WO2019075403A1

WO2019075403A1 - Prognostic characterization of biological sample

Info

Publication number: WO2019075403A1
Application number: PCT/US2018/055714
Authority: WO
Inventors: Christopher James Sevinsky; Fiona Mary Ginty; Alberto Santamaria-Pang; Yunxia SUI
Original assignee: General Electric Company
Priority date: 2017-10-12
Filing date: 2018-10-12
Publication date: 2019-04-18

Abstract

The present disclosure relates to prognostic characterization of biological samples. By way of example, systems and methods are provided for performing a prognostic characterization of a biological sample from a cancer patient. The method includes measuring marker characteristics of individual markers in a predefined marker panel that include at least a first marker specifically measured in tumor cells in the epithelial region and a second, different marker specifically measured in the non-tumor cells in the stromal region where the marker characteristics of the first and the second markers are, individually or in combination, clinically predictive of a risk of cancer reoccurrence. A prognostic biosignature is also developed based on the measured marker characteristics.

Description

PROGNOSTIC CHARACTERIZATION OF BIOLOGICAL SAMPLE

TECHNICAL FIELD

[0001] The subject matter disclosed herein generally relates to systems and methods for prognostic characterization of biological samples obtained from a tumor region of a patient diagnosed with a cancer. More particularly, to systems and methods for characterizing a biological sample and developing a prognostic biosignature of the biological sample, by measuring marker characteristics of cell markers in the biological sample. The marker characteristics are clinically predictive of cancer prognosis. The resulting prognostic biosignature may be used to provide a cancer prognosis, direct cancer therapy, monitor and/or predict responsiveness to a cancer therapy.

BACKGROUND

[0002] For certain cancers, surgery alone is the standard treatment option, with adjuvant therapy considered on a case-by-case basis. For example, surgery is the primary treatment option for patients who are diagnosed with CRC, a second leading cause of cancer related death in the US. Adjuvant therapy such as adjuvant chemotherapy has not been a standard treatment option because CRC is a heterogeneous disease on many levels and efficacy of adjuvant therapy greatly varies between patient populations. In early stages, >90% stage I and >80% stage II CRC patients may be cured by surgery alone, and the benefit of chemotherapy is limited (e.g. only <4% stage II patients benefits from the chemotherapy). In stage III disease, adjuvant chemotherapy treatment may benefit about 15%-20% of patients, but >50% patients relapse or develop distant metastases within five years. Therefore, it is important to determine, especially for stage II and stage III CRC patients, which sub- population of patients might have a high risk of cancer reoccurrence (e.g. cancer reoccurrence within 5 years after diagnosis), thus making them more suitable candidates for adjuvant therapy. And, for these suitable candidates for adjuvant therapy, predict responsiveness to such treatment, and/or monitor or evaluate the applied adjuvant therapy.

[0003] While some progress has been made in understanding sub-populations of CRC patients that might have a high risk of cancer reoccurrence and/or benefit from adjuvant therapy, limited clinical impact has been achieved.

[0004] One of the challenges is that knowledge and tools for selecting clinically relevant markers and their associated characteristics that would provide for stratification and identification of CRC patients, for example, the high risk and chemo responsive stage II patients, are limited. The NCCN guidelines provides guidance on high risk clinical markers in stage II (e.g. T4 staging, poorly differentiated tumor, positive margin, inadequately sampled lymph nodes (<12), histological signs of vascular, lymphatic or perineural invasion and micros atellite instability (MSI) status), which are believed to correlate with poorer prognosis. In addition, unsupervised cluster analyses of global gene expression data identified four molecular subtypes, CMS 1-4, that may help direct therapy selection.

[0005] Other reported gene-based approaches of colorectal cancer prognosis include PCR for microsatellite instability analysis, Sanger sequencing to characterize high frequency somatic mutations, or cDNA microarrays and qPCR for mRNA expression analysis. For example, two qPCR-based tests, Coloprint and Oncotype Dx Colon tests, are currently marketed for the stratification of the risk of CRC reoccurrence. Coloprint test uses a profile of 18 undisclosed gene to stratify risk of reoccurrence for stage II patients. Oncotype Dx Colon test is a 12-gene test which includes 7 prognostic genes and 5 reference genes used for normalization.

[0006] While gene-based methods might have found utility in providing CRC prognosis, morphological features and subcellular information of the cells are lost in the application of genomic techniques. For example, samples processed for DNA or mRNA analysis by PCR or sequencing methods are destroyed to form homogenate, and certain cellular information such as exact location of origin of the cells responsible for the marker characteristics observed can only be inferred indirectly.

[0007] The field is still in need of an improved system and method for prognostic characterization of biological samples obtained from tumor regions of patients diagnosed with cancer, to stratify and identify high risk and/or chemo-responsive cancer patients and to guide therapy decisions such as adjuvant chemotherapy. The improved system and method would also allow preservation of the biological sample and obtain subcellular information of clinically relevant markers in the biological sample while performing such prognostic characterization.

BRIEF DESCRIPTION

[0008] Certain embodiments commensurate in scope with the originally claimed subject matter are summarized below. These embodiments are not intended to limit the scope of the claimed subject matter, but rather these embodiments are intended only to provide a brief summary of possible embodiments. Indeed, the disclosure may encompass a variety of forms that may be similar to or different from the embodiments set forth below.

[0009] In one embodiment, a method is provided for performing a prognostic characterization of a biological sample obtained from a tumor region of a patient diagnosed with a cancer, the method comprising: providing an image data of the biological sample comprising an epithelial region having one or more tumor cells therein and a stromal region having one or more non-tumor cells therein; segmenting the epithelial and the stromal regions in the image data to identify individual cells; measuring, from the identified individual cells, one or more marker characteristics of each marker in a predefined marker panel, the marker panel comprising at least a first marker specifically measured in the tumor cells in the epithelial region and a second, different marker specifically measured in the non-tumor cells in the stromal region; developing a prognostic biosignature based on the measured marker characteristics of each marker in the predefined marker panel; wherein the marker characteristics of the first and the second markers in the marker panel are, individually or in combination, clinically predictive of a risk of cancer reoccurrence.

[0010] In another embodiment, a method is provided for determining a treatment for a patient diagnosed with a cancer, based on a prognostic biosignature of a biological sample obtained from a tumor region of the patient, the method comprising: determining if the patient has a high risk of cancer reoccurrence by: providing an image data of the biological sample comprising an epithelial region having one or more tumor cells therein and a stromal region having one or more non-tumor cells therein; segmenting the epithelial and the stromal regions in the image data to identify a plurality of single cells; measuring, from the identified single cells, one or more marker characteristics of each marker in a predefined marker panel comprising at least a first marker specifically measured in the tumor cells in the epithelial region and a second, different marker specifically measured in the non-tumor cells in the stromal region; developing the prognostic biosignature based on a combination of the measured marker characteristics of each marker in the predefined marker panel wherein the marker characteristics of the first and the second markers in the marker panel are, individually or in combination, clinically predictive of a risk of cancer reoccurrence; determining a prognostic value for the risk of cancer reoccurrence of the patient based on the prognostic biosignature; and if the prognostic value of the patient exceeds an optimal threshold, then determining the patient is in need of an adjuvant therapy, and if the prognostic value of the patient does not exceed the optimal threshold, then determining the patient is in need of a surgery treatment alone.

[0011] In another embodiment, a method is provided for performing a prognostic characterization of a biological sample obtained from a tumor region of a patient diagnosed with a stage II or stage III colorectal cancer and with an unknown prognosis, the method comprising: providing an image data of the biological sample comprising an epithelial region having one or more tumor cells therein and a stromal region having one or more non-tumor cells therein; segmenting the epithelial and the stromal regions in the image data to identify single cells; measuring, from the identified single cells, one or more marker characteristics of each marker in a marker panel comprising COX2, beta-catenin, p21, NaKATPase, phospho-ERK and CD8 markers; developing a prognostic biosignature by selecting a combination of the one or more marker characteristics of each marker wherein the marker characteristics of each marker in the marker panel are, individually or in combination, clinically predictive of a risk of cancer reoccurrence.

BRIEF DESCRIPTION OF THE DRAWINGS

[0012] These and other features, aspects, and advantages of the present disclosure will become better understood when the following detailed description is read with reference to the accompanying drawings in which like characters represent like parts throughout the drawings, wherein:

[0013] FIG. 1 is a block diagram illustrating an embodiment of a system for assessing a biological sample according to an embodiment of the present disclosure;

[0014] FIG. 2A is a flow diagram illustrating an embodiment of a method for obtaining an image data of a sample, according to an embodiment of the present disclosure;

[0015] FIG. 2B is a flow diagram illustrating an embodiment of a method for processing of image data of a biological sample, according to an embodiment of the present disclosure;

[0016] FIGs. 3A-3H show a panel of epithelial, immune and stromal cell markers imaged on the same field of view (FOV), where the markers are, respectively, E- Cadherin (FIG. 3A), SMA (FIG. 3B), CD20 (FIG. 3C), CD3 (FIG. 3D), CD68 (FIG. 3E), CD8 (FIG. 3F), CD31 (FIG. 3G), and a combination of all markers from FIGs. 3A-3G (FIG. 3H), according to an embodiment of the present disclosure;

[0017] FIGs. 4A-4D are plots of Log2 mean expression level of stromal cell markers in stromal cells of tissue samples of the whole cohort, stage I, stage II, and stage III patients, in the tissue microarray (TMA). The markers used are CD3+ (FIG. 4A), CD8+ (FIG. 4B), CD20+ (FIG. 4C), and CD68+ (FIG. 4D);

[0018] FIGs. 4E-4H show % of stromal cells positive for the respective cell markers used in FIGs. 4A-4D;

[0019] FIG. 5A a plot of possibility of survival vs. days post-diagnosis for 7 clusters of marker features identified with a multivariate clustering analysis for a CRC cancer cohort; according to embodiments of the present disclosure;

[0020] FIG. 5B shows a comparison of a plot of possibility of survival vs. days post- diagnosis between cluster 3 and cluster 7 of FIG. 5A;

[0021] FIG. 5C shows a comparison of a plot of possibility of survival vs. days post- diagnosis between cluster 4 and cluster 7 of FIG. 5A;

[0022] FIG. 5D shows a comparison of a plot of possibility of survival vs. days post- diagnosis between cluster 3 and cluster 6 of FIG. 5A;

[0023] FIG. 6A shows a plot of possibility of survival for a CRC patient whole cohort vs. days post-diagnosis using a stromal marker panel including markers specifically measured in non-tumor cells in the stromal region. The marker panel used in FIG. 6A includes CD3, CD8 and Ρ13Κρ110α, according to embodiments of the present disclosure;

[0024] FIG. 6B shows a plot of possibility of survival for stage I patients only in the patient cohort of FIG. 6A vs. days post-diagnosis based on the stromal marker panel used in FIG. 6A;

[0025] FIG. 6C shows a plot of possibility of survival for stage II patients only in the patient cohort of FIG. 6A vs. days post-diagnosis based on the stromal marker panel used in FIG. 6A; [0026] FIG. 6D shows a plot of possibility of survival for stage III patients only in the patient cohort of FIG. 6 A vs. days post-diagnosis based on the stromal marker panel used in FIG. 6A;

[0027] FIG. 6E shows a multivariate modeling result with hazard ratio and p- value, using the stromal marker panel of FIG. 6A, and clinical variables (age, stage, and tumor grade);

[0028] FIG. 7A shows a plot of possibility of survival for a CRC patient whole cohort vs. days post-diagnosis based on an epithelial model using an epithelial marker panel including markers specifically measured in cancer cells in the epithelial region. The marker panel used in FIG. 7A includes COX2, beta-catenin, p21, Na⁺K⁺ATPase (NaKATPase), and phospho-ERK (pERK), according to embodiments of the present disclosure;

[0029] FIG. 7B shows a plot of possibility of survival for stage I patients only in the patient cohort of FIG. 7A vs. days post-diagnosis based on the epithelial marker panel used in FIG. 7A;

[0030] FIG. 7C shows a plot of possibility of survival for stage II patients only in the patient cohort of FIG. 7A vs. days post-diagnosis based on the epithelial d marker panel used in FIG. 7A;

[0031] FIG. 7D shows a plot of possibility of survival for stage III patients only in the patient cohort of FIG. 7A vs. days post-diagnosis based on the epithelial marker panel used in FIG. 7A;

[0032] FIG. 7E shows a multivariate modeling result with a hazard ratio and p-value, using the epithelial marker panel of FIG. 7A, and clinical variables (age, stage, and tumor grade);

[0033] FIG. 8A shows a plot of possibility of survival for a CRC patient whole cohort vs. days post-diagnosis based on a combined epithelial and stromal model using a combined marker panel including markers specifically measured in cancer cells in the epithelial region and markers for non-tumor cells in the stromal region. The marker panel used in the plot includes COX2, beta-catenin, p21, NaKATPase, phospho-ERK, and CD8, according to embodiments of the present disclosure;

[0034] FIG. 8B shows a plot of possibility of survival for stage I patients only in the patient cohort of FIG. 8A vs. days post-diagnosis based on the combined marker panel used in FIG. 8A;

[0035] FIG. 8C shows a plot of possibility of survival for stage II patients only in the patient cohort of FIG. 8A vs. days post-diagnosis based on the combined marker panel used in FIG. 8A;

[0036] FIG. 8D shows a plot of possibility of survival for stage III patients only in the patient cohort of FIG. 8A vs. days post-diagnosis based on the combined marker panel used in FIG. 8A;

[0037] FIG. 8E shows a multivariate modeling result with hazard ratio and p- value, using the combined marker panel used in FIG. 8A, and clinical variables (age, stage, and tumor grade);

[0038] FIG. 9A shows a plot of patient risk scores derived from the stromal and the epithelial models; FIG. 9B shows a plot of patient risk scores derived from the stromal and the combined models; FIG. 9C shows a plot of patient risk scores derived from the epithelial and the combined models; FIG. 9D shows a plot of patient risk scores derived from a stromal model; FIG. 9E shows a plot of patient risk scores derived from the epithelial model; FIG. 9F shows a plot of patient risk scores derived from the combined stromal and epithelial model, and FIG. 9G shows pairwise correlations of patient risk scores (with p value) derived from epithelial, stromal and the combined models, according to embodiments of the present disclosure;

[0039] FIG. 10A illustrates a characterization of a tissue sample obtained from a tumor region of a stage II CRC patient. E-Cadherin is used as markers for the epithelial region of the tissue sample. CD3 and CD8 are used for non-tumor cells in the stromal region, according to embodiments of the present disclosure;

[0040] FIG. 10B illustrates differences in morphology and distribution of markers in the tissue samples of patients at different stages, according to embodiments of the present disclosure.

DETAILED DESCRIPTION

[0041] Embodiments of the present disclosure relate to system and methods for prognostic characterization of biological samples obtained from a tumor region of a patient diagnosed with a cancer. Techniques disclosed in the present disclosure may be used to assess the sample and based on the assessment, develop a prognostic biosignature of the sample that provides a cancer prognosis (e.g., presence or absence of a cancer reoccurrence), direct cancer therapy, monitor and/or predict responsiveness to a cancer therapy (i.e. adjuvant chemotherapy) based on the developed prognostic biosignature. The system and method of the current disclosure also allows for preserving of the biological sample and obtaining subcellular information of clinically relevant markers in the biological sample while performing such prognostic characterization.

[0042] The present techniques may be performed in situ, for example, in intact organ or tissue or in a representative segment of an organ or tissue. In some embodiments, in situ analysis of targets may be performed on cells derived from a variety of sources, including an organism, an organ, tissue sample, or a cell culture. In situ analysis provides contextual information that may be lost when the target is removed from its site of origin. Accordingly, in situ analysis of targets describes analysis of target-bound probe located within a whole cell or a tissue sample, whether the cell membrane is fully intact or partially intact where target-bound probe remains within the cell. Furthermore, the methods disclosed herein may be employed to analyze targets in situ in cell or tissue samples that are fixed or unfixed. In situ techniques, such as those provided herein, permit assessment of cell profiles in a particular microenvironment. Such methods may be in contrast to techniques in which the cell microenvironments are disrupted to conduct analysis.

[0043] The present techniques provide systems and methods for image analysis. In certain embodiments, it is envisioned that the present techniques may be used in conjunction with previously acquired images, for example, digitally stored images, in retrospective studies. In other embodiments, the images may be acquired from a physical sample. In such embodiments, the present techniques may be used in conjunction with an image acquisition system. An exemplary imaging system 10 capable of operating in accordance with the present technique is depicted in FIG. 1. Generally, the imaging system 10 includes an imager 12 that detects signals and converts the signals to data that may be processed by downstream processors. The imager 12 may operate in accordance with various physical principles for creating the image data and may include a fluorescent microscope, a bright field microscope, or devices adapted for suitable imaging modalities. In general, however, the imager 12 creates image data indicative of a biological sample including a population of cells 14, shown here as being multiple samples on a tissue micro array, either in a conventional medium, such as photographic film, or in a digital medium. As used herein, the term "biological material" or "biological sample" refers to material obtained from, or located in, a biological subject, including biological tissue or fluid obtained from a subject. Such samples can be, but are not limited to, body fluid (e.g., blood, blood plasma, serum, or urine), organs, tissues, biopsies, fractions, and cells isolated from, or located in humans, such as cancer patients. Biological samples and/or biological materials also may include sections of the biological sample including tissues (e.g., sectional portions of an organ or tissue).

[0044] In some embodiments, the biological sample used may be a tissue sample obtained from a tumor region of a patient diagnosed with a cancer. In one embodiment, the biological sample contains an epithelial region having one or more tumor cells therein and a stromal region having one or more non-tumor cells therein. The patient diagnosed with the cancer may have an unknown prognosis or a known prognosis based on a previous cancer treatment. The biological samples may be imaged as part of a slide.

[0045] The imager 12 operates under the control of system control circuitry 16. The system control circuitry 16 may include a wide range of circuits, such as illumination source control circuits, timing circuits, circuits for coordinating data acquisition in conjunction with sample movements, circuits for controlling the position of light sources and detectors, and so forth. In the present context, the system control circuitry 16 may also include computer-readable memory elements, such as magnetic, electronic, or optical storage media, for storing programs and routines executed by the system control circuitry 16 or by associated components of the system 10. The stored programs or routines may include programs or routines for performing all or part of the present technique.

[0046] Image data acquired by the imager 12 may be processed by the imager 12, for a variety of purposes, for example to convert the acquired data or signal to digital value provided to data acquisition circuitry 18. The data acquisition circuitry 18 may perform a wide range of processing functions, such as adjustment of digital dynamic ranges, smoothing or sharpening of data, as well as compiling of data streams and files, where desired.

[0047] The data acquisition circuitry 18 may also transfer acquisition image data to data processing circuitry 20, where additional processing and analysis may be performed. Thus, the data processing circuitry 20 may perform substantial analyses of image data, including ordering, sharpening, smoothing, feature recognition, and so forth. In addition, the data processing circuitry 20 may receive data for one or more sample sources, (e.g. multiple wells of a multi-well plate). The processed image data may be stored in short or long term storage devices, such as picture archiving communication systems, which may be located within or remote from the imaging system 10 and/or reconstructed and displayed for an operator, such as at the operator workstation 22. [0048] In addition to displaying the reconstructed image, the operator workstation 22 may control the above-described operations and functions of the imaging system 10, typically via an interface with the system control circuitry 16. The operator workstation 22 may include one or more processor-based components, such as general purpose or application specific computers 24. In addition to the processor-based components, the computer 24 may include various memory and/or storage components including magnetic and optical mass storage devices, internal memory, such as RAM chips. The memory and/or storage components may be used for storing programs and routines for performing the techniques described herein that are executed by the operator workstation 22 or by associated components of the system 10. Alternatively, the programs and routines may be stored on a computer accessible storage and/or memory remote from the operator workstation 22 but accessible by network and/or communication interfaces present on the computer 24. The computer 24 may also comprise various input/output (I/O) interfaces, as well as various network or communication interfaces. The various I/O interfaces may allow communication with user interface devices, such as a display 26, keyboard 28, mouse 30, and printer 32, that may be used for viewing and inputting configuration information and/or for operating the imaging system 10. The various network and communication interfaces may allow connection to both local and wide area intranets and storage networks as well as the Internet. The various I/O and communication interfaces may utilize wires, lines, or suitable wireless interfaces, as appropriate or desired.

[0049] More than a single operator workstation 22 may be provided for an imaging system 10. For example, an imaging scanner or station may include an operator workstation 22 which permits regulation of the parameters involved in the image data acquisition procedure, whereas a different operator workstation 22 may be provided for manipulating, enhancing, and viewing results and reconstructed images. Thus, the image processing, segmenting, and/or enhancement techniques described herein may be carried out remotely from the imaging system, as on completely separate and independent workstations that access the image data, either raw, processed or partially processed and perform the steps and functions described herein to improve the image output or to provide additional types of outputs (e.g., raw data, intensity values, cell profiles).

[0050] The computer analysis method 40 used to analyze images is shown in FIG. 2A. It should be understood that the method 40 may also be used with stored images that are retrospectively analyzed. Typically, one or more images of the same sample may be obtained or provided. In step 42, the biological sample is prepared by applying a plurality of probes. In one embodiment, the probes are applied in a sequential manner. The probes may include probes for identifying cellular regions such as the cell membrane, cytoplasm and nuclei. In such an embodiment, a mask of the stromal region may be generated, and using curvature and geometry based segmentation, the image of the compartment marker or markers is segmented. For example, the membrane and nuclear regions of a given tumor region may be demarcated. The cytoplasm may be designated as the area between the membrane and nucleus or within the membrane space. Any number and type of morphological markers for segmentation may be used.

[0051] FIG. 2A is a flow diagram of one embodiment of a technique 40 for providing image data of a biological sample. At step 42, one or more probes is applied to the biological sample 14. The probe may be applied as part of a multi-molecular, multiplexing imaging technology such as the GE Healthcare Cell-DIVE™ platform. For example, the probe may be applied, and an image of the probe may be acquired at step 44 by the imaging system 10. The image may be in the form of image data that is representative of the probe bound to the target or marker of interest on the sample. Rather than use a separate slide or section to then assess a second probe relative to the first probe, e.g., via image registration techniques on the acquired images, the probe may be inactivated, e.g., via a chemical inactivation, at step 46 before application of a subsequent second probe. The method 40 then returns to step 42 for sequential probe application, image acquisition, and probe inactivation until all the desired probes have been applied. In some embodiments, the disclosed techniques may be used in conjunction with any number of desired probes, including 2, 3, 4, 5, 6, 7, 8, 9, 10, 20, 30, 40, 50, 60, 70 or more probes per sample. Accordingly, the acquired image data 48 represents a plurality of images, and individual images within the image data may be associated with a detected intensity of a particular probe. In one embodiment, the sequential probe imaging may be performed as disclosed in U.S. Patent Nos. 7,639,125 and 9,201,063 which are incorporated by reference herein in their entirety for all purposes. During the sample handling, certain quality control steps may be taken to account for marker staining variability. For example, replicates may be stained.

[0052] FIG. 2B is a flow diagram of one embodiment of a technique 50 for image processing of image data of a biological sample as provided herein. Steps in FIG. 2B may be performed after the steps of FIG. 2A, or as a separate work flow. At step 52, image data 48 is segmented to identify individual cells. For example, a biological sample may be obtained from a tumor region of a cancer patient and includes an epithelial region having one or more tumor cells therein and a stromal region having one or more non-tumor cells therein, the image data of the biological sample may be segmented into epithelial and stromal regions to identify individual cells such as tumor cells in the epithelial region and non-tumor cells in the stromal region. In some embodiments, respective probes including probes specific for cell markers in the epithelial tumor cells or stromal cells may be used. In some embodiments, probes may also include probes specific for segmentation markers and morphological markers, e.g., epithelium probes, membrane probes, cytoplasmic probes, and/or nuclear probes. Accordingly, the image data 48 may include information to facilitate segmenting as well as information to identify cell types. The method 50 may optionally include one or more quality control features to exclude poorly stained markers or poorly segmented spots. Further, the identification of individual cells may include quality control features such as thresholds to exclude certain cells based on staining or signal quality. At step 54, one or more marker characteristics (e.g. intensity, distribution, location within subcellular compartments, or a ratio of intensities between subcellular compartments, etc.) for individual markers are measured. The individual markers may be markers in a predefined marker panel comprising at least a first marker specifically measured in the epithelial tumor cells and a second, different marker specifically measured in the non- tumor cells in the stromal region. In some embodiments, a first marker specifically measured in the epithelial tumor cells means that for first marker, only measurements of marker characteristics of the first marker that are associated with or measured in the epithelial tumor cells are obtained or selected for analysis. And a second, different marker specifically measured in the non-tumor cells in the stromal region means that for the second marker, only measurements of marker characteristics of the second marker that are associated with or measured in the non-tumor cells in the stromal region are obtained or selected for analysis.

[0053] The method 50 may include developing a prognostic biosignature based on the marker characteristics measured at step 54. The marker characteristics may be selected to be clinically predictive of cancer reoccurrence. In some embodiments, the marker characteristics of the first and the second markers in the predefined marker panel are, individually or in combination, clinically predictive of a risk of cancer reoccurrence.

[0054] In some embodiments, the predefined marker panel is developed by: measuring marker characteristics of a plurality of markers in reference samples obtained from respective tumor regions of a group of cancer patients with known prognosis conditions, applying a multivariate model to the marker characteristics of the plurality of markers; and developing the predefined marker panel comprising at least the first marker and the second marker, wherein their respective marker characteristics are, individually or in combination, clinically associated with known prognosis of each patient. In one embodiment, the multivariate model is a clustering model. In another embodiment, the multivariate model is a combined epithelial model and stroma model, as described in detail in the non-limiting examples. It is to be understood that other multivariate models known to one skilled person in the art can be used and are within the scope of the present disclosure.

[0055] In some embodiments, the predefined marker panel comprises at least two markers selected from a marker group comprising 4E-BP1, Akt, Albumin, ALDH1, bcl-2, CA9, CD163, CD20, CD24, CD3,CD31, CD44v6, CD68, CD79, CD8, Claudinl, Cleaved-Caspase3, c-Myc, Collagen IV, COX2, Cyclin B l, Cyclin Dl, Cytokeratin 19, Cytokeratinl5, E-Cadherin, EGFR, EPCAM, ERK1/2, estrogen receptor-a, EZH2, Fibronectin, FOXOl, FOX03a, GLUT1, HER2, IHH, Lamin A/C, MET, MLH1, MSH2, Na⁺K⁺ATPase, NDRG1, p21, p53, pan-keratin, pan-keratin 8/18, PCNA, pEGFR, phospho-4E-BPl, phospho-GSK3a, phospho-GSK3 , phospho- MAPKAPK2, phospho-MET, phospho-NDRGl, phospho-p38MAPK, phospho-S6 (Ser 235/236), phsopho-ERKl/2, PDKpl lOa, PTEN, S100, S6, SMA, TKLP1, Vimentin, Wnt5a, xCT, β-actin, β-Catenin, β-Catenin, and β-tubulin.

[0056] In some embodiments, the predefined marker panel comprises at least two markers selected from a marker group comprising CD8, NaKATPase, Claudinl, CD3, pMAPKAPK2, COX2, CA9, phospho-ERK, β-catenin, p21, MSH2, E-Cad, Glutl, p53, CD20, MLH1, LaminA_C and pERK.

[0057] In some embodiments, the predefined marker panel comprises COX2, beta- catenin, p21, NaKATPase, phospho-ERK and CD8.

[0058] In some embodiments, the first marker of the predefined marker panel is selected from a marker group comprising NaKATPase, Claudinl, MSH2, pMAPKAPK2, COX2, CA9, E-Cad, Glutl, pERK, β-catenin, p21, or any combinations thereof.

[0059] In some embodiments, the second marker is selected from a marker group comprisingCD31, SMA, Collagen IV, Fibronectin, CD3, CD8, CD68, Lamin A/C, vimentin, SI 00, or any combinations thereof.

[0060] Optionally, the method 50 may also provide an output related to the prognostics characterization, for example via a display associated with the system 10 or stored in a memory of the system 10. The output may include one or more of a histogram, boxplot, density plot, violin plots, or numerical values corresponding to such plots. In one embodiment, the output may be a prognostic signature of the sample. The prognostic signature may include a total number of cells in the sample and/or in the epithelial and stromal regions, a total number of each type of cells (epithelial tumor cells, stromal cells, immune cells etc.) in the sample and/or in the epithelial and stromal regions, location information, distribution, and/or a representation of marker expression level of various cell types in the sample and/or in the epithelial and stromal regions. In one embodiment, the output may be marker expression average for epithelial, stromal, and whole image regions. The output may also include metrics such as mean, median, skewness, a standard deviation, or coefficient of variation of the marker distribution or a binary classification of marker being present or absent based on a user defined threshold. In one embodiment, the output may be in a form of a risk score or a prognostic score that include coefficient values for weighted contributions from measured value of each marker characteristics in the prognostic biosignature.

[0061] In some embodiments, the prognostic biosignature further includes one or both of a molecular subtype characteristic based on a genomic analysis, and a patient clinical characteristic selected from a group comprising histologic types, stage, age, gender, cancer grade, or any combinations thereof. In one embodiment, the technique may be used to assess a sample from a cancer patient with unknown prognosis condition by obtaining a risk score derived from the prognostic biosignature and comparing the obtained risk score with a predetermined risk score threshold value developed with reference prognostic biosignatures of samples from a group of cancer patients with known prognosis conditions. The reference prognostic biosignatures may be stored in the memory of the system 10. In such an example, the prognostic biosignature may be used for providing a prognosis. In another embodiment, assessment of the prognostic biosignature may be used to determine if a patient is in need of an adjuvant therapy, for example, an adjuvant chemotherapy or an adjuvant immunotherapy. For example, if a patient is assigned with a high risk of cancer reoccurrence based on an assessment of the prognosis biosignature, it may be determined that the patient is in need of an adjuvant therapy. The patient may be assigned with the high risk of cancer reoccurrence when the assessment of the prognostic biosignature provides a risk or prognostic score exceeding a predetermined threshold value. The predetermined threshold value may be determined based on assessment of prognostic biosignatures of reference samples from a group of cancer patients with known prognosis conditions. In some embodiments, "low risk" may be assigned to patients without cancer recurring within 5 years post diagnosis, and "high risk" may be assigned to patients with cancer recurring within 5 years post diagnosis.

[0062] In one embodiment, the cancer patient with an unknown prognosis condition has not been subjected to an adjuvant therapy. In another embodiment, the patient is being subjected or has been subjected to an adjuvant therapy. In another embodiment, the patient is diagnosed with a stage II or a stage III colorectal cancer (CRC).

[0063] In one embodiment, the prognostic biosignature developed before and after a cancer therapy may be used to determine if the therapy is achieving a favorable outcome. In another embodiment, prognostic biosignature is developed to predict a patient response to a given therapy and determine if the given therapy is appropriate for the patient, and/or is in need of adjustments. In another embodiment, the prognostic biosignature may be used to design personalized treatment monitoring schedules. For example, for patients who have received a treatment including adjuvant therapy, determine if the patient has a high risk of post-treatment cancer reoccurrence, and accordingly, in need of more frequent post-treatment monitoring.

EXAMPLES

Study Design and Participants

[0064] The colorectal cancer cohort was collected from the Clearview Cancer Institute of Huntsville Alabama from 1993 until 2002, with 747 patient tumor samples collected as paraffin embedded specimens. Tissue microarrays (TMA) were constructed to facilitate large scale biomarker analysis, and cores from patient samples were distributed across three slides. The median follow-up time of patients in the cohort is 4.2 years, with a maximum of over ten years. Stage II patients are included at 38% of the cohort, with stage I and stage III patients included at 26.5% and 35%, respectively. The cohort is relatively balanced in gender, and spans a wide range of ages, tumor histology, and anatomical locations. Summary of cohort clinical characteristics is provided in Table 1. After quality control measures were taken to remove specimens with insufficient tumor fraction in the representative TMA core, 694 specimens remained for analysis, in which 450 patients were treated with surgery alone, and the remaining 244 patients were treated with 5- fluorouracil (5-FU) based chemotherapy regimens.

Table 1. Cohort clinical characteristics

Multiplexed analysis of TMA

[0065] Sixty-one antibodies were included in this study. The antibodies used include4E-BPl, Akt, Albumin, ALDHl, androgen receptor, bcl-2, CA9, CD163, CD20, CD24, CD3, CD31, CD44v6, CD68, CD79, CD8, Claudinl, Cleaved-Caspase3, c-Myc, Collagen IV, COX2, Cyclin Bl, Cyclin Dl, Cytokeratin 19, Cytokeratinl5, E- Cadherin, EGFR, EPCAM, ERK1/2, estrogen receptor-a, EZH2, Fibronectin, FOXOl, FOX03a, GLUT1, HER2, IHH, Ki67, Lamin A/C, MET, MLH1, MSH2, Na+K+ATPase (NaKATPase), NDRG1, p21, p53, pan-keratin, pan-keratin 8/18, PCNA, pEGFR, phospho-4E-BPl, phospho-GSK3a, phospho-GSK3 , phospho- MAPKAPK2, phospho-MET, phospho-NDRGl, phospho-p38MAPK, phospho-S6 (Ser 235/236), phsopho-ERKl/2, PDKpl lOa, PTEN, S100, S6, smooth muscle actin alpha (aSMA), TKLP1, Vimentin, Wnt5a, xCT, β-actin, β-Catenin, and β-tubulin.

[0066] TMA slides were analyzed using a biological multiplexing immunofluorescent workflow as previously described. This allows iterative sequential staining of FFPE histopathology specimens and quantification of signal intensity in single 5 micrometer (μιη) sections at cellular/subcellular levels. All samples were stained and imaged in a single batch for 2-3 biomarkers at a time, and 4',6-diaminidino- 2-phenylindole (DAPI).

[0067] Stained nuclei enabled reimaging of identical regions of the specimen and image alignment (registration) prior to downstream image analysis steps. Registered images were flattened to normalize and eliminate optical intensity gradients, and auto- fluorescence (AF) removed by subtracting unstained from stained images. Images were segmented into epithelial and nonepithelial (stromal) regions, differentiated by epithelial E cadherin (E-Cad) staining. It is to be understood that other markers known to a skilled person for segmentation of epithelial and nonepithelial regions may also be used. This was followed by segmentation of individual cells in both the epithelium and stroma. Epithelial cells were segmented using cell membrane staining (Na+K+ATPase) to delineate cell borders and membrane regions, the cytosolic ribosomal protein small subunit S6 (S6) for cytoplasm, and DAPI stain for nuclear regions. Due to the complex morphologies of many stromal cell types the stromal cells were segmented based on nuclear stain alone, for example, DAPI. However, other nuclear stains known to a skilled person may be used as well. Marker expression level and standard deviation was subsequently quantified in each cell, including the whole cell and subcellular compartments of epithelial cells.

[0068] FIGs. 3A-3H show a non-limiting example of a panel of cells in the epithelial and stromal regions imaged on the same field of view (FOV) and characterized with respective cell markers. The markers used are, E-Cadherin (FIG. 3A), SMA (FIG. 3B), CD20 (FIG. 3C), CD3 (FIG. 3D), CD68 (FIG. 3E), CD8 (FIG. 3F), CD31 (FIG. 3G), and a combined image (FIG. 3H) of all markers from FIGs. 3A-3G.

Quality control and data normalization

[0069] Following the single cell segmentation, several data pre-processing steps may be optionally conducted including, but not limited to, one or more steps of cell filtering, spots exclusion, log2 transformation and slide to slide normalization. In some embodiments, cells having a subcellular compartment size >10 pixels were included for downstream analysis. During the multiplexing process, the sample might be damaged, folded or lost. Image registration issues can also result in a reduced quality cell data. Therefore, a tissue quality index based on the correlation of the image with DAPI was calculated for each cell at each round. Only those cells whose quality index equals to or greater than 0.9 (meaning that at least 90% of the cells overlapped with DAPI) were included. All the slides for all the biomarkers were adjusted to a common exposure time per channel. The data were then log2 transformed to generate a more normal distribution. A median normalization that equalizes the median cell intensities of all slides was performed to remove slide to slide non-biological variability. This normalization assumes that the median cell intensity on each slide is the same across all three slides. This is reasonable to assume as each TMA slide has around 250 patients in this study and 250 subjects represent good random sample of population.

Markers and marker characteristics for statistical analysis

[0070] Once the single cell segmentation data is quality controlled, normalized and transformed, cell marker intensity is quantified per patient. The marker intensities may form a distribution for the many cells of the patient. In some embodiments, metrics for quantifying cell marker intensities may include, but not limited to, overall membrane intensity, cytoplasmic intensity, nuclear intensity, a ratio of membrane to the average of cytoplasm and nuclei compartments, a ratio of nuclei to the average of membrane and cytoplasm compartments, a ratio of cytoplasm to the average of membrane and nuclei compartments, or any combinations thereof. In one embodiment, 54 probes selected from the 61 antibody probes and 6 metrics are used for the characterization of the epithelium region, resulting in 54*6=324 individual marker features. In the stromal regions, overall cell marker intensity is determined, and average is calculated for each patient (total 54 stromal features). In total, 378 marker features/characteristics were quantified and evaluated for correlation with risk of reoccurrence.

[0071] In some embodiments, markers for the immune cells in the stromal regions may be included in the statistical prognosis analysis. In one embodiment, a machine learning approach may be developed to identify or classify cells as one of the cell classes selected from CD3+, CD8+, CD20+, CD68+, and negative. Experts annotate the cells according to the above cell classes, and the resulting annotations were used to train a linear support vector machines model. The trained model was then applied to predict class of each cell from all samples in the study. Within the training set, the accuracy of the model was 99% for CD20+, CD8+ and CD68+ cells, while the CD3+ cells were classified with 97% accuracy. Properties, such as density of the diverse cell types (e.g. cells of interest/total stromal cells), may be analyzed in each patient's tissue sample. Similar trends were observed, in either a continuous or discrete scoring system, with smaller immune cell populations as stage of the disease increased. It is to be noted that while the machine learning approach is described here, other methods known to a skilled person in the art may be used to identify and evaluate the immune cell and markers.

[0072] FIGs. 4A-4D are exemplary plots of Log2 mean expression level of various cell markers of cells in the stromal region of tissue samples from the whole cohort, stage I, stage II, and stage III CRC patients in a tissue microarray (TMA). The markers used are general T-lymphocyte marker CD3 (FIG. 4A), the cytotoxic T-lymphocyte marker CD8 (FIG. 4B), the B-lymphocyte marker CD20 (FIG. 4C), and the macrophage marker CD68 (FIG. 4D). FIGs. 4E-4H show % of cells in the stromal region that stained positive (i.e. measured expression level exceeding a predetermined threshold).

[0073] In FIGs. 4A-4D, for each cell marker used, the marker expression levels or intensities of the cells in the stromal region spans a range of at least ten-fold (FIGs 4A- 4D, cohort data). More particularly, in some embodiments, marker expression level decreases as the stage advances (e.g. from stage I to III). For example, in FIG. 4B, CD8 expression level decreases as the stage advances from stage I to III. In some embodiments, a sub-population may exist within a specific stage that may have lower risk of cancer reoccurrence compared to the rest of the population within the same stage. For example, for stages II CRC subjects in FIG. 4B, a portion of the plot as indicated by an arrow corresponds to cases with elevated CD8 stromal expression significantly deviate from the rest of the population. This indicates that there is a sub- population within stage II patients that has higher cytotoxic T-lymphocyte marker CD8 expression. It is therefore hypothesized that the higher expression of CD8 would lead to lower risk of cancer reoccurrence and may be clinically predictive of cancer prognosis and used to identify sub-populations within stage II patients for chemotherapy. The markers used in FIGs. 4A-4D are included in a subsequent statistical analysis in identifying markers and their associated characteristics clinically predictive of cancer prognosis.

Statistical analysis models

[0074] In some embodiments, Cox Proportional Hazards model is used to assess the association of marker characteristics with the cancer reoccurrence in patients. Cox Proportional Hazards model may be applied in a univariate and/or a multivariate analysis. [0075] In some embodiments, each measured marker characteristic subjected to the model analysis is inputted as a variable in the Cox Proportional Hazards model, and the model output provides a regression coefficient (coef) for each marker characteristic. Further, each regression coefficient is associated with a sign. A positive sign indicates that the risk of cancer reoccurrence is higher, for patients with higher values of the corresponding marker characteristic. For example, in one embodiment, a higher intensity measured for NaKATPase marker is associated with a higher risk when its corresponding coefficient is positive.

[0076] The model may also provide exponentiated coefficient (exp(coef)), referred to as a hazard ratio (HR), for each marker characteristic. Upper and lower 95% confidence intervals for the hazard ratio may also be provided. Table 2 shows an example of a multivariate modeling result, according to one embodiment of the present disclosure.

Table 2. A multivariate modeling result based on Cox Proportional Hazards model

[0077] In some embodiments, Cox Proportional Hazards model may provide, as an output, a risk score for every patient subjected to the model analysis. In one embodiment, the risk score may be the linear term in the cox model as represented in eq. (1) below:

[0078] Let Xi= {Xz^'l, ...Xz^'p} be the realized or measured values of the covariates for subject i, and b= {bl, ...bp} be the values of corresponding coefficients for each realized values of the covariates. The Cox proportional hazards model can be written as eq. (1):

[0079] where h(t) is the expected hazard at time t, h0(t) is the baseline hazard and represents the hazard when all of the predictors (or independent variables) XI, X2, Xp are equal to zero. The linear term of the model may be defined as (bXz).

[0080] In one embodiment, one or more marker characteristics of markers are covariates in the model. The linear term may be determined based on measured value of the one or more marker characteristics, and corresponding coefficients for each measured marker characteristics. The risk score may be ranked. Based on the risk scores, estimation of a plurality of paired value of sensitivity and specificity may be determined. Each paired value of sensitivity and specificity may correspond to an applied threshold selected from a plurality of threshold value. A cut off threshold value may be determined or selected, from the plurality of threshold, in which the cut off threshold value corresponds to the maximum J as described in the paragraph below, according to one embodiment.

[0081] A ROC curve may be developed to estimate the sensitivity and specificity corresponding to every risk score value. In one embodiment, the Youden's J statistic, in which J is computed as J= Sensitivity+ Specificity- 1 , may be applied to determine or select an optimal threshold value that corresponds to the maximum J. It is to be understood that the embodiments described in the present disclosure are non-limiting examples, and other models and statistical tools known to a skilled person in the art may also be used in determining a risk score(s) and/or an optimal threshold value.

[0082] In some embodiments, a sample from a cancer patient with unknown prognosis condition may be assessed by obtaining a risk score derived from the prognostic biosignature as described throughout the disclosure and comparing the obtained risk score with a risk score threshold value developed with reference prognostic biosignatures of samples from a group of cancer patients with known prognosis conditions.

[0083] In some embodiments, the Kaplan- Meier curve is examined for the entire no chemo cohort, stage I, stage II and stage III patients, respectively, using the same Youden's J cutoff, of 5-year reoccurrence.

[0084] In some embodiments, the most suitable model is selected with highest AUC, smallest AIC and best split of KM curves. It is to be understood that the models described in the present disclosure are non-limiting examples, and other statistical analysis model known to a skilled person in the art may also be used.

[0085] In one embodiment, to identify markers and marker characteristics that are clinically associated with cancer reoccurrence, a univariate model is applied to assess the association of individual marker features or characteristics with cancer reoccurrence data of a patient cohort with known prognosis conditions (e.g. as illustrated in Table 1). In one embodiment, for each marker characteristic, a univariate Cox Proportional Hazards model was fitted to assess the association of the marker characteristic and the colon cancer 5 -year reoccurrence time in patients who did not receive chemotherapy (n=450). The obtained p-values of all the marker features were corrected by Benjamin Hochberg method resulting in false discovery rates (FDR, Q values) for each marker characteristic.

[0086] For each multivariate model, the 5 years Area under Receiver Operating Characteristic (ROC) curve (AUC) was estimated and ranked. The best subset model is selected if AUC is highest, AIC (discussed below) is minimum and features of exclusive markers are present (e.g. only a single marker feature/characteristic of each marker that forms the model, is present).

[0087] In one embodiment, Akaike information criteria (AIC) is used for model selection, where AIC = -2Log likelihood + 2K, where K is the number of effective degrees of freedom measuring the model complexity. [0088] In one embodiment, to assess that if the selected models can differentiate patients in terms of their reoccurrence time, patients were split into two groups using Youden's J statistic and their empirical Kaplan-Meier curves compared. The selected model with multiple markers is fitted to the data using cox proportional hazards methods (R package survival) and it provides every patient a risk score (the linear term in cox model). The R package survival ROC estimates the sensitivity and specificity of 5 -year reoccurrence corresponding to every risk score value. The Youden's J statistic is computed as J= Sensitivity-i- Specificity- 1 and the cut point is chosen at maximum J. The Kaplan-Meier curve is examined for the entire no chemo cohort, stage I, stage II and stage III patients, respectively, using the same Youden's J cutoff. The most suitable model is selected with highest AUC, smallest AIC and best split of KM curves.

[0089] Table 3 provides a non-limiting example of the markers and their exemplary marker characteristics (e.g. marker location of epithelium or stroma within the tissue sample, subcellular localization of markers) selected as a result of a statistical analysis using a univariate model. Each selected marker characteristics, taken individually, is clinically predictive of a risk of cancer reoccurrence.

TABLE 3. Selected markers and associated marker characteristics based on a univariate analysis

Multivariate analysis

[0090] In some embodiments, at least one marker characteristics of each marker selected in univariate analysis are subjected to a multivariate analysis. Marker characteristics subjected to the multivariate analysis include both Marker characteristics of a first marker specifically measured in the tumor cells in the epithelial region and a second, different marker specifically measured in the non-tumor cells in the stromal region of a sample obtained from a tumor region of a cancer patient.

[0091] In some embodiments, the first marker specifically measured in the tumor cells in the epithelial region may be selected from a marker group including, but not limited to, beta-catenin (β-catenin), CA9, Claudinl, COX2, E-cadherin, Glutl, CA9, MSH2, NaKATPase, NDRGl, p21, p53, pERK, pMAPKAPK2, or a combination thereof. The second marker specifically measured in non-tumor cells in the stromal regions may be selected from a marker group including, but not limited to, CD3, CD8, CD68, smooth muscle actin alpha (aSMA), Collagen IV, Lamin A/C, or a combination thereof.

[0092] In one embodiment, a multivariate cluster analysis may be used. For example, FIGs. 5A-5D shows a result of an exemplary K-means cluster analysis based on marker characteristics of each marker in a marker panel including beta-catenin (β-catenin), E- cadherin, Glutl, CA9, NDRGl, p53, CA9 in the epithelial region and markers CD3, CD8, CD68, smooth muscle actin alpha (aSMA), Collagen IV, Lamin A/C in the stromal region.

[0093] FIG. 5A shows prognostic profiles of seven distinctive groups or clusters (clusters 1-7) with different prognostic characteristics developed based on the cluster analysis, as represented by probability of survival within 5 years vs. days post cancer diagnosis (Dx). Clusters 1 and 7 have high aSMA and Collagen IV expression levels and low immune cell infiltration levels in the stroma. Cluster 1 is a small cluster with very little expression of the other markers. Cluster 7 also has elevated aSMA and Collagen IV levels and low immune cell infiltration, but is further characterized by elevated p53 expression level, and intermediate to high levels of various tumor cell markers. Cluster 6 shows partial stromal character with elevated aSMA and Lamin A/C, low T-lymphocyte infiltration (CD8), but high macrophage infiltration (CD68), suggesting tumors with high fibroblast and innate immune cell stromal composition. Two larger clusters (clusters 3 and 4) are characterized by elevated stromal and cytotoxic T-lymphocyte profiles. The marker characteristics of clusters 1, 6, and 7 suggest that the corresponding cells may have profiles consistent with genomic classifiers reporting increased mesenchymal character (CMS4) and high risk cancer reoccurrence. FIGs. 5B-5E show the comparisons between clusters 3 and 7 (FIG. 5B), clusters 4 and 7 (FIG. 5C), and clusters 3 and 6 (FIG. 5D), respectively. It is shown that clusters 6 and 7 have the worst overall prognosis and separate well from clusters 3 and 4.

[0094] In some embodiments, a model is applied to develop a predefined marker panel in a two-step process. The first step (SI) includes down-selecting, from a plurality of markers in reference samples obtained from respective tumor regions of a group of cancer patients with known prognosis conditions, individual markers and associated marker characteristics clinically associated with the prognosis conditions of each patient, based on a univariate analysis. The univariate analysis was described in the earlier section of the present disclosure and will be illustrated in more detail hereinafter. At the second step (S2), the predefined marker panel is developed, by applying a multivariate model to the down-selected marker characteristic of the plurality of markers. Both steps SI and S2 are based on assessment of reference biological samples obtained from respective tumor regions of a group of cancer patients with known prognosis conditions. In one embodiment, multivariate model may be a cox proportional hazards model. [0095] The resulting predefined marker panel comprising at least a first marker specifically measured in the tumor cells in the epithelial region and a second, different marker specifically measured in non-tumor cells in the stromal region. The respective marker characteristics of the first and the second markers in the predefined marker panel are, individually or in combination, clinically predictive of a risk of cancer reoccurrence. A prognostic biosignature may be developed based on a combination of measured marker characteristics of each marker.

[0096] In some embodiments, the prognostic biosignature is developed for a biological sample of a cancer patient with unknown prognosis. The techniques of the present disclosure provide prognostic characterization and prognostic biosignature that allows risk of cancer reoccurrence assessment of such sample to be performed, which is not previously obtainable with other reported techniques. The developed prognostic biosignature may be applied to provide cancer prognosis, to determine if a patient is in need of an adjuvant therapy, to direct and/or monitor therapy, or any combinations thereof.

[0097] Getting back to the first step (SI), the down-selected marker characteristics are, taken individually, each clinically predictive of a risk of cancer reoccurrence of the patient. Recognizing that a single marker and associated marker characteristics is unlikely to describe cancer reoccurrence and mortality risk, it is hypothesized that a combination of markers with partial, non-overlapping relationships with disease outcome may provide more robust biosignatures. Therefore, at the second step (S2), the identified individual markers are assessed to evaluate the correlation between marker characteristics and cancer reoccurrence based on known prognosis conditions of the group of cancer patients in the cohort. In one embodiment, a Cox proportional hazards model is used but other univariate models known to a skilled person in the art may be used as well. In some embodiments, a significant amount of potential marker candidates was retrieved. Metrics may then be applied to filter the potential marker candidates to develop a preferred marker panel with markers and associated marker characteristics that provides closer association with caner reoccurrence that exceeds an optimal threshold. In one embodiment, metrics including, but not limited to, area under the receiver operator characteristic curve (ROC AUC) and false discovery rates (FDR) used to define an optimal threshold value for selecting prognostic markers clinically predictive of a risk of cancer reoccurrence. For example, in one embodiment, ROC AUC and FDR threshold are defined to be 0.6 and 0.1, respectively. The threshold value for ROC AUC may be in a range of 0.5-1, in a range of 0.6-1, in a range of 0.7- 1, or in a range of 0.8-1. The threshold of FDR may be less than 0.1. In one example, thirty-two marker characteristics derived from 12 markers met the ROC AUC and FDR criteria for association with reoccurrence (Table 4). While each individual marker characteristic was associated with cancer reoccurrence in this analysis, certain markers were measured in different subcellular locations (e.g. membrane and cytoplasm measures of COX2). Those markers were further filtered to isolate the most relevant characteristic(s) per marker, based on biological evidence (e.g. cytoplasmic COX2 was selected based on its localization to the endoplasmic reticulum). This yielded 12 markers, nine from the epithelial tumor cells and three from the non-tumor cells in the stroma region.

[0098] The down-selected marker characteristics include marker characteristics for markers specifically measured in the epithelial tumor cells. Including features not previously described such as the localization of phosphorylated ERK, levels of phosphorylated MAPKAPK2, and the membrane localization ratio of Na+K+ATPase. These reoccurrence-associated marker features were subsequently assessed as variables in the second step (S2) of the multivariate analysis.

[0099] Table 4 shows result of the univariate analysis, according to one embodiment of the present disclosure.

MARKER FEATURE

a e : esu s o a un var a e ana ys s, accor ng o em o men s o e presen disclosure

Optimization of multivariate prognostic models

Stromal multivariate prognostic model

[0100] In one embodiment, three stromal marker characteristics (i.e. marker characteristics of markers specifically measured in the non-tumor cells in the stromal region) met the criteria (e.g. ROC AUC) for significant univariate associations with outcome (CD8, CD3, Ρ13Κρ110α) in the univariate analysis. These three stromal features were combined to fit a multivariate model, for example, a Cox proportional hazards model. The combination of these markers stratifies patients (ROC AUC = 0.68), as shown in FIGs. 6A-6E.

[0101] FIG. 6A shows a plot of possibility of survival for a CRC patient whole cohort vs. days post-diagnosis using a stromal marker panel including markers specifically measured in the non-tumor cells in the stromal region. The marker panel used in FIG. 6A includes CD3, CD8 and Ρ13Κρ110α, according to one embodiment of the present disclosure. FIG. 6B shows a plot of possibility of survival for stage I patients only in the patient cohort of FIG. 6A vs. days post-diagnosis based on the same stromal marker panel used in FIG. 6A. FIG. 6C shows a plot of possibility of survival for stage II patients only in the patient cohort of FIG. 6A vs. days post-diagnosis based on the stromal marker panel used in FIG. 6A. FIG. 6D shows a plot of possibility of survival for stage III patients only in the patient cohort of FIG. 6 A vs. days post-diagnosis based on the stromal marker panel used in FIG. 6A. [0102] FIG. 6E shows a multivariate modeling result with hazard ratio and p- value, using the stromal marker panel of FIG. 6A, and clinical variables (age, stage, and tumor grade). FIG. 6E shows that these markers were independent of clinical variables (e.g. stage, age, and grade). In Kaplan-Meier analysis (using Youden's J statistic derived cutoff), the stromal markers show a very strong correlation with disease outcome on the untreated patient cohort (Log Rank p<0.0001). However, Kaplan Meier analysis within each stage analysis revealed weaker associations (Log Rank p = 0.4; 0.06; 0.04; in stages I, II, and III, respectively). This suggests that the stromal marker panel confers some prognostic information but are not well-suited for use as a standalone prognostic marker panel, especially in early stage disease.

[0103] Table 5 shows an example in which a stromal model is used with only stromal cell marker characteristics specifically measured in the stromal region. The three- marker panel provides a concordance score = 0.67 and AUC= 0.686.

Table 5: Kaplan-Meier analysis with stromal model using a stromal marker panel with three markers (CD8/CD3/P13Kpl l0a)

Epithelial multivariate prognostic model

[0104] In some embodiments, to derive epithelial protein specific expression patterns with prognostic potential, a Cox proportional hazards model was used to build multivariate models focusing on epithelial cell expression patterns. Markers were first selected based on specific expression in epithelial cells, then filtered based on significance in univariate models as described hereinabove. This yielded a rich set of marker combinations (number of combination or n=182). The best performers were selected for evaluation based on metrics including ROC AUC, Akaike information criterion, and Log rank tests in individual stages (see entry 1-10, Table SI). It is to be noted that other statistical metrics may be used for the evaluation.

[0105] Table 6 shows an example in which an epithelial cell model is used with only marker characteristics of markers specifically measured in the epithelial tumor cell . The five-marker model gives concordance score = 0.69 and AUC = 0.702.

Table 6: Kaplan-Meier analysis with epithelial model using a marker panel with five markers (COX2, NaKATPase, pERK, beta- Catenin, p21)

[0106] FIG. 7A shows a plot of possibility of survival for a CRC patient whole cohort vs. days post-diagnosis based on an epithelial model using an epithelial marker panel including markers specifically measured in cancer cells in the epithelial region. The five-marker panel (COX2, beta-catenin, p21, Na+K+ATPase (NaKATPase), and phospho-ERK (pERK) as shown in Table 6 showed a result of ROC analysis (AUC =0.70). Youden's J statistic was used to select a cutoff or optimal threshold value for Kaplan-Meier analysis of the entire cohort (Log Rank p< 0.0001). However, as shown in FIGs. 7B-7D, in stage specific analyses the performance of the five-marker model was best in stage II (Log Rank p=0.0015), with a trend toward significance in stage III (Log Rank p=0.07), and no discrimination of reoccurrence in stage I (p=0.77). The results show that the marker characteristics of markers specifically measured in tumor cells in the epithelial region confer more early stage prognostic information relative to the stromal model. The results also suggest that the prognostic model performance might be improved by combing marker characteristics used in the stroma and epithelial model, as well as available clinical covariates.

Combined epithelial and stromal multivariate model

[0107] It is therefore envisioned, in the present disclosure, that it is advantageous to model the risk of reoccurrence by developing a prognostics biosignature or profile that is statistically associated with risk of reoccurrence.

[0108] FIG. 8A shows a plot of possibility of survival for a CRC patient whole cohort vs. days post-diagnosis based on a combined epithelial and stromal model using a combined marker panel including markers specifically measured in cancer cells in the epithelial region and markers for non-tumor cells in the stromal region. The marker panel used in the plot includes COX2, beta-catenin, p21, NaKATPase, phospho-ERK, and CD8, according to embodiments of the present disclosure;

[0109] FIG. 8B shows a plot of possibility of survival for stage I patients only in the patient cohort of FIG. 8A vs. days post-diagnosis based on the combined marker panel used in FIG. 8A;

[0110] FIG. 8C shows a plot of possibility of survival for stage II patients only in the patient cohort of FIG. 8A vs. days post-diagnosis based on the combined marker panel used in FIG. 8A;

[0111] FIG. 8D shows a plot of possibility of survival for stage III patients only in the patient cohort of FIG. 8A vs. days post-diagnosis based on the combined marker panel used in FIG. 8A;

[0112] FIG. 8E shows a multivariate modeling result with hazard ratio and p- value, using the combined marker panel used in FIG. 8A, and clinical variables (age, stage, and tumor grade); [0113] As shown in FIGs. 9A-9C, to evaluate the potential for improved prognostic modeling with a combined epithelial and stromal multivariate model, the concordance and correlation between the epithelial and stromal models was examined. It was hypothesized that the strength of correlation between the risk scores assigned by these models would inform the likelihood that the marker characteristics measured in and epithelial tumor cells and non-tumor cells in the stromal region provide complementary information regarding patient prognosis (e.g. if the two were highly correlated, then the likelihood of complementarity would be diminished). In a pairwise analysis of patient risk scores derived from the epithelial and stromal multivariate models, a correlation is observed (FIG. 9A: R= 0.366, p=1.5 x 10-15). However, despite statistically significant correlation, sufficient discordance was present to warrant further examination of combining these factors, as the correlation plot showed considerable scatter. Therefore, consistent with the approach described above, models of disease reoccurrence with combinations of stromal and epithelial biology as independent variables were built, tested and selected.

[0114] Similar to the approach taken for the epithelial model, several candidate models were developed. Selection of the final or optimized model was based on criteria including, but not limited to, AIC, ROC AUC, and stage specific Kaplan-Meier analysis.

[0115] In some embodiments, the final model includes marker characteristics such as membrane localization of Na+K+ATPase, nuclear localization of phospho-ERK, expression of COX2, phospho-MAPKAPK2, and p21. Membrane localization of Na+K+ATPase is associated with negative prognosis (i.e. cancer reoccurs within 5 years of diagnosis), while the remaining markers are associated with positive prognosis.

[0116] Table 7 below shows an example of selected marker characteristics of markers in a panel of six markers as a result of the combined epithelial and stromal model analysis. For example, in Table 7, the entry of "Median.Cyt.COX2_mean_epi" refers to a mean value of median intensity of cytoplasmic COX2 marker expression of individual tumor cells in the epithelial region.

Table 7: Selected marker characteristics of markers in a combined marker panel from multivariate analysis

[0117] The combined model yielded an improved ROC AUC (0.74) relative to the stromal or epithelial models alone. Kaplan Meier analysis of Youden's J derived cutoff or optimal threshold value yielded strong risk segmentation (Log Rank p< 0.0001). Within a certain stage (e.g., stage II), Kaplan-Meier analysis is improved when compared to the stromal or epithelial model alone. While stage I remained insignificant (Log Rank p=0.56), stage II and III exhibited strong separation/stratification of patients based on association with risk of reoccurrence (Log Rank p = 0.0003 and Log Rank p = 0.01, respectively).

[0118] FIG. 8E shows the addition of clinical factors as independent variables in a Cox model along with the six markers/molecular features improved the performance of the model (AUC = 0.792), demonstrating that the molecular features confer additional information useful in prognostic assessment of CRC reoccurrence.

[0119] It is to be understood that the example describing a six-marker panel including COX2, beta-catenin, p21, Na⁺K⁺ATPase, phospho-ERK (pERK), and CD8 is provided as a non-limiting example and by no means should be treated as the only marker panel applicable. There are various combinations of marker characteristics that may form a marker panel with comparable performance for prognostic characterization, as illustrated in Table 8.

[0120] In some embodiments, the marker panel, for example, the six-marker panel of Table 7 was applied to stratify patients in the cohort that have been subjected to the adjuvant therapy. In this analysis, the marker model derived from the untreated patients was applied directly to the patients who received chemotherapy. In Kaplan Meier analysis, the six-marker model is applied to assess risk of reoccurrence for patients in the chemotherapy treated group, with a precision slightly lower than a precision in assessing risk of reoccurrence for patients in the untreated group (Log Rank p=0.016).

[0121] In the present disclosure, a highly multiplexed immunofluorescence (MxlF) single cell proteomics technology was applied to develop prognostic models of CRC. Several cell marker characteristics are determined to show associations with patient prognosis. Importantly, the risk model of combined epithelial and stromal model resulted in clinically relevant indicators of disease reoccurrence in stage II disease, providing a novel technical solution to the stratification and identification of high risk CRC and chemo-responsive stage II patients, a technical challenge associated with the greatest unmet clinical need.

[0122] The current disclosure demonstrates the potential of performing prognostic characterization of biological samples obtained from cancer patients, and of providing more informative treatment management, the highly multiplexed and quantitative nature of MxlF extends the in situ analysis advantages of traditional IHC over approaches that homogenize samples. Indeed, over one billion metrics derived from over 1.5 million cells was analyzed in an exemplary workflow. The intact sample morphology confers advantages including the direct observation of the cells responsible for expression of the measured analytes. The techniques described in the present disclosure further allow the assessment of cell type, cell to cell expression heterogeneity, subcellular protein localization, spatial aspects of cell arrangement and biological compartment residency (e.g. stroma, epithelium, vascular), and morphometric analysis. [0123] FIG. 10A illustrates a characterization of a tissue sample obtained from a tumor region of a stage II CRC patient. By subjecting the sample to the highly multiplexed and quantitative MxlF workflow, the epithelial and the stromal regions in the image data may be segmented and individual cells identified. Multiple markers may be therefore specially measured in cells in certain regions of a single sample. For example, in FIG. 10A, marker characteristics of E-Cadherin marker can be specifically measured (e.g. intensity measured using a corresponding probe specific for E-Cad marker) and quantified in the tumor cells in epithelial region of the sample, and marker characteristics of CD3 and CD8 markers can be specifically measured and quantified in the non-tumor cells in the stromal region, according to embodiments of the present disclosure;

[0124] FIG. 10B illustrates that MxlF technique enables the assessment and characterization of differences in other marker characteristics such as morphology and distribution of markers in the tissue samples of patients at different stages, according to embodiments of the present disclosure

[0125] Techniques described in the present disclosure show that the patient populations may be stratified into sub-populations, based on measured marker characteristics of markers which, when taken individually or in combination, are clinically predictive of risk of cancer reoccurrence. The system and method described herein also enables the characteristics of markers measured in specific cell populations, for example, tumor cells in the epithelial region and non-tumor cells in the stromal region and provides marker characteristic measurements in subcellular compartments of individual cells involved. This is an improvement compared to the gene-based techniques. While gene-based techniques may find utility in CRC prognosis, implementation of genomic techniques would lead to the samples being destroyed and morphological features and subcellular information of the cells being lost.

[0126] This written description uses examples as part of the disclosure, including the best mode, to enable any person skilled in the art to practice the disclosed implementations, including making and using any devices or systems and performing any incorporated methods. The patentable scope is defined by the claims, and may include other examples that occur to those skilled in the art. Such other examples are intended to be within the scope of the claims if they have structural elements that do not differ from the literal language of the claims, or if they include equivalent structural elements with insubstantial differences from the literal languages of the claims.

[0127]

Table 8. Optimization of combined multivariate models, according to embodiments of the present disclosure

Claims

We claim:

1. A method for performing a prognostic characterization of a biological sample obtained from a tumor region of a patient diagnosed with a cancer, the method comprising:

providing an image data of the biological sample comprising an epithelial region having one or more tumor cells therein and a stromal region having one or more non-tumor cells therein;

segmenting the epithelial and the stromal regions in the image data to identify individual cells;

measuring, from the identified individual cells, one or more marker characteristics of each marker in a predefined marker panel, the marker panel comprising at least a first marker specifically measured in the tumor cells in the epithelial region and a second, different marker specifically measured in the non-tumor cells in the stromal region;

developing a prognostic biosignature based on the measured marker characteristics of each marker in the predefined marker panel;

wherein the marker characteristics of the first and the second markers in the marker panel are, individually or in combination, clinically predictive of a risk of cancer reoccurrence.

2. The method of claim 1, wherein the biological sample is obtained from the tumor region of the patient diagnosed with the cancer and with an unknown prognosis.

3. The method of claim 1, further comprising determining the patient is in need of an adjuvant therapy, if the patient is assigned with a high risk of cancer reoccurrence based on an assessment of the prognostic biosignature.

4. The method of claim 3, wherein the adjuvant therapy is an adjuvant chemotherapy or an adjuvant immunotherapy.

5. The method of claim 3, wherein the patient is assigned with the high risk of cancer reoccurrence when the assessment of the prognostic biosignature provides a prognostic score exceeding an optimal threshold value.

6. The method of claim 5, wherein the optimal threshold value is determined based on assessment of prognostic biosignatures of reference samples from a group of cancer patients with known prognosis conditions .

7. The method of claim 1, wherein the patient has not been subjected to an adjuvant therapy.

8. The method of claim 1, wherein the patient is being subjected or has been subjected to an adjuvant therapy and the method further comprising monitoring responsiveness to the adjuvant therapy.

9. The method of claim 1, wherein the patient is diagnosed with a stage II or a stage III colorectal cancer (CRC).

10. The method of claim 1, wherein the measured marker characteristics comprise one or more of a measured intensity, a distribution, a type, a location within subcellular compartments, or a ratio of intensities between subcellular compartments of individual markers in the predefined marker panel.

11. The method of claim 1, wherein the predefined marker panel comprises at least two markers selected from a marker group comprising CD8, NaKATPase, Claudinl, CD3, pMAPKAPK2, COX2, CA9, phospho-ERK, β-catenin, p21, MSH2, E-Cad, Glutl, p53, CD20, MLH1, LaminA_C and pERK.

12. The method of claim 1, wherein the predefined marker panel comprises COX2, beta-catenin, p21, NaKATPase, phospho-ERK and CD8.

13. The method of claim 1, wherein the first marker is selected from a marker group comprising beta-catenin (β-catenin), CA9, Claudinl, COX2, E- cadherin, Glutl, CA9, MSH2, NaKATPase, NDRG1, p21, p53, pERK, pMAPKAPK2, or any combinations thereof.

14. The method of claim 1, wherein the second marker is selected from a marker group comprisingCD31, SMA, Collagen IV, Fibronectin, CD3, CD8, CD68, Lamin A/C, vimentin, S100, or any combinations thereof.

15. The method of claim 1, wherein the prognostic biosignature further comprise one or both of a molecular subtype characteristic based on a genomic analysis, and a patient clinical characteristic selected from a group comprising histologic types, stage, age, gender, cancer grade, or any combinations thereof.

16. The method of claim 1, further comprises developing the predefined marker panel by:

measuring marker characteristics of a plurality of markers in reference samples obtained from respective tumor regions of a group of cancer patients with known prognosis conditions,

applying a multivariate model to the marker characteristics of the plurality of markers; and

developing the predefined marker panel comprising at least the first marker and the second marker, wherein their respective marker characteristics are, individually or in combination, clinically associated with known prognosis of each patient.

17. The method of claim 12, wherein the multivariate model is a combined epithelial model and stroma model.

18. A method for determining a treatment for a patient diagnosed with a cancer, based on a prognostic biosignature of a biological sample obtained from a tumor region of the patient, the method comprising:

determining if the patient has a high risk of cancer reoccurrence by: providing an image data of the biological sample comprising an epithelial region having one or more tumor cells therein and a stromal region having one or more non-tumor cells therein;

segmenting the epithelial and the stromal regions in the image data to identify a plurality of single cells;

measuring, from the identified single cells, one or more marker characteristics of each marker in a predefined marker panel comprising at least a first marker specifically measured in the tumor cells in the epithelial region and a second, different marker specifically measured in the non-tumor cells in the stromal region;

developing the prognostic biosignature based on a combination of the measured marker characteristics of each marker in the predefined marker panel wherein the marker characteristics of the first and the second markers in the marker panel are, individually or in combination, clinically predictive of a risk of cancer reoccurrence;

determining a prognostic value for the risk of cancer reoccurrence of the patient based on the prognostic biosignature; and

if the prognostic value of the patient exceeds an optimal threshold, then determining the patient is in need of an adjuvant therapy, and

if the prognostic value of the patient does not exceed the optimal threshold, then determining the patient is in need of a surgery treatment alone.

19. A method for performing a prognostic characterization of a biological sample obtained from a tumor region of a patient diagnosed with a stage II or stage III colorectal cancer and with an unknown prognosis, the method comprising: providing an image data of the biological sample comprising an epithelial region having one or more tumor cells therein and a stromal region having one or more non-tumor cells therein;

segmenting the epithelial and the stromal regions in the image data to identify single cells;

measuring, from the identified single cells, one or more marker characteristics of each marker in a marker panel comprising COX2, beta- catenin, p21, NaKATPase, phospho-ERK and CD8 markers;

developing a prognostic biosignature by selecting a combination of the one or more marker characteristics of each marker wherein the marker characteristics of each marker in the marker panel are, individually or in combination, clinically predictive of a risk of cancer reoccurrence.