DNA methylation test for prostate cancer

Description

Technical Field

[0001] The invention relates to the detection of a biomarker in a biological sample to test for the presence of prostate cancer. Specifically, the invention relates to the detection of a plurality of biomarkers in a biological sample to distinguish the presence of aggressive prostate cancer from non-aggressive prostate cancer or no cancer.Background to the Invention

[0002] Prostate cancer (PCa) is the most common non-cutaneous malignancy in men in the Western world. An estimated 1.1 million new cases were diagnosed in 2012, accounting for 15% of all male cancers worldwide. Ireland is currently experiencing one of the highest incidences of PCa in Europe, with approximately 3,000 new cases diagnosed per annum, representing 30% of all invasive cancers in men. With an ageing Western population and spread of Western culture (particularly diet), the global incidence is predicted to rise dramatically; the National Cancer Registry predicts the incidence in Ireland to rise by between 104-288% by 2040.

[0003] It is often said that "most men die with and not because of their prostate cancer". This is explained by the fact that most prostate tumours have a slow, long natural trajectory, posing little likelihood of clinical manifestation, and deemed indolent in nature; 10-year survival rates for PCa are close to 100%. Nevertheless, a proportion of prostate tumours are highly aggressive, and are associated with the lethal form of the disease. Whilst PSA (prostate specific antigen) screening and improvements in treatments have reduced PCa mortality, this disease accounted for an ~307,000 deaths in 2012, making it the 5 th< leading cause of male cancer-related deaths worldwide. Identifying molecular correlates to discern between aggressive and indolent tumors at an early stage (whilst potentially curable), is one of the greatest unmet clinical needs in this field. This will become even more pressing as the differential between the total number of PCa cases diagnosed and the number of lethal PCa cases grows.

[0004] Early detection and diagnosis of PCa involves a combination of a PSA blood test, a digital rectal examination (DRE) and histological examination of transrectal ultrasound (TRUS)-guided biopsy cores, respectively. Several major problems confound the early detection of PCa. There are an estimated 25-45 million PSA tests performed worldwide every year, Widespread PSA testing has significantly increased PCa incidence and led to overtreatment of low-risk disease with little likelihood of clinical manifestation. A further problem with PSA is its poor tumour-specificity; its high false-positive rate means that two-thirds of men who undergo invasive TRUS-biopsy have no tumour diagnosed. There are an estimated 10 million prostate biopsies performed worldwide / annum. Unnecessary TRUS-biopsies create an enormous burden on our healthcare system and cause significant anxiety, trauma and co-morbidities for patients. Finally, TRUS-biopsies are needle biopsies that sample <5% of the prostate and can thus miss tumour foci or indeed miss high-grade aggressive tumours. Studies addressing the economic burden of cancer in the EU, have estimated costs for PCa diagnosis and treatment over the next 20 years per 100,000 men at €30,284,000 (unscreened population) and €60,695,000 (screened population), €23,669,000 of which can be attributed to over-detected cancers.

[0005] Currently, there are no commercially available molecular diagnostics for PCa in widespread clinical practice. Progensa ®< (Gen-Probe) is a urine-based test of PCA3 gene expression performed after DRE, with FDA approval for use in men who have had ≥1 previous negative biopsies and for whom a repeat biopsy would be recommended based on current standard of care. The test is used to guide the decision to perform a repeat biopsy only. Its prognostic value is debated and research efforts combining it with the fusion-transcript TMPRSS2-ERG are underway in an attempt to address this.

[0006] Prolaris ®< (Myriad Genetics) and oncotypeDX ®< Prostate Cancer Assay (Genomic Health) are two examples of prognostic gene expression signatures (46 genes and 17 genes, respectively) that are analysed on biopsy tissues to aid prediction of PCa aggressiveness in conjunction with other clinical parameters (Gleason score, PSA). Both tests provide a more individualised risk-assessment of the underlying biology of the patient's tumour and are therefore aimed at guiding the decision between active surveillance and radical treatment in men diagnosed with PCa.

[0007] MDxHealth's product ConfirmMDx ™< is a PCR-based assay, which measures methylation of a 3-gene panel (GSTP1, RARβ, APC) in biopsy cores. It is positioned to distinguish patients with a true-negative prostate biopsy from those with occult cancer and akin to Progensa ®< , is used to guide the decision to perform a repeat biopsy. This same 3-gene panel (ProCaM ™< ) has also been investigated as a urine test to predict biopsy results for PCa, although these studies were inadequately powered. US 2013 / 022974 describes methylation markers and different sets of methylation markers associated with prostate cancer. WO 2014 / 012176 discloses methods for assessing prostate cancer comprising diagnosis of prostate cancer; prognosis of prostate cancer; staging assessment of prostate cancer; prostate cancer aggressiveness classification using methylation markers.

[0008] It is an object of the subject invention to overcome at least one of the above-mentioned problems.Statements of Invention

[0009] In contrast to these prior art technologies, the test presented herein (called epigenetic Cancer of the Prostate test in urine or epiCaPture) is an example of a "first in field" for urine diagnostics of potentially lethal, high-risk PCa. The panel of genes encompasses multiple dysregulated pathways in PCa, which is necessary to address the heterogeneity of the disease. These pathways include intracellular detoxification, the IGF axis, the Wnt axis and inflammation. The test presented herein addresses the unmet clinical needs confounding early detection of PCa. The test is a non-invasive DNA methylation test performed using urine or urine cell-sediment. It comprises a panel of at least 6 genes and an internal control gene. The test described herein offers significant commercial potential as a liquid biopsy for early non-invasive detection of high-risk, potentially lethal PCa. The data show that the test offers the unique advantages of i) better tumour-specificity than PSA, and ii) selective identification of high-risk PCa.

[0010] According to the invention, there is provided, as set out in the appended claims a method of determining the presence of high-risk prostate cancer in an individual, the method comprising a step of assaying a biological sample obtained from the individual for the presence of methylated regulatory DNA sequences as defined by SEQ ID NO 7, SEQ ID NO 8, SEQ ID NO 9, SEQ ID NO 10, SEQ ID NO 11, SEQ ID NO 12, SEQ ID NO 13, SEQ ID NO 14, SEQ ID NO 15 and SEQ ID NO 16 and calculating a normalized index of methylation (NIM) score, wherein the presence of the methylated regulatory DNA sequences as defined by SEQ ID NOs 7 to 16, and the calculated NIM score indicates a high-risk prostate cancer.

[0011] In one embodiment, the present invention discloses a method for determining an aggressive prostate cancer in an individual, the method comprising the step of assaying the biological sample obtained from the individual for the presence of at least one sequence selected from SEQ ID NO 17 and SEQ ID NO 18, wherein detection of the methylated regulatory DNA sequences as defined by SEQ ID NOs 7, SEQ ID NO 8, SEQ ID NO 9, SEQ ID NO 10, SEQ ID NO 11, SEQ ID NO 12, SEQ ID NO 13, SEQ ID NO 14, SEQ ID NO 15 and SEQ ID NO 16 and one sequence from SEQ ID NO 17 and SEQ ID NO 18 in a sample and calculating a NIM score indicates the presence of an aggressive prostate cancer, and wherein the sensitivity of the assay for detecting the methylated regulatory DNA sequences as defined by SEQ ID NOs 7 to 16 is at least 80%.

[0012] In one embodiment, the at least one sequence selected from SEQ ID NOs 17 and 18 is SEQ ID NO: 18 prostate-specific antigen (PSA).

[0013] In one embodiment, the sample is urine or a urine derivative from the individual.

[0014] The two controls may be used in the method of the invention: 1) ACTB (SEQ ID NO. 17): ACTB is measured by quantitative PCR (qPCR) and verifies and quantifies the presence of bisulfite modified DNA in each test sample. The quantity of ACTB is used to calculate an epiCaPture score (a score derived from the method of the invention). The amount of each gene in the method must be normalised relative to the amount of input bisulfite modified DNA in each test sample. 2) KLK3 (SEQ ID NO. 18). Expression of the KLK3 gene (the gene encoding PSA, Prostate Specific Antigen) is measured by quantitative RT-PCR and is used as a positive control to confirm the presence of prostate-derived nucleic acids in the test sample. This is important to carry out, in order to show that a test sample which appears negative for prostate cancer as determined by the method of the invention, is indeed truly negative and it is not simply a virtue of no prostate-derived material present in the bio-specimen. The expression of the KLK3 gene is measured using a commercially available qPCR assay, such as Integrated DNA Technologies (Assay ID Hs.PT.58.38546086).

[0015] The invention also relates to a kit for detecting the presence of prostate cancer in a sample from an individual, the kit comprising a control oligonucleotide as defined by SEQ ID NO 19, 20 or 21, or a variant thereof, and a set of oligonucleotides for detecting SEQ ID NOs 1 to 16. In one embodiment, the kit further comprises an oligonucleotide for detecting the presence of PSA. In one embodiment, the set of oligonucleotides is defined by SEQ ID NOs. 22 to 72. In one embodiment, the kit further comprises a support having at least one oligonucleotide selected from group SEQ ID No's 1 to 16 anchored thereon.

[0016] The kit preferably comprises a pair of forward and reverse oligonucleotide primers (SEQ ID NOs. 22 to 55) designed to specifically hybridise with bisulfite modified hypermethylated DNA sequences at the regulatory regions of each specific gene as defined by SEQ ID NOs 1 to 16; a fluorescently labelled oligonucleotide probe designed to specifically hybridise with bisulfite modified hypermethylated DNA sequences at the regulatory region of each specific gene (SEQ ID NO. 56 to 72), a set of forward and reverse oligonucleotide primers and a fluorescently labelled probe to specifically hybridise with bisulfite modified DNA contained as part of the human ACTB gene, regardless of DNA methylation patterns of this gene (Positive control 1), a qRT-PCR assay for the KLK3 gene (Positive control 2) to control for the presence of prostate-derived nucleic acids in the bio-specimen, and a gBlock ®< synthetic gene fragments for construction of standard curves (SEQ ID NO. 19, 20 or 21), necessary for quantification of methylation levels at individual DNA sequences contained within the panel.

[0017] As indicated above, the methods, assays and kits of the invention employ biomarkers (methylated regulatory DNA sequences of specific genes or oligonucleotides specific to those regulatory DNA sequences of those genes) as a means of assessing the risk of an aggressive or metastatic prostate cancer in an individual. In one preferred embodiment of the invention, the methods, assays, and kits may be employed as a clinical screening tool to assist in the identification of individuals with an aggressive form of or a high risk metastatic prostate cancer, especially symptomatic individuals, who should be subjected to more invasive investigations, such as a prostate biopsy. In this regard, it should be noted that many patients who present with symptoms of prostate cancer (i.e. the need to urinate frequently, difficulty in starting urination, weak or interrupted flow of urine, painful / burning urination; blood in the urine etc.) can turn out to be negative for prostate cancer, yet still have to undergo a prostate biopsy to reach that diagnosis. In this regard, the present invention provides a useful clinical decision making tool which can assist a clinician in identifying those symptomatic patients that are most at risk of having the cancer, thereby potentially reducing the numbers of patients who have to undergo a prostate biopsy needlessly.

[0018] In this specification, the term "biological sample" or "biological fluid" may be a sample obtained from an individual such as, for example, urine or urine cell-sediment, blood or a prostate tissue sample from a biopsy or a radical prostatectomy. In many cases, the individual will be a person suspected of having prostate cancer, or pre-disposed to developing prostate cancer as determined by other phenotypic, genotypic or hereditary traits.

[0019] In this specification, the term "prostate cancer status" when used with reference to an individual primarily refers to the risk of the individual having the cancer. Depending on the number of biomarkers detected in the individual, the assay and methods of the invention will assist a clinician is determining the risk that the individual is positive for prostate cancer. Thus, in one embodiment, the methods, assays and kits of the invention provide a means for screening male patients to identify those patients that should undergo further investigative procedures, such as a biopsy. However, the term also encompasses prognostic evaluation of the cancer, identification of predisposition to developing the cancer, staging of the cancer, and evaluation or monitoring of the progress of the cancer, in the individual. The latter evaluation is typically employed as a means of monitoring the effectiveness of a treatment for the cancer.

[0020] A "variant" of one of SEQUENCE ID No's 1 to 16 shall be taken to mean at least 70% sequence identity, preferably at least 80% sequence identity, more preferably at least 90% sequence identity, and ideally at least 95%, 96%, 97%, 98% or 99% sequence identity with the native sequence.

[0021] mRNA expression of the KLK3 gene (positive control 2 - SEQ ID NO. 18) may be measured by any suitable method including, but not limited to, a Northern Blot or detection by hybridisation to a oligonucleotide probe. A variety of hybridization assays using a variety of technologies for hybridization and detection are available. For example, a TaqMan assay (PE Biosystems, Foster City, CA; See e.g., U.S. Patent Nos. 5,962,233 and 5,538,848,) is utilized. The assay is performed during a PCR reaction. The TaqMan assay exploits the 5'-3' exonuclease activity of the AMPLITAQ GOLD DNA polymerase. A probe consisting of an oligonucleotide with a 5 '-reporter dye (e.g., a fluorescent dye) and a 3 '-quencher dye is included in the PCR reaction. During PCR, if the probe is bound to its target, the 5'-3' nucleolytic activity of the AMPLITAQ GOLD polymerase cleaves the probe between the reporter and the quencher dye. The separation of the reporter dye from the quencher dye results in an increase of fluorescence. The signal accumulates with each cycle of PCR and can be monitored with a fluorometer.

[0022] DNA methylation may be measured by any suitable method, such as quantitative methylation specific PCR (PMID: 10734209).

[0023] In other embodiments, reverse-transcriptase PCR (RT-PCR) is used to detect the expression of RNA where RNA is enzymatically converted to complementary DNA or "cDNA" using a reverse transcriptase enzyme. The cDNA is then used as a template for a PCR reaction. PCR products can be detected by any suitable method, including but not limited to, gel electrophoresis and staining with a DNA specific stain or hybridization to a labelled probe. In some embodiments, the quantitative reverse transcriptase PCR with standardized mixtures of competitive templates method described in U.S. Patents 5,639,606, 5,643,765, and 5,876,978 is utilized.

[0024] In the specification, the term "high-risk prostate cancer", "high-risk disease" or "aggressive prostate cancer" or "metastatic prostate cancer" should be understood mean a prostate cancer that is categorised by the D'Amico Risk Stratification criteria. The D'Amico criteria are used to define low, intermediate and high-risk prostate cancer. For example, (i) Low risk: having a PSA less than or equal to 10, a Gleason score less than or equal to 6, or are in clinical stage T1-2a; (ii) Intermediate risk: having a PSA between 10 and 20, a Gleason score of 7, or are in clinical stage T2b; and (iii) High-risk: having a PSA more than 20, a Gleason score equal or larger than 8, or are in clinical stage T2c-3a. The terms high-risk, aggressive and metastatic can be used interchangeably. The terms high-risk and aggressive describe a cancer of high tumour grade (according to the Gleason scale, >=8) and a highly likelihood of metastasising.

[0025] In the specification, the term "gBlock ®< " should be understood to mean a doublestranded DNA molecule of 125-2000 bp in length. In this instance, the gBlock ®< is defined by SEQ ID NO. 19 and contains sequences for (A) an internal control ACTB, and the genes (B) GSTP1 (C), SFRP2, (D) IGFBP3, (E) IGFBP7, (F) APC and (G) PTGS2. The gBock ®< defined by SEQ ID NO: 20 was designed to contain the nucleotide sequences of bisulfite converted fully methylated internal control sequence (ACTB) and seven DNA regulatory sequences (LXN, MAGPIE-1B, DNAH10, ZMIZ1, CENPV and OR2L13). The gBlock ®< defined by SEQ ID NO. 21 was designed to contain the nucleotide sequences of bisulfite converted fully methylated internal control sequence (ACTB) and four DNA regulatory sequences (MTMR8, F3, CDH8 and GALNTL6).

[0026] Some of the uses of the invention include: To test for the presence of prostate cancer. Use as a novel screening test for any male at risk of having prostate cancer. Use as a non-invasive test using urine to determine which male requires an invasive trans-rectal biopsy to confirm a diagnosis of prostate cancer by histological review of a biopsy specimen. The test can be carried out on any biological sample that harbours prostate DNA, including blood plasma / serum, prostate tissue and metastatic lesions, either visceral or bone.

[0027] Some of the advantages of the invention is to: Reduce / eliminate unnecessary invasive biopsies in men who don't need them; Identify which men require a trans-rectal prostate biopsy Alleviate over-treatment of low-risk disease; Inform the clinician about the molecular biology of the disease; and Aid risk-stratification for selection of subsequent treatments / active surveillance. Brief Description of the Figures

[0028] The invention will be more clearly understood from the following description of an embodiment thereof, given by way of example only, with reference to the accompanying drawings, in which:- Figure 1 illustrates Feasibility Study data (n=156). Results on a panel of 156 pre-biopsy urine samples, shown by biopsy outcome. Positive-biopsy men are further categorised into low (LR), intermediate (IR) and high (HR) risk groups using the CAPRA score. Each row represents a gene, each column represents a patient. Methylation is measured as a continuous variable from 0-1000 (Normalised Index of Methylation, NIM). Black squares indicate high methylation with a normalised index of methylation (NIM)>1, white squares indicate NIM=0 and shades of grey indicate intermediate NIM. Figure 2 illustrates results reduced into categorical results: men with methylation of at least one gene and men with an NIM>1 in any one gene. Figure 3 illustrates ROC curves for (A) PSA alone, (B)-(C) using the invention, and (D) the test of the invention and PSA>4 ng / ml, which achieve an AUC of 0.54, 0.87 (average) and 0.96, respectively. Figure 4 illustrates that the positive predictive value of invention which indicates its utility for reducing number of unnecessary biopsies by selectively detecting aggressive PCa. Figure 5 illustrates standard curves constructed over a 6-log range using 5 independent qMSP measurements of a gBlock ®< fragment containing sequences for (A) internal control, (B) GSTP1 (C), SFRP2, (D) IGFBP3, (E) IGFBP7, (F) APC and (G) PTGS2 (SEQ ID NO. 19). Each qMSP assay has a slope of -3.3 (+ / - 10%) and an R 2< >0.99, indicating a PCR efficiency close to 100%. Figure 6 illustrates graphs showing quantitative methylation-specific PCR data for SEQ ID NOs 7-16 in a radical prostatectomy cohort reveal quantitatively higher levels of DNA methylation in aggressive tumours (PC-A) compared with significant (PC-S) and indolent (PC-I) tumours and benign tissue. Figure 7 illustrates graphs showing descriptive statistics of the study cohort used in epiCaPture analysis with an increased cohort size from 156 men to 283 men. The data presented includes the original cohort of patients. A) Age, B) PSA levels of men (Whiskers indicate the minimum and maximum levels and the mean is indicated by a horizontal line) and statistics of the biopsy-positive cohort used in epiCaPture analysis. Patients were stratified according to C) D'Amico risk group and D) tumour grade (Gleason score, GS). Figure 8 illustrates a heat map of epiCaPture NIM scores for increased cohort of 283 men. Positive-biopsy men are further categorised by Gleason score. Each row represents a gene, each column represents a patient. Methylation is measured as a continuous variable from 0-1000 (Normalised Index of Methylation, NIM). White squares represents absence of methylation (NIM=0), with increasing shades of grey and black representing quantitatively higher amounts of methylation for a gene. Figure 9 is a graph illustrating performance of epiCaPture versus predicate at non-invasive detection of high-risk prostate cancer. Figure 10 are graphs illustrating GSTP1 methylation as detected in three independent cohorts of radical prostatectomy samples (A) a cohort of 44 men studied by Infinium HM450k methylation Beadchip (Table 11); (B) cohort of 125 men studied by quantitative PCR (Table 12); and (C) cohort of 178 men extracted from The Cancer Genome Atlas, for whom Infinium HM450k methylation Beadchip are publically available (Table 13). Panel (D) shows the methylation values detected in the urine samples from men undergoing trus-biopsy (n=283). In all cohorts, significantly higher levels of methylation are detected in the high-risk and high grade disease. Figure 11 are graphs illustrating SFRP2 methylation as detected in A) prostate tissues and B) urine samples from men undergoing TRUS-biopsy. In both cohorts, significantly higher levels of methylation are detected in the high-risk and high grade disease. Figure 12 are graphs illustrating IGFBP3 methylation as detected in (A) prostate tissues from men undergoing radical prostatectomy (Perry et al, British Journal of Cancer, 2007). Abbreviations: HGPIN: high grade prostatic intraepithelial neoplasia, HB: histologically benign and BPH: benign prostatic hyperplasia. (B) urine samples from men undergoing trus-biopsy, n=283. In both prostate tissue and urine, significantly higher levels of methylation are detected in the high-risk and high grade disease patients. Figure 13 are graphs illustrating IGFBP7 methylation as detected in (A) prostate tissues from men undergoing radical prostatectomy (Sullivan et al, Journal of Urology, 2012). Abbreviations: HGPIN: high grade prostatic intraepithelial neoplasia, HB: histologically benign and BPH: benign prostatic hyperplasia. (B) urine samples from men undergoing trus-biopsy, n=283. In both prostate tissue and urine, significantly higher levels of methylation are detected in the high-risk and high grade disease patients. Figure 14 are graphs illustrating APC methylation as detected in (A) prostate tissues from men undergoing radical prostatectomy (Murphy et al, Epigenetic Diagnosis and Therapy, 2015). Abbreviations: HGPIN: high grade prostatic intraepithelial neoplasia, TA: tumour associated benign. (B) urine samples from men undergoing trus-biopsy, n=283. In both prostate tissue and urine, significantly higher levels of methylation are detected in the high-risk and high grade disease patients. Figure 15 are graphs illustrating supporting data for LXN. DNA methylation was measured in three independent cohorts of radical prostatectomy samples (Tables 11 to 13), and in each case is significantly higher in high-risk / aggressive prostate cancer compared benign prostate tissue and / or low-risk or indolent prostate cancer. Two methodologies were used to quantitatively measure DNA methylation: Infinium HM450K BeadChip (A and C) and quantitative PCR (B). For simplicity, significance values are only shown for comparisons with aggressive prostate cancer. Figure 16 are graphs illustrating supporting data for MAGPIE-1B. DNA methylation was measured in three independent cohorts of radical prostatectomy samples (Tables 11 to 13), and in each case is significantly higher in high-risk / aggressive prostate cancer compared benign prostate tissue and / or low-risk or indolent prostate cancer. Two methodologies were used to quantitatively measure DNA methylation: Infinium HM450K BeadChip (A and C) and quantitative PCR (B). For simplicity, significance values are only shown for comparisons with aggressive prostate cancer. Figure 17 are graphs illustrating supporting data for DNAH10. DNA methylation was measured in three independent cohorts of radical prostatectomy samples (Tables 11 to 13), and in each case is significantly higher in high-risk / aggressive prostate cancer compared benign prostate tissue and / or low-risk or indolent prostate cancer. Two methodologies were used to quantitatively measure DNA methylation: Infinium HM450K BeadChip (A and C) and quantitative PCR (B). For simplicity, significance values are only shown for comparisons with aggressive prostate cancer. Figure 18 are graphs illustrating supporting data for ZMIZ1. DNA methylation was measured in three independent cohorts of radical prostatectomy samples, and in each case is significantly higher in high-risk / aggressive prostate cancer compared benign prostate tissue and / or low-risk or indolent prostate cancer. Two methodologies were used to quantitatively measure DNA methylation: Infinium HM450K BeadChip (A and C) and quantitative PCR (B). For simplicity, significance values are only shown for comparisons with aggressive prostate cancer. Figure 19 are graphs illustrating supporting data for CENPV. DNA methylation was measured in three independent cohorts of radical prostatectomy samples, and in each case is significantly higher in high-risk / aggressive prostate cancer compared benign prostate tissue and / or low-risk or indolent prostate cancer. Two methodologies were used to quantitatively measure DNA methylation: Infinium HM450K BeadChip (A and C) and quantitative PCR (B). For simplicity, significance values are only shown for comparisons with aggressive prostate cancer. Figure 20 are graphs illustrating supporting data for OR2L13. DNA methylation was measured in three independent cohorts of radical prostatectomy samples, and in each case is significantly higher in high-risk / aggressive prostate cancer compared benign prostate tissue and / or low-risk or indolent prostate cancer. Two methodologies were used to quantitatively measure DNA methylation: Infinium HM450K BeadChip (A and C) and quantitative PCR (B). For simplicity, significance values are only shown for comparisons with aggressive prostate cancer. Figure 21 are graphs illustrating supporting data for (A-C) F3. DNA methylation was measured in three independent cohorts of radical prostatectomy samples, and in each case is significantly higher in high-risk / aggressive prostate cancer compared benign prostate tissue and / or low-risk or indolent prostate cancer. Two methodologies were used to quantitatively measure DNA methylation: Infinium HM450K BeadChip (A, and C) and quantitative PCR (B). For simplicity, significance values are only shown for comparisons with aggressive (high-risk) prostate cancer Detailed Description Materials and MethodsStatistical methods:

[0029] Logistic regression is a standard method for modelling the relationship between a binary variable, in this case high-risk versus low-risk prostate cancer, and a set of continuous or categorical variables. For this analysis, the variables used for prediction consist of gene methylation values, as well as patient variables age and PSA. Mathematically this relationship is expressed as log p i 1 − p i = β 1 X 1 i + β 2 X 2 i + … + β m X mi where p i is the probability that the nth patient is high-risk based on their methylation profile and clinical characteristics, which are represented by the X mi 's. The β m coefficients give the effect that each incremental change in methylation, age or PSA has on the log-odds of the patient being high risk of prostate metastasis.

[0030] Due to the cost of collection of biomarkers, and the general principle that simpler models lead to more robust predictions, one aim of the analysis is to choose the smallest number of predictor variables, the X m 's, which will yield the best performing prediction model.

[0031] A LASSO 1< logistic regression (discussed below), along with a standard logistic regression incorporating six genes were fit to the data. Logistic regression models for each of the separate genes were also fitted for comparison. All models were trained using repeated 5-fold cross-validation with bootstrap resampling. The optimal cut-off value for prediction was then chosen using the entire data set.

[0032] A common problem with building a prediction model on the entire dataset is that the model will tend to over fit the current data set and will then underperform when new data is predicted from the model. In general, more complex models will tend to adapt to the training data and will not generalise as well to new data. Therefore sparser models are preferred.

[0033] In an ideal case a model is fitted to some training data and then its performance is estimated on an independent test set. A model can be selected by choosing the model that performs best on the test dataset which consists of new unseen observations. For small datasets, a single split of the data into testing and training sets is often not possible.

[0034] Cross-validation is a method for performing multiple random training-test splits of a dataset. The process is as follows: 1. Split the data into K equally sized portions (K is usually chosen to be between 5 and 10); 2. Leave aside one of the K portions and train the model(s) on the all of the remaining K-1 portions together; 3. Test the performance of the model(s) on the portion of the data that was set aside in step 2; and 4. Repeat iteratively leaving out each of the K portions in succession.

[0035] This method gives a more accurate assessment of the out-of-sample prediction performance of the models than simply fitting the model to the entire dataset. To further account for uncertainty in the cross-validation process, the dataset is bootstrapped, i.e. resample the entire dataset with replacement, and perform cross-validation on each bootstrap iteration. For the analysis here, K= 5 and 2000 bootstrap iterations were used.

[0036] All of the data is used in the model-building and assessment stages. The model building process used aims to mitigate against any optimistic bias. Due to the relatively small number of high-risk cases, splitting the data is not an efficient option. Ideally a test or hold-out set of data which were not used in the training step should be used on which to test performance. This should be done at a later stage using an independently collected test data set.

[0037] The LASSO is a penalised regression method for building prediction models which mitigates against over-fitting on a data set. The LASSO is a method for both model selection and estimation. In standard logistic regression, the model parameters are fitted using iterative maximum likelihood. This estimates the parameters which fit the data the best, and as such can over-fit to the training data. Penalised regression methods add a penalty term to the estimation equation which penalises large values of the coefficients.

[0038] This is a form of shrinkage which can yield more robust results. In the case of LASSO, the penalty term shrinks some coefficients to zero, acting as a form of variable selection.

[0039] The strength of penalisation is determined by a parameter, λ. The optimal lambda value is found by running a further cross-validation iteration within each iteration of the outer cross-validation loop.

[0040] For the final assessment of the LASSO model, models chosen in the resampling and cross-validation iterations were searched and selected as the final model, the most frequently occurring model. The performance assessments and model parameters were then based on the iterations where this model occurred. Regression coefficients were then obtained by averaging over these iterations.Samples

[0041] The epiCaPture test (method described herein) is performed on urine or urine cell-sediment. The urine cell-sediment is obtained by centrifugation (10,000xg for 10 minutes) of a first-void urine sample (up to 50ml) following a digital rectal examination (DRE). The DRE consists of three strokes per lobe of the prostate gland. Enough pressure is applied to the prostate to depress the surface approximately 1cm, from the base toe the apex and from the lateral to the median line for each lobe. Total nucleic acid is extracted from the cell sediment using a standard silica-membrane based extraction protocol (using a Qiagen total nucleic acid isolation kit, or similar commercially available product). Purified DNA (100ng) is subject to bisulfite conversion (using a Qiagen epitect kit, or similar commercially available product).

[0042] Expression of the KLK3 gene is measured by qRT-PCR using a commercially available primer and probe assay (Assay ID Hs.PT.58.38546086 available from Integrated DNA Technologies). Positive expression of the KLK3 gene relative to a housekeeper gene (ACTB) indicates the presence of prostate cells in the urine sediment and validity of the urine sample for epiCaPture analysis.

[0043] A 648 bp synthetic gBlock ®< DNA sequence (IDT - SEQ ID NO: 19) was designed to contain the nucleotide sequences of bisulfite converted fully methylated internal control sequence (ACTB) and six DNA regulatory sequences (GSTP1, SFRP2, IGFBP3, IGFBP7, APC, PTGS2):

[0044] An 885 bp synthetic gBlock ®< DNA sequence (IDT - SEQ ID NO: 20) was designed to contain the nucleotide sequences of bisulfite converted fully methylated internal control sequence (ACTB) and six DNA regulatory sequences (LXN, MAGPIE-1B, DNAH10, ZMIZ1, CENPV and OR2L13):

[0045] A 643 bp synthetic gBlock ®< DNA sequence (IDT - SEQ ID NO: 21) was designed to contain the nucleotide sequences of bisulfite converted fully methylated internal control sequence (ACTB) and four DNA regulatory sequences (MTMR8, F3, CDH8 and GALNTL6): Table 1: The gBlock ®< (SEQ ID NO. 19) was used to construct to-fold serial dilutions over a 6-log template concentration range to determine the dynamic range and PCR efficiency of each epiCaPture assay:StandardCopy numberVol. of gBlock (µl)Vol. of molecular grade H 2 O (µl)DNA conc. (pg / µl)11,000,00010 (WS)173.80.5442100,00010 (1)900.0544310,00010 (2)900.0054441,00010 (3)900.000544510010 (4)900.000054461010 (5)900.00000544

[0046] Quantitative methylation specific PCR (qMSP) is performed, as previously described 3-5< . The PCR efficiency of each of the assays (internal control and 6 targets) was rigorously evaluated by performing 5 independent replicates (each with 3 technical replicates) over a 6-log template concentration range (Figure 5). Bisulfite treated DNA is amplified in parallel TaqMan ®< PCR reactions performed with oligonucleotides specific for each of the target methylated DNA regulatory sequences (SEQ ID NOs 1 to 6) and the endogenous control gene ACTB (SEQ ID NO: 17). Samples are considered positively amplified when a comparative threshold cycle (C T ) of <50 was detected in at least two out of three replicates. A normalized index of methylation (NIM) was calculated, as previously described 6< , to determine the ratio of the normalized amount of methylated target to the normalized amount of ACTB in any given sample, by applying the formula: NIM = T A R G E T sample / T A R G E T MC / A C T B sample / A C T B MC x 1000

[0047] Where TARGET sample is the quantity of fully methylated copies of each of the sequences being sampled in any individual sample, TARGET MC is the quantity of fully methylated copies of each of the sequences being sampled in a commercially available fully methylated bisulfite converted human DNA sample (Qiagen product number 59655), ACTB sample is the quantity of bisulfite modified templates in any individual sample and ACTB MC is the quantity of bisulfite modified templates in the universally methylated control DNA.

[0048] All genomic sequences for GSTP1, SFRP2, IGFBP3, IGFBP7, APC, PTGS-2, LXN, MAGPIE-1B, DNAH10, ZMIZ1, CENPV, OR2L13, and F3 were obtained from the UCSC Human Genome Browser (http: / / genome-euro.ucsc.edu).Results

[0049] Results from the study on 156 men (Figures 3(A) to 3(D)) demonstrate that the invention can non-invasively discriminate high-risk (metastatic) PCa from low-risk (less chance of metastasis) disease and benign enlargement of the prostate (AUC=0.86). In this cohort, a high score (NIM>1) had 100% specificity for PCa, and greatly outperformed PSA, which yielded a PCa-specificity of only 11.63%. Table 2 is data relating to the cohort of 156 TRUS-biopsy patients. The exclusion criteria for the cohort were (1) metastases on an MRI and / or a bone scan and (2) not post-DRE.Biopsy positive Biopsy negative P value N10848Age (years) Mean69.9564.850.008Median69.5066Range53-8542-82PSA (ng / ml) Mean16.537.06<0.0001Median106.10Range4.1-95.90.2-30.30Risk group LR14IR58HR36 Urinary detection of prostate cancer

[0050] For non-invasively distinguishing men who have prostate cancer form those who do not (or more strictly speaking, men with a positive biopsy from men with a negative biopsy), the best combination of biomarkers is GSTP1 used in conjunction with PSA. This is calculated using a LASSO model (Table 3, Figure 3A). This achieved a positive predictive value (PPV) of 92%, with a negative predictive value (NPV) of 52%, with a sensitivity and specificity for prostate cancer of 81% and 77%, respectively. The combination of six methylated DNA regulatory sequences (as defined by SEQ ID NOs 1 to 6) in the method described herein also performs well at non-invasive detecting prostate cancer: PPV = 92%, NPV = 51%, sensitivity = 60% and specificity = 89%. Table 3: Biopsy positive versus biopsy negative SEQ (Gene) AUCSensitivitySpecificityPPVNPV1 (GSTP1) 0.72 (0.66 - 0.77)0.440.980.980.452 (SFRP2) 0.66 (0.60 - 0.72)0.370.940.930.413 (IGFBP3) 0.65 (0.58 - 0.73)0.410.890.890.414 (IGFBP7) 0.64 (0.58 - 0.70)0.29110.405 (APC) 0.68 (0.60 - 0.75)0.490.870.890.446 (PTGS2) 0.63 (0.55 - 0.72)0.340.980.970.41PSA 0.76 (0.67 - 0.84)0.880.510.800.67All 0.75 (0.68 - 0.81)0.600.890.920.51LASSO (1 (GSTP1) + PSA) 0.83 (0.74 - 0.89)0.810.770.920.50 Urinary detection of high-risk prostate cancer

[0051] However, as stated already, the dilemma for prostate cancer detection is not in the ability to detect the entire spectrum of disease, for which PSA is already adequately doing, but to specifically detect high-risk disease with high likelihood to metastasise. For predicting high-risk prostate cancer according to D'Amico criteria 2< , the LASSO, which is the selection method used here, determines that GSTP1 and IGFBP3 are the best fit (Table 4, Figure 3B). This combination delivers a PPV 56% of and NPV of 94% for high-risk disease, with a sensitivity and specificity both at 82%. The combination of all 6 genes, performs slightly less well, delivering a sensitivity of 52% and a specificity of 92%, for high risk disease. The method described herein (and derivations of it) outperforms current clinical practice (PSA), which in this cohort was found to have a sensitivity of 100% and specificity of only 21% (at the 4ng / ml cut-off) for high-risk disease. Table 4. Detection of high-risk disease SEQ (Gene) AUCSensitivitySpecificityPPVNPV1 (GSTP1) 0.78 (0.68 - 0.87)0.700.870.610.922 (SFRP2) 0.77 (0.68 - 0.86)0.670.850.550.903 (IGFBP3) 0.76 (0.66 - 0.86)0.610.820.490.884 (IGFBP7) 0.77 (0.67 - 0.86)0.580.910.630.885 (APC) 0.76 (0.66 - 0.86)0.670.840.540.906 (PTGS2) 0.71 (0.60 - 0.82)0.520.920.630.87All six 0.84 (0.75 - 0.93)0.790.820.550.93LASSO (1 (GSTP1) + 3 (IGFBP3)) 0.83 (0.75 - 0.92)0.820.820.560.94 Urinary detection of high-grade prostate cancer.

[0052] The Gleason grading system is the strongest prognostic indicator for prostate cancer. It is a histological grading system based on the glandular pattern of the tumour. A Gleason score is obtained by the addition of the primary and secondary grades. The presence of Gleason grade 4 or higher, or a Gleason score of 7 or higher predicts a poor prognosis.

[0053] For predicting tumours with a high Gleason score (>=8), the combination of all 6 biomarkers outlined above outperforms all biomarkers assessed individually (Table 5, Figure 3C), with a PPV of 48%, a NPV of 96% and a sensitivity and specificity of 76% and 87%, respectively. Combining all six markers with PSA gives some improvement again, with a sensitivity and specificity of 86% and 82%. Table 5: Detection of high-grade disease SEQ (Gene) AUCSensitivitySpecificityPPVNPV1 (GSTP1) 0.74 (0.63 - 0.86)0.670.820.380.942 (SFRP2) 0.78 (0.67 - 0.89)0.670.880.480.943 (IGFBP3) 0.78 (0.67 - 0.90)0.670.820.380.944 (IGFBP7) 0.77 (0.66 - 0.89)0.710.780.350.945 (APC) 0.75 (0.62 - 0.87)0.710.800.380.946 (PTGS2) 0.68 (0.54 - 0.83)0.570.890.460.93PSA 0.79 (0.71 - 0.87)0.950.630.290.99All 6 0.83 (0.73 - 0.94)0.760.870.480.96All 6 + PSA 0.86 (0.76 - 0.96)0.860.820.440.97LASSO (3 (IGFBP3)) 0.78 (0.67 - 0.90)0.670.820.380.94 Supplementary Data

[0054] The cohort size was increased from 156 men to 283 men. Table 6 is data relating to the cohort of 283 TRUS-biopsy patients. The exclusion criteria for the cohort were (1) metastases on an MRI and / or a bone scan and (2) not post-DRE.Biopsy negative Biopsy positive P value n (283) 135 148 Age (years) mean64.4368.44<0.0001median6568<0.0001range42-8347-85PSA (ng / ml) mean6.5911.78<0.0001median6.058.90<0.0001range0.2-63.800.6-144D'Amico Risk Group LR29IR73HR48Gleason Score 641774811921101

[0055] epiCaPture was performed on the cohort of 283 men, consisting of 135 biopsy-positive men and 148 biopsy-negative men. The age and PSA characteristics of the cohort are presented in Table 6. Although the biopsy-positive group were significantly older and had a significantly higher median PSA level (8.90 versus 6.05), there is considerable overlap in the range of ages and PSA levels for both groups (Figure 7A and 7B). Indeed, the mean and median PSA levels for the biopsy-negative group are above the 4ng / ml threshold widely used for indicating need for prostate-biopsy. The biopsy-positive cohort were considered in terms of risk-group stratification (according to the D'Amico criteria), which encompasses tumour grade (Gleason score), PSA level and clinical stage) and tumour grade stratification (Table 6, Figures 7C and 7D).

[0056] Each of the 6 gene panel was analysed individually in each patient, and a normalised index of methylation (NIM) score was generated for each gene (Figure 8). Different approaches were studied to determine the best performing method to (1) discriminate biopsy positive from biopsy negative and (2) selectively detect high-risk and high-grade disease. The performance of individual genes versus different combinations was studied using LASSO and tree mathematical models. In each instance, the performance of an NIM threshold (equations 3 to 5 below) produced the best performance indices (positive and negative predictive power) (Table 7-9). The NIM equation normalises for the amount if input bisulfite modified DNA present in the sample and calculates the proportion of the target sequence which is methylated relative to a 100% fully methylated DNA sequence. NIM = T A R G E T sample / T A R G E T MC / A C T B sample / A C T B MC x 1000

[0057] NIM threshold for discriminating biopsy positive from biopsy negative was determined as 0.73: NIM SUM POSITIVE : NIM Gene 1 + NIM Gene 2 + NIM Gene 3 + NIM Gene 4 + NIM Gene 5 + NIM Gene 6 > 0.73

[0058] Data from the 283 men show that for the 6 gene panel, the NIM threshold for detecting high-risk / high-grade disease was determined as 1.25 across the 6 gene panel. NIM SUM HIGHT RISK : NIM Gene 1 + NIM Gene 2 + NIM Gene 3 + NIM Gene 4 + NIM Gene 5 + NIM Gene 6 > 1.25Detection of high-grade Prostate Cancer

[0059] By applying this model (NIM_SUM >1.25), epiCaPture has a comparable sensitivity for high-grade prostate cancer (>=Gleason score 8) compared with the predicate test, PSA (Table 9, Figure 9). In this cohort of men, epiCaPture detected 84.85% of men with high-grade disease, as compared with 90.91% detected by PSA. The specificity and negative predictive value (Table 9, Table 10) of epiCaPture is far superior to PSA. Almost 98% of men with a negative biopsy tested negative for epiCaPture. Comparably, only 24.44% of the 135 men with a negative biopsy did not have an elevated PSA. This high false-positive rate (76%) of PSA is the reason why so many men undergo unnecessary biopsy. Table 7. epiCaPture performance characteristics: biopsy positive versus biopsy negative SEQ ID NO (Gene) AUCSensitivitySpecificityPPVNPV1 (GSTP1)0.690.390.980.950.592 (SFRP2)0.640.350.930.840.573 (IGFBP3)0.650.350.930.840.574 (IGFBP7)0.660.330.970.920.575 (APC)0.690.430.930.860.606 (PTGS2)0.660.400.930.870.59Best 4 (1,2, 3, 4)0.750.451.001.000.63Best 5 (1, 3, 4, 5, 6)0.760.490.990.970.64All 60.770.500.980.960.64NIM Sum 4 (1,3,4,5)0.450.990.990.62NIM Sum 5 (1,2,4,5,6)0.460.990.970.62NIM Sum > 0.73 0.55 0.93 0.9 0.65 Tree (G1 + G3) 0.760.90.50.97LASSO (G1 + G3 +G4 + G5) 0.760.560.90.860.65

[0060] Individual genes (targets) varied in their ability to discriminate presence of prostate cancer (biopsy-positive) from absence (biopsy-negative), ranging from a sensitivity of 33% (Gene 4) to 40% (Gene 6) (Table 7). Increasing the number of markers, for example, the best 4 or 5 or all 6, improved the sensitivity of urinary detection of prostate cancer to 45%, 49% and 50%, respectively.

[0061] However, summing the NIM across the gene panel and applying an NIM sum threshold of >0.73 improved the sensitivity to 55% of men with prostate cancer. The positive and negative predictive values for prostate cancer by applying an NIM sum threshold >0.73 are 90% and 65%, respectively. Table 8. epiCaPture performance characteristics: detection of high risk disease SEQ ID NO (Gene) AUCSensitivitySpecificityPPVNPV1 (GSTP1)0.790.630.920.620.932 (SFRP2)0.780.590.930.610.923 (IGFBP3)0.780.460.970.780.904 (IGFBP7)0.790.570.930.620.925 (APC)0.780.610.920.600.926 (PTGS2)0.740.480.950.630.90Best 4 (2,3,5,6)0.810.870.630.310.96Best 5 (1, 2,3,4,6)0.800.870.520.260.95All 60.780.890.130.170.86NIM Sum 4 (1,3,5,6)0.800.880.560.96NIM Sum 5 (1,2,3,4,5)0.830.850.520.96NIM Sum >1.25 0.83 0.85 0.52 0.96 Tree (G3 + G5) 0.740.910.610.95LASSO (G3 + G4 + G5 + G6) 0.860.890.700.360.97

[0062] Individual genes (targets) varied in their ability to detect high-risk prostate cancers, ranging from a sensitivity of 46% (Gene 3) to 63% (Gene 1) (Table 8). Increasing the number of markers, for example, the best 4 or 5 or all 6, does not markedly improve the accuracy of detecting high-risk prostate cancer, over individual markers, which can be attributed to the molecular heterogeneity of prostate cancer.

[0063] However, summing the NIM across the panel of best 4 or best 5 or applying an NIM sum threshold of 1.25 improved the sensitivity to 80% and 83%, respectively. The positive and negative predictive value for high-risk prostate cancer by applying an NIM sum threshold >1.25 are 52% and 96%, respectively. Table 9. epiCaPture performance characteristics: detection of high -grade disease SEQ ID NO (Gene) AUCSensitivitySpecificityPPVNPV1 (GSTP1)0.7730910.6060610.8920.4255320.9449152 (SFRP2)0.792970.6363640.920.5121950.9504133 (IGFBP3)0.7905450.5151520.9680.680.9379844 (IGFBP7)0.8032120.6060610.9120.476190.9460585 (APC)0.7832120.6363640.8960.4468090.9491536 (PTGS2)0.7284240.5151520.9320.50.935743Best 4 (3, 4, 5, 6)0.820.880.620.240.98Best 5 (2, 3, 4, 5, 6)0.820.880.510.190.97All 60.760.880.080.110.83NIM Sum 4 (2,3,4,5)0.820.870.450.97NIM Sum 5 (2, 3, 4, 5, 6)0.820.850.420.97NIM Sum >1.25 0.85 0.82 0.38 0.98 Tree (G3 + G6) 0.730.920.550.96LASSO (G3 + G5 + G6) 0.830.780.820.360.97

[0064] Individual genes also varied in their ability to detect high-grade (Gleason score >=8) prostate cancers, ranging from a sensitivity of 52% (Gene 3 and 6) to 64% (Gene 2 and 5) (Table 9). Increasing the number of markers, for example, the best 4 or 5 or all 6, does not markedly improve the accuracy of detecting high-grade prostate cancer, over individual markers, which can be attributed to the molecular heterogeneity of prostate cancer.

[0065] However, summing the NIM across the panel of best 4 or best 5 or applying an NIM sum threshold of 1.25 improved the sensitivity to 82% and 85%, respectively. The positive and negative predictive values for high-grade prostate cancer by applying an NIM sum threshold >1.25 are 38% and 98%, respectively.

[0066] By applying this model (NIM_SUM>1.25), epiCaPture has a comparable sensitivity for high-grade prostate cancer (>=Gleason score 8) compared with the predicate test, PSA (Table 10, Figure 9). In this cohort of men, epiCaPture detected 84.85% of men with high-grade disease, as compared with 90.91% detected by PSA. The specificity and negative predictive value (Table 8, Table 9) of epiCaPture is far superior to PSA. Almost 98% of men with a negative biopsy tested negative for epiCaPture. Comparably, only 24.44% of the 135 men with a negative biopsy did not have an elevated PSA. This high false-positive rate (76%) of PSA is the reason why so many men undergo unnecessary biopsy. Table 10. Relative Sensitivity & Specificity SENSITIVITY n epiCaPture high n (%) PSA >= 4 ng / ml n (%) Biopsy positive 14870 (47.30)136 (91.89)False-negative rate0.530.08Gleason >=8 3328 (84.85)30 (90.91)False-negative rate0.150.09SPECIFICITY n epiCaPture negative n (%) PSA < 4 ng / ml n (%) Biopsy negative 135132 (97.78)33 (24.44)False-positive rate0.020.76

[0067] Quantitative analysis of DNA methylation at six gene loci in prostate tissues and urine samples indicates that high levels of methylation detected in high-risk tumour tissues can be measured in urine as a surrogate. Examples of this are shown for five of the six gene panel, Target 1 (GSTP1; Figure 10), Target 2 (SFRP2; Figure 11), Target 3 (IGFBP3; Figure 12), Target 4 (IGFBP7; Figure 13) and Target 5 (APC; Figure 14).

[0068] Quantitative analysis of DNA methylation at the seven of the remaining ten gene loci that were analysed in prostate tissues indicates that high levels of methylation detected in high-risk tumour tissues can be measured. The genes were analysed on 3 independent cohorts of prostate tissue samples and all show consistent patterns of significant methylation in high-risk / aggressive prostate cancer. Examples of this are shown for Target 7 (LXN; Figure 15), Target 8 (MAGPIE-1B; Figure 16), Target 9 (DNAH10; Figure 17), Target 10 (ZMIZ1; Figure 18), Target 11 (CENPV; Figure 19), Target 12 (OR2L13; Figure 20) and Target 13 (F3; Figure 21). The details of the three different cohorts used for the study relating to those genes listed above are provided below.Cohort A

[0069] Benign prostate tissue was obtained from radical cystoprostatectomy or trans-urethral resection of the prostate, from men with no clinical or histopathological evidence of prostate cancer. Precursor lesions proliferative inflammatory atrophy (PIA) and high grade prostatic intra-epithelial neoplasia (HGPIN), (HGPIN) and primary tumours (indolent (PCI) and aggressive (PCA)) were all obtained from radical prostatectomy specimens. PCI was defined as Gleason 6, pT2 disease, with a pre-operative PSA <10 ng / ml and no evidence of biochemical or clinical recurrence (5-year follow-up). PCA was defined as primary Gleason ≥4, pT3 disease, with evidence of biochemical or clinical recurrence. Metastatic lesions were obtained from visceral metastases (liver and or lymph node), obtained during rapid autopsy. All patient samples were obtained retrospectively with ethical approval granted by the associated institutions: benign (St. James's Hospital (SJH), Ireland; Adelaide and Meath Hospital incorporating the National Children's Hospital (AMNCH), Ireland); PIA (SJH); HGPIN (AMNCH); PCI (SJH; Mater Misericordiae (MM), Ireland; Beaumont Hospital (BH), Ireland); PCA (SJH; MM; BH); PCM (University of Washington, USA).

[0070] In each case, H&E slides were reviewed by a consultant pathologist, who identified and marked the relevant target areas. Six serial 8µm sections were cut from the respective formalin fixed paraffin embedded (FFPE) blocks and mounted onto PEN membrane glass slides (Life Technologies) for laser capture microdissection (LCM). The sixth section was H&E stained and reviewed to ensure a consistent percentage of target cells. LCM was performed to enrich for target epithelia as previously described, using the Arcturus XT system (Life Technologies). DNA and total RNA were isolated from LCM caps (harboring microdissected tissue) in parallel, using the QIAamp DNA micro kit (Qiagen) and RecoverAll Total Nucleic Acids isolation kit (Ambion), respectively. Table 11 - Clinicopathologic data for Cohort A Benign* PIA* HGPIN* PCI* PCA* PCM Number of cases 1076786Mean age (years) 66.1061.3061.5057.5058.7572.83median 64.5062.0062.0058.0059.5074.00range 48-7949-6856-6650-6646-6960-81Mean PSA (ng / ml) NA7.748.055.238.0062.48median 8.947.905.407.2551.95range 3.18-9.895-11.603.60-7.104.50-13.6041-105Gleason score (n) 6 -047007 (3+4) -400007 (4+3) -310508 -000329 -0100310 -00001TNM stage (n) pT2a -10000pT2b -01000pT2c -54700pT3a -11040pT3b -00040pT3c -00000pT4 -00006BCR (n) -NANA086Abbreviations: PIA: proliferative inflammatory atrophy, HGPIN: high-grade prostatic intraepithelial neoplasia, PCI: indolent prostate cancer, PCA: aggressive prostate cancer, PCM: metastatic prostate cancer, BCR: biochemical recurrence, NA = not available. *cohorts are age-matched. Cohort B

[0071] A retrospective cohort of radical prostatectomy cases was used to validate potentially prognostic differentially methylated regions identified in cohort 1. All patient samples were obtained retrospectively with ethical approval granted by the associated institutions: benign (SJH, AMNCH) and tumor (SJH, MM and BH). Tumor samples were assigned as low-risk (Gleason score 3+3, pT2; n=23); significant (Gleason score 7, pT2; n=42); or high-risk (Gleason score ≥ 4+3, pT3; n=39), based on histopathological review of radical prostatectomy specimens. For control purposes, histologically benign prostate tissues (n=21) were procured from radical prostatectomy or trans-urethral resection of the prostate. Tumor and benign foci were marked by a consultant histopathologist (SPF, BL) and targeted macro-dissection with a scalpel was carried out on four serial 5µm sections. DNA and total RNA were isolated using the RecoverAll Total Nucleic Acids Isolation kit (Ambion). Table 12 - Clinicopathologic data for Cohort B Benign* Tumor low-risk significant high-risk Number of cases 21234239Mean age (years) 65.959.76062range 44-8749-7048-7349-74Mean PSA (ng / ml) 5.56.66.88.2range 0.42-10.41.2-12.32.4-14.73.1-18.7Gleason score (n) ≤6 -23107 (3+4) -037177 (4+3) -0414≥8 -008TNM stage (n) pT2 -23420pT3 -0039pT4 -000 Cohort C

[0072] In June 2014, The Cancer Genome Atlas (TCGA) database was mined for HM450k data for patient specimens corresponding to low-risk (n=9), significant (n=68) and high-risk (n=67) PCa as defined already for cohort 2. Histologically benign HM450k data were also retrieved (n=29). For each sample, raw *.IDAT files were extracted and processed through an abridged run of RnBeads (including pre-filtering, BMIQ normalization and post-filtering). β-values for probes contained within the 13 potentially prognostic DMRs were extracted and a mean DMR β-value was calculated for each sample. Methylation differences between cohorts were assessed using an unpaired T test with Welch's correction. Significance was ascribed as P<0.05. Table 13 - Clinicopathologic data for Cohort C Benign* Tumor low-risk significant high-risk Number of cases 3496867Mean age (years) 60.85860.161.4range 44 - 7147 - 7244 - 7344 - 77Mean PSA (ng / ml) 12.76.87.113.6range 1.8 - 873.6 - 100.7 - 26.61.6 - 87Gleason score (n) ≤6 -9007 (3+4) -05607 (4+3) -01235≥8 -0032TNM stage (n) pT2 -9680pT3 -0063pT4 -004

[0073] In the specification the terms "comprise, comprises, comprised and comprising" or any variation thereof and the terms "include, includes, included and including" or any variation thereof are considered to be totally interchangeable and they should all be afforded the widest possible interpretation and vice versa.References

[0074] 1. Tibshirani R. Regression shrinkage and selection via the LASSO. J. Royal. Statist. Soc B., 1996 vol. 58(1): 267-288. 2. Bastian PJ, Boorjian SA, Bossi A, et al. High-risk prostate cancer: from definition to contemporary management. European urology 2012; 61(6): 1096-106. 3. Eads CA DK, Kawakami K, Saltz LB, Blake C, Shibata D, Danenberg PV, Laird PW. MethyLight: a high-throughput assay to measure DNA methylation. Nucleic Acids Research 2000; 28(8): E32. 4. Perry AS, Loftus B, Moroose R, et al. In silico mining identifies IGFBP3 as a novel target of methylation in prostate cancer. British journal of cancer 2007; 96(10): 1587-94. 5. Perry AS, O'Hurley G, Raheem OA, et al. Gene expression and epigenetic discovery screen reveal methylation of SFRP2 in prostate cancer. International journal of cancer Journal international du cancer 2013; 132(8): 1771-80. 6. Yegnasubramanian S, Kowalski J, Gonzalgo ML, et al. Hypermethylation of CpG islands in primary and metastatic human prostate cancer. Cancer Res 2004; 64(6): 1975-86. 7. Sullivan L, Murphy TM, Barrett C, Loftus B, Thornhill J, Hollywood D, Lynch T, Perry AS. IGFBP7 promoter methylation and gene expression analysis in prostate cancer. Journal of Urology (2012) 188(4) 1354-60. PMID: 22906661. 8. Murphy TM, Tuzova AV, O'Rourke CJ, Greene C, Sullivan L, Thornhill J, Barrett C, Loftus B, Lynch T, Perry AS. Multigene Methylation Biomarker Analysis in Prostate Cancer. Epigenetic Diagnosis and Therapy, in press

Claims

1. A method of determining the presence of high-risk prostate cancer in an individual, the method comprising a step of assaying a biological sample obtained from the individual for the presence of methylated regulatory DNA sequences as defined by SEQ ID NO 7, SEQ ID NO 8, SEQ ID NO 9, SEQ ID NO 10, SEQ ID NO 11, SEQ ID NO 12, SEQ ID NO 13, SEQ ID NO 14, SEQ ID NO 15 and SEQ ID NO 16 and calculating a normalized index of methylation (NIM) score, wherein the presence of the methylated regulatory DNA sequences as defined by SEQ ID NOs 7 to 16, and the calculated NIM score indicates a high-risk prostate cancer.

2. A method according to Claim 1, wherein detection of all of the methylated regulatory DNA sequences from SEQ ID NOs 7 to 16, having a sensitivity of at least 80% ,and the calculated NIM score, indicates the presence of high-risk prostate cancer.

3. A method according to Claim 1 in which the sample is urine or a urine derivative from the individual.

4. A method according to Claim 1 for determining an aggressive prostate cancer in an individual, the method further comprising the step of assaying the biological sample obtained from the individual for the presence of at least one sequence selected from SEQ ID NO 17 and SEQ ID NO 18, wherein detection of the methylated regulatory DNA sequences as defined by SEQ ID NOs 7, SEQ ID NO 8, SEQ ID NO 9, SEQ ID NO 10, SEQ ID NO 11, SEQ ID NO 12, SEQ ID NO 13, SEQ ID NO 14, SEQ ID NO 15 and SEQ ID NO 16 and one sequence from SEQ ID NO 17 and SEQ ID NO 18 in a sample and calculating a NIM score indicates the presence of an aggressive prostate cancer, and wherein the sensitivity of the assay for detecting the methylated regulatory DNA sequences as defined by SEQ ID NOs 7 to 16 is at least 80%.

5. A method according to Claim 4, in which the at least one sequence selected from SEQ ID NO 17 and SEQ ID NO 18, is SEQ ID NO: 18, encoding prostate-specific antigen (PSA).

6. A method according to Claim 4 or Claim 5, in which detection of at the methylated DNA regulatory sequences as defined by SEQ ID NOs 7, SEQ ID NO 8, SEQ ID NO 9, SEQ ID NO 10, SEQ ID NO 11, SEQ ID NO 12, SEQ ID NO 13, SEQ ID NO 14, SEQ ID NO 15 and SEQ ID NO 16, having a sensitivity of at least 80% and the calculated NIM score indicates the presence of an aggressive (metastatic) prostate cancer.

7. A kit for detecting the presence of prostate cancer in a sample from an individual, the kit comprising a control oligonucleotide as defined by SEQ ID NO 19, 20 or 21, and a set of oligonucleotides for detecting SEQ ID NOs 1 to 16.

8. A kit as claimed in Claim 7, wherein the kit further comprises an oligonucleotide for detecting the presence of PSA.

9. A kit as claimed in Claim 7 or Claim 8, wherein the set of oligonucleotides is defined by SEQ ID NOs. 22 to 72.

10. A kit as claimed in any one of Claims 7 or 8, wherein the kit further comprises a support having at least one oligonucleotide selected from group SEQ ID No's 1 to 16 anchored thereon.