, DSS, and RFS with just a modest number of genes, underscoring both the clinical utility and applicability of the MRS. The MRS of the present disclosure is the first externally validated multigene set to prognosticate outcomes after metastasectomy not only for liver metastasis of colorectal cancer but for any solid tumor (DSS was not validated in the Dutch cohort because this information was not available for this cohort— however DSS closely correlates with OS). Thus, the prognostic utility of the MRS score of the present disclosure was validated using a "validation cohort" comprising patients with surgically resected colorectal liver metastases at a different institution, thereby confirming that the developed clinical risk scoring system is applicable to patients regardless of the practices and biases of the institution in which they are being treated.

Moreover, the MRS of the present disclosure can also be used to predict outcome of adjuvant chemotherapy in patients with liver metastases of colorectal cancer and has potential for clinical application as a biomarker in resected colorectal liver metastases to monitor the efficacy of therapy. In other words, the MRS can be derived from clinically applicable PCR assays to monitor and prognosticate outcomes not only of metastasectomy (whether the patient would benefit from surgery) but also after metastasectomy (whether the patient would benefit from adjuvant chemotherapy).

In one embodiment, the present disclosure provides a method of predicting the likelihood of a positive clinical outcome of surgical resection of colorectal liver metastases. The clinical outcome can be expressed as OS, DSS, or RFS.

In another embodiment, the present disclosure provides a method of using an MRS score to predict the likelihood of a specific clinical outcome in a colorectal cancer patient, such as likelihood of long- term survival without disease recurrence. For example, a likelihood score can be calculated by determining the level of mRNA or its expression product, corresponding to at least 10 or at least 15 or all 20 of the 20 genes included in Table 2. In one embodiment, the present disclosure is used for determining outcome after liver resection in a patient diagnosed with resectable colorectal cancer liver metastasis. Knowing a patient's predisposition to benefit from liver resection can be used in the decision making process regarding the most appropriate therapy regimen. Thus, in view of the methods provided by the present disclosure, clinicians can weigh therapy options regarding a specific patient, and choose the therapy most likely to be beneficial based on the predicted outcome of CRLM resection.

In another embodiment, the present disclosure provides a method for characterizing a patient as a suitable candidate for surgical resection of colorectal liver metastases or as not suitable for such resection.

Therefore, in one embodiment, the present disclosure relates to a method for predicting a patient's response to colorectal liver metastases resection, comprising (i) comparing expression levels within tumor cells of at least 10 or at least 11, 12, 13, 14, 15, 16, 17, 18, or at least 19 or all 20 of the following 20 genes: RBBP8, DKK1, LRRC42, REG4, RAD23B, FGFBP1, NUP62CL, RNF135, HOXC6, DNAJC12, SMIM24, LRP8, RPS24, TYMS, PLA2G2A, CES2, ODC1, SERPINB1, PLCB4, and STEAP1, to expression levels of one or more reference genes or to one or more predetermined values correlating to expression levels of said one or more reference genes; (ii) calculating a molecular risk score (MRS) for the patient based on differential expression of said genes compared to the one or more reference genes or value and; (iii) making a therapy decision with respect to said patient taking into account whether the MRS is high or low compared to a predetermined median score.

In some embodiments the tumor cells can be obtained from a biological sample obtained from the patient.

In methods described herein, a biological sample containing tumor cells is assayed for levels of an RNA transcript, or its expression product. The biological sample may comprise any clinically relevant tissue sample, such as a tumor biopsy sample, fine needle aspiration biopsy, a tissue, cell, blood or a bodily fluid containing cancer cells. The biological sample can be obtained from a solid tumor tissue or cells, such as from a liver metastatic lesion. The sample may be collected in any clinically acceptable manner, such as core needle biopsy, fine needle biopsy, surgical biopsy, surgical resection, etc. In one embodiment of the invention, the biological sample is obtained from a patient with colorectal cancer liver metastases. In another embodiment, the sample is an archival pathological sample that can be preserved, e.g., paraffin-embedded or frozen.

In a particular embodiment, the level of an RNA transcript of one of the gens in the foregoing panel, or the expression product of such transcript, is normalized relative to the level of one or more reference polynucleotides or RNA transcripts, or expression products. In another embodiment, normalized levels of a particular RNA transcript or its expression product isolated from tumor tissue (or cells), are compared to the levels of the same RNA transcript or its expression product isolated from normal tissue (or cells). Patients' gene expression levels may be quantified by any means known in the art. A detection mechanism can be any comparison mechanism such as whole transcriptome sequencing, or a reverse transcription polymerase chain reaction (RT-PCR) for detecting at least 10, 1 1, 12, 13, 14, 15, 16, 17, 18, 19 or ail of the genes included in Table 2.

As would be understood by the skilled person, detection of expression of nucleic acids may be performed by the detection of expression of any appropriate portion or fragment of these nucleic acids, or the entire nucleic acids. Preferably, the portions are sufficiently large to contain unique sequences relative to other sequences expressed in a sample. Moreover, the skilled person would recognize that either strand of a nucleic acid may be detected as an indicator of expression of the nucleic acid. This follows because the nucleic acids are expressed as RNA molecules in cells, which may be converted to cDNA molecules for ease of manipulation and detection. The resultant cDNA molecules may have the sequences of the expressed RNA as well as those of the complementary strand. In one embodiment, the method comprises performing a reverse transcription of mRNA molecules present in a sample; and amplifying the target cDNA and the one or more control cDNAs using primers hybridizing to the cDNAs.

In another embodiment, the expression level of the genes listed in Table 2 can be determined by measuring their protein levels. The determination of the expression level of the protein can be carried out by immunological techniques such as e.g. immunohistochemistry, Western blot, immunofluorescence, etc. One skilled in the art will recognize that there are many statistical methods that may be used to determine whether there is a correlation between an outcome of interest (e.g., likelihood of OS, RFS, or DSS) and levels of RNA transcripts or their expression products as described here. Patients may be classified based on the calculated MRS. For example, the MRS of a patient cohort may be generated using a Cox proportional hazard model. Patients with a MRS less than the median are defined as good candidates for surgery (responders), whereas patients with a risk score greater than the median are classified as poor candidates for surgery (non- responders). A patient's MRS can also be determined by using a statistical model or a machine learning algorithm, which computes the MRS based on the patient's gene expression profiles. Cutoffs can be defined for patient stratification based on the specific clinical setting in accordance with the skill in the art in light of the present disclosure. In addition, patients may be defined into two risk groups based on the tumor gene set defined above.

In some embodiments described herein, the MRS is calculated using a Cox Proportional Hazards Model Analysis (T.J. Cleophas and A.H. Zwinderman, Statistics Applied to Clinical Studies, 2012.), which is a regression method for survival data that provides an estimate of the hazard ratio and its confidence interval. Briefly, the Cox model is a well-recognized statistical technique for exploring the relationship between the survival of a patient and particular variables. This statistical method allows for estimation of the hazard (i.e. , risk) of individuals having a particular expression profile of the gene panel according to the present disclosure. The "hazard ratio" according to the Cox model is the risk of death at any given time point for patients displaying particular prognostic variables.

The tumor gene set described above and corresponding MRS calculated according to the methods disclosed herein can additionally be used or at least taken into account to stratify cancer patients for inclusion in (or exclusion from) clinical studies. For example, MRS may be used on samples collected from patients in a clinical trial and the results of the test used in conjunction with patient outcomes in order to determine whether subgroups of patients are more or less likely to demonstrate a benefit from adjuvant chemotherapy or from a new therapy than the whole group or other subgroups. As another example, patients with a high MRS, who would not benefit from surgery, can be guided to enter clinical trials whereas those with a low MRS who would benefit from surgery can be subjected to metastasectomy and not considered for clinical trial entry unless there is recurrence post-surgery. On the other hand, those with low MRS, especially those who have not received preoperative chemotherapy are likely to benefit from post-operative (adjuvant) chemotherapy. The data presented herein supports this for this patient subgroup (p<0.02).

In some embodiments, the time of prognosis assessment via MRS begins at any time a sample is collected containing cancer cells from colorectal liver metastases. In another embodiment, the MRS is calculated based on a sample obtained at the time of diagnosis of colorectal cancer metastasis to the liver. In another embodiment, the methods disclosed here can be performed and used by a variety of agencies, as well as private individuals, such as clinical laboratories, experimental laboratories, or medical practitioners. Thus, the present methods need not be practiced by the institution in which surgery would be performed. Of course, most often, the MRS calculation and underlying gene expression measurements would be performed at the request of the treating physician or pursuant to an institutional protocol for treating patients with colorectal cancer liver metastases.

Reports

The methods of the present disclosure are suited for the preparation of reports summarizing the predicted clinical outcome resulting from the methods of the present disclosure. A "report," as described herein, is an electronic or tangible document that includes report elements that provide information relating to patient's response to colorectal cancer liver metastases resection. A report can include an assessment or estimate of one or more of OS, DSS, and RFS. In one embodiment, the present disclosure provides methods for creating reports and in some aspects it is directed to the reports resulting therefrom. The reports may include a summary of the expression levels of the R A transcripts, or the expression products of such RNA transcripts, for genes in the samples obtained from the patient's tumor. The report can further include a calculated MRS. The report can include information relating to prognostic outcome of the patient determined using the MRS. The report may include information, such as comorbidities or performance status of the patient relevant to assist with decisions about the surgery or adjuvant chemotherapy or other further treatment for the patient such as referral to a clinical study.

Thus, in some embodiments, the methods of the present disclosure further include generating a report that includes information regarding the patient's likely clinical outcome, e.g. OS, DSS, and RFS. Such report can be provided in the form of an electronic medium (e.g., an electronic display on a computer monitor), or in the form of a tangible medium (e.g., a report printed on paper or other tangible medium). A report that includes information regarding the patient's likely prognosis

(e.g., the likelihood that a patient having colorectal cancer liver metastasis will have a positive clinical outcome in response to surgery and/or treatment) is provided to a user. A person or entity that prepares a report may also perform the MRS calculation and interpretation of the risk score. The report generator may also perform one or more of sample gathering, sample processing, and data generation. Alternatively, an entity other than the report generator can perform one or more sample gathering, sample processing, and data generation.

Research Uses

Gene expression deregulation followed by disruption of cellular networks is a hallmark of cancer. Altered gene expression can lead to defects in essential cellular processes such as proliferation, apoptosis, etc., ultimately leading to cancer progression, tumor recurrence, metastatic progression, and colonization of distant organs. However, many of the signaling mechanisms that govern tumorigenesis and mediate the steps necessary for cancer progression remain unclear. While in vivo human data are an excellent indicator of the disease status and response to therapy, they provide limited insight into the mechanisms of human disease. On the contrary, studies using in vivo animal and in vitro cell culture models allow researchers to directly investigate links between molecular signals and pathogenesis of various diseases. Thus, the 20-gene signature (including orthologs of the corresponding human gene) described in the present disclosure can serve as a valuable in vivo and in vitro research tool for studying diverse processes associated with colorectal cancer and more specifically its liver metastasis. For example, the gene set disclosed herein can be used to gain a better understanding of the molecular mechanisms responsible for the metastatic progression and recurrence in the context of colorectal cancer. Gene profiling of a large panel of colorectal cancer cell lines can lead to identification of cell lines that resemble cancer cells or tissues of patients expected to positively respond to metastatic liver resection.

Comparison analysis of colorectal cancer cell lines distinguished based on the lack or presence of the 20-gene signature disclosed herein can be employed to reveal the cellular processes associated with the signature. For example, standard molecular assays used in basic research (such as proliferation assays, colony formation assays, apoptosis assays, etc.) can point to cellular functions linked to the presence or absence of the gene signature described herein.

Drug Development

In addition to research purposes, profiled colorectal cancer cell lines can further be utilized in drug development. The process of drug development, from start to commercialization is very long and involves numerous steps including identifying in vitro lead drug candidates from a million of compounds, pre-clinical development using in vivo animal models and, finally, clinical trials in humans. High-throughput in vitro screening is a widely used method during the initial stages of drug development, and allows for the simultaneous evaluation of millions of compounds under a given condition. It involves the screening of the candidate therapy compound library against the specific drug target directly or in a more complex assay system. For example, high throughput screen can involve screening a candidate therapy compound library in a cell-based assay, where the activity of the candidate compound is intended to affect a specific cellular process and/or pathway. This type of screen can be carried out using ceils cuitured in multi-well plates with automated operation.

Since the ultimate outcome desired by the candidate compound is sensitivity of cancer cells to a candidate drug, the selection of cancer cell type is dictated by specific goals or a target drug. Because different cell types have different susceptibility to test compounds that are cytotoxic or cause apoptosis, choosing a biologically representative cell line and appropriate assay conditions is crucial for obtaining relevant results. Thus, profiling of colorectal cancer cell lines according to the 20-gene signature disclosed herein could aid in drug design for colorectal cancer drug therapies. Similarly, distinguishing colorectal cancer cell lines based on the 20-gene signature can aid in the selection of ceil types used in xenograft animal models of colorectal cancer.

Kits

The present disclosure provides kits useful for providing prognostic information regarding patients diagnosed with colorectal cancer liver metastasis. In one embodiment, the present disclosure relates to a kit comprising at least one reagent for detecting the level of expression of at least 10, at least 11, 12, 13, 14, 15, 16, 17, 18 19, or 20 genes selected from a group consisting of RBBP8, DKK1, LRRC42, REG4, RAD23B, FGFBP1, NUP62CL, RNF135, HOXC6, DNAJC12, SMIM24, LRP8, RPS24, TYMS, PLA2G2A, CES2, ODC1, SERPINB1, PLCB4, and STEAP1, and any combination of these genes. In one method embodiment, the kit of the present disclosure is used for determining a patient's suitability for colorectal cancer liver metastases resection; in another embodiment, the kit is used for determining a patient's likely response to adjuvant chemotherapy. In one embodiment, the kit comprises a set of capture probes and/or primers specific for at least 10 and as man}' as all the genes listed in Table 2 with intermediate gene numbers possible as described elsewhere herein, as well as reagents sufficient to facilitate detection and/or quantitation of the gene expression product. In another embodiment of the present disclosure, the kit comprises capture probes immobilized on an array. By "array" is intended a solid support or a substrate with peptide or nucleic acid probes attached to the support or substrate. Arrays typically comprise a plurality of different capture probes that are coupled to a surface of a substrate in different, known locations. The arrays comprise a substrate having a plurality of capture probes (for example a pair of such probes) that can specifically bind a gene expression product. The arrays may contain at least 10 or more pluralities (e.g., pairs) of capture probes suitable for the detection of genes listed in Table 2.

In another embodiment, the kit comprises a set of oligonucleotide primers designed for the detection and/or quantitation of the genes listed in Table 2. The oligonucleotide primers may be provided in a lyophilized or reconstituted form, or may be provided as a set of nucleotide sequences. In one embodiment, the primers are provided for example in a micropiate format, where each primer set occupies a well (or multiple wells, as in the case of replicates) in the micropiate. The micropiate may further comprise primers designed for the detection of one or more housekeeping genes or of one or more genes from a non-cancerous cell or both. The kit may further comprise other reagents as well as instructions sufficient for the amplification (e.g., through one or more PGR methods) of expression products from the genes listed in Table 2 or fragments thereof.

Additional components as necessary or desirable may include one or more of buffers, labels, lysis buffer, and optionally a software package of the statistical methods of the invention.

Additionally, the kits of the invention can contain additional instructions for the simultaneous, sequential or separate use of the different components which are in the kit. Methods for Assaying Levels of RNA Transcripts or their Expression

Products

Exemplary methods known in the art for the quantification of mRNA expression in a sample include northern blotting and in situ hybridization (Parker & Barnes, Methods in Molecular Biology 106:247-283 (1999)); RNAse protection assays (Hod, Biotechniques 13:852-854 (1992)); and PCR-based methods, such as reverse transcription polymerase chain reaction (RT-PCR) (Weis et al, Trends in Genetics 8:263-264 (1992)).

Reverse Transcription PCR (RT-PCR) The starting material is typically total RNA isolated from a biological sample, usually from a human tumor. Optionally, normal tissues or cells from the same patient can be used as an internal control. RNA can be extracted from cells or tissue sample, e.g., from a sample that is fresh, frozen (e.g. fresh frozen), or paraffin-embedded and fixed (e.g. formalin-fixed).

General methods for RNA extraction are well known in the art and are disclosed in standard textbooks of molecular biology, including Ausubel et al, Current Protocols of Molecular Biology, John Wiley and Sons (1997). Methods for RNA extraction from paraffin embedded tissues are disclosed, for example, in Rupp and Locker, Lab Invest. 56:A67 (1987), and De Andres et al, BioTechniques 18:42044 (1995). In particular, RNA isolation can be performed using a purification kit, buffer set and protease from commercial manufacturers, such as Qiagen, according to the manufacturer's instructions. For example, total RNA from cells in culture can be isolated using Qiagen RNeasy mini-columns. Other commercially available RNA isolation kits include MasterPure™ Complete DNA and RNA Purification Kit (EPICENTRE®, Madison, WI), and Paraffin Block RNA Isolation Kit (Ambion, Inc.). Total RNA from biological samples can be isolated using RNA Stat-60 (Tel-Test). RNA prepared from a biological sample can be isolated, for example, by cesium chloride density gradient centrifugation. The isolated RNA may then be depleted of ribosomal RNA as described in U.S. Pub. No. 2011/0111409.

The sample containing the RNA is then subjected to reverse transcription to produce cDNA from the RNA template, followed by exponential amplification in a PCR reaction. The reverse transcription step is typically primed using specific primers, random hexamers, or oligo-dT primers, depending on the circumstances and the goal of expression profiling. For example, extracted RNA can be reverse- transcribed using a Gene Amp RNA PCR kit (Perkin Elmer, CA, USA), following the manufacturer's instructions. The derived cDNA can then be used as a template in the subsequent PCR reaction.

PCR-based methods use a thermostable DNA-dependent DNA polymerase, such as a Taq DNA polymerase. For example, TaqMan® PCR typically utilizes the 5 '-nuclease activity of Taq or Tth polymerase to hydrolyze a hybridization probe bound to its target amplicon, but any enzyme with equivalent 5' nuclease activity can be used. Two oligonucleotide primers are used to generate an amplicon typical of a PCR reaction product. A third oligonucleotide, or probe, can be designed to facilitate detection of a nucleotide sequence of the amplicon located between the hybridization sites of the two PCR primers. The probe can be detectably labeled, e.g., with a reporter dye, and can further be provided with both a fluorescent dye, and a quencher fluorescent dye, as in a Taqman® probe configuration. Where a Taqman® probe is used, during the amplification reaction, the Taq DNA polymerase enzyme cleaves the probe in a template-dependent manner. The resultant probe fragments disassociate in solution, and signal from the released reporter dye is free from the quenching effect of the second fluorophore. One molecule of reporter dye is liberated for each new molecule synthesized, and detection of the unquenched reporter dye provides the basis for quantitative interpretation of the data.

TaqMan® RT-PCR can be performed using commercially available equipment, such as, for example, ABI PRISM 7700™ Sequence Detection System™ (Perkin-Elmer- Applied Biosystems, Foster City, CA, USA), or Lightcycler (Roche Molecular Biochemicals, Mannheim, Germany). In a preferred embodiment, the 5' nuclease procedure is run on a real-time quantitative PCR device such as the ABI PRISM 7700™ Sequence Detection System™. The system consists of a thermocycler, laser, charge-coupled device (CCD), camera and computer. The system amplifies samples in a 384-well format on a thermocycler. The RT- PCR may be performed in triplicate wells. During amplification, laser-induced fluorescent signal is collected in real-time through fiber optics cables for all wells, and detected at the CCD. The system includes software for running the instrument and for analyzing the data. 5'-Nuclease assay data are generally initially expressed as a threshold cycle ("Ct").

Fluorescence values are recorded during every cycle and represent the amount of product amplified to that point in the amplification reaction. To minimize errors and the effect of sample-to-sample variation, RT-PCR is usually performed using an internal standard. Gene expression measurements can be normalized relative to the mean of one or more (e.g., 2, 3, 4, 5, or more) reference genes. Real time PCR is compatible both with quantitative competitive PCR, where an internal competitor for each target sequence is used for normalization, and with quantitative comparative PCR using a normalization gene contained within the sample, or a housekeeping gene for RT-PCR. For further details see, e.g. Held et al, Genome Research 6:986-994 (1996).

Design of PCR Primers and Probes

PCR primers and probes can be designed based upon exon, intron, or intergenic sequences present in the RNA transcript of interest. Primer/probe design can be performed using publicly available software, such as the DNA BLAT software developed by Kent, W.J., Genome Res. 12(4):656-64 (2002), or by the BLAST software including its variations.

Factors that can influence PCR primer design include primer length, melting temperature (Tm), and G/C content, specificity, complementary primer sequences, and 3 '-end sequence. In general, optimal PCR primers are generally 17-30 bases in length, and contain about 20-80%, such as, for example, about 50-60% G+C bases, and exhibit Tm's between 50 and 80 °C, e.g. about 50 to 70 °C.

Primers that can be used to detect expression levels of each of the 20 genes disclosed herein are commercially available. Examples of commercially available primers designed to detect genes listed in Table 2 are provided in Table 9.

Table 9. Commercially available primer sets for detection of gene expression levels

RNF135 http://www.bio-rad.com/en-us/prime-pcr- assays/assay/qhsaced0038213-primepcr-sybr-green-assay-rnfl35- human

HOXC6 http://www.sabiosciences.com/primerinfo.php?pcatn=PPH06331E

DNAJC12 http://sabiosciences.com/primerinfo. php?pcatn=PPH15303A

SMIM24 http://www.labome.com/product/OriGene/HP230106.html

LRP8 http://www.sabiosciences.com/primerinfo.php?pcatn=PPH06224A

RPS24 http://www.labome.com/product/OriGene/HP227873.html

PLA2G2A http://www.sabiosciences.com/primerinfo.php?pcatn=PPH05823B

CES2 http://www.sabiosciences.com/primerinfo.php?pcatn=PPH15866B

ODC1 http://www.sabiosciences.com/primerinfo.php?pcatn=PPH00987C

SERPI B1 http://www.sabiosciences.com/primerinfo.php?pcatn=PPH18653A

PLCB4 http://www.bio-rad.com/en-us/prime-pcr- assays/assay/qhsacid0014933-primepcr-sybr-green-assay-plcb4- human

TYMS http ://www. sabiosciences . com/primerinfo. php?pcatn=PPHO 1011 A

STEAP1 http://www.sabiosciences.com/primerinfo.php?pcatn=PPH02268C

Other PCR-based Methods

Further PCR-based techniques that can find use in the methods disclosed herein include, for example, BeadArray® technology (Illumina, San Diego, CA; Oliphant et a!. , Discovery of Markers for Disease (Supplement to Biotechniques), June 2002; Ferguson et al, Analytical Chemistry 72:5618 (2000)); BeadsArray for Detection of Gene Expression® (BADGE), using the commercially available LurninexlOO LabMAP® system and multiple color-coded microspheres (Luminex Corp., Austin, TX) in a rapid assay for gene expression (Yang et al., Genome Res. 1 1 : 1888-1898 (2001)); and high coverage expression profiling (HiCEP) analysis (Fukumura et al. , Nucl. Acids. Res, 31( 16) e94 (2003).

Microarrays

In this method, polynucleotide sequences of interest (including cDNAs and oligonucleotides) are arrayed on a substrate. The arrayed sequences are then contacted under conditions suitable for specific hybridization with detectably labeled cDNA generated from RNA of a sample. The source of RNA typically is total RNA isolated from a biological sample, and optionally from normal tissue of the same patient as an internal control. RNA can be extracted, for example, from frozen or archived paraffin-embedded and fixed (e.g. formalin-fixed) tissue samples.

For example, PCR amplified inserts of cDNA clones of a gene to be assayed are applied to a substrate in a dense array. Usually at least 10,000 nucleotide sequences are applied to the substrate. For example, the microarrayed genes, immobilized on the microchip at 10,000 elements each, are suitable for hybridization under stringent conditions. Fluorescently labeled cDNA probes may be generated through incorporation of fluorescent nucleotides by reverse transcription of RNA extracted from tissues of interest. Labeled cDNA probes applied to the chip hybridize with specificity to each spot of DNA on the array. After washing under stringent conditions to remove non-specifically bound probes, the chip is scanned by confocal laser microscopy or by another detection method, such as a CCD camera. Quantitation of hybridization of each arrayed element allows for assessment of corresponding mRNA abundance. Measuremen i of protein expression

In certain embodiments, the present invention concerns determining the expression level of a protein, corresponding to one or more genes listed in Table 2.

In some embodiments, the present disclosure relates to the detection of proteins, polypeptides, or peptides using immunodetection assays. Immunodetection assays include but are not limited to: enzyme linked immunosorbent assay (ELISA), radioimmunoassay (RIA), immunoradiometric assay, fluoroimmunoassay, chemiluminescent assay, bioiuminescent assay, immunohistochemistiy, and Western blot. The steps of various immunodetection assays have been described in the literature, e.g. Silva, J. ML, McMahon, M. The Fastest Western, in Town: A Contemporary Twist on the Classic Western Blot Analysis. J. Vis. Exp. (84), e51149, doi: 10.3791/51 149 (2014), Am J Physiol Regul Integr Comp Phy siol. 2011 September; 301(3): R632-R640.

In general, the immunobinding methods include obtaining a sample containing a protein, polypeptide and/or peptide, and contacting the sample with a first (or primary) antibody, monoclonal or polyclonal, under conditions effective to allow the formation of detectable immunocomplexes. The immunobinding methods include methods for detecting and quantifying the amount of an antigen component in a sample and the detection and quantification of any immune complexes formed during the binding process. Here, one would obtain a sample suspected of containing an antigen or antigenic domain, and contact the sample with an antibody against the antigen or antigenic domain, and then detect and quantify the amount of immune complexes formed under the specific conditions.

A biological sample analyzed may be any sample that is suspected of containing an antigen or antigenic domain. A biological sample, e.g., a physiological sample that comprises cancer ceils from a patient may be lysed to yield an extract, which comprises cellular proteins. Alternatively, intact cells, e.g., a tissue sample such as paraffin embedded and/or frozen sections of biopsies, are permeabiiized in a manner that permits macromolecules, e.g., antibodies, to enter the cell. The antibodies are then incubated with cells, including permeabiiized cells, e.g., prior to flow cytometry, nuclei or the protein extract, e.g., in a western blot, so as to form a complex. The presence, amount and location of the complex is then determined or delected.

Antibodies are incubated with a biological sample under effective conditions and for a period of time sufficient to allow the formation of immune complexes (primary immune complexes), i.e., for a period of time long enough for the antibodies to bind to any antigens present. Following the incubation step, the sample-antibody composition, such as a tissue section, ELISA plate, dot blot or western, blot, is generally washed to remove any non-specifically bound antibody species, allowing only those antibodies specifically bound within the primary immune complexes to be detected.

In general, the detection of immunocomplexes is well known in the art and may be achieved through the application of numerous approaches. These methods are generally based upon the detection of a label or marker, such as any of those radioactive, fluorescent, biological and enzymatic tags. Additionally, detection using a secondary binding ligand such as a second antibody and/or a biotin/avidin ligand binding arrangement can also be utilized. Method Overview

Referring to Figure 7, an example method begins at operation 600 by taking a biological sample to be tested. Reagents sufficient for the detection of expression levels of genes present in the biological sample are then added to the sample at operation 602. Expression levels of genes can be evaluated my measuring for example the levels of a particular RNA transcript or expression product thereof. The reagents can be designed to detect expression levels of more than one gene from the sample or, alternatively, a single gene at operation 604. Expression products of genes of interest are detected and the level of their expression is measured. Reagents used for the detection of gene expression levels can include, but are not limited to: primers, DNA polymerase, labeled probe, antibodies, antibody detection reagents, etc. At operation 606, a reference expression level for the genes identified in operation 604 is determined. The reference level is a level of expression of the same gene as that found in a cancer cell and identified in operation 604, but in a non-cancerous cell. At operation 608, the measured level of expression for each of the identified genes is compared to the reference level of expression for the identified gene. Following operation 608, a molecular risk score (MRS) is calculated for biological sample based on the difference between the measured level of gene expression and the reference level of expression at operation 610. The MRS calculate in operation 610 is then compared to a predetermined MRS median score at operation 612. Operation 614 determines, based at least in part on the calculated MRS, whether a patient will benefit from undergoing surgical resection of colorectal cancer liver metastasis. In view of the disclosure herein, one skilled in the art will appreciate there are many alternative operations and sequence of operations that can be used to measure gene expression and ultimately determine the outcome after liver resection based on the measured gene expressions.

Systems

The methods and operations disclosed herein can be performed using a variety of different apparatuses and systems. For example, FIG. 8 illustrates an exemplary architecture of a computing device 112 that can be used to implement aspects of the present disclosure, including servers and client devices. The computing device 112 is used to execute the operating system, application programs, and software modules (including the software engines) described herein.

The computing device 112 is capable of generating an MRS based on the expression values obtained from the biological sample. The computing device 112 includes, in at least some embodiments, at least one programmable circuit such as a processing device 120. Examples of processing devices include a central processing unit (CPU) and a microprocessor. A variety of processing devices are available from a variety of manufacturers, for example, Intel, Advanced Micro Devices, Qualcomm, and others. In this example, the computing device 1 12 also includes a system memory 122, and a system bus 124 that couples various system components including the system memory 122 to the processing device 120. The system bus 124 is one of any number of types of bus structures including a memory bus, or memory controller; a peripheral bus; and a local bus using any of a variety of bus architectures. The computing device 112 also can include a graphical processing unit separate from the processing device 120.

Examples of computing devices suitable for the computing device 112 include a desktop computer, a laptop computer, a tablet computer, a mobile phone device such as a smart phone, or other devices configured or programmed to process digital instructions. Each of the above mentioned computing devices is capable of performing functions necessary for the generation of MRS, data storage, data manipulation, etc.

The system memory 122 includes read only memory 186 and random access memory 188. A basic input/output system 130 containing the basic routines that act to transfer information within computing device 1 12, such as during start up, is typically stored in the read only memory 126.

The computing device 112 also includes a secondary storage device 132 in at least some embodiments, such as a hard disk drive, including magnetic and solid state drives, for storing digital data. The secondary storage device 132 is connected to the system bus 124 by a secondary storage interface 134. The secondary storage devices and their associated computer readable media provide nonvolatile storage of computer readable instructions (including application programs and program modules), data structures, and other data for the computing device 1 12.

Although the exemplary environment described herein employs a hard disk drive as a secondary storage device, other types of computer readable storage media are used in other embodiments. Examples of these other types of computer readable storage media include magnetic cassettes, flash memory cards, digital video disks, Bernoulli cartridges, compact disc read only memories, digital versatile disk read only memories, random access memories, or read only memories. Some embodiments include non-transitory media. A number of program modules can be stored in secondary storage device

132 or memory 122, including a system BIOS 190, an operating system 136, one or more application programs 138, other program modules 140, and program data 142.

In at least some embodiments, the data stored in program data 142 can be represented in one or more files having any format usable by a computer. Examples include text files formatted according to a markup language and having data items and tags to instruct computer programs and processes how to use and present the data item. Examples of such formats include markup languages such as html, xml, and xhtml, although other formats for text files can be used. Additionally, the data can be represented using formats other than those conforming to a markup language. In at least some embodiments, the data stored in program data 142 can be represented in one or more files having any format usable by a computer. Examples include text files formatted according to a markup language and having data items and tags to instruct computer programs and processes how to use and present the data item. Examples of such formats include markup languages such as html, xml, and xhtml, although other formats for text files can be used. Additionally, the data can be represented using formats other than those conforming to a markup language.

In at least some embodiments, computing device 112 includes input devices to enable the caregiver to provide inputs to the computing device 1 12. Examples of input devices 144 include a keyboard 146, pointer input device 148, microphone 150, and touch sensitive display 156. Various embodiments also may include other input devices 144. The input devices are often connected to the processing device 120 through an input/output interface 154 that is coupled to the system bus 124. These input devices 144 can be connected by any number of input/output interfaces, such as a parallel port, serial port, game port, or a universal serial bus. At least some embodiments also include wireless communication between input devices and interface 154 such as infrared, BLUETOOTH® wireless technology, 802.11a/b/g/n, cellular, or other radio frequency or optical communication systems.

In this example embodiment, a touch sensitive display device 156 is also connected to the system bus 124 via an interface, such as a video adapter 158. The touch sensitive display device 156 includes touch sensors 144 for receiving input from a user when the user touches or hovers a finger or pointer proximal to the display. Such sensors can be capacitive sensors, pressure sensors, or other touch sensors. The sensors not only detect contact with the display, but also the location of the contact and movement of the contact over time. For example, a user can move a finger or stylus across the screen to provide written inputs. The written inputs are evaluated and, in at least some embodiments, converted into text inputs. It is understood that all user selections described herein may be conducted by utilizing a finger to select or move an item on the touch sensitive display device 156. The touch sensitive display can use various different technologies such as resistive, surface acoustic wave, capacitive, infrared grids, projected optical imaging, dispersive signaling, and any other suitable touch technology. User interfaces displayed on the touch sensitive display device 156 can be operated with other types of input devices such as a mouse, touchpad, or keyboard. Other embodiments can use a non-touch display that is operated with an input device such as a mouse, touchpad, keyboard, or other type of input device. In addition to the display device 156, the computing device 112 can include various other peripheral devices (not shown), such as speakers or a printer.

When used in a local area networking environment or a wide area networking environment (such as the Internet), the computing device 112 is typically connected to a network 110 through a network interface, such as a wireless network interface 160. Other possible embodiments use other communication devices. For example, some embodiments of the computing device 112 include an Ethernet network interface, or a modem for communicating across the network. The computing device 112 typically includes at least some form of computer-readable media. Computer readable media includes any available media that can be accessed by the computing device 112. By way of example, computer- readable media include computer readable storage media and computer readable communication media.

Computer readable storage media includes volatile and nonvolatile, removable and non-removable media implemented in any device arranged and configured to store information such as computer readable instructions, data structures, program modules or other data. Computer readable storage media includes, but is not limited to, random access memory, read only memory, electrically erasable programmable read only memory, flash memory or other memory technology, compact disc read only memory, digital versatile disks or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and that can be accessed by the computing device 112.

Computer readable communication media typically embodies computer readable instructions, data structures, program modules or other data in a modulated data signal such as a carrier wave or other transport mechanism and includes any information delivery media. The term "modulated data signal" refers to a signal that has one or more of its characteristics set or changed in such a manner as to encode information in the signal. By way of example, computer readable communication media includes wired media such as a wired network or direct-wired connection, and wireless media such as acoustic, radio frequency, optical such as infrared, and other wireless media. Combinations of any of the above are also included within the scope of computer readable media.

Experimental Procedures

RNA Isolation and Microarray

RNA was extracted using Trizol (Invitrogen, Carlsbad, CA), quality analyzed using an Agilent Bioanalyzer (Agilent Technology, Palo Alto, CA), and included for microarray analysis if RNA integrity number (RIN) > 7 in both cohorts. For the derivation cohort, extracted total RNA was reverse-transcribed by a previously published method and the resulting complimentary DNA (cDNA) template was applied to gene expression analysis (Ito et al. PLoS One, 8 (12), 2013). The target cDNAs were hybridized to Illumina Human HT-12 Gene Chip containing a total of 47,231 annotated gene probe sets (Illumina, San Diego, CA). Arrays were scanned by using standard Illumina protocols and scanners. For the validation cohort, gene expression was assessed on mRNA expression measured on frozen tumor using a Qiagen microarray (Snoeren et al. PLoS One, 7(11), 2012). All microarray data were log transformed and quantile normalized. Raw microarray data for the derivation cohort are published in http://www.ebi.ac.uk/arrayexpress/experiments/E-MTAB-1951/.

Identification of prognostic genes

The semi-supervised method was used to identify prognostic genes. Briefly, the derivation cohort is randomly partitioned into training (60%) and test (40%) sets, with a fixed ratio of high (40%) and low risk (60%) patients in each group. Risk for this purpose was calculated as the sum of the Fong clinical risk score + any liver recurrence + any recurrence + cancer death, with 1 point assigned for recurrence and 0 for none. Data from the training set was used to identify prognostic genes through the supervised principal components method (SPCM) and the test set was used for its intrainstitutional validation. A gene was considered differentially expressed between cancer and normal tissue if its expression in liver tumor was at least 50% higher or 50% lower than in normal tissue. Genes were selected based on a pre- specified threshold selection frequency of 20% after iterative application of SPCM 1000 times. Following gene selection, standardized gene expression (SGE) was calculated as:

SGE = (Absolute gene expression-mean gene expression)/standard deviation Molecular Risk Score (MRS) was calculated as

MRS = Sum [(log univariate Cox proportional hazards coefficients) * SGE]

The derivation cohort was then partitioned into a low and high risk group based on median MRS, to eliminate the effect of extreme values in the training set thereby ensuring equal numbers of patients in the high and low risk groups. Gene expression in tumor liver was normalized to GAPDH expression and compared to normal human liver obtained from normal primary liver tissue (n=60). Gene expression data were analyzed using R statistical software (version 3.0). Survival probabilities were estimated using the Kaplan-Meier method. Statistical methods:

Categorical variables were compared using the χ² test and continuous variables using the t-test. Univariate analysis was performed using the log-rank test and multivariate analysis using the Cox regression model.

All statistical analysis was performed using StataSE (version 13.1, TX, USA) and Prism (version 6, CA, USA). P<0.05 was considered statistically significant.

EXAMPLES

Example 1

Gene expression analysis in 96 patients

A gene expression analysis was conducted in order to identify a gene set that can be used to assess risk and clinical outcome in patients treated with surgery for colorectal liver metastasis. 96 patients formed the derivation cohort (Table 1 and Fig. 5). Patients in derivation cohort underwent liver resection between January 2000 and October 2007 for metastatic colorectal cancer at Memorial Sloan Kettering Cancer Center (MSKCC). Patients with extrahepatic metastasis, macroscopic residual disease (R2), missing CRS scores, inadequate follow-up, and insufficient RNA were excluded from the study. The primary endpoint evaluated was overall survival (OS), while the secondary endpoints included disease-specific survival (DSS), and recurrence-free survival (RFS).

69 patients (72%) received neoadjuvant chemotherapy (prior to surgery), 79 patients (83%) adjuvant chemotherapy (after surgery), and 34 (35%) hepatic artery infusion (HAI) chemotherapy. 37 patients (39%) had a Fong clinical risk score > 3,7 underscoring the high risk clinically-derived profile of this cohort. 52 patients (54%) had > 3 segments resected, and 58 patients (60%) had > 1 tumor. Median follow-up was 48 months. 66 patients (69%) developed recurrence, 46 patients (48%) died of cancer related causes, and a total of 70 patients (73%) died during the study period (whether of causes known to be cancer related or not). Median overall survival (OS), disease specific survival (DSS), and recurrence free survival (RFS) in the derivation cohort was 52.1, 80.2, and 13 months respectively (FIG. 1).

Table 1: Clinicopathologic variables in derivation and validation cohorts

MSKCC (derivation) Dutch (validation) p-value

(%) (%)

Total number of patients 96 119

Age (median, range) 60 (29-88) 62 (33-85)

Sex < 0.0001

Male 33 (34%) 77 (65%)

Female 63 (66%) 42 (35%)

Tumor Size > 5cm 22 (23%) 38 (32%) NS

Primary Nodal Status NS

N+ 56 (58%) 59 (50%)

N- 40 (42%) 51 (43%)

Missing 0 9 (8%)

> 1 tumor 58 (60%) 63 (53%) NS

DFI < 12 months 51 (53%) 72 (61%) NS

CEA > 200 8 (8.3%) 13 (11%) NS

Neoadjuvant chemotherapy 69 (72%) 64 (54%) < 0.01

Adjuvant chemotherapy 79 (83%) 68 (57%) < 0.001

HAI chemotherapy 34 (35%) 0 (0%) < 0.001

Type of resection < 0.01

Minor (£ 3 segments) 44 (46%) 76 (64%)

Major (> 3 segments) 52 (54%) 43 (36%)

Median follow UD (months) 48 24 < 0.01

RNA was isolated from frozen tumor specimens, reverse-transcribed, and the resulting complimentary DNA (cDNA) template was subjected to gene expression analysis. All microarray data were log-transformed and quantile-normalized. Following a gene expression microarray to assess individual gene expression in the entire derivation cohort, the derivation cohort was randomly partitioned into a training (60 patients) and test (36 patients) set. To identify genes correlating with OS, a supervised principal component analysis (Bair et al., PLoS Biology, 2:4, 2004) was performed first using the training set, with subsequent cross validation in the test set (FIG. 2). To ensure a highly stringent data analysis, this method was iteratively applied 1000 times with randomly generated training and test sets. Genes that significantly correlated with OS with a frequency > 20% after 1000 iterations were selected for construction of the molecular risk score (MRS) (20 genes, Table 2).

The MRS was subsequently validated using validation cohorts from different institutions. Table 2. Genes selected for molecular risk score

CELL CYCLE REGULATOR GENES RBBP8 (Retinoblastoma Binding Protein 8) DKK1 (Dickkopf-related Protein 1) LRRC42 (Leucine Rich Repeat Containing 42) REG4 (Regenerating islet-derived protein 4) RAD23B (UV excision repair protein RAD23 homolog B) FGFBP1 (Fibroblast Growth Factor-Binding Protein 1) NUP62CL (Nucleoporin 62kDa C-terminal Lke)

RNF135 (Ring Finger Protein 135)

HOXC6 (Homeobox protein Hox-C6)

DNAJC12 (DnaJ (Hsp40) Homolog, Subfamily C, Member 12) SMIM24 (Small Integral Membrane Protein 24)

LRP8 (Low Density Lipoprotein Receptor-Related Protein 8), Apolipoprotein E Receptor 2)

RPS24 (Ribosomal Protein S24)

ENZYME REGULATOR GENES PLA2G2A (Phospholipase A2) CES2 (Carboxylesterase 2) ODC1 (Ornithine decarboxylase) SERPINB 1 (Leukocyte elastase inhibitor) PLCB4 (Phospholipase C, Beta 4)

PROTEIN SYNTHESIS GENES TYMS (Thymidylate Synthetase)

DNA DAMAGE REPAIR GENES

STEAP 1 (Six Transmembrane Epithelial Antigen of the Prostate 1) Following gene selection, standardized gene expression (SGE) was calculated as:

SGE = (Absolute gene expression - mean gene expression)/standard deviation Molecular Risk Score (MRS) was calculated as:

MRS = Sum [(log univariate Cox proportional hazards coefficients) * SGE]

The MRS was subsequently calculated for each patient as described above. Genes included in the MRS included predominantly genes controlling cell cycle and enzymatic regulation (FIG. 2). Expression of the 20 identified genes differed significantly in tumor compared to normal liver.

The following table ranks the 20 genes according to frequency of identification as significantly differentially expressed in the training cohort. This can be taken into consideration in deciding to use fewer than all 20 genes identified in Table 2.

Note that of the 20 genes, only 6 genes overlapped with the previously reported 19-gene signature for DSS, and only 5 genes overlapped with the 251 previously reported 115-gene signature for LRFS (Ito H. et al, supra). Only 3 genes were common to all three signatures. Gene expression hierarchical clustering revealed 13 first-order clustering groups (data not shown).

Table 3. Genes ranked according to frequency of identification as significantly differentially expressed in the training cohort.

selected)

RBBP8 1

DNAJC12 2

PLCB4 3

PLA2G2A 4

DKK1 5

TYMS 6

LOC284422/SMIM24 7

LRRC42 8

CES2 9

LRP8 10

REG4 11

RAD23B 12

ODC1 13

STEAP1 14

SERPINB l 15

FGFBP1 16

NUP62CL 17 RNF135 18

HOXC6 19

RPS24 20

Example 2

MRS is the strongest independent prognosticator of OS In order to test the prognostic ability of MRS, the existence of an association between MRS and OS was investigated. The derivation cohort was stratified into low (n=48) and high (n=48) based on the median MRS. In this study, but also generally in the practice of the methods described herein, an MRS higher than the median classified the patient as high risk. Univariate analysis confirmed stratifying by median MRS resulted in groups with different OS (median OS 84 months vs 25 months, HR 3.8, 95% CI 2.2-6.4, pO.001, FIG.4, Table 4). Other variables associated with OS were analyzed, including 3 of the most commonly used clinical risk scores (Fong, Nordlinger, and Iwatsuki). Among the parameters examined, only adjuvant chemotherapy (HR .5, 95% CI .3-.9, P=0.01), and the Fong score (HR 2.7, 95% CI 1.6- 4.4, P<0.001) significantly stratified patients (Table 4). The MRS univariate HR was greater than the Fong score (3.8 vs. 2.7). Neoadjuvant HAI chemotherapy was not significantly associated with OS (HR 2, 95% CI .6-6.6, P=0.2), and adjuvant HAI chemotherapy approached but did not reach statistical significance (HR .6, 95%CI .4-1, P=0.05). Notably, the Nordlinger score and Iwatsuki score that stratified patients into multiple risk categories did not significantly risk stratify for all categories (Table 4) either. On multivariate analysis, only adjuvant chemotherapy (HR .3, 95% CI .1-.6,

P=.001) and MRS (HR 4.2, 95% CI 2.3-7.6, PO.001) remained independently associated with OS. MRS was the only score that was independently associated with OS, and retained a high level of prognostic ability with a magnitude of effect exceeding that observed in the univariate analysis (multivariate HR 4.2 vs. univariate HR 3.8). On contrary, none of the multiple categories proposed by Fong, Nordlinger, or Iwatsuki scores independently correlated with OS (Table 4). Hence, MRS is the strongest prognosticator of OS, is independent of perioperative chemotherapy, and outperforms previously reported clinical risk scores.

Table 4: Univariate and multivariate analysis of clinicopathologic features, clinical risk scores, and molecular risk score with overall survival in MSKCC derivation cohort.

Univariate Multivariate

HR ^CI P-value HR ^L1 P-value

1 .7-1.9 .6

1 1-1.02 .1

Neoadjuvant chemotherapy 1.1 .7-1.9 .6

Neoadjuvant HAI chemotherapy 2 .6-6.6 .2

Adjuvant Chemotherapy .5 .3-.9 .01 .3 .1-.6 .001

Adjuvant HAI chemotherapy .6 .4-1 .05 .9 .6-1.7 .9

Fong Score

Low Ref

High 2.7 1.6-4.4 < 001 1.7 .9-3.3 .1

Nordlinger Score

1 Ref Ref

2 2.1 1.3-3.5 .002 1.8 1-3.1 .05 3 2.2 .5-9.5 .3 2.2 .5-9.6 .3

Iwatsuki Score

1 Ref Ref

2 1.9 .7-5.6 .2 1.1 .4-3.1 .9 3 2.9 1-8.3 .05 1.1 .4-3.7 .8 4 11.4 1.9-66 .007 1 .1-7.4 .9

Molecular Risk Score

Low Ref Ref

High 3.8 2.2-6.4 <0.001 4.2 2.3-7.6 <0.001

Example 3

MRS is the only independent prognosticator of DSS and RFS

The ability of the MRS to stratify patients based by DSS and RFS was next examined in the derivation cohort. MRSs for DSS and RFS were calculated based on their respective univariate iTRs. Similar to OS, the MRS stratified patients into categories with remarkably different DSS (92 months vs 25 months, HR 4.7, 95% CI 2.5-8.9, P<0.0001; FIG. 4, Table 5). Notably, none of the clinical risk scores correlated with DSS on univariate analysis, including the Fong score (HR 1.4, 95% CI .8-2.5, P=0.3). On multivariate analysis, MRS remained the only covariate that was highly independently correlated with DSS (HR 5.9, 95% CI 3-11.6, P<0.001).

Similar to DSS, MRS stratified patients into categories with different RFS (median 28 months vs 7 months, HR 2.2, 95% CI 1.4-3.7, P<0.001; Figure 4, Table 6). On univariate analysis, among the clinical scores, only the Fong score correlated but to a lesser extent than the MRS (HR 1.4, 95% CI 1-2.8, P=0.03). Similar to OS and DSS, the Nordlinger and Iwatsuki scores did not correlate with RFS (Table 6). Multivariate analysis revealed that similar to OS and DSS, MRS remained the only independent predictor of RFS HR 3.7, 95% CI 1.6-4.8, PO.001, Table 6) along with age outperforming the 3 clinical risk scores (Table 6). The HR of age however was 0.9, suggesting minimal prognostic ability (Table 6). Hence, MRS is the exclusive independent prognosticator of OS, DSS and RFS.

Table 5: Univariate and multivariate analysis of clinicopathologic features, clinical risk scores, and molecular risk score with disease-specific (DSS) survival.

Univariate Multivariate

HR CI P-value HR CI P-value

Sex .9 .5-1.7 .9

Age .9 .9-1 .5 .9

Neoadjuvant chemotherapy 1 .5-2 1.1

Neoadjuvant HAI chemotherapy 1.3 .3-5.3

Adjuvant Chemotherapy .9 .4-2

Adjuvant HAI chemotherapy 1.3 .7-2.4

Fong Score

Low Ref

High 1.4 .8-2.5

Nordlinger Score

1 Ref Ref

2 3 1.6-5.5 <0.001 3 1.6-5.7 <0.001 3 3.8 .9-17 .08 3.1 .7-13.8 .1

Iwatsuki Score

1

2 .2-1.6 .3

3 .3-2.1 .7

4 .1-9.9 .9

Molecular Risk Score

Low Ref Ref

High 4.7 2.5-8.9 <0.0001 5.9 3-11.6 <0.001

Table 6: Univariate and multivariate analysis of clinicopathologic features, clinical risk scores, and molecular risk score with recurrence free survival in derivation cohort. Multivariate analyses represent individual comparisons of molecular risk score with each clinical risk score.

Univariate Multivariate

HR CI P- HR CI P- HR CI P- HR CI P- value value value value

Age

≤75 Ref - - - - - Ref - - Ref - -

>75 0.3 0.1-0.9 0.03 0.5 0.2-1.2 0.1 0.3 0.1-0.7 0.008 0.3 0.1-0.8 0.02

Sex 1 0.6-1.8 0.8 - - -

Neoadjuvant chemotherapy 1.8 1-3.3 0.04 1.1 0.6-2 0.9 1.5 0.8-2.8 0.2 1.5 0.8-2.7 0.2

HAI chemotherapy 1 0.6-1.7 0.9 - - -

Adjuvant chemotherapy 0.7 0.4-1.3 0.3

Fong Score

Low Ref - - Ref - - - -

High 4.7 2.8-7.9 <0.001 3.8 2.1-6.7 <0.001

Nordlinger Score

1 Ref - - Ref - -

2 1.8 1.1-2.9 0.02 - 2.0 1.2-3.3 0.005 -

3 6.5 1.5-28 0.01 4.5 1-20 0.04

Iwatsuki Score

1 Ref - - Ref - -

2 1.2 0.5-3.2 0.7 - - 1 0.4-3 0.8

3 2 0.8-5.2 0.1 1.6 0.6-4.1 0.4

4 8.6 1.6-47 0.01 3.8 0.7-22 0.1

Molecular Risk Score

Low Ref - - Ref - - Ref - - - - -

High 2.5 1.5-4.2 <0.001 2.6 1.5-4.3 <0.001 2.9 1.8-4.9 <0.001 2.8 1.7-4.6 <0.00

Example 4

MRS is the only independent prognosticator of OS in the validation cohort

After generation of MRS described in Example 1, its ability to stratify clinical outcomes in patients treated at a different institution was assessed. The MRS was validated in 119 patients (Dutch validation cohort) with surgically resected colorectal liver metastases, treated at the Paul Brousse Hospital (Villejuif, France) and UMC Utrecht (Utrecht, Netherlands) between November 2000 and August 2010.

The validation cohort comprised patients with a similar high risk profile as those included in the derivation cohort (Table 1). 64 patients (54%) received neoadjuvant chemotherapy, 68 patients (57%) adjuvant chemotherapy, and 0 patients HAI chemotherapy. 35 patients (29%) had a Fong clinical risk score > 3. 43 patients (36%) had > 3 segments resected, and 63 (53%) had > 1 tumor. Median follow-up was 24 months. 98 patients (82%) recurred, and 29 patients (24%) died during follow-up. Median OS was 55 months, and median RFS was 10 (FIG. 5A and 5B) months. Although the validation cohort comprised patients that significantly differed from the derivation cohort (Table 1), univariate analysis of OS demonstrated that MRS remained the only covariate that significantly stratified patients (median not reached vs. 39 months; HR 3.7, 95% CI 1.5-9.1, P=0.004 (FIG. 6A and B), Table 7. None of the Fong score, Nordlinger score, or Iwatsuki score were able to stratify patients into statistically different groups (Table7). Adjuvant chemotherapy had a similar effect in the validation cohort as it did in the derivation cohort (HR .5, 95% CI .2-1), however did not reach statistical significance (p=.07). Multivariate analysis was remarkable in that only the MRS remained independently prognostic of OS (HR 3.5, 95% CI 1.4-8.9, p=.01, Table 7), and demonstrated an equivalent magnitude of stratification as it achieved in the derivation cohort (multivariate HR for OS derivation cohort 4.2 vs. multivariate HR for OS validation cohort 3.5). The patient stratification provided by the present MRS score based on the present 20 genes in the derivation cohort is in good agreement with that of the validation cohort. If MRS is plotted against OS as a continuous variable, a scatter plot is obtained. The slope of the "best fit" straight line for that scatter plot is quite close (-11.33 for the derivation cohort vs -10.51 for the external cohort) to that of the validation cohort even though the range of MRS values is different (plots not shown). This can be attributed to differences in methodology.

Next, the efficacy of the MRS in stratifying RFS was evaluated. On univariate analysis, neoadjuvant chemotherapy (HR 1.7, 95% CI 1.1-2.6, p= 009), adjuvant chemotherapy (HR .5, 95% CI .4-.8, p=.003), Fong score (HR 1.8, 95% CI 1.2-2.8, p=.006) and the MRS (median 12 months vs. 8 months, HR 1.5, 05% CI 1-2.3, p=.03) were significant (FIG. 6A and B, Table 8). On multivariate analysis, the Fong score (HR 1.6, 95% CI 1-2.6, p=.02), MRS (1.6, 1-2.4, p=.04), and adjuvant chemotherapy (HR .5, HR .3-.8, p=.001) demonstrated independent prognostic ability (Table 8).

DSS values were not available for the validation cohort so no external validation of DSS was performed.

Table 7: Univariate and multivariate analysis of clinicopathologic features, clinical risk scores, and molecular risk score with overall survival (OS) in Dutch validation cohort.

Univariate Multivariate

Sex HR CI P-value HR CI P-value

1.1 .5-2.5 .7 - - -

Age 1.0 .9-1 .9 - - -

Neoadjuvant chemotherapy .8 .4-1.7 .5 - - -

Adjuvant Chemotherapy .5 .2-1 .07 .7 .3-1.5 .3

Fong Score

Low Ref Ref Ref Ref

High 1 .4-2.4 .9 .8 .3-2.1 .7

Nordlinger Score Ref Ref

1

2 1.7 .5-5.1 .4 1.6 .5-5.4 .4

3 3.4 1.1-10.9 .04 3.3 .8-13.1 .09

Iwatsuki Score

1 Ref Ref

2 .7 .2-2.5 .6 .5 .1-2 .3

3 .9 .2-2.9 .8 .8 .2-3.2 .8

4 1 .3-3.5 .9 1 .2-4.1 1

Molecular Risk Score

Low Ref Ref

High 3.7 1.5-9.1 .004 3.5 1.4-8.9 .01

Table 8: Univariate and multivariate analysis of clinicopathologic features, clinical risk scores, and molecular risk score with recurrence free survival in validation cohort. Multivariate analyses represent individual comparisons of molecular risk score with each clinical risk score.

Univariate Multivariate

HR CI P- HR CI P- HR CI P- HR CI P- value value value value

Age

≤75 Ref - -

>75 1.8 0.9-3.6 0.09

Sex 0.9 0.6-1.4 0.6 - - -

Neoadjuvant chemotherapy 1.7 1.1-2.6 0.009 1.5 0.9-2.3 0.07 1.5 1-2.3 0.04 1.3 0.9-2 0.2

Adjuvant chemotherapy 0.5 0.4-0.8 0.003 0.5 0.3-0.8 0.001 0.5 0.4-0.8 0.003 0.4 0.3-0.7 < 0.001

Fong Score

Low Ref - - Ref - - - -

High 1.8 1.2-2.8 0.009 1.8 1.1-2.8 0.009

Nordlinger Score

1 Ref - - Ref - -

2 1.3 0.8-2.3 0.3 - 1.2 0.7-2.1 0.5 -

3 2.0 1.1-3.5 0.01 2.1 1.1-3.7 0.01

Iwatsuki Score

1 Ref - - Ref - -

2 1.6 0.8-3 0.2 - - 1.8 0.9-3.5 0.08

3 1.7 0.8-3.3 0.1 2 1-4.3 0.04

4 2.2 1-4.2 0.02 3.2 1.5-6.7 0.002

Molecular Risk Score

Low Ref - - Ref - - Ref - - Ref - -

High 1.5 1-2.3 0.04 1.4 0.9-2.1 0.1 1.6 1-2.4 0.03 2.5 1.5-4.1 <0.001

Prophetic Example Evaluation of Differential Expression of 20-Gene Signature

To obtain prognostic information regarding which patients with colorectal cancer liver metastases would benefit from liver resection, MRS is calculated as described herein using the differential expression levels of genes listed in Table 2. In order to determine the extent of differential gene expression for 10 or more of the 20 genes described herein, a comparison of expression levels in cancer versus normal (noncancerous) cells of a subject is performed. This process involves: (i) obtaining a sample containing metastatic liver cells from a patient; (ii) obtaining a sample containing normal liver cells from the same patient; (iii) measuring gene expression quantitatively in both the cancer and normal cells for 10 or more genes from the signature described in Table 2; and (iv) comparing the gene expression levels of genes in the cancer cells to the levels of the same genes in the normal cells. Gene expression levels can be determined either at mRNA or protein level. For example, mRNA is isolated from each biological sample and reverse-transcribed to yield cDNA. cDNA is then used as a template in subsequent PCR amplification reaction. Since housekeeping genes, such as GAPDH, are expressed in a wide variety of tissues and cells with minimal variations in their expression, one or more of them can be used as internal control to normalize the data and thus avoid variability stemming from the measurement process. Finally, mRNA levels of signature genes from cancer cells are compared to the same genes in corresponding normal cells (cells obtained from the same subject). One of the foregoing at least 10 of the 20 genes described herein is considered differentially expressed if its expression level is at least 1.5 fold lower or higher than the expression level of the same gene in a noncancerous cell.

Alternatively, protein levels can be used to determine whether one or more genes is differentially expressed in cancer cells compared to normal cells. For example, following extraction of proteins from each biological sample, Western blot is performed in order to evaluate protein levels of one or more of the genes listed in Table 2. Similarly to RNA analysis, housekeeping genes are used to normalize for the variations among sample handling, such as loading.

Prophetic Example 2

Further Studies of MRS vs. OS Calculation of MRS for a large cohort of patients diagnosed with CRLM

(including both MSKCC and another cohort from another institution, e.g., the Dutch cohort described herein in parallel) using the 20-gene signature described herein would result in a formula of enhanced comparability in terms not only of slope of the best fit scatter plot of MRS v. OS, but also in terms of the range of MRS values. Gene

expression levels for 10 or more of the 20 genes described herein are determined for 96 patients in the MSKCC cohort and for 119 patients in the Dutch cohort using the same procedure to determine expression levels. Housekeeping genes such as GAPDH or β- actin are used for data normalization. All measurements, as well as calculations, are performed in parallel. A plot of MRS (calculated according to the formula provided above) (x axis) versus OS (y-axis) provides a scatterplot. A "best fit" straight line through this scatterplot provides a slope value that is used in stratifying patients.

However, repeating the retrospective study in this manner is anticipated to provide more closely matching MRS range of values. Optionally, a third prospective study arm could also be included, performed in the same manner as outlined for the retrospective studies, using samples from a new cohort of patients diagnosed with CRLM and using MRS as a biomarker to confirm the predictive value of MRS as to outcome of therapy both pre- and post-metastasectomy.The method of data collection will be as outlined above using kits described in this disclosure.

All documents cited herein are incorporated by reference in their entirety for all purposes.

Claims

1. A method for making a therapy decision involving a colorectal cancer patient diagnosed with one or more liver metastases comprising:

a) comparing expression levels within tumor cells from said patient of at least 10 of the following 20 genes: RBBP8, DKK1, LRRC42, REG4, RAD23B, FGFBP1, NUP62CL, RNF135, HOXC6, DNAJC12, SMIM24, LRP8, RPS24, TYMS, PLA2G2A, CES2, ODC1, SERPINB1, PLCB4, and STEAPl, to expression levels of one or more reference genes or to one or more

predetermined values correlating to expression levels of said one or more reference genes;

b) calculating a molecular risk score (MRS) for the patient based on differential expression of said genes compared to the one or more reference genes or value;

c) making a therapy decision with respect to whether to refer said patient to

surgery for resection of said metastases taking into account whether the MRS is high or low compared to a predetermined median score.

2. The method of claim 1 comprising measuring RNA or protein expression in order to determine the expression levels of said at least 10 genes.

3. The method of claim 1 wherein the expression level of said at least 10 genes is normalized to the expression level of one or more reference genes prior to the comparing step.

4. The method of claim 1 comprising comparing expression levels of at least 10 of said genes.

5. The method of claim 1 comprising comparing expression levels of at least 15 of said genes.

6. The method of claim 1 comprising comparing expression levels of 20 of said genes.

7. The method of claim 1 wherein the biological sample is selected from isolated fresh tumor tissue or cells, paraffin embedded tumor tissue, or frozen tumor tissue.

8. The method of claim 1 comprising making a therapy decision selected from the group consisting of one or more of surgical resection of said metastases, surgical resection plus adjuvant chemotherapy, referral to a clinical trial and no further treatment.

9. The method of claim 8 wherein the therapy decision is to perform resection of said liver metastases if the MRS is low or to forgo said resection if the MRS is high.

10. The method of claim 1 wherein said reference genes are the same genes as the at least 10 of the 20 genes but expressed in normal tissue.

11. The method of claim 1 wherein said reference value is an average or median level of expression of the same genes as the at least 10 of the 20 genes but in the tumor cells of a patient cohort diagnosed with colorectal cancer liver metastasis

12. The method of claim 1 wherein the tumor cells are contained in a biological sample obtained from said patient.

13. The method of claim 2, wherein said expression is measured at the mRNA level.

14. The method of claim 2 wherein said mRNA is reverse transcribed to cDNA prior to the measuring step.

15. The method of claim 13 where said mRNA is amplified prior to said measuring.

16. The method of claim 14 wherein said cDNA is amplified prior to said measuring.

17. The method of treating a colorectal cancer patient diagnosed with one or more liver metastases comprising:

a) obtaining a comparison of expression levels within tumor cells from said patient of at least 10 of the following 20 genes: RBBP8, DKK1, LRRC42, REG4, RAD23B, FGFBP1, NUP62CL, RNF135, HOXC6, DNAJC12, SMIM24, LRP8, RPS24, TYMS, PLA2G2A, CES2, ODC1, SERPINBl, PLCB4, and STEAP1, with expression levels of one or more reference genes or to one or more predetermined values correlating to expression levels of said one or more reference genes;

b) calculating a molecular risk score (MRS) for the patient based on differential expression of said genes compared to the one or more reference genes or value;

c) treating the patient with surgical resection of said metastases when the

patient's MRS is low compared to a predetermined median score provided the patient has the physiologic ability to withstand said surgery; or refraining from subjecting the patient to surgical resection of said metastases when the patient's MRS is high compared to the predetermined median.

18. A kit for comparing expression levels of at least 10 of the following 20 genes:

RBBP8, DKK1, LRRC42, REG4, RAD23B, FGFBP1, NUP62CL, R F135, HOXC6, DNAJC12, SMIM24, LRP8, RPS24, TYMS, PLA2G2A, CES2, ODCl, SERPINB l, PLCB4, and STEAPl,in cancer cells within a biological sample from a patient diagnosed with one or more colorectal cancer liver metastases, to expression levels of reference genes or to a predetermined reference value correlating with the expression level of each of said at least 10 genes, the kit comprising:

a. reagents for measuring the expression levels of said at least 10 genes from cancer cells of said patient and optionally for measuring expression levels of the same genes from noncancerous liver cells from the patient if the comparison will be to the same genes from noncancerous cells; b. instructions for effecting the comparison;

c. instructions for deriving an MRS score for the patient and comparing it to a predetermined mean MRS score.

19. The kit of claim 18 wherein the reagents comprise reagents for determining

expression levels for the at least five genes from the patient's cancerous cells, the kit further comprising instructions regarding predetermined reference values for said genes for the comparison of said expression levels.

20. The kit of claim 18, wherein the reagents comprise one or more of nucleic acid primers, buffers, labels, lysis buffer, antibodies, and optionally a software package of statistical methods for calculating the MRS score.

21. A computer-readable storage medium storing one or more programs, the one or more programs comprising instructions, which when executed by a device comprising a processor and a memory, cause the device to:

a) obtaining differential expression levels within tumor cells of at least 10 of the following 20 genes from a colorectal cancer patient: RBBP8, DKK1, LRRC42, REG4, RAD23B, FGFBP1, NUP62CL, R F135, HOXC6, DNAJC12, SMIM24, LRP8, RPS24, TYMS, PLA2G2A, CES2, ODC1, SERPINB 1, PLCB4, and STEAPl;

b) calculating a molecular risk score (MRS) for the patient based on the

differential expression levels;

c) providing a report, wherein the report comprises the MRS.

22. The computer-readable storage medium of claim 21, wherein the MRS is

calculated according to the following expression:

MRS = Sum [(log univariate Cox proportional hazards coefficients) * (Absolute gene expression - mean gene expression) / standard deviation]

23. The computer-readable storage medium of claim 21, the report further comprising a clinical risk score.

24. The computer-readable storage medium of claim 23, wherein the clinical risk score is based on lymph node status of a primary tumor, disease-free interval, serum carcinoembryonic antigen level prior to liver resection, number of hepatic tumors, and tumor size.

25. The computer-readable storage medium of claim 23, wherein the report further comprises a recommendation of no surgery, surgery, or surgery and adjuvant chemotherapy.