Predicting prostate cancer recurrence using a prognostic model that combines immunohistochemical staining and gene expression profiling
A method that provides a graphical indication of whether a patient will have cancer recurrence uses univariate and bivariate prognostic features that were generated as part of a minimal spanning tree (MST). The method determines the values of first and second features. A first value is measured by detecting objects in an image of tissue from the cancer patient stained with a protein-specific IHC biomarker. A second value is measured using objects marked with an mRNA-specific probe biomarker detected in the tissue. The first feature is the univariate prognostic feature for cancer recurrence in a cohort of cancer patients. A combination of the first and second features is the bivariate prognostic feature for cancer recurrence in the cohort. The first and second features are elements of the MST. Nodes of the MST represent the univariate features, edges represent the bivariate features, and edge weights represent prognostic significance of bivariate features.
1. A method for providing a graphical indication of whether a patient will have a recurrence of prostate cancer, comprising:
(a) measuring a first feature to obtain a first value based on an object detected in a digital image of a first tissue slice that has been stained with a protein-specific immunohistochemical (IHC) biomarker, wherein the first tissue slice was cut from a formalin fixed paraffin embedded (FFPE) tissue sample of prostate cancer from the patient, wherein the first feature is a univariate prognostic feature for the recurrence of prostate cancer in a cohort of patients exhibiting prostate cancer, and wherein the univariate prognostic feature is taken from the group consisting of: IHC DIST CD163(+) CD8(+) STROMA, IHC DIST CD34(+) CD163(+) STROMA, IHC DIST CD163(+) CD3(+)CD8(−), CD59, MAGEC2, ATF2, NFKBIA, CXCL13, NUP107, IL12A, JAK2, SYCP1, MAPK1, TBX21, SYK, ANXA1, PMCH, MBL2, C4B, BMI1, CCL5, STAT4, IRF1, CD47, PSEN2, CD96, CTSH, IL17B, DDX43, KLRK1, LGALS3, and PTGDR2;
(b) measuring a second feature to obtain a second value based on objects marked with an mRNA-specific probe biomarker detected in a second tissue slice that was cut from the FFPE tissue sample, wherein a combination of the first feature and the second feature is a bivariate prognostic feature for the recurrence of prostate cancer in the cohort, wherein the first feature and the second feature are elements of a minimal spanning tree that is constructed from univariate and bivariate prognostic features for the recurrence of prostate cancer in the cohort, wherein nodes of the minimal spanning tree represent the univariate prognostic features, wherein edges of the minimal spanning tree represent the bivariate prognostic features, and wherein weights of the edges represent prognostic significance of the bivariate prognostic features; and
(c) displaying the graphical indication of the combination of the first value and the second value on a graphical user interface to illustrate whether the patient will have a recurrence of prostate cancer.
2. The method of claim 1 , wherein the minimal spanning tree is generated by Prim's algorithm, and the edge weights are related to a Kaplan Meier log rank p value of the bivariate prognostic features in the cohort.
3. The method of claim 1 , wherein the bivariate prognostic feature is generated by a combination of two features f1 and f2 using a formula taken from the group consisting of: f1×f2 and (1-f1)×f2 and f1×(1-f2) and (1-f1)×(1-f2).
4. The method of claim 1 , further comprising:
generating the bivariate prognostic feature by combining normalized rankings of the first feature and the second feature according to how the patients in the cohort exhibited the recurrence of prostate cancer.
5. The method of claim 1 , wherein the graphical indication illustrates a risk of prostate cancer recurrence for an individual patient based on statistics of the risks of the cohort of patients.
6. A system for providing a graphical indication of whether a patient will have a recurrence of prostate cancer, comprising:
(a) code for measuring a first feature to obtain a first value based on an object detected in a digital image of a first tissue slice that has been stained with a protein-specific immunohistochemical (IHC) biomarker, wherein the first tissue slice was cut from a formalin fixed paraffin embedded (FFPE) tissue sample of prostate cancer from the patient;
(b) code for measuring a second feature to obtain a second value based on objects marked with an mRNA-specific probe biomarker detected in a second tissue slice that was cut from the FFPE tissue sample, wherein the first feature is a univariate prognostic feature for the recurrence of prostate cancer in a cohort of patients exhibiting prostate cancer, wherein a combination of the first feature and the second feature is a bivariate prognostic feature for the recurrence of prostate cancer in the cohort, wherein the first feature and the second feature are elements of a minimal spanning tree that is constructed from univariate and bivariate prognostic features for the recurrence of prostate cancer in the cohort, wherein nodes of the minimal spanning tree represent the univariate prognostic features, wherein edges of the minimal spanning tree represent the bivariate prognostic features, and wherein weights of the edges represent prognostic significance of the bivariate prognostic features, and wherein the univariate prognostic feature is taken from the group consisting of: IHC DIST CD163(+) CD8(+) STROMA, IHC DIST CD34(+) CD163(+) STROMA, IHC DIST CD163(+) CD3(+)CD8(−), CD59, MAGEC2, ATF2, NFKBIA, CXCL13, NUP107, IL12A, JAK2, SYCP1, MAPK1, TBX21, SYK, ANXA1, PMCH, MBL2, C4B, BMI1, CCL5, STAT4, IRF1, CD47, PSEN2, CD96, CTSH, IL17B, DDX43, KLRK1, LGALS3, and PTGDR2; and
(d) a graphical user interface that displays the graphical indication of the combination of the first value and the second value so as to illustrate whether the patient will have a recurrence of prostate cancer.
7. The system of claim 6 , wherein the minimal spanning tree is generated by Prim's algorithm, and the edge weights are related to a Kaplan Meier log rank p value of the bivariate prognostic features in the cohort.
8. The system of claim 6 , wherein the bivariate prognostic feature is generated by a combination of two features f1 and f2 using a formula taken from the group consisting of: f1×f2 and (1-f1)×f2 and f1×(1-f2) and (1-f1)×(1-f2).
9. The system of claim 6 , wherein the bivariate prognostic feature is generated by combining a normalized ranking of the first feature with a normalized ranking of the second feature according to how the recurrence of prostate cancer is exhibited in the cohort of patients.
10. The system of claim 6 , wherein the graphical indication illustrates a risk of prostate cancer recurrence for an individual patient based on statistics of the risks of the cohort of patients.
11. The system of claim 6 , wherein the graphical indication of the combination of the first value and the second value is an indication of whether measurable prostate-specific antigen (PSA) will be present in the future in blood of the patient.
12. The system of claim 6 , wherein the combination of the first value and the second value are indicative of whether there will be a recurrence of measurable prostate-specific antigen (PSA) in blood of the patient.
13. A method of using both phenomic features and genomic features to indicate whether a patient will have a recurrence of a cancer, comprising:
(a) measuring a phenomic feature to obtain a first value based on an object detected in a digital image of a first tissue slice that has been stained with a protein-specific immunohistochemical (IHC) biomarker, wherein the first tissue slice was cut from a formalin fixed paraffin embedded (FFPE) tissue sample of the cancer from the patient, wherein the phenomic feature is empirically determined to be indicative of the recurrence of the cancer in a cohort of patients exhibiting the cancer, and wherein the phenomic feature is a univariate prognostic feature taken from the group consisting of: IHC DIST CD163(+) CD8(+) STROMA, IHC DIST CD34(+) CD163(+) STROMA, IHC DIST CD163(+) CD3(+)CD8(−), CD59, MAGEC2, ATF2, NFKBIA, CXCL13, NUP107, IL12A, JAK2, SYCP1, MAPK1, TBX21, SYK, ANXA1, PMCH, MBL2, C4B, BMI1, CCL5, STAT4, IRF1, CD47, PSEN2, CD96, CTSH, IL17B, DDX43, KLRK1, LGALS3, and PTGDR2;
(b) measuring a genomic feature to obtain a second value based on objects marked with an mRNA-specific probe biomarker detected in a second tissue slice that was cut from the FFPE tissue sample, wherein the genomic feature is empirically determined to be indicative of the recurrence of the cancer in the cohort of patients exhibiting the cancer, wherein a combination of the phenomic feature and the genomic feature is a combination feature for the recurrence of the cancer in the cohort, wherein a minimal spanning tree is constructed from phenomic features and combination features for the recurrence of the cancer in the cohort, wherein nodes of the minimal spanning tree represent the phenomic features, wherein edges of the minimal spanning tree represent the combination features, and wherein weights of the edges represent prognostic significance of the combination features; and
(c) displaying a graphical indication of the phenomic feature and the combination feature on a graphical user interface to illustrate whether the patient will have a recurrence of the cancer.
14. The method of claim 13 , further comprising:
generating the combination feature by combining normalized rankings of the phenomic feature and the genomic feature according to how the patients in the cohort exhibited the recurrence of the cancer.