Novel biomarkers and diagnostic profiles for prostate cancer

A method of classifying prostate cancer in a test subject or identifying a test subject with a poor prognosis for cancer based on the expression status of a plurality of genes comprising:

• (a) providing a plurality of patient expression profiles each comprising the expression status of the plurality of genes in at least one sample obtained from each patient, wherein each of the patient expression profiles is associated with one or more cancer risk groups, wherein each cancer risk group is associated with a different cancer prognosis or cancer diagnosis, optionally wherein each patient expression profile is normalised relative to (i) the expression status of one or more normalising genes in the same patient sample, (ii) an average expression status of one or more normalising genes in a reference population and/or (iii) the status of one or more control-probes;
• (b) counting the number (n) of different cancer risk groups to which the patient expression profiles belong, optionally wherein at least one cancer risk group is associated with an absence of cancer;
• (c) applying a cumulative link model to the patient expression profiles to select a subset of one or more genes from the plurality of genes in the patient expression profile that are significantly associated with the n cancer risk groups;
• (d) inputting the expression values of the selected subset of one or more genes to a constrained continuation ratio logistic regression model comprising n modifier coefficients such that the model generates n risk scores for each patient expression profile, wherein for each patient expression profile, a risk score is provided for each of the n cancer risk groups and wherein each of the n risk scores for a given patient expression profile is associated with the clinical outcome of the corresponding cancer risk group and wherein the regression model generates regression coefficients associated with each of the selected genes based on the plurality of patient expression profiles;
• (e) providing a test subject expression profile comprising the expression status of the same selected subset of one or more genes as in step (c) in at least one sample obtained from the test subject, optionally wherein the test subject expression profile is normalised relative to (i) the expression status of one or more normalising genes in the test subject sample, (ii) an average expression status of one or more normalising genes in a reference population, and/or (iii) the status of one or more control-probes;
• (f) inputting the test subject expression profile to the constrained continuation ratio logistic regression model comprising the n modifier coefficients and gene regression coefficients generated in step (d) to generate n risk scores for the test subject expression profile, wherein each of the n risk scores for the test subject expression profile is associated with the likelihood of membership to the corresponding cancer risk group;
• and
• (g) classifying the cancer of the test subject or determining whether the test subject has a poor prognosis based on the value of a risk score associated with a poor prognosis cancer risk group for the test subject expression profile, wherein the higher the risk score associated with a poor prognosis cancer risk group, the worse the predicted outcome.

A method of classifying prostate cancer in a test subject or identifying a test subject with a poor prognosis for cancer comprising:

• (a) providing a test subject expression profile comprising the expression status of a subset of one or more genes selected by a method according to the first aspect of the invention in a sample obtained from the test subject, optionally wherein the test subject expression profile is normalised relative to (i) the expression status of one or more normalising genes in the test subject sample, (ii) an average expression status of one or more normalising genes in a reference population, and/or (iii) the status of one or more control-probes;
• (b) inputting the test subject expression profile to a constrained continuation ratio logistic regression model comprising the n modifier coefficients and gene regression coefficients generated using a method according to the first aspect of the invention, thereby generating n risk scores, wherein each of the n risk scores for a given test subject expression profile is associated with the likelihood of membership to the corresponding cancer risk group, wherein the n modifier coefficients and corresponding gene regression coefficients are generated by applying the regression model to patient expression profiles comprising the expression status of the same subset of one or more genes;
• and
• (c) classifying the cancer of the test subject or determining whether the test subject has a poor prognosis based on the value of a risk score associated with a poor prognosis cancer risk group for the test subject expression profile, wherein the higher the risk score associated with a poor prognosis cancer risk group, the worse the predicted outcome.

A method of classifying prostate cancer in a test subject or identifying a test subject with a poor prognosis for cancer comprising:

• (a) providing a test subject expression profile comprising the expression status of a plurality of the 37 genes in Table 3 in a sample obtained from the test subject, optionally wherein the test subject expression profile is normalised relative to (i) the expression status of one or more normalising genes in the test subject sample, (ii) an average expression status of one or more normalising genes in a reference population, and/or (iii) the status of one or more control-probes;
• (b) inputting the test subject expression profile to a constrained continuation ratio logistic regression model comprising the 4 modifier coefficients (Cp1, Cp2, Cp3 and the intercept) and 36 gene regression coefficients in Table 8, thereby generating 4 risk scores (PUR-1, PUR-2, PUR-3 and PUR-4), wherein the risk scores indicate the likelihood of non-cancerous tissue (PUR-1), low-risk of cancer or cancer progression (PUR-2), intermediate-risk of cancer or cancer progression (PUR-3) and high-risk of cancer or cancer progression (PUR-4) in the test subject;
• and
• (c) classifying the cancer of the test subject or determining whether the test subject has a poor prognosis based on the value of a risk score associated with a poor prognosis cancer risk group for the test subject expression profile, wherein the higher the risk score associated with a poor prognosis cancer risk group, the worse the predicted outcome.

A method of classifying prostate cancer in a test subject or identifying a test subject with a poor prognosis for cancer comprising:

• (a) providing a test subject expression profile comprising the expression status of a plurality of the 33 genes in Table 4 in a sample obtained from the test subject, optionally wherein the test subject expression profile is normalised relative to (i) the expression status of one or more normalising genes in the test subject sample, (ii) an average expression status of one or more normalising genes in a reference population, and/or (iii) the status of one or more control-probes;
• (b) inputting the test subject expression profile to a constrained continuation ratio logistic regression model comprising the 4 modifier coefficients (Cp1, Cp2, Cp3 and the intercept) and 33 gene regression coefficients in Table 9, thereby generating 4 risk scores (PUR-1, PUR-2, PUR-3 and PUR-4), wherein the risk scores indicate the likelihood of non-cancerous tissue (PUR-1), low-risk of cancer or cancer progression (PUR-2), intermediate-risk of cancer or cancer progression (PUR-3) and high-risk of cancer or cancer progression (PUR-4) in the test subject;
• and
• (c) classifying the cancer of the test subject or determining whether the test subject has a poor prognosis based on the value of a risk score associated with a poor prognosis cancer risk group for the test subject expression profile, wherein the higher the risk score associated with a poor prognosis cancer risk group, the worse the predicted outcome.

A method of classifying prostate cancer in a test subject or identifying a test subject with a poor prognosis for cancer comprising:

• (a) providing a test subject expression profile comprising the expression status of a plurality of the 29 genes in Table 5 in a sample obtained from the test subject, optionally wherein the test subject expression profile is normalised relative to (i) the expression status of one or more normalising genes in the test subject sample, (ii) an average expression status of one or more normalising genes in a reference population, and/or (iii) the status of one or more control-probes;
• (b) inputting the test subject expression profile to a constrained continuation ratio logistic regression model comprising the 4 modifier coefficients (Cp1, Cp2, Cp3 and the intercept) and 29 gene regression coefficients in Table 10, thereby generating 4 risk scores (PUR-1, PUR-2, PUR-3 and PUR-4), wherein the risk scores indicate the likelihood of non-cancerous tissue (PUR-1), low-risk of cancer or cancer progression (PUR-2), intermediate-risk of cancer or cancer progression (PUR-3) and high-risk of cancer or cancer progression (PUR-4) in the test subject;
• and
• (c) classifying the cancer of the test subject or determining whether the test subject has a poor prognosis based on the value of a risk score associated with a poor prognosis cancer risk group for the test subject expression profile, wherein the higher the risk score associated with a poor prognosis cancer risk group, the worse the predicted outcome.

A method of classifying prostate cancer in a test subject or identifying a test subject with a poor prognosis for cancer comprising:

• (a) providing a test subject expression profile comprising the expression status of a plurality of the 25 genes in Table 6 in a sample obtained from the test subject, optionally wherein the test subject expression profile is normalised relative to (i) the expression status of one or more normalising genes in the test subject sample, (ii) an average expression status of one or more normalising genes in a reference population, and/or (iii) the status of one or more control-probes;
• (b) inputting the test subject expression profile to a constrained continuation ratio logistic regression model comprising the 4 modifier coefficients (Cp1, Cp2, Cp3 and the intercept) and 25 gene regression coefficients in Table 11, thereby generating 4 risk scores (PUR-1, PUR-2, PUR-3 and PUR-4), wherein the risk scores indicate the likelihood of non-cancerous tissue (PUR-1), low risk of cancer or cancer progression (PUR-2), intermediate-risk of cancer or cancer progression (PUR-3) and high-risk of cancer or cancer progression (PUR-4) in the test subject;
• and
• (c) classifying the cancer of the test subject or determining whether the test subject has a poor prognosis based on the value of a risk score associated with a poor prognosis cancer risk group for the test subject expression profile, wherein the higher the risk score associated with a poor prognosis cancer risk group, the worse the predicted outcome.

A method of classifying prostate cancer in a test subject or identifying a test subject with a poor prognosis for cancer based on the expression status of a plurality of the genes in Table 2 comprising:

• (a) providing a plurality of patient expression profiles each comprising the expression status of the plurality of genes in at least one sample obtained from each patient, wherein each of the patient expression profiles is associated with one of four cancer risk groups, wherein each of the four cancer risk groups is associated with (i) non-cancerous tissue, (ii) low-risk of cancer or cancer progression, (iii) intermediate-risk of cancer or cancer progression and (iv) high-risk of cancer or cancer progression;
• optionally wherein each patient expression profile is normalised relative to (i) the expression status of one or more normalising genes in the same patient sample, (ii) an average expression status of one or more normalising genes in a reference population and/or (iii) the status of one or more control-probes;
• (b) applying a cumulative link model to the patient expression profiles to select a subset of one or more genes from the plurality of genes in the patient expression profile that are significantly associated with the four cancer risk groups, optionally wherein the subset of one or more genes is the list of 37 genes in Table 3, the 29 genes in Table 5 or the 25 genes in Table 6;
• (c) inputting the expression values of the selected subset of one or more genes to a constrained continuation ratio logistic regression model comprising three modifier coefficients such that the model generates four risk scores for each patient expression profile, wherein for each patient expression profile, a risk score is provided for each of the four cancer risk groups and wherein each of the four risk scores for a given patient expression profile is associated with the likelihood of membership to the corresponding cancer risk group and wherein the regression model generates regression coefficients associated with each of the selected genes based on the plurality of patient expression profiles;
• (d) providing a test subject expression profile comprising the expression status of the same selected subset of one or more genes as in step (c) in at least one sample obtained from the test subject, optionally wherein the test subject expression profile is normalised relative to (i) the expression status of one or more normalising genes in the test subject sample, (ii) an average expression status of one or more normalising genes in a reference population, and/or (iii) the status of one or more control-probes;
• (e) inputting the test subject expression profile to the constrained continuation ratio logistic regression model comprising the three modifier coefficients and gene regression coefficients generated in step (d) to generate four risk scores (PUR-1, PUR-2, PUR-3 and PUR-4) for the test subject expression profile, wherein each of the four risk scores for the test subject expression profile is associated with the likelihood of membership to the corresponding cancer risk group (i) non-cancerous tissue (PUR-1), (ii) low risk of cancer or cancer progression (PUR-2), (iii) intermediate-risk of cancer or cancer progression (PUR-3) and (iv) high-risk of cancer or cancer progression (PUR-4);
• and
• (f) classifying the cancer of the test subject or determining whether the test subject has a poor prognosis based on the value of a risk score associated with a poor prognosis cancer risk group for the test subject expression profile, wherein the higher the risk score associated with a poor prognosis cancer risk group, the worse the predicted outcome.

A method of diagnosing or testing for prostate cancer comprising determining the expression status of:

• (i) one or more genes selected from the group consisting of AMACR, AMH, ANKRD34B, APOC1, AR (exons 4-8), DPP4, ERG (exons 4-5), GABARAPL2, GAPDH, GDF15, HOXC6, HPN, IGFBP3, IMPDH2, ITGBL1, KLK2, KLK4, MARCH5, MED4, MEMO1, MEX3A, MME, MMP11, MMP26, NKAIN1, PALM3, PCA3, PPFIA2, SIM2-short, SMIM1, SSPO, SULT1A1, TDRD1, TMPRSS2:ERG, TRPM4, TWIST1 and UPK2;
• (ii) one or more genes selected from the group consisting of AMACR, AMH, ANKRD34B, APOC1, ARexons4-8, CD10, DPP4, GABARAPL2, GAPDH, HOXC6, HPN, IGFBP3, IMPDH2, ITGBL1, KLK4, MED4, MEMO1, MEX3A, MIC1, MMP26, NKAIN1, PALM3, PCA3, PPFIA2, SIM2.short, SMIM1, SSPO, SULT1A1, TDRD, TMPRSS2/ERG fusion, TRPM4, TWIST1, UPK2;
• (iii) one or more genes selected from the group consisting of AMACR, AMH, ANKRD34B, APOC1, AR (exons 4-8), CD10, DPP4, GAPDH, HOXC6, IGFBP3, IMPDH2, KLK2, KLK4, MARCH5, MED4, MEMO1, MEX3A, MIC1, MMP11, MMP26, PALM3, PCA3, PPFIA2, SIM2-short, SLC12A1, SSPO, SULT1A1, TDRD, TMPRSS2:ERG and UPK2;
• or
• (iv) one or more genes selected from the group consisting of AMACR, AMH, ANKRD34B, APOC1, ARexons4-8, CD10, DPP4, ERG 3 ex 4-5, GABARAPL2, HOXC6, HPN, IGFBP3, ITGBL1, MEMO1, MEX3A, MIC1, PALM3, PCA3, SIM2.short, SMIM1, TDRD, TMPRSS2:ERG, TRPM4, TWIST1 and UPK2,
• in a biological sample.

The method according to any preceding claim, wherein the method can be used to predict:

• (i) the volume of Gleason 4 or Gleason ≥4 prostate cancer;
• (ii) significant Intermediate- or High-risk disease (based on, for example, the D'Amico grades);
• and/or
• (iii) low risk disease that will not require treatment for 1, 2, 3, 4, 5 or more years.