DIAGNOSTIC FOR COLORECTAL CANCER

Abstract

The present invention provides a method for diagnosing or detecting colorectal cancer in a subject, the method comprising determining the presence and/or level of biomarkers selected from IL-8, IGFBP2, MAC2BP, M2PK, IL-13, DKK-3, EpCam, MIP1β, TGFβ1, and TIMP-1. The invention also relates to diagnostic kits comprising reagents for determining the presence and/or level of the biomarkers and methods of detecting or diagnosing colorectal cancer.

Description

FIELD OF THE INVENTION

The present invention relates to determining the presence and/or level of biomarkers for detecting or diagnosing colorectal cancer. The invention also relates to diagnostic kits comprising reagents for determining the presence and/or level of the biomarkers and methods of detecting or diagnosing colorectal cancer.

BACKGROUND OF THE INVENTION

Colorectal cancer, also referred to as colon cancer or bowel cancer, is the second most common cause of cancer worldwide. There is an annual incidence of almost a million colorectal cancer cases with an annual mortality around 500,000 (Cancer in Australia: an overview, 2008). Unfortunately, 30-50% of patients have occult or overt metastases at presentation and once tumours have metastasized prognosis is very poor with a five year survival of less than 10% (Etzioni et al., 2003). By contrast, greater than 90% of patients who present while the tumour is still localised will still be alive after 5 years and can be considered cured. The early detection of colorectal lesions would therefore significantly reduce the impact of colon cancer (Etzioni et al., 2003).

The current screening assays in widespread use for the diagnosis of colorectal cancer are the faecal occult blood test (FOBT), flexible sigmoidoscopy, and colonoscopy (Lieberman, 2010). FOBT has relatively low specificity resulting in a high rate of false positives. All positive FOBT must therefore be followed up with colonoscopy. Sampling is done by individuals at home and requires at least two consecutive faecal samples to be analysed to achieve optimal sensitivity. Some versions of the FOBT also require dietary restrictions prior to sampling. FOBT also lacks sensitivity for early stage cancerous lesions that do not bleed into the bowel and as stated above, these are the lesions for which treatment is most successful.

While FOBT screening does result in reduction of mortality due to colorectal cancer it suffers from a low compliance rate (30-40%), most likely due to the unpalatable nature of the test, which limits its usefulness as a screening tool. Colonoscopy is the current gold standard and has a specificity of greater then 90% but it is intrusive and costly with a small but finite risk of complications (2.1 per 1000 procedures) (Levin, 2004). Development of a rapid, specific, cheap blood based assay would overcome compliance issues commonly seen with other screening tests (Tonus, 2006; Hundt et al., 2007) and would be more acceptable as part of a large screening assay.

SUMMARY OF THE INVENTION

The present inventors investigated over sixty biomarkers associated with colorectal cancer, but found that none of the biomarkers alone would be suitable as a diagnostic test. Surprisingly, it was found that determining the presence and/or level of at least two biomarkers associated with colorectal cancer in a sample from a subject allowed for the detection or diagnosis of colorectal cancer at any of the stages of disease. Determining the presence and/or level of at least two biomarkers advantageously provides a diagnostic test that is at least comparable in sensitivity and specificity to the FOBT.

Accordingly, in one aspect, the present invention provides a method for diagnosing or detecting colorectal cancer in a subject, the method comprising:

i) determining the presence and/or level of at least two biomarkers selected from IL-8, IGFBP2, MAC2BP, M2PK, IL-13, DKK-3, EpCam, MIP1β, TGFβ1, and TIMP-1 in a sample from the subject,

wherein the presence and/or level of the two biomarkers is indicative of colorectal cancer.

In one embodiment, the method comprises determining the presence and/or level of two biomarkers selected from M2PK, EpCam, IL-13, DKK-3, IL-8 and IGFBP2.

In another embodiment, the method comprises determining the presence and/or level of expression of at least three of the biomarkers.

In one embodiment, the three biomarkers are selected from M2PK, EpCam, IL-13, DKK-3, IL-8, IGFBP2, MIP1β, TGFβ1 and MAC2BP.

In one particular embodiment, the method comprises determining the presence and/or level of three biomarkers, wherein the three biomarkers are:

i) DKK-3, M2PK, and IGFBP2;

ii) M2PK, IGFBP2, and EpCAM;

iii) M2PK, MIP1β, and TGFβ1; or

iv) IL-8, IL-13, and MAC2BP.

In another embodiment, the method comprises determining the presence and/or level of expression of at least four of the biomarkers.

In one particular embodiment, the method comprises determining the presence and/or level of four biomarkers, wherein the four biomarkers are:

i) DKK-3, M2PK, MAC2BP, and IGFBP2;

ii) IL-8, IL-13, MAC2BP, and EpCam;

iii) DKK3, M2PK, TGFβ1, and TIMP-1;

iv) M2PK, MIP1β, IL-13, and TIMP-1; or

v) IL-8, MAC2BP, IGFBP2, and EpCam.

In yet another embodiment, the method comprises determining the presence and/or level of at least five of the biomarkers.

In one particular embodiment, the five biomarkers are IL-8, IGFBP2, MAC2BP, M2PK, and IL-13.

In another embodiment, the method comprises determining the presence and/or level of at least six of the biomarkers.

In another embodiment, the method comprises determining the presence and/or level of at least seven of the biomarkers.

In one particular embodiment, the seven biomarkers are:

i) IL-8, IGFBP2, MAC2BP, M2PK, IL-13, DKK-3, and TGF β1; or

ii) IL-8, IGFBP2, MAC2BP, M2PK, IL-13, EpCam, and MIP1β.

In yet another embodiment, the method comprises determining the presence and/or level of at least eight of the biomarkers.

In one embodiment, the method comprises determining the presence and/or level of at least nine of the biomarkers.

In yet another embodiment, the method comprises determining the presence and/or level of at least ten of the biomarkers.

In another embodiment, the method comprises determining the presence and/or level of a combination of biomarkers as provided in any of Tables 7 to 18.

In another embodiment, the method comprises detecting the presence and/or level of least one additional biomarker selected from IGF-I, IGF-II, IGF-BP2, Amphiregulin, VEGFA, VEGFD, MMP-1, MMP-2, MMP-3, MMP-7, MMP-9, TIMP-1, TIMP-2, ENA-78, MCP-1, MIP-1β, IFN-γ, IL-10, IL-13, IL-1β, IL-4, IL-8, IL-6, MAC2BP, Tumor M2 pyruvate kinase, M65, OPN, DKK-3, EpCam, TGFβ-1, and VEGFpan.

In one embodiment, the method diagnoses or detects colorectal cancer with a sensitivity of at least 50%.

In another embodiment, the method diagnoses or detects colorectal cancer with a sensitivity of at least 66%.

In yet another embodiment, the method diagnoses or detects colorectal cancer with a sensitivity of at least 77%.

In one embodiment, the method diagnoses or detects colorectal cancer with a specificity of at least 75%.

In one embodiment, the method diagnoses or detects colorectal cancer with a specificity of at least 80%.

In another embodiment, the method diagnoses or detects colorectal cancer with a specificity of at least 90%.

In yet another embodiment, the method diagnoses or detects colorectal cancer with a specificity of at least 95%.

In another embodiment, the method diagnoses or detects Dukes Stage A colorectal cancer with a sensitivity of at least 50% and a specificity of at least 95%.

In yet another embodiment, the method diagnoses or detects Dukes Stage A colorectal cancer with a sensitivity of at least 60% and a specificity of at least 80%.

In another embodiment, the method diagnoses or detects Dukes Stage A colorectal cancer with a sensitivity of at least 50% and a specificity of at least 90%.

The skilled person will understand that Dukes Stage A corresponds to TNM Classifications T1, N0, M0 and T2, N0, M0.

Thus in one embodiment, the method diagnoses or detects TNM Classification T1, N0, M0 or T2, N0, M0 colorectal cancer with a sensitivity of at least 50% and a specificity of at least 95%.

In yet another embodiment, the method diagnoses or detects TNM Classification T1, N0, M0 or T2, N0, M0 colorectal cancer with a sensitivity of at least 60% and a specificity of at least 80%.

In another embodiment, the method diagnoses or detects TNM Classification T1, N0, M0 or T2, N0, M0 colorectal cancer with a sensitivity of at least 50% and a specificity of at least 90%.

Any suitable technique for the detection of polypeptides may be used in the methods of the invention. In one embodiment, the method comprises contacting the sample with at least one compound that binds a biomarker polypeptide. Alternatively, the method comprises detecting the polypeptides by mass spectrometry.

In one particular embodiment, the compound is detectably labelled.

In another embodiment, the compound is an antibody.

In one embodiment, the compound is bound to a solid support.

In the methods of the invention, determining the presence and/or level of the biomarker may comprise determining the presence and/or level of a polynucleotide encoding the biomarker, such as a biomarker gene transcript. Thus, in one embodiment, the biomarkers are polynucleotides.

In yet another embodiment of the methods of the invention, the method comprises:

i) determining the presence and/or level of the biomarkers in the sample from the subject; and

ii) comparing the presence and/or level of the biomarkers to a control, wherein a presence and/or level in the sample that is different to the control is indicative of colorectal cancer.

In one embodiment, the sample comprises blood, plasma, serum, urine, platelets, magakaryocytes or faeces.

In another aspect, the present invention provides a method of treatment comprising:

(i) diagnosing or detecting colorectal cancer according to the method of the invention; and

(ii) administering or recommending a therapeutic for the treatment of colorectal cancer.

In yet another aspect, the present invention provides a method for monitoring the efficacy of treatment of colorectal cancer in a subject, the method comprising treating the subject for colorectal cancer and then detecting the presence and/or level of at least two biomarkers selected from IL-8, IGFBP2, MAC2BP, M2PK, IL-13, DKK-3, EpCam, MIP1β, TGFβ1, and TIMP-1 in a sample from the subject, wherein an absence of and/or reduction in the level of expression of the polypeptides after treatment when compared to before treatment is indicative of effective treatment.

In another aspect, the present invention provides an array of at least two compounds for the diagnosis or detection of colorectal cancer, wherein each of the compounds binds a different biomarker polypeptide selected from IL-8, IGFBP2, MAC2BP, M2PK, IL-13, DKK-3, EpCam, MIP1β, TGFβ1, and TIMP-1.

In yet another aspect, the present invention provides a kit for diagnosing or detecting colorectal cancer in a subject, the kit comprising two compounds that each binds a different biomarker polypeptide selected from IL-8, IGFBP2, MAC2BP, M2PK, IL-13, DKK-3, EpCam, MIP1β, TGFβ1, and TIMP-1.

Throughout this specification the word “comprise”, or variations such as “comprises” or “comprising”, will be understood to imply the inclusion of a stated element, integer or step, or group of elements, integers or steps, but not the exclusion of any other element, integer or step, or group of elements, integers or steps.

As will be apparent, preferred features and characteristics of one aspect of the invention are applicable to many other aspects of the invention.

The invention is hereinafter described by way of the following non-limiting Examples and with reference to the accompanying figures.

BRIEF DESCRIPTION OF THE ACCOMPANYING DRAWINGS

FIG. 1. In Study 3 an optimum combination of the 46 potential protein biomarkers was found using logistic regression modelling, resulting in a panel of seven biomarkers and is illustrated as a ROC curve (black curve). The performance of this “panel” on independent data was estimated using “leave one out” cross-validation (grey curve). The vertical lines are drawn at points of 80% and 90% specificity—operating points of interest in screening tests. Performance statistics are given in Table 5.

FIG. 2. Performance of a seven biomarker model identifying colorectal cancer patients from normals at each Dukes Stage illustrated by ROC curves for each stage. A (red)—Stage A, B (green)—Stage B, C (blue)—Stage C, and D (black)-Stage D from Study 3a. Performance characteristics are given in Table 6.

FIG. 3. When biomarker results from Study 4 (also referred to as Study 3 remeasured) were modelled in pairs a total of 5 pairs (out of a possible 45 combinations selected from the list of 10 biomarkers above) could be shown to produce a sensitivity above 52% at a specificity of 95. The performance of these pair wise biomarker combinations is illustrated as ROC curves (n=5 curves). Performance characteristics are given in Table 7.

FIG. 4. An example of a 3 biomarker model generated from Study 4 data which had a sensitivity of at least 50% at 95% specificity. There were 968 possible 3-biomarker combinations and approximately half of those combinations showed a performance of at least 50% sensitivity at 90% specificity.

FIG. 5. ROC curves are illustrated for all combinations of 3-10 biomarkers generated from Study 4 data which have a sensitivity of at least 50% at 95% specificity (n=485 cross validated curves out of a possible 968 models).

FIG. 6. Frequency of each biomarker in the best 485 models. These BMs represent all serum models that gave a sensitivity of at least 50% at 95% The high representation of all 10 biomarkers in the useful models demonstrates the unity of our selection of these 10 biomarkers.

FIG. 7. A 5 biomarker model generated from Study 4 data is illustrated as a ROC curve (black) and cross validated ROC curve (grey). This model shows a sensitivity of 68% at 95% specificity when all stages of disease are included and when cross validated gave a sensitivity of 64%. Biomarkers included are [IL-8, IGFBP2, Mac2BP, DKK-3 and M2PK].

FIG. 8. A 6 biomarker model generated from Study 4 data is illustrated as a ROC curve (black) and cross validated ROC curve (Grey). This model shows a sensitivity of 77% at a specificity of 95% when all stages of disease are included and when cross validated gave a sensitivity of 67%. Biomarkers included are [IL-8, IGFBP2, Mac2BP, DKK-3, TGFbeta1&M2PK].

FIG. 9. Two alternative seven biomarker models generated from Study 3a data are shown. One was optimised for high specificity (black/new) and an alternative or model optimised for area under the curve is shown (grey/old). At 90% specificity the sensitivity was 72% for the new model and 77% for the older model. Biomarkers included were as follows:

New: IL8, IGFBP2, s90MAC2BP, M2PK, DKK-3, IL-13 & TGFbeta,

Old: IL8, IGFBP2, s90MAC2BP, M2PK, EpCAM, IL13 & MIP-1b.

FIG. 10. A seven biomarker model generated from Study 4 data is illustrated as a ROC curve (black) and cross validated ROC curve (grey). This model shows, a sensitivity of 84% at a specificity of 95%. Biomarkers included are [M2PK serum, IL8.plasma, TGF beta1.serum, IGFBP2.plasma, Mac2BP.serum, TIMP1.plasma and Dkk3 plasma.

FIG. 11. Cross validated ROC curves showing the performance of a 3 biomarker model for each Dukes stage is illustrated. This data demonstrates the validity of the choice of three biomarkers (DKK-3, M2PK and IGFBP2) for detecting cancer at different stages of the disease progression. The data indicates that at Stage A if the three markers are used, the test still will achieve a significant sensitivity of 64% at 95% specificity which is comparable to the sensitivity achieved at late stage disease (79%). That is the biomarker panel of three will pick up early disease states allowing early detection. Biomarkers included are Dkk3, M2PK and IGFBP2.

KEY TO THE SEQUENCE LISTING

• SEQ ID NO:1—amino acid sequence of IL-8
• SEQ ID NO:2—amino acid sequence of IGFBP2
• SEQ ID NO:3—amino acid sequence of MAC2BP
• SEQ ID NO:4—amino acid sequence of M2PK variant 1
• SEQ ID NO:5—amino acid sequence of M2PK variant 2
• SEQ ID NO:6—amino acid sequence of M2PK variant 3
• SEQ ID NO:7—amino acid sequence of IL-13
• SEQ ID NO:8—amino acid sequence of DKK-3 variant 1
• SEQ ID NO:9—amino acid sequence of DKK-3 variant 2
• SEQ ID NO:10—amino acid sequence of DKK-3 variant 3
• SEQ ID NO:11—amino acid sequence of EpCam
• SEQ ID NO:12—amino acid sequence of MIP1β
• SEQ ID NO:13—amino acid sequence of TGFβ1
• SEQ ID NO:14—amino acid sequence of TIMP-1

DETAILED DESCRIPTION

General Techniques and Definitions

Unless specifically defined otherwise, all technical and scientific terms used herein shall be taken to have the same meaning as commonly understood by one of ordinary skill in the art (e.g., in cell culture, molecular genetics, immunology, immunohistochemistry, protein chemistry, and biochemistry).

Unless otherwise indicated, the recombinant protein, cell culture, and immunological techniques utilized in the present invention are standard procedures, well known to those skilled in the art. Such techniques are described and explained throughout the literature in sources such as, J. Perbal, A Practical Guide to Molecular Cloning, John Wiley and Sons (1984), J. Sambrook et al., Molecular Cloning: A Laboratory Manual, 3^rd edn, Cold Spring Harbour Laboratory Press (2001), R. Scopes, Protein Purification—Principals and Practice, 3^rd edn, Springer (1994), T. A. Brown (editor), Essential Molecular Biology: A Practical Approach, Volumes 1 and 2, IRL Press (1991), D. M. Glover and B. D. Hames (editors), DNA Cloning: A Practical Approach, Volumes 1-4, IRL Press (1995 and 1996), and F. M. Ausubel et al. (editors), Current Protocols in Molecular Biology, Greene Pub. Associates and Wiley-Interscience (1988, including all updates until present), Ed Harlow and David Lane (editors) Antibodies: A Laboratory Manual, Cold Spring Harbour Laboratory, (1988), and J. E. Coligan et al. (editors) Current Protocols in Immunology, John Wiley & Sons (including all updates until present).

As used herein, the term “colorectal cancer”, also known as “colon cancer”, “bowel cancer” or “rectal cancer”, refers to all forms of cancer originating from the epithelial cells lining the large intestine and/or rectum.

As used herein, “biomarker” refers to any molecule, such as a gene, gene transcript (for example mRNA), peptide or protein or fragment thereof produced by a subject which is useful in differentiating subjects having colorectal cancer from normal or healthy subjects.

As used herein, the term “diagnosis”, and variants thereof such as, but not limited to, “diagnose”, “diagnosed” or “diagnosing” shall not be limited to a primary diagnosis of a clinical state, but should be taken to include diagnosis of recurrent disease.

As used herein, the term “subject” refers to any animal that may develop colorectal cancer and includes animals such as mammals, e.g. humans, or non-human mammals such as cats and dogs, laboratory animals such as mice, rats, rabbits or guinea pigs, and livestock animals. In a preferred embodiment, the subject is a human.

The “sample” may be of any suitable type and may refer, e.g., to a material in which the presence or level of biomarkers can be detected. Preferably, the sample is obtained from the subject so that the detection of the presence and/or level of biomarkers may be performed in vitro. Alternatively, the presence and/or level of biomarkers can be detected in vivo. The sample can be used as obtained directly from the source or following at least one step of (partial) purification. The sample can be prepared in any convenient medium which does not interfere with the method of the invention. Typically, the sample is an aqueous solution, biological fluid, cells or tissue. Preferably, the sample is blood, plasma, serum, urine, platelets, megakaryocytes or faeces. Pre-treatment may involve, for example, preparing plasma from blood, diluting viscous fluids, and the like. Methods of treatment can involve filtration, distillation, separation, concentration, inactivation of interfering components, and the addition of reagents. The selection and pre-treatment of biological samples prior to testing is well known in the art and need not be described further.

As used herein the terms “treating”, “treat” or “treatment” include administering a therapeutically effective amount of a compound sufficient to reduce or delay the onset or progression of colorectal cancer, or to reduce or eliminate at least one symptom of colorectal cancer.

Biomarkers

The present inventors have shown that determining the presence and/or level of least two biomarkers in a sample from a subject allows for the detection or diagnosis of colorectal cancer, either early detection at Dukes Stage A or at some later stage such as Dukes Stage B or C or D, with specificity and sensitivity comparable to or greater than that achieved with the FOBT. The at least two biomarkers that are useful in the methods of the present invention are selected from IL-8 (interleukin-8), IGFBP2 (insulin-like growth factor binding protein-2), MAC2BP (MAC2-binding protein; serum protein 90K), M2PK (pyruvate kinase muscle 2, pyruvate kinase 3), IL-13 (interleukin-13), DKK-3 (dickkopf homolog, 3), EpCAM (epithelial cell adhesion molecule), MIP1β (macrophage inflammatory protein 1β, CCL4, MIP1beta), TGFβ1 (transforming growth factor β1 , TGFbeta1) and TIMP-1 (tissue inhibitor of metalloproteinase 1). Reference to any of these biomarkers includes reference to all polypeptide and polynucleotide variants such as isoforms and transcript variants as would be known by the person skilled in the art. NCBI accession numbers of representative sequences for each of the biomarkers are provided in Table 1.

TABLE 1
NCBI accession numbers for representative biomarker sequences.
Biomarker Representative NCBI Accession Numbers
IL-8      NM_000584.2 (SEQ ID NO: 1)
IGFBP2    NM_000597.2 (SEQ ID NO: 2)
MAC2BP    NM_005567.3 (SEQ ID NO: 3)
M2PK      NM_002654.3; NM_182470.1; NM_182471.1 (SEQ ID
          NOs: 4-6)
IL-13     NM_002188.2 (SEQ ID NO: 7)
DKK-3     NM_015881.5; NM_013253; NM_001018057.1 (SEQ ID
          NOs: 8-10)
EpCam     NM_002354.2 (SEQ ID NO: 11)
MIP1β     NM_002984.2 (SEQ ID NO: 12)
TGFβ1     NM_000660.4 (SEQ ID NO: 13)
TIMP-1    NM_003254.2 (SEQ ID NO: 14)

Detecting or Diagnosing Colorectal Cancer

It will be apparent from the preceding description that the diagnostic methods of the present invention may involve a degree of quantification to determine levels biomarkers in patient samples. Such quantification is readily provided by the inclusion of appropriate control samples.

In one embodiment, internal controls are included in the methods of the present invention. A preferred internal control is one or more samples taken from one or more healthy individuals.

In the present context, the term “healthy individual” shall be taken to mean an individual who is known not to suffer from colorectal cancer, such knowledge being derived from clinical data on the individual, including, but not limited to, a different diagnostic assay to that described herein.

As will be known to those skilled in the art, when internal controls are not included in each assay conducted, the control may be derived from an established data set.

Data pertaining to the control subjects are preferably selected from the group consisting of:

1. a data set comprising measurements of the presence or level of expression of biomarkers for a typical population of subjects known to have colorectal cancer;

2. a data set comprising measurements of the presence or level of biomarkers for the subject being tested wherein said measurements have been made previously, such as, for example, when the subject was known to be healthy or, in the case of a subject having colorectal cancer, when the subject was diagnosed or at an earlier stage in disease progression;

3. a data set comprising measurements of the presence or level of biomarkers for a healthy individual or a population of healthy individuals; and

4. a data set comprising measurements of the presence or level of biomarkers for a normal individual or a population of normal individuals.

In the present context, the term “typical population” with respect to subjects known to have colorectal cancer shall be taken to refer to a population or sample of subjects diagnosed with colorectal cancer that is representative of the spectrum of colorectal cancer patients. This is not to be taken as requiring a strict normal distribution of morphological or clinicopathological parameters in the population, since some variation in such a distribution is permissible. Preferably, a “typical population” will exhibit a spectrum of colorectal cancer at different stages of disease progression. It is particularly preferred that a “typical population” exhibits the expression characteristics of a cohort of subjects as described herein.

The term “normal individual” shall be taken to mean an individual that does not express a biomarker, or expresses a biomarker at a low level in a sample. As will be known to those skilled in the art, data obtained from a sufficiently large sample of the population will normalize, allowing the generation of a data set for determining the average level of a particular biomarker.

Those skilled in the art are readily capable of determining the baseline for comparison in any diagnostic assay of the present invention without undue experimentation, based upon the teaching provided herein.

Compounds that bind a biomarker when used diagnostically may be linked to a diagnostic reagent such as a detectable label to allow easy detection of binding events in vitro or in vivo. Suitable labels include radioisotopes, dye markers or other imaging reagents for detection and/or localisation of target molecules. Compounds linked to a detectable label can be used with suitable in vivo imaging technologies such as, for example, radiology, fluoroscopy, nuclear magnetic resonance imaging (MRI), CAT-scanning, positron emission tomography (PET), computerized tomography etc.

The diagnostic methods of the present invention are able to diagnose or detect colorectal cancer with a sensitivity and specificity that is at least comparable to FOBT, or greater. As would be understood by the person skilled in the art, sensitivity refers to the proportion of actual positives in the diagnostic test which are correctly identified as having colorectal cancer. Specificity measures the proportion of negatives which are correctly identified as not having colorectal cancer. In one embodiment, the methods of the invention are able to diagnose or detect colorectal cancer with a sensitivity of at least 50%, 60% or 66%, or at least 77%, 80%, 83%, 85%, 86%, 87%, 88%, 89%, 90%, or at least 93%. In another embodiment, the methods of the invention are able to diagnose or detect colorectal cancer with a sensitivity of at least 80%, or at least 85% or at least 90%, or at least 95%.

In one embodiment, the methods of the invention are able to diagnose or detect colorectal cancer with a specificity of at least 75%, 80%, 85%, 90%, 91%, 92%, 93%, 94% or at least 95%.

Advantageously, the methods of the present invention are able to detect colorectal cancer at all of the Dukes Stages with greater sensitivity than the FOBT. In Dukes Stage A, the tumor has penetrated into, but not through, the bowel wall. In Dukes Stage B, the tumor has penetrated through the bowel wall but there is not yet any lymph node involvement. In Dukes Stage C, the cancer involves regional lymph nodes. In Dukes Stage D, there is distant metastasis, for example, to the liver or lung. In one embodiment, the methods of the present invention are able to diagnose or detect colorectal cancer at any Dukes Stage with a sensitivity of at least 80%.

As known to the skilled person, there are other systems for staging cancer that are know in the art. One example is the TMN Classification of Malignant Tumors (TNM) that is used by the American Joint Committee on Cancer (AJCC: Colon and rectum. in Edge et al., eds; AJCC Cancer Staging Manual, 7^th ed. New York, N.Y.: Springer, 2010, pp: 143-164). Another example is the Modified Astler-Coller classification (MAC).

Accordingly, the skilled person will appreciate that the Dukes Stages correspond to certain TNM Classifications. For example, Dukes Stage A corresponds to T1, T2, N0 and M0; Dukes Stage B corresponds to T3, T4a, T4b, N0 and M0; and Dukes Stage C corresponds to i) T1-T2, N1/N1c, M0; ii) T1, N2a and M0; iii) T3-T4a, N1/N1c and M0; iv) T2-T3, N2a and M0; v) T1-T2, N2b and M0; vi) T4a, N2a and M0; vii) T3-T4a, N2b and M0; and viii) T4b, N1-N2 and M0. Thus, the skilled person will understand that reference to a Dukes Stage as used herein includes reference to the corresponding TMN classification as known in the art.

Protein Detection Techniques

In one embodiment, biomarker polypeptide is detected in a patient sample, wherein the presence and/or level of the polypeptide in the sample is indicative of colorectal cancer. For example, the method may comprise contacting a biological sample derived from the subject with a compound capable of binding to a biomarker polypeptide, and detecting the formation of complex between the compound and the biomarker polypeptide. The term “biomarker polypeptide” as used herein includes fragments of biomarker polypeptides, including for example, immunogenic fragments and epitopes of the biomarker polypeptide.

In one embodiment, the compound that binds the biomarker is an antibody.

The term “antibody” as used herein includes intact molecules as well as molecules comprising or consisting of fragments thereof, such as, for example Fab, F(ab′)2, Fv and scFv, as well as engineered variants including diabodies, triabodies, mini-bodies and single-domain antibodies which are capable of binding an epitopic determinant. Thus, antibodies may exist as intact immunoglobulins, or as modifications in a variety of forms.

In another embodiment, an antibody to a biomarker polypeptide is detected in a patient sample, wherein the presence and/or level of the antibody in the sample is indicative of colorectal cancer.

Preferred detection systems contemplated herein include any known assay for detecting proteins or antibodies in a biological sample isolated from a human subject, such as, for example, SDS/PAGE, isoelectric focussing, 2-dimensional gel electrophoresis comprising SDS/PAGE and isoelectric focussing, an immunoassay, flow cytometry e.g. fluorescence-activated cell sorting (FACS), a detection based system using an antibody or non-antibody compound, such as, for example, a small molecule (e.g. a chemical compound, agonist, antagonist, allosteric modulator, competitive inhibitor, or non-competitive inhibitor, of the protein). In accordance with these embodiments, the antibody or small molecule may be used in any standard solid phase or solution phase assay format amenable to the detection of proteins. Optical or fluorescent detection, such as, for example, using mass spectrometry, MALDI-TOF, biosensor technology, evanescent fiber optics, or fluorescence resonance energy transfer, is clearly encompassed by the present invention. Assay systems suitable for use in high throughput screening of mass samples, e.g. a high throughput spectroscopy resonance method (e.g. MALDI-TOF, electrospray MS or nano-electrospray MS), are also contemplated. Another suitable protein detection technique involves the use of Multiple Reaction Monitoring (MRM) in LC-MS (LC/MRM-MS) (Anderson and Hunter, 2006).

Immunoassay formats are particularly suitable, e.g., selected from the group consisting of, an immunoblot, a Western blot, a dot blot, an enzyme linked immunosorbent assay (ELISA), radioimmunoassay (RIA), enzyme immunoassay. Modified immunoassays utilizing fluorescence resonance energy transfer (FRET), isotope-coded affinity tags (ICAT), matrix-assisted laser desorption/ionization time of flight (MALDI-TOF), electrospray ionization (ESI), biosensor technology, evanescent fiber-optics technology or protein chip technology are also useful.

Nucleic Acid Detection Techniques

Any suitable technique that allows for the qualitative and/or quantitative assessment of the level of a biomarker polynucleotide in a sample may be used. The terms “nucleic acid molecule” or “polynucleotide” as used herein refer to an oligonucleotide, polynucleotide or any fragment thereof.

Comparison may be made by reference to a standard control, or to a control level that is found in healthy tissue. For example, levels of a transcribed gene can be determined by Northern blotting, and/or RT-PCR. With the advent of quantitative (real-time) PCR, quantitative analysis of gene expression can be achieved by using appropriate primers for the gene of interest. The nucleic acid may be labelled and hybridised on a gene array, in which case the gene concentration will be directly proportional to the intensity of the radioactive or fluorescent signal generated in the array.

Methods for direct sequencing of nucleotide sequences are well known to those skilled in the art and can be found for example in Ausubel et al., eds., Short Protocols in Molecular Biology, 3rd ed., Wiley, (1995) and Sambrook et al., Molecular Cloning, 3rd ed., Cold Spring Harbor Laboratory Press, (2001). Sequencing can be carried out by any suitable method, for example, dideoxy sequencing, chemical sequencing or variations thereof. Direct sequencing has the advantage of determining variation in any base pair of a particular sequence.

Other PCR methods that may be used in carrying out the invention include hybridization based PCR detection systems, TaqMan assay (U.S. Pat. No. 5,962,233) and the molecular beacon assay (U.S. Pat. No. 5,925,517).

The nucleic acid may be separated from the sample for testing. Suitable methods will be known to those of skill in the art. For example, RNA may be isolated from a sample to be analysed using conventional procedures, such as are supplied by QIAGEN technology. This RNA is then reverse-transcribed into DNA using reverse transcriptase and the DNA molecule of interest may then be amplified by PCR techniques using specific primers.

Diagnostic procedures may also be performed directly upon patient samples. Hybridisation or amplification assays, such as, for example, Southern or Northern blot analysis, immunohistochemistry, single-stranded conformational polymorphism analysis (SSCP) and PCR analyses are among techniques that are useful in this respect. If desired, target or probe nucleic acid may be immobilised to a solid support such as a microtitre plate, membrane, polystyrene bead, glass slide or other solid phase.

Kits

The present invention provides kits for the diagnosis or detection of colorectal cancer. Such kits may be suitable for detection of nucleic acid species, or alternatively may be for detection of a polypeptide gene product, as discussed above.

For detection of polypeptides, antibodies will most typically be used as components of kits. However, any agent capable of binding specifically to a biomarker gene product will be useful in this aspect of the invention. Other components of the kits will typically include labels, secondary antibodies, substrates (if the gene is an enzyme), inhibitors, co-factors and control gene product preparations to allow the user to quantitate expression levels and/or to assess whether the diagnosis experiment has worked correctly. Enzyme-linked immunosorbent assay-based (ELISA) tests and competitive ELISA tests are particularly suitable assays that can be carried out easily by the skilled person using kit components.

Optionally, the kit further comprises means for the detection of the binding of an antibody to a biomarker polypeptide. Such means include a reporter molecule such as, for example, an enzyme (such as horseradish peroxidase or alkaline phosphatase), a dye, a radionucleotide, a luminescent group, a fluorescent group, biotin or a colloidal particle, such as colloidal gold or selenium. Preferably such a reporter molecule is directly linked to the antibody.

In yet another embodiment, a kit may additionally comprise a reference sample. In one embodiment, a reference sample comprises a polypeptide that is detected by an antibody. Preferably, the polypeptide is of known concentration. Such a polypeptide is of particular use as a standard. Accordingly, various known concentrations of such a polypeptide may be detected using a diagnostic assay described herein.

For detection of nucleic acids, such kits may contain a first container such as a vial or plastic tube or a microtiter plate that contains an oligonucleotide probe. The kits may optionally contain a second container that holds primers. The probe may be hybridisable to DNA whose altered expression is associated with colorectal cancer and the primers are useful for amplifying this DNA. Kits that contain an oligonucleotide probe immobilised on a solid support could also be developed, for example, using arrays (see supplement of issue 21(1) Nature Genetics, 1999).

For PCR amplification of nucleic acid, nucleic acid primers may be included in the kit that are complementary to at least a portion of a biomarker gene as described herein. The set of primers typically includes at least two oligonucleotides, preferably four oligonucleotides, that are capable of specific amplification of DNA. Fluorescent-labelled oligonucleotides that will allow quantitative PCR determination may be included (e.g. TaqMan chemistry, Molecular Beacons). Suitable enzymes for amplification of the DNA, will also be included.

Control nucleic acid may be included for purposes of comparison or validation. Such controls could either be RNA/DNA isolated from healthy tissue, or from healthy individuals, or housekeeping genes such as β-actin or GAPDH whose mRNA levels are not affected by colorectal cancer.

Regression Algorithms and Statistics

In order to develop a panel of biomarkers suitable for diagnosing or detecting colorectal cancer, the present inventors have analysed numerous biomarkers in a statistical model. Such an improvement in the performance of a test is sometimes referred to as the “in-sample” performance. A fair evaluation of a test requires its assessment using out-of-sample subjects, that is, subjects not included in the construction of the initial predictive model. This is achieved by assessing the test performance using cross validation.

Tests for statistical significance include linear and non linear regression, including ANOVA, Kruskal-Wallis, Wilcoxon, Mann-Whitney and odds ratio, Baysian probability algorithms. As the number of biomarkers measured increases however, it can be generally more convenient to use a more sophisticated technique such as Random Forests, simple logistic, Bayes Net to name a few.

For example, Bayesian probability may be adopted. In this circumstance a 10-fold cross-validation can be used to estimate the “out-of-sample” performance of the models in question. For each combination of biomarkers under consideration, the data can be divided randomly into 10 sub-samples, each with similar proportions of healthy subject and subjects at each stage of disease. In turn, each subsample can be excluded, and a logistic model built using the remaining 90% of the subjects. This model can then be used to estimate the probability of cancer for the excluded sub-sample, providing an estimate of “out-of-sample” performance. By repeating this for the remaining 9 subsamples, “out-of-sample” performance can be estimated from the study data itself. These out-of sample predicted probabilities can then be compared with the actual disease status of the subjects to create a Receiver Operating Characteristic (ROC) Curve, from which the cross-validated sensitivity at 95% specificity may be estimated.

Each estimate of “out-of-sample” performance using cross-validation (or any other method), whilst unbiased, has an element of variability to it. Hence a ranking of models (based on biomarker combinations) can be indicative only of the relative performance of such models. However a set of biomarkers which is capable of being used in a large number of combinations to generate a diagnostic test as demonstrated via “out-of-sample” performance evaluations, almost certainly contains within itself combinations of biomarkers that will withstand repeated evaluation.

Many different combinations can qualify as diagnostic tests which prove useful and cost effective and have acceptable sensitivity for a given specificity. As an example, consider the five biomarkers: IL-8, IGFBP2, MAC2BP, M2PK and DKK-3. A model discriminating subjects with cancer from healthy controls can be as follows:

$\log \left(\frac{p}{1-p}\right)={\beta }_0+{\beta }_{\mathrm{IL}8}C_{\mathrm{IL}8}+{\beta }_{\mathrm{IGFBP}2}C_{\mathrm{IGFBP}2}+{\beta }_{\mathrm{MAC}2\mathrm{BP}}C_{\mathrm{MAC}2\mathrm{BP}}+{\beta }_{M2\mathrm{PK}}C_{M2\mathrm{PK}}+{\beta }_{\mathrm{DKK}3}C_{\mathrm{DKK}3}$

Here p represents the probability that a person has colorectal cancer. Each C_i is the logarithm of concentration biomarker i in the plasma (or serum) of a person. Each beta (β) is a coefficient applying to that biomarker in the concentration units in which it is measured—β₀ is an “offset” or “intercept”. This linear logistic model is common to all results presented herein, but is far from the only way in which a combination of biomarker concentrations may be modelled to predict the probability of cancer.

Other non linear or linear logistic algorithms that would be equally applicable include Random Forest, ANOVA, t-Test, Fisher analysis, Support Vector Machine, Linear Models for MicroArray data (LIMMA) and/or Significance Analyses of Microarray Data (SAM), Best First, Greedy Stepwise, Naive Bayes, Linear Forward Selection, Scatter Search, Linear Discriminant Analysis (LDA), Stepwise Logistic Regression, Receiver Operating Characteristic and Classification Trees (CT).

Thus, in light of the teachings of the present specification, the person skilled in the art will appreciate that the sensitivity and specificity of a test for diagnosing colorectal cancer may be modulated by selecting a different combination of the biomarkers as described herein

Knowledge-Based Systems

It will be apparent from the discussion herein that knowledge-based computer software and hardware for implementing an algorithm also form part of the present invention. Such computer software and/or hardware are useful for performing a method of diagnosing or detecting colorectal cancer according the invention. Thus, the present invention also provides software or hardware programmed to implement an algorithm that processes data obtained by performing the method of the invention via a multivariate analysis to provide a disease score and provide or permit a diagnosis or detection of colorectal cancer and/or determine progression or status of a colorectal cancer or determine whether or not a colorectal cancer has progressed or determine whether or not a subject is responding to treatment for colorectal cancer in accordance with the results of the disease score in comparison with predetermined values.

In one example, a method of the invention may be used in existing knowledge-based architecture or platforms associated with pathology services. For example, results from a method described herein are transmitted via a communications network (e.g. the internet) to a processing system in which an algorithm is stored and used to generate a predicted posterior probability value which translates to the score of disease probability or risk of recurrence or metastasis or responsiveness to treatment which is then forwarded to an end user in the form of a diagnostic or predictive report.

The method of the invention may, therefore, be in the form of a kit or computer-based system which comprises the reagents necessary to detect the concentration of the biomarkers and the computer hardware and/or software to facilitate determination and transmission of reports to a clinician.

The assay of the present invention permits integration into existing or newly developed pathology architecture or platform systems. For example, the present invention contemplates a method of allowing a user to determine the status of a subject with respect to colorectal cancer, the method including:

(a) receiving data in the form of levels at least two biomarkers selected from IL-8, IGFBP2, MAC2BP, M2PK, IL-13, DKK-3, EpCam, MIP1β, TGFβ1, and TIMP-1 in a readily obtained sample, optionally in combination with another marker of colorectal cancer;
(b) processing the subject data via multivariate analysis (for example, regression analysis) to provide a disease score;
(c) determining the status of the subject in accordance with the results of the disease score in comparison with predetermined values; and
(d) transferring an indication of the status of the subject to the user via the communications network reference to the multivariate analysis includes an algorithm which performs the multivariate analysis function.

In one embodiment, the method for diagnosing or detecting colorectal cancer of the invention may be performed by taking a blood sample from a patient and determining the presence and/or level of any one or more of the biomarkers as described herein. If desired, the measurements may be made, for example, on a biochip so that a single analysis can be used to measure the presence and/or level of multiple biomarkers. The results of this analysis may then be inputted into into a computer program that subjects them to linear regression analysis. The computer could also contain information as to control values or expected ranges, or the clinician, nurse, medical administrator or general practitioner could input such data. This analysis wold then provide a score or likelihood of having colorectal cancer. If a second test for the patient is performed, the regression analysis may indicate a change in the score, thus indicating that the patient's disease has progressed or worsened.

EXAMPLES

Materials and Methods

Patient Samples

A collection of plasma and serum samples was taken and processed from a cohort of colorectal cancer patients (Dukes Stages A-D) that were being treated at several hospitals.

Blood was also collected and processed from a group of about 50 healthy volunteers over the age of 65 and from a group of 15 over the age of 50.

Four separate studies were undertaken with slightly different biomarkers. Study 1 looked at 52 cancer samples and 50 controls, study 2 looked at 55 cancer samples and 53 controls, study 3 and 4 looked at 96 cancer samples and 50 controls. In study 2, 3 and 4 the patients were age and gender matched across Dukes Stages, see Table 2 for summary statistics.

TABLE 2
Characteristics of normal volunteers and colorectal cancer
patients used in studies 2, 3 and 4.
	Controls	Cancers
	n = 50	n = 96
Gender
male	25	48
female	25	48
Mean Age (yr)	68	68
Dukes stage
A		22
B		30
C		30
D		14
Tumour site
colon		73
rectum		17
unknown		6
Proximal (includes caecum, ascending, hepatic		43
flexure and transverse colon)
Distal (includes splenic flexure, descending,		47
sigmoid and rectum)

Biomarker Analysis

Analysis of biomarkers was done with commercial kits and sourced antibodies (DSL, R&D Duoset, Calbiotech, Millipore, Abnova, Genway, Peviva, Schebo, Bender) and using ELISA or Luminex assays.

Statistical Evaluation and Panel Biomarker Modelling

Results for each assay were analysed using the statistical software packages Prism and “R”. Individual performance of markers was evaluated using the non-parametric Mann-Whitney t-test and individual receiver operator characteristic (ROC) curves were generated.

Logistic regression and related modelling strategies were used to find combinations of biomarkers that best separated controls and colorectal cancer patients. Four separate studies were performed with the same samples/aliquots. The results of each of these is given below.

Results of Study 1, 2 and 3

Biomarkers chosen to be measured in Study 1 and 2 and 3 are listed in Table 3. Biomarkers in bold were those identified as promising from each study (i.e. they were significantly different in samples from colorectal cancer patients versus control and/or they were identified in panels of combined biomarkers that distinguish colorectal cancer from controls).

TABLE 3
Biomarkers analysed in studies.
	Study 1	Study 2	Study 3
	IGF-I	IGF-BP2	IGF-BP2 (DSL)
	IGF-II	IGF-II	IGF-II
	IGF-BP2	IGF-BP3	IFNg
	IGF-BP3	Her2	TNFa
	BTC	VegFA	IL-10
	Amphiregulin	VegFC	IL-6
	VegFA	VegFD	GM-CSF
	VegFC	TIMP-1	IL-12
	VegFD	TIMP-2	IL-13
	MMP-2	MMP-1	IL-8
	MMP-7	MMP-2	IL-4
	MMP-9	MMP-3	Il-2
	TIMP-2	MMP-7	IL-1b
	Her2	MMP-8	MMP-1
		MMP-12	MMP-2
		MMP-13	MMP-3
		ENA-78	MMP-7
		MCP-1	MMP-8
		MIP-1beta	MMP-9
		GM-CSF	ENA-78
		IFN-gamma	MIP-1alpha
		IL-10	MIP-1beta
		IL-12	MCP-1
		IL-13	Mac-2BP
		IL-1beta	TIMP-1
		IL-2	TIMP-2
		IL-4	Gro-alpha
		IL-6	Tumour M2 pyruvate
			kinase
		IL-8	M30-apoptosense
		TNF-alpha	M65
		Cripto	Trail-R2
			P-cadherin
			OPN
			Dkk-3
			EpCam
			TGFbeta1
			REG IV
			CEA
			DcR3
			CA19.9
			Amphiregulin
			CEACAM6
			VegFA pan
			VegFA165b
			Spondin-2
			survivin

Statistical Evaluation and Panel Biomarker Modelling

To find combinations of biomarkers that best separated controls and colorectal cancer patients, forward variable selection with Bayesian Information Criteria to penalize log-likelihood to prevent over-fitting was adopted. To estimate the likely performance of the panel of biomarkers on an independent dataset, “N-Fold” or “leave-one-out” cross validation was used. In this procedure one observation at a time was excluded whilst the entire model fitting algorithm was applied to the remaining observations.

The resulting model was then used to estimate the probability that the excluded observation is a case. This was repeated for each observation in the dataset. In this way each observation in turn acted as an independent test of the model-building algorithm. The resulting dataset consisting of cases and controls each with an “independently predicted” case probability can then be compared with the original model. The ability to choose from numerous biomarkers and weight them appropriately allows a search strategy which optimises performance in regions of interest on the ROC curve. The cost of poor specificity is large numbers of unnecessary colonoscopies.

In study 3, 48 potential biomarkers were evaluated to select a candidate panel of colorectal cancer biomarkers, using block randomization within plates to avoid bias. From this list of 48 only 42 showed measurable levels. Individually 14 biomarkers showed significant difference between controls and CRC as assessed by t-tests; (IGFII, IGFBP2, IL-8, IL-6, MMP-1, MMP-7, s90/Mac2BP, M2PK, EpCam, TIMP-1 (serum and plasma), M65, OPN, TGFβ1, VEGFpan. As expected, none had sufficient sensitivity or specificity to be useful as a biomarker by itself (not shown). However, using a variety of modelling strategies, including use of logarithmic values, several different panels of biomarkers were found that exceeded the performance of FOBT especially for early to late stage disease.

FIG. 1 shows the results from a 7 biomarker panel which included IL8 (serum), IL-13 (serum), EpCAM (plasma), M2PK (plasma), IGFBP2 (serum) and Mac2BP (serum) and which was cross validated to predict its performance on independent samples.

This 7 biomarker model, which is described at least conceptually as

$\log \left(\frac{p}{1-p}\right)={\beta }_0+{\beta }_{\mathrm{IL}8}C_{\mathrm{IL}8}+{\beta }_{\mathrm{IGFBP}2}C_{\mathrm{IGFBP}2}+{\beta }_{\mathrm{MAC}2\mathrm{BP}}C_{\mathrm{MAC}2\mathrm{BP}}+{\beta }_{M2\mathrm{PK}}C_{M2\mathrm{PK}}+{\beta }_{\mathrm{DKK}3}C_{\mathrm{DKK}3}$

provided good performance at high specificity and was robust under cross validation. The coefficients estimated to give the best model for this biomarker combination in plasma are listed in Table 4. Performance statistics are provided in Table 5. This performance exceeds that quoted for FOBT (sensitivity 65.8%, specificity 95%) (Morikawa et al., 2005).

TABLE 4
Coefficients for the biomarker combination.
		Measured	Concentration
	Biomarker	in	Units	Coefficient
	Intercept	NA	NA	−37.74
	IL-8	serum	pg/mL	1.07
	IL-13	serum	pg/mL	−0.28
	EpCAM	plasma	pg/mL	−0.33
	M2PK	plasma	units/mL	1.40
	IGFBP2	serum	ng/mL	1.99
	Mac2BP	serum	ng/mL	2.39
	MIP1beta	Serum	Pg/ml	−1.19

TABLE 5
Performance of the 7 biomarker model and cross-validation.
	Model
	estimate	Cross validation
	Area Under the ROC Curve (AUC)	0.91	0.86
	Sensitivity at 80% specificity	0.84	0.78
	Sensitivity at 90% specificity	0.81	0.69

This model was also applied separately to patients from each stage of colorectal cancer (Dukes Stage A, B, C, D) and shown to perform equally well within each stage (FIG. 2). The AUCs were 0.88-0.93 and were almost equally good at discriminating all Stages of colorectal cancer. The model shows the highest sensitivity of 90% at 90% specificity for Stage C and the lowest sensitivity of 73% at 90% specificity for Stage B (Table 6).

TABLE 6
Performance of the Model by Dukes Stage
	Stage A	Stage B	Stage C	Stage D
Area Under the ROC Curve	0.89	0.88	0.93	0.91
(AUC)
Sensitivity at 80% specificity	0.82	0.77	0.90	0.93
Sensitivity at 90% specificity	0.77	0.73	0.90	0.86

Study 4 (Also Referred to as “Study 3 Remeasured”)

In study 4, 10 biomarkers were remeasured in the same cohort as Study 3. Blood was collected from 96 colorectal cancer patients and 50 normal subjects (the controls). In this study the focus was on 10 biomarkers, namely IGFBP2, IL8, IL13, Mac2BP, M2PK, Dkk3, EpCam, TGFbeta1, TIMP-1, MIP1beta. Assays were performed as described previously. Both serum and plasma levels of each of the biomarkers were measured and compared with control values.

When modelled in pairs (two markers), a total of 5 pairs (out of a possible 45 combinations selected from the list of 10 biomarkers above) could be shown to produce a sensitivity above 52% at a specificity of 95%. See Table 7 and FIG. 3.

TABLE 7
Biomarker Pairs Producing Useful Screening Tests
on Cross-Validation.
		Estimated	Estimated
		In-Sample	Out-of-Sample
		(Test)	(Cross-Validated)
		Sensitivity at 95%	Sensitivity at 95%
Biomarker 1	Biomarker 2	Specificity	Specificity
M2PK	EpCAM	58.3%	58.3%
M2PK	IL13	56.3%	57.3%
Dkk3	M2PK	55.2%	55.2%
M2PK	IL8	60.4%	54.2%
M2PK	IGFBP2	58.3%	52.1%

In analysing combinations of three to ten of the nominated biomarkers, there are 968 possible combinations. The 968 combinations of between 3 and 10 biomarkers consist of the 120 combinations of 3 marker; 210 combinations of 4 markers; 252 combinations of 5 markers; 210 combinations of 6 markers; 120 combinations of 7 markers; 45 combinations of 8 markers; 10 combinations of 9 markers and the single combination that includes all 10 biomarkers. When they were modelled using a linear logistic model, and then tested via 10-fold cross validation, about half of the 968 combinations had a sensitivity of 50% at a specificity of 95%, see FIG. 4 which shows the results for a three biomarker combination. More than half of these combinations would have a specificity of 90% and a sensitivity of 50%.

FIG. 5 shows all 485 of the estimated out-of sample (10-fold cross-validated) ROC curves for tests out of a total possible 968 models based on all possible combinations of 3 to 10 of the biomarkers. Note that many individual segments of the 485 ROC curves are coincident, due to as each horizontal segment represents one control and each vertical segment one case. In this instance 50.1% of the combinations have exceeded the 50% sensitivity, 95% specificity, The best estimated “out-of-sample” performance is a sensitivity of 76% at 95% specificity. Repeating the cross-validation will select a different set of models—the sensitivity of any one combination may vary by 10% at 95% specificity due to random sampling—but result in a similar proportion of useful “useful screening tests”. Precise validation of individual models requires repeated experiments and larger sample sizes.

FIG. 6 shows how many of the 485 combinations with 50% sensitivity, 95% specificity, include any given biomarker. At the high end, 432 of the chosen “useful” combinations include M2PK. At the low end 227 of the chosen “useful” combinations include MIP1beta. This high representation of all 10 biomarkers in “useful” models shows the unity and self-complementarity of the selection of these 10 biomarkers.

FIG. 7 to FIG. 11 demonstrate some of the results from this last study (Study 4) for combinations of 5 and 7 biomarkers, including a model where the samples are either from plasma or serum cluster. FIG. 11 demonstrates the validity of the choice of three biomarkers (DKK-3, M2PK and IGFBP2) at different stages of the disease progression. The data indicates that at Stage A if the three markers are used, the test still will achieve a significant sensitivity (64%) at 95% specificity which is comparable to the sensitivity achieved at late stage disease, 79%). That is the biomarker panel of three will pick up early disease states allowing early detection.

Tables 8 to 16 list results from various combinations of various biomarker panel sets. Depending on the linear regression that is used, as well as the cohort control and other factors such as sample derivation and assay kit technique, there may be a variation on the actual figures or order of the markers. Regardless, many of these combinations will achieve good selectivity at high specificities so as to be useful for diagnosing or detecting colorectal cancer at any stage of the disease progression.

10

TABLE 8
Combination of three biomarkers in serum that equal or exceed
50% sensitivity at 95% specificity.
					Cross
				Test	Validated
				Sensitivity	Sensitivity
				at 95%	at 95%
	BM1	BM2	BM3	Specificity	Specificity
	Dkk3	M2PK	IGFBP2	72.9%	70.8%
	Dkk3	M2PK	IL8	62.5%	61.5%
	M2PK	IL13	IGFBP2	65.6%	61.5%
	M2PK	IGFBP2	EpCAM	63.5%	61.5%
	M2PK	IL8	IGFBP2	65.6%	60.4%
	M2PK	IL8	IL13	61.5%	58.3%
	M2PK	MIP1beta	IL13	55.2%	57.3%
	M2PK	IL8	Mac2BP	59.4%	57.3%
	M2PK	MIP1beta	TGFbeta1	57.3%	56.3%
	M2PK	IL8	EpCAM	64.6%	56.3%
	Dkk3	M2PK	IL13	59.4%	55.2%
	Dkk3	M2PK	EpCAM	56.3%	55.2%
	M2PK	MIP1beta	EpCAM	59.4%	55.2%
	M2PK	IL8	TGFbeta1	58.3%	55.2%
	M2PK	IL8	TIMP1	57.3%	55.2%
	TGFbeta1	Mac2BP	TIMP1	54.2%	55.2%
	M2PK	Mac2BP	IGFBP2	58.3%	54.2%
	Dkk3	IL8	Mac2BP	55.2%	53.1%
	M2PK	MIP1beta	Mac2BP	52.1%	53.1%
	M2PK	MIP1beta	IGFBP2	57.3%	53.1%
	M2PK	TIMP1	EpCAM	58.3%	53.1%
	M2PK	MIP1beta	IL8	56.3%	52.1%
	M2PK	IL13	TIMP1	58.3%	52.1%
	Dkk3	M2PK	Mac2BP	57.3%	51.0%
	Dkk3	M2PK	TIMP1	55.2%	51.0%
	M2PK	IL13	Mac2BP	58.3%	51.0%
	M2PK	TGFbeta1	IGFBP2	52.1%	51.0%
	M2PK	TGFbeta1	TIMP1	57.3%	51.0%
	M2PK	TGFbeta1	Mac2BP	56.3%	50.0%
	IL8	IL13	Mac2BP	61.5%	50.0%
	IL8	TGFbeta1	Mac2BP	53.1%	50.0%
	M2PK	MIP1beta	TIMP1	49.0%	49.0%
	M2PK	TGFbeta1	EpCAM	49.0%	47.9%
	IL8	IL13	IGFBP2	49.0%	47.9%
	IL8	Mac2BP	IGFBP2	57.3%	47.9%
	Dkk3	M2PK	MIP1beta	55.2%	46.9%
	TGFbeta1	Mac2BP	IGFBP2	46.9%	46.9%
	Dkk3	Mac2BP	IGFBP2	49.0%	45.8%
	M2PK	IL13	EpCAM	50.0%	45.8%
	IL8	Mac2BP	TIMP1	52.1%	45.8%
	IL13	Mac2BP	IGFBP2	45.8%	44.8%
	Dkk3	IL8	TGFbeta1	41.7%	43.8%
	MIP1beta	IL8	Mac2BP	42.7%	43.8%
	MIP1beta	IL8	EpCAM	44.8%	43.8%
	IL8	TGFbeta1	EpCAM	46.9%	43.8%
	Dkk3	IL8	EpCAM	51.0%	42.7%
	IL8	IGFBP2	EpCAM	43.8%	42.7%
	M2PK	Mac2BP	TIMP1	51.0%	41.7%

TABLE 9
Combination of four biomarkers including DKK-3 in serum that
equal or exceed 50% sensitivity at 95% specificity.
					Cross
				Test	Validated
				Sensitivity	Sensitivity
				at 95%	at 95%
BM1	BM2	BM3	BM4	Specificity	Specificity
Dkk3	M2PK	Mac2BP	IGFBP2	68.8%	69.8%
Dkk3	M2PK	IL8	IL13	71.9%	68.8%
Dkk3	M2PK	IL8	EpCAM	70.8%	67.7%
Dkk3	M2PK	TGFbeta1	Mac2BP	67.7%	65.6%
Dkk3	M2PK	IL8	IGFBP2	69.8%	64.6%
Dkk3	M2PK	IL8	Mac2BP	69.8%	63.5%
Dkk3	M2PK	MIP1beta	TGFbeta1	65.6%	61.5%
Dkk3	M2PK	IL8	TIMP1	68.8%	61.5%
Dkk3	M2PK	IL13	IGFBP2	63.5%	61.5%
Dkk3	M2PK	MIP1beta	IL8	59.4%	60.4%
Dkk3	M2PK	IGFBP2	EpCAM	68.8%	60.4%
Dkk3	M2PK	MIP1beta	IGFBP2	69.8%	59.4%
Dkk3	M2PK	TGFbeta1	IGFBP2	61.5%	59.4%
Dkk3	M2PK	IL13	Mac2BP	56.3%	58.3%
Dkk3	M2PK	TGFbeta1	TIMP1	65.6%	58.3%
Dkk3	M2PK	MIP1beta	IL13	57.3%	57.3%
Dkk3	IL8	Mac2BP	IGFBP2	62.5%	56.3%
Dkk3	M2PK	IL8	TGFbeta1	65.6%	55.2%
Dkk3	M2PK	IL13	TIMP1	57.3%	55.2%
Dkk3	M2PK	Mac2BP	TIMP1	57.3%	55.2%
Dkk3	M2PK	MIP1beta	Mac2BP	57.3%	54.2%
Dkk3	IL8	IL13	Mac2BP	61.5%	54.2%
Dkk3	M2PK	TGFbeta1	EpCAM	61.5%	53.1%
Dkk3	M2PK	IGFBP2	TIMP1	61.5%	53.1%
Dkk3	M2PK	IL13	TGFbeta1	63.5%	52.1%
Dkk3	M2PK	TIMP1	EpCAM	56.3%	52.1%
Dkk3	IL8	Mac2BP	TIMP1	57.3%	52.1%
Dkk3	M2PK	MIP1beta	EpCAM	60.4%	51.0%
Dkk3	TGFbeta1	Mac2BP	TIMP1	59.4%	51.0%

TABLE 10
Combination of four biomarkers including M2PK in serum that
equal or exceed 50% sensitivity at 95% specificity.
					Cross
				Test	Validated
				Sensitivity	Sensitivity
				at 95%	at 95%
BM1	BM2	BM3	BM4	Specificity	Specificity
M2PK	IL8	Mac2BP	TIMP1	63.5%	65.6%
M2PK	Mac2BP	IGFBP2	EpCAM	70.8%	65.6%
M2PK	IL8	IL13	Mac2BP	66.7%	64.6%
M2PK	IL8	TGFbeta1	Mac2BP	65.6%	64.6%
M2PK	MIP1beta	IL13	IGFBP2	64.6%	62.5%
M2PK	IL8	IL13	TIMP1	64.6%	62.5%
M2PK	IL8	IL13	EpCAM	65.6%	62.5%
M2PK	IL13	Mac2BP	IGFBP2	69.8%	62.5%
M2PK	MIP1beta	IL8	IL13	63.5%	61.5%
M2PK	IL8	Mac2BP	EpCAM	65.6%	61.5%
M2PK	IL13	IGFBP2	EpCAM	69.8%	61.5%
M2PK	MIP1beta	IL8	TGFbeta1	58.3%	58.3%
M2PK	IL8	IL13	IGFBP2	67.7%	58.3%
M2PK	IL8	Mac2BP	IGFBP2	61.5%	58.3%
M2PK	IL8	IGFBP2	EpCAM	64.6%	58.3%
M2PK	IL13	TGFbeta1	IGFBP2	64.6%	58.3%
M2PK	IL13	TGFbeta1	EpCAM	62.5%	58.3%
M2PK	IL13	IGFBP2	TIMP1	62.5%	58.3%
M2PK	TGFbeta1	Mac2BP	TIMP1	62.5%	58.3%
M2PK	TGFbeta1	IGFBP2	EpCAM	61.5%	58.3%
M2PK	MIP1beta	IL13	TIMP1	57.3%	57.3%
M2PK	MIP1beta	TGFbeta1	TIMP1	58.3%	57.3%
M2PK	MIP1beta	IGFBP2	EpCAM	65.6%	57.3%
M2PK	MIP1beta	IL8	Mac2BP	54.2%	56.3%
M2PK	IL8	IL13	TGFbeta1	64.6%	56.3%
M2PK	IL8	TIMP1	EpCAM	62.5%	56.3%
M2PK	IL13	Mac2BP	TIMP1	57.3%	56.3%
M2PK	TGFbeta1	Mac2BP	IGFBP2	60.4%	56.3%
M2PK	IGFBP2	TIMP1	EpCAM	63.5%	56.3%
M2PK	MIP1beta	IL8	TIMP1	57.3%	55.2%
M2PK	IL8	TGFbeta1	TIMP1	57.3%	55.2%
M2PK	MIP1beta	IL13	Mac2BP	57.3%	54.2%
M2PK	MIP1beta	IL8	EpCAM	62.5%	53.1%
M2PK	MIP1beta	TIMP1	EpCAM	59.4%	52.1%
M2PK	IL13	TGFbeta1	TIMP1	64.6%	52.1%
M2PK	IL13	Mac2BP	EpCAM	51.0%	52.1%
M2PK	MIP1beta	IL13	TGFbeta1	57.3%	51.0%
M2PK	MIP1beta	Mac2BP	IGFBP2	57.3%	51.0%
M2PK	TGFbeta1	Mac2BP	EpCAM	52.1%	51.0%
M2PK	TGFbeta1	TIMP1	EpCAM	59.4%	51.0%
M2PK	MIP1beta	IL8	IGFBP2	52.1%	50.0%
M2PK	Mac2BP	TIMP1	EpCAM	52.1%	50.0%
M2PK	MIP1beta	TGFbeta1	Mac2BP	63.5%	49.0%
M2PK	MIP1beta	IL13	EpCAM	50.0%	47.9%
M2PK	MIP1beta	TGFbeta1	EpCAM	53.1%	47.9%
M2PK	IL13	TGFbeta1	Mac2BP	60.4%	46.9%
M2PK	MIP1beta	TGFbeta1	IGFBP2	49.0%	44.8%

TABLE 11
Combination of five biomarkers in serum that equal
or exceed 50% sensitivity at 95% specificity.
					Test	Cross Validated
					Sensitivity at	Sensitivity at
BM1	BM2	BM3	BM4	BM5	95% Specificity	95% Specificity
Dkk3	M2PK	IL8	IL13	Mac2BP	74.0%	70.8%
Dkk3	M2PK	IL8	IL13	TIMP1	71.9%	70.8%
M2PK	TGFbeta1	Mac2BP	IGFBP2	EpCAM	69.8%	70.8%
Dkk3	M2PK	MIP1beta	IL8	IL13	71.9%	69.8%
Dkk3	M2PK	MIP1beta	IGFBP2	EpCAM	71.9%	69.8%
Dkk3	M2PK	IL8	IGFBP2	EpCAM	78.1%	69.8%
Dkk3	M2PK	TGFbeta1	Mac2BP	IGFBP2	68.8%	69.8%
M2PK	MIP1beta	Mac2BP	IGFBP2	EpCAM	69.8%	68.8%
Dkk3	M2PK	IL8	TGFbeta1	Mac2BP	70.8%	67.7%
Dkk3	M2PK	Mac2BP	IGFBP2	EpCAM	70.8%	67.7%
M2PK	MIP1beta	IL8	IL13	Mac2BP	66.7%	67.7%
Dkk3	M2PK	IL8	IL13	IGFBP2	70.8%	66.7%
Dkk3	M2PK	IL13	TGFbeta1	IGFBP2	68.8%	66.7%
Dkk3	M2PK	IL13	IGFBP2	EpCAM	66.7%	66.7%
M2PK	IL13	Mac2BP	IGFBP2	EpCAM	71.9%	66.7%
Dkk3	M2PK	IL8	IL13	TGFbeta1	69.8%	65.6%
Dkk3	M2PK	IL8	Mac2BP	TIMP1	67.7%	65.6%
Dkk3	M2PK	TGFbeta1	Mac2BP	EpCAM	66.7%	65.6%
M2PK	IL8	IL13	TGFbeta1	Mac2BP	71.9%	65.6%
M2PK	IL8	IL13	TGFbeta1	IGFBP2	66.7%	65.6%
M2PK	IL8	TGFbeta1	Mac2BP	IGFBP2	64.6%	65.6%
M2PK	IL8	TGFbeta1	Mac2BP	EpCAM	69.8%	65.6%
Dkk3	M2PK	MIP1beta	IL8	IGFBP2	66.7%	64.6%
Dkk3	M2PK	MIP1beta	Mac2BP	IGFBP2	70.8%	64.6%
M2PK	MIP1beta	IL8	IL13	TIMP1	65.6%	64.6%
M2PK	MIP1beta	IL8	TGFbeta1	Mac2BP	65.6%	64.6%
M2PK	IL8	IL13	Mac2BP	EpCAM	76.0%	64.6%
M2PK	IL8	IL13	IGFBP2	EpCAM	72.9%	64.6%
M2PK	IL8	TGFbeta1	Mac2BP	TIMP1	64.6%	64.6%
M2PK	IL8	Mac2BP	IGFBP2	EpCAM	71.9%	64.6%
Dkk3	M2PK	IL13	TGFbeta1	TIMP1	68.8%	63.5%
Dkk3	M2PK	IGFBP2	TIMP1	EpCAM	67.7%	63.5%
M2PK	MIP1beta	IL13	IGFBP2	EpCAM	67.7%	63.5%
M2PK	IL13	TGFbeta1	Mac2BP	IGFBP2	66.7%	63.5%
M2PK	IL13	TGFbeta1	IGFBP2	EpCAM	70.8%	63.5%
Dkk3	M2PK	MIP1beta	IL13	TGFbeta1	69.8%	62.5%
Dkk3	M2PK	IL8	IL13	EpCAM	65.6%	62.5%
Dkk3	M2PK	IL8	TGFbeta1	IGFBP2	70.8%	62.5%
Dkk3	M2PK	IL8	Mac2BP	IGFBP2	69.8%	62.5%
Dkk3	M2PK	IL13	TGFbeta1	Mac2BP	68.8%	62.5%
Dkk3	M2PK	TGFbeta1	Mac2BP	TIMP1	68.8%	62.5%
Dkk3	M2PK	TGFbeta1	TIMP1	EpCAM	67.7%	62.5%
M2PK	IL8	IL13	Mac2BP	TIMP1	64.6%	62.5%
M2PK	IL8	IL13	IGFBP2	TIMP1	66.7%	62.5%
M2PK	IL13	IGFBP2	TIMP1	EpCAM	68.8%	62.5%
Dkk3	M2PK	MIP1beta	IL13	IGFBP2	63.5%	61.5%
Dkk3	M2PK	IL8	TGFbeta1	TIMP1	64.6%	61.5%

TABLE 12
Combination of seven biomarkers in serum that equal or exceed 50% sensitivity at 95% specificity.
							Test	Cross Validated
							Sensitivity at	Sensitivity at
BM1	BM2	BM3	BM4	BM5	BM6	BM7	95% Specificity	95% Specificity
Dkk3	M2PK	IL8	Mac2BP	IGFBP2	TIMP1	EpCAM	78%	56%
Dkk3	M2PK	MIP1beta	IL8	Mac2BP	IGFBP2	EpCAM	76%	64%
Dkk3	M2PK	IL8	IL13	Mac2BP	IGFBP2	TIMP1	73%	69%
Dkk3	M2PK	IL8	TGFbeta1	Mac2BP	IGFBP2	TIMP1	72%	68%
Dkk3	M2PK	IL8	TGFbeta1	Mac2BP	IGFBP2	EpCAM	71%	62%
Dkk3	M2PK	MIP1beta	IL8	Mac2BP	IGFBP2	TIMP1	70%	64%
Dkk3	M2PK	IL8	IL13	Mac2BP	IGFBP2	EpCAM	70%	67%
Dkk3	M2PK	IL8	IL13	TGFbeta1	Mac2BP	IGFBP2	69%	66%
M2PK	MIP1beta	IL8	Mac2BP	IGFBP2	TIMP1	EpCAM	69%	55%
Dkk3	M2PK	MIP1beta	IL8	IL13	Mac2BP	IGFBP2	67%	68%
Dkk3	M2PK	MIP1beta	IL8	TGFbeta1	Mac2BP	IGFBP2	67%	64%
Dkk3	M2PK	MIP1beta	TGFbeta1	Mac2BP	IGFBP2	TIMP1	65%	59%
Dkk3	M2PK	IL8	IL13	TGFbeta1	IGFBP2	TIMP1	65%	52%
Dkk3	M2PK	IL8	TGFbeta1	IGFBP2	TIMP1	EpCAM	64%	56%
Dkk3	M2PK	IL13	TGFbeta1	Mac2BP	IGFBP2	TIMP1	64%	57%
M2PK	MIP1beta	IL8	IL13	IGFBP2	TIMP1	EpCAM	64%	50%
Dkk3	M2PK	MIP1beta	IL8	IL13	Mac2BP	TIMP1	63%	55%
Dkk3	M2PK	MIP1beta	IL8	TGFbeta1	IGFBP2	EpCAM	63%	58%
Dkk3	M2PK	MIP1beta	IL13	TGFbeta1	Mac2BP	IGFBP2	63%	53%
Dkk3	M2PK	IL8	IL13	TGFbeta1	Mac2BP	TIMP1	63%	59%
M2PK	MIP1beta	IL8	TGFbeta1	Mac2BP	IGFBP2	EpCAM	63%	59%
Dkk3	M2PK	MIP1beta	IL8	TGFbeta1	Mac2BP	TIMP1	62%	51%
Dkk3	M2PK	MIP1beta	IL8	IGFBP2	TIMP1	EpCAM	62%	56%
Dkk3	M2PK	MIP1beta	IL13	Mac2BP	IGFBP2	TIMP1	62%	39%
Dkk3	M2PK	IL8	IL13	TGFbeta1	TIMP1	EpCAM	62%	43%
Dkk3	M2PK	IL8	TGFbeta1	Mac2BP	TIMP1	EpCAM	62%	50%
Dkk3	MIP1beta	IL8	IL13	Mac2BP	IGFBP2	TIMP1	62%	55%
Dkk3	MIP1beta	IL8	Mac2BP	IGFBP2	TIMP1	EpCAM	62%	47%
M2PK	MIP1beta	IL8	IL13	Mac2BP	IGFBP2	TIMP1	62%	56%
M2PK	IL8	TGFbeta1	Mac2BP	IGFBP2	TIMP1	EpCAM	62%	49%
Dkk3	M2PK	IL8	IL13	TGFbeta1	IGFBP2	EpCAM	60%	56%
Dkk3	M2PK	IL13	TGFbeta1	Mac2BP	IGFBP2	EpCAM	60%	55%
M2PK	MIP1beta	IL8	IL13	Mac2BP	IGFBP2	EpCAM	60%	53%

TABLE 13
Seven biomarker combinations with Sensitivity between 60% and 52%.
							Test	Cross Validated
							Sensitivity at	Sensitivity at
BM1	BM2	BM3	BM4	BM5	BM6	BM7	95% Specificity	95% Specificity
Dkk3	M2PK	MIP1beta	IL8	IL13	TGFbeta1	EpCAM	59%	54%
Dkk3	M2PK	MIP1beta	IL8	IL13	IGFBP2	EpCAM	59%	52%
Dkk3	M2PK	MIP1beta	IL8	TGFbeta1	IGFBP2	TIMP1	59%	52%
Dkk3	M2PK	TGFbeta1	Mac2BP	IGFBP2	TIMP1	EpCAM	59%	47%
Dkk3	MIP1beta	IL8	IL13	TGFbeta1	Mac2BP	IGFBP2	59%	52%
Dkk3	MIP1beta	IL8	TGFbeta1	Mac2BP	IGFBP2	TIMP1	59%	45%
Dkk3	M2PK	MIP1beta	IL8	IL13	TGFbeta1	Mac2BP	58%	41%
Dkk3	M2PK	MIP1beta	IL8	TGFbeta1	TIMP1	EpCAM	58%	50%
Dkk3	M2PK	IL8	IL13	IGFBP2	TIMP1	EpCAM	58%	55%
Dkk3	MIP1beta	IL8	IL13	Mac2BP	IGFBP2	EpCAM	58%	50%
Dkk3	IL8	IL13	TGFbeta1	Mac2BP	IGFBP2	TIMP1	58%	52%
M2PK	IL8	IL13	Mac2BP	IGFBP2	TIMP1	EpCAM	58%	52%
Dkk3	M2PK	MIP1beta	IL8	IL13	TGFbeta1	IGFBP2	57%	49%
Dkk3	M2PK	MIP1beta	IL8	IL13	TIMP1	EpCAM	57%	48%
Dkk3	M2PK	MIP1beta	IL13	Mac2BP	IGFBP2	EpCAM	57%	51%
Dkk3	M2PK	MIP1beta	TGFbeta1	Mac2BP	IGFBP2	EpCAM	57%	46%
Dkk3	M2PK	MIP1beta	TGFbeta1	IGFBP2	TIMP1	EpCAM	57%	49%
Dkk3	M2PK	IL13	Mac2BP	IGFBP2	TIMP1	EpCAM	57%	49%
Dkk3	MIP1beta	IL8	TGFbeta1	Mac2BP	IGFBP2	EpCAM	57%	50%
M2PK	MIP1beta	IL8	IL13	TGFbeta1	TIMP1	EpCAM	57%	43%
M2PK	IL8	IL13	TGFbeta1	IGFBP2	TIMP1	EpCAM	57%	52%
Dkk3	M2PK	MIP1beta	Mac2BP	IGFBP2	TIMP1	EpCAM	56%	47%
Dkk3	IL8	IL13	Mac2BP	IGFBP2	TIMP1	EpCAM	56%	48%
MIP1beta	IL8	IL13	TGFbeta1	Mac2BP	IGFBP2	TIMP1	56%	54%
Dkk3	M2PK	MIP1beta	IL8	TGFbeta1	Mac2BP	EpCAM	55%	35%
Dkk3	M2PK	MIP1beta	IL13	IGFBP2	TIMP1	EpCAM	55%	44%
Dkk3	M2PK	IL8	IL13	TGFbeta1	Mac2BP	EpCAM	55%	35%
Dkk3	IL8	IL13	TGFbeta1	Mac2BP	IGFBP2	EpCAM	55%	50%
Dkk3	IL8	IL13	TGFbeta1	IGFBP2	TIMP1	EpCAM	55%	54%
Dkk3	M2PK	MIP1beta	IL8	IL13	Mac2BP	EpCAM	54%	49%
Dkk3	M2PK	MIP1beta	IL8	Mac2BP	TIMP1	EpCAM	54%	39%
Dkk3	M2PK	IL13	TGFbeta1	IGFBP2	TIMP1	EpCAM	54%	42%
Dkk3	IL8	TGFbeta1	Mac2BP	IGFBP2	TIMP1	EpCAM	54%	51%
M2PK	MIP1beta	IL8	IL13	TGFbeta1	IGFBP2	EpCAM	54%	37%
M2PK	MIP1beta	IL13	Mac2BP	IGFBP2	TIMP1	EpCAM	54%	41%
M2PK	IL8	IL13	TGFbeta1	Mac2BP	IGFBP2	TIMP1	54%	43%
MIP1beta	IL8	IL13	TGFbeta1	Mac2BP	IGFBP2	EpCAM	54%	42%
Dkk3	M2PK	MIP1beta	IL8	IL13	IGFBP2	TIMP1	53%	52%
Dkk3	M2PK	IL8	IL13	Mac2BP	TIMP1	EpCAM	53%	42%
M2PK	MIP1beta	IL8	TGFbeta1	IGFBP2	TIMP1	EpCAM	53%	34%
M2PK	MIP1beta	IL13	TGFbeta1	Mac2BP	IGFBP2	EpCAM	53%	45%
	IL8	IL13	TGFbeta1	Mac2BP	IGFBP2	EpCAM	53%	51%
MIP1beta	IL8	IL13	Mac2BP	IGFBP2	TIMP1	EpCAM	53%	39%
MIP1beta	IL8	TGFbeta1	Mac2BP	IGFBP2	TIMP1	EpCAM	53%	51%

TABLE 14
Seven biomarker combinations with sensitivity <53%.
							Test	Cross Validated
							Sensitivity at	Sensitivity at
BM1	BM2	BM3	BM4	BM5	BM6	BM7	95% Specificity	95% Specificity
M2PK	MIP1beta	IL8	IL13	TGFbeta1	Mac2BP	EpCAM	52%	42%
Dkk3	M2PK	MIP1beta	IL13	TGFbeta1	IGFBP2	EpCAM	51%	41%
Dkk3	MIP1beta	IL8	IL13	TGFbeta1	TIMP1	EpCAM	51%	35%
M2PK	MIP1beta	IL8	IL13	TGFbeta1	IGFBP2	TIMP1	51%	38%
M2PK	MIP1beta	IL8	TGFbeta1	Mac2BP	IGFBP2	TIMP1	51%	30%
M2PK	IL13	TGFbeta1	Mac2BP	IGFBP2	TIMP1	EpCAM	51%	37%
Dkk3	M2PK	MIP1beta	IL8	IL13	TGFbeta1	TIMP1	50%	48%
Dkk3	MIP1beta	IL8	IL13	TGFbeta1	Mac2BP	TIMP1	50%	41%
Dkk3	MIP1beta	IL8	IL13	TGFbeta1	IGFBP2	TIMP1	50%	42%
M2PK	MIP1beta	IL8	IL13	TGFbeta1	Mac2BP	IGFBP2	50%	49%
M2PK	MIP1beta	IL8	IL13	TGFbeta1	Mac2BP	TIMP1	50%	35%
MIP1beta	IL8	IL13	TGFbeta1	IGFBP2	TIMP1	EpCAM	50%	39%
Dkk3	M2PK	MIP1beta	IL13	TGFbeta1	IGFBP2	TIMP1	49%	44%
Dkk3	MIP1beta	IL8	IL13	TGFbeta1	Mac2BP	EpCAM	49%	33%
M2PK	IL8	IL13	TGFbeta1	Mac2BP	TIMP1	EpCAM	49%	43%
Dkk3	M2PK	MIP1beta	IL13	Mac2BP	TIMP1	EpCAM	48%	43%
Dkk3	MIP1beta	IL8	IL13	Mac2BP	TIMP1	EpCAM	48%	38%
Dkk3	MIP1beta	IL8	TGFbeta1	Mac2BP	TIMP1	EpCAM	48%	40%
Dkk3	MIP1beta	IL13	Mac2BP	IGFBP2	TIMP1	EpCAM	48%	31%
Dkk3	IL8	IL13	TGFbeta1	Mac2BP	TIMP1	EpCAM	48%	38%
M2PK	MIP1beta	IL8	IL13	Mac2BP	TIMP1	EpCAM	48%	33%
M2PK	MIP1beta	IL13	TGFbeta1	IGFBP2	TIMP1	EpCAM	48%	45%
M2PK	MIP1beta	TGFbeta1	Mac2BP	IGFBP2	TIMP1	EpCAM	48%	37%
M2PK	MIP1beta	IL13	TGFbeta1	Mac2BP	TIMP1	EpCAM	47%	41%
IL8	IL13	TGFbeta1	Mac2BP	IGFBP2	TIMP1	EpCAM	47%	40%
Dkk3	MIP1beta	IL13	TGFbeta1	Mac2BP	IGFBP2	TIMP1	46%	42%
M2PK	MIP1beta	IL8	TGFbeta1	Mac2BP	TIMP1	EpCAM	46%	30%
Dkk3	M2PK	MIP1beta	TGFbeta1	Mac2BP	TIMP1	EpCAM	45%	37%
Dkk3	M2PK	IL13	TGFbeta1	Mac2BP	TIMP1	EpCAM	45%	41%
Dkk3	MIP1beta	IL8	IL13	TGFbeta1	IGFBP2	EpCAM	45%	33%
Dkk3	MIP1beta	IL8	IL13	IGFBP2	TIMP1	EpCAM	44%	40%
Dkk3	MIP1beta	IL8	TGFbeta1	IGFBP2	TIMP1	EpCAM	44%	43%
Dkk3	MIP1beta	IL13	TGFbeta1	Mac2BP	TIMP1	EpCAM	44%	43%
MIP1beta	IL8	IL13	TGFbeta1	Mac2BP	TIMP1	EpCAM	44%	28%
Dkk3	M2PK	MIP1beta	IL13	TGFbeta1	Mac2BP	TIMP1	43%	31%
Dkk3	MIP1beta	IL13	TGFbeta1	IGFBP2	TIMP1	EpCAM	42%	40%
MIP1beta	IL13	TGFbeta1	Mac2BP	IGFBP2	TIMP1	EpCAM	42%	31%
Dkk3	M2PK	MIP1beta	IL13	TGFbeta1	Mac2BP	EpCAM	41%	23%
Dkk3	M2PK	MIP1beta	IL13	TGFbeta1	TIMP1	EpCAM	41%	33%
Dkk3	MIP1beta	IL13	TGFbeta1	Mac2BP	IGFBP2	EpCAM	41%	41%
Dkk3	MIP1beta	TGFbeta1	Mac2BP	IGFBP2	TIMP1	EpCAM	41%	39%
Dkk3	IL13	TGFbeta1	Mac2BP	IGFBP2	TIMP1	EpCAM	41%	37%
M2PK	MIP1beta	IL13	TGFbeta1	Mac2BP	IGFBP2	TIMP1	32%	27%

TABLE 15
Sensitivity of nine biomarker combinations in plasma and serum samples at 95% specificity.
BM1	BM2	BM3	BM4	BM5	BM6	BM7	BM8	BM9	Plasma	Serum
Dkk3	M2PK	MIP1beta	IL8	IL13	TGFbeta1	Mac2BP	IGFBP2	TIMP1	73%	77%
Dkk3	M2PK	MIP1beta	IL8	IL13	TGFbeta1	Mac2BP	IGFBP2	EpCAM	73%	77%
Dkk3	M2PK	MIP1beta	IL8	IL13	TGFbeta1	Mac2BP	TIMP1	EpCAM	54%	72%
Dkk3	M2PK	MIP1beta	IL8	IL13	TGFbeta1	IGFBP2	TIMP1	EpCAM	58%	74%
Dkk3	M2PK	MIP1beta	IL8	IL13	Mac2BP	IGFBP2	TIMP1	EpCAM	74%	70%
Dkk3	M2PK	MIP1beta	IL8	TGFbeta1	Mac2BP	IGFBP2	TIMP1	EpCAM	71%	78%
Dkk3	M2PK	MIP1beta	IL13	TGFbeta1	Mac2BP	IGFBP2	TIMP1	EpCAM	57%	72%
Dkk3	M2PK	IL8	IL13	TGFbeta1	Mac2BP	IGFBP2	TIMP1	EpCAM	67%	76%
Dkk3	MIP1beta	IL8	IL13	TGFbeta1	Mac2BP	IGFBP2	TIMP1	EpCAM	54%	55%
M2PK	MIP1beta	IL8	IL13	TGFbeta1	Mac2BP	IGFBP2	TIMP1	EpCAM	58%	69%

TABLE 16
Sensitivity of ten biomarker combinations in plasma and serum samples at 95% specificity.
BM1	BM2	BM3	BM4	BM5	BM6	BM7	BM8	BM9	BM1o	Plasma	Serum
Dkk3	M2PK	MIP1beta	IL8	IL13	TGFbeta1	Mac2BP	IGFBP2	TIMP1	EpCAM	70%	55%
Dkk3	M2PK	MIP1beta	IL8	IL13	TGFbeta1	Mac2BP	IGFBP2	TIMP1	EpCAM	73%	68%

It will be appreciated by persons skilled in the art that numerous variations and/or modifications may be made to the invention as shown in the specific embodiments without departing from the scope of the invention as broadly described.

The present embodiments are, therefore, to be considered in all respects as illustrative and not restrictive.

All publications discussed and/or referenced herein are incorporated herein in their entirety.

The present application claims priority from AU 2010903140, the entire contents of which are incorporated herein by reference.

Any discussion of documents, acts, materials, devices, articles or the like which has been included in the present specification is solely for the purpose of providing a context for the present invention. It is not to be taken as an admission that any or all of these matters form part of the prior art base or were common general knowledge in the field relevant to the present invention as it existed before the priority date of each claim of this application.

REFERENCES

• Anderson and Hunter (2006) Mol Cell Proteomics, 5:573-588.
• Cancer in Australia, an overview (2008) AIHW (Australian Institute of Health and Welfare) & AACR (Australian Association of Cancer Registries), Cancer series no. 46, Cat. No: CAN 42, Can berra: AIHW.
• Etzioni et al. (2003) Nat Rev Cancer, 3:243-252.
• Hundt et al. (2007) Cancer Epidemiol Biomarkers Prev, 16:1935-1953.
• Kimmel (1987) Methods Enzymol, 152:507-511.
• Kwoh et al. (1989) Proc Natl Acad Sci USA. 86:1173-1177.
• Levin (2004) Gastroenterology, 127:1841-1844.
• Lieberman (2010) Gastroenterology, 138:2115-2126.
• Morikawa et al. (2005) Gastroenterology, 129:422-428.
• Notomi et al. (2000) Nucleic Acids Res. 28:E63.
• Tonus (2006) World J Gastroenterol, 12:7007-7011.
• Wahl and Berger (1987) Methods Enzymol, 152:399-407.
• Walker et al. (1992a) Proc Natl Acad Sci USA. 89:392-396.
• Walker et al. (1992b) Nucleic Acids Res. 20:1691-1696.

Claims

1. A method for diagnosing or detecting colorectal cancer in a subject, the method comprising:

i) determining the presence and/or level of at least two biomarkers selected from the group consisting of IL-8, IGFBP2, MAC2BP, M2PK, IL-13, DKK-3, EpCam, MIP1β, TGFβ1, and TIMP-1 in a sample from the subject,

wherein the presence and/or level of the two biomarkers is indicative of colorectal cancer.

2. The method of claim 1, wherein the two biomarkers selected from M2PK, EpCam, IL-13, DKK-3, IL-8 and IGFBP2.

3. The method of claim 1, wherein the method comprises determining the presence and/or level of expression of at least three of the biomarkers.

4. The method of claim 3, wherein the three biomarkers are selected from the group consisting of M2PK, EpCam, IL-13, DKK-3, IL-8, IGFBP2, MIP1β, TGFβ1 and MAC2BP.

5. The method of claim 4, wherein the three biomarkers are as follows:

i) DKK-3, M2PK, and IGFBP2;

ii) M2PK, IGFBP2, and EpCAM;

iii) M2PK, MIP1β, and TGFβ1; or

iv) IL-8, IL-13, and MAC2BP.

6. The method of claim 1, wherein the method comprises determining the presence and/or level of expression of at least four of the biomarkers.

7. The method of claim 6, wherein four biomarkers are as follows:

i) DKK-3, M2PK, MAC2BP, and IGFBP2;

ii) IL-8, IL-13, MAC2BP, and EpCam;

iii) DKK3, M2PK, TGFβ1, and TIMP-1;

iv) M2PK, MIP1β, IL-13, and TIMP-1; or

v) IL-8, MAC2BP, IGFBP2, and EpCam.

8. The method of claim 1, wherein the method comprises determining the presence and/or level of at least five of the biomarkers.

9. The method of claim 1, wherein the method comprises determining the presence and/or level of at least six of the biomarkers.

10. The method of claim 1, wherein the method comprises determining the presence and/or level of at least seven of the biomarkers.

11. The method of claim 1, wherein the method comprises determining the presence and/or level of at least eight of the biomarkers.

12. The method of claim 1, wherein the method comprises determining the presence and or level of at least nine of the biomarkers.

13. The method of claim 1, wherein the method comprises determining the presence and/or level of at least ten of the biomarkers.

14. The method of claim 1, wherein the method comprises detecting the presence and/or level of least one additional biomarker selected from the group consisting of ICF-I, IGF-II, IGF-BP2, Amphiregulin, VEGFA, VEGFD, MMP-1, MMP-2, MMP-3, MMP-7, MMP-9, TIMP-1, TIMP-2, ENA-78, MCP-1, MIP-1β, IFN-γ, IL-10, IL-13, IL-1β, IL-4, IL-8, IL-6, MAC2BP, Tumor M2 pyruvate kinase, M65, OPN, DKK-3, EpCam, TGFβ-1, and VEGFpan.

15. The method of claim 1, wherein the method diagnoses or detects colorectal cancer with a sensitivity of at least 50%.

16. (canceled)

17. (canceled)

18. The method of claim 1, wherein the method diagnoses or detects Dukes Stage A colorectal cancer with a sensitivity of at least 50% and a specificity of at least 95%; a sensitivity of at least 60% and a specificity of at least 80%; or a sensitivity of at least 50% and a specificity of at least 90%.

19. (canceled)

20. (canceled)

21. (canceled)

22. (canceled)

23. (canceled)

24. (canceled)

25. (canceled)

26. The method of claim 1, wherein the method comprises:

i) determining the presence and/or level of the biomarkers in the sample from the subject; and

ii) comparing the presence and/or level of the biomarkers relative to a control, wherein a presence and/or level in the sample that is different to the control is indicative of colorectal cancer.

27. The method of claim 1, wherein the sample comprises blood, plasma, serum, urine, platelets, magakaryocytes or faeces.

28. A method of treatment comprising:

(i) diagnosing or detecting colorectal cancer according to the method of claim 1; and

(ii) administering a therapeutic suitable for the treatment of colorectal cancer.

29. A method for monitoring the efficacy of treatment of colorectal cancer in a subject, the method comprising treating the subject for colorectal cancer and then detecting the presence and/or level of at least two biomarkers selected from IL-8, IGFBP2, MAC2BP, M2PK, IL-13, DKK-3, EpCam, MIP1β, TGFβ1, and TIMP-1 in a sample from the subject, wherein an absence of and/or reduction in the level of expression of the biomarkers after treatment as compared with before treatment is indicative of an effective treatment.

30. (canceled)

31. (canceled)