The present invention relates to methods in diagnosing of prostate cancer patients.

Afflicting one out of nine men over age 65, prostate cancer (PCa) is a leading cause of male cancer-related death, second only to lung cancer. It is the most common type of cancer in men in the U.S., with 186, 000 new cases in 2008 and 28, 600 deaths.

Proper and early diagnosis of PCa is paramount for appropri ^¬ ate medical treatment to preserve the life of the patient. Cur ^¬ rently, PCa is typically diagnosed with a digital rectal exam and/or prostate specific antigen (PSA) screening. An elevated serum PSA level can indicate the presence of PCa. PSA is used as a marker for PCa because it is secreted only by prostate cells. While healthy men have a low PSA serum it is raised in men with prostate cancer and further increased with the extent and the severity of the tumour. An elevated PSA level - especially when above 3 ng/ml - may raise a doctor's suspicion that a patient has PCa, while amounts above 50 in the majority of cancer cases indicate spread elsewhere in the body. Although PSA is consid ^¬ ered as an effective tumour marker, it is not cancer specific. There is a considerable overlap in PSA concentrations in men with PCa and men with benign prostatic diseases such as prosta ^¬ titis. Elevated levels of PSA are detectable in a large percent ^¬ age of patients with non-malignant disorders, thus limiting the diagnostic specificity of serum PSA levels in PCa patients. Moreover, there is also variation in the levels of PSA in indi ^¬ viduals from different races. Based on these findings, PSA serum levels alone are not exclusively used for the diagnosis of PCa. Recently, attempts have been made to improve the sensitivity of the PSA assay by relating PSA levels to the size of the prostate (PSA density) , or to measure the change of PSA levels over time (PSA velocity) or the ratio of bound to free PSA as an addition ^¬ al indication of the presence of clinically relevant PCa.

However, there is still need for more reliable diagnostic methods or tools for the identification and management of PCa. Nowadays, when PSA or digital rectal examination (DRE) indicates a strong likelihood that cancer is present, a transrectal ultra ^¬ sound (TRUS) is used to map the prostate and show any suspicious areas. Biopsies of various sectors of the prostate are used to determine if PCa is present as well as to determine the stage of the cancer.

In recent years, the advances in technologies in genomics and proteomics have led to the identification of several new candidate markers for PCa. Many new genes involved in PCa have been reported since the first microarray studies. Consequently, the long list of genes implicated in PCa includes Hepsin, a transmembrane serine protease, which is over-expressed in PCa compared to benign hyperplasia of normal prostate; RabGTPase- activating protein (PRC17), calcium-binding protein (S100-P) , an actin polycomb group protein enhancer of zeste homologue 2 (EZH2), alpha-methylacyl-CoA racemase (AMACR) , prostate-specific membrane antigen (PSMA) , selenoprotein-P, Wnt signalling (Wnt 5A) , Enlongin-C, androgen regulated gene (FKBP5, largely un ^¬ known) , ANKH (regulates transportation of pyrophosphate, pros ^¬ tate A regulated transcript-1 (PART-1), prostate short chain DR1 (PSDR-1, Claudin-8, Gro2 and specific granule protein-28 (SG028). However, most of these investigations have concentrated on establishing unique associations of PCa with genetic or epi- genetic determinants. In addition, linkage studies have a weak power to associate changes of small to moderate effects to dis ^¬ ease outcomes.

Recently, metabolites, such as sarcosine, cysteine, gluta- mate, asparagine, glycine, leucine, proline, threonine, histi- dine, n-acetyl-aspartic acid (N-acetylaspartate (NAA) ) , inosine, inositol, adenosine, taurine, creatine, uric acid, glutathione, uracil, kynurenine, glycerol-s-phosphate, glycocholic acid, su ^¬ beric acid, thymine, glutamic acid, xanthosine, 4- acetamidobutyric acid, citrate, malate, and N-acetylyrosine or thymine were identified in prostate tissue samples and/or in blood of PCa patients using mass spectrometry. In fact, the di ^¬ agnostic value of biomarkers for PCa has been hampered by their poor sensitivity and specificity, and the non-conclusive under ^¬ standing of the mechanisms controlling their differential levels in PCa. As a result of these limitations, the use of biomarkers is currently limited to research settings, and none has been recommended for routine assessment.

Early diagnosis of PCa is very important because this in ^¬ creases the chance of survival and influences the kind of treat ^¬ ment, because treatment of PCa depends quite often on the stage of the cancer. Patients with a normal life expectancy of 10 years or less, whose Gleason score is low, and who have a local ^¬ ized tumour are often treated with watchful waiting (no treat ^¬ ment) . Treatment options for more aggressive cancers include surgery (radical prostatectomy (RP) ) , chemotherapy and radiation application to kill or eliminate the cancer cells locally. Other methods of treating PCa include hormone therapy such as anti- androgen alone or in combination with surgery or radiation. Furthermore, luteinizing hormone-releasing hormone (LH-RH) analogs, which block the release of pituitary hormones stimulating testosterone production, have also been used for PCa treatment. However, the treatment of patients with hormones confines pa ^¬ tients to hormonal injections for the rest of their lives, which may cause other side effects and could lead to hormone resistant tumours. For example, androgen ablation is the most common therapy for advanced PCa, leading to massive apoptosis of androgen- dependent malignant cells and temporary tumour regression. In most cases, however, the tumour relapses and can proliferate in ^¬ dependently of androgen supply. While surgical and radiation treatments are often effective for localized PCa, advanced dis ^¬ ease remains incurable. Therefore, it is essential to have means for prognosticating the risk or probability that a patient has with respect to such PCa relapses. Currently, the standard diagnostic means for detecting relapse of PCa is a steady meas ^¬ urement of PSA. Increases in postoperative serum PSA at levels of 0.001 to 0.1 yg/l after radical prostatectomy are associated with poor prognosis. Monitoring postoperative cases with a high ^¬ ly sensitive PSA assay (detection limit 0.001 yg/l) are suggest ^¬ ed to detect clinically important biochemical relapse early af ^¬ ter radical prostatectomy. These patients may be suitable for early intervention (e.g. radiation, hormone ablation therapy or chemotherapy) .

Estimation of risk for relapse before RP is usually made on the basis of PSA (a higher PSA level indicates a higher risk for relapse) or tumour volume (a higher tumour volume (as well as tumour grade and tumour stage) directly correlates with the risk for relapse) (Cordon-Cardo et al . , J.Clin. Invest. 117 (2007), 1876-1883) . Other studies suggest time of diagnosis and PSA dou ^¬ bling time as major risk factors for PCa relapse - besides PSA values (Slovin et al . , Clin. Cancer Res. 11 (2005), 8669-8673). In addition, it could also be shown that the surgeon's expertise correlates with the risk for PCa relapse.

Currently, however, PCa relapse is detected by PSA recur ^¬ rence (concordance index being 0.77, sensitivity of 77% and specificity of 72% (Cordon-Cardo et al . , 2007)). Kumar-Sinha et al. (Urology 62 (2003): 19-35) and Shulka-Dave et al . (Clin. Cancer Res. 15 (2009): 3842-3849) refer to proteinaceous mark ^¬ ers). Glunde et al. (Exp. Rev. Mol. Diagnos. 6 (2006): 821-829), Jordan et al . (Exp. Rev. Proteom. 4 (2007): 389-400), Glunde et al. (Pharmacogenomics 7 (2007): 1109-1123) and WO 2009/026152 Al disclose the use of small molecule markers for diagnosis in fields significantly differing from PCa relapse diagnosis.

Accordingly, there is still an urgent need for an early, rapid and reliable prognosis for PCa relapse which should pref ^¬ erably be present before RP, ideally requiring only minute amounts of samples which can easily be taken from a patient, such as blood. This is of great relevance for selection of prop ^¬ er type and time of treatment.

Therefore, the present invention provides the use of at least one substance selected from the group consisting of Phos ^¬ phatidylcholine with diacyl residue sum C24:0 (PC aa C24:0); Phosphatidylcholine with diacyl residue sum C40:3 (PC ae C40:3); Phosphatidylcholine with diacyl residue sum C40:4 (PC ae C40:4); Lysophosphatidylcholine with acyl residue sum C26:0 (lysoPC a C26:0); Lysophosphatidylcholine with acyl residue sum C6:0 (lysoPC a C6:0); 13 (S) -hydroxy-9Z, 1 lE-octadecadienoic acid (13S- HODE) ; 12 (S) -hydroxy-5Z, 8Z, 10E, 14Z-eicosatetraenoic acid (12S- HETE) ; 15 (S) -hydroxy-5Z, 8Z, 11Z, 13E-eicosatetraenoic acid (15S- HETE) ; Leukotriene B4 (LTB4) ; Prostaglandin E2 (PGE2) ; Prostaglandin D2 (PGD2); 7a-Hydroxycholesterol (7aOHC) ; 7- Ketocholesterol ( 7KC) ; 5B , 6B-Epoxycholesterol ( 5b, 6b, EPC) ; 5a, 6a-Epoxycholesterol ( 5a, 6a, EPC) ; and 4B-Hydroxycholesterol (4BOHC) ; for prognosticating the relapse of a prostate cancer (PCa) in a sample of a body fluid or a tissue sample of a PCa patient .

With the present invention, a superior set of small molecule biomarkers is provided, which allows a reliable prognosis for PCa relapse at an early stage, i.e. even before tumour surgery or tumour irradiation therapy. For example, the prognostic mole ^¬ cules according to the present invention have proven to allow a proper and statistically significant prognosis whether the tu ^¬ mour patient will show relapse within one, two or three years

(or whether the patient will remain relapse-free for more than one, two or three years) . The markers according to the present invention are molecules which are known to be involved in bio ^¬ chemical processes in the body. However, these molecules have never been discussed before as being relevant in connection with PCa relapse. The present invention provides for the first time a class of PCA relapse markers which structurally belong to the group of lipid molecules including metabolites thereof (in con ^¬ trast to the protein markers of the prior art) . The lipid mark ^¬ ers according to the present invention are diligently selected from very specific groups of molecules (metabolites) relevant in lipid metabolism. These molecules are known to regulate cell function via a complex network involving G-protein linked receptors. This goes along with the concept that the majority of the metabolites behaved in a chemical-class specific pattern. This agrees with findings that related classes of metabolites are generally clustered together, thus revealing related functional ^¬ ity. In contrast to that, prior art documents refer - if at all - to protein markers, including enzymes, which are significantly larger than metabolites being claimed in the present application which have formal molecular masses well below 1000. This is a significant part for enabling a robust and easy prognosis tool for PCa relapse which provides an early information, because tests which involve protein markers are always less robust and reliable due to the relative differences in stabilities and mo ^¬ lecular masses, especially if the detection should be also ena ^¬ bled by MALDI techniques. The markers according to the present invention are of similar nature and can fairly be substituted for one another (at least to a certain extent which is also made clear with the experimental details disclosed herein) . All al ^¬ ternatives have the common diagnostic property with respect to the present invention and they are structurally related in so far as they are members of the lipid metabolism.

The determination of "fingerprints" of small molecules

("metabolomic profiles") is increasingly important in the field of diagnostics. The generation and the analysis of small mole ^¬ cule profiles of cells, cellular compartments, and specific or ^¬ ganelles (e.g., mitochondria, Golgi, endoplasmic reticulum, cy- toplasm, nucleus, etc.) is generally referred to as "metabolom- ics" ("the systematic study of the unique chemical fingerprints that specific cellular processes leave behind"; specifically, the study of their small-molecule metabolite profiles) . Metabo ^¬ lites have the advantage that they directly reflect the biologi ^¬ cal processes in the cells, are easier to be profiled, and gen ^¬ erally offer a new route for the identification of diagnostic targets .

Whereas such metabolomic profiles have already been investi ^¬ gated in PCa (Osl et al . , Bioinform. 24 (2008), 2908-2914); Lokhov et al . , Metabolomics DOI 10.1007/sll306-009-0187-x; Sreekumar et al . , Nature 457 (2009), 910-914 (identifying sarco- sine as one of the major differential metabolites distinguishing between tumour and non-tumour tissue) ) , nothing is known about metabolomics and differentiated metabolomics for PCa relapse risk .

Interestingly, the markers according to the present invention completely differ from the metabolites known to be indica ^¬ tive for PCa as indicated e.g. by Sreekumar et al . (2009) . For example, sarcosine is definitely not a suitable marker for the present invention.

The marker set according to the present invention includes phosphaditylcholines (Phosphatidylcholine with diacyl residue sum C24:0 (PC aa C24:0); Phosphatidylcholine with diacyl residue sum C40:3 (PC ae C40:3); Phosphatidylcholine with diacyl residue sum C40:4 (PC ae C40:4)) lysophosphaditylcholines (Lysophospha- tidylcholine with acyl residue sum C26:0 (lysoPC a C26:0); Lyso- phosphatidylcholine with acyl residue sum C6:0 (lysoPC a C6:0)) prostaglandins (13 (S) -hydroxy-9Z, llE-octadecadienoic acid (13S- HODE) ; 12 (S) -hydroxy-5Z, 8Z, 10E, 14Z-eicosatetraenoic acid (12S- HETE) ; 15 (S) -hydroxy-5Z, 8Z, 11Z, 13E-eicosatetraenoic acid (15S- HETE) ; Leukotriene B4 (LTB4) ; Prostaglandin E2 (PGE2) ; Prostaglandin D2 (PGD2)), and oxysterols ( 7a-Hydroxycholesterol (7aOHC); 7-Ketocholesterol ( 7KC) ; 5B , 6B-Epoxycholesterol

(5b,6b,EPC); 5a, 6a-Epoxycholesterol (5a,6a,EPC); and 4β- Hydroxycholesterol (4BOHC) ) . (Glycero-phospholipids are differ ^¬ entiated with respect to the presence of ester (a) and ether (e) bonds in the glycerol moiety, where two letters (aa, ea, or ee) denote that the first and the second position of the glycerol scaffold are bound to a fatty acid residue, whereas a single letter (a or e) indicates a bond with only one fatty acid resi ^¬ due; e.g. PC_aa_C24:0 denotes a phosphatidylcholine with 24 car ^¬ bons in the two fatty acid side chains and no double bond in one of them) .

Importantly, it has been shown with the present invention that other members of the compound classes phosphaditylcholines , lysophosphaditylcholines , prostaglandines and oxysterols have no marker function with respect to the PCa relapse status.

With the present invention it is possible to reliably pre ^¬ dict whether a given patient has a certain risk for developing PCa relapse (i.e. to prognosticate whether he will develop PCa relapse or not) . Knowing this prognosis, this risk can then be used for designing specific treatment of these patients (i.e. to decide on the optimal available treatment option and time of treatment) which could include more severe measures typically applied in PCa treatment (i.e. instead of "only RP" or "only ir ^¬ radiation" or "only hormone therapies", combinations of these measures or higher doses of medication/irradiation are applied). The results obtained by the present invention may also have im ^¬ plications for the RP itself which could be carried out in larg ^¬ er areas if the patient has a prognosis for PCa relapse. On the other hand, patients with no PCa relapse risk can be treated in a more convenient way.

Accordingly, the present invention also relates to a method for prognosticating relapse of a prostate cancer (PCa) patient in a sample of a body fluid or a tissue sample of said PCa pa ^¬ tient which is characterised in that at least one substance se ^¬ lected from the group consisting of Phosphatidylcholine with di- acyl residue sum C24:0 (PC aa C24:0); Phosphatidylcholine with diacyl residue sum C40:3 (PC ae C40:3); Phosphatidylcholine with diacyl residue sum C40:4 (PC ae C40:4); Lysophosphatidylcholine with acyl residue sum C26:0 (lysoPC a C26:0); Lysophosphatidyl ^¬ choline with acyl residue sum C6:0 (lysoPC a C6:0); 13 (S)- hydroxy- 9Z , 1 lE-octadecadienoic acid (13S-HODE); 12 (S) -hydroxy- 5Z, 8Z, 10E, 14Z-eicosatetraenoic acid (12S-HETE); 15 (S) -hydroxy- 5Z, 8Z, 11Z, 13E-eicosatetraenoic acid (15S-HETE); Leukotriene B4 (LTB4); Prostaglandin E2 (PGE2); Prostaglandin D2 (PGD2); 7a- Hydroxycholesterol (7aOHC) ; 7-Ketocholesterol ( 7KC) ; 5B, 6β- Epoxycholesterol ( 5b, 6b, EPC) ; 5a, 6a-Epoxycholesterol

( 5a, 6a, EPC) ; and 4B-Hydroxycholesterol (4BOHC) is detected and quantified in said sample and the result of this quantification is compared with the amount of the at least one detected and quantified substance in a corresponding sample of known relapse status .

The samples for the present invention are derived from PCa tumour patients (which are, of course, always male human pa ^¬ tients) , especially patients that have already been confirmed to have PCa and expect RP . Preferably, the samples are taken and preferably also analysed according to the present invention be ^¬ fore RP so that the results are available for the RP so that ap ^¬ propriate further measures can be foreseen even before or during surgery (e.g. appropriate planning of the kind and mode of sur ^¬ gery) .

A specific advantage of the present invention is that the measurements can be carried out in blood or urine samples which can be obtained much easier than the prostate biopsy samples (or the excised prostate) which are usually investigated for full diagnosis of PCa patients. Therefore, the present method is preferably performed on a blood (or blood derived) sample (espe ^¬ cially a blood, plasma or serum sample) or a urine sample. How ^¬ ever, also tissue samples (e.g. prostate or liver tissue), cere ^¬ bral spinal fluid samples, lymph samples, semen samples, etc. are also useable according to the present invention.

The results of the quantification of the substances accord ^¬ ing to the present invention are compared to the levels of ref ^¬ erence for the substances tested, i.e. to levels which either prognose the development of PCa relapse or to levels which do not prognose the development of PCa relapse. Typically, refer ^¬ ence levels for the metabolites according to the present inven ^¬ tion are given, which are indicative for PCa relapse (or for having a reduced risk for PCa relapse) . Preferably, the quanti ^¬ fied levels for the samples according to the present invention are compared with such known reference levels. Therefore, in a preferred embodiment of the present invention, the amount of the at least one detected and quantified substance in a correspond ^¬ ing sample of known relapse status is a numerical limit for the amount of this substance.

The reference levels applied in the present method may also be tailored to specific techniques that are used to measure lev ^¬ els of metabolites in biological samples (e.g., LC-MS, GC-MS, etc.) , where the levels of metabolites may differ based on the specific technique that is used. Establishment of such reference values for a given technique is easily possible for a person skilled in the art.

The sample analysis according to the present invention can be done with any method known and applied for metabolite finger ^¬ printing. Usually, metabolomic analyses include the following important steps, which are well available to the person skilled in the art: First, the metabolites have to be extracted from the biological source (tissue sample or sample of a body fluid) in an efficient and unbiased manner. Specific care must be taken to prevent unwanted processes (such as oxidation) in the samples or at least take these processes into consideration when further processing and analysing those samples. Then the analytes can be separated, usually by chromatographical methods or electrophore ^¬ sis, particularly capillary electrophoresis, liquid chromatog ^¬ raphy, especially HPLC, or gas chromatography. Detection, identification and quantification of the analytes are usually done by mass spectrometry (MS) or nuclear magnetic resonance (NMR) spectroscopy .

LC- or GC-MS is the most preferred way to develop metabolic profiles. In addition, mass spectral fingerprint libraries exist or can be developed that allow identification and ( interlabora- torial) comparison of a metabolite according to its fragmenta ^¬ tion pattern. MS is both sensitive and can be very specific. There are also a number of studies which use MS as a stand-alone technology: the sample is infused directly into the mass spec ^¬ trometer with no prior separation, and the MS serves to both separation and to detection of metabolites.

NMR does not rely on separation of the analytes, and the sample can thus be recovered for further analyses. All kinds of small molecule metabolites can be measured simultaneously - in this sense, NMR is close to being a universal detector. Practi ^¬ cally, however, it is relatively insensitive compared to mass spectrometry-based techniques; additionally, NMR spectra can be very difficult to interpret for complex mixtures.

MS and NMR are by far the two leading technologies for metabolomics . Other methods of detection that have been used in ^¬ clude electrochemical detection (coupled to HPLC) , radiolabel (when combined with thin-layer chromatography) , chemical modifi- cation assays (e.g. radioactive labeling), and enzymatic assays.

Preferably, the sample analysis with respect to the metabo ^¬ lites according to the present invention is performed by mass spectroscopy. Therefore, the method according to the present in ^¬ vention is preferably detected and quantified by mass spectros ^¬ copy, preferably liquid chromatography mass spectroscopy (LC- MS) , especially high performance liquid chromatography mass spectroscopy (HPLC-MS) or reverse phase liquid chromatography mass spectroscopy (RPLC-MS) ; electro-spray ionisation mass spec ^¬ troscopy (ESI-MS) , gas chromatography mass spectroscopy (GC-MS) , atmospheric pressure chemical ionisation mass spectroscopy (AP- CI-MS) , capillary electrophoresis mass spectroscopy (CE-MS) , tandem mass spectroscopy (MS-MS) ; or combinations of these mass spectroscopy techniques.

Each of the PCa marker substances according to the present invention is useable by itself (i.e. without the need for other markers) to establish the prognosis according to the present in ^¬ vention. Each of them shows statistically significant increase in pre-RP samples of patients with a risk of PCa relapses (ex ^¬ cept 12S-HETE, however, 12S-HETE shows performance as a marker by other selection approaches) . A widely accepted threshold for statistical significance is generally "p value below 0.05". For -omics applications, the "p value" is replaced by "q value" to take into account that the generation of many "p values" will increase the number of falsely significant results. A more ro ^¬ bust approach consists in basing the selection on two or more statistical measures. In the present case, a metabolite is pref ^¬ erably entering the final selection if passing at least two criteria amongst: "q value below 0.1", "AUC greater than 0.7", "fold change greater than 20%". This is a preferred selection criterion for the present invention and additional modelling did come up with similar conclusions) . However, due to the possibil ^¬ ities of mass spectrometers, it is very easy to analyse more than one marker in a sample. This allows an even more reliable and sensitive estimation for the PCa relapse prognosis. Accord ^¬ ingly, it is preferred that at least two, preferably at least three, especially at least four of the substances are deter ^¬ mined, quantified and compared. Even more preferred, at least five, preferably at least ten, especially all sixteen, of the metabolites according to the present invention are analysed. The prognosis developed according to the present invention may also further be combined with known prognosis factors for PCa relapse, such as pre-RP-PSA values, tumour size, grade and status, etc.. The methods according to the present invention can also be combined with alpha-KGA (alpha-keto glutaric acid) , an ^¬ other marker which has significance in detection of PC relapse. Alpha-KGA can even be applied independently from the other (li ^¬ pid) markers according to the present invention.

As a preferred reference value, amounts of the metabolites according to the present invention are compared with the values in a corresponding sample of a PCa relapse patient (i.e. a sam ^¬ ple of known prognosis for PCa relapse) . The "corresponding sam ^¬ ple" is, of course, the sample that qualitatively corresponds to the sample measured according to the present invention, e.g. a blood sample corresponds to a reference blood sample or a urine sample corresponds to a reference urine sample. Preferably, the reference sample also corresponds to the measured sample con ^¬ cerning the detection method or even the detection apparatus (i.e. the same or the same type of mass spectrometer) . Prefera ^¬ bly, such a "real world" reference value is not only a single sample, but an average value based on a sample collective.

According to a preferred embodiment of the present inven ^¬ tion, the amount of the at least one detected and quantified substance in a corresponding sample of known relapse status is therefore the amount of this substance in a sample of a PCa re ^¬ lapse patient or the average amount of this substance of a pool of samples of PCa relapse patients.

The prognosis markers according to the present invention are differentially upregulated in samples with expectation of PCa relapse .

This means that there is a statistically significant corre ^¬ lation between the increase of the substances (metabolites, PCa relapse markers) according to the present invention in samples of PCa patients and an increased risk of PCa relapse after RP . Otherwise, if these substances are not upregulated, the PCa pa ^¬ tient has a good chance of not developing a relapse (e.g. in more than two or more than three years; there is even the chance that these patients do have significantly less relapse in PCa within a higher year number (e.g. at least five or at least ten years) . Accordingly, in a preferred embodiment of the present method the sample is regarded as having a prognosis for PCa re ^¬ lapse, if the at least one substance is contained in an amount which is at least 10%, preferably at least 20%, more preferably at least 50 %, even more preferred at least 100%, especially at least 200%, increased compared to the amount of this substance in a sample of a PCa patient (or a group or even population of such patients in a given country, region or genetic pool) who does not have a prognosis for PCa relapse ("med FC") . Some of the marker substances provided according to the present inven ^¬ tion have even significantly higher "-fold differences" (com ^¬ pared to the non PCa relapse) . Any preferred combination of me ^¬ tabolite marker substances according to the present invention can be elaborated based on the information contained herein. Op ^¬ timisation with respect to sensitivity, significance, specifici ^¬ ty, etc. can be made by the person skilled in the art. In doing so, also the data contained herein, especially according to the disclosure in the example section, is specifically helpful. Preferably, the statistical methods mentioned and referred to in the example section are applied to the analysis according to the present invention, specifically for selecting reference values and statistical significance for attribution PCa relapse progno ^¬ sis (or not having PCa relapse prognosis) .

Preferred markers have at least two of the following proper ^¬ ties: AUC of at least 0.70; med FC of at least 20, especially at least 50; and a p-value of 0.05 or lower, especially of 0.01 or lower .

When the amount (s) or level (s) of the one or more metabo ^¬ lites in the sample of the PCa patient is (are) determined, the ^¬ se amount (s) or level (s) may be compared to PCa relapse metabo ^¬ lite-reference levels, such as PCa relapse-positive and/or PCa relapse-negative reference levels to aid in diagnosing or to di ^¬ agnose whether the subject has a risk of PCa relapse. To this end, it has also to be borne in mind that biological measure ^¬ ments always show natural variations which will be attended to by appropriate calibration of the individual systems. Objective values are therefore often problematic to establish whereas rel ^¬ ative differences between the relapse/non relapse groups are easily providable for a given technique and/or apparatus set-up without the actual need for objective standard values being in ^¬ dependent of the detection and quantification tools. Preferably, the markers according to the present invention are selected from 15S-HETE, PGD2 , PC ae C40:3, LTB4 , PGE2 , 12S- HETE, PC ae C40:4, lysoPC a C26:0, 13S-HODE, 5b, 6b-EpoxyC or combinations (of two, three, four, five, six, etc.) thereof. Specifically preferred are 15S-HETE, PGD2, LTB4, PGE2 and 12S- HETE . Another preferred group of markers are PC aa C24:0, PC ae C40:3, PC ae C40:4, lysoPC a C26:0, 13S-HODE, 12S-HETE, 15S- HETE, LTB4, PGE2, PGD2 , 7aOHC, 7KC, 5B,6B,EPC, 5a,6a,EPC and 4BOHC. Another preferred group of markers are PC aa C24:0, PC ae C40:3, lysoPC a C26:0, lysoPC a C6:0, 13S-HODE, 12S-HETE, 15S- HETE, LTB4, PGE2, PGD2 , 7aOHC, 7KC, 5B,6B,EPC, and 5a,6a,EPC.

The method according to the present invention is in a pre ^¬ ferred embodiment characterised in that the comparison is per ^¬ formed using software-based statistical and bioinformatic data analyses .

According to another aspect, the present invention relates to the use of a mass spectrometer for carrying out the present method by measuring one or more (or all) of the metabolites ac ^¬ cording to the present invention.

A further aspect of the present invention relates to a kit for carrying out the method according to the present invention. The kit according to the present invention comprises

- a mass spectrometer,

- a standard sample containing a known amount of at least one substance from the group of PCa relapse marker substances according to the present invention (and therefore a known status with respect to PCa relapse) and

- a sample of a PCa patient containing an unknown amount of at least one substance from the group of PCa relapse marker substances according to the present invention.

According to a preferred embodiment, the kit of the present invention is characterised in that it further comprises one or more of the following components: quality control samples, ana- lyte standards, internal standards or data analysis software.

In some embodiments of the present invention, a computer- based analysis program is used to translate the raw data gener ^¬ ated by the detection assay (e.g., the presence, absence, or amount of a PCa relapse marker metabolite) into data of predic ^¬ tive value for a clinician. The clinician can access the predic ^¬ tive data using any suitable means. Thus, in some embodiments, the present invention provides the further benefit that the cli ^¬ nician, who is not likely to be trained in metabolite analysis, needs not to understand the raw data. The data is presented di ^¬ rectly to the clinician in its most useful form. The clinician is then able to immediately utilize the information in order to optimize the care of the subject.

The profile data is then prepared in a format suitable for interpretation by a treating clinician. For example, rather than providing raw data, the prepared format may represent a diagno ^¬ sis or risk assessment (e.g., likelihood of PCa relapse being present) for the PCa patient, along with recommendations for particular treatment options. The data may be displayed to the clinician by any suitable method. For example, in some embodi ^¬ ments, the profiling service generates a report that can be printed for the clinician (e.g., at the point of care) or dis ^¬ played to the clinician on a computer monitor.

The present invention is further described by the following examples, yet without being limited thereto.

Examples :

1.: Prostate Cancer Early Detection program (Innsbruck) Materials and Methods

Blood serum sample collection and processing

Serum samples were obtained from the PCa serum archive in Innsbruck, Austria. The serum sample procurement, data manage ^¬ ment and blood collection protocols were approved by the local Ethical Review Board. Blood samples from patients diagnosed with PCa were obtained from the Prostate Cancer Early Detection pro ^¬ gram open to the general public in Tyrol (Bartsch et al . , BJU International 101 (2008), 809-816) with informed consent of sam ^¬ ple donors. Blood samples were drawn by venous puncture using Sarstedt 9 ml z-gel serum monovettes, serum was obtained by cen- trifugation (4 min, 1800g) and frozen in 2ml cryovials (Simport) at -80°C. At first use the sera were distributed into aliquots to minimize freeze-thaw cycles.

Patient and control cohorts Serum samples were obtained prior to treatment from PCa pa ^¬ tients who underwent radical prostatectomy after cancer diagno ^¬ sis. The inclusion criteria for PCa patients required the histo- pathological assessment of radical prostatectomy specimens. Tu ^¬ mour patients with progression after radical prostatectomy were defined as patients in follow-up control with rising PSA serum levels (increase in at least two consecutive measurements from 0 to above 0,2 ng/ml) or patients who's PSA nadir (i.e. the lowest PSA reading achieved after any treatment for PC; a PSA nadir of 0.1 ng/ml or lower indicates undetectable PSA) was >0 after rad ^¬ ical prostatectomy and increased thereafter. Clinical data of the patients were retrieved from the clinical databases as well as patients' health records.

General Analytics:

Sample preparation and metabolomic analyses were performed at BIOCRATES Life Sciences AG, Innsbruck, Austria. A multi- parametric, highly robust, sensitive and high-throughput target ^¬ ed metabolomic platform was used consisting of flow injection analysis (FIA) -MS/MS and LC-MS/MS methods for the simultaneous quantification of a broad range of endogenous intermediates namely from the panel disclosed in table 1. All procedures (sam ^¬ ple handling, analytics) were performed by co-workers blinded to the groups .

Plasma homogenization

Plasma samples were prepared by standard procedures and stored at (-70°C) . To enable analysis of all samples simultane ^¬ ously within one batch, samples were thawed on ice (1 h) on the day of analysis and centrifuged at 18000 g at 2°C for 5 min. All tubes were prepared with 0.001% BHT (butylated hydroxytoluene ; Sigma-Aldrich, Vienna, Austria) to prevent artificial formation of prostaglandins caused by autooxidation.

Acylcarnitines , Sphingomyelins, Hexoses, Glycerophospholipids (FIA-MS/MS)

To determine the concentration of acylcarnitines, sphingomy ^¬ elins and glycerophospholipids in brain homogenates and in plas ^¬ ma, the AbsolutelDQ kit pl50 (Biocrates Life Sciences AG) was prepared as described in the manufacturer's protocol. In brief, 10 yL of brain homogenate was added to the center of the filter on the upper 96-well kit plate, and the samples were dried using a nitrogen evaporator (VLM Laboratories) . Subsequently, 20 yL of a 5 % solution of phenyl-isothiocyanate was added for derivati- zation. After incubation, the filter spots were dried again using an evaporator. The metabolites were extracted using 300 yL of a 5 mM ammonium acetate solution in methanol. The extracts were obtained by centrifugation into the lower 96-deep well plate followed by a dilution step with 600 yL of kit MS running solvent. Mass spectrometric analysis was performed on an API4000 QTrap® tandem mass spectrometry instrument (Applied Biosys ^¬ tems/MDS Analytical Technologies) equipped with an electro-spray ionization (ESI) -source using the analysis acquisition method as provided in the AbsolutelDQ kit. The standard FIA-MS/MS method was applied for all measurements with two subsequent 20 yL in ^¬ jections (one for positive and one for negative mode analysis) . Multiple reaction monitoring (MRM) detection was used for quantification applying the spectra parsing algorithm integrated into the MetlQ software (Biocrates Life Sciences AG) . Concentra ^¬ tion values for 148 metabolites (all analytes determined with the metabolomics kit besides of the amino acids, which were de ^¬ termined by a different method) obtained by internal calibration were exported for comprehensive statistical analysis.

Amino acids, Biogenic amines (LC-MS/MS)

Amino acids and biogenic amines were quantitatively analysed by reversed phase LC-MS/MS to obtain chromatographic separation of isobaric (same MRM ion pairs) metabolites for individual quantitation performed by external calibration and by use of in ^¬ ternal standards. 10 yL sample volume (plasma, brain homogenate) is required for the analysis using the following sample prepara ^¬ tion procedure. Samples were added on filter spots placed in a 96-solvinert well plate (internal standards were placed and dried down under nitrogen before) , fixed above a 96 deep well plate (capture plate) . 20 yL of 5% phenyl-isothiocyanate deri- vatization reagent was added. The derivatized samples were ex ^¬ tracted after incubation by aqueous methanol into the capture plate. Sample extracts were analyzed by LC-ESI-MS/MS in positive MRM detection mode with an API4000 QTrap® tandem mass spectrome ^¬ try instrument (Applied Biosystems/MDS Analytical Technologies) . The analyzed individual metabolite concentrations (Analyst 1.4.2 software, Applied Biosystems) were exported for comprehensive statistical analysis.

Bile acids (LC-MS/MS)

A highly selective reversed phase LC-MS/MS analysis method in negative MRM detection mode was applied to determine the con ^¬ centration of bile acids in plasma samples. Samples were ex ^¬ tracted via dried filter spot technique in 96 well plate format, which is well suitable for high throughput analysis. For highly accurate quantitation internal standards and external calibra ^¬ tion were applied. In brief, internal standards and 20 yL sample volume placed onto filter spots were extracted and simultaneous ^¬ ly protein precipitated with aqueous methanol. These sample ex ^¬ tracts were measured by LC-ESI-MS/MS with an API4000 QTrap® tan ^¬ dem mass spectrometry instrument (Applied Biosystems/MDS Analyt ^¬ ical Technologies) . Data of bile acids were quantified with Ana ^¬ lyst 1.4.2 software (Applied Biosystems) and finally exported for comprehensive statistical analysis.

Prostanoids, oxidised fatty acids (LC-MS/MS)

Prostanoids - a term summarizing prostaglandins (PG) , throm ^¬ boxanes (TX) and prostacylines - and oxidised fatty acid metabo ^¬ lites were analyzed in plasma extracts by LC-ESI-MS/MS (Unter- wurzacher et al . Clin Chem Lab Med 46 (11) (2008), 1589-1597) and in brain homogenate extracts by online solid phase extrac ^¬ tion (SPE) -LC-MS/MS with an API4000 QTrap® tandem mass spectrometry instrument (Applied Biosystems/MDS Analytical Technologies) in negative MRM detection mode. The sample preparation was the same for both, plasma and brain homogenates. In brief, filter spots in a 96 well plate were spiked with internal standard; 20 yL of plasma or tissue homogenates were added and extracted with aqueous methanol, the individual extracts then were analysed. Data of prostanoids and oxidised fatty acids were quantified with Analyst 1.4.2 software (Applied Biosystems) and finally ex ^¬ ported for statistical analysis.

Oxysterols

Oxysterols are determined after extraction and saponifica ^¬ tion by HPLC-Tandem mass spectrometer (HPLC-API-MS/MS) in posi- tive detection mode using Multiple Reaction Mode (MRM) .

Samples (20 yL) , calibrators and internal standard mixture were placed into a capture plate and were protein precipitated in the first step by means of addition of 200 yL acetonitrile and centrifugation . 180 yL of the appropriate supernatants were transferred on a new filter plate with 7 mm filter spots, dried down, hydrolysed with 0.35 M KOH in 95 % Ethanol and after wash ^¬ ing steps extracted with 100 yL aqueous MeOH. An aliquot of the extracted sample is injected onto the HPLC-MS/MS system. Chroma ^¬ tographic separation and detection is performed by using a Zorbax Eclipse XDB C18, 150 x 2.0 mm, 3.5 ym HPLC-Column at a flow rate of 0.3 mL/min followed by electrospray ionization on the API4000/QTRAP4000 tandem mass spectrometer. For the quanti ^¬ tation the Analyst Quantitation software from Applied Bioystems was used.

Energy metabolism (Organic Acids) (LC-MS/MS)

For the quantitative analysis of energy metabolism interme ^¬ diates (glycolysis, citrate cycle, pentose phosphate pathway, urea cycle) hdyrophilic interaction liquid chromatography (HILIC) -ESI-MS/MS method in highly selective negative MRM detec ^¬ tion mode was used. The MRM detection was performed using an API4000 QTrap® tandem mass spectrometry instrument (Applied Bio- systems/MDS Analytical Technologies) . 20 yL sample volume (plas ^¬ ma, brain homogenate) was protein precipitated and extracted simultaneously with aqueous methanol in a 96 well plate format. Internal standards (ratio external to internal standard) and ex ^¬ ternal calibration were used for highly accurate quantitation. Data were quantified with Analyst 1.4.2 software (Applied Bio- systems) and finally exported for statistical analysis.

Table 1: investigated metabolites

Name Family Common me

CO Ac. . Ca. Carnitine (free)

CIO Ac. . Ca. Decanoylcarnitine [Caprylcarnitine] ( Fumarylcarnitine )

CIO : 1 Ac. . Ca. Decenoylcarnitine

CIO : 2 Ac. . Ca. Decadienoylcarnitine

C12 Ac. . Ca. Dodecanoylcarnitine [Laurylcarnitine ]

C12- ^■ DC Ac. . Ca. Dodecanedioylcarnitine

C12 : 1 Ac. . Ca. Dodecenoylcarnitine

C14 Ac. . Ca. Tetradecanoylcarnitine

C14 : 1 Ac. . Ca. Tetradecenoylcarnitine [Myristoleylcarnitine ]

C14 : 1-OH Ac. . Ca. 3-Hydroxytetradecenoylcarnitine [ 3-Hydroxymyristoleylcarnitine ]

C14 : 2 Ac. . Ca. Tetradecadienoylcarnitine

C14 : 2-OH Ac. . Ca. 3-Hydroxytetradecadienoylcarnitine

C16 Ac. . Ca. Hexadecanoylcarnitine [ Palmitoy1carnitine ]

C16- ^■ OH Ac. . Ca. 3-Hydroxyhexadecanolycarnitine [ 3-Hydroxypalmitoy1carnitine ] C16: 1 Ac Ca Hexadecenoylcarnitine [ Palmitoleylcarnitine ]

016 : 1-OH Ac Ca 3-Hydroxyhexadecenoylcarnitine [3-Hydroxypalmitoleylcarnitine ]

C16:2 Ac Ca Hexadecadienoylcarnitine

C16 : 2-OH Ac Ca 3-Hydroxyhexadecadienoylcarnitine

C18 Ac Ca Octadecanoylcarnitine [ Stearylcarnitine ]

C18 : 1 Ac Ca Octadecenoylcarnitine [Oleylcarnitine ]

C18 : 1-OH Ac Ca 3-Hydroxyoctadecenoylcarnitine [ 3-Hydroxyoleylcarnitine ]

C18 : 2 Ac Ca Octadecadienoylcarnitine [Linoleylcarnitine ]

C2 Ac Ca Acetylcarnitine

C3 Ac Ca Propionylcarnitine

C3-0H Ac Ca Hydroxypropionylcarnitine

C3: 1 Ac Ca Propenoylcarnitine

04 Ac Ca Butyrylcarnitine / Isobutyrylcarnitine

C4-0H (C3 -DC) Ac Ca 3-Hydroxybutyrylcarnitine / Malonylcarnitine

C4 : 1 Ac Ca Butenoylcarnitine

C5 Ac Ca Isovalerylcarnitine / 2-Methylbutyrylcarnitine / Valerylcarnitine

C5-DC (C6 -OH) Ac Ca Glutarylcarnitine / Hydroxycaproylcarnitine

C5-M-DC Ac Ca Methylglutarylcarnitine

C5-0H (CS- DC-M) Ac Ca 3-Hydroxyisovalerylcarnitine / 3-Hydroxy-2-methylbutyryl

CS : 1 Ac Ca Tiglylcarnitine / 3-Methyl-crotonylcarnitine

05 : 1-DC Ac Ca Tiglylcarnitme / 3-Methy1-crotonylcarnitine

06 (04:1- DC) Ac Ca Hexanoylcarnitine [Caproylcarnitine]

06: 1 Ac Ca Hexenoylcarnitine

C7-DC Ac Ca Pimelylcarnitine

08 Ac Ca Octanoylcarnitine [Caprylylcarnitine ]

08 : 1 Ac Ca Octenoylcarnitine

09 Ac Ca Nonoylcarnitme [Pelargonylcarnitine

SM (OH) 014 : 1 Sphingomyelin with acyi residue sum (OH) 014 : : 1

SM (OH) 016: 1 Sphingomyelin with acyi residue sum (OH) C16: : 1

SM (OH) C22 : 1 Sphingomyelin with acyi residue sum (OH) 022 : : 1

SM (OH) 022 : 2 Sphingomyelin with acyi residue sum (OH) C22 : : 2

SM (OH) C24 : 1 Sphingomyelin with acyi residue sum (OH) 024 : : 1

SM 026 : 0 Sphingomyelin with acyi residue sum 026: 0

SM 026 : 1 Sphingomyelin with acyi residue sum C26: 1

PC aa C24 Phosphatidylcholine with diacyl residue sum 024

PC aa C26 Phosphatidylcholine with diacyl residue sum C26

PC aa C28 Phosphatidylcholine with diacyl residue sum 028

PC aa C32 Phosphatidylcholine with diacyl residue sum C32

PC aa C34 Phosphatidylcholine with diacyl residue sum 034

PC aa C36 Phosphatidylcholine with diacyl residue sum C36

PC aa C38 Phosphatidylcholine with diacyl residue sum 038 0

PC aa C40 Phosphatidylcholine with diacyl residue sum C40 1

PC aa C40 Phosphatidylcholine with diacyl residue sum C40 2

PC aa C40 Phosphatidylcholine with diacyl residue sum C40 3

PC aa C42 Phosphatidylcholine with diacyl residue sum 042 0

PC aa C42 Phosphatidylcholine with diacyl residue sum C42 1

PC aa C42 Phosphatidylcholine with diacyl residue sum 042 2

PC aa C42 Phosphatidylcholine with diacyl residue sum C42 4

PC aa C42 Phosphatidylcholine with diacyl residue sum 042 5

PC aa C42 Phosphatidylcholine with diacyl residue sum C42 6

PC ae C30 Phosphatidylcholine with acyl lkyl residue sum C30

PC ae C30 Phosphatidylcholine with acyl lkyl residue sum C30

PC ae C30 Phosphatidylcholine with acyl lkyl residue sum C30

PC ae C32 Phosphatidylcholine with acyl lkyl residue sum C32

PC ae C36 Phosphatidylcholine with acyl lkyl residue sum 036

PC ae C38 Phosphatidylcholine with acyl lkyl residue sum C38

PC ae C40 Phosphatidylcholine with acyl lkyl residue sum C40

PC ae C40 Phosphatidylcholine with acyl lkyl residue sum C40

PC ae C40 Phosphatidylcholine with acyl lkyl residue sum C40

PC ae C40 Phosphatidylcholine with acyl lkyl residue sum C40

PC ae C40 Phosphatidylcholine with acyl lkyl residue sum C40

PC ae C40 Phosphatidylcholine with acyl lkyl residue sum C40

PC ae C42 0 Phosphatidylcholine with acyl lkyl residue sum 042

PC ae C42 1 Phosphatidylcholine with acyl lkyl residue sum C42

PC ae C42 2 Phosphatidylcholine with acyl lkyl residue sum 042

PC ae C42 3 Phosphatidylcholine with acyl lkyl residue sum C42

PC ae C42 4 Phosphatidylcholine with acyl lkyl residue sum 042

PC ae C42 5 Phosphatidylcholine with acyl lkyl residue sum C42

PC ae C44 3 Phosphatidylcholine with acyl lkyl residue sum 044

PC ae C44 4 Phosphatidylcholine with acyl lkyl residue sum C44

PC ae C44 5 Phosphatidylcholine with acyl lkyl residue sum 044

PC ae C44 6 Phosphatidylcholine with acyl lkyl residue sum C44

lysoPC C14 Lysophosphatidylcholine with acyl residue sum 014 0

lysoPC C16 Lysophosphatidylcholine with acyl residue sum C16 1

lysoPC C17 Lysophosphatidylcholine with acyl residue sum 017

lysoPC C20 Lysophosphatidylcholine with acyl residue sum C20

lysoPC C24 Lysophosphatidylcholine with acyl residue sum 024

lysoPC C26 Lysophosphatidylcholine with acyl residue sum C26

lysoPC C26 Lysophosphatidylcholine with acyl residue sum 026

lysoPC C28 Lysophosphatidylcholine with acyl residue sum C28

lysoPC C28 Lysophosphatidylcholine with acyl residue sum 028

lysoPC C6: 0 Lysophosphatidylcholine with acyl residue sum C6: Gly Am. Ac. Glycine

Ala Am. Ac. Alanine

Ser Am. Ac. Serine

Pro Am. Ac. Proline

Val Am. Ac. Valine

Thr Am. Ac. Threonine

Xle Am. Ac. Leucine + Isoleucine

Leu Am. Ac. Leucine

He Am. Ac. Isoleucine

Asn Am. Ac. Asparagine

Asp Am. Ac. Aspartate

Gin Am. Ac. Glutamine

Glu Am. Ac. Glutamate

Met Am. Ac. Methionine

His Am. Ac. Histidine

Phe Am. Ac. Phenylalanine

Arg Am. Ac. Arginine

Cit Am. Ac. Citrulline

Tyr Am. Ac. Tyrosine

Trp Am. Ac. Tryptophan

Orn Am. Ac. Ornithine

Lys Am. Ac. Lysine

ADMA B.Am. asymmetrical Dimethylarginin

totalDMA B.Am. Total dimethylarginine : sum ADMA + SDMA

Met-SO Am. Ac. Methionine-Sulfoxide

Kynurenine B.Am. Kynurenine

Putrescine B.Am. Putrescine

Spermidine B.Am. Spermidine

Spermine B.Am. Spermine

Creatinine B.Am. Creatinine

13S-HODE P.G. 13 (S) -hydroxy-9Z, 1 lE-octadecadienoic acid

12S-HETE P.G. 12 (S) -hydroxy-5Z, 8Z, 10E, 14Z-eicosatetraenoic acid

15S-HETE P.G. 15 (S) -hydroxy-5Z, 8Z, HZ, 13E-eicosatetraenoic acid

LTB4 P.G. Leukotriene B4

DHA P.G. Docosahexaenoic acid

PGE2 P.G. Prostaglandin E2

PGD2 P.G. Prostaglandin D2

AA P.G. Arachidonic acid

Lac En. Met. Lactate

Sue En. Met. Succinic acid (succite)

Hex En. Met. Hexose pool

22ROHC O.St. 22-R-Hydroxycholesterol

24SOHC O.St. 24-S-Hydroxycholesterol

250HC O.St. 25-Hydroxycholesterol

270HC O.St. 27-Hydroxycholesterol

3B, 5a, 6BTHC O.St. 3β , 5α, 6B-Trihydroxycholestan

7aOHC O.St. 7a-Hydroxycholesterol

7KC O.St. 7-Ketocholesterol

5a, 6a, EPC O.St. 5a, 6a-Epoxycholesterol

4BOHC O.St. 4B-Hydroxycholesterol

Desmosterol O.St. Desmosterol

7DHC O.St. 7-Dehydrocholesterol (Vitamin D3)

Lanosterol O.St. Lanosterol

20aOHC O.St. 2 Oa-Hydroxycholesterol

22SOHC O.St. 22S-Hydroxycholesterol

24, 25EC O.St. 24, 25-Epoxycholesterol

5B, 6B, EC O.St. 5β , 6B-Epoxycholesterol

5a, 6a, EC O.St. 5a, 6a-Epoxycholesterol

Cholestenone O.St. Cholestenone

PE C16 GP.L. Lysophosphatidylethanolamine with acyl residue sum C16 PE C18 GP.L. Lysophosphatidylethanolamine with acyl residue sum C18 PE C18 GP.L. Lysophosphatidylethanolamine with acyl residue sum C18 PE C18 GP.L. Lysophosphatidylethanolamine with acyl residue sum C18 PE C20 GP.L. Lysophosphatidylethanolamine with acyl residue sum C20 PE C22 GP.L. Lysophosphatidylethanolamine with acyl residue sum C22 PE C22 GP.L. Lysophosphatidylethanolamine with acyl residue sum C22 PE C22 GP.L. Lysophosphatidylethanolamine with acyl residue sum C22 PE C18 GP.L. Lysophosphatidylethanolamine with alkyl residue sum CI 8 : 0 PG C14 GP.L. Lysophosphatidylglycerol with alkyl residue sum C14:2 PE C20 0 GP.L. Phosphatidylethanolamine with diacyl residue sum C20: PE C22 2 GP.L. Phosphatidylethanolamine with diacyl residue sum C22: PE C26 4 GP.L. Phosphatidylethanolamine with diacyl residue sum C2 PE C28 4 GP.L. Phosphatidylethanolamine with diacyl residue sum C2 PE C28 5 GP.L. Phosphatidylethanolamine with diacyl residue sum C2 PE C34 0 GP.L. Phosphatidylethanolamine with diacyl residue sum C3 PE C34 1 GP.L. Phosphatidylethanolamine with diacyl residue sum C3 PE C34 2 GP.L. Phosphatidylethanolamine with diacyl residue sum C3 PE C34 3 GP.L. Phosphatidylethanolamine with diacyl residue sum C3 PE C36 0 GP.L. Phosphatidylethanolamine with diacyl residue sum C3 PE C36 1 GP.L. Phosphatidylethanolamine with diacyl residue sum C3 PE C36 2 GP.L. Phosphatidylethanolamine with diacyl residue sum C3 PE C36 3 GP.L. Phosphatidylethanolamine with diacyl residue sum C3 PE aa C36 4 GP L. Phosphatidylethanolamine with diacyl residue sum C36 4

PE aa C36 5 GP L. Phosphatidylethanolamine with diacyl residue sum C36 5

PE aa C38 0 GP L. Phosphatidylethanolamine with diacyl residue sum C38 0

PE aa C38 1 GP L. Phosphatidylethanolamine with diacyl residue sum C38 1

PE aa C38 2 GP L. Phosphatidylethanolamine with diacyl residue sum C38 2

PE aa C38 3 GP L. Phosphatidylethanolamine with diacyl residue sum C38 3

PE aa C38 4 GP L. Phosphatidylethanolamine with diacyl residue sum C38 4

PE aa C38 5 GP L. Phosphatidylethanolamine with diacyl residue sum C38 5

PE aa C38 6 GP L. Phosphatidylethanolamine with diacyl residue sum C38 6

PE aa C38 7 GP L. Phosphatidylethanolamine with diacyl residue sum C38 7

PE aa C40 2 GP L. Phosphatidylethanolamine with diacyl residue sum C40 2

PE aa C40 3 GP L. Phosphatidylethanolamine with diacyl residue sum C40 3

PE aa C40 4 GP L. Phosphatidylethanolamine with diacyl residue sum C40 4

PE aa C40 5 GP L. Phosphatidylethanolamine with diacyl residue sum C40 5

PE aa C40 6 GP L. Phosphatidylethanolamine with diacyl residue sum C40 6

PE aa C40 7 GP L. Phosphatidylethanolamine with diacyl residue sum C40 7

PE aa C48 1 GP L. Phosphatidylethanolamine with diacyl residue sum C48 1

PE ae C34 1 GP L. Phosphatidylethanolamine with acyl-alkyl residue sum C34

PE ae C34 2 GP L. Phosphatidylethanolamine with acyl-alkyl residue sum C34

PE ae C34 3 GP L. Phosphatidylethanolamine with acyl-alkyl residue sum C34

PE ae C36 1 GP L. Phosphatidylethanolamine with acyl-alkyl residue sum C36

PE ae C36 2 GP L. Phosphatidylethanolamine with acyl-alkyl residue sum C36

PE ae C36 3 GP L. Phosphatidylethanolamine with acyl-alkyl residue sum C36

PE ae C36 4 GP L. Phosphatidylethanolamine with acyl-alkyl residue sum C36

PE ae C36 5 GP L. Phosphatidylethanolamine with acyl-alkyl residue sum C36

PE ae C38 1 GP L. Phosphatidylethanolamine with acyl-alkyl residue sum C38

PE ae C38 2 GP L. Phosphatidylethanolamine with acyl-alkyl residue sum C38

PE ae C38 3 GP L. Phosphatidylethanolamine with acyl-alkyl residue sum C38

PE ae C38 4 GP L. Phosphatidylethanolamine with acyl-alkyl residue sum C38

PE ae C38 5 GP L. Phosphatidylethanolamine with acyl-alkyl residue sum C38

PE ae C38 6 GP L. Phosphatidylethanolamine with acyl-alkyl residue sum C38

PE ae C40 1 GP L. Phosphatidylethanolamine with acyl-alkyl residue sum C40

PE ae C40 2 GP L. Phosphatidylethanolamine with acyl-alkyl residue sum C40

PE ae C40 3 GP L. Phosphatidylethanolamine with acyl-alkyl residue sum C40

PE ae C40 4 GP L. Phosphatidylethanolamine with acyl-alkyl residue sum C40

PE ae C40 5 GP L. Phosphatidylethanolamine with acyl-alkyl residue sum C40

PE ae C40 6 GP L. Phosphatidylethanolamine with acyl-alkyl residue sum C40

PE ae C42 1 GP L. Phosphatidylethanolamine with acyl-alkyl residue sum C42

PE ae C42 2 GP L. Phosphatidylethanolamine with acyl-alkyl residue sum C42

PE ae C46 5 GP L. Phosphatidylethanolamine with acyl-alkyl residue sum C46

PE ae C46 6 GP L. Phosphatidylethanolamine with acyl-alkyl residue sum C46

PG aa C30 0 GP L. Phosphatidylglycerol with diacyl residue sum C30 0

PG aa C32 0 GP L. Phosphatidylglycerol with diacyl residue sum C32 0

PG aa C32 1 GP L. Phosphatidylglycerol with diacyl residue sum C32 1

PG aa C33 6? GP L. Phosphatidylglycerol with diacyl residue sum C33 6?

PG aa C34 0 GP L. Phosphatidylglycerol with diacyl residue sum C34 0

PG aa C34 1 GP L. Phosphatidylglycerol with diacyl residue sum C34 1

PG aa C34 2 GP L. Phosphatidylglycerol with diacyl residue sum C34 2

PG aa C34 3 GP L. Phosphatidylglycerol with diacyl residue sum C34 3

PG aa C36 0 GP L. Phosphatidylglycerol with diacyl residue sum C36 0

PG aa C36 1 GP L. Phosphatidylglycerol with diacyl residue sum C36 1

PG aa C36 2 GP L. Phosphatidylglycerol with diacyl residue sum C36 2

PG aa C36 3 GP L. Phosphatidylglycerol with diacyl residue sum C36 3

PG aa C36 4 GP L. Phosphatidylglycerol with diacyl residue sum C36 4

PG aa C38 5 GP L. Phosphatidylglycerol with diacyl residue sum C38 5

PG ae C32 0 GP L. Phosphatidylglycerol with acyl -alkyl residue sum C32 0

PG ae C34 0 GP L. Phosphatidylglycerol with acyl -alkyl residue sum C34 0

PG ae C34 1 GP L. Phosphatidylglycerol with acyl -alkyl residue sum C34 1

PG ae C36 1 GP L. Phosphatidylglycerol with acyl -alkyl residue sum C36 1

PS aa C34 1 GP L. Phosphatidylserine with diacyl residue sum C34 : 1

PS aa C34 2 GP L. Phosphatidylserine with diacyl residue sum C34 : 2

PS aa C36 0 GP L. Phosphatidylserine with diacyl residue sum C36: 0

PS aa C36 1 GP L. Phosphatidylserine with diacyl residue sum C36: 1

PS aa C36 2 GP L. Phosphatidylserine with diacyl residue sum C36:2

PS aa C36 3 GP L. Phosphatidylserine with diacyl residue sum C36:3

PS aa C36 4 GP L. Phosphatidylserine with diacyl residue sum C36:4

PS aa C38 1 GP L. Phosphatidylserine with diacyl residue sum C38 : 1

PS aa C38 2 GP L. Phosphatidylserine with diacyl residue sum C38 : 2

PS aa C38 3 GP L. Phosphatidylserine with diacyl residue sum C38 : 3

PS aa C38 4 GP L. Phosphatidylserine with diacyl residue sum C38 : 4

PS aa C38 5 GP L. Phosphatidylserine with diacyl residue sum C38 : 5

PS aa C40 1 GP L. Phosphatidylserine with diacyl residue sum C40 : 1

PS aa C40 2 GP L. Phosphatidylserine with diacyl residue sum C40 : 2

PS aa C40 3 GP L. Phosphatidylserine with diacyl residue sum C40 : 3

PS aa C40 4 GP L. Phosphatidylserine with diacyl residue sum C40 : 4

PS aa C40 5 GP L. Phosphatidylserine with diacyl residue sum C40 : 5

PS aa C40 6 GP L. Phosphatidylserine with diacyl residue sum C40 : 6

PS aa C40 7 GP L. Phosphatidylserine with diacyl residue sum C40 : 7

PS aa C42 1 GP L. Phosphatidylserine with diacyl residue sum C42 : 1

PS aa C42 2 GP L. Phosphatidylserine with diacyl residue sum C42 : 2

PS aa C42 4 GP L. Phosphatidylserine with diacyl residue sum C42 : 4

PS aa C42 5 GP L. Phosphatidylserine with diacyl residue sum C42 : 5 PS ae C34 2 GP.L Phosphatidylsenne with yl-alkyl esidue sum C34

PS ae C36 1 GP.L Phosphatidylserine with yl-alkyl esidue sum C36

PS ae C36 2 GP.L Phosphatidylserine with yl-alkyl esidue sum C36

PS ae C38 4 GP.L Phosphatidylserine with yl-alkyl esidue sum C38

SM C14 : 0 S.L. Sphingomyelin with acyi sidue sum C14 : 0

SM C16 : 0 S.L. Sphingomyelin with acyi sidue sum C16

SM C16 : 1 S.L. Sphingomyelin with acyi sidue sum C16

SM C17 : 0 S.L. Sphingomyelin with acyi sidue sum C17

SM C18 : 0 S.L. Sphingomyelin with acyi sidue sum C18

SM C18 : 1 S.L. Sphingomyelin with acyi sidue sum C18

SM C19 : 0 S.L. Sphingomyelin with acyi sidue sum C19

SM C19 : 1 S.L. Sphingomyelin with acyi sidue sum C19

SM C19 : 2 S.L. Sphingomyelin with acyi sidue sum C19:2

SM C20 : 0 S.L. Sphingomyelin with acyi sidue sum C20 : 0

SM C20 : 1 S.L. Sphingomyelin with acyi sidue sum C20 : 1

SM C20 : 2 S.L. Sphingomyelin with acyi sidue sum C20:2

SM C21 : 0 S.L. Sphingomyelin with acyi sidue sum C21 : 0

SM C21 : 1 S.L. Sphingomyelin with acyi sidue sum C21 : 1

SM C21 : 2 S.L. Sphingomyelin with acyi sidue sum C21:2

SM C21 : 3 S.L. Sphingomyelin with acyi sidue sum C21:3

SM C22 : 0 S.L. Sphingomyelin with acyi sidue sum C22 : 0

SM C22 : 1 S.L. Sphingomyelin with acyi sidue sum C22 : 1

SM C22 : 2 S.L. Sphingomyelin with acyi sidue sum C22:2

SM C22 : 3 S.L. Sphingomyelin with acyi sidue sum C22:3

SM C23 : 0 S.L. Sphingomyelin with acyi sidue sum C23:0

SM C23 : 1 S.L. Sphingomyelin with acyi sidue sum C23:l

SM C23 : 2 S.L. Sphingomyelin with acyi sidue sum C23:2

SM C23 : 3 S.L. Sphingomyelin with acyi sidue sum C23:3

SM C24 : 0 S.L. Sphingomyelin with acyi sidue sum C24 : 0

SM C24 : 1 S.L. Sphingomyelin with acyi sidue sum C24 : 1

SM C24 : 2 S.L. Sphingomyelin with acyi sidue sum C24:2

SM C24 : 3 S.L. Sphingomyelin with acyi sidue sum C24:3

SM C24 : 4 S.L. Sphingomyelin with acyi sidue sum C24:4

SM C26 : 3 S.L. Sphingomyelin with acyi sidue sum C26:3

SM C26 : 4 S.L. Sphingomyelin with acyi sidue sum C26:4

SM C3: 0 S.L. Sphingomyelin with acyi sidue sum C3: 0

lysoPC a C16 0 GP.L Lysophosphatidylcholine with acyl residue sum C16 0 lysoPC a C18 0 GP.L Lysophosphatidylcholine with acyl sidue sum C18 0 lysoPC a C18 1 GP.L Lysophosphatidylcholine with acyl sidue sum C18 1 lysoPC a C18 2 GP.L Lysophosphatidylcholine with acyl sidue sum C18 2 lysoPC a C20 4 GP.L Lysophosphatidylcholine with acyl sidue sum C20 4

PC e C18 : 0 GP.L Lysophosphatidylcholine with alkyl esidue sum CI 3 :

PC aa C30 0 GP.L Phosphatidylcholine with diacyl residue sum C30 0

PC aa C30 1 GP.L Phosphatidylcholine with diacyl residue sum C30 1

PC aa C30 2 GP.L Phosphatidylcholine with diacyl residue sum C30 2

PC aa C32 0 GP.L Phosphatidylcholine with diacyl residue sum C32 0

PC aa C32 1 GP.L Phosphatidylcholine with diacyl residue sum C32 1

PC aa C32 2 GP.L Phosphatidylcholine with diacyl residue sum C32 2

PC aa C34 0 GP.L Phosphatidylcholine with diacyl residue sum C34 0

PC aa C34 1 GP.L Phosphatidylcholine with diacyl residue sum C34 1

PC aa C34 2 GP.L Phosphatidylcholine with diacyl residue sum C34 2

PC aa C34 3 GP.L Phosphatidylcholine with diacyl residue sum C34 3

PC aa C36 0 GP.L Phosphatidylcholine with diacyl residue sum C36 0

PC aa C36 1 GP.L Phosphatidylcholine with diacyl residue sum C36 1

PC aa C36 2 GP.L Phosphatidylcholine with diacyl residue sum C36 2

PC aa C36 3 GP.L Phosphatidylcholine with diacyl residue sum C36 3

PC aa C36 4 GP.L Phosphatidylcholine with diacyl residue sum C36 4

PC aa C36 5 GP.L Phosphatidylcholine with diacyl residue sum C36 5

PC aa C38 1 GP.L Phosphatidylcholine with diacyl residue sum C38 1

PC aa C38 2 GP.L Phosphatidylcholine with diacyl residue sum C38 2

PC aa C38 3 GP.L Phosphatidylcholine with diacyl residue sum C38 3

PC aa C38 4 GP.L Phosphatidylcholine with diacyl residue sum C38 4

PC aa C38 5 GP.L Phosphatidylcholine with diacyl residue sum C38 5

PC aa C38 6 GP.L Phosphatidylcholine with diacyl residue sum C38 6

PC aa C40 4 GP.L Phosphatidylcholine with diacyl residue sum C40 4

PC aa C40 5 GP.L Phosphatidylcholine with diacyl residue sum C40 5

PC aa C40 6 GP.L Phosphatidylcholine with diacyl residue sum C40 6

PC aa C40 7 GP.L Phosphatidylcholine with diacyl residue sum C40 7

PC aa C40 8 GP.L Phosphatidylcholine with diacyl residue sum C40 8

PC ae C32 0 GP.L Phosphatidylcholine with cyl lkyl residue sum C3 2 : 0

PC ae C32 1 GP.L Phosphatidylcholine with cyl lkyl residue sum C32 : 1

PC ae C32 6 GP.L Phosphatidylcholine with cyl lkyl residue sum C32 : 6

PC ae C34 0 GP.L Phosphatidylcholine with cyl lkyl residue sum C34 : 0

PC ae C34 1 GP.L Phosphatidylcholine with cyl lkyl residue sum C34 : 1

PC ae C34 2 GP.L Phosphatidylcholine with cyl lkyl residue sum C34 : 2

PC ae C34 3 GP.L Phosphatidylcholine with cyl lkyl residue sum C34 : 3

PC ae C34 6 GP.L Phosphatidylcholine with cyl lkyl residue sum C34 : 6

PC ae C36 1 GP.L Phosphatidylcholine with cyl lkyl residue sum C36:1

PC ae C36 2 GP.L Phosphatidylcholine with cyl lkyl residue sum C36:2

PC ae C36 3 GP.L Phosphatidylcholine with cyl lkyl residue sum C36:3

PC ae C36 4 GP.L Phosphatidylcholine with cyl lkyl residue sum C36:4

PC ae C36 5 GP.L Phosphatidylcholine with cyl lkyl residue sum C36:5 PC ae C38 1 GP.L Phosphatidylcholine with acyl -alkyl residue sum C38 1

PC ae C38 2 GP.L Phosphatidylcholine with acyl -alkyl residue sum C38 2

PC ae C38 3 GP.L Phosphatidylcholine with acyl -alkyl residue sum C38 3

PC ae C38 4 GP.L Phosphatidylcholine with acyl -alkyl residue sum C38 4

PC ae C38 5 GP.L Phosphatidylcholine with acyl -alkyl residue sum C38 5

PC ae C38 6 GP.L Phosphatidylcholine with acyl -alkyl residue sum C38 6

PC ae C40 5 GP.L Phosphatidylcholine with acyl -alkyl residue sum C40 5

Histamine B.Am Histamine

Serotonin B.Am Serotonin

Phenylethyla B.Am Phenylethylamine

TXB2 P.G. Tromboxane B2

PGF2a P.G. Prostaglandin F2alph

24, 25,EPC 0. St 24, 25-Epoxycholesterol

5B, 6B, EPC 0. St 5β, 6B-Epoxycholesterol

24DHLan 0. St 24-Dihydrolanosterol

SDMA B.Am Symmetrical Dimethylarginine

OH-kynurenine B.Am Hydroxykynurenine

9S-HODE P.G. (±) 9-hydroxy-10E, 12Z -octadecadienoic acid

14 (15) -EpETE P.G. (±) 14 (15) -epoxy-5Z, 8Z, 11Z, 17Z-eicosatetraenoic acid

15S-HpETE P.G. 15 (S)- hydroperoxy-5Z, 8Z, 11Z, 13E eicosatetraenoic acid

5S-HpETE P.G. 5 (S) -hydroperoxy-6E, 8Z, 11Z, 14Z-eicosatetraenoic acid

LTD4 P.G. Leukotriene D4

6-keto-PGFla P.G. 6-keto-Prostaglandin Flalpha

8-isoPGF2a P.G. 8-iso-Prostaglandin F2alpha

Nitrotyrosine B.Am Nitrotyrosine

Dopamine B.Am Dopamine

Sarcosine B.Am

Cholesterol 0. St

3-PG EM 3-Phosphoglycerate

alpha-KGA EM alpha-Ketoglutaric acid

AMP EM Adenosine-5 ' -monophosphate

CAMP EM 3 '-5 '-cyclic Adenosine monophosphate

DHAP+3-PGA EM Dihydroxyacetonephosphate + 3-Phosphoglyceraldehyde

Fum EM Fumaric acid

Glt-6-P EM Gluconate-6-phosphate

Lac EM Lactate

PEP EM Phosphoenolpyruvate

Pyr+OAA EM Pyruvate + Oxaloacetate

Sue EM Succinic acid

Table 1 summarizes analyzed metabolites and respective abbreviations; Glycero-phospholipids are further differentiated with respect to the presence of ester (a) and ether (e) bonds in the glycerol moiety, where two letters (aa, ea, or ee) denote that the first and the second position of the glycerol scaffold are bound to a fatty acid residue, whereas a single letter (a or e) indicates a bond with only one fatty acid residue; e.g. PC_ea_33:l denotes a plasmalogen phosphatidylcholine with 33 carbons in the two fatty acid side chains and a single double bond in one of them.

Statistical analysis

All statistical calculations have performed using R [R] . An- alytes that were detected in at least 50% of the samples are in ^¬ cluded resulting in a final list of 221 metabolites.

The concentrations of metabolites in the 72 samples were de ^¬ termined in two analytical batches. A correction factor calcu ^¬ lated from the independent QC samples is initially applied to circumvent plate discrepancies before statistical analysis.

The metabolic data is left censored due to the threshold of the mass spectrometer data resulting in non detected peak/signals. Due to a combination of metabolic pathway dyna ^¬ mism, complex sample molecular interaction and overall efficiency of the analytical protocol, replacement of missing data by means of a multivariate algorithm are preferred over a naive im ^¬ putation by zero. For any statistical treatment, metabolite missing concentrations are replaced by average value of the 6 closest samples to the one where the measurement is missing [KIM] using the function llslmpute in the R package pcaMethods [PCAM] .

Except for fold change determination, all data analyses are performed on log basis 2 transformed data.

To test for concentration changes between relapse and non progressive patients for each analyte, a linear model is fitted specifying group and analytical batch using the lmFit function in the R package limma (version 2.16.5) [LIMMA] . Coefficients from the linear model are then combined with equal weight to es ^¬ timate moderated statistics related to the difference between the two patient cohorts including p value and log 2 fold changes (i.e. log2FC, in the text) . Resulting significance levels are presented in form of q values after adjustment by the method de ^¬ scribed in Benjamini & Hochberg [FDR] . q value threshold of 0.1 is used to pick analytes undergoing to subsequent selection.

Fold changes correspond to the relative change in percent between the median concentration in the Relapse group and the median concentration in the Non Progressive samples. A selection cut off of 20 percent is chosen.

Sensitivity/specificity properties of a classifier compris ^¬ ing one analyte or a combination of analytes are summarized in term of Area Under the Receiver Operating Characteristic Curve (AUC) . The function colAUC [CATOOLS] too is used for this pur ^¬ pose and AUO0.7 is used as selection threshold.

Regardless direction of change, dedicated function vennDia- gram in R package limma [LIMMA] is employed to display metabo ^¬ lite classification counts under three univariate techniques: q value, fold change and AUC.

Metabolites selected by at least two methods are making up potential marker composite. Their performance markers are as ^¬ sessed by Diagonal Discriminant Analysis [DLDA] with a contrib ^¬ uted sfsmisc library within R [SFSMISC] . Predictive abilities of the models are computed with stratified bootstrap (B=50) strate ^¬ gy were left-out samples are used to compute predicted AUC. Re ^¬ sults are presented in form of mean and 10/90 quantiles. For comparison purposes, same strategy is applied to all metabolites (i.e 221) and to the whole set of metabolites without the ones selected in the composite of markers. For graphing, individual ROC curves from the 50 bootstraps are combined by threshold av ^¬ eraging using the function mc.roc in the package FIEmspro [FIEMS]

Selection of marker combinations

For the set of the biomarkers according to the present in ^¬ vention, parsimonious multi metabolite panels can be used for predicting PCa-relapse. Predictive models using all possible combinations of 2 to 6 markers are built using logistic regres ^¬ sion with the dedicated function multinom in the nnet package [NNET] in the R environment. Each model is then further subject ^¬ ed to backward elimination using the Akaike Information Criterion (AIC) to penalize final models having too many parameters. Table 2 provides preferred list of various individual markers and composites from the proposed set of metabolites according to the present invention. While the invention described herein may comprise composite with adequate other than those shown, combi ^¬ nation of metabolites would show sensitivities greater than 85% and specificities greater than 75%.

Results :

In the present invention, metabolic signatures in serum were identified to predict whether a PCa would lead to relapse or not. Thus the sera used in this study were all obtained during the tumour stages of the patients.

Comparisons were performed to generate the data for >2y and

>3 y:

(a) Relapsed vs non progressive (patients without relapse for less than 2 years were excluded; only patients without re ^¬ lapse for 2 year or more after radical prostatectomy were included, 28 samples in all)

(b) Relapsed vs non progressive (patients without relapse for less than 3 years were excluded; only patients without re ^¬ lapse for 3 year or more after radical prostatectomy were included, 18 samples in all)

Patients were classified as non progressive (relapse free) based on results from follow up examination after radical prostatectomy. The duration of relapse fee (non progressive) for pa ^¬ tients' classification was determined with the last date of fol ^¬ low up examination after treatment. Table 2: markers and (exemplary) marker combinations according to the present invention (> 2 years relapse/no relapse; this was also the total group of markers in the >1 year investigation)

Table 2A 0 Relapse free 2 years and more

Metabolites pval q-value Auc log2FC medFC

PC aa C24:0 2, 88E-05 0, 001327 0, 77 0,42 21,7

PC ae C40:3 4, 61E-08 4, 24E-06 0, 74 0, 54 26,9

PC ae C40:4 0,000831 0,013047 0, 66 0, 32 20,2 lysoPC a C26:0 0,000809 0,013047 0, 75 0,2 11, 6 lysoPC a C6:0 7, 24E-05 0, 002222 0, 74 0,4 38

13S-HODE 0,000249 0, 000497 0, 73 1,24 166

12S-HETE 0,248173 0,310216 0, 71 0,42 119,3

15S-HETE 3, 07E-07 2, 73E-06 0, 77 1,49 165, 5

LTB4 5, 33E-06 1, 78E-05 0, 76 2,19 233

PGE2 0,000177 0,000443 0, 69 0,72 50,1

PGD2 5, 45E-07 2, 73E-06 0, 81 2,72 339, 5

7aOHC 0, 005579 0,022316 0, 67 0,72 66,5

7KC 9, 42E-05 0, 000565 0, 74 0, 66 39,7

5B, 6B, EPC 2, 98E-07 3, 58E-06 0, 79 1,2 180,2

5a, 6a, EPC 0,03128 0, 062559 0, 64 0,42 29,9

4BOHC 0,012236 0, 036707 0, 73 0, 37 31

Table 2A: Avg. relapse free for two years and more

Table 2B: preferred marker combinations > 2y

Table 2B: preferred marker combinations > 2y

Table 3: markers and (exemplary) marker combinations according to the present invention (> 3 years relapse/no relapse)

Table 3A: Relapse free 3 years and more

pval q- -value auc log2FC medFC

PC aa C24:0 0, 001337 o , 037199 0,77 0,43 19,7

PC ae C40:3 8, 79E-05 o , 008091 0,72 0, 52 21,7 lysoPC a C26:0 0, 006707 0, 088149 0,71 0,19 9,8

lysoPC a C6:0 0, 000539 0, 024783 0,73 0,41 38

13S-HODE 0,006151 0, 013242 0,72 1,29 196, 1

12S-HETE 0, 431194 0, 538992 0,74 0,31 119,3 15S-HETE 2, 31E-05 0, 000194 0, 76 1, 52 165, 5

LTB4 0, 000632 0, 002106 0, 75 2, 19 233

PGE2 0, 006621 0, 013242 0, 66 0, 61 38,4

PGD2 3, 89E-05 0, 000194 0, 85 2, 99 366,3

7aOHC 0, 029709 0, 071303 0, 67 0, 72 65, 6

7KC 0, 001064 0, 006385 0, 75 0, 75 39,7

5B, 6B, EPC 1, 8E-07 2, 16E-06 0, 86 1, 46 219,4

5a, 6a, EPC 0, 010567 0, 042268 0, 68 0, 53 35, 7

Table 3B: preferred marker combinations > 3y

Table 3B: preferred marker combinations > 3y p value, q value etc... are general terms that will be under ^¬ stood by anyone in the field. Exact definitions are:

"p value": probability of obtaining a result greater than the observed quantity assuming that this result is obtained by chance in statistical hypothesis testing

"q value": minimum expected false positive rate at which the test may be called significant

"sensitivity": proportion of positive examples which are cor ^¬ rectly identified as positive in a binary classification test "specificity": proportion of negative examples which are cor ^¬ rectly identified as negative in a binary classification test "accuracy" : proportion of true results (both true positives and true negatives) in the population

"AUC" : area under the receiver operating characteristic (ROC) curve. ROC curve is a graphical representation of the relation ^¬ ship of the sensitivity and (1 - specificity) across all possi ^¬ ble threshold values that define the decision boundaries of a binary classifier.

"fold change": Fold change is a mathematical operation describ ^¬ ing how much two groups differ.

"FC(log2)": is defined as the difference between the group means calculated from the log basis 2 transformed data

"medFC" : is defined as the ratio between the group medians mi ^¬ nus 1

"AIC" : Akaike Information Criterion is a measure of the goodness of fit of an estimated statistical model. In the general case, AIC = 2k - 2*ln(L), where k is the number of parameters in the statistical model, and L is the maximized value of the likeli ^¬ hood function for the estimated model.

Those skilled in the art would understand that each assay of the multi anaylte panel needs to undergo fit for purpose assay.

Those skilled in the art would understand that a suitable classification algorithm could also comprise (diagonal) linear or quadratic discriminant analysis (LDA, QDA, DLDA, DQDA) , per- ceptron, shrunken centroids, regularized discriminant analysis (RDA) , random forests (RF) , neural networks (NN) , Bayesian net ^¬ works, hidden Markov models, support vector machines (SVM) , gen ^¬ eralized partial least squares (GPLS) , inductive logic program ^¬ ming (ILP), generalized additive models, gaussian processes, regularized least square regression, recursive partitioning and regression trees, K-nearest neighbour classifiers (K-NN) , fuzzy classifiers, bagging, boosting, and naive Bayes.

2.: Partial Multicentre Follow-up study (Berlin/ Innsbruck) Blood serum sample collection and processing

The serum sample procurement, data management and blood col ^¬ lection protocols were approved by the local Ethical Review Board at Medical University, Innsbruck, Austria and Charite- Universitatsmedizin Berlin, Germany. Blood samples were drawn by venous puncture using Sarstedt 9 ml z-gel serum monovettes, se ^¬ rum was obtained by centrifugation (4 min, 1800g) and frozen in 2ml cryovials (Simport) at -80°C. At first use the sera were distributed into aliquots to minimize freeze-thaw cycles.

Two independent cohorts consisting of 23 samples (15 non progressive and 8 relapse) from Charite-Universitatsmedizin Berlin and 38 (25 non progressive and 13 relapse) samples from Innsbruck were combined for this analysis.

Analytics and statistical analyses were carried out for the markers according to Table 4 as described under 1. (although with adapted statistical settings due to the lower number of samples) and confirmed their status as appropriate markers ac ^¬ cording to the present invention.

Results :

In the present invention, metabolic signatures in serum were identified to predict whether a PCa would lead to relapse or not. Thus the sera used in this study were all obtained dur ^¬ ing the tumour stages of the patients.

Comparisons relapse vs. non progressive were performed to generate the data for Table 4 (the two independent cohorts con ^¬ sisting of 23 samples (15 non progressive (NP) and 8 relapse (R) ) and 38 (25 non progressive and 13 relapse) samples gave the fol ^¬ lowing results for the markers according to the present inven ^¬ tion:

Metabolites pvalue R vs NP AUC log2FC Median % cha

15S-HETE 0. 007 0.666 1.212 193. 685

PGD2 0. 009 0.688 1.406 138. 939

PC ae C40:3 0. 022 0.631 0.233 16. 654

LTB4 0. 045 0.626 1.158 90. 562

PGE2 0. 125 0.614 0.479 35. 656

12S-HETE 0. 146 0.620 0.692 83. 841

PC ae C40:4 0. 180 0.591 0.135 13. 611 lysoPC a C26:0 0.456 0.541 0.068 4.340

13S-HODE 0.551 0.549 0.172 11.337

5b,6b-EpoxyC 0.874 0.514 0.053 5.678

Table 4 lysoPC a 24:0, 7a-OH-C, 7-KC, 5a, 6a-EpoxyC also showed mark ^¬ er properties according to the present invention in this follow- up-study; however, due to the lower sample number than in example 1, did not reach the significance as the markers in table 4 for the patient panel as a whole (nevertheless, reached signifi ^¬ cance in the regional panels) . Accordingly, these results con ^¬ firmed the data already gathered in Example 1.

An interesting observation was made in the course of these tests with respect to another marker, alpha-KGA (alpha-keto glu- taric acid). This marker also reached statistical significance: pval_RvsNP AUC log2FC medFC

alpha-KGA 0.006 0.680 -0.370 -19.145

Table 5

This shows that alpha-KGA can also be used as PC relapse marker according to the present invention, be it either as independent marker or as an additional marker for combination with the lipid markers according to the present invention. In the present study, additional markers for PCa relapse were identi ^¬ fied as statistically significant:

Metabolites pval RvsNP AUC log2FC medFC

lysoPC a C17 0 0. 007 0.675 0.355 31.782

lysoPC a C16 0 0. 012 0.655 0.245 18.565

lysoPC a C18 0 0. 048 0.621 0.226 14.981

lysoPC a C16 1 0. 049 0.621 0.232 18.895

lysoPC a C18 1 0. 056 0.624 0.217 15.406

PC aa C40:6 0. 060 0.613 -0.241 -22.357

PC ae C38:3 0. 073 0.615 0.167 11.821

Table 6

According to these data, the markers according to table 6 are further markers according to the present invention and can be used either alone or in combination with one or more of the other PCa relapse markers described herein. References :

[R] R Development Core Team (2009) . R: A language and environ ^¬ ment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3-900051-07-0, URL http://www.R-project.org ORGINAL SOFTWARE WEBPAGE

[KIM] H. Kim, G.H. Golub and H. Park (2005) Missing value estimation for DNA microarray gene expression data: local least squares imputation, Bioinformatics , 21 (2) : 187-198.

[PCAM] W. Stacklies and H. Redestig (2007) pcaMethods : A collec ^¬ tion of PCA methods, R package version 1.18.0 (ONLINE DOCUMENT) [LIMMA] G.K. Smyth (2005) . Limma: linear models for microarray data. In: ' Bioinformatics and Computational Biology Solutions using R and Bioconductor ' . R. Gentleman, V. Carey, S. Dudoit, R. Irizarry, W. Huber (eds) , Springer, New York, 397-420.

[FDR] Y. Benjamini and Y. Hochberg (1995) Controlling the false discovery rate: a practical and powerful approach to multiple testing, Journal of the Royal Statistical Society Series B, 57, 289-300

[DLDA] S. Dudoit, J. Fridlyand, and T. P. Speed. (2000) Compari ^¬ son of Discrimination Methods for the Classification of Tumors Using Gene Expression Data. (Statistics, UC Berkeley, June 2000, Tech Report #576)

[SFSMISC] M. Maechler and many others. (2009) sfsmisc: Utilities from Seminar fuer Statistik ETH Zurich, R package version 1.0-8 [CATOOLS] Tuszynski J. (2008) caTools: Tools: moving window statistics, GIF, Base64, ROC AUC, etc., R package version 1.9 ONLINE DOCUMENT

[FIEMS] M. Beckmann, D. Enot and W. Lin (2007) FIEmspro: Flow Injection Electrospray Mass Spectrometry Processing: \\ data processing, classification modelling and variable selection in metabolite fingerprinting, R package version 1.1-0

[NNET] Venables, W. N. & Ripley, B. D. (2002) Modern Applied Statistics with S. Fourth Edition. Springer, New York. ISBN 0- 387-95457-0