Methods and systems for predicting response to PD-1 axis-directed therapy

Abstract

The present invention provides a scoring function that is developed and used to identify patients who may respond to PD-1 axis-directed therapy. The scoring function is obtained by extracting features from multiple stained sections, selecting features related to response to the therapy using a feature selection function, and fitting one or more of the selected features to multiple candidate scoring functions. Candidate scoring functions that demonstrate a desired balance between predictive sensitivity and specificity can then be selected for inclusion in a scoring system that includes at least an image analysis system.

Description

[0001] Cross-references to related patent applications

[0002] This is an international application under the Patent Cooperation Treaty, claiming priority to U.S. Provisional Patent Application No. 62 / 739,828 filed on October 1, 2018 and U.S. Provisional Patent Application No. 62 / 742,934 filed on October 9, 2018, the contents of each of which are incorporated herein by reference in their entirety.

[0003] References to sequence listings submitted as ASCII-compliant text files (.txt)

[0004] In accordance with the EFS-Web legal framework and 37 CFR §§1.821-825 (see MPEP §2442.03(a)), a sequence listing in the form of an ASCII-compliant text file (titled "Sequence_Listing_3000022-004977_ST25.txt") created on September 30, 2019 and 23,475 bytes in size) is submitted concurrently with this application, and the entire contents of the sequence listing are incorporated herein by reference. Technical Field

[0005] The present invention relates to the detection, characterization and enumeration of biomarkers in tumor samples for predicting response to checkpoint inhibitor therapy. Background Art

[0006] Cancers can evade immune surveillance and eradication by upregulating the programmed death 1 (PD-1) pathway and its ligand programmed death ligand 1 (PD-L1) on tumor cells and in the tumor microenvironment. Blocking this pathway with antibodies against PD-1 or PD-L1 has led to significant clinical responses in some cancer patients. However, identifying predictive biomarkers for patient selection is a major challenge.

[0007] PD-L1 is the most widely used predictive biomarker for selecting patients to receive PD-1 axis-directed therapy. However, inconsistent results have been observed. See Yi.

[0008] Mismatch repair (MMR) deficiency predicts response of solid tumors to PD-1 blockade. See Le(I) and Le(II). However, not all patients with mismatch repair deficiency respond to PD-1 blockade. The predictive value is limited due to variable strength of association between studies and tumor types.

[0009] Recent studies have shown that the spatial arrangement and interactions between cancer cells and immune cells can affect patient prognosis, survival, and response to therapy. Wang and Barrera.

[0010] There is an increasing need to understand the tumor microenvironment and associated biomarkers to guide cancer immunotherapy. Summary of the invention

[0011] The present disclosure generally relates to systems and methods for identifying and using novel biomarkers to predict the response of solid tumors to PD-1 axis-directed therapies.

[0012] In a certain embodiment, a method for developing a scoring function for predicting a tumor's response to a PD-1 axis-directed therapy comprises: (a) obtaining: (a1) a set of digital images of tumor tissue samples obtained from a plurality of patients before treatment with the PD-1 axis-directed therapy, wherein at least one digital image of each patient is a digital image of a tissue section stained with multiple affinity histochemical stains for each of one or more epithelial markers, one or more immune cell markers, and one or more PD-1 axis pathway markers; and (a2) post-treatment response data for each patient; (b) extracting a plurality of features from the digital images of the multiplexed stained tissue sections; (c) applying a feature selection function to the extracted plurality of features and the post-treatment response data to obtain a ranking of each feature for strength of correlation with the response to the PD-1 axis-directed therapy; (d) applying a modeling function to one or more of the ranked features and the post-treatment response data to generate a plurality of candidate models for predicting a response to a checkpoint inhibitor therapy, and testing the consistency of each candidate model with the response; and (e) selecting the candidate model with the highest consistency with the response as the scoring function. In a certain embodiment, the multiplex affinity histochemical staining includes histochemical staining using a biomarker-specific reagent for each of PD-L1, CD8, CD3, CD68, and PanCK (Group 1), and the feature is selected from the group consisting of the features in the left column of Table 4. In a certain embodiment, the multiplex affinity histochemical staining includes histochemical staining using a biomarker-specific reagent for each of PanCK, PD-L1, PD1, CD8, and LAG3 (Group 2), and the feature is selected from the group consisting of the features in the right column of Table 4. In a certain embodiment, the feature selection function is selected from the group consisting of: an ensemble feature selection method (including, for example, a random forest function), a filtering method (including, for example, a mutual information-based function (mRMR) / a correlation coefficient-based function and a relief-based function) and / or an embedded feature selection function (such as an elastic network / least absolute shrinkage function or selection operator (LASSO) function). In a certain embodiment, the candidate model is made using one or more of the first 25, first 20, first 15, first 10, first 9, first 8, first 7, first 6, first 5, first 4, or first 3 features identified by the feature selection function. In another embodiment, the candidate model uses at least 1, at least 2, at least 3, at least 4, or at least 5 features identified in the first 10 features of the feature selection function. In another embodiment, the candidate model includes at least one feature present in the first 5 features of at least 2 feature selection functions. In a certain embodiment, the modeling function is selected from the group consisting of: quadrant discriminant analysis (QDA), linear discriminant analysis (LDA), support vector machine (SVM), and artificial neural network (ANN).In a certain embodiment, the PD-1 axis-directed therapy is a PD-1 specific monoclonal antibody or a PD-L1 specific monoclonal antibody. In a certain embodiment, the PD-1 axis-directed therapy is selected from the group consisting of: pembrolizumab, nivolumab, atezolizumab, avelumab, durvalumab, cemiplizumab, tislelizumab and LY3300054.

[0013] In a certain embodiment, a method for scoring the possibility that a tumor sample will respond to PD-1 axis-directed therapy is provided, the method comprising: (a) obtaining a digital image of a tumor section from the tumor sample, wherein the tumor section is stained with multiple affinity histochemical staining for each of one or more epithelial markers, one or more immune cell markers, and one or more PD-1 axis pathway markers; (b) identifying a region of interest (ROI) in the digital image; (c) extracting one or more features of cells stained for corresponding biomarkers from the ROI; and (d) applying a scoring function to a feature vector including one or more extracted features of (c) to generate a score, wherein the score indicates the possibility that the tumor will respond to the PD-1 axis-directed therapy. In a certain embodiment, the ROI is derived from a digital image of a morphologically stained section of the tumor sample, wherein the morphologically stained section and the sample of the multiple affinity histochemical staining are continuous sections. In a certain embodiment, the ROI is identified by a user in a digital image of the morphologically stained section and automatically registered in a digital image of the multiple affinity histochemically stained section. In a certain embodiment, the multiplex affinity histochemical staining comprises histochemical staining using a biomarker-specific reagent for each of PD-L1, CD8, CD3, CD68, and PanCK (Group 1), the ROI is an ROI according to Table 3, and the features include at least one feature selected from the group consisting of the features in the left column of Table 4. In a certain embodiment, the multiplex affinity histochemical staining comprises histochemical staining using a biomarker-specific reagent for each of PD-L1, CD8, CD3, CD68, and PanCK (Group 1), the ROI is an ROI according to Table 3, and the features include at least one feature determined to be important for predicting a patient's response to PD-1 axis-directed therapy by ReliefF and / or Random Forests. In a certain embodiment, the multiplex affinity histochemical staining comprises histochemical staining using biomarker-specific reagents for each of PD-L1, CD8, CD3, CD68, and PanCK (Group 1), the ROI is an ROI according to Table 3, and the features comprise at least one feature determined to be one of the top 10 most important features for predicting patient response to PD-1 axis-directed therapy by ReliefF and / or Random Forests.In a certain embodiment, the multiplex affinity histochemical staining comprises histochemical staining using a biomarker-specific reagent for each of PD-L1, CD8, CD3, CD68, and PanCK (Group 1), the ROI is an ROI according to Table 3, and the features include at least 1, 2, 3, 4, 5, 6, 7, 8, 9, or 10 of the top 10 most important features determined for predicting a patient's response to PD-1 axis-directed therapy by ReliefF and / or Random Forest. In a certain embodiment, the multiplex affinity histochemical staining comprises histochemical staining using a biomarker-specific reagent for each of PD-L1, CD8, CD3, CD68, and PanCK (Group 1), the ROI is an ROI according to Table 3, and the features include at least one feature selected from the group consisting of: the fraction of PD-L1+ macrophages in the stroma, the fraction of PD-L1+CD3+CD8- cells in the stroma, and the fraction of PD-L1+CD3+ cells in the stroma. In a certain embodiment, the multiplex affinity histochemical staining comprises histochemical staining using a biomarker-specific reagent for each of PD-L1, CD8, CD3, CD68, and PanCK (Group 1), the ROI is an ROI according to Table 3, and the features include each of the following items: the fraction of PD-L1+ macrophages in the stroma, the fraction of PD-L1+CD3+CD8- cells in the stroma, and the fraction of PD-L1+CD3+ cells in the stroma. In a certain embodiment, the multiplex affinity histochemical staining comprises histochemical staining using a biomarker-specific reagent for each of PanCK, PD-L1, PD1, CD8, and LAG3 (Group 2), the ROI is an ROI according to Table 3, and the features are selected from the group consisting of the features in the right column of Table 4. In a certain embodiment, the multiplex affinity histochemical staining comprises Group 2, the ROI is an ROI according to Table 3, and the features comprise at least one feature in the right column of Table 4 determined to be important for predicting a patient's response to PD-1 axis-directed therapy by ReliefF and / or Random Forest. In a certain embodiment, the multiplex affinity histochemical staining comprises Group 2, the ROI is an ROI according to Table 3, and the features comprise at least one feature in the right column of Table 4 determined to be one of the top 10 most important features for predicting a patient's response to PD-1 axis-directed therapy by ReliefF and / or Random Forest. In a certain embodiment, the multiplex affinity histochemical staining comprises Group 2, the ROI is an ROI according to Table 3, and the features comprise at least 1, 2, 3, 4, 5, 6, 7, 8, 9 or 10 of the features in the right column of Table 4 determined to be one of the top 10 most important features for predicting a patient's response to PD-1 axis-directed therapy by ReliefF and / or Random Forest.In a certain embodiment, the multiple affinity histochemical staining includes Group 2, the ROI is an ROI according to Table 3, and the features include at least one feature selected from the group consisting of: the maximum number of CD8+ / PD-1 low intensity cells within 20μm of PD-L1+ cells in epithelial tumors, the average # of PD-1 low intensity CD8+ cells within a 20μm radius of PD-L1+ cells, the maximum value of Lag3 intensity in CD8+Lag3+ cells, the average # of PD-1+ cells within a 20μm radius of PD-L1+ cells, and the maximum value of Lag3+ intensity on CD8+ cells. In a certain embodiment, the multiple affinity histochemical staining includes Group 2, the ROI is an ROI according to Table 3, and the features include each of the following items: the maximum number of CD8+ / PD-1 low intensity cells within 20 μm of PD-L1+ cells in epithelial tumors, the average value of PD-1 low intensity CD8+ cells within a 20 μm radius of PD-L1+ cells #, the maximum value of Lag3 intensity in CD8+Lag3+ cells, the average value of PD-1+ cells within a 20 μm radius of PD-L1+ cells #, and the maximum value of Lag3+ intensity on CD8+ cells. In a certain embodiment, the scoring function is derived from a modeling function selected from the group consisting of: quadrant discriminant analysis (QDA), linear discriminant analysis (LDA), support vector machine (SVM) and artificial neural network (ANN). In a certain embodiment, the PD-1 axis-guided therapy is a PD-1-specific monoclonal antibody or a PD-L1-specific monoclonal antibody. In a certain embodiment, the PD-1 axis-directed therapy is selected from the group consisting of pembrolizumab, nivolumab, atezolizumab, avelumab, durvalumab, cemiplizumab, tislelizumab, and LY3300054.

[0014] In a certain embodiment, a method for selecting a patient to receive PD-1 axis-directed therapy is provided, the method comprising: (a) obtaining a digital image of a tumor section from the tumor sample, wherein the tumor section is stained with multiple affinity histochemical staining for each of one or more epithelial markers, one or more immune cell markers, and one or more PD-1 axis pathway markers; (b) identifying a region of interest (ROI) in the digital image; (c) extracting one or more features of cells stained for the corresponding biomarkers from the ROI; (d) applying a scoring function to a feature vector including one or more extracted features of (c) to generate a score, wherein the score indicates the likelihood that the tumor will respond to the PD-1 axis-directed therapy; (e) comparing the score with a predetermined cutoff value; and (f) selecting a patient to receive the PD-1 axis therapy or an alternative therapy based on the comparison of (e). In a certain embodiment, the ROI is derived from a digital image of a morphologically stained section of the tumor sample, wherein the morphologically stained section and the sample stained with multiple affinity histochemical staining are continuous sections. In a certain embodiment, the ROI is identified by a user in a digital image of the morphologically stained slice and automatically registered in a digital image of the multiplex affinity histochemically stained slice. In a certain embodiment, the multiplex affinity histochemical staining includes histochemical staining using a biomarker-specific reagent for each of PD-L1, CD8, CD3, CD68, and PanCK (Group 1), the ROI is an ROI according to Table 3, and the features include at least one feature selected from the group consisting of the features in the left column of Table 4. In a certain embodiment, the multiplex affinity histochemical staining includes histochemical staining using a biomarker-specific reagent for each of PD-L1, CD8, CD3, CD68, and PanCK (Group 1), the ROI is an ROI according to Table 3, and the features include at least one feature determined to be important for predicting a patient's response to PD-1 axis-directed therapy by ReliefF and / or Random Forests. In a certain embodiment, the multiplex affinity histochemical staining comprises histochemical staining using biomarker-specific reagents for each of PD-L1, CD8, CD3, CD68, and PanCK (Group 1), the ROI is an ROI according to Table 3, and the features comprise at least one feature determined to be one of the top 10 most important features for predicting patient response to PD-1 axis-directed therapy by ReliefF and / or Random Forests.In a certain embodiment, the multiplex affinity histochemical staining comprises histochemical staining using a biomarker-specific reagent for each of PD-L1, CD8, CD3, CD68, and PanCK (Group 1), the ROI is an ROI according to Table 3, and the features include at least 1, 2, 3, 4, 5, 6, 7, 8, 9, or 10 of the top 10 most important features determined for predicting a patient's response to PD-1 axis-directed therapy by ReliefF and / or Random Forest. In a certain embodiment, the multiplex affinity histochemical staining comprises histochemical staining using a biomarker-specific reagent for each of PD-L1, CD8, CD3, CD68, and PanCK (Group 1), the ROI is an ROI according to Table 3, and the features include at least one feature selected from the group consisting of: the fraction of PD-L1+ macrophages in the stroma, the fraction of PD-L1+CD3+CD8- cells in the stroma, and the fraction of PD-L1+CD3+ cells in the stroma. In a certain embodiment, the multiplex affinity histochemical staining comprises histochemical staining using a biomarker-specific reagent for each of PD-L1, CD8, CD3, CD68, and PanCK (Group 1), the ROI is an ROI according to Table 3, and the features include each of the following items: the fraction of PD-L1+ macrophages in the stroma, the fraction of PD-L1+CD3+CD8- cells in the stroma, and the fraction of PD-L1+CD3+ cells in the stroma. In a certain embodiment, the multiplex affinity histochemical staining comprises histochemical staining using a biomarker-specific reagent for each of PanCK, PD-L1, PD1, CD8, and LAG3 (Group 2), the ROI is an ROI according to Table 3, and the features are selected from the group consisting of the features in the right column of Table 4. In a certain embodiment, the multiplex affinity histochemical staining comprises Group 2, the ROI is an ROI according to Table 3, and the features comprise at least one feature in the right column of Table 4 determined to be important for predicting a patient's response to PD-1 axis-directed therapy by ReliefF and / or Random Forest. In a certain embodiment, the multiplex affinity histochemical staining comprises Group 2, the ROI is an ROI according to Table 3, and the features comprise at least one feature in the right column of Table 4 determined to be one of the top 10 most important features for predicting a patient's response to PD-1 axis-directed therapy by ReliefF and / or Random Forest. In a certain embodiment, the multiplex affinity histochemical staining comprises Group 2, the ROI is an ROI according to Table 3, and the features comprise at least 1, 2, 3, 4, 5, 6, 7, 8, 9 or 10 of the right column of Table 4 determined to be one of the top 10 most important features for predicting a patient's response to PD-1 axis-directed therapy by ReliefF and / or Random Forest.In a certain embodiment, the multiple affinity histochemical staining includes Group 2, the ROI is an ROI according to Table 3, and the features include at least one feature selected from the group consisting of: the maximum number of CD8+ / PD-1 low intensity cells within 20μm of PD-L1+ cells in epithelial tumors, the average # of PD-1 low intensity CD8+ cells within a 20μm radius of PD-L1+ cells, the maximum value of Lag3 intensity in CD8+Lag3+ cells, the average # of PD-1+ cells within a 20μm radius of PD-L1+ cells, and the maximum value of Lag3+ intensity on CD8+ cells. In a certain embodiment, the multiple affinity histochemical staining includes Group 2, the ROI is an ROI according to Table 3, and the features include each of the following items: the maximum number of CD8+ / PD-1 low intensity cells within 20 μm of PD-L1+ cells in epithelial tumors, the average value of PD-1 low intensity CD8+ cells within a 20 μm radius of PD-L1+ cells #, the maximum value of Lag3 intensity in CD8+Lag3+ cells, the average value of PD-1+ cells within a 20 μm radius of PD-L1+ cells #, and the maximum value of Lag3+ intensity on CD8+ cells. In a certain embodiment, the scoring function is derived from a modeling function selected from the group consisting of: quadrant discriminant analysis (QDA), linear discriminant analysis (LDA), support vector machine (SVM) and artificial neural network (ANN). In a certain embodiment, the PD-1 axis-guided therapy is a PD-1-specific monoclonal antibody or a PD-L1-specific monoclonal antibody. In a certain embodiment, the PD-1 axis-directed therapy is selected from the group consisting of pembrolizumab, nivolumab, atezolizumab, avelumab, durvalumab, cemiplizumab, tislelizumab, and LY3300054.

[0015] In a certain embodiment, a method of treating a patient having a tumor is provided, the method comprising: (a) obtaining a digital image of a tumor section from the tumor sample, wherein the tumor section is stained with a multiplex affinity histochemical stain for each of one or more epithelial markers, one or more immune cell markers, and one or more PD-1 axis pathway markers; (b) identifying a region of interest (ROI) in the digital image; (c) extracting one or more features related to cells stained for the corresponding biomarkers from the ROI; (d) applying a scoring function to a feature vector including one or more extracted features of (c) to generate a score, wherein the score indicates a likelihood that the tumor will respond to the PD-1 axis-directed therapy; (e) comparing the score to a predetermined cutoff value; and (f) if the comparison of (e) indicates that the patient is likely to respond to the PD-1 axis-directed therapy, administering the PD-1 axis-directed therapy to the patient, or if the comparison of (e) indicates that the patient is likely to not respond to the PD-1 axis-directed therapy, administering a course of treatment that does not include the PD-1 axis-directed therapy to the patient. In a certain embodiment, the ROI is derived from a digital image of a morphologically stained slice of the tumor sample, wherein the morphologically stained slice and the sample of the multiple affinity histochemical staining are continuous slices. In a certain embodiment, the ROI is identified by a user in the digital image of the morphologically stained slice and is automatically registered in the digital image of the multiple affinity histochemically stained slice. In a certain embodiment, the multiple affinity histochemical staining includes histochemical staining using a biomarker-specific reagent for each of PD-L1, CD8, CD3, CD68 and PanCK (Group 1), the ROI is an ROI according to Table 3, and the features include at least one feature selected from the group consisting of the features in the left column of Table 4. In a certain embodiment, the multiplex affinity histochemical staining includes histochemical staining using a biomarker-specific reagent for each of PD-L1, CD8, CD3, CD68, and PanCK (Group 1), the ROI is an ROI according to Table 3, and the features include at least one feature determined to be important for predicting a patient's response to PD-1 axis-directed therapy by ReliefF and / or Random Forests. In a certain embodiment, the multiplex affinity histochemical staining includes histochemical staining using a biomarker-specific reagent for each of PD-L1, CD8, CD3, CD68, and PanCK (Group 1), the ROI is an ROI according to Table 3, and the features include at least one feature determined to be one of the top 10 most important features for predicting a patient's response to PD-1 axis-directed therapy by ReliefF and / or Random Forests.In a certain embodiment, the multiplex affinity histochemical staining comprises histochemical staining using a biomarker-specific reagent for each of PD-L1, CD8, CD3, CD68, and PanCK (Group 1), the ROI is an ROI according to Table 3, and the features include at least 1, 2, 3, 4, 5, 6, 7, 8, 9, or 10 of the top 10 most important features determined for predicting a patient's response to PD-1 axis-directed therapy by ReliefF and / or Random Forest. In a certain embodiment, the multiplex affinity histochemical staining comprises histochemical staining using a biomarker-specific reagent for each of PD-L1, CD8, CD3, CD68, and PanCK (Group 1), the ROI is an ROI according to Table 3, and the features include at least one feature selected from the group consisting of: the fraction of PD-L1+ macrophages in the stroma, the fraction of PD-L1+CD3+CD8- cells in the stroma, and the fraction of PD-L1+CD3+ cells in the stroma. In a certain embodiment, the multiplex affinity histochemical staining comprises histochemical staining using a biomarker-specific reagent for each of PD-L1, CD8, CD3, CD68, and PanCK (Group 1), the ROI is an ROI according to Table 3, and the features include each of the following items: the fraction of PD-L1+ macrophages in the stroma, the fraction of PD-L1+CD3+CD8- cells in the stroma, and the fraction of PD-L1+CD3+ cells in the stroma. In a certain embodiment, the multiplex affinity histochemical staining comprises histochemical staining using a biomarker-specific reagent for each of PanCK, PD-L1, PD1, CD8, and LAG3 (Group 2), the ROI is an ROI according to Table 3, and the features are selected from the group consisting of the features in the right column of Table 4. In a certain embodiment, the multiplex affinity histochemical staining comprises Group 2, the ROI is an ROI according to Table 3, and the features comprise at least one feature in the right column of Table 4 determined to be important for predicting a patient's response to PD-1 axis-directed therapy by ReliefF and / or Random Forest. In a certain embodiment, the multiplex affinity histochemical staining comprises Group 2, the ROI is an ROI according to Table 3, and the features comprise at least one feature in the right column of Table 4 determined to be one of the top 10 most important features for predicting a patient's response to PD-1 axis-directed therapy by ReliefF and / or Random Forest. In a certain embodiment, the multiplex affinity histochemical staining comprises Group 2, the ROI is an ROI according to Table 3, and the features comprise at least 1, 2, 3, 4, 5, 6, 7, 8, 9 or 10 features in the right column of Table 4 determined to be one of the top 10 most important features for predicting a patient's response to PD-1 axis-directed therapy by ReliefF and / or Random Forest.In a certain embodiment, the multiple affinity histochemical staining includes Group 2, the ROI is an ROI according to Table 3, and the features include at least one feature selected from the group consisting of: the maximum number of CD8+ / PD-1 low intensity cells within 20μm of PD-L1+ cells in epithelial tumors, the average # of PD-1 low intensity CD8+ cells within a 20μm radius of PD-L1+ cells, the maximum value of Lag3 intensity in CD8+Lag3+ cells, the average # of PD-1+ cells within a 20μm radius of PD-L1+ cells, and the maximum value of Lag3+ intensity on CD8+ cells. In a certain embodiment, the multiple affinity histochemical staining includes Group 2, the ROI is an ROI according to Table 3, and the features include each of the following items: the maximum number of CD8+ / PD-1 low intensity cells within 20 μm of PD-L1+ cells in epithelial tumors, the average value of PD-1 low intensity CD8+ cells within a 20 μm radius of PD-L1+ cells #, the maximum value of Lag3 intensity in CD8+Lag3+ cells, the average value of PD-1+ cells within a 20 μm radius of PD-L1+ cells #, and the maximum value of Lag3+ intensity on CD8+ cells. In a certain embodiment, the scoring function is derived from a modeling function selected from the group consisting of: quadrant discriminant analysis (QDA), linear discriminant analysis (LDA), support vector machine (SVM) and artificial neural network (ANN). In a certain embodiment, the PD-1 axis-guided therapy is a PD-1-specific monoclonal antibody or a PD-L1-specific monoclonal antibody. In a certain embodiment, the PD-1 axis-directed therapy is selected from the group consisting of pembrolizumab, nivolumab, atezolizumab, avelumab, durvalumab, cemiplizumab, tislelizumab, and LY3300054.

[0016] In a certain embodiment, the method provided includes: (a) marking a region of interest (ROI) on a digital image of a test sample of a tumor, wherein the digital image is a digital image of a sample subjected to multiple affinity staining for PD-L1, CD8, CD3, CD68, and PanCK (Group 1); (b) extracting one or more features of Table 9 from the ROI; (c) applying a scoring function to a feature vector including one or more features of (b), wherein the output value of the scoring function is a value predicting the patient's response to PD-1 axis-guided therapy. In a certain embodiment, the one or more features are determined to be important for predicting the patient's response to PD-1 axis-guided therapy by ReliefF and / or Random Forest. In a certain embodiment, the feature is determined to be one of the top 10 most important features for predicting the patient's response to PD-1 axis-guided therapy by ReliefF and / or Random Forest. In a certain embodiment, the at least one feature is selected from the group consisting of: the fraction of PD-L1+ macrophages in the stroma, the fraction of PD-L1+CD3+CD8- cells in the stroma, and the fraction of PD-L1+CD3+ cells in the stroma. In a certain embodiment, the feature vector includes each of the following: the fraction of PD-L1+ macrophages in the stroma, the fraction of PD-L1+CD3+CD8- cells in the stroma, and the fraction of PD-L1+CD3+ cells in the stroma. In a certain embodiment, the feature vector includes at least 1, 2, 3, 4, 5, 6, 7, 8, 9 or 10 of the top 10 most important features for predicting the patient's response to PD-1 axis-directed therapy by ReliefF and / or random forests. In a certain embodiment, the scoring function is derived by fitting a quadrant discriminant classification model to the selected features to predict the response to treatment. In a certain embodiment, the treatment outcomes for fitting the quadrant discriminant classification model are combined in a configuration selected from the group consisting of: PD relative to SD relative to PR+CR; PD relative to SD+PR+CR; and PD+SD relative to PR+CR. In a certain embodiment, the ROI is identified in a digital image of a first continuous section of the test sample, wherein the first continuous section is stained with hematoxylin and eosin, and wherein the ROI is automatically displayed in a digital image of at least a second continuous section of the test sample, wherein the second continuous section is stained with the first group. In a certain embodiment, the method is computer-implemented. In a certain embodiment, the PD-1 axis-directed therapy is a PD-1-specific monoclonal antibody or a PD-L1-specific monoclonal antibody. In a certain embodiment, the PD-1 axis-directed therapy is selected from the group consisting of: pembrolizumab, nivolumab, atezolizumab, avelumab, durvalumab, cemiplizumab, tislelizumab, and LY3300054.

[0017] In a certain embodiment, the method provided includes: (a) marking a region of interest (ROI) on a digital image of a test sample of a tumor, wherein the digital image is a digital image of a sample subjected to multiple affinity staining for PanCK, PD-L1, PD1, CD8, and LAG3 (Group 2); (b) extracting one or more features of Table 15 from the ROI; (c) applying a scoring function to a feature vector including one or more features of (b), wherein the output value of the scoring function is a value for predicting a patient's response to PD-1 axis-guided therapy. In a certain embodiment, the one or more features are features determined to be important for predicting a patient's response to PD-1 axis-guided therapy by ReliefF and / or Random Forest. In a certain embodiment, the one or more features are features determined to be one of the top 10 most important features for predicting a patient's response to PD-1 axis-guided therapy by ReliefF and / or Random Forest. In a certain embodiment, the feature vector includes at least one feature selected from the group consisting of: "maximum number of CD8+ / PD-1 low intensity cells within 20 μm of PD-L1+ cells in epithelial tumors," the average # of PD-1 low intensity CD8+ cells within a 20 μm radius of PD-L1+ cells, the maximum value of Lag3 intensity in CD8+Lag3+ cells, the average # of PD-1+ cells within a 20 μm radius of PD-L1+ cells, and the maximum value of Lag3+ intensity on CD8+ cells. In a certain embodiment, the feature vector includes each of the maximum number of CD8+ / PD-1 low intensity cells within 20 μm of PD-L1+ cells in epithelial tumors, and optionally further includes one or more additional features selected from the group consisting of: the average # of PD-1 low intensity CD8+ cells within a 20 μm radius of PD-L1+ cells, the maximum value of Lag3 intensity in CD8+Lag3+ cells, the average # of PD-1+ cells within a 20 μm radius of PD-L1+ cells, and the maximum value of Lag3+ intensity on CD8+ cells. In a certain embodiment, the feature vector includes at least 1, 2, 3, 4, 5, 6, 7, 8, 9 or 10 features determined to be one of the top 10 most important features for predicting a patient's response to PD-1 axis-directed therapy by ReliefF and / or Random Forest. In a certain embodiment, the scoring function is derived by fitting a quadrant discriminant classification model to the selected features to predict the response to treatment. In a certain embodiment, the treatment outcomes for fitting the quadrant discriminant classification model are combined in a configuration selected from the group consisting of: PD relative to SD relative to PR+CR; PD relative to SD+PR+CR; and PD+SD relative to PR+CR.In a certain embodiment, the ROI is identified in a digital image of a first serial section of the test sample, wherein the first serial section is stained with hematoxylin and eosin, and wherein the ROI is automatically displayed in a digital image of at least a second serial section of the test sample, wherein the second serial section is stained with set 2. In a certain embodiment, the PD-1 axis-directed therapy is a PD-1-specific monoclonal antibody or a PD-L1-specific monoclonal antibody. In a certain embodiment, the PD-1 axis-directed therapy is selected from the group consisting of: pembrolizumab, nivolumab, atezolizumab, avelumab, durvalumab, cemiplizumab, tislelizumab, and LY3300054.

[0018] In a certain embodiment, a system for predicting a patient's response to PD-1 axis therapy is provided, the system comprising: a processor; and a memory coupled to the processor, the memory being used to store computer executable instructions, which, when executed by the processor, cause the processor to perform operations including one or more methods of predicting a patient's response to PD-1 guided therapy described herein. In a certain embodiment, the system further comprises a scanner or microscope, the scanner or microscope being suitable for capturing a digital image of a slice of the tissue sample and transmitting the image to a computer device. In a certain embodiment, the system further comprises an automatic slide stainer, the automatic slide stainer being programmed to perform histochemical staining on the slice of the tissue sample with group 1 or group 2. In a certain embodiment, the system further comprises an automatic hematoxylin and eosin stainer, the automatic hematoxylin and eosin stainer being programmed to stain one or more consecutive slices of the slice stained by the automatic slide stainer. In a certain embodiment, the system further includes a laboratory information system (LIS) for tracking sample and image workflow and diagnostic information, the LIS including a central database configured to receive and store information related to the tissue sample, the information including at least one of the following items: the processing steps to be performed on the tumor tissue sample, the processing steps to be performed on the digital image of the slice of the tumor tissue sample, the processing history of the tumor tissue sample and the digital image; and one or more clinical variables (e.g., MMR or MSI status) associated with the likelihood that the patient will respond to the therapy. In a certain embodiment, the PD-1 axis-directed therapy is a PD-1-specific monoclonal antibody or a PD-L1-specific monoclonal antibody. In a certain embodiment, the PD-1 axis-directed therapy is selected from the group consisting of: pembrolizumab, nivolumab, atezolizumab, avelumab, durvalumab, cemiplizumab, tislelizumab, and LY3300054.

[0019] In a certain embodiment, a non-transitory computer-readable storage medium for storing computer-executable instructions is provided, and the computer-executable instructions are executed by a processor to perform operations, the operations including one or more methods of predicting the response of a patient to the PD-1-guided therapy described herein. In a certain embodiment, the PD-1 axis-guided therapy is a PD-1-specific monoclonal antibody or a PD-L1-specific monoclonal antibody. In a certain embodiment, the PD-1 axis-guided therapy is selected from the group consisting of: pembrolizumab, nivolumab, atezolizumab, avelumab, durvalumab, cemiplizumab, tislelizumab and LY3300054. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] This patent or application file contains at least one drawing executed in color. Copies of this patent or patent application publication with color drawing(s) will be provided by the Office upon request and payment of the necessary fee.

[0021] Figure 1 is a flow chart illustrating an exemplary method of deriving a scoring function disclosed herein.

[0022] Figure 2 is a flow chart illustrating an exemplary method for scoring a test sample using the scoring functions described herein.

[0023] Figure 3 An exemplary scoring system incorporating the scoring functions described herein is shown.

[0024] Figure 4A An exemplary workflow implemented on the image analysis system disclosed herein is shown, wherein an object recognition function is performed on the entire image before performing a ROI generator function.

[0025] Figure 4B An exemplary workflow implemented on the image analysis system disclosed herein is shown, wherein after executing the ROI generator function, the object recognition function is performed on only the ROI.

[0026] Figure 5 The multiple staining process using Group 1 as described in Example 1 is shown.

[0027] Figure 6 An exemplary slide stained with Set 1 is shown.

[0028] Figure 7 Importance rankings of all features stained with Group 1 are shown in all cases.

[0029] Figure 8 The importance ranking of features stained with group 1 in the context of MMR deficiency is shown.

[0030] Fig. 9 is an IHC image of a sample stained with Panel 1 from a patient who showed a complete response to treatment with PD-1 axis directed therapy.

[0031] Fig.10 are IHC images of a sample stained with Panel 1 from a patient with progressive disease following treatment with PD-1 axis directed therapy.

[0032] Fig.11 The staining and image analysis procedures used for Group 2 in Example II are shown.

[0033] Fig.12 The ranking of the top 15 features for Group 2 using the ReliefF feature selection function is shown. The features are as follows: (1) PD-L1 + / CD8 + PD-1 within a 20 μm radius of cells 低强度 / CD8 + The maximum number of cells; (2) from PD-L1 + / CD8 + The maximum PD-L1 intensity of the cells; (3) PD-1 + / PD-L1- / Lag3 + / CD8 + Cells and CD8 + The ratio of the number of cells; (4) PD-L1 + PD-1 within a 20 μm radius of cells 低强度 / CD8 + Spatial variance of the number of cells; (5) from all PD-L1 + The maximum PD-L1 intensity of the cell; (6) Lag3 in the panCK– region + Number of cells; (7) PD-L1 in the panCK– region + / panCK – cell density; (8)panCK + PD-L1 in the region + Number of cells; (9) panCK – PD-L1 in the region + Number of cells; (10) panCK – PD-1 in the region + Number of cells; (11) PD-L1 + PD-1 within a 20 μm radius of cells 低强度 / CD8 + Maximum number of cells; (12) PD-L1 + PD-1 within a 20 μm radius of cells 低强度 / CD8+ The average number of cells; (13) from CD8 + / Lag3 + The maximum value of Lag3 intensity of the cell; (14) panCK – CD8 + The number of cells; and (15) PD-L1 + / CD8 + PD-1 within a 20 μm radius of cells 低强度 / CD8 + The variance of the number of cells.

[0034] Fig.13 The ranking of the top 15 features for group 2 using the random forest feature selection function is shown. The features are as follows: (1) from PD-L1 + cells to their nearest PD-1 + Average distance of cells; (2) PD-L1 + / CD8 + PD-1 within a 20 μm radius of cells 低强度 / CD8 + Maximum number of cells; (3) PD-L1 + / panCK + PD-1 within a 20 μm radius of cells 低强度 / CD8 + The maximum number of cells; (4) from PD-L1 + Cells to the nearest PD-1 + Standard distance of cells; (5) panCK – Lag3 + / CD8 + Cells and CD8 + The ratio of the number of cells; (6) panCK – PD-L1 in the region + / CD8 + Cells and CD8 + The ratio of the number of cells; (7) PD-L1 + PD-1 within a 10 μm radius of cells 低强度 / CD8 + Average number of cells; (8) PD-L1 + PD-1 within a 10 μm radius of cells 低强度 / CD8 + Variance of the number of cells; (9) from all PD-1 + The average PD-1 intensity of cells; (10) from all Lag3 + Minimum Lag3 intensity of cells; (11) panCK –PD-L1 in the region + Cell density; (12) PD-L1 + / CD8 + PD-1 within a 10 μm radius of cells 低强度 / CD8 + Maximum number of cells; (13) panCK + PD-1 in the region + Number of cells; (14) panCK – PD-1 in the region + / PD-L1 – / Lag3 + / CD8 + Cells and CD8 + ratio of the number of cells; and (15) from Lag3 + / CD8 + The maximum value of Lag3 intensity of the cell.

[0035] Fig.14A Contains graphs illustrating the following: Spatial variance within a 10 μm radius of PD-L1+ cells # Predicted value of PD-1+ cells (F1), PD-L1 + Average #PD-1 within a 20 μm radius of the cell 低强度 CD8 + Predicted value of cells (F2), and CD8 + Lag3 + Predicted value of the maximum Lag3 intensity in the cell (F3).

[0036] Fig. 14B is a scatter plot showing the predicted values ​​of: spatial variance within a 10 μm radius of PD-L1+ cells #PD-1+ cells (F1) and PD-L1 + Average #PD-1 within a 20 μm radius of the cell 低强度 CD8 + Cells (F2).

[0037] Fig. 14C is a scatter plot showing the predicted values ​​of: spatial variance within a 10 μm radius of PD-L1+ cells #PD-1+ cells (F1), PD-L1 + Average #PD-1 within a 20 μm radius of the cell 低强度 CD8 + cells (F2), and CD8 + Lag3 + The maximum value of cellular Lag3 intensity (F3).

[0038] Fig.15Example IHC images of non-responders and responders are shown, as well as PD-L1 + Cell location (gray dots), PD-1 within 20 μm of PD-L1+ cells 低 Cells (white dots), and PD-1 within 10 μm of PD-L1+ cells + Figure 2: Graphical reconstruction of cells (black dots). Box (a) is the original fluorescence image of a non-responder. Box (b) is a graphical reconstruction of the white box from box (a), showing PD-L1+ cells and PD-1 within 20 μm of PD-L1+ cells in non-responders. 低 Spatial relationships between cells. Box (c) is a graphical reconstruction of the gray box from box (a), showing PD-L1 in non-responders. + PD-L1+ cells and PD-1 within 10μm of cells + Spatial relationship between cells. Box (d) is the original fluorescence image of the responder. Box (e) is a graphic reconstruction of the white box from box (d), showing the PD-L1+ cells and PD-1 within 20 μm of the PD-L1+ cells of the responder. 低 Spatial relationships between cells. Box (f) is a graphical reconstruction of the gray box from box (d), showing the responder PD-L1 + PD-L1+ cells and PD-1 within 10μm of cells + The spatial relationship between cells.

[0039] Fig.16A is a Kaplan-Meier survival curve for predicting overall survival after pembrolizumab treatment based on the expression of Lag3 in the panCK-negative region. + / CD8 + The number of cells and CD8 + ratio to the total number of cells.

[0040] Fig. 16B is a Kaplan-Meier survival curve for predicting overall survival after pembrolizumab treatment based on the expression of Lag3 in the panCK-negative region. - / CD8 + Cells and CD8 + The ratio of the number of cells.

[0041] Fig. 16C The Kaplan-Meier survival curve for predicting overall survival after pembrolizumab treatment is based on Lag3 + / panCK – The number of cells was divided by the panCK-negative area.

[0042] Fig.16Dis a Kaplan–Meier survival curve for predicting overall survival after pembrolizumab treatment based on the number of Lag3-positive cells in panCK-negative areas.

[0043] Fig.16E Kaplan-Meier survival curves for predicting overall survival after pembrolizumab treatment based on CD8 + The maximum value of Lag3 intensity in the cell.

[0044] Fig.16F is a Kaplan-Meier survival curve for predicting overall survival after pembrolizumab treatment based on the number of Lag3+ cells in panCK-positive areas.

[0045] Figure 16G is a Kaplan-Meier survival curve for predicting overall survival after pembrolizumab treatment based on PD-L1 + / panCK + PD-1 within a 10 μm radius of cells 中 / The average number of CD8+ cells.

[0046] Fig.16H is a Kaplan-Meier survival curve for predicting overall survival after pembrolizumab treatment based on PD-L1 + / panCK + PD-1 within a 20 μm radius of cells 中 / The average number of CD8+ cells.

[0047] Fig.16I is a Kaplan-Meier survival curve for predicting overall survival after pembrolizumab treatment based on PD-L1 + / CD8 + PD-1 within a 20 μm radius of cells 中 / Variance of the number of CD8+ cells.

[0048] Fig.16J is a Kaplan-Meier survival curve for predicting overall survival after pembrolizumab treatment based on PD-L1 + / panCK + PD-1 within a 20 μm radius of cells 中 / Variance of the number of CD8+ cells.

[0049] Figure 16K is a Kaplan-Meier survival curve for predicting overall survival after pembrolizumab treatment based on PD-L1 +PD-1 within a 10 μm radius of cells 低 / Variance of the number of CD8+ cells.

[0050] Figure 16L is a Kaplan-Meier survival curve for predicting overall survival after pembrolizumab treatment based on PD-L1 + / CD8 + PD-1 within a 20 μm radius of cells 中 / Maximum number of CD8+ cells.

[0051] Figure 16M is a Kaplan-Meier survival curve for predicting overall survival after pembrolizumab treatment based on PD-L1 + PD-1 within a 20 μm radius of cells 中 / Maximum number of CD8+ cells. For each feature metric, the median of the feature metric distribution was used as the cutoff value to divide the cohort into two groups. DETAILED DESCRIPTION

[0052] I. Definitions

[0053] Unless otherwise indicated, the scientific and technical terms used herein have the same meaning as those of ordinary skill in the art generally understand. See, for example, Lackie, DICTIONARY OF CELL AND MOLECULAR BIOLOGY, Elsevier (4th edition 2007); Sambrook et al., MOLECULAR CLONING, A LABORATORY MANUAL, Cold Springs Harbor Press (Cold Springs Harbor, NY 1989). The term "one" or "a kind of" is intended to mean "one (kind) or more (kind)". When preceding the description of a step or element, the term "comprise (comprises, comprises and comprising)" is intended to mean that adding additional steps or elements is optional and is not excluded.

[0054] Antibody: The term "antibody" herein is used in the broadest sense and includes various antibody structures, including but not limited to monoclonal antibodies, polyclonal antibodies, multispecific antibodies (e.g., bispecific antibodies), and antibody fragments, so long as they exhibit the desired antigen-binding activity.

[0055] Antibody fragment: "Antibody fragment" refers to a molecule other than an intact antibody that comprises a portion of an intact antibody and binds to the antigen to which the intact antibody binds. Examples of antibody fragments include, but are not limited to, Fv, Fab, Fab', Fab'-SH, F(ab')2; diabodies; linear antibodies; single-chain antibody molecules (e.g., scFv); and multispecific antibodies formed from antibody fragments.

[0056] Biomarker: As used herein, the term "biomarker" shall refer to any molecule or group of molecules found in a biological sample that can be used to characterize the biological sample or the subject from which the biological sample was obtained. For example, a biomarker can be a molecule or a group of molecules whose presence, absence, or relative abundance is characteristic of a particular cell or tissue type or state; or is characteristic of a particular pathological condition or state; or is indicative of the severity of a pathological condition, the likelihood of progression or regression of a pathological condition, and / or the likelihood of a pathological condition responding to a particular treatment. As another example, a biomarker can be a cell type or microorganism (e.g., bacteria, mycobacteria, fungi, viruses, etc.), or a substituent molecule or group of molecules thereof.

[0057] Biomarker-specific reagent: A specific detection reagent, such as a primary antibody, that can directly and specifically bind to one or more biomarkers in a cell sample.

[0058] Cell sample: As used herein, the term "cell sample" refers to any sample containing intact cells, such as a cell culture, a body fluid sample, or a surgical specimen sampled for pathological, histological, or cytological interpretation.

[0059] Detection reagent: A "detection reagent" is any reagent used to deposit a stain in the vicinity of a biomarker-specific reagent in a cell sample. Non-limiting examples include a biomarker-specific reagent (e.g., a primary antibody), a secondary detection reagent (e.g., a secondary antibody capable of binding to a primary antibody), a tertiary detection reagent (e.g., a tertiary antibody capable of binding to a secondary antibody), an enzyme directly or indirectly associated with a biomarker-specific reagent, a chemical reactive with such an enzyme to affect the deposition of a fluorescent or chromogenic stain, a washing reagent used between staining steps, and the like.

[0060] Detectable moiety: A molecule or material that can generate a detectable signal (e.g., visual, electronic, or other) that indicates the presence (i.e., qualitative analysis) and / or concentration (i.e., quantitative analysis) of a detectable moiety deposited on a sample. Detectable signals can be generated by any known or yet to be discovered mechanism, including absorption, emission, and / or scattering of photons (including radio frequency, microwave frequency, infrared frequency, visible light frequency, and ultraviolet frequency photons). The term "detectable moiety" includes chromogenic, fluorescent, phosphorescent, and luminescent molecules and materials, catalysts (e.g., enzymes) that convert one substance into another to provide a detectable difference (e.g., by converting a colorless substance into a colored substance or vice versa, or by producing a precipitate or increasing the turbidity of a sample). In some instances, the detectable moiety is a fluorophore, which belongs to several common chemical categories, including coumarin, fluorescein (or fluorescein derivatives and analogs), rhodamine, resorufin, luminophores, and cyanines. Additional examples of fluorescent molecules can be found in Molecular Probes Handbook—A Guide to Fluorescent Probes and Labeling Technologies, Molecular Probes, Eugene, OR, Thermo Fisher Scientific, 11th edition. In other embodiments, the detectable moiety is a molecule detectable via bright field microscopy, such as dyes including diaminobenzidine (DAB), 4-(dimethylamino)azobenzene-4'-sulfonamide (DABSYL), tetramethylrhodamine (DISCOVERY violet), N,N'-dicarboxypentyl-5,5'-disulfonic acid-indole-dicarbocyanine (Cy5), and rhodamine 110 (rhodamine).

[0061] Feature metric: A value indicating the level of a biomarker in a sample or the relationship between biomarkers. Examples include: expression intensity (e.g., on a scale of 0+, 1+, 2+, 3+), the number of cells that are positive for a biomarker, cell density (e.g., the number of biomarker-positive cells in the area of ​​the ROI, the number of biomarker-positive cells in the linear distance defining the edge of the ROI, etc.), pixel density (i.e., the number of biomarker-positive pixels in the area of ​​the ROI, the number of biomarker-positive pixels in the linear distance defining the edge of the ROI, etc.), the mean or median between cells expressing one or more biomarkers, etc. Feature metrics can be total metrics or overall metrics.

[0062] Histochemical detection: A process that involves labeling a biomarker or other structure in a tissue sample with a biomarker-specific reagent and a detection reagent in a manner that allows microscopic detection of the biomarker or other structure in the context of the cross-sectional relationship between the structures of the tissue sample. Examples include immunohistochemistry (IHC), chromogenic in situ hybridization (CISH), fluorescent in situ hybridization (FISH), silver in situ hybridization (SISH), and hematoxylin and eosin (H&E) stained formalin-fixed paraffin-embedded tissue sections.

[0063] Immune checkpoint molecules: Proteins expressed by immune cells whose activation downregulates cytotoxic T cell responses. Examples include PD-1, TIM-3, LAG-4, and CTLA-4.

[0064] Immune escape biomarkers: Biomarkers expressed by tumor cells that help the tumor avoid T cell-mediated immune responses. Examples of immune escape biomarkers include PD-L1, PD-L2, and IDO.

[0065] Immunological biomarkers: Biomarkers that characterize or influence an immune response to abnormal cells include, but are not limited to, biomarkers that indicate a particular class of immune cells (e.g., CD3), characterize an immune response (e.g., the presence, absence, or amount of a cytokine protein or one or more specific immune cell subtypes), or are expressed by, presented by, or otherwise located on non-immune cell structures that influence the type or extent of an immune cell response.

[0066] Monoclonal antibody: an antibody obtained from a substantially homogeneous antibody population, that is, except for possible variant antibodies (e.g., comprising naturally occurring mutations or produced during the production of monoclonal antibody preparations, such variants are usually present in trace amounts), the individual antibodies comprising the population are identical and / or bind the same epitope. Contrary to polyclonal antibody preparations that typically include different antibodies for different determinants (epitopes), each monoclonal antibody in a monoclonal antibody preparation is directed to a single determinant on an antigen. Therefore, the modifier "monoclonal" indicates that the characteristic of an antibody is obtained from a substantially homogeneous antibody population, and should not be interpreted as requiring antibodies to be produced by any particular method. For example, the monoclonal antibodies used in accordance with the present invention can be prepared by a variety of techniques, including but not limited to hybridoma methods, recombinant DNA methods, phage display methods, and methods utilizing transgenic animals comprising all or part of or a combination of human immunoglobulin loci.

[0067] Multiplex histochemical staining: A histochemical staining method in which multiple biomarker-specific reagents that bind to different biomarkers are applied to a single section and stained with different colored stains.

[0068] PD-1 axis-directed therapy: A therapeutic agent that disrupts the ability of PD-1 to downregulate T cell activity. Exemplary PD-1 axis-directed therapies include PD-1-specific monoclonal antibodies (e.g., pembrolizumab, nivolumab, cemiplimab, and tislelizumab), PD-L1-specific monoclonal antibodies (e.g., atezolizumab, avelumab, durvalumab, and LY3300054), and PD-1 small molecule inhibitors (e.g., CA-170, others reviewed by Li and Tian in preclinical development).

[0069] Sample: As used herein, the term "sample" shall refer to any material obtained from a subject that can be tested for the presence or absence of a biomarker.

[0070] Secondary detection reagents: Specific detection reagents are able to specifically bind to biomarker-specific reagents.

[0071] Section: When used as a noun, a thin slice of a tissue sample suitable for microscopic analysis, usually cut using a microtome. When used as a verb, the process of producing a section.

[0072] Serial section: As used herein, the term "serial section" shall refer to any of a series of sections cut one after another from a tissue sample by a microtome. For two sections to be considered "serial sections" of each other, they do not necessarily need to be consecutive sections of the tissue, but they should generally contain sufficiently similar tissue structures with the same spatial relationship so that the structures can be matched to each other after histological staining.

[0073] Single histochemical staining: A histochemical staining method in which a single biomarker-specific reagent is applied to a single section and stained with a single color stain.

[0074] Specific detection reagent: A composition of any substance that can specifically bind to a target chemical structure in the context of a cell sample. As used herein, the phrase "specific binding", "specific binding to" or "specific for" or other similar repetitions refer to a measurable and reproducible interaction between a target and a specific detection reagent, which determines the presence of a target in the presence of a heterogeneous population of molecules (including biomolecules). For example, an antibody that specifically binds to a target is an antibody that binds to the target with greater affinity, avidity, greater ease and / or longer duration than it binds to other targets. In one embodiment, the degree of binding of a specific detection reagent to an unrelated target is less than about 10% of the binding of the antibody to the antigen, for example, as measured by radioimmunoassay (RIA). In certain embodiments, the dissociation constant (Kd) of a biomarker-specific reagent that specifically binds to a target is ≤1 μM, ≤100 nM, ≤10 nM, ≤1 nM, or ≤0.1 nM. In another embodiment, specific binding may include but does not require exclusive binding. Exemplary specific detection reagents include nucleic acid probes specific for a particular nucleotide sequence; antibodies and antigen-binding fragments thereof; and engineered specific binding compositions, including ADNECTIN (a scaffold based on 10FN3 fibronectin; Bristol-Myers-Squibb Co.), AFFIBODY (a scaffold based on the Z domain of protein A from Staphylococcus aureus; Affibody AB, Solna, Sweden), AVIMER (a scaffold based on the domain A / LDL receptor; Amgen, Thousand Oaks, CA), dAb (a scaffold based on the VH or VL antibody domain; GlaxoSmithKline PLC, Cambridge, UK), DARPin (a scaffold based on ankyrin repeat proteins; Molecular Partners AG, Zürich, CH), ANTICALIN (a scaffold based on lipocalin; Pieris AG, Freising, DE), NANOBODY (a scaffold based on VHH (camel Ig); Ablynx N / V, Ghent, BE), TRANS-BODY (a transferrin-based scaffold; Pfizer Inc., New York, NY), SMIP (Emergent Biosolutions, Inc., Rockville, MD), and TETRANECTIN (a C-type lectin domain (CTLD), tetranectin-based scaffold; Borean Pharma A / S, Aarhus, DK).The description of such engineered specific binding structures is reviewed by Wurch et al. in Development of Novel Protein Scaffolds as Alternatives to Whole Antibodies for Imaging and Therapy: Status on Discovery Research and Clinical Validation, Current Pharmaceutical Biotechnology, Vol. 9, pp. 502-509 (2008), the contents of which are incorporated by reference.

[0075] Stain: When used as a noun, the term "stain" shall refer to any substance that can be used to visualize specific molecules or structures in a cell sample for microscopic analysis, including bright field microscopy, fluorescence microscopy, electron microscopy, etc. When used as a verb, the term "stain" shall refer to any process that results in the deposition of a stain on a cell sample.

[0076] Subject: As used herein, the term "subject" or "individual" is a mammal. Mammals include, but are not limited to, domesticated animals (e.g., cattle, sheep, cats, dogs, and horses), primates (e.g., humans and non-human primates such as monkeys), rabbits, and rodents (e.g., mice and rats). In certain embodiments, the individual or subject is a human.

[0077] Test Sample: A tumor sample is obtained from a subject with a result of the location at the time the sample was obtained.

[0078] Tissue sample: As used herein, the term "tissue sample" shall refer to a sample of cells that preserves the cross-sectional spatial relationship between cells as they exist in the subject from which the sample is obtained.

[0079] Tumor Sample: A tissue sample is obtained from a tumor.

[0080] II. Description of Biomarkers

[0081] CD3: CD3 is a cell surface receptor complex that is often used as a defining biomarker for cells with a T cell lineage. The CD3 complex consists of four distinct polypeptide chains: CD3-γ chain, CD3-δ chain, CD3ε chain, and CD3-ζ chain. CD3-γ and CD3-δ each form heterodimers with CD3ε (εγ-homodimer and εδ-heterodimer), and homodimers with the CD3-ζ chain (ζζ-homodimer). Functionally, the εγ-homodimer, εδ-heterodimer, and ζζ-homodimer form a signaling complex with the T cell receptor complex. Exemplary sequences for human CD3-γ chain, CD3-δ chain, CD3 epsilon chain, and CD3-zeta chain (and isoforms and variants thereof) can be found in Uniprot Accession Nos. P09693 (for the canonical amino acid sequence of the sequence disclosed herein in SEQ ID NO: 1), P04234 (for the canonical amino acid sequence of the sequence disclosed herein in SEQ ID NO: 2), P07766 (for the canonical amino acid sequence of the sequence disclosed herein in SEQ ID NO: 3), and P20963 (for the canonical amino acid sequence of the sequence disclosed herein in SEQ ID NO: 4), respectively. As used herein, the term "human CD3 protein biomarker" encompasses any CD3-γ chain, CD3-δ chain, CD3ε chain, and CD3-ζ chain polypeptides having the classical human sequence and natural variants thereof that retain the function of the classical sequence; εγ-homodimers, εδ-heterodimers, and ζζ-homodimers comprising one or more of the polypeptides of the CD3-γ chain, CD3-δ chain, CD3ε chain, and CD3-ζ chain having the classical human sequence and natural variants thereof that retain the function of the classical sequence; and any signaling complex comprising one or more of the aforementioned CD3 homodimers or heterodimers. In some embodiments, human CD3 protein biomarker-specific agents encompass any biomarker-specific agent that specifically binds to a structure (e.g., an epitope) within a CD3-γ chain polypeptide (e.g., a polypeptide in SEQ ID NO: 1), a CD3-δ chain polypeptide (e.g., a polypeptide in SEQ ID NO: 2), a CD3-ε chain polypeptide (e.g., a polypeptide in SEQ ID NO: 3), or a CD3-ζ chain polypeptide (e.g., a polypeptide in SEQ ID NO: 4), or to a structure (e.g., an epitope) located within an εγ-homodimer, an εδ-heterodimer, or a ζζ-homodimer.

[0082] CD8: CD8 is a heterodimeric, disulfide-linked transmembrane glycoprotein present in cytotoxic suppressor T cell subsets, thymocytes, certain natural killer cells, and myeloid subsets. Exemplary sequences for the α- and β-chains (and their subtypes and variants) of the human CD8 receptor can be found in Uniprot Accession Nos. P01732 (for the canonical amino acid sequence of the sequence disclosed herein in SEQ ID NO: 5) and P10966 (for the canonical amino acid sequence of the sequence disclosed herein in SEQ ID NO: 6), respectively. The term "human CD8 protein biomarker" as used herein encompasses any CD8-α chain polypeptide having a canonical human sequence and its natural variants that retain the function of the canonical sequence; any CD8-β chain polypeptide having a canonical human sequence and its natural variants that retain the function of the canonical sequence; including any CD8-α chain polypeptide having a canonical human sequence and its natural variants that retain the function of the canonical sequence and / or any CD8-β chain polypeptide having a canonical human sequence and its natural variants that retain the function of the canonical sequence. In some embodiments, human CD8 protein biomarker-specific agents encompass any biomarker-specific agent that specifically binds to a structure (e.g., an epitope) within a CD8-α chain polypeptide (e.g., a polypeptide in SEQ ID NO: 5), a CD8-β chain polypeptide (e.g., a polypeptide in SEQ ID NO: 6), or to a structure (e.g., an epitope) located within a CD8 dimer.

[0083] CD68: CD68 is a glycoprotein encoded by the CD68 gene located at 17p13.1 on chromosome 17. The CD68 protein is present in the cytoplasmic granules of various blood cells and muscle cells and is often used as a biomarker for macrophage lineage cells (including monocytes, tissue cells, giant cells, Kupffer cells and osteoclasts). Exemplary sequences for human CD68 (and its subtypes and variants) can be found in Uniprot Accession No. P34810 (for the canonical amino acid sequence of the sequence disclosed in SEQ ID NO: 7 herein). As used herein, the term "human CD68 protein biomarker" encompasses any CD68 polypeptide having a canonical human sequence and its natural variants that retain the function of the canonical sequence. In some embodiments, human CD20 protein biomarker-specific agents encompass any biomarker-specific agents that specifically bind to a structure (e.g., an epitope) within a human CD68 polypeptide (e.g., a polypeptide in SEQ ID NO: 7).

[0084] Pan-cytokeratin: As used herein, "pan-cytokeratin" and "PanCK" refer to any biomarker-specific reagent or group of biomarker-specific reagents that specifically binds to a sufficient number of cytokeratins to specifically stain epithelial tissue in a tissue sample. Exemplary pan-cytokeratin biomarker-specific reagents generally include: (a) a single cytokeratin-specific reagent that recognizes an epitope common to multiple cytokeratins, wherein a majority of epithelial cells of a tissue express at least one of the multiple cytokeratins; or (b) a mixture of biomarker-specific reagents such that the mixture is specifically reactive with multiple cytokeratins, wherein a majority of epithelial cells of a tissue express at least one of the multiple cytokeratins. References to "mixtures" in this definition include both a single composition comprising each member of a plurality of components and providing each member of a plurality of components as separate compositions, but staining them with a single dye or a combination thereof. Review of PanCK mixtures by NordiQC. In some embodiments, the PanCK biomarker specific reagent includes an antibody mixture containing two or more antibody clones selected from the group consisting of 5D3, LP34, AE1, AE2, AE3, MNF116, and PCK-26. In a certain embodiment, the PanCK mixture is selected from the group consisting of: a mixture of AE1 and AE3, a mixture of AE1, AE3, and 5D3, and a mixture of AE1, AE3, and PCK26. A mixture of AE1 and AE3 is commercially available from Agilent Technologies (Catalog Nos. GA05361-2, IS05330-2, IR05361-2, M351501-2, and M351529-2). A mixture of AE1, AE3, and 5D3 is commercially available from BioCare (Catalog Nos. CM162, IP162, OAI162, and PM162) and Abcam (Catalog No. ab86734). A mixture of AE1, AE3, and PCK26 is available from Roche (Cat. No. 760-2135).

[0085] PD-1: Programmed death-1 (PD-1) is a member of the CD28 receptor family encoded by the PDCD1 gene on chromosome 2. Exemplary sequences for human PD-1 proteins (and isoforms and variants thereof) can be found in Uniprot Accession No. Q15116 (canonical amino acid sequence for the sequence disclosed herein in SEQ ID NO: 8). In some embodiments, human PD-1 protein biomarker-specific agents encompass any biomarker-specific agent that specifically binds to a structure (e.g., an epitope) within a human PD-1 polypeptide (e.g., a polypeptide in SEQ ID NO: 8).

[0086] PD-L1: Programmed death ligand 1 (PD-L1) is a type 1 transmembrane protein encoded by the CD274 gene on chromosome 9. PD-L1 acts as a ligand for PD-1 and CD80. Exemplary sequences for human PD-L1 proteins (and isoforms and variants thereof) can be found in Uniprot accession number Q9NZQ7 (canonical amino acid sequence for the sequence disclosed herein in SEQ ID NO: 9). In some embodiments, human PD-L1 protein biomarker-specific agents encompass any biomarker-specific agent that specifically binds to a structure (e.g., an epitope) within a human PD-L1 polypeptide (e.g., a polypeptide in SEQ ID NO: 9).

[0087] LAG3: Lymphocyte activation gene 3 protein (LAG3) is a member of the immunoglobulin (Ig) superfamily encoded by the LAG3 gene on human chromosome 12. Exemplary sequences for human LAG3 proteins (and isoforms and variants thereof) can be found in Uniprot Accession No. P18627 (canonical amino acid sequence for the sequence disclosed herein in SEQ ID NO: 10). In some embodiments, human LAG3 protein biomarker-specific agents encompass any biomarker-specific agent that specifically binds to a structure (e.g., an epitope) within a human LAG3 polypeptide (e.g., a polypeptide in SEQ ID NO: 10).

[0088] III. Generation of Scoring Function

[0089] Figure 1is a flowchart showing an exemplary method for deriving a scoring function disclosed herein. The scoring function of the present method and system is typically derived from a tumor sample obtained from a patient cohort before treatment with a PD-1 axis-guided therapy, and its outcome data (e.g., 3-year or 5-year overall survival, progression-free survival, relapse-free survival, progressive disease, stable disease, partial response, complete response, etc.) 101 is available. A set of biomarkers to be tested is selected, and the cohort samples are stained for biomarkers 102 and imaged (usually with serial sections of morphological staining) 103. Regions of interest (ROIs) are identified in one or more digital images 104 and multi-biomarker features are extracted from one or more ROIs 105. The extracted features are evaluated by a feature selection function to identify those features associated with the response to the PD-1 axis by using a feature selection function 106. One or more selected features are modeled for the outcome using one or more modeling functions, candidate scoring functions are identified, and one or more cutoff values ​​are optionally selected to group the cohort according to their scores (e.g., "likely to respond" and "unlikely to respond" groups or "high likelihood of response" and "low likelihood of response") (e.g., using a ROC curve), and the cutoff values ​​are tested against each group using a Kaplan-Meier curve 107. Scoring function and cutoff value combinations that exhibit the desired separation between the groups are then selected for inclusion in the scoring systems and methods described herein.

[0090] III.A. Samples and Sample Preparation for Generating Scoring Functions

[0091] The scoring function is typically modeled 101 on tissue sections obtained from a cohort of subjects with tumors and known to respond to PD-1 axis-guided therapy. In some embodiments, the tumor is a solid tumor, such as cancer, lymphoma, or sarcoma. In a certain embodiment, the tumor is a tumor of skin, breast, head and / or neck, lung, upper gastrointestinal tract (including esophagus and stomach), female reproductive system (including uterine, fallopian tube and ovarian tumors), lower gastrointestinal tract (including colon, rectum and anus tumors), urogenital tract, exocrine, endocrine, kidney, nerve or lymphocyte origin. In a certain embodiment, the subject suffers from melanoma, breast cancer, ovarian cancer, pancreatic cancer, head and neck cancer, lung cancer, esophageal cancer, gastric cancer, colorectal cancer (including colon cancer, rectal cancer and anal cancer), prostate cancer, urothelial carcinoma or lymphoma. In a specific embodiment, the tumor is non-small cell lung cancer, head and neck squamous cell carcinoma, Hodgkin's lymphoma, urothelial carcinoma, gastric cancer, renal cell carcinoma, hepatocellular carcinoma, or colorectal cancer.

[0092] The sample 101 obtained is typically a tissue sample processed in a manner compatible with histochemical staining, including, for example, fixation, embedding in a wax matrix (e.g., paraffin), and slicing (e.g., using a microtome). The present disclosure does not require specific processing steps, as long as the sample obtained is compatible with the histochemical staining of the sample for the biomarker of interest and a digital image of the stained sample is generated. In a specific embodiment, the scoring function is modeled using a microtome slice of a formalin-fixed paraffin-embedded (FFPE) sample. In addition, in order to generate a scoring function, the sample 101 of the queue should be a sample with a known result, such as recurrence of the disease, progression of the disease, death caused by the disease, overall death, progressive disease, stable disease, partial response, and / or complete response.

[0093] III.B. Biomarker Panel

[0094] When generating the scoring function, at least one section of the sample is stained with a set of biomarker-specific reagents 102. These groups generally include at least one epithelial marker-specific reagent (e.g., Pan-CK-specific reagent), at least one immune cell-specific reagent (e.g., CD3-, CD8-, and / or CD68-specific reagent), and at least one PD-1 axis biomarker-specific reagent (e.g., PD-1-, PD-L1-, and / or PD-L2-specific reagent). In some embodiments, the group may further include one or more additional immune checkpoint biomarker-specific reagents, such as LAG3-specific reagents. In a certain embodiment, the biomarker-specific reagent group is selected from the group consisting of the following items: Group 1, including CD8, epithelial markers (EM), CD68, CD3, and PD-L1; and Group 2, including CD8, epithelial markers (EM), PD-L1, PD-1, and LAG3. Examples of epithelial markers useful in Groups 1 and 2 include cytokeratins. In a certain embodiment, the epithelial marker is a panel of cytokeratins stained with a panel of PanCK biomarker-specific reagents.

[0095] The group of biomarker specific reagents is used in combination with a group of appropriate detection reagents to generate biomarker-stained slices. Biomarker staining is usually accomplished by contacting the slice of the sample with a biomarker-specific reagent under conditions that specifically bind between the biomarker and the biomarker-specific reagent. The sample is then contacted with a group of detection reagents that interact with the biomarker-specific reagent to promote the deposition of the detectable part in the vicinity of the biomarker, thereby generating a detectable signal for positioning the biomarker. Typically, a washing step is performed between different reagents to prevent tissue from having unwanted nonspecific staining. The biomarker-stained slices can be optionally stained with a contrast agent (e.g., hematoxylin stain) to visualize macromolecular structures. In addition, the continuous slices of the biomarker-stained slices can be stained with morphological staining to promote the identification of ROI.

[0096] III.C.1. Labeling Protocols and Related Reagents

[0097] The biomarker-specific reagent facilitates detection of the biomarker by mediating the deposition of the detectable moiety in close proximity to the biomarker-specific reagent.

[0098] In some embodiments, the detectable portion is directly conjugated to the biomarker-specific reagent and is therefore deposited on the sample when the biomarker-specific reagent binds to its target (commonly referred to as a direct labeling method). Direct labeling methods can generally be more directly quantitative, but often lack sensitivity. In other embodiments, the deposition of the detectable portion is achieved by using a detection reagent associated with the biomarker-specific reagent (commonly referred to as an indirect labeling method). Indirect labeling methods increase the number of detectable portions that can be deposited near the biomarker-specific reagent, so indirect labeling methods are generally more sensitive than direct labeling methods, especially when used in combination with dyes.

[0099] In some embodiments, an indirect method is used, wherein the detectable moiety is deposited via an enzymatic reaction that positions a biomarker-specific reagent. Suitable enzymes for such reactions are well known and include, but are not limited to, oxidoreductases, hydrolases, and peroxidases. Specific enzymes explicitly included are horseradish peroxidase (HRP), alkaline phosphatase (AP), acid phosphatase, glucose oxidase, β-galactosidase, β-glucuronidase, and β-lactamase. The enzyme may be directly conjugated to a biomarker-specific reagent, or may be indirectly associated with a biomarker-specific reagent via a labeled conjugate. As used herein, a "labeled conjugate" includes:

[0100] (a) specific detection reagents; and

[0101] (b) An enzyme conjugated to a specific detection reagent, wherein the enzyme reacts with a chromogenic substrate, a signaling conjugate or an enzyme-reactive dye under appropriate reaction conditions to achieve in situ generation of the dye and / or deposition of the dye on the tissue sample.

[0102] In non-limiting examples, the specific detection reagent of the labeled conjugate can be a secondary detection reagent (e.g., a species-specific secondary antibody that binds to a primary antibody, an anti-hapten antibody that binds to a hapten-conjugated primary antibody, or a biotin-binding protein that binds to a biotinylated primary antibody), a tertiary detection reagent (e.g., a species-specific tertiary antibody that binds to a secondary antibody, an anti-hapten antibody that binds to a hapten-conjugated secondary antibody, or a biotin-binding protein that binds to a biotinylated secondary antibody), or other such arrangements. The enzyme thus localized to the sample-bound biomarker-specific reagent can then be used in a variety of protocols to deposit the detectable moiety.

[0103] In some cases, the enzyme reacts with a chromogenic compound / substrate. Specific non-limiting examples of chromogenic compounds / substrates include 4-nitrophenyl phosphate (pNPP), Fast Red, bromochloroindolyl phosphate (BCIP), nitro blue tetrazolium (NBT), BCIP / NBT, Fast Red, AP Orange, AP Blue, tetramethylbenzidine (TMB), 2,2'-azino-di-[3-benzothiazoline sulfonate] (ABTS), o-dianisidine, 4-chloronaphthol (4-CN), nitrophenyl-β-D-galactoside (O-D-galactoside), 4- ... NPG), o-phenylenediamine (OPD), 5-bromo-4-chloro-3-indolyl-β-galactoside (X-Gal), methylumbelliferyl-β-D-galactoside (MU-Gal), p-nitrophenyl-α-D-galactoside (PNP), 5-bromo-4-chloro-3-indolyl-β-D-glucuronide (X-Gluc), 3-amino-9-ethylcarbazole (AEC), basic fuchsin, iodonitrotetrazolium (INT), tetrazolium blue, or tetrazolium violet.

[0104] In some embodiments, the enzyme can be used in a metallographic detection scheme. The metallographic detection method includes combining an enzyme such as alkaline phosphatase with a water-soluble metal ion and a redox-inactive substrate of the enzyme. In some embodiments, the substrate is converted into a redox-active agent by the enzyme, and the redox-active agent reduces the metal ion to form a detectable precipitate. (See, for example, U.S. Patent Application No. 11 / 015,646, PCT Publication No. 2005 / 003777, and U.S. Patent Application Publication No. 2004 / 0265922, filed on December 20, 2004; each of which is incorporated herein by reference in its entirety.) The metallographic detection method includes using an oxidoreductase (such as horseradish peroxidase) and a water-soluble metal ion, an oxidant, and a reductant to form a detectable precipitate again. (See, for example, U.S. Patent No. 6,670,113, which is incorporated herein by reference in its entirety.)

[0105] In some embodiments, the enzymatic action occurs between the enzyme and the dye itself, wherein the reaction converts the dye from an unbound species to a species that is deposited on the sample. For example, the reaction of DAB with a peroxidase (e.g., horseradish peroxidase) oxidizes the DAB, causing it to precipitate.

[0106] In other embodiments, the detectable portion is deposited via a signaling conjugate comprising a latent reactive portion, which is configured to react with an enzyme to form a reactive substance that can be combined with a sample or other detection components. These reactive substances can react with the sample at the proximal end where they are generated, i.e., near the enzyme, but are rapidly converted into non-reactive substances so that the signaling conjugate is not deposited at the distal end away from the site of the deposited enzyme. Examples of potential reactive portions include: quinone methide (QM) analogs (such as those described in WO2015124703A1), and tyramine conjugates (such as those described in WO2012003476A2), each of which is incorporated herein by reference in its entirety. In some examples, the latent reactive moiety is directly conjugated to a dye, such as N,N'-dicarboxypentyl-5,5'-disulfonic acid-indole-dicarbocyanine (Cy5), 4-(dimethylamino)azobenzene-4'-sulfonamide (DABSYL), tetramethylrhodamine (DISCO violet), and rhodamine 110 (rhodamine). In other examples, the latent reactive moiety is conjugated to one member of a specific binding pair, and the dye is attached to the other member of the specific binding pair. In other examples, the latent reactive moiety is attached to one member of the specific binding pair, and the enzyme is attached to the other member of the specific binding pair, wherein the enzyme (a) is reactive with a chromogenic substrate to achieve the formation of the dye, or (b) is reactive with the dye to achieve the deposition of the dye (e.g., DAB). Examples of specific binding pairs include:

[0107] (1) biotin or a biotin derivative (e.g., desthiobiotin) linked to a potentially reactive moiety, and a biotin-binding entity (e.g., avidin, streptavidin, deglycosylated avidin (e.g., NEUTRAVIDIN), or a biotin-binding protein having a nitrotyrosine at the biotin-binding site (e.g., CAPTAVIDIN)) linked to a dye or reactive with a chromogenic substrate or reactive with a dye (e.g., peroxidase linked to a biotin-binding protein when the dye is DAB); and

[0108] (2) A hapten linked to a potentially reactive moiety and an anti-hapten antibody linked to a dye or an enzyme reactive with a chromogenic substrate or a dye (eg, peroxidase linked to a biotin-binding protein when the dye is DAB).

[0109] Specifically included are non-limiting examples of biomarker-specific reagents and detection reagent combinations listed in Table 1.

[0110] Table 1

[0111]

[0112]

[0113]

[0114]

[0115]

[0116]

[0117] In certain embodiments, the biomarker-specific reagents and specific detection reagents listed in Table 1 are antibodies. As will be appreciated by one of ordinary skill in the art, the detection protocols for each biomarker-specific reagent may be the same, or may be different.

[0118] Non-limiting examples of commercially available detection reagents or kits comprising detection reagents suitable for use in the methods of the invention include: VENTANA ultraView Detection System (secondary antibodies conjugated to enzymes including HRP and AP); VENTANA iVIEW Detection System (biotinylated anti-species secondary antibodies and streptavidin-conjugated enzymes); VENTANA OptiView Detection System (OptiView) (anti-species secondary antibodies conjugated to haptens and anti-hapten tertiary antibodies conjugated to enzyme polymers); VENTANA Amplification Kit (unconjugated secondary antibodies that can be used with any of the aforementioned VENTANA Detection Systems to amplify the amount of enzyme deposited at the binding site of the primary antibody); VENTANA OptiView Amplification System (anti-species secondary antibody conjugated to a hapten, anti-hapten tertiary antibody conjugated to an enzyme polymer, and tyramide conjugated to the same hapten. In use, the secondary antibody is contacted with the sample to achieve binding to the primary antibody. The sample is then incubated with the anti-hapten antibody to achieve association of the enzyme with the secondary antibody. The sample is then incubated with tyramide to achieve deposition of additional hapten molecules. The sample is then incubated again with the anti-hapten antibody to achieve deposition of additional enzyme molecules. The sample is then incubated with a detectable moiety to achieve dye deposition); VENTANA DISCOVERY, DISCOVERY OmniMap, DISCOVERY UltraMap anti-hapten antibodies, secondary antibodies, chromogens, fluorophores and dye kits, each of which is available from Ventana Medical Systems, Inc. (Tucson, Arizona); PowerVision and PowerVision+ IHC Detection Systems (secondary antibodies polymerized directly with HRP or AP into compact polymers carrying a high ratio of enzyme to antibody); and DAKO EnVision TM + system (enzyme-labeled polymer conjugated to a secondary antibody).

[0119] III.C.2. Multiple labeling scheme

[0120] In certain embodiments, the biomarker-specific reagents and detection reagents are applied in a multiplex staining approach. In a multiplex approach, biomarker-specific reagents and detection reagents are applied in a manner that allows for differential labeling of different biomarkers.

[0121] One method of accomplishing differential labeling of different biomarkers is to select combinations of biomarker-specific reagents, detection reagents, and enzyme combinations that do not result in off-target cross-reactivity between different antibodies or detection reagents (referred to as "combined staining"). For example, in the case of using secondary detection reagents, each secondary detection reagent is only capable of binding to one of the primary antibodies used on the slice. For example, primary antibodies derived from different animal species (e.g., mouse, rabbit, rat, and goat antibodies) can be selected, in which case species-specific secondary antibodies can be used. As another example, each primary antibody can include a different hapten or epitope tag, and the secondary antibody is selected to specifically bind to the hapten or epitope tag. In addition, each set of detection reagents should be suitable for depositing different detectable entities on the slice, such as by depositing different enzymes near each biomarker-specific reagent. Examples of such arrangements are shown in US 8,603,765. Such an arrangement has the potential advantage of enabling each set of biomarker-specific reagents and associated specific binding reagents to be present on the sample at the same time and / or to be stained with a mixture of biomarker-specific reagents and detection reagents, thereby reducing the number of staining steps. However, such an arrangement may not always be possible because reagents may cross-react with different enzymes and various antibodies may cross-react with each other, resulting in aberrant staining.

[0122] Another method for accomplishing differential labeling of different biomarkers is to stain the sample for each biomarker in turn. In such embodiments, a first biomarker-specific reagent is reacted with the slice, followed by a second detection reagent to the first biomarker-specific reagent and other detection reagents reacting with the slice, resulting in the deposition of a first detectable entity. The slice is then treated to remove the biomarker-specific reagent and the associated detection reagent from the slice, while appropriately retaining the deposited stain. The process is repeated for subsequent biomarker-specific reagents. Examples of methods for removing biomarker-specific reagents and associated detection reagents include heating the sample in the presence of a buffer, which elutes the antibody from the sample (referred to as a "heat-killing method"), such as those disclosed by Stack et al. in Multiplexed immunohistochemistry, imaging, and quantitation: A review, with an assessment of Tyramide signal amplification, multispectral imaging and multiplex analysis, Methods, Vol. 70, No. 1, pp. 46–

[0123] 58 (November 2014), and PCT / EP2016 / 057955, the contents of which are incorporated herein by reference.

[0124] As will be appreciated by those skilled in the art, combined staining and sequential staining methods can be combined. For example, in the case where only a subset of primary antibodies are compatible with combined staining, the sequential staining method can be modified, wherein the antibodies compatible with combined staining are applied to the sample using the combined staining method, and the remaining antibodies are applied using the sequential staining method.

[0125] III.C.3. Counterstaining

[0126] If desired, biomarker-stained slides can be counterstained to aid in identifying morphologically relevant regions for manual or automated ROI identification. Examples of counterstains include chromogenic nuclear counterstains such as hematoxylin (stains blue to purple), methylene blue (stains blue), toluidine blue (stains nuclei dark blue and polysaccharides pink to red), Nuclear Fast Red (also known as Kernechtrot dye, stains red), and methyl green (stains green); non-nuclear chromogenic stains such as eosin (stains pink); silver and stains including 4',6-diamino-2-phenylindole (DAPI, stains blue), propidium iodide (stains red), Hoechst stain (stains blue), Nuclear Green DCS1 (stains green), and Nuclear Yellow (Hoechst S769121, stains yellow at neutral pH and blue at acidic pH), DRAQ5 (stains red), DRAQ7 (stains red); fluorescent non-nuclear stains such as fluorophore-labeled phalloidin, (stains for fibrillar actin, the color depends on the conjugated fluorophore).

[0127] III.C.4. Morphological staining of samples

[0128] In certain embodiments, it is also desirable to perform morphological staining on a series of biomarker-stained slices 102. The slices can be used to identify the ROI 103 from which the score is to be performed. Basic morphological staining techniques typically rely on staining the nuclear structure with a first dye and staining the cytoplasmic structure with a second stain. Many morphological stains are known, including but not limited to hematoxylin and eosin (H&E) stains and Li's stains (methylene blue and basic fuchsin). In a particular embodiment, at least one continuous slice of each biomarker-stained slide is H&E-stained. Any method for applying H&E staining can be used, including manual and automatic methods. In a certain embodiment, at least one slice of the sample is a H&E-stained sample stained on an automatic staining system. Automated systems for performing H&E staining typically operate based on one of the following two staining principles: batch staining (also known as "immersion basket") or single slide staining. Batch stainers typically use barrels or cylinders of reagents, in which many slides are immersed at the same time. On the other hand, a single slide stainer applies reagents directly to each slide, and no two slides share the same equal parts of reagents. Examples of commercially available H&E stainers include the VENTANA SYMPHONY (single slide stainer) and VENTANA HE 600 (single slide stainer) series H&E stainers from Roche; CoverStainer (batch stainer) from Agilent Technologies; Leica ST4020 Small Linear Stainer (batch stainer), Leica ST5020 Multistainer (batch stainer), and Leica ST5010 Autostainer XL series (batch stainer) H&E stainers from Leica Biosystems Nussloch GmbH.

[0129] III.D. ROI, Objects, and Features

[0130] In a certain embodiment, one or more objects associated with a group of biomarkers are identified in a digital image of a biomarker-stained sample 104. The number of objects and / or the relationship between different objects and another object is used to define features to be evaluated for the development of a scoring function. A non-limiting example group of potential objects that can be detected from each group is listed in Table 2 below:

[0131]

[0132]

[0133] Table 2

[0134] In some embodiments, one or more regions of interest (ROIs) are also identified in the digital image of the biomarker-stained sample 104. The ROIs encompass biologically relevant locations of the tissue slice from which relevant objects are identified for feature calculation. In one embodiment, the ROIs contain morphological regions of the tissue slice of the tumor, such as the tumor region (TR), the invasive front, and the peritumoral (PT) region.

[0135] The ROI may be limited to the morphological region, may be extended to include regions outside the morphological region (i.e., by extending the edge of the ROI to a defined distance outside the morphological region), or may be limited to a subregion of the morphological region (e.g., by shrinking the ROI to a defined distance within the circumference of the morphological region, or by identifying regions within the ROI that have certain features (e.g., baseline density of certain cell types). At a morphological region defined by the edge (e.g., an invasive front), the ROI may be defined as, for example, all points within a defined distance of any point at the edge, all points to one side of a defined distance of any point at the edge, a minimal geometric region (e.g., a circle, ellipse, square, rectangle, etc.) encompassing the entire edge region, all points within a circle of a defined radius centered at a center point of the edge region, etc.

[0136] In some embodiments, the same ROI can be used for all slices and biomarkers. For example, a morphologically defined ROI can be identified in an H&E stained slice of a sample and used for all biomarker stained slices. In other embodiments, different ROIs can be used for different biomarkers. For example, an H&E stained slide can be used to identify a specific morphological region, such as a tumor region, used as a first ROI. A second ROI or multiple ROIs can then be identified in one of the biomarker stained slices, for example, to identify an area with a class of cells at a certain threshold density (e.g., epithelium versus stromal region). The second ROI or multiple ROIs can then be used for feature calculations.

[0137] Table 3 shows non-limiting examples of different ROIs:

[0138]

[0139] Table 3

[0140] In some embodiments, ROIs are manually identified in digital images. For example, a trained expert can manually delineate one or more morphological regions (e.g., tumor regions and / or invasion fronts) on a digital image of a sample. The one or more regions delineated in the image can then be used as ROIs for feature calculations or as reference points for calculating ROIs.

[0141] In other embodiments, the computer-implemented system can assist the user in annotating the ROI (referred to as "semi-automatic ROI annotation"). For example, a user can depict one or more regions on a digital image, which are then automatically converted by the system into a complete ROI. For example, if the desired ROI is the PI, PO, and / or PR regions, the user can depict the tumor region and the infiltration front, and the system automatically draws the PI, PO, and PR regions according to the user's definition. In another embodiment, where the ROI is EA or SA, the user can draw the tumor region and optionally the infiltration front in the image, which is then registered to the biomarker-stained image, and the system creates the associated EA and SA ROIs by marking all cells within a predetermined distance of the EM+ cells as within the EA and marking all cells beyond the predetermined distance as within the SA. In another embodiment, the system can also apply a pattern recognition function that uses computer vision and machine learning to identify regions with morphological features similar to the depicted and / or automatically generated regions. Thus, for example, tumor regions can be annotated in a semi-automated manner by the following method:

[0142] (a) The user annotates the tumor region by outlining the tumor region in the H&E image of the sample;

[0143] as well as

[0144] (b) The computer system applies pattern recognition functions to identify additional regions of the sample having morphological features of the contour region, wherein the entire tumor region includes the region annotated by the user and the region automatically identified by the system.

[0145] In another example, PR, PI and / or PO ROIs may be annotated in a semi-automated manner by a method comprising:

[0146] (a) The user annotates the tumor region by outlining the tumor region and the invasion front in the H&E image of the sample; and

[0147] (b) the computer system automatically defines PR, PI and / or PO regions that encompass all pixels within a defined distance of the annotated invasion front; and

[0148] (c) The computer system applies pattern recognition functions to identify additional regions of the sample that have morphological characteristics of the PI, PO and / or PR regions identified in step (b).

[0149] Many other arrangements may also be used. In the case of semi-automatic generation of the ROI, the user may be given the option to modify the ROI annotated by the computer system, for example by zooming in on the ROI, annotating areas of the ROI or objects within the ROI to exclude them from analysis, etc.

[0150] In other embodiments, the computer system may automatically suggest ROIs without any direct input from the user (referred to as "automatic ROI annotation"). For example, a previously stained tissue segmentation function or other pattern recognition function may be applied to an unannotated image to identify a desired morphological region to use as an ROI. The user may be given the option to modify the ROI annotated by the computer system, such as by zooming in on the ROI, annotating a region of the ROI, or an object within the ROI to exclude it from analysis, etc.

[0151] One or more features are extracted from one or more ROIs and quantified to obtain a feature measure for each sample 105. Exemplary features include, for example, the total number of objects in the ROI, the density of a particular object in the ROI, the spatial relationship between different objects in the ROI, the spatial distribution of a particular object within the ROI, the ratio of the number and / or density of different objects within the ROI, the ratio of the same objects in different ROIs (e.g., the ratio of a particular cell in the EA ROI relative to the SA ROI or the ratio of a particular cell in the PI ROI relative to the PO ROI), the fraction of the total objects in the larger ROI that fall into the smaller ROI that fall into the larger ROI (e.g., the fraction of a particular cell type in the TA ROI that falls into the EA, SA, PI, PO, or PT ROI). Table 4 lists specific exemplary features for each group:

[0152] Table 4

[0153]

[0154]

[0155]

[0156]

[0157]

[0158]

[0159]

[0160]

[0161]

[0162] Unless otherwise stated, the ROIs of the features in Table 4 are tumor areas. Unless otherwise stated, any density listed in Table 4 is an area density (i.e., the number of positive cells over the entire ROI area). As used in Table 4, "PD1 low", "PD1 medium" and "PD1 high" refer to individual cells with low, medium and high PD-1 staining intensities. In a certain embodiment, "PD1 low" cells are PD-1+ cells with the lowest staining intensity of all measured PD-1+ cells in all tested samples, "PD1 medium" cells are PD-1+ cells with a medium staining intensity of all measured PD-1+ cells in all tested samples, and "PD1 high" cells are PD-1+ cells with the highest staining intensity of all measured PD-1+ cells in all tested samples.

[0163] III.F. Modeling Scoring Function

[0164] To identify scoring functions, the ability of features to predict their relative likelihood of response to a course of PD-1 axis directed therapy was modeled.

[0165] In a certain embodiment, features can be selected by executing feature selection function 106. The feature measurement and result data of each member in the queue are input into the feature selection function, and then the data is used to rank these features according to the relative relevance of different features to the desired results. Exemplary feature selection functions include set feature selection functions (including, for example, random forest functions), filtering method functions (including, for example, mutual information-based functions (mRMR) / functions based on correlation coefficients and functions based on relief) and / or embedded feature selection functions (such as elastic network / minimum absolute shrinkage functions or selection operators (LASSO) functions). In a certain embodiment, the candidate model is made using the first 25, first 20, first 15, first 10, first 9, first 8, first 7, first 6, first 5, first 4 or first 3 features identified by the feature selection function. In another embodiment, the candidate model uses at least 1, at least 2, at least 3, at least 4 or at least 5 features identified in the first 10 features of at least two feature selection functions. In another embodiment, the candidate model includes at least one feature present in the first 5 features of at least 2 feature selection functions. In certain embodiments, "responders" are considered to be patients with a partial response or a complete response. In certain embodiments, "responders" are considered to be patients with stable disease, a partial response, or a complete response.

[0166] Candidate model is generated by inputting the selected feature measurement and result data of each member in the queue into the modeling function. The model with the highest consistency with the response is selected as the scoring function. Exemplary modeling functions include quadrant discriminant analysis (QDA), linear discriminant analysis (LDA), support vector machine (SVM) and artificial neural network (ANN). In a certain embodiment, the candidate function is modeled only on the features extracted from the digital image. In other embodiments, the candidate function includes other clinical variables, such as age, sex, mismatch repair status, and / or microsatellite instability status. In a certain embodiment, the model is used to predict the possibility of a partial or complete response to therapy for progressive disease after treatment relative to a stable disease after treatment. In a certain embodiment, the model is used to predict the possibility that a patient will suffer from a progressive disease after treatment relative to the possibility that the patient will suffer from a stable disease, partially respond to therapy or respond completely. In a certain embodiment, the model is used to predict the possibility that a patient will suffer from a progressive disease or a stable disease after treatment relative to the possibility that the patient has a partial or complete response to therapy.

[0167] In addition, one or more stratification cutoffs can be selected to differentiate patients into “risk intervals” (e.g., “high risk” and “low risk” quartiles, deciles, etc.) based on relative risk 107. In one example, the stratification cutoffs are selected using a receiver operating characteristic (ROC) curve. The ROC curve allows the user to balance the sensitivity of the model (i.e., preferentially capturing as many “positive” or “likely to respond” candidates as possible) with the specificity of the model (i.e., minimizing false positives for “likely to respond” candidates). In one embodiment, a cutoff is selected between the risk intervals of likely response and unlikely response, and the selected cutoff has a balanced sensitivity and specificity. In one embodiment, the stratification cutoff distinguishes (a) patients who are likely to have progressive disease after treatment and (b) patients who are likely to have stable disease, partial response, or complete response to therapy. In one embodiment, the stratification cutoff distinguishes (a) patients who are likely to have progressive disease after treatment, (b) patients who are likely to have stable disease after treatment, and (c) patients who are likely to have a partial response or complete response to therapy. In certain embodiments, the stratification cutoff value distinguishes between (a) patients who are likely to have progressive disease or stable disease following treatment and (b) patients who are likely to have a partial or complete response to therapy.

[0168] If necessary, the model can be performed using a computer statistical analysis software suite (e.g., R Project for Statistical Computing (available from https: / / www.r-project.org / ), SAS, MATLAB, etc.).

[0169] IV. Scoring with Scoring Function

[0170] After the scoring function is modeled and the optional stratification cutoffs are selected, the scoring function can be applied to an image of a test sample to calculate a score for the test sample. Figure 2 is a flow chart illustrating an exemplary method for scoring a test sample using the scoring function described above. First, a tumor tissue slice is obtained from a patient for whom PD-1 axis-directed therapy is being considered 201. Tissue slices are generally similar to the sample type used to model the scoring function, except that the results are not yet clear. If ROI selection is required, at least one tissue slice is stained for a biomarker associated with the scoring function, and its serial sections are stained with a morphological stain (e.g., H&E) 202. The stained slices are imaged 203, and one or more ROIs associated with the scoring function and any objects used in the calculation of relevant feature metrics are annotated in the biomarker-stained image 204. Relevant features are extracted from the ROI, and feature metrics for each feature are calculated 205. A feature vector including all variables used by the scoring function 206 is then assembled, and the scoring function is applied to the feature vector 207. In some cases, the variables are simply feature metrics extracted from the ROI. In other cases, additional clinical variables may include, for example, age, sex, mismatch repair status (e.g., the patient has defective MMR (dMMR) or non-defective MMR (pMMR)), microsatellite instability status (e.g., the patient is MSI 高 or MSI 低 If stratified cutoffs are used, the output score may also estimate relevant risk intervals. The score may then be integrated by a clinician into diagnostic and / or treatment decisions, including, for example, by integrating the score with other clinical variables that may weigh in the decision whether to administer PD-1 axis-directed therapy.

[0171] In one embodiment, the scoring function is integrated into a scoring system. An exemplary scoring system is shown in Figure 3 .

[0172] The scoring system includes an image analysis system 300. The image analysis system 300 may include one or more computing devices, such as a desktop computer, a laptop computer, a tablet computer, a smart phone, a server, a dedicated computing device, or any other one or more types of one or more electronic devices capable of performing the techniques and / or operations described herein. In some embodiments, the image analysis system 300 may be implemented as a single device. In other embodiments, the image analysis system 300 may be implemented as a combination of two or more devices that together implement the various functions discussed herein. For example, the image analysis system 300 may include one or more server computers and one or more client computers that are communicatively coupled to each other via one or more local area networks and / or a wide area network such as the Internet.

[0173] like Figure 3 As shown, the image analysis system 300 may include a memory 314, a processor 315, and a display 316. The memory 314 may include any combination of any type of volatile or non-volatile memory, such as random access memory (RAM), read-only memory such as electrically erasable programmable read-only memory (EEPROM), flash memory, hard disk drives, solid-state drives, optical disks, etc. For simplicity, the memory 314 may include any combination of any type of volatile or non-volatile memory, such as random access memory (RAM), read-only memory such as electrically erasable programmable read-only memory (EEPROM), flash memory, hard disk drives, solid-state drives, optical disks, etc. Figure 3 Memory 314 is depicted as a single device in FIG. 3 , but it should be understood that memory 314 may also be distributed across two or more devices.

[0174] Processor 315 may include one or more processors of any type, such as a central processing unit (CPU), a graphics processing unit (GPU), a dedicated signal or image processor, a field programmable gate array (FPGA), a tensor processing unit (TPU), etc. For simplicity, Figure 3 The processor 315 is depicted as a single device in FIG. 3 , but it should be understood that the processor 315 may also be distributed across any number of devices.

[0175] The display 316 can be implemented using any suitable technology, such as LCD, LED, OLED, TFT, plasma, etc. In some implementations, the display 316 can be a touch-sensitive display (touch screen).

[0176] like Figure 3 As shown, the image analysis system 300 may also include an object identifier 310, a region of interest (ROI) generator 311, a user interface module 312, and a scoring engine 313. Figure 3 100 is shown as an independent module, it is obvious to those skilled in the art that each module can be implemented as a sub-module alternatively, and in some embodiments, any two or more modules can be combined into a single module. In addition, in some embodiments, the system 100 may include Figure 3 For the sake of brevity, additional engines and modules (e.g., input devices, network and communication modules, etc.) are not shown. In addition, in some embodiments, Figure 3 Some of the boxes depicted in the system 100 may be disabled or omitted. As will be discussed in more detail below, the functionality of some or all of the modules of the system 100 may be implemented in hardware, software, firmware, or any combination thereof. Exemplary commercially available software packages that can be used to implement the modules disclosed herein include VENTANA VIRTUOSO; Definiens TISSUE STUDIO, DEVELOPER XD, and IMAGE MINER; and Visopharm BIOTOPIX, ONCOTOPIX, and STEREOTOPIX software packages.

[0177] After acquiring the image, the image analysis system 300 can pass the image to an object identifier 310, which is responsible for identifying and marking relevant objects and other features in the image, which will then be used for scoring. The object identifier 310 can extract (or generate for each image) multiple image features representing each object in the image and pixels representing the expression of one or more biomarkers. The extracted image features can include, for example, texture features, such as Haralick features, bag-of-words features, etc. The values ​​of multiple image features can be combined into a high-dimensional vector, hereinafter referred to as a "feature vector", which represents the expression of biomarkers related to the features of the scoring function. For example, if M features are extracted for each object and / or pixel, each object and / or pixel can be represented by an M-dimensional feature vector. The output of the object identifier 310 is effectively a map of the image, which marks the location of objects and pixels of interest and associates those objects and pixels with feature vectors describing the objects or pixels.

[0178] For biomarkers that are scored based on the association of the biomarker with a specific type of object (e.g., membrane, nucleus, cell, etc.), the features extracted by the object identifier 310 may include features or feature vectors sufficient to classify the objects in the sample as biomarker-positive objects of interest or biomarker-negative markers of interest and / or features or feature vectors classified by the level of intensity of the biomarker staining of the object. In the case where the biomarker can be weighted differently according to the type of object expressing the biomarker, the features extracted by the object identifier 310 may include features related to determining the type of object associated with the biomarker-positive pixel. Therefore, the object can then be classified based at least on the expression of the biomarker (e.g., biomarker-positive or biomarker-negative cells) and the subtype of the object (e.g., tumor cells, immune cells, etc.). In the case where the degree of biomarker expression is scored without considering the association with the object, the features extracted by the object identifier 310 may include, for example, the location and / or intensity of the biomarker-positive pixel. The precise features extracted from the image will depend on the type of classification function applied, and will be well known to those of ordinary skill in the art.

[0179] Examples of objects identified for certain biomarker panels are listed in Table 5 below:

[0180]

[0181] Table 5

[0182] The image analysis system 300 may also pass the image to a ROI generator 311. The ROI generator 311 is used to identify one or more ROIs of the image from the image for which the immune content score is calculated. In the event that the object identifier 310 is not applied to the entire image, the one or more ROIs generated by the ROI generator 311 may also be used to define a subset of the image on which the object identifier 310 is performed.

[0183] In one embodiment, the ROI generator 311 can be accessed through the user interface module 312. An image of a sample stained with a biomarker (or a continuous section of a morphological stain of a sample stained with a biomarker) is displayed on the graphical user interface of the user interface module 112, and the user marks one or more regions in the figure that are considered as ROIs. In this example, the ROI annotation can take a variety of forms. For example, the user can manually define the ROI (hereinafter referred to as "manual ROI annotation"). In other examples, the ROI generator 311 can assist the user in annotating the ROI (referred to as "semi-automatic ROI annotation"). For example, the user can depict one or more regions on the digital image, and the system automatically converts it into a complete ROI. For example, if the desired ROI is a tumor region, the user depicts the tumor region, and the system, for example, uses computer vision and machine learning to identify similar morphological regions. As another example, the user can annotate the edges in the image (e.g., by drawing a line defining the infiltration front of the tumor), and the ROI generator 311 can automatically define the ROI based on the user-defined edges. For example, a user can mark the edge of an invasion front or tumor region in the user interface module 312, and the ROI generator 311 uses the edge as a guide to create an ROI, for example, by drawing an ROI that encompasses all objects within a predetermined distance of the edge (e.g., a PT ROI), or within a predefined distance on one side of the edge (e.g., a PO or PI ROI), or within a first predefined distance on a first side of the edge and within a second predefined distance on a second side of the edge (e.g., a PT ROI with different standard distances from the invasion front inside and outside of it).

[0184] In other embodiments, the ROI generator 311 may automatically suggest ROIs without any direct input from the user (eg, by applying a tissue segmentation function to an unlabeled image), which the user may then choose to accept, reject, or edit as appropriate.

[0185] In some embodiments, the ROI generator 311 may also include a registration function, whereby a ROI annotated in one slice of a set of continuous slices is automatically registered in the other slices of the set of continuous slices. This function is particularly useful when H&E stained continuous slices are provided simultaneously with biomarker labeled slices. In such embodiments, a user can, for example, draw a tumor region in a digital image of an H&E stained slice. The ROI generator 311 then displays the ROI from the H&E image into the image of the biomarker stained continuous slice, thereby matching the tissue structure from the H&E image with the corresponding tissue structure in the continuous slice. Exemplary display methods can be found in, for example, WO2013 / 140070 and US2016-0321809.

[0186] The object identifier 310 and ROI generator 311 may be implemented in any order. For example, the object identifier 310 may be applied to the entire image first. Then, when the ROI generator 311 is implemented, the locations and features of the identified objects may be stored and recalled later. In such an arrangement, the scoring engine 313 may generate scores immediately when the ROI is generated. This workflow is described in detail in the following sections. Figure 4A As shown in Figure 4A As can be seen in , the obtained image has a mixture of different objects (shown by the dark ovals and dark diamonds). After implementing the object recognition task, all diamonds in the image will be identified (represented by the hollow diamonds). When a ROI is attached to the image (represented by the dashed line), only the diamonds located in the ROI region are included in the metric calculations for the ROI. The feature vector is then calculated, including the feature metrics and any additional metrics used by the scoring function executed by the scoring engine 313. Alternatively, the ROI generator 311 can be implemented first. In this workflow, the object identifier 310 can be implemented only on the ROI (which can minimize computation time), or it can still be implemented on the entire image (which can allow on-the-fly adjustments without rerunning the object identifier 310). This workflow is described in detail in detail in the accompanying drawings. Figure 4B As shown in Figure 4B As can be seen in the figure, the obtained image has a mixture of different objects (shown by the dark ovals and dark diamonds). The ROI is attached to the image (indicated by the dotted line), but the objects are not yet labeled. After implementing the object recognition task on the ROI, all diamonds in the ROI are identified (indicated by the hollow diamonds) and included in the feature metric calculation of the ROI. A feature vector is then calculated, including one or more feature metrics and any additional metrics used by the scoring function performed by the scoring engine 313. It is also possible to implement the object identifier 310 and the ROI generator 311 simultaneously.

[0187] After the object identifier 310 and the ROI generator 311 have been implemented, a scoring engine 313 is implemented. The scoring engine 313 calculates one or more feature metrics for the ROI and, if used, calculates predetermined maximum and / or minimum cutoff values. A feature vector including the calculated feature metrics and any other variables used by the scoring function is assembled by the scoring engine, and the scoring function is applied to the feature vector.

[0188] Table 6 lists specific exemplary characteristics of each group:

[0189] Table 6

[0190]

[0191]

[0192]

[0193]

[0194]

[0195]

[0196]

[0197]

[0198]

[0199]

[0200] like Figure 3 As shown in , in some embodiments, the image analysis system 300 can be communicatively coupled to an image acquisition system 320. The image acquisition system 320 can acquire images of the sample and provide those images to the image analysis system 300 for analysis and presentation to a user.

[0201] The image acquisition system 320 may include a scanning platform 325, such as a slide scanner, which can scan the stained slide at 20X, 40X, or other magnifications to produce a high-resolution digital image of the whole slide, including, for example, a slide scanner. At a basic level, a typical slide scanner includes at least: (1) a microscope with an objective lens, (2) a light source (e.g., halogen, light emitting diode, white light, and / or a multi-spectral light source, depending on the dye), (3) a robot to move the slide around (or to move the optics around the slide, (4) one or more digital cameras for image capture, and (5) a computer and associated software to control the robot and manipulate, manage, and view the digital slides. The camera's charge-coupled device (CCD) captures digital data at many different XY locations on the slide (in some cases, in multiple Z planes) and merges the images together to form a composite image of the entire scanned surface. Common methods for achieving this include:

[0202] (1) Tile-based scanning, where the stage or optics are moved in very small increments to capture square image frames that slightly overlap adjacent squares. The captured squares are then automatically matched to each other to construct a composite image; and

[0203] (2) Line-based scanning, in which the stage moves on a single axis during acquisition to capture many composite image “strips.” The image strips can then be matched to each other to form a larger composite image.

[0204] A detailed overview of various scanners (fluorescence and bright field) can be found in Farahani et al., Whole slide imaging in pathology: advantages, limitations, and emerging perspectives, Pathology and Laboratory Medicine Int'l, Vol. 7, pp. 23-33 (June 2015), the contents of which are incorporated herein by reference in their entirety. Examples of commercially available slide scanners include: 3DHistech PANNORAMICSCAN II; DigiPath PATHSCOPE; Hamamatsu NANOZOOMER RS, HT, and XR; Huron TISSUESCOPE4000, 4000XT, and HS; Leica SCANSCOPE AT, AT2, CS, FL, and SCN400; Mikroscan D2; Olympus VS120-SL; Omnyx VL4, and VL120; PerkinElmer LAMINA; Philips ULTRA-FAST SCANNER; Sakura Finetek VISIONTEK; Unic PRECICE 500, and PRECICE 600x; VENTANA ISCAN COREO and ISCAN Ht; and Zeiss AXIO SCAN.Z1. Other exemplary systems and features can be found, for example, in WO2011-049608) or U.S. Patent Application No. 61 / 533,114, filed on September 9, 2011, entitled “IMAGING SYSTEMS, CASSETTES, AND METHODS OF USING THESAME,” the contents of which are incorporated herein by reference in their entirety.

[0205] The images generated by the scanning platform 325 may be transmitted to the image analysis system 300 or a server or database accessible to the image analysis system 300. In some embodiments, the images may be automatically transmitted via one or more local area networks and / or wide area networks. In some embodiments, the image analysis system 300 may be integrated with or included in the scanning platform 325 and / or other modules of the image acquisition system 320, in which case the images may be transmitted to the image analysis system, for example, via a memory accessible to the platform 325 and the system 320. In some embodiments, the image acquisition system 320 may not be communicatively coupled to the image analysis system 300, in which case the images may be stored on any type of non-volatile storage medium (e.g., a flash drive) and may be downloaded from the medium to the image analysis system 300 or a server or database communicatively coupled thereto. In any of the above examples, the image analysis system 300 may obtain an image of a biological sample, wherein the sample may have been fixed on a slide and stained by a histochemical staining platform 323, and the slide may have been scanned by a slide scanner or other type of scanning platform 325. However, it should be understood that in other embodiments, the techniques described below may also be applied to images of biological samples acquired and / or stained by other means.

[0206] The image acquisition system 320 may also include an automated histochemical staining platform 323, such as an automated IHC / ISH slide stainer. An automated IHC / ISH slide stainer typically includes at least: a reservoir of various reagents used in the staining protocol, a reagent dispensing unit in fluid communication with the reservoir for dispensing reagents onto a slide, a waste removal system for removing used reagents or other waste from the slide, and a control system for coordinating the actions of the reagent dispensing unit and the waste removal system. In addition to performing the staining steps, many automated slide stainers can also perform staining auxiliary steps (or are compatible with separate systems that perform such auxiliary steps), including: slide baking (for adhering samples to slides), dewaxing (also known as deparaffinization), antigen retrieval, counterstaining, dehydration and clearing, and coverslipping. Prichard, Overview of Automated Immunohistochemistry, Arch Pathol Lab Med., Vol. 138, pp. 1578–1582 (2014), which is incorporated herein by reference in its entirety, describes several specific examples of automated IHC / ISH slide stainers and their various features, including the intelliPATH (Biocare Medical), WAVE (Celerus Diagnostics), DAKOOMNIS and DAKO AUTOSTAINER LINK 48 (Agilent Technologies), BENCHMARK (Ventana Medical Systems, Inc.), Leica BOND, and Lab Vision Autostainer (Thermo Scientific) automated slide stainers. In addition, Ventana Medical Systems, Inc. is the assignee of multiple U.S. patents disclosing systems and methods for performing automated analysis, including U.S. Patent Nos. 5,650,327, 5,654,200, 6,296,809, 6,352,861, 6,827,901, and 6,943,029, and U.S. Published Patent Application Nos. 20030211630 and 20040052685, each of which is incorporated herein by reference in its entirety.Commercially available staining devices generally operate according to one of the following principles: (1) open single slide staining, in which the slide is placed horizontally and the reagent is dispensed as a puddle onto the slide surface containing the tissue sample (such as implemented on the DAKO AUTOSTAINER Link 48 (Agilent Technologies) and intelliPATH (Biocare Medical) stainers); (2) liquid overlay technique, in which the reagent is covered or dispensed by a layer of inert fluid deposited on the sample (such as implemented on the VENTANA BenchMark and DISCOVERY stainers); (3) capillary gap staining, in which the slide surface is placed adjacent to another surface (possibly another slide or a cover plate) to form a narrow gap that draws and maintains the liquid reagent in contact with the sample by capillary forces (such as the staining principle used by the DAKOTECHMATE, Leica BOND, and DAKO OMNIS stainers). Some iterations of capillary gap staining do not mix the fluid in the gap (such as, on the DAKO TECHMATE and Leica BOND). In the variant of capillary gap staining called dynamic gap staining, the sample is applied to the slide using capillary force, and then the parallel surfaces are translated relative to each other to stir the reagent during incubation to achieve reagent mixing (such as, the staining principle implemented on the DAKO OMNIS stainer). In the translation gap staining, the translatable head is located on the slide. The lower surface of the head is spaced apart from the slide by a first gap, and the first gap is small enough to allow the meniscus of the liquid to be formed by the liquid on the slide during the slide translation. A mixing extension with a lateral dimension less than the width of the slide extends from the lower surface of the translatable head to define a second gap less than the first gap between the mixing extension and the slide. During the translation of the head, the lateral dimension of the mixing extension is enough to generate lateral motion in the liquid on the slide along the direction extending from the second gap to the first gap generally. Referring to WO 2011-139978 A1. Recently it is suggested that reagents are deposited on the slide using inkjet technology. See WO2016-170008 A1. The list of staining techniques is not intended to be comprehensive, and any fully automated or semi-automated system for performing biomarker staining may be integrated into the histochemical staining platform 323.

[0207] The image acquisition system 320 may include an automated H&E staining platform 324. Automated systems for performing H&E staining typically operate based on one of two staining principles: batch staining (also known as "immersion basket") or single slide staining. Batch stainers typically use a barrel or vat of reagent into which many slides are immersed simultaneously. Single slide stainers, on the other hand, apply the reagent directly to each slide, and no two slides share the same equal portion of reagent. Examples of commercially available H&E stainers include the VENTANA SYMPHONY (single slide stainer) and VENTANA HE 600 (single slide stainer) series H&E stainers from Roche; CoverStainer (batch stainer) from Agilent Technologies; Leica ST4020 Small Linear Stainer (batch stainer), Leica ST5020 Multistainer (batch stainer), and Leica ST5010 Autostainer XL series (batch stainer) H&E stainers from Leica Biosystems Nussloch GmbH. The H&E staining platform 324 is typically used in the workflow of serial sections that require morphological staining of biomarker-stained sections.

[0208] The scoring system may further include a laboratory information system (LIS) 330. The LIS 330 typically performs one or more functions selected from the following: recording and tracking processes performed on samples and slides and images derived from the samples, instructing different components of the scoring system to perform specific processes on samples, slides and / or images, and tracking information about specific reagents applied to samples and / or slides (e.g., batch number, expiration date, dispensed volume, etc.). The LIS 330 typically includes at least one database containing information about the sample; a label associated with the sample, slide and / or image file (e.g., a barcode (including a one-dimensional barcode and a two-dimensional barcode), a radio frequency identification (RFID) label, an alphanumeric code affixed to the sample, etc.); and a communication device that reads the label on the sample or slide and / or transmits information about the slide between the LIS 330 and other components of the immunoenvironment scoring system. Thus, for example, a communication device may be placed on each of the sample processing station, the automated histochemical stainer 323, the H&E staining platform 324, and the scanning platform 325. When the sample is initially processed into sections, information about the sample (e.g., patient ID, sample type, procedures to be performed on one or more sections) can be entered into the communication device and a label is created for each section generated from the sample. At each subsequent station, the label is entered into the communication device (e.g., by scanning a barcode or RFID tag or by manually entering an alphanumeric code), and the station communicates electronically with the database, for example, instructing the station or station operator to perform a specific procedure on the section and / or recording the procedure being performed on the section. At the scanning platform 325, the scanning platform 325 can also encode each image with a computer readable label or code that is associated back to the section or sample from which the image originated, so that when the image is sent to the image analysis system 300, the image processing steps to be performed can be sent from the database of the LIS 330 to the image analysis system, and / or the image processing steps performed on the image by the image analysis system 300 are recorded in the database of the LIS 330. Commercially available LIS systems useful in the present methods and systems include, for example, the VENTANA Vantage Workflow system (Roche).

[0209] Examples

[0210] I. Characterization of PD-L1, CD8, CD3, CD68, and PanCK in the Tumor Microenvironment of Gastrointestinal Tumors with Respect to Patient Mismatch Repair Status and Anti-PD-1 Therapy Outcomes Using 5Plex IHC and Whole Slide Image Analysis

[0211] IA Background Technology

[0212] There is an increasing need to understand tumor microenvironmental markers to guide cancer immunotherapy. Multiplexed immunohistochemistry (IHC) enables characterization of the tumor microenvironment while preserving tissue morphology by detecting multiple biomarkers and their co-expression on a single slide. Extracting information about the co-expression of multiple biomarkers and their spatial relationships requires whole slide image analysis algorithms customized for the individual assay and its intended use. Cancers can evade immune surveillance and eradication by upregulating the programmed death 1 (PD-1) pathway and its ligand programmed death ligand 1 (PD-L1) on tumor cells and in the tumor microenvironment. Blocking this pathway with antibodies against PD-1 or PD-L1 has led to significant clinical responses in some cancer patients.

[0213] Mismatch repair (MMR) deficiency predicts response to PD-1 blockade in solid tumors. However, not all patients with mismatch repair deficiency respond to PD-1 blockade therapy. To understand the differential responses, we assessed the tumor microenvironment by measuring PD-L1 expression associated with tumor cells and tumor-infiltrating immune cells.

[0214] IB samples, staining, and image acquisition

[0215] A cohort of 60 pre-treatment (anti-PD-1 pembrolizumab) patient gastrointestinal tumor samples with acceptable image and tissue quality for automated analysis were available for this study. After eliminating unevaluable responses, 54 cases remained. Table 7 shows the attenuation of responses with respect to mismatch repair deficiency.

[0216] Mismatch repair without defects Mismatch repair deficiency Progressive disease 13 5 Stable disease 3 10 Partial response + complete response 1 22

[0217] Table 7

[0218] Samples were formalin fixed, paraffin embedded, sectioned, and mounted on microscope slides.

[0219] Slides were stained in a multiplex format on the BenchMark ULTRA IHC / ISH Automated Slide Stainer with fluorescent tyramide dye conjugates in the tyramide signal amplification procedure as shown in Table 8:

[0220]

[0221] Table 8

[0222] The general concept of tyramide signal amplification is described by US 6,593,100. The staining steps are essentially the same as the procedure described by Zhang I. Figure 5 The stains were applied in the order described in . After each stain was deposited, a heat kill step was applied, which included the process described by Zhang I. An exemplary stained slide is shown in Figure 6Serial sections of each sample were also stained with H&E using a VENTANA HE 600 automated slide stainer.

[0223] IHC-stained slides were scanned on a Zeiss AxioScan Z1 slide scanner and H&E-stained slides were scanned on a VENTANA ISCAN COREO slide scanner. All images were exported to DPath, a proprietary digital pathology image analysis suite software suite from Roche.

[0224] IC image annotation and ROI generation

[0225] The pathologist annotated the tumor region ROI and the invasion front of the tumor (if applicable) in the images. In addition, the pathologist annotated the necrotic areas and other areas that were excluded from the analysis.

[0226] The DPath system automatically annotates epithelial tumor ROIs from aggregates of panCK+ cells, and stromal regions. First, the tumor region is subdivided into tiles. For each tile, a panCK mask is first generated by labeling each panCK+ cell and finding the union between labeled cells. Post-processing is then performed on the mask to compensate for infiltrating lymphocytes by:

[0227] (a) A morphological closing operation is performed on a disk-shaped structure element with a radius of 10 pixels (resolution per pixel is 0.325 μms);

[0228] (b) Close the mask with more than 80 pixels (8.5 μm 2 ) to generate a “PanCK mask to fill the holes”.

[0229] (c) Convert the PanCK mask with holes into polygons.

[0230] The polygons of all tiles were then assembled together to create polygons at the whole slide level. The whole slide level polygons were then converted to masks (lower resolution (reduced in size by 3^3)) and zoomed in by 8.8 μm to generate the final mask for “PanCK+ cell aggregates” at the whole slide level.

[0231] In addition, the inner ROI of the tumor is automatically generated from the region where the invasion front enters the tumor by 0.5 mm, and the outer ROI of the tumor is automatically generated from the region where the invasion front is away from the tumor by 0.5 mm.

[0232] IC characterization and data analysis

[0233] The following features were calculated for each ROI: area density of all phenotypes;+ panCK + The fraction of cells that are PD-L1 + CD3 + The fraction of cells that are PD-L1 + CD8 + The fraction of cells; also PD-L1 + CD3 + CD8 - The fraction of cells; CD8+ cells and their nearest neighbor PD-L1 + / CD68 + Descriptive statistics of the distance between a CD8+ cell and its nearest neighbor, PD-L1 + / panCK + Descriptive statistics of the distance between a CD8+ cell and its nearest neighbor, PD-L1 + / CD3 + Descriptive statistics of distance; panCK + The cell and its nearest neighbor CD8 + Descriptive statistics of the distance; and CD8 + Average #PD-L1 within 10 and 30 μm of cells + / panCK + The complete list of calculated features is given in Table 9:

[0234] Table 9

[0235]

[0236]

[0237]

[0238]

[0239]

[0240]

[0241]

[0242] Features were subjected to ReliefF feature selection to determine the importance of each feature in classifying cases according to anti-PD-1 treatment outcome, and then the 10 most important features were selected and fitted with a quadrant discriminant classification model to predict response to treatment. Treatment outcomes were classified into 3 phenotypes (PD = progressive disease; SD = stable disease; PR = partial response; CR = complete response); PD vs. SD vs. PR+CR; PD vs. SD+PR+CR; PD+SD vs. PR+CR. Most of the 54 samples did not have a clear tumor invasion front, so this feature was excluded from the analysis of the medial and lateral areas of the peritumoral area. Each configuration was analyzed using 1) all features, 2) only tumor features, and 3) only epithelial and stromal features. The purpose of this was to remove features that were highly associated with to explore the impact on the final classification. Due to the small sample size, no cross-validation was performed, and the reported classification results represent the classification accuracy on the staining set.

[0243] Table 10 summarizes the classification accuracy from different configurations and different feature sets.

[0244]

[0245] Table 10

[0246] The shaded cells show that for the third configuration (binary response), multiplex (Mpx) IHC data could achieve 89% accuracy, while mismatch repair (MMR) status could only achieve 70%. Figure 7 The importance ranking of all features in all cases is shown. The following features were identified as the most important:

[0247] (1) The fraction of PD-L1+ macrophages in the stroma,

[0248] (2) the fraction of PD-L1+ T helper cells in the stroma, and

[0249] (3) The fraction of PD-L1+ T cells in the stroma.

[0250] Mismatch repair (MMR) deficiency has been shown in the past to predict response to anti-PD-1 therapy. An analysis focusing only on cases with mismatch repair deficiency was performed to determine whether multiplex (Mpx) IHC data could identify which MMR-deficient cases responded to anti-PD-1 therapy. Figure 8 The importance ranking of features in the analysis configuration in the case of MMR deficiency is shown. Fig. 9 is an image of a stained sample showing a patient who had a complete response to treatment, and Fig.10are patients with progressive disease after treatment. Table 11 shows that using epithelial and stromal Mpx IHC data, a classification accuracy of 92% can be achieved to distinguish PD+SD results from PR+CR.

[0251]

[0252] Table 11

[0253] The confusion matrix for the classification is shown in Table 12. Of the 37 deficient cases, 34 were correctly identified as responders and non-responders by Mpx IHC data, and only 3 were misclassified.

[0254] In contrast, 53.7% of MMR-deficient cases responded to anti-PD-1 therapy.

[0255] answer SD+PD predicted by Mpx PR+CR predicted by Mpx Actual SD+PD(15) 13 2 Actual PR+CR(22) 1 21

[0256] Table 12

[0257] II. Exploring the spatial interaction of PD-1 / PD-L1 by automated multiplex IHC quantitative image analysis to predict the response of gastrointestinal tumors to immunotherapy

[0258] II.A. Background Art

[0259] Blockade of the PD-1 / L1 axis is an effective immunotherapy in some cancer patients. However, identifying predictive biomarkers for patient selection is a major challenge. Current clinical practice based on PD-L1 expression levels measured by IHC and the emerging biomarkers tumor mutational burden and mismatch repair (MMR) status is inadequate. The predictive value is limited by the variable strength of association across studies and tumor types. Recent studies have shown that the spatial arrangement and interactions between cancer cells and immune cells can affect patient prognosis, survival, and response to therapy. Multiplexed immunohistochemistry (IHC) tissue staining can provide a detailed profile of the tumor microenvironment based on specific tumor and immune molecular signatures.

[0260] II.B. Samples, Staining, and Image Acquisition

[0261] A cohort of 50 pre-treatment (anti-PD-1 pembrolizumab) patient gastrointestinal tumor samples with acceptable image and tissue quality for automated analysis were available for this study. Table 13 shows the attenuation of response.

[0262] Anti-PD-1 Response # of patients (all pre-treatment samples) PD (progressive disease) 15 SD (stable disease) 17 NE (Not Evaluable) 3 PR (Partial Response) 11 CR (Complete Response) 4 total 50

[0263] Table 13

[0264] Samples were formalin fixed, paraffin embedded, sectioned, and mounted on microscope slides.

[0265] An overview of staining and image analysis is provided in Fig.11 As shown in Table 14, slides were stained with fluorescent tyramide dye conjugates in multiplex IHC for PanCK, PD-L1, PD1, CD8, LAG3 on the BenchMark ULTRA IHC / ISH automated slide stainer in the tyramide signal amplification procedure:

[0266]

[0267] Table 14

[0268] The stains were applied sequentially as in Example 1. After deposition of each stain, a heat kill step was applied, which included the procedure described by Zhang (I). Serial sections of each sample were also stained with H&E using a VENTANA HE 600 automated slide stainer.

[0269] The entire slide was scanned with a Zeiss AXIO Z1 scanner, and the pathologist annotated the tumor area on it. HaloHi-Plex software was used for image analysis. MatLab Computer Vision, Image Processing, and Machine Learning Toolbox were used to (a) reconstruct the map of each cell type from the spatial position in the csv file output by Halo; (b) develop quantitative metrics to characterize the mutual influence between different cell signals; (c) rank and mine the most predictive feature combinations related to anti-PD-1 response; and (d) build and optimize the prediction model based on the selected features.

[0270] II.C. Image Annotation and ROI Generation

[0271] The pathologist annotates the tumor region ROI and the invasion front of the tumor (if applicable) in the image. As described in Example 1, the DPath system automatically annotates the epithelial tumor ROI from the aggregates of panCK+ cells and the stromal region.

[0272] II.C. Feature Calculation and Data Analysis

[0273] Each feature in Table 15 is analyzed for each image and ranked by ReliefF and Random Forest:

[0274]

[0275]

[0276]

[0277]

[0278]

[0279]

[0280] Table 15

[0281] The rankings of the top 15 features from each of ReliefF and Random Forest are Fig.12 and Fig.13 . Of the 190 features analyzed, Relieff and Random Forest ranked the following features in the top two in predicting response to treatment: “Maximum number of CD8+ / PD-1 low intensity cells within 20 μm of PD-L1+ cells in epithelial tumors”. Quadrant discriminant analysis (QDA) with 5-fold cross validation yielded 85% prediction accuracy. When combined with “mean # PD-1+ cells within 20 μm radius of PD-L1+ cells” and “maximum value of Lag3+ intensity on CD8+ cells”, it achieved 90.2% accuracy regardless of MMR status (see Figures 14A-14C and Tables 16 and 17).

[0282] Anti-PD-1 Response MMR is defect-free MMR deficiency PD+SD+NE 17 18 PR+CR 1 14

[0283] Table 16

[0284] Predictors MPX MMR MPX+MMR Accuracy 90.2％ 62％ 80％

[0285] Table 17

[0286] The accuracy of other features is lower (eg, 60%-70%).

[0287] Fig.14A –14C shows: (a) the spatial variance of PD-L1+ cells within a 10 μm radius # predicted value of PD-1+ cells; (b) PD-L1 + Average #PD-1 within a 20 μm radius of the cell 低强度 CD8 + Predictive value of cells; (c) CD8 + Lag3 + Predicted values ​​of the maximum Lag3 intensity in cells; and (d) a scatter plot showing (a) relative to (b).

[0288] Fig.15 Example IHC images of non-responders are shown, as well as images showing PD-L1 + Cell location (gray dots), PD-1 within 20 μm of PD-L1+ cells 低Cells (white dots), and PD-1 within 10 μm of PD-L1+ cells + Graphical reconstruction of cells (black dots). It can be seen that there is a similar presence of PD-L1+ cells in non-responders and responders. However, PD-L1 + Within 20 μm of the cell, responders showed CD8+PD-1 低 In addition, PD-1 within 10 μm of PD-L1+ cells was higher in responders compared with non-responders. + The cells are more evenly distributed.

[0289] II.D. Survival analysis

[0290] Overall survival (OS) data were available for a subset of 46 patients. Survival analysis was performed for each variable in Table 14. The following variables were significantly predictive of survival benefit: (a) Lag3 in panCK-negative regions + / CD8 + The number of cells and CD8 + (b) Lag3 in panCK-negative areas – / CD8 + Cells and CD8 + The ratio of the number of cells, (c) Lag3 + / panCK – The number of cells was divided by the panCK-negative area, (d) the number of Lag3-positive cells in the panCK-negative area, (e) the number of CD8 + The maximum intensity of Lag3 in cells, (f) the number of Lag3+ cells in panCK-positive areas, (g) PD-L1 + / panCK + PD-1 within a 10 μm radius of cells 中 / Average number of CD8+ cells, (h)PD-L1 + / panCK + PD-1 within a 20 μm radius of cells 中 / Average number of CD8+ cells, (i) PD-L1 + / CD8 + PD-1 within a 20 μm radius of cells 中 Variance of the number of CD8+ cells / (j)PD-L1 + / panCK + PD-1 within a 20 μm radius of cells 中 Variance of the number of CD8+ cells, (k) PD-L1 + PD-1 within a 10 μm radius of cells 低Variance of the number of CD8+ cells / (l)PD-L1 + / CD8 + PD-1 within a 20 μm radius of cells 中 / maximum number of CD8+ cells, and (m)PD-L1 + PD-1 within a 20 μm radius of cells 中 / Maximum number of CD8+ cells. For each characteristic indicator, the median of the characteristic metric distribution was used as the cutoff value to divide the cohort into two groups. Kaplan-Meier survival curves are Fig.16A – shown in 16M.

[0291] IV. Exemplary Image Analysis System and Clinical Workflow

[0292] In clinical practice, the scoring function can be integrated into prognostic analysis and treatment decision making. After a tumor biopsy or surgical resection, a representative tissue block showing a tumor cross-section from the patient's tumor sample is selected for analysis. At least three 4 μm thick sections are cut from this tissue block and transferred to glass slides. The sections are stained as:

[0293] 1. IHC negative control (i.e., staining with primary antibody diluent instead of primary antibody);

[0294] 2. Multiplex IHC (including at least PD-L1, PD1, CD8, and LAG3 primary antibodies);

[0295] and

[0296] 3.H&E.

[0297] All sections are scanned on a slide scanner. The images are transferred to the digital pathology system along with the slide metadata. The slide metadata includes the identification of the tumor sample and the staining of the slide (H&E, IHC, or negative control), which can be entered by the user when scanning the slide or automatically obtained from the laboratory information system. The digital pathology system uses the slide metadata to trigger the automatic calculation of one or more features of Table 9 or Table 15. For example, the feature includes at least the maximum number of CD8+ / PD-1 low intensity cells within 20μm of PD-L1+ cells in epithelial tumors, and optionally further includes "Average #PD-1+ cells within a 20μm radius of PD-L1+ cells" and "Maximum value of Lag3+ intensity on CD8+ cells".

[0298] In a digital pathology system, a pathologist or professional observer opens a digital image of an H&E slide in a viewing software to understand the relevant morphological areas to be scored. The user then annotates the tumor using the annotation tools provided by the viewing software. Typically, the tumor is defined by creating one or more contours and identifying them as the tumor contours. To do this, the user creates additional contours that intersect the tumor contours. The intersections define the beginning and end of the slice on the tumor contour involved in the invasive process. The new contours are identified as the invasive margins.

[0299] The user then triggers the automatic transfer of the annotations to an adjacent IHC slide. The digital pathology system provides a display function that transfers the annotations to an adjacent slide, taking into account the position, orientation, and local deformation of the tissue section. The user opens the IHC slide image in the viewer software and controls the position of the automatically displayed annotations. The viewer software provides tools for modifying and editing the annotations when necessary. Editing functions include moving the annotations, rotating the annotations, and locally modifying their outlines. The user further inspects the IHC slide image for tissue, staining, or imaging artifacts in the viewer software. The user delineates such artifact areas with annotations and identifies them to exclude from the analysis.

[0300] In the digital pathology system, users can select one or more IHC slides and trigger report generation. Users can obtain a quality control report that may include the following components:

[0301] 1. Slides show low to medium resolution images of all tissues on the slide

[0302] 2. The same low- to medium-resolution image is overlaid with outlines and / or transparent colored areas indicating morphological areas of interest, such as tumor margins. In addition, areas annotated to be excluded from the analysis are overlaid in this image.

[0303] 3. The same low- to medium-resolution image with automatically generated small rectangular markers indicating the location of the high-resolution FOV used for quality control

[0304] 4. Each high-resolution FOV

[0305] 5. Each high-resolution FOV is overlaid with markers that indicate the presence of each cell phenotype as determined by automated cell counting.

[0306] Alternatively, morphological areas of interest and markers indicating cells from the automated cell count can also be displayed in the viewer software.

[0307] The user reviews the QC data and decides to accept or reject the case. For accepted cases, the digital pathology system reports quantitative readouts and passes them to the scoring module. These quantitative readouts may include:

[0308] 1. The area of ​​each morphological region of interest (in mm) 2 express).

[0309] 2. The number of cells in the morphological area of ​​interest.

[0310] 3. Descriptive statistics describe the spatial distribution of cells of different phenotypes and / or the spatial relationship between cells in each morphological region of interest.

[0311] Additional information about the sample can be further input into the digital pathology system, for example, MMR status (before the subject is exposed to therapy (e.g., chemotherapy, radiotherapy, and / or targeted therapy)), tumor scores using the TNM staging system and / or overall tumor staging, and clinical variables (e.g., age, tumor directionality, number of lymph nodes harvested, and patient sex), and the system can use the above information to, for example, select an appropriate scoring function to apply to the image. Additionally or alternatively, the user can select an appropriate scoring function based on such criteria or other criteria. The scoring module calculates the score based on the extracted features, which can be reported as a raw number. In addition, an interval function can be applied to the score to assess the risk interval to which the patient belongs (e.g., by applying a cutoff value between groups based on "likely response" or "unlikely response" to checkpoint inhibitors) and / or a population stratification interval (e.g., based on a quartile or decile interval of the score); and / or a feature selection function to rank the score. The clinician reviews the report and discusses the results with the patient based on the results determined by the clinical pathologist, which can then be used to make treatment decisions for the patient.

[0312] References

[0313] Barrera et al., Computer-extracted features relating to spatial arrangement of tumor infiltrating lymphocytes to predict response to tonivolumab in non-small cell lung cancer (NSCLC). ASCO Annual Meeting 2018:Abstract#:12115.

[0314] Le et al., PD-1Blockade in Tumors with Mismatch-Repair Deficiency, NEngl JMed. 2015Jun 25; 372(26):2509-20 ("Le(I)").

[0315] Le et al., Mismatch-repair deficiency predicts response of solid tumors to PD-1 blockade, Science, 10.1126 / science.aan6733(2017) (“Le(II)”).

[0316] Li & Tian, ​​Development of small-molecule immune checkpoint inhibitors of PD-1 / PD-L1 as a new therapeutic strategy for tumour immunotherapy, J. of Drug Targeting, DOI: 10.1080 / 1061186X.2018.1440400 (online publication date February 20, 2018).

[0317] Nordic Immunohistochemical Quality Control, CK-Pan run 47 (2016), referenced at http: / / www.nordiqc.org / downloads / assessments / 82_85.pdf (last visited October 4, 2018) (“NordiQC”).

[0318] Topalian, Suzanne L. et al. "Safety, activity, and immune correlates ofanti–PD-1antibody in cancer." New England Journal of Medicine 366.26(2012):2443-2454.

[0319] Wang et al., Prediction of recurrence in early stage non-small cell lung cancer using computer extracted nuclear features from digital H&E images., Scientific Reports 7.1(2017):13543.

[0320] Woodcock-Mitchell et al. Immunolocalization of keratin polypeptides in human epidermis using monoclonal antibodies. J Cell Biol. 1982;95(2):580-588.

[0321] Yi et al., Biomarkers for predicting efficacy of PD-1 / PD-L1 inhibitors, Mol Cancer. 2018;17:129. (Online publication date August 23, 2018)

[0322] Zhang et al. "Automated 5-plex fluorescent immunohistochemistry with tyramide signal amplification using antibodies from the same species." J Immunother Cancer. 2015;3(Suppl 2):P111(“Zhang(I)”).

[0323] Zhang et al. "An automated 5-plex fluorescent immunohistochemistry enabled characterization of PD-L1 expression and tumor infiltrating immune cells in lung and bladder cancer specimens." Cancer Research 2016,76(14 Supplement):5117(“Zhang(II)”). Sequence Listing <110> Ventana Medical Systems, Inc. Johns Hopkins University Memorial Sloan Kettering Cancer Center <120> Methods and systems for predicting response to PD-1 axis-directed therapy <130> 3000022-004977 <150> US 62 / 739,828 <151> 2018-10-01 <150> US 62 / 742,934 <151> 2018-10-09 <160> 10 <170> PatentIn Version 3.5 <210> 1 <211> 182 <212> PRT <213> Homo sapiens <400> 1 Met Glu Gln Gly Lys Gly Leu Ala Val Leu Ile Leu Ala Ile Ile Leu 1 5 10 15 Leu Gln Gly Thr Leu Ala Gln Ser Ile Lys Gly Asn His Leu Val Lys 20 25 30 Val Tyr Asp Tyr Gln Glu Asp Gly Ser Val Leu Leu Thr Cys Asp Ala 35 40 45 Glu Ala Lys Asn Ile Thr Trp Phe Lys Asp Gly Lys Met Ile Gly Phe 50 55 60 Leu Thr Glu Asp Lys Lys Lys Trp Asn Leu Gly Ser Asn Ala Lys Asp 65 70 75 80 Pro Arg Gly Met Tyr Gln Cys Lys Gly Ser Gln Asn Lys Ser Lys Pro 85 90 95 Leu Gln Val Tyr Tyr Arg Met Cys Gln Asn Cys Ile Glu Leu Asn Ala 100 105 110 Ala Thr Ile Ser Gly Phe Leu Phe Ala Glu Ile Val Ser Ile Phe Val 115 120 125 Leu Ala Val Gly Val Tyr Phe Ile Ala Gly Gln Asp Gly Val Arg Gln 130 135 140 Ser Arg Ala Ser Asp Lys Gln Thr Leu Leu Pro Asn Asp Gln Leu Tyr 145 150 155 160 Gln Pro Leu Lys Asp Arg Glu Asp Asp Gln Tyr Ser His Leu Gln Gly 165 170 175 Asn Gln Leu Arg Arg Asn 180 <210> 2 <211> 171 <212> PRT <213> Homo sapiens <400> 2 Met Glu His Ser Thr Phe Leu Ser Gly Leu Val Leu Ala Thr Leu Leu 1 5 10 15 Ser Gln Val Ser Pro Phe Lys Ile Pro Ile Glu Glu Leu Glu Asp Arg 20 25 30 Val Phe Val Asn Cys Asn Thr Ser Ile Thr Trp Val Glu Gly Thr Val 35 40 45 Gly Thr Leu Leu Ser Asp Ile Thr Arg Leu Asp Leu Gly Lys Arg Ile 50 55 60 Leu Asp Pro Arg Gly Ile Tyr Arg Cys Asn Gly Thr Asp Ile Tyr Lys 65 70 75 80 Asp Lys Glu Ser Thr Val Gln Val His Tyr Arg Met Cys Gln Ser Cys 85 90 95 Val Glu Leu Asp Pro Ala Thr Val Ala Gly Ile Ile Val Thr Asp Val 100 105 110 Ile Ala Thr Leu Leu Leu Ala Leu Gly Val Phe Cys Phe Ala Gly His 115 120 125 Glu Thr Gly Arg Leu Ser Gly Ala Ala Asp Thr Gln Ala Leu Leu Arg 130 135 140 Asn Asp Gln Val Tyr Gln Pro Leu Arg Asp Arg Asp Asp Ala Gln Tyr 145 150 155 160 Ser His Leu Gly Gly Asn Trp Ala Arg Asn Lys 165 170 <210> 3 <211> 207 <212> PRT <213> Homo sapiens <400> 3 Met Gln Ser Gly Thr His Trp Arg Val Leu Gly Leu Cys Leu Leu Ser 1 5 10 15 Val Gly Val Trp Gly Gln Asp Gly Asn Glu Glu Met Gly Gly Ile Thr 20 25 30 Gln Thr Pro Tyr Lys Val Ser Ile Ser Gly Thr Thr Val Ile Leu Thr 35 40 45 Cys Pro Gln Tyr Pro Gly Ser Glu Ile Leu Trp Gln His Asn Asp Lys 50 55 60 Asn Ile Gly Gly Asp Glu Asp Asp Lys Asn Ile Gly Ser Asp Glu Asp 65 70 75 80 His Leu Ser Leu Lys Glu Phe Ser Glu Leu Glu Gln Ser Gly Tyr Tyr 85 90 95 Val Cys Tyr Pro Arg Gly Ser Lys Pro Glu Asp Ala Asn Phe Tyr Leu 100 105 110 Tyr Leu Arg Ala Arg Val Cys Glu Asn Cys Met Glu Met Asp Val Met 115 120 125 Ser Val Ala Thr Ile Val Ile Val Asp Ile Cys Ile Thr Gly Gly Leu 130 135 140 Leu Leu Leu Val Tyr Tyr Trp Ser Lys Asn Arg Lys Ala Lys Ala Lys 145 150 155 160 Pro Val Thr Arg Gly Ala Gly Ala Gly Gly Arg Gln Arg Gly Gln Asn 165 170 175 Lys Glu Arg Pro Pro Pro Val Pro Asn Pro Asp Tyr Glu Pro Ile Arg 180 185 190 Lys Gly Gln Arg Asp Leu Tyr Ser Gly Leu Asn Gln Arg Arg Ile 195 200 205 <210> 4 <211> 164 <212> PRT <213> Homo sapiens <400> 4 Met Lys Trp Lys Ala Leu Phe Thr Ala Ala Ile Leu Gln Ala Gln Leu 1 5 10 15 Pro Ile Thr Glu Ala Gln Ser Phe Gly Leu Leu Asp Pro Lys Leu Cys 20 25 30 Tyr Leu Leu Asp Gly Ile Leu Phe Ile Tyr Gly Val Ile Leu Thr Ala 35 40 45 Leu Phe Leu Arg Val Lys Phe Ser Arg Ser Ala Asp Ala Pro Ala Tyr 50 55 60 Gln Gln Gly Gln Asn Gln Leu Tyr Asn Glu Leu Asn Leu Gly Arg Arg 65 70 75 80 Glu Glu Tyr Asp Val Leu Asp Lys Arg Arg Gly Arg Asp Pro Glu Met 85 90 95 Gly Gly Lys Pro Gln Arg Arg Lys Asn Pro Gln Glu Gly Leu Tyr Asn 100 105 110 Glu Leu Gln Lys Asp Lys Met Ala Glu Ala Tyr Ser Glu Ile Gly Met 115 120 125 Lys Gly Glu Arg Arg Arg Gly Lys Gly His Asp Gly Leu Tyr Gln Gly 130 135 140 Leu Ser Thr Ala Thr Lys Asp Thr Tyr Asp Ala Leu His Met Gln Ala 145 150 155 160 Leu Pro Pro Arg <210> 5 <211> 235 <212> PRT <213> Homo sapiens <400> 5 Met Ala Leu Pro Val Thr Ala Leu Leu Leu Pro Leu Ala Leu Leu Leu 1 5 10 15 His Ala Ala Arg Pro Ser Gln Phe Arg Val Ser Pro Leu Asp Arg Thr 20 25 30 Trp Asn Leu Gly Glu Thr Val Glu Leu Lys Cys Gln Val Leu Leu Ser 35 40 45 Asn Pro Thr Ser Gly Cys Ser Trp Leu Phe Gln Pro Arg Gly Ala Ala 50 55 60 Ala Ser Pro Thr Phe Leu Leu Tyr Leu Ser Gln Asn Lys Pro Lys Ala 65 70 75 80 Ala Glu Gly Leu Asp Thr Gln Arg Phe Ser Gly Lys Arg Leu Gly Asp 85 90 95 Thr Phe Val Leu Thr Leu Ser Asp Phe Arg Arg Glu Asn Glu Gly Tyr 100 105 110 Tyr Phe Cys Ser Ala Leu Ser Asn Ser Ile Met Tyr Phe Ser His Phe 115 120 125 Val Pro Val Phe Leu Pro Ala Lys Pro Thr Thr Thr Pro Ala Pro Arg 130 135 140 Pro Pro Thr Pro Ala Pro Thr Ile Ala Ser Gln Pro Leu Ser Leu Arg 145 150 155 160 Pro Glu Ala Cys Arg Pro Ala Ala Gly Gly Ala Val His Thr Arg Gly 165 170 175 Leu Asp Phe Ala Cys Asp Ile Tyr Ile Trp Ala Pro Leu Ala Gly Thr 180 185 190 Cys Gly Val Leu Leu Leu Ser Leu Val Ile Thr Leu Tyr Cys Asn His 195 200 205 Arg Asn Arg Arg Arg Val Cys Lys Cys Pro Arg Pro Val Val Lys Ser 210 215 220 Gly Asp Lys Pro Ser Leu Ser Ala Arg Tyr Val 225 230 235 <210> 6 <211> 210 <212> PRT <213> Homo sapiens <400> 6 Met Arg Pro Arg Leu Trp Leu Leu Leu Ala Ala Gln Leu Thr Val Leu 1 5 10 15 His Gly Asn Ser Val Leu Gln Gln Thr Pro Ala Tyr Ile Lys Val Gln 20 25 30 Thr Asn Lys Met Val Met Leu Ser Cys Glu Ala Lys Ile Ser Leu Ser 35 40 45 Asn Met Arg Ile Tyr Trp Leu Arg Gln Arg Gln Ala Pro Ser Ser Asp 50 55 60 Ser His His Glu Phe Leu Ala Leu Trp Asp Ser Ala Lys Gly Thr Ile 65 70 75 80 His Gly Glu Glu Val Glu Gln Glu Lys Ile Ala Val Phe Arg Asp Ala 85 90 95 Ser Arg Phe Ile Leu Asn Leu Thr Ser Val Lys Pro Glu Asp Ser Gly 100 105 110 Ile Tyr Phe Cys Met Ile Val Gly Ser Pro Glu Leu Thr Phe Gly Lys 115 120 125 Gly Thr Gln Leu Ser Val Val Asp Phe Leu Pro Thr Thr Ala Gln Pro 130 135 140 Thr Lys Lys Ser Thr Leu Lys Lys Arg Val Cys Arg Leu Pro Arg Pro 145 150 155 160 Glu Thr Gln Lys Gly Pro Leu Cys Ser Pro Ile Thr Leu Gly Leu Leu 165 170 175 Val Ala Gly Val Leu Val Leu Leu Val Ser Leu Gly Val Ala Ile His 180 185 190 Leu Cys Cys Arg Arg Arg Arg Ala Arg Leu Arg Phe Met Lys Gln Phe 195 200 205 Tyr Lys 210 <210> 7 <211> 354 <212> PRT <213> Homo sapiens <400> 7 Met Arg Leu Ala Val Leu Phe Ser Gly Ala Leu Leu Gly Leu Leu Ala 1 5 10 15 Ala Gln Gly Thr Gly Asn Asp Cys Pro His Lys Lys Ser Ala Thr Leu 20 25 30 Leu Pro Ser Phe Thr Val Thr Pro Thr Val Thr Glu Ser Thr Gly Thr 35 40 45 Thr Ser His Arg Thr Thr Lys Ser His Lys Thr Thr Thr His Arg Thr 50 55 60 Thr Thr Thr Gly Thr Thr Ser His Gly Pro Thr Thr Ala Thr His Asn 65 70 75 80 Pro Thr Thr Thr Ser His Gly Asn Val Thr Val His Pro Thr Ser Asn 85 90 95 Ser Thr Ala Thr Ser Gln Gly Pro Ser Thr Ala Thr His Ser Pro Ala 100 105 110 Thr Thr Ser His Gly Asn Ala Thr Val His Pro Thr Ser Asn Ser Thr 115 120 125 Ala Thr Ser Pro Gly Phe Thr Ser Ser Ala His Pro Glu Pro Pro Pro 130 135 140 Pro Ser Pro Ser Pro Ser Pro Thr Ser Lys Glu Thr Ile Gly Asp Tyr 145 150 155 160 Thr Trp Thr Asn Gly Ser Gln Pro Cys Val His Leu Gln Ala Gln Ile 165 170 175 Gln Ile Arg Val Met Tyr Thr Thr Gln Gly Gly Gly Glu Ala Trp Gly 180 185 190 Ile Ser Val Leu Asn Pro Asn Lys Thr Lys Val Gln Gly Ser Cys Glu 195 200 205 Gly Ala His Pro His Leu Leu Leu Ser Phe Pro Tyr Gly His Leu Ser 210 215 220 Phe Gly Phe Met Gln Asp Leu Gln Gln Lys Val Val Tyr Leu Ser Tyr 225 230 235 240 Met Ala Val Glu Tyr Asn Val Ser Phe Pro His Ala Ala Gln Trp Thr 245 250 255 Phe Ser Ala Gln Asn Ala Ser Leu Arg Asp Leu Gln Ala Pro Leu Gly 260 265 270 Gln Ser Phe Ser Cys Ser Asn Ser Ser Ile Ile Leu Ser Pro Ala Val 275 280 285 His Leu Asp Leu Leu Ser Leu Arg Leu Gln Ala Ala Gln Leu Pro His 290 295 300 Thr Gly Val Phe Gly Gln Ser Phe Ser Cys Pro Ser Asp Arg Ser Ile 305 310 315 320 Leu Leu Pro Leu Ile Ile Gly Leu Ile Leu Leu Gly Leu Leu Ala Leu 325 330 335 Val Leu Ile Ala Phe Cys Ile Ile Arg Arg Arg Pro Ser Ala Tyr Gln 340 345 350 Ala Leu <210> 8 <211> 288 <212> PRT <213> Homo sapiens <400> 8 Met Gln Ile Pro Gln Ala Pro Trp Pro Val Val Trp Ala Val Leu Gln 1 5 10 15 Leu Gly Trp Arg Pro Gly Trp Phe Leu Asp Ser Pro Asp Arg Pro Trp 20 25 30 Asn Pro Pro Thr Phe Ser Pro Ala Leu Leu Val Val Thr Glu Gly Asp 35 40 45 Asn Ala Thr Phe Thr Cys Ser Phe Ser Asn Thr Ser Glu Ser Phe Val 50 55 60 Leu Asn Trp Tyr Arg Met Ser Pro Ser Asn Gln Thr Asp Lys Leu Ala 65 70 75 80 Ala Phe Pro Glu Asp Arg Ser Gln Pro Gly Gln Asp Cys Arg Phe Arg 85 90 95 Val Thr Gln Leu Pro Asn Gly Arg Asp Phe His Met Ser Val Val Arg 100 105 110 Ala Arg Arg Asn Asp Ser Gly Thr Tyr Leu Cys Gly Ala Ile Ser Leu 115 120 125 Ala Pro Lys Ala Gln Ile Lys Glu Ser Leu Arg Ala Glu Leu Arg Val 130 135 140 Thr Glu Arg Arg Ala Glu Val Pro Thr Ala His Pro Ser Pro Ser Pro 145 150 155 160 Arg Pro Ala Gly Gln Phe Gln Thr Leu Val Val Gly Val Val Gly Gly 165 170 175 Leu Leu Gly Ser Leu Val Leu Leu Val Trp Val Leu Ala Val Ile Cys 180 185 190 Ser Arg Ala Ala Arg Gly Thr Ile Gly Ala Arg Arg Thr Gly Gln Pro 195 200 205 Leu Lys Glu Asp Pro Ser Ala Val Pro Val Phe Ser Val Asp Tyr Gly 210 215 220 Glu Leu Asp Phe Gln Trp Arg Glu Lys Thr Pro Glu Pro Pro Val Pro 225 230 235 240 Cys Val Pro Glu Gln Thr Glu Tyr Ala Thr Ile Val Phe Pro Ser Gly 245 250 255 Met Gly Thr Ser Ser Pro Ala Arg Arg Gly Ser Ala Asp Gly Pro Arg 260 265 270 Ser Ala Gln Pro Leu Arg Pro Glu Asp Gly His Cys Ser Trp Pro Leu 275 280 285 <210> 9 <211> 290 <212> PRT <213> Homo sapiens <400> 9 Met Arg Ile Phe Ala Val Phe Ile Phe Met Thr Tyr Trp His Leu Leu 1 5 10 15 Asn Ala Phe Thr Val Thr Val Pro Lys Asp Leu Tyr Val Val Glu Tyr 20 25 30 Gly Ser Asn Met Thr Ile Glu Cys Lys Phe Pro Val Glu Lys Gln Leu 35 40 45 Asp Leu Ala Ala Leu Ile Val Tyr Trp Glu Met Glu Asp Lys Asn Ile 50 55 60 Ile Gln Phe Val His Gly Glu Glu Asp Leu Lys Val Gln His Ser Ser 65 70 75 80 Tyr Arg Gln Arg Ala Arg Leu Leu Lys Asp Gln Leu Ser Leu Gly Asn 85 90 95 Ala Ala Leu Gln Ile Thr Asp Val Lys Leu Gln Asp Ala Gly Val Tyr 100 105 110 Arg Cys Met Ile Ser Tyr Gly Gly Ala Asp Tyr Lys Arg Ile Thr Val 115 120 125 Lys Val Asn Ala Pro Tyr Asn Lys Ile Asn Gln Arg Ile Leu Val Val 130 135 140 Asp Pro Val Thr Ser Glu His Glu Leu Thr Cys Gln Ala Glu Gly Tyr 145 150 155 160 Pro Lys Ala Glu Val Ile Trp Thr Ser Ser Asp His Gln Val Leu Ser 165 170 175 Gly Lys Thr Thr Thr Thr Asn Ser Lys Arg Glu Glu Lys Leu Phe Asn 180 185 190 Val Thr Ser Thr Leu Arg Ile Asn Thr Thr Thr Asn Glu Ile Phe Tyr 195 200 205 Cys Thr Phe Arg Arg Leu Asp Pro Glu Glu Asn His Thr Ala Glu Leu 210 215 220 Val Ile Pro Glu Leu Pro Leu Ala His Pro Pro Asn Glu Arg Thr His 225 230 235 240 Leu Val Ile Leu Gly Ala Ile Leu Leu Cys Leu Gly Val Ala Leu Thr 245 250 255 Phe Ile Phe Arg Leu Arg Lys Gly Arg Met Met Asp Val Lys Lys Cys 260 265 270 Gly Ile Gln Asp Thr Asn Ser Lys Lys Gln Ser Asp Thr His Leu Glu 275 280 285 Glu Thr 290 <210> 10 <211> 525 <212> PRT <213> Homo sapiens <400> 10 Met Trp Glu Ala Gln Phe Leu Gly Leu Leu Phe Leu Gln Pro Leu Trp 1 5 10 15 Val Ala Pro Val Lys Pro Leu Gln Pro Gly Ala Glu Val Pro Val Val 20 25 30 Trp Ala Gln Glu Gly Ala Pro Ala Gln Leu Pro Cys Ser Pro Thr Ile 35 40 45 Pro Leu Gln Asp Leu Ser Leu Leu Arg Arg Ala Gly Val Thr Trp Gln 50 55 60 His Gln Pro Asp Ser Gly Pro Pro Ala Ala Ala Pro Gly His Pro Leu 65 70 75 80 Ala Pro Gly Pro His Pro Ala Ala Pro Ser Ser Trp Gly Pro Arg Pro 85 90 95 Arg Arg Tyr Thr Val Leu Ser Val Gly Pro Gly Gly Leu Arg Ser Gly 100 105 110 Arg Leu Pro Leu Gln Pro Arg Val Gln Leu Asp Glu Arg Gly Arg Gln 115 120 125 Arg Gly Asp Phe Ser Leu Trp Leu Arg Pro Ala Arg Arg Ala Asp Ala 130 135 140 Gly Glu Tyr Arg Ala Ala Val His Leu Arg Asp Arg Ala Leu Ser Cys 145 150 155 160 Arg Leu Arg Leu Arg Leu Gly Gln Ala Ser Met Thr Ala Ser Pro Pro 165 170 175 Gly Ser Leu Arg Ala Ser Asp Trp Val Ile Leu Asn Cys Ser Phe Ser 180 185 190 Arg Pro Asp Arg Pro Ala Ser Val His Trp Phe Arg Asn Arg Gly Gln 195 200 205 Gly Arg Val Pro Val Arg Glu Ser Pro His His His Leu Ala Glu Ser 210 215 220 Phe Leu Phe Leu Pro Gln Val Ser Pro Met Asp Ser Gly Pro Trp Gly 225 230 235 240 Cys Ile Leu Thr Tyr Arg Asp Gly Phe Asn Val Ser Ile Met Tyr Asn 245 250 255 Leu Thr Val Leu Gly Leu Glu Pro Pro Thr Pro Leu Thr Val Tyr Ala 260 265 270 Gly Ala Gly Ser Arg Val Gly Leu Pro Cys Arg Leu Pro Ala Gly Val 275 280 285 Gly Thr Arg Ser Phe Leu Thr Ala Lys Trp Thr Pro Pro Gly Gly Gly 290 295 300 Pro Asp Leu Leu Val Thr Gly Asp Asn Gly Asp Phe Thr Leu Arg Leu 305 310 315 320 Glu Asp Val Ser Gln Ala Gln Ala Gly Thr Tyr Thr Cys His Ile His 325 330 335 Leu Gln Glu Gln Gln Leu Asn Ala Thr Val Thr Leu Ala Ile Ile Thr 340 345 350 Val Thr Pro Lys Ser Phe Gly Ser Pro Gly Ser Leu Gly Lys Leu Leu 355 360 365 Cys Glu Val Thr Pro Val Ser Gly Gln Glu Arg Phe Val Trp Ser Ser 370 375 380 Leu Asp Thr Pro Ser Gln Arg Ser Phe Ser Gly Pro Trp Leu Glu Ala 385 390 395 400 Gln Glu Ala Gln Leu Leu Ser Gln Pro Trp Gln Cys Gln Leu Tyr Gln 405 410 415 Gly Glu Arg Leu Leu Gly Ala Ala Val Tyr Phe Thr Glu Leu Ser Ser 420 425 430 Pro Gly Ala Gln Arg Ser Gly Arg Ala Pro Gly Ala Leu Pro Ala Gly 435 440 445 His Leu Leu Leu Phe Leu Ile Leu Gly Val Leu Ser Leu Leu Leu Leu 450 455 460 Val Thr Gly Ala Phe Gly Phe His Leu Trp Arg Arg Gln Trp Arg Pro 465 470 475 480 Arg Arg Phe Ser Ala Leu Glu Gln Gly Ile His Pro Pro Gln Ala Gln 485 490 495 Ser Lys Ile Glu Glu Leu Glu Gln Glu Pro Glu Pro Glu Pro Glu Pro 500 505 510 Glu Pro Glu Pro Glu Pro Glu Pro Glu Pro Glu Gln Leu 515 520 525

Claims

1. A method for scoring a tumor sample for the likelihood of responding to a PD-1 axis-directed therapy, the method comprising: (a) obtaining a digital image of a tumor section from the tumor sample, wherein the tumor section is stained with multiplex affinity histochemical staining for each of PD-L1, CD8, CD3, CD68, and an epithelial marker (EM); (b) extracting one or more feature metrics from a region of interest (ROI) in the image, the one or more feature metrics selected from the group consisting of: (xx) PD-L1 in the ROI + / CD3 + The number of cells and CD3 + The ratio between the total number of cells, wherein the ROI is the interstitial ROI, (xxii) PD-L1 in the ROI + / CD8 + The number of cells and CD8 + The ratio between the total number of cells, wherein the ROI is the interstitial ROI, (xxiii) PD-L1 in the ROI + / CD68 + The number of cells and CD68 in the ROI + The ratio between the total number of cells, where the ROI is the interstitial ROI, and (xxiv) PD-L1 in the ROI + / CD3 + / CD8 – The number of cells and CD3 + / CD8 – The ratio between the total number of cells and the total number of cells, wherein the ROI is an interstitial ROI; and (c) applying a scoring function to a feature vector comprising the one or more feature metrics to generate a score indicative of a likelihood that the tumor will respond to the PD-1 axis-directed therapy, wherein the scoring function is obtained from a cohort of patients prior to PD-1 axis-directed therapy treatment for which outcome data is available.

2. The method according to claim 1, wherein the feature vector comprises each of the following: (xxiii) PD-L1 in stromal ROI + / CD68 + The number of cells and CD68 in the interstitial ROI + The ratio between the total number of cells, (xxiv) PD-L1 in stromal ROI + / CD3 + / CD8 – The number of cells and CD3 + / CD8 – The ratio between the total number of cells, and (xx) PD-L1 in stromal ROI + / CD3 + The number of cells and CD3 + The ratio between the total number of cells.

3. The method of claim 1, wherein the feature vector further comprises one or more feature metrics selected from the group consisting of: (i) CD8 in the ROI + Cells with recent PD-L1 + / CD68 + The average distance between cells, (ii) CD8 in the ROI + Cells with recent PD-L1 + / CD3 + The average distance between cells, (iii) Epithelial cells in the ROI and the nearest CD8 + The average distance between cells, (iv) CD8 + PD-L1 within 10 μm of cells + The number of epithelial cells, (v) CD8 + PD-L1 within 30 μm of cells + The number of epithelial cells (vi) CD8 within 10 μm of the epithelial cells in the ROI + (vii) Number of CD8 cells within 30 μm of epithelial cells in the ROI + (viii) The number of cells with PD-L1 in the ROI + / CD3 + Cell density (ix) PD-L1 in the ROI + / CD3 + / CD8 – Cell density (x) PD-L1 in the ROI + / CD8 + Cell density (xi) PD-L1 in the ROI + / CD68 + Cell density (xii) PD-L1 in the ROI + The density of epithelial cells (xiii) CD3 in the ROI + Cell density (xiv) CD8 in the ROI + Cell density (xv) CD68 in the ROI + Cell density (xvi) The density of epithelial cells in the ROI (xvii) PD-L1 in the ROI + The ratio between the area occupied by epithelial cells and the total area of ​​the ROI (xviii) The ratio of the area occupied by epithelial cells in the ROI to the total area of ​​the ROI (xix) PD-L1 in the ROI + The ratio between the number of epithelial cells and the total number of epithelial cells in the ROI (xx) PD-L1 in the ROI + / CD3 + The number of cells and CD3 + The ratio of the total number of cells (xxi) PD-L1 in the ROI + / CD3 + / CD8 – The number of cells and CD3 + / CD8 – The ratio of the total number of cells (xxii) PD-L1 in the ROI + / CD8 + The number of cells and CD8 + The ratio of the total number of cells (xxiii) PD-L1 in the ROI + / CD68 + The number of cells and CD68 in the ROI + The ratio of the total number of cells (xxiv) PD-L1 in the ROI + / CD3 + / CD8 – The number of cells and CD3 + / CD8 – The ratio between the total number of cells, (xxv) CD3 in the ROI + / CD8 – The number of cells and CD3 + The ratio between the total number of cells, and (xxvi)CD3 + The total area occupied by cells.

4. The method of claim 3, wherein the ROI is selected from the group consisting of: tumor ROI, stromal ROI, epithelial marker positive (EM + )ROI, epithelial marker negative (EM – )ROI, peritumoral inner (PI)ROI, peritumoral outer (PO)ROI and peritumoral region (PR)ROI.

5. The method according to claim 4, wherein the ROI is the EM – ROI, and the feature vector comprises one or more feature metrics selected from the group consisting of: (xx) PD-L1 in the ROI + / CD3 + The number of cells and CD3 + The ratio between the total number of cells, (xxii) PD-L1 in the ROI + / CD8 + The number of cells and CD8 + The ratio between the total number of cells, (xxiii) PD-L1 in the ROI + / CD68 + The number of cells and CD68 in the ROI + The ratio between the total number of cells, and (xxiv) PD-L1 in the ROI + / CD3 + / CD8 – The number of cells and CD3 + / CD8 – The ratio between the total number of cells.

6. The method according to claim 5, wherein the feature vector includes each of the following: – PD-L1 in ROI + / CD68 + The number of cells and the EM – CD68 in ROI + The ratio between the total number of cells, The EM – PD-L1 in ROI + / CD3 + / CD8 – The number of cells and the EM – CD3 in ROI + / CD8 – The ratio between the total number of cells, and The EM – PD-L1 in ROI + / CD3 + The number of cells and the EM – CD3 in ROI + The ratio between the total number of cells.

7. The method according to claim 4, wherein the ROI is the tumor ROI or the EM + ROI, and the feature vector comprises one or more feature metrics selected from the group consisting of: (i) The EM + CD8 in ROI or tumor ROI + Cells with recent PD-L1 + / CD68 + The average distance between cells, (ix) PD-L1 in the ROI + / CD3 + / CD8 – density, (xx) PD-L1 in the ROI + / CD3 + The number of cells and CD3 + The ratio between the total number of cells, (xxii) the EM + PD-L1 in ROI or tumor ROI + / CD8 + The number of cells and the EM + CD8 in ROI or tumor ROI + The ratio between the total number of cells, (xxiii) PD-L1 in the ROI + / CD68 + The number of cells and CD68 in the ROI + The ratio between the total number of cells, (xxiv) PD-L1 in the ROI + / CD3 + / CD8 – The number of cells and CD3 + CD8 – The ratio between the total number of cells, and (xxv) EM + CD3 in ROI or tumor ROI + / CD8 – The number of cells and the EM + CD3 in ROI or tumor ROI + The ratio between the total number of cells.

8. The method according to claim 1, wherein the ROI is derived from a digital image of a morphologically stained section of the tumor sample, wherein the morphologically stained section and the tumor section stained by multiple affinity histochemistry are consecutive sections.

9. The method of claim 8, wherein the ROI is identified by a user in the digital image of the morphologically stained section and automatically registered in the digital image of the multiplex affinity histochemically stained tumor section.

10. The method according to any one of claims 1 to 9, wherein the scoring function is derived from a modeling function selected from the group consisting of: quadrant discriminant analysis (QDA), linear discriminant analysis (LDA), support vector machine (SVM) and artificial neural network (ANN).

11. The method of claim 10, wherein the scoring function is a QDA model fitted to a selected feature metric to predict response to treatment, and wherein the treatment outcomes used to fit the QDA model are combined together in a configuration selected from the group consisting of: progressive disease (PD), stable disease (SD) versus partial response (PR) + complete response (CR); Comparison of PD with SD+PR+CR; and Comparison of PD+SD and PR+CR.

12. Use of reagents specific for PD-L1, CD8, CD3, CD68 and epithelial markers (EM) to prepare a kit for use in a method of selecting a patient to receive PD-1 axis-directed therapy, the method comprising: Generating a score according to the method of any one of claims 1 to 11, comparing the score to a predetermined cutoff value, and Based on whether the score is above or below the predetermined cutoff value, patients are selected to receive the PD-1 axis-directed therapy or alternative therapy.

13. The use according to claim 12, wherein the PD-1 axis-directed therapy is a PD-1-specific monoclonal antibody or a PD-L1-specific monoclonal antibody.

14. The use according to claim 12, wherein the PD-1 axis-directed therapy is selected from the group consisting of: pembrolizumab, nivolumab, atezolizumab, avelumab, durvalumab, cemiplizumab, tislelizumab and LY3300054.

15. A system for predicting a patient's response to PD-1 axis-directed therapy, the system comprising: processor; and a memory coupled to the processor, the memory being used to store computer executable instructions which, when executed by the processor, cause the processor to perform operations including the method according to any one of claims 1 to 11.

16. The system of claim 15, wherein the system further comprises a scanner or microscope adapted to capture digital images of sections of the tumor sample and transmit the images to a computer device.

17. The system of claim 15, wherein the system further comprises an automated slide stainer programmed to perform histochemical staining on sections of the tumor sample.

18. The system of claim 17, wherein the system further comprises an automated hematoxylin and eosin stainer programmed to stain one or more consecutive sections of the sections stained by the automated slide stainer.

19. The system of any one of claims 15 to 18, wherein the system further comprises a laboratory information system (LIS) for tracking sample and image workflow and diagnostic information, the LIS comprising a central database configured to receive and store information related to the tumor sample, the information comprising at least one of: processing steps to be performed on the tumor sample, processing steps to be performed on digital images of slices of the tumor sample, a processing history of the tumor sample and digital images; and one or more clinical variables associated with the likelihood that the patient will respond to the therapy.

20. The system of claim 19, wherein the clinical variable is MMR or MSI status.

21. A non-transitory computer-readable storage medium for storing computer-executable instructions, the computer-executable instructions being executed by a processor to perform operations, the operations comprising the method according to any one of claims 1 to 11.

Images