Protein signature/markers for the detection of adenocarcinoma

[0001] The present invention relates to methods for the diagnosis of pancreatic adenocarcinoma and biomarkers and arrays for use for this purpose.

BACKGROUND OF THE INVENTION [0002] One of the constant challenges in oncology is the possibility of patient stratification in terms of the likelihood of tumor recurrence or resistance to drug treatment, or expected survival.

[0003] Adenocarcinoma of the Pancreas is characterized by the greatest lethal malignancy due to the anatomical site, with an annual detection of over 30,000 new cases and the same number of deaths in the United States alone, with 5 years of survival

- 5%. This extraordinary mortality is due to the lack of effective diagnostic methods at an early stage of cancer development (5) and the low effectiveness of existing therapies in advanced disease. Even patients (10-20%) who have been diagnosed with a surgically removable tumor eventually die of recurrent and metastatic disease. Hence, the increased ability to detect and predict cancer is critical to the treatment of individual patients.

[0004] Antibody microarray technique (3) has the ability to provide highly multi-component analysis (6, 7) and has been suggested to be a technology platform that will ultimately provide a specific protein signature, i.e. a combination of serum proteins that will distinguish cancer patients from individuals healthy. Microarray technique has now matured to the point where the initial defects have been overcome and small amounts of proteins in complex proteomes can be analyzed (8-12). However, gene expression profiling in tumors has shown the ability to predict survival only in a few cases (1, 2), and so far no combination of serum proteins has been associated with any of the above clinical parameters.

[0005] Analysis of serum samples to predict survival could enable more individualized treatment of cancer. This was emphasized, for example, in pancreatic adenocarcinomas in which there are no tumor-specific markers, although most patients have elevated CA 19-9 levels at the time of diagnosis, it has been shown that individual prognostic markers are inconclusive (4). In addition, non-invasive tests, such as computed tomography, have insufficient sensitivity to detect small tumors, while e.g. endoscopic ultrasound can be used to assess high-risk individuals for changes in the pancreas (5).

[0006] WO 2006/113210 A discloses a method of diagnosing the presence of a pancreatic cancer in a patient, comprising measuring serum concentration of markers in a blood marker panel containing two or more, three or more, four or more, five or more, six or more , seven or more, eight or more, nine or more of IP-10, HGF, IL-8, ^ FGF, IL-12p40, TNFRI, TNFRII, eotaxin, MCP-1 and CA 19-9, with a significant increase in concentration in serum IP-10, HGF, IL-8, // FGF, IL-12p40, TNFRI, TNFRII and CA 19-9 in a patient, compared to healthy, matched controls, and a significant decrease in eotaxin and MCP-1 serum levels in a patient, compared to healthy, matched controls (purportedly) indicate a likely diagnosis of a patient's pancreatic cancer (see paragraph 12).

[0007] In the light of this information, the inventors have now developed a proteomic approach to the prognostic diagnosis of cancer and have isolated the first set of serum biomarkers for detecting pancreatic cancer and predicting survival.

Summary of the Invention [0008] Thus, in a first aspect, the invention provides a method of determining the presence of pancreatic adenocarcinoma in an individual, comprising the following steps:

a) providing a protein sample for testing (e.g. serum or plasma);

b) determination of the protein signature of the test sample by measuring the presence and / or amount in the test sample of two or more proteins selected from the group specified in Table 1;

wherein the two or more proteins selected from the group specified in Table 1 comprise IL-5, and wherein the presence and / or amount in the test sample of two or more proteins selected from the group specified in Table 1 indicates the presence of pancreatic adenocarcinoma.

[0009] The term "protein signature" includes the meaning of a combination of the presence and / or amount of serum proteins present in a subject with cancer and which can be distinguished from a combination of the presence and / or amount of serum proteins present in a non-cancer subject (e.g. pancreatic adenocarcinoma) - i.e. in a normal or healthy individual.

[0010] As shown in the accompanying Examples, the presence and / or amount of some serum proteins present in the test sample may indicate the presence of a tumor, for example pancreatic adenocarcinoma, in the subject. For example, the relative presence and / or amount of some serum proteins in a single test sample may indicate the presence of a cancer, such as pancreatic adenocarcinoma, in an individual.

[0011] Preferably the subject is a human, but may be any mammal, for example a domesticated mammal (preferably of agricultural or commercial importance, including horse, pig, cow, sheep, dog and cat).

[0012] Preferably, the method of the first aspect of the invention further comprises the steps of:

c) providing a control sample of serum or plasma from an individual who does not suffer from pancreatic adenocarcinoma;

d) determining the protein signature of the control sample by measuring the presence and / or amount in the control of two or more proteins measured in step (b);

wherein the presence of pancreatic adenocarcinoma is found when the presence and / or amount in the test sample of two or more proteins measured in step (b) is different from the presence and / or amount in the control sample of two or more proteins measured in step (b) .

Preferably, the presence and / or amount in the test sample of two or more proteins measured in step (b) is significantly different (i.e. statistically different) from the presence and / or amount in the control sample of two or more proteins measured in step ( b). For example, as discussed in the attached Examples, a significant difference between the presence and / or amount of a particular protein in the test and control sample can be classified as a difference for which p <0.05.

[0014] Typically, the method of the first aspect comprises measuring the presence and / or amount in the test sample of all the proteins set out in Table 1 - i.e. all 19 proteins in Table 19.

[0015] Alternatively, the method of the first aspect may comprise measuring the presence and / or amount in a test sample of 2 or 3, or 4, or 5, or 6, or 7, or 8, or 9, or 10, or 11, or 12 , or 13, or 14, or 15, or 16, or 17, or 18, or 19 proteins referred to in Table 1.

[0016] In a preferred embodiment, the method of the first aspect comprises measuring the presence and / or amount in a sample of Rantes and / or eotaxin, and / or EI, and / or TNFb (1) and / or TNF-b (2), and / or GLP-1, and / or VEGF, and / or IL-13, and / or CD40.

[0017] Also described herein is a second method of determining the survival time of an individual suffering from pancreatic adenocarcinoma, comprising the steps of:

i) providing a serum or plasma sample for testing;

ii) determination of the protein signature of the test sample by measuring the presence and / or amount in the test sample of one or more proteins selected from the group specified in Table 2;

wherein the survival time of the subject is determined if the presence and / or amount in the test sample of one or more proteins selected from the group specified in Table 2 indicates survival time shorter than 12 months or longer than 12 months, or longer than 24 months.

[0018] Preferably the second method comprises the steps of:

iii) providing a first serum or plasma control sample from a subject having a survival time of less than 12 months and / or a second serum or plasma control sample from a subject having a survival time of more than 12 months and / or more than 24 months;

iv) determining the protein signature of the first and / or second control sample by measuring the presence and / or amount of one or more proteins measured in step (ii);

wherein the subject's survival is determined by comparing the presence and / or amount of one or more proteins in the test sample, measured in step (ii) with the presence and / or amount of one or more proteins in the first and / or second control measured in stage (iv).

[0019] By comparing the presence and / or amount of one or more proteins selected in the test sample and the control sample, it is possible to determine the survival time of an individual suffering from pancreatic adenocarcinoma. For example, if the test sample has the same (i.e. identical) or substantially similar or substantially similar presence and / or amount of one or more proteins selected as a control sample from a patient known to have a survival time of more than 24 months, such a test sample will be determined as a sample from a patient having Survival longer than 24 months. Other such comparisons will be understood by a specialist in the field of diagnostics.

[0020] Typically, the second method involves measuring the presence and / or amount in the test sample of all the proteins set out in Table 2 - i.e. all 22 proteins from Table 2. [0021] Alternatively, the second method may include measuring the presence and / or amount in the tested sample 1 or 2, or 3, or 4, or 5, or 6, or 7, or 8, or 9, or 10, or 11, or

12, or 13, or 14, or 15, or 16, or 17, or 18, or 19, or 20, or 21 or 22 of the proteins referred to in Table 2.

[0022] Preferably, the second method comprises measuring the presence and / or amount of a CD40 ligand and / or mucin, and / or IL-16, and / or Rantes, and / or eotaxin and / or MCP-4 in the sample, and / or or IL-11, and / or TNF-b, and / or IL-1ra, and / or MCP-3, and / or IL-1a, and / or IL-3, and / or C3, and / or LDL ( 1), and / or LDL (2), and / or Lewis Y.

Preferably the first aspect of the invention provides a method in which step (b) and / or step (d) is performed using a first binding agent capable of binding to two or more proteins. Preferably, in the second method step (ii) and / or step (iv) is performed using a first binding agent capable of binding to one or more proteins.

[0024] Binding agents (also called binding molecules) may be selected from the library based on the binding ability of the motif, as discussed below. [0025] At least one type, especially all types, of the binding molecules may be an antibody or fragments or variants thereof.

[0026] Thus, the fragment may contain one or more heavy chain (VH) or light chain (VL) variable domains. For example, the term antibody fragment includes Fab-like molecules (Better et al (1988) Science 240, 1041); Fv molecules (Skerra et al (1988) Science 240, 1038); single chain Fv molecules (ScFv), where the VH and VL partner domains are joined by a flexible oligopeptide (Bird et al. (1988) Science 242, 423; Huston et al. (1988) Proc. Natl. Acad Sci. USA 85, 5879) and single domain antibodies (dAb) containing isolated V domains (Ward et al (1989) Nature 341, 544).

[0027] The term "antibody variation" includes all synthetic antibodies, recombinant antibodies or antibody hybrids such as, but not limited to, a single chain antibody molecule produced by phage-displayed variable and / or constant regions of light and / or heavy chains of immunoglobulins or other immuno-interactive molecules capable of binding an antigen in an immunoassay format known to those skilled in the art.

[0028] A general overview of techniques used in the synthesis of antibody fragments that retain specific binding sites can be found in Winter & Milstein (1991) Nature 349, 293-299.

[0029] Additionally or alternatively, at least one type, especially all types of binding molecules are aptamers.

[0030] Molecular libraries such as antibody libraries (Clackson et al., 1991, Nature 352, 624-628; Marks et al., 1991, J Mol Biol 222 (3): 581-97), peptide libraries (Smith, 1985, Science 228 (4705): 1315-7), libraries of expressed cDNA (Santi et al. (2000) J Mol Biol 296 (2): 497-508), scaffold libraries other than antibody framework regions such as high-protein proteins affinity affibodies) (Gunneriusson et al., 1999, Appl Environ Microbiol 65 (9): 4134-40) or aptamer-based libraries (Kenan et al., 1999, Methods Mol Biol 118, 217-31) can be used as a source, from which binding molecules that are specific for a given motif are selected for use in the methods of the invention.

[0031] Molecular libraries can be expressed in vivo in prokaryotic (Clackson et al., 1991, op. Cit .; Marks et al., 1991, op. Cit.) Or eukaryotic (Kieke et al., 1999, Proc Natl Acad Sci USA, 96 (10): 5651-6) or can be expressed in vitro without the participation of cells (Hanes & Pluckthun, 1997, Proc Natl Acad Sci USA 94 (10): 4937-42; He & Taussig, 1997, Nucleic Acids Res 25 (24): 5132-4; ​​Nemoto et al., 1997, FEBS Lett, 414 (2): 405-8).

[0032] In cases where protein-based libraries are used, often genes encoding libraries of potential binding molecules are inserted into viruses and the potential binding molecule is displayed on the surface of the virus (Clackson et al., 1991, op. Cit .; Marks and et al., 1991, op. cit., Smith, 1985, op. cit.).

[0033] The most commonly used system currently is filamentous bacteriophage displaying antibody fragments on the surface, and the antibody fragments are expressed as a fusion with a small bacteriophage envelope protein (Clackson et al., 1991, op. Cit .; Marks et al., 1991, op. cit). However, other presentation systems are also used, using other viruses (EP 39578), bacteria (Gunneriusson et al., 1999, op. cit .; Daugherty et al., 1998, Protein Eng 11 (9): 825-32; Daugherty et al., 1999, Protein Eng 12 (7): 613-21) and yeast (Shusta et al., 1999, J Mol Biol 292 (5): 949-56).

[0034] In addition, presentation systems using a combination of a polypeptide product and its encoding mRNA in so-called ribosomal display systems have recently been introduced (Hanes & Pluckthun, 1997, op.cit .; He & Taussig, 1997, op.cit .; Nemoto et al. , 1997, op. Cit.) Or alternatively a combination of a polypeptide product with its encoding DNA (see US Patent No. 5,856090 and WO 98/37186).

[0035] When potential binding molecules are selected from libraries, one or more selection peptides having specific motifs are usually used.

Amino acid residues that provide structure, reducing flexibility in peptide or charged, polar or hydrophobic side chains, allowing interaction with a binding molecule, can be used in the design of motifs for selection peptides. For example, (i) Proline can stabilize the structure of a peptide because its side chain is attached to both the alpha carbon and nitrogen;

(ii) Phenylalanine, tyrosine and tryptophan have aromatic side chains and are highly hydrophobic, while leucine and isoleucine have aliphatic side chains and are also hydrophobic;

(iii) Lysine, arginine and histidine have basic side chains and are positively charged at neutral pH, while aspartate and glutamate have acid side chains and are negatively charged at neutral pH;

(iv) Asparagine and glutamine are neutral at neutral pH, but contain an amide group that can participate in hydrogen bonds;

(v) Serine, threonine and tyrosine side chains contain hydroxyl groups that may participate in hydrogen bonds.

[0036] Typically, the selection of binding molecules may involve the use of matrix techniques and systems to analyze binding to sites corresponding to types of binding molecules.

[0037] Preferably, the first binding agent is an antibody or fragment thereof;

more preferably a recombinant antibody or fragment thereof. Suitably, the antibody or fragment thereof is selected from the group consisting of: scFv; Fab; binding domain of an immunoglobulin molecule.

[0038] The heavy (VH) and light (VL) antibody variable domains are involved in antigen recognition; this fact was made known by the experience with early digestion with protease. Additional confirmation was obtained due to the "humanization" of rodent antibodies. Rodent-derived variable domains can be fused to human constant domains, such that the resulting antibody retains the antigen specificity of the parent rodent antibody (Morrison et al. (1984) Proc. Natl. Acad. Sci. USA 81, 6851-6855).

[0039] Antigen specificity is conferred by variable domains and is independent of the constant domains, as is known from experiments involving bacterial expression of antibody fragments, all containing one or more variable domains. These molecules include Fab-like molecules (Better et al (1988) Science 240, 1041); Fv molecules (Skerra et al. (1988) Science 240, 1038); single chain Fv molecules (ScFv), where the VH and VL partner domains are joined by a flexible oligopeptide (Bird et al. (1988) Science 242, 423; Huston et al. (1988) Proc. Natl. Acad Sci. USA 85, 5879) and single domain antibodies (dAb) containing isolated V domains (Ward et al (1989) Nature 341, 544). A general review of techniques used in the synthesis of antibody fragments that retain specific binding sites can be found in Winter and Milstein (1991) Nature 349, 293-299.

[0040] By "ScFv molecules" we mean molecules in which the VH and VL partner domains are connected by a flexible oligopeptide.

[0041] The benefits of using antibody fragments, not whole antibodies, are manifold. Smaller fragment sizes can lead to improved pharmacological properties, for example better penetration through solid tissue. The effector functions of whole antibodies, such as complement fixation, are eliminated. Each of the Fab, Fv, ScFv and dAb antibody fragments can be expressed in and secreted by E. coli, thus allowing easy production of large amounts of these fragments.

[0042] Whole F (ab ') 2 antibodies and fragments are "divalent". By "bivalent" we understand that these antibodies and F (ab ') 2 fragments have two sites that bind to the antigen. In contrast, the Fab, Fv, ScFv and dAb fragments are monovalent, having only one site binding to the antigen. [0043] Antibodies can be monoclonal or polyclonal. Suitable monoclonal antibodies can be produced by known techniques, for example disclosed in "Monoclonal Antibodies: A manual of techniques", H Zola (CRC Press, 1988) and in "Monoclonal Hybridoma Antibodies: Techniques and applications", JGR Hurrell (CRC Press, 1982) , both of which are incorporated herein by reference.

[0044] In a preferred embodiment, the invention provides a method in which two or more proteins in a test sample are labeled with a detectable moiety. Preferably, the first aspect provides a method in which two or more proteins in a control sample are labeled with a detectable moiety. Alternatively, in the second method one or more proteins in the first and / or second control sample are labeled with a detectable moiety.

[0045] By the term "detectable moiety" we mean that such a moiety is one that can be detected and the relative amount and / or location of the moiety (e.g. location in the matrix) can be determined.

[0046] Detectable moieties are well known in the art.

[0047] The detectable moiety may be a fluorescent and / or luminescent and / or chemiluminescent moiety that, when exposed to particular conditions, can be detected. For example, a fluorescent moiety may need to be exposed to radiation (i.e., light) of a certain wavelength and intensity in order to induce excitation of the fluorescent moiety and thus to allow the emission of detectable fluorescence at a certain wavelength that can be detected.

[0048] Alternatively, the detectable moiety may be an enzyme that is capable of converting the (preferably undetectable) substrate into a detectable product that can be visualized and / or detected. Examples of suitable enzymes are discussed in more detail below with respect to, for example, ELISA tests.

[0049] Alternatively, the detectable moiety may be a radioactive atom that is useful in imaging. Suitable radioactive atoms include^99mTc and ¹²³And for scintigraphic research. Other readily detectable moieties include, for example, magnetic resonance imaging (MRI) spin labels, such as again¹²³AND, ¹³¹AND, ¹¹¹in, ¹⁹F ¹³C ¹⁵N ¹⁷Oh, gadolinium, manganese or iron. Undoubtedly, the detectable agent (such as, for example, one or more proteins in the test sample and / or the control sample described herein, and / or the antibody molecule to be used in detecting the selected protein) must have a sufficient amount of appropriate atomic isotopes for the detectable moiety to be able to be easily detected.

[0050] Radiolabels and other labels may be embedded in known means in the agents of the invention (i.e., proteins present in the samples of the methods of the invention and / or binding agents of the invention). For example, if the binding moiety is a polypeptide, it can be biosynthesized or can be synthesized using chemical amino acid synthesis using appropriate amino acid precursors containing, for example, fluorine-19 instead of hydrogen. Tags such as^99mtc ¹²³AND, ¹⁸⁸rh ¹⁸⁸Rh and ¹¹¹In, for example, they can be attached via cysteine ​​residues on the binding moiety. Yttrium-90 can be attached via a lysine residue. Built-in¹²³And the IODOGEN method can be used (Fraker et al. (1978) Biochem. Biophys. Res. Comm. 80, 49-57). Literature reference ("Monoclonal

Antibodies in Immunoscintigraphy ". JF Chatal, CRC Press, 1989) describes in detail other methods. Methods for coupling other detectable moieties (such as enzymatic, fluorescent, luminescent, chemiluminescent or radioactive moieties) to proteins are well known in the art.

[0051] The attached Examples provide an example of methods and detectable moieties for labeling agents of the invention (such as, for example, proteins in a sample of the methods of the invention and / or binding molecules) for use in the methods of the invention.

[0052] Preferably, the detectable moiety is selected from the group consisting of: fluorescent moiety; luminescent grouping; chemiluminescent moiety; radioactive moiety; enzyme moiety.

[0053] Preferably, the first aspect provides a method in which step (b) and / or step (d) is performed using a matrix. Suitably, in the second method, step (ii) and / or step (iv) is performed using a matrix.

[0054] Matrices as such are well known in the art. They are usually formed by a linear or two-dimensional structure having separated (ie, discrete) regions ("spots"), each having a limited surface area formed on the surface of a solid support. The matrix may also have a ball structure, each ball may be identified by a molecular code or a color code, or identified in a continuous flow. The analysis can also be performed sequentially when the sample is passed over a series of spots, each of which adsorbs a specific class of molecules from the solution. The solid support is usually glass or polymer, the most commonly used polymers are cellulose, polyacrylamide, nylon, polystyrene, polyvinyl chloride or polypropylene. The solid support can be in the form of tubes, beads, discs, silicon chips, microplates, polyvinylidene difluoride (PVDF) membranes, nitrocellulose film, nylon film, other porous film, non-porous film (among others, e.g. plastic, polymer, persepex, silicone), multiple polymer pins or multiple microtiter wells or any other surface suitable for immobilizing proteins, polynucleotides and other suitable molecules and / or performing an immunoassay. Binding processes are well known in the art and generally include cross-linking, covalent binding or physical adsorption of a protein molecule, polynucleotide or the like into a solid support. The location of each spot can be determined using well-known techniques such as contact or contactless printing, masking or photolithography. A review can be found in Jenkins, RE, Pennington, SR (2001, Proteomics, 2,13-29) and Lal et al. (2002, Drug Discov Today 15; 7 (supplement 18): S143-9).

[0055] Typically, the matrix is ​​a microarray. By "microarrays" we mean a matrix of regions having a density of discrete regions of at least about 100 / cm²and preferably at least about 1000 / cm². Regions in the microarray have typical dimensions, e.g., diameters, in the range between about 10-250 μm and are separated from other regions in the matrix by the same distance. The matrix may also be a macroarray or nanomarray.

[0056] Other suitable binding molecules (discussed above) have been identified and isolated, and one of skill in the art can produce the matrix using methods well known in the field of molecular biology.

[0057] Alternatively, the first aspect of the invention provides a method in which step (b) and / or step (d) is performed using a test comprising a second binding agent capable of binding two or more proteins, and the second binding agent has a detectable moiety.

[0058] Typically, in the second method, step (ii) and / or step (iv) is performed using a test comprising a second binding agent capable of binding one or more proteins, and the second binding agent has a detectable moiety.

[0059] Binders are described in detail above.

[0060] In a preferred embodiment, the second binding agent is an antibody or fragment thereof; usually a recombinant antibody or fragment thereof. Suitably, the antibody or fragment thereof is selected from the group consisting of: scFv; Fab; binding domain of an immunoglobulin molecule. Antibodies are described in detail above.

Preferably, when using the assay, the invention provides a method in which the detectable moiety is selected from the group consisting of: a fluorescent moiety; luminescent grouping; chemiluminescent moiety; radioactive moiety; enzyme moiety. Examples of suitable detectable moieties for use in the methods of the invention are described above. [0062] Preferred tests for detecting serum or plasma proteins include an enzyme-linked immunosorbent assay (ELISA), radioimmunoassay (RIA), immunoradiometric assay (IRMA) and enzyme-linked immunoassay (IEMA), including double binding assays using monoclonal antibodies and / or polyclonal. Exemplary double binding assays are described by David et al. in US Patent Nos. 4,376110 and 4,486,530, hereby incorporated by reference. Staining cells with antibodies on slides can be used in methods well known in cytological laboratory diagnostic tests well known to those skilled in the art.

[0063] Typically, the assay is ELISA (enzyme immunosorbent assay), which usually involves the use of enzymes that give colored reaction products, usually in solid phase assays. Enzymes such as horseradish peroxidase and phosphatase are widely used. A way to amplify phosphatase reactions is to use NADP as a substrate to produce NAD, which acts as a coenzyme for the second enzyme system. Escherichia coli pyrophosphatase provides a good conjugate because this enzyme is not present in the tissues, it is stable and gives a good reaction color. Chemiluminescent systems based on enzymes such as luciferase may also be used.

[0064] Vitamin, biotin conjugation is often used because it can be easily detected by reaction with avidin or streptavidin conjugated with an enzyme with which it binds with high specificity and affinity.

[0065] In a second aspect, the invention provides a matrix for determining the presence of pancreatic adenocarcinoma in a subject, comprising two or more binding agents of the invention. Preferably, two or more binders are capable of binding all the proteins defined in Table 1.

[0066] Also described herein is a matrix for determining the survival time of an individual suffering from pancreatic adenocarcinoma, comprising one or two binders according to the other described herein. Preferably, the one or more binding agents are capable of binding all the proteins defined in Table 2.

[0067] Arrays suitable for use in the method of the invention are discussed above. In a third aspect, the invention provides the use of an array for use in the method of the first aspect of the invention. The use of arrays for use in the second method described is also described herein.

[0068] In a fourth aspect, the invention provides the use of two or more proteins selected from the group specified in Table 1 as a diagnostic marker for determining the presence of pancreatic adenocarcinoma in a subject in which two or more proteins selected from the group specified in Table 1 include IL-5 . Conveniently, all the proteins set out in Table 1 are used as a diagnostic marker to determine the presence of pancreatic adenocarcinoma in a subject.

[0069] Also described herein is the use of one or more proteins selected from the group defined in Table 2 as a diagnostic marker for determining the survival time of an individual suffering from pancreatic adenocarcinoma. Preferably, all the proteins set out in Table 2 are used as a diagnostic marker to determine the survival time of an individual suffering from pancreatic adenocarcinoma.

[0070] In a fifth aspect, there is provided a kit for determining the presence of pancreatic adenocarcinoma, comprising:

A) two or more binders or a matrix according to the first aspect of the invention;

B) instructions for carrying out the method of the first aspect of the invention, wherein the two or more proteins selected from the group specified in Table 1 comprise IL-5.

[0071] In a sixth aspect of the invention there is provided a kit for determining the presence of pancreatic adenocarcinoma, comprising:

A) two or more binding agents specified in the description,

B) instructions for performing the method of the first aspect of the invention.

[0072] Also described herein is a kit for determining the survival time of an individual suffering from pancreatic adenocarcinoma, comprising:

1) one or more of the first binding agents specified herein;

2) instructions for performing the method according to the second method.

[0073] Also described herein is a kit for determining the survival time of an individual suffering from pancreatic adenocarcinoma, comprising:

1) one or more second binders as defined herein;

2) instructions for performing the method as specified in the second method.

[0074] The list and discussion of previously published documents in this description of the invention should not be taken as an acknowledgment that the document is part of the state of the art or is of general knowledge.

[0075] Preferred, non-limiting examples that illustrate some aspects of the invention will now be described with reference to the following tables and figures:

Table 1: Serum protein profile for distinguishing healthy individuals from patients with pancreatic cancer

Table 2: Protein signature of serum proteins to distinguish people with short survival (<12 months) versus people with long survival (> 24 months) among patients with pancreatic cancer

Table 3: 129 recombinant antibody fragments directed against 60 serum proteins for use in the microarray according to the invention

Table 4: Demographics of patients from whom serum samples were obtained. PC = pancreatic adenocarcinoma [0076] Figure 1: Detection of pancreatic adenocarcinomas using serum protein expression analysis using recombinant antibody microarrays. (a) The scanned image of the microarray with antibodies contained 1280 data points; (b) The multivariate analysis corresponded to the unattended Sammon plot based on all 129 antibody fragments, which showed that cancer patients were completely separated (red) from healthy individuals (blue); (c) dendrogram in which cancer patients (PA) were completely separated from normal individuals (N); Bidirectional hierarchical grouping was based on 19 serum biomarkers that were significantly (p &lt;0.05) differently expressed in tumor-bearing individuals compared to normal subjects, using a training kit consisting of 28 serum samples. Subsequently, the tested set of 16 serum samples was classified 100% correctly (marked *). The columns correspond to donors; blue indicates normal individuals (N) and red indicates cancer patients (PA). Each row corresponds to the biomarker in serum as shown on the right, and each pixel indicates the level of expression of a given biomarker in each donor (overexpression (red), reduced expression (green) or no change (black) in serum from cancer patients pancreas compared to serum from normal individuals); (d) A number of serum biomarkers such as IL-4, IL-5, IL-13 and MCP-3 were also determined by ELISA to confirm the results of the microarray analysis. A representative data set for IL-13 is shown showing that the usual ELISA and antibody microarray analysis gave similar results. The sensitivity of the microarray analysis was equal to or better than that of the ELISA (data not shown).

[0077] Figure 2: Identification of the predictive protein signature of biomarkers in serum to distinguish two patient cohorts with short (<12 months) and long (> 24 months) survival. (a) Area under the characteristic curve (ROC) as a function of the number of analytes included in the predictive signature, which clearly shows that two patient cohorts with different survival times can be accurately distinguished by using a signature of> 29 analytes; (b) ROC surface area of ​​signature of predictive biomarkers in serum, based on 29 analytes identified by antibodies; (c) SVM was trained using the biomarker signatures selected by the training kit. A test set of 10 randomly selected patients (samples marked with *) was then classified using the SVM predictive value; (d) heat map based on 22 non-redundant serum proteins in the predictive signature. The columns indicate cancer patients, where blue means long (> 24 months) survival and red means short (<12 months) survival. Colored codes as in the explanations of Figure 1c.

[0078] Figure 3: Basis of microarray technique with recombinant antibodies.

EXAMPLE

Overview [0079] The driving force of oncoproteomics is the identification of protein signatures that are associated with a particular malignant tumor. Based on the microarray analysis with recombinant antibodies of unfractionated human serum proteomes obtained from pancreatic cancer patients and normal healthy donors, we identified a protein signature based on 22 non-redundant analytes to distinguish cancer patients from healthy individuals. The predicted specificity and sensitivity were 99.7% and 99.9%, respectively. In addition, a protein signature of 19 protein analytes was identified that has the ability to predict the survival of cancer patients. This new predictive signature distinguished patients with a survival time of <12 months and> 24 months and suggests that new possibilities exist in individualized medicine.

[0080] The present study describes an affinity proteomics approach for predicting cancer diagnosis based on recombinant antibody microarrays using recombinant scFv fragments adapted to the matrix (12, 13). The results show that an array with antibody fragments specific for immunoregulatory proteins can distinguish human serum proteosomes obtained from cancer patients from serum proteosomes taken from healthy individuals. We present the first set of serum biomarkers for detecting pancreatic cancer and predicting patient survival.

Materials and methods [0081] Production and purification of scFv - 129 human recombinant scFv antibody fragments directed against 60 different proteins, mainly involved in immunoregulation, were selected from the n-CoDeR library (13) and kindly provided by BioInvent International AB (Lund, Sweden). Hence, each antigen was recognized by a maximum of four different scFv fragments. All scFv antibodies were produced in

100 ml E. coli cultures and purified from expression supernatants using Ni-NTA agarose affinity chromatography (Qiagen, Hilden, Germany). Bound molecules were washed with 250 mM imidazole, extensively dialyzed against PBS and stored at 4 ° C until further use. Protein concentration was determined by measuring absorbance at 280 nm (mean concentration 210 μg / ml, range 60-1090 μg / ml). Purity was assessed using 10% SDS-PAGE (Invitrogen, Carlsbad, CA, United States).

[0082] Serum Samples - A total of 44 serum samples provided by the Southern General Hospital in Stockholm (Sweden) and the University of Lund Hospital (Lund, Sweden) were included in this study. 24 serum samples (PA1-PA30) were taken from patients with pancreatic cancer at the time of diagnosis. 20 serum samples (N1-N20) (no clinical symptoms) were taken from healthy donors. Table 4 presents patient demographics. All samples were aliquoted and stored at -80 ° C according to standard operating procedure. [0083] Labeling of serum samples - Serum samples were labeled using previously optimized serum proteome labeling protocols (9,14,15). All serum samples were biotinylated using EZ-Link Sulfo-NHS-LC-Biotin (Pierce, Rockford, IL, United States). 50 Pl aliquots of serum were centrifuged at 16,000 xg for 20 minutes at 4 ° C and diluted 1:45 in PBS to give a concentration of about 2 mg / ml. The samples were then biotinylated by the addition of Sulfo-NHS-biotin to a final concentration of 0.6 mM for 2 h on ice, with careful mixing with a vortex mixer every 20 minutes. Unreacted biotin was removed by dialysis against PBS for 72 hours using a 3.5 kDa dialysis membrane (Spectrum Laboratories, Rancho Dominguez, CA). Samples were aliquoted and stored at -20 ° C.

[0084] Enzyme-linked immunosorbent assay - Serum concentration of 4 protein analytes (MCP-3, IL-4, IL-5 and IL-13) was measured in all samples using commercially available ELISA kits (Quantikine, R&amp;D Systems, Minneapolis, MN, United States). The measurements were carried out in accordance with the instructions provided by the supplier.

[0085] Preparation and processing of microarrays with antibodies - We used a previously optimized and validated kit (9,12,14,15) to produce microarrays with antibodies. Briefly, scFv microarrays were produced using a contactless printer (Biochip Arrayer, Perkin Elmer Life & Analytical Sciences) that deposited about 330 pl / drop using piezo technique. An scFv antibody matrix was created by instillation of 2 drops in each position and the first drop was allowed to dry before instillation of the second drop. Antibodies were dropped onto black polymer slides of MaxiSab microarray (NUNC A / S, Roskilde, Denmark), obtaining an average of 5 fmol scFv per site (range 1.5-25 fmol). Each scFv clone was instilled eight times to get the appropriate statistics. A total of 160 antibodies and controls were printed on the slide, arranged in two columns, 8 x 80 antibodies in each column. To facilitate grid setting during the assay, a row containing Cy5 conjugated to streptavidin (2 μg / ml) was instilled every tenth row. A hydrophobic pen (DakoCytomation Pen, DakoCytomation, Glostrup, Denmark) was used to draw the hydrophobic barrier around the matrix. The arrays were blocked with 500 μΐ 5% (w / v) nonfat dry milk (Semper AB, Sundbyberg, Sweden) in PBS overnight. All incubations were carried out in a humidity chamber at room temperature. The matrices were then washed four times with 400 μΐ 0.05% Tween-20 in PBS (PBS-T) and incubated with 350 μΐ biotinylated serum, diluted 1:10 (obtaining a total dilution of serum 1: 450) in 1% (wt. / vol) nonfat dry milk and 1% Tween in PBS (PBSMT) for 1 h. Then the matrices were washed four times with 400 μΐ PBS-T and incubated with 350 μΐ 1 μg / ml Alexa-647 conjugated with streptavidin, diluted in PBSMT by 1 hour Finally, the arrays were washed four times with 400 μΐ PBS-T, immediately dried in a gas and nitrogen stream, and scanned using a confocal microarray scanner (ScanArray Express, Perkin Elmer Life & Analytical Sciences) with a resolution of 5 μm, using six different scanner settings. ScanArray Express version 2.0 software (Perkin Elmer Life & Analytical Sciences) was used to quantify the intensity of each site using the fixed wheel method. The local background was subtracted and to compensate for possible local losses, the two highest and two lowest results for a given analyte were automatically rejected; each data point represents the average value of the other four determinations. The correlation coefficient was> 0.99 for tests in one set and> 0.96 for tests performed in different sets, respectively.

[0086] Data normalization - Only unsaturated sites were used for further data analysis. The data sets were normalized within a given chip using a semi-global standardization strategy conceptually similar to the standardization developed for DNA microarrays. Hence, the coefficient of variation (CV) was first calculated for each analyte and ordered. Fifteen percent of the analytes that had the lowest CV values ​​among all samples were determined, which corresponds to 21 analytes, and were used to calculate the normalization factor for individual chips. The normalization coefficient Ni was calculated from the formula Ni = Si / μ, where Si is the sum of signal intensities for 21 analytes for each sample, and μ is the sum of signal intensities for 21 anaiits averaged for all samples. Each data set obtained from one sample was divided by the normalization factor Nl. For intensities, log2 values ​​were used in the analysis.

[0087] Data analysis - Sammon mapping was made using the Euclidean distance in the space of all 129 analytes. Supervised classification was determined using a support vector machine (SVM) using a linear nucleus (16-18). It was determined that the cost of violation of restrictions (parameter C in SVM) is 1, which is the default value in the function R svm, and no attempts were made to adjust it. It was decided not to adjust this parameter to avoid over-fitting the model and make the classification procedure easier to understand. The SVM result of the sample being tested is the SVM decision value, which is indicated by the distance to the hyperplane. In Figures 1C and 2C, the division into training and test kit was made once at random and then kept until the end. Figure 2A uses the n-fold cross-validation procedure. For each K number between 1 and 129, the following procedure was used. For the training set, i.e., all samples except one, K was selected in the order of analytes in the Wilcoxon test and SVM was trained with these K analytes. The SVM decision value for the left sample was then calculated using this classifier. As is common practice, this was done with all samples in n-fold cross-validation.

[0088] The Characteristic-Operational Curve (ROC) was constructed using the SVM decision value and the area under the curve was determined. Figure 2A shows the ROC surface area as a function of K. Figure 2B shows the ROC curve for K = 29. All statistics were performed in R (19).

Results [0089] Ductal pancreatic adenocarcinoma is a cancer with poor prognosis and an improved diagnostic tool that facilitates making clinical decisions could bring significant benefits to patients. One strategy to improve diagnostics is to identify a set of biomarkers that can enable cancer detection and predict clinical prognosis. As a result, to be able to identify the protein signature associated with pancreatic cancer with high sensitivity, we designed the first large-scale microarray (Figure 1A) based on 129 recombinant antibody fragments (12, 14, 15) directed against 60 serum proteins, mainly immunoregulatory ( Table 2). In this study, labeled sera from 24 people with pancreatic cancer and 20 healthy people were incubated on antibody microarrays, which were then determined using a confocal scanner. Initially, to test our ability to detect cancer, microarray data were presented in the form of an unattended Sammon plot based on all antibodies, and two separate populations could be clearly distinguished (Figure 1B). This indicates a clear difference between the tumor and normal proteomes in terms of serum analytes analyzed using microarrays. We then performed N-fold cross-validation using a support vector machine (SVM) and decision values ​​were collected for each sample. The decision value is the result of the predictor, and it is anticipated that samples with a predictive value above (below) the threshold are taken from (healthy) pancreatic cancer patients. The threshold is a compromise between sensitivity and specificity and is often, but not always, zero. 24 samples from people with pancreatic cancer gave decision values ​​in the range of 0.30 to 1.93, and samples from healthy people in the range from -1.84 to -0.30. Hence, with a threshold value of zero or any other value between -0.30 and 0.30, the sensitivity and specificity in our data set is 100%. However, to extrapolate sensitivity and specificity to a larger population, we first checked that the decision values ​​were indicative of a normal distribution in the normal and cancer groups, respectively, and we calculated the means and variances. By setting the classification threshold in the middle between two means and using normal distributions, we obtained a sensitivity of 99.9% and a specificity of 99.3%, which indicates excellent classification power even in a larger population. To illustrate the clear separation between normal and cancer groups, we randomly selected a training kit consisting of 18 cancer and 10 normal samples. This training set of tumor and normal serum proteomes determined a smaller bi omarker set consisting of 19 non-redundant serum proteins that differed significantly (p &lt;0.05) between the two samples. These differently expressed proteins were then used to construct a dendrogram of 28 training samples and 16 other samples that were used as the test set. As can be seen in Figure 1C, the tumor samples are completely separated from the normal samples for both the training kit and the test kit (100% sensitivity and specificity).

[0090] An interesting observation was the fact that, using a blank method, we took three serum samples from one patient (PA14) at different time points with an interval of several weeks, but 11-12 months before the diagnosis of this patient's pancreatic cancer. Despite this, all samples were correctly classified as cancerous when used in the test kit (data not shown). Importantly, the protein signature, specified by the training kit and used to classify the samples tested, is specific for pancreatic adenocarcinomas and differs from the serum signatures found in our microarray in other cancers, such as gastric (9) and breast adenocarcinomas (manuscript in preparation) . Some of the data obtained in the microarray was also confirmed by analyzing a number of serum proteins by a simple enzyme immunosorbent assay (Figure 1D). However, analysis based on microarray measurements is often more sensitive compared to traditional enzyme immunoassays. As a result, the analytes could be verified when the ELISA test was sufficiently sensitive, but then our microarray data was confirmed.

[0091] While early detection of cancer is justified, especially in pancreatic cancer, it has also been suggested that profiling of serum proteins is a way of determining signatures that, in addition to the classification of people with cancer against normal people, may also be associated with clinical parameters (16) . Predicting the expected survival time would be very important as it could affect the treatment regimens assigned to individual patients. As a result, to further explore our microarray platform with recombinant antibodies, we compared two cohorts of cancer patients divided into individuals with short survival (<12 months) and long survival (> 24 months). We first calculated the area under the Characteristic Surgical Curves (ROC) as a function of the total number of analytes determined by the antibodies in the predictive signature, using the Wilcoxon test to filter analytes, and then using a support vector machine (SVM) (Figure 2A). All 129 antibodies were included in these calculations, and because we had 1 to 4 antibody / serum analytes, there was some degree of redundancy in the size of the predictive signature biomarker. Based on these calculations, it is clear that it was possible to distinguish these two cohorts, with a surface area (AUC) ROC> 0.80. Importantly, this curve also showed that the protein signature of <26 analytes provided a more variable and less resistant predictor. As a result, we selected a predictor signature for 29 analytes for further analysis. The ROC curve for 29 analytes has an area under the curve of 0.86 (Figure 2B).

[0092] Again, to illustrate the predictive ability of this biomarker signature, pancreatic cancer patients (n = 23), including people with short and long survival, were randomly divided into a training set of 13 patients and a study set of 10 patients. Because the average survival time for patients with inoperable disease is between 5 and 6 months, there was an unavoidable systematic error in terms of cohort size and the cohort of patients with long survival included only 5 patients. SVM was trained with the biomarker signature selected by the training kit, and then the test kit could be classified as shown in Figure 2C. All patients with &lt;12 months survival were correctly classified using the SVM predictive value &lt;0, which was considered the most important classification. One patient with long survival was incorrectly classified. 29 the most important analytes separating long-term and short-term survivors among all 23 patients in the Wilcoxon test corresponded to 22 non-redundant serum proteins (7 of 29 analytes were duplicate but determined by different antibody clones). This new predictor signature, represented by 22 non-redundant proteins and differentiating analyte response among individuals with short and long survival, respectively, is shown as a heat map in Figure 2D. Analysis of individual proteins did not show a strict consensus on the pattern among serum proteins, although it is clear that cytokines such as IL-1a, IL-3, IL-8 and IL-11 were upregulated in people with short survival, whereas Rantes, IL-16, IL-4 and eotaxin were most upregulated in people with long survival (Figure 2D). The importance of this observation still requires validation, but may potentially indicate a more active T cell population in this second patient population.

Discussion of results [0093] Microarrays with antibodies, as a tool in affinity proteomics, have evolved in recent years from a promising tool to a strategy that is beginning to yield promising results in oncoproteomics (3, 12, 20, 21). The main purpose of these activities is to detect cancer at an early stage of development, predict cancer recurrence and resistance to treatment, or select patients for a specific treatment regimen (3). This is especially important in the case of tumors characterized by poor prognosis, which also applies to pancreatic cancer, because it gives metastases quickly, e.g. to lymph nodes, lungs, peritoneum (4, 23), and its diagnosis at an early stage of disease development is difficult. However, until now there have been difficulties in distinguishing between different cancers or between cancer and inflammation based on the biomarker signature (for review, see item 3) (20). The observed difference between serum proteomes in cancer and normal individuals obtained in this study most likely depends on the range of specificity of the microarray antibodies, which in turn is also supported by the rational matrix design described by Sanchez-Carbayo et al. (21). These researchers were able to stratify patients with bladder cancer based on overall survival using antibodies produced against differently expressed gene products. In recent years, we have developed a highly efficient microarray platform with recombinant antibodies for complex analysis of proteomes (6, 9, 11, 12, 14, 15), assessing and optimizing key technical parameters (24) such as probe and substrate design (13, 25 ), matrix / test design (9, 15) and sample format (9, 14, 15). This enabled us to perform the first differentiating expression profile of human plasma proteome proteins using optimized scFv microarrays targeted primarily at immunoregulatory proteins. In line with previous results, these antibody microarrays had sensitivity in the pM to fM range, easily detecting low-concentration cytokines. In addition, we maintained the test repeatability with a correlation coefficient in the range of 0.96 - 0.99, which is a key feature of multiplex analyzes and looks favorable compared to previous reports (12, 26). In addition, antibody microarray data were compared with ELISA tests if they were of sufficient sensitivity and this normal test confirmed our results.

[0094] Patients with pancreatic cancer are often diagnosed late, which leads to poor prognosis. Due to the rarity, it is difficult to collect a large number of samples, especially from patients with long survival, i.e.> 24 months. In this study, we had access to 25 patients, which, thanks to rigorous statistical evaluation, allowed us to classify cancer and normal proteomes. This is a supervised classification and we used the carrier vector machine (SVM) as the classifier, although we obtained very similar results for this data set using a naive Bayer classifier (data not shown). SVM separated these two groups by defining a hyperplane in the space of all analytes and assigned samples on one side of the hyperplane to one group and those on the other side to the other group. The distance from the hyperplane is called the predictive or decision value (Figure 2C). The hyperplane and with it the group classification was determined using our training kit. Then, the classifier characteristics were evaluated using a test kit, with no overlap between the training kit and test kit. In contrast, the data set can be randomly divided into different training and test sets, which are then used to train and test the classifier accordingly. The disadvantage of this approach is that the final result depends on the division into the training and test set. As a result, we used cross-validation as a procedure to perform a number of divisions of our data set and we used the average characteristics of the test sets as a measure of the accuracy of data classification. Therefore, in the N-fold cross-validation that was carried out, the test kit contains one sample and the training kit contains the remaining samples. SVM characteristics can be measured using an ROC curve, and in particular, the area under the ROC curve. Samples from normal people and people with pancreatic cancer were surprisingly well separated, because SVM classified all samples correctly, with a gap between the two groups. Extrapolation of decision values ​​gave very high sensitivity (99.9%) and specificity (99.3%), showing that misclassification occurs in one out of several hundred samples.

[0095] In this study, we were unable to compare pancreatic cancer patients with pancreatitis patients, which could be a desirable comparison, but instead we used serum samples from normal people. Importantly, this biomarker signature associated with pancreatic cancer, however, only had eotaxin, IL-5 and IL-13 as common elements with fourteen bacterial infection biomarkers associated with another gastrointestinal tumor (12), indicating that the pancreatic signature was not associated with general inflammation. In addition, this signature was not similar to biomarkers found in systemic lupus erythematosus, an autoimmune disorder with significant inflammatory component (Wingren et al., Manuscript in preparation). The signature was also completely different from the one described by Orchekowski et al. (26), obtained when profiling serum samples from patients with pancreatic cancer using microarrays based on monoclonal and polyclonal antibodies. These authors relied primarily on high concentration serum proteins such as albumin, transferrin and hemoglobin, and more common inflammatory markers such as C-reactive protein (CRP), serum amyloid A and immunoglobulins, with only eight cytokines analyzed. On the other hand, our current cancer signature contained a number of overexpressed TH2 cytokines (IL-4, -5, -10 and -13), and classic TH1 cytokines (IL-12 and TNF-b) were down regulated, which was also consistent with the results of the Belone et al. study, which showed that TGF-b and IL-10 were upregulated in the sera of patients with pancreatic cancer (27). These authors also showed that monocytes obtained from the blood of patients with pancreatic cancer were stimulated to develop a TH2-like, rather than TH1-like response, with increased IL-4 expression and decreased IL-12 expression.

[0096] Finally, we examined the possibility of identifying a signature that, in addition to being able to classify cancer and normal samples, could also be used to predict patient survival. Initially, SVM could classify patients with short and long survival with an ROC 0.81 surface area, using all analytes (data not shown), which was very promising. A classifier was then made for each number of biomarkers, selecting the most relevant analytes, which was later used to distinguish between two groups of samples in the training set. As shown in Figure 2A, the classifier characteristics were stable above 26 analytes and we could show that the signature of 29 biomarkers (22 non-redundant analytes) gave an ROC of 0.86. However, a study involving more than 18 patients with short survival and 5 patients with long survival is necessary in order to firmly determine the profile of survival classifying proteins, but this study clearly shows the possibility of establishing such a profile.

[0097] In summary, by using a microarray with recombinant antibodies against immunoregulatory proteins, we were able to specifically detect pancreatic adenocarcinoma and completely differentiate between normal and serum cancer proteomas. More importantly, the first attempt to determine a signature capable of predicting the survival of cancer patients is presented, indicating the strength of oncoproteomic affinity in making clinical decisions.

References [0098]

1. Rosenwald A. et al. Cancer Cell 3, 195-197 (2003).

2. van de Vijver, et al. N. Eng. J. Med 347, 1999-2009 (2002).

3. Borrebaeck, CAK Expert Opin. Biol. Ther 6, 833-838 (2006).

4. Garcea, G., Neal, CP, Pattenden, CJ, Steward, WP and Berry, DP Eur. J. Cancer 41, 2213-2236 (2005).

5. Yeo, TP et al. Curr. Probl. Cancer 26, 176-275 (2002).

6. Wingren, C. and Borrebaeck, CAK. Exp. Rev. Proteomics 1, 355-364 (2004).

7. Pavlickova, P., Schneider, ME, and Hug, H., Clin. Chim. Acta 342, 17-35 (2004).

8. Haab, BB, et al. Genome Biol. 2, 1-13 (2001).

9. Wingren, C., et al., Microarrays based on affinity-tagged single-chain Fv antibodies: sensitive detection of analyte in complex proteomes. Proteomic 5, 1281-1291 (2005).

10. Pawlak, M. et al. Proteomics 2, 383-393 (2002).

11. Wingren, C. and Borrebaeck, CAK OMICS 3, 411-427 (2006).

12. Ellmark, P., et al., Identification of protein expression signatures associated with H. pylori infection and gastric adenocarcinoma using recombinant antibody microarrays. Mol Cell Proteomics 5, 1638-1646 (2006).

13. Soderlind, E., et al., Recombining germline-derived CDR sequences for creating diverse single-framework antibody libraries. Nat. Biotechnol, 18, 852-856 (2000)

14. Ingvarsson, J .; Larsson, A .; Sjoholm, L .; Truedsson, L .; Jansson, B .; Borrebaeck, CAK and Wingren, C. Design of recombinant antibody microarrays for serum protein profiling: Targeting of complement proteins. J. Proteome Res.in press

15. Wingren, C., Ingvarsson, J., Dexlin, L., Szul, D. and Borrebaeck, CAK. Design of recombinant antibody microarrays for complex proteome analysis: choice of sample labeling-tag and solid support. Proteomics in press

16. Eisen, MB, et al., Cluster analysis and display of genome-wide expression patters. Proc Natl Acad Sci USA. 95, 14863-14868 (1998).

17. N. Cristianini and J. Shawe-Taylor, An introduction to support vector machines (and other kernelbased learning methods), Cambridge University Press (2000).

18. Chih-Chung Chang and Chih-Jen Lin, LIBSVM: a library for support vector machines, available at <a href="http://www.csie.ntu.edu.tw/~cjlin/libsvm">http://www.csie.ntu.edu.tw/~cjlin/libsvm</a>

19. R. Ihaka and R. Gentleman, R: A language for data analysis and graphics, J. Comp. Graph Stat. 5, 299-314, (1996)

twenty. Sanchez-Carbayo, M., Socci, ND, Lozano, JJ, Haab, BB and Cordon-Cardo,

C. Am. J. Pathol. 168, 93-103 (2006).

21. Schafer, MW, Mangold, L., Partin, AW and Haab, BB (2007) Antibody array profiling reveals serum TSP-I as a marker to distinguish benign from malignant prostatic disease, The Prostate 67, 255-267.

22. Rustgi, AK Gastroenterology 129, 1344-1347 (2005).

23. More, G., et al. Natl. Acad. Sci (USA) 102, 7677-7682 (2005).

24. Wingren, C. and Borrebaeck, CAK (2007) nya reviewn

25. Steinhauer, C., et al., Biocompatibility of surfaces for antibody microarrays: design of macroporous silicon substrates. Anal Biochem 341, 204-13 (2005)

26. Orchekowski, R., et al., Antibody microarray profiling reveals individual and combined serum proteins associated with pancreatic cancer. Cancer Res. 65,

11193-202 (2005)

27. Bellone, G., et al., Tumor-associated transforming growth factor-beta and interleukin-10 contribute to a systemic Th2 immune phenotype in pancreatic carcinoma patients. Am J Pathol. 155, 537-47 (1999)