Methods of predicting prognosis and treatment response

Abstract

The present invention provides a computer-implemented method for predicting a treatment response of a mammalian cancer patient to an anti-cancer immunotherapy, the method comprising: a) providing a plurality of features comprising gene expression levels of at least 8 of the following cancer promoting genes: PTGS2, CXCL1, CXCL2, CXCL5, CXCL6, IL1A, HAS2, IL11, IL1B, EREG, IL6, CXCL8 (IL8), MMP12, NRG1, VCAN, AREG previously measured in a sample obtained from a tumor of the mammalian cancer patient; b) calculating a score using the features provided in a), and c) predicting a treatment response of the mammalian cancer patient based on the score calculated in step b) Related methods and systems for predicting a treatment response are also provided.

Description

Technical Field

[0001] This invention relates to materials and methods for predicting the response of cancer patients to cancer treatment and overall survival, and particularly, but not exclusively, to patients undergoing immunotherapy. Background Technology

[0002] The concept that cancer-induced inflammation and inflammatory cells at the tumor site can support cancer progression is well-established (Coussens et al., 2013; Hanahan and Weinberg, 2011; Mantovani et al., 2008). Several cellular and molecular inflammatory mediators commonly found in clinically apparent tumors are known to have pro-tumorigenic effects and are associated with many characteristics of invasive and aggressive tumors in both preclinical models and cancer patients. These include ubiquitous tumor-infiltrating leukocytes such as macrophages, neutrophils, immature myeloid cells, or regulatory T cells, as well as molecules produced by these and other leukocytes, stromal cells, or directly by cancer cells. IL-6, IL-8, CCL2, CXCL1, or vascular endothelial growth factor (VEGF) are typical examples of soluble factors with pleiotropic effects that can promote cancer growth and spread (Coussens et al., 2013; Mantovani et al., 2008).

[0003] Inflammation at the tumor site can also exert an anti-cancer effect, partly by promoting immune recognition and elimination of cancer cells. Specifically, cytotoxic T cells (CTLs) have been identified as anti-tumor effectors in preclinical cancer models, and their intratumoral abundance has been associated with improved patient outcomes and response to cancer therapy (Binnewies et al., 2018; Fridman et al., 2012; Galon, 2006; Thorsson et al., 2018). Therefore, a more favorable prognosis is also associated with high intratumoral levels of CTL chemical inducers (such as CXCL9 or CXCL10) or cytokines that promote type I immunity, CTL differentiation, and effector function (such as IL-12 or type I and II interferons (IFN)) (Gajewski et al., 2013; Spranger and Gajewski, 2018; Vesely et al., 2011). The levels of these factors in the tumor microenvironment (TME) can vary depending on the overall systemic and / or local inflammatory state, and their combined effects contribute to the strength and extent of the antitumor immune response. In addition to traditional helper CD4+...+ and cytotoxic CD8 + Beyond T cells, further evidence suggests that tumor infiltration via other inflammatory cells also favors immune attack and is associated with a favorable prognosis. Natural killer (NK) cells, γδ T cells, intrinsic lymphoid cells, and Batf3-dependent conventional type I dendritic cells (cDC1) constitute some of the immune subsets typically associated with improved outcomes (Böttcher et al., 2018; Brozet al., 2014; Gentles et al., 2015; Mittal et al., 2017; Morvan and Lanier, 2016; Ruffell et al., 2014; Sánchez-Paulete et al., 2016; Spranger et al., 2015; 2017). This is true in both spontaneous and treatment-induced antitumor responses, such as those induced by the administration of immune checkpoint inhibitors, which are revolutionary in cancer treatment and have evoked beneficial responses (including long-term remissions) across multiple cancer types (Ribas and Wolchok, 2018). Particularly in immune checkpoint blockade (ICB) therapy, the abundance of selected immune cells or inflammatory mediators is positively or negatively correlated with the treatment response (Ayers et al., 2017; De Henau et al., 2016; Herbst et al., 2014; Roh et al., 2017; Rooney et al., 2015; Tumeh et al., 2014).

[0004] However, it remains unclear how antagonistic cancer-promoting or cancer-suppressing inflammatory TMEs are established during cancer development and progression. Furthermore, the signaling and pathways regulating the quality and quantity of different elements of inflammatory infiltration are not fully elucidated. Improving our understanding of these processes has significant clinical relevance, as it can enhance our ability to predict treatment responses and further identify attractive therapeutic targets to improve the efficacy of anticancer therapies.

[0005] The idea of ​​immune-based cancer control has been historically controversial, but it has gained undeniable support in the last decade due to the remarkable clinical success of cancer therapies utilizing the cancer-suppressive function of the immune system (Alfaro et al., 2016; Van Allen et al., 2015). In particular, so-called immune checkpoint blockade (ICB) therapies targeting the CTLA-4 and PD(L)-1 axes have transformed the landscape of cancer treatment (Andrea et al., 2012). Despite its widespread use in clinical practice, including as a first-line treatment for many cancer types, many patients derive little benefit from ICBs, and some experience severe side effects (Auslander et al., 2018). Therefore, there is an urgent unmet need for sensitive and accurate biomarkers to predict responses to ICBs and guide treatment selection, especially in the context of the potential benefit of alternative therapies (Ayers et al., 2017; Azizi et al., 2018).

[0006] Established and emerging immunobiomarkers used for ICB therapy fall into two categories, broadly evaluating T-cell activity or (neo)antigenicity of cancer cells. Tumors described as "hot" or "T-cell inflammatory" are those exhibiting T-cell activity (especially CD8+). + Tumors exhibiting signs of T-cell infiltration and / or elevated levels of inflammatory mediators associated with the differentiation and function of cytotoxic T-cell effectors are more likely to respond better to ICB. Similarly, measures of cancer cell antigenicity, such as tumor mutational burden (TMB) and neoantigen clonality, are also associated with ICB response. Importantly, T-cell inflammation and TMB are poorly correlated, but when combined, they exhibit combined predictive power (Cristescu et al., 2018). However, even when combined, these biomarkers lack sufficient accuracy, especially with high CD8+. + Numerous instances of T-cell-infiltrated tumors failed to respond to ICB, while conversely, cold and poorly antigenic tumors sometimes responded strongly. Furthermore, in some cancer types, such as clear cell renal cell carcinoma (ccRCC), a malignancy with one of the highest ICB response rates, currently available immunobiomarkers have not shown meaningful predictive utility (Bhattacharya et al., 2020; Bisawas et al., 2019).

[0007] A common feature of past and present tumor immune profiling studies, as well as of established and emerging immunotherapy biomarkers, is the focus on T-cell inflammation. However, inflammation is arguably characteristic of all solid tumors. Even in so-called “immune desert” tumors, which are extensively lacking infiltrate by immune cells, oncogene-driven inflammatory signaling is still typically upregulated in cancerous or stromal cells. However, in contrast to T-cell-mediated inflammation, the types of inflammatory responses most commonly found in clinically apparent tumors are associated with malignant cancer growth and treatment resistance. Therefore, tumors are not simply inflamed or non-inflammatory, but rather exhibit inflammatory profiles that differ in both quantity and quality, possessing either cancer-promoting or anti-inflammatory effects.

[0008] Multiple studies have highlighted the close link between oncogenic transformation and tumor progression and inflammation and immune escape (vanWeverwij Pellyk and de Visser, 2023; Spranger and Gajewski, 2018). In these studies, recent work has revealed the central role of the cyclooxygenase (COX)-2 / prostaglandin E2 (PGE2) pathway, which is upregulated in many cancers and involved in multiple aspects of malignant growth. As a major pro-tumorigenic inflammatory axis, it enables cancer to escape immune responses and drive resistance to immunotherapy (Zelenay et al., 2015; Bonavita et al., 2020; Pelly et al., 2021). Therefore, in preclinical models, anti-inflammatory drugs or antagonists that inhibit PGE2 synthesis or signaling enhance the potency of the combination of ICB and chemoimmunotherapy (Pelly et al., 2021; Bell et al., 2022). Furthermore, by querying large cancer patient datasets, the inventors' previous work demonstrated the prognostic and predictive power of mouse-derived inflammatory gene signatures, which integrate both COX-2-related aspects of tumorigenesis and T-cell activation. Therefore, the activity levels of the COX-2 / PGE2 pathway, upregulated in many cancers and involved in multiple aspects of malignant tumor growth, constitute candidates for the molecular stratification of tumor inflammatory phenotypes.

[0009] While the previously described cancer prediction and prognostic models have shown promise, there remains an unmet need for other human-centered models capable of predicting treatment response and / or survival in cancer patients. This invention seeks to address these needs and provide further related advantages. Summary of the Invention

[0010] The inventors conducted a systematic meta-analysis of thousands of patient samples and cancer cell lines to investigate the value of monitoring tumor-promoting inflammation as a prognostic and predictive immune-related biomarker. Using COX-2 expression as a key inferred indicator of intratumoral pro-tumorigenic inflammatory activity, the inventors defined extracellularly mediated human-derived pro-tumorigenic inflammatory (PTI) gene signatures that exhibit profound and consistent prognostic value across molecularly defined immune subtypes, pan-cancer, and individual malignancies. By querying genomic, transcriptomic, and proteomic data, the inventors revealed the effects of cellular origin, oncogenicity, and PTI on T-cell inflammation. Notably, PTI exhibited synergistic prognostic and predictive utility when combined with the strongest predictors of ICB response currently available. Furthermore, in clear cell renal cell carcinoma (ccRCC)—a cancer type in which established immune biomarkers have not shown predictive utility—PTI selectively predicted immunotherapy outcomes. Molecular profiling analysis of archived ccRCC samples and validation through analysis of two independent ICB-treated patient cohorts demonstrated that patient stratification based on pre-treatment PTI levels has the potential to guide immunotherapy decisions. PTI was identified as an independent prognostic and predictive immunobiomarker with the potential to guide treatment selection in clinical practice. Overall, pro-tumorigenic inflammation was identified as a key determinant of cancer patient outcomes, and the PTI signature was identified as a robust prognostic and predictive biomarker in response to immune checkpoint inhibitor therapy.

[0011] In a first aspect of the invention, a computer-implemented method is provided for predicting the therapeutic response to anticancer immunotherapy in a mammalian cancer patient, the method comprising: a) providing a plurality of features including gene expression levels of at least 8, at least 9, at least 10, at least 11, at least 12 or more (e.g., all) of the following cancer-promoting genes previously measured in a sample obtained from a tumor of a mammalian cancer patient: PTGS2, CXCL1, CXCL2, CXCL5, CXCL6, IL1A, HAS2, IL11, IL1B, EREG, IL6, CXCL8 (IL8), MMP12, NRG1, VCAN, AREG; b) calculating a score using the features provided in a); and c) predicting the therapeutic response of the mammalian cancer patient based on the score calculated in step b).

[0012] In some implementations, the features also include clone_TMB and CXCL9.

[0013] In some implementations, the features include all of the following gene expression levels: PTGS2, CXCL1, CXCL2, CXCL5, CXCL6, IL1A, HAS2, IL11, IL1B, EREG, IL6, CXCL8 (IL8), MMP12, NRG1, VCAN, and AREG.

[0014] In some implementations, the gene expression level is a normalized gene expression level.

[0015] In some implementations, the score is calculated by z-normalizing the individual gene expression values ​​of the feature and calculating the average of the normalized measurements.

[0016] In some implementation schemes, the score is calculated according to the following formula:

[0017] Rating = Where n p It is the number of gene expression characteristics. Genes are positively correlated, within the unit interval (i), and (e) represents gene expression value, expressed as log2 counts per million (CPM); optionally, where It is the average expression per million counts (reads per kilobase million) of positive gene log2 transformations.

[0018] In other implementations, the score is a probability score calculated using the following formula:

[0019] Probability Score

[0020] Where β0 is the intercept weight, β is the weight vector for each k variable, and x is the sample-related variable vector, wherein the variables include the multiple features.

[0021] In some implementations, the computer-implemented method also includes assessing tumor burden and / or neoantigen prevalence in cancer patients.

[0022] In some implementations, predicting treatment response in mammalian cancer patients includes comparing the calculated score from step b) with a reference median score from a sample group of responders and a reference median score from a sample group of non-responders, wherein the sample groups of responders and non-responders are i) cancer patients with the same type of cancer as the said cancer patient or ii) cancer patients with a tumor type unrelated to the said cancer patient.

[0023] In some implementations, the score calculated in step b) indicates that the cancer is predicted to be in response to anticancer immunotherapy, and the method further includes selecting mammalian cancer patients for anticancer immunotherapy.

[0024] In some implementations, the gene expression levels of at least eight cancer-promoting genes are obtained by measuring the gene expression levels of at least eight, at least nine, at least ten, at least eleven, at least twelve, or more (e.g., all) of the following cancer-promoting genes in samples obtained from tumors of mammalian cancer patients: PTGS2, CXCL1, CXCL2, CXCL5, CXCL6, IL1A, HAS2, IL11, IL1B, EREG, IL6, CXCL8 (IL8), MMP12, NRG1, VCAN, and AREG.

[0025] In a second aspect of the invention, a method is provided for predicting the therapeutic response to anticancer immunotherapy in mammalian cancer patients, the method comprising: a) measuring a plurality of characteristics, including measuring the expression of at least 8, at least 9, at least 10, at least 11, at least 12 or more (e.g., all) of the following cancer-promoting genes in a sample obtained from the patient's tumor: PTGS2, CXCL1, CXCL2, CXCL5, CXCL6, IL1A, HAS2, IL11, IL1B, EREG, IL6, CXCL8 (IL8), MMP12, NRG1, VCAN, AREG; b) determining a score using the characteristics measured in a); and c) predicting the patient's therapeutic response based on the score determined in step b).

[0026] In some implementations, the features also include clone_TMB and CXCL9.

[0027] In some implementations, determining a score to predict a subject's treatment response includes: inputting data representing the gene expression levels of at least eight cancer-promoting genes containing the subject into a machine learning classifier, wherein the machine learning classifier has been trained with a training dataset containing gene expression levels of at least the same cancer-promoting genes as those forming the subject's score, and wherein the training dataset is derived from at least two sample cohorts; one cohort containing multiple samples from multiple mammalian cancer patients (responders) known to respond to anticancer immunotherapy, and the other cohort containing multiple samples from mammalian cancer patients (non-responders) known not to respond to anticancer immunotherapy; and having the machine learning classifier classify the subject's score as belonging to the responder group or the non-responder group.

[0028] In some implementations, the training dataset includes gene expression levels of at least eight cancer-promoting genes from at least 50, at least 100, at least 150, at least 200, at least 250, and at least 300 samples from mammalian cancer patients (responders) known to respond to anticancer immunotherapy, and gene expression levels of at least eight cancer-promoting genes from at least 50, at least 100, at least 150, at least 200, at least 250, at least 300, at least 350, at least 400, at least 450, at least 500, and at least 550 samples from mammalian cancer patients (non-responders) known not to respond to anticancer immunotherapy.

[0029] In some implementations, the machine learning classifier has been trained with multiple scores from at least three sample groups, one group containing multiple samples from multiple mammalian cancer patients known to have fully responded to anticancer immunotherapy (full responders), one group containing multiple samples from multiple mammalian cancer patients known to have partially responded to anticancer immunotherapy (partial responders), and another group containing multiple samples from multiple mammalian cancer patients known to have not responded to anticancer immunotherapy (non-responders).

[0030] In some implementations, the machine learning classifier is a logistic regression model, an elastic network regression model, or a tree-based model, such as a random forest classifier.

[0031] In a third aspect of the invention, a method is provided for classifying a sample obtained from a mammalian cancer subject, the sample comprising nucleic acids and / or proteins, the method comprising: a) analyzing the sample to obtain gene expression levels of at least eight cancer-promoting genes selected from the following: PTGS2, CXCL1, CXCL2, CXCL5, CXCL6, IL1A, HAS2, IL11, IL1B, EREG, IL6, CXCL8 (IL8), MMP12, NRG1, VCAN, and AREG; and b) using the gene expression levels obtained in step a) to carry out the method of the first or second aspect of the invention.

[0032] In some implementations, the sample is derived from one or more cancer cells of the subject.

[0033] In some implementations, the sample obtained from the patient's tumor is a formalin-fixed-paraffin embedded (FFPE) sample.

[0034] In a fourth aspect of the invention, a method for treating cancer in a mammalian cancer patient is provided, the method comprising: a) performing a method according to any of the foregoing aspects of the invention to predict the therapeutic response of the mammalian cancer patient to anticancer immunotherapy; b) determining that the mammalian cancer patient is predicted to respond to the anticancer immunotherapy in step a); and c) administering a therapeutically effective amount of the anticancer immunotherapy to a subject in need.

[0035] In a fifth aspect of the invention, a method of treating cancer in a mammalian cancer patient is provided, comprising: a) performing the method of any of the first to fourth aspects of the invention; b) determining a score of the mammalian cancer patient indicating that the mammalian cancer patient is predicted to be responsive to anticancer immunotherapy; and c) administering immunotherapy to the mammalian cancer patient in need.

[0036] In some implementations, the method further includes analyzing one or more additional biomarkers of anticancer immunotherapy response derived from the mammalian cancer patient, thereby supplementing and / or corroborating the prediction of antitherapy response or non-response.

[0037] In a sixth aspect of the invention, a system for predicting the therapeutic response of a mammalian cancer patient to anticancer immunotherapy is provided, the system comprising: at least one processor; and at least one non-transitory computer-readable medium containing instructions that, when executed by the at least one processor, cause the at least one processor to perform the method of the first or second aspect of the invention.

[0038] In a seventh aspect of the invention, one or more computer-readable media are provided, comprising instructions that, when executed by one or more processors, cause the one or more processors to perform the steps of the method of the first or second aspect of the invention.

[0039] In an eighth aspect of the invention, a pharmaceutical composition is provided for treating a mammalian cancer patient, the pharmaceutical composition comprising anticancer immunotherapy, wherein the patient has been predicted to respond to the anticancer immunotherapy by any of the methods of the first to third aspects of the invention.

[0040] In some embodiments of the first to eighth aspects of the present invention, the mammalian cancer patient suffers from the following and / or the cancer is: kidney cancer, such as renal cell carcinoma or kidney renal cell carcinoma (KIRC); gastroesophageal tumor; esophageal cancer (ECA); melanoma; bladder cancer (BLCA); gastric cancer; lung cancer, such as lung adenocarcinoma (LUAD), lung squamous cell carcinoma (LUSC), or non-small cell lung cancer (NSCLC); pancreatic cancer (PAAD); liver cancer (LIHC); glioblastoma (GBM); adrenocortical carcinoma (ACC); cervical cancer (CESC); head and neck cancer, such as head and neck squamous cell carcinoma (HNSC); thyroid cancer (THCA); uterine cancer. Renal cell carcinoma (UCEC); tenosynovial giant cell tumor (TGCT); melanoma; BRCA; sarcoma and pan-carcinoma. In some embodiments, renal cell carcinoma is clear cell renal cell carcinoma (ccRCC). In some embodiments, the immune subtype of the cancer is one of the following: wound healing, IFNγ dominant, TGFβ dominant, inflammatory and / or lymphocyte depleted subtypes.

[0041] In some embodiments, the anticancer immunotherapy comprises IL-2 immunotherapy alone or in combination with antiVEGF therapy, or immune checkpoint blockade therapy. In some embodiments, immune checkpoint blockade therapy comprises programmed death-1 (PD-1) blockade, programmed death-ligand 1 (PD-L1) blockade, and / or cytotoxic T-lymphocyte-associated protein 4 (CTLA-4) blockade. In some embodiments, the immune checkpoint blockade therapy comprises treatment with nivolumab, pembrolizumab, atezolizumab, and / or ipilimumab. In some embodiments, the antiVEGF therapy is bevacizumab.

[0042] The present invention includes combinations of the aforementioned aspects and preferred features, except where such combinations are clearly not permitted or explicitly avoided. Attached Figure Description

[0043] The following description, with reference to the accompanying drawings, will illustrate some embodiments and experiments demonstrating the principles of the present invention, wherein:

[0044] Figure 1 Derivation of pan-cancer pro-tumorigenic inflammatory signatures associated with COX-2. (A) COX-2 mRNA (PTGS2) expression in 31 TCGA (Cancer Genome Atlas) datasets. (B) Bioinformatics workflow for identifying the genetic signatures of COX-2-centric PTIs. (C) COX-2 高 Protein-protein interaction networks of genes upregulated in pan-cancer tumors. (D) In ​​COX-2 高 Gene ontology biological processes significantly enriched in protein-protein interaction networks. (E) Kaplan-Meier (KM) plots and associated risk tables of pan-solid tumor TCGA patients stratified by quantile based on PTI characteristic scores.

[0045] Figure 2Inflammatory mediators that promote tumorigenesis are strong indicators of poor prognosis. (A) HR and 95% confidence intervals for continuous PTI and IFNγ features and their interaction terms from a multivariate CoxPH model of pan-cancer with varying cancer type and disease stage. (B) Risk rate simulations for patients with typical high and low IFNγ values ​​across a range of PTI values. Lines represent median risk rates. Darker shaded areas represent quartiles, and lighter shaded areas represent minimum and maximum risk rates. (C) HR and 95% confidence intervals from univariate Cox regression survival analysis of target biomarkers across six immune subtypes. (D) KM survival curves (left panel) for patients stratified into four groups based on PTI and IFNγ features in the “IFNγ-dominant” subtype. Asterisks indicate FDR values. Comparison of CD8 between groups. + Box plot of T-cell scores (right). Unless otherwise stated, all P-values ​​are indicated by an asterisk (...). (E) HR, 95% confidence intervals, and Wald test p-values ​​for multivariate CoxPH models of OS from ccRCC (n= 531), LIHC (n= 374), PAAD (n= 176), glioblastoma (GBM, n= 168), CESC (n= 306), adrenocortical carcinoma (ACC, n= 79), UCEC (n= 277), THCA (n= 513), and HNSC (n= 240).

[0046] Figure 3 MEK inhibition attenuates the secretion of tumorigenesis-promoting inflammatory mediators by cancer cells. (A) Pairwise Spearman correlation coefficients of PTI genes in non-hematologic, non-fibroblastic cancer cell lines from CCLE. Non-significant (P>0.05) relationships are marked with “X”. (B) PTI state coefficients from a linear mixed-effects model, where PTI state (low / high) is used to predict cell line drug sensitivity adjusted for cancer type. Drug name (black) and target (light / dark gray) are labeled on the “y” axis. All bars shown are from statistically significant associations (P<0.05). (C) Correlation between cell line sensitivity to refametinib and cancer-inherent PTI transcriptional programs in all cell lines (top) and NSCLC lines (bottom), based on the natural logarithm of the drug IC50 value. (D) Concentrations of secreted proteins in NCIH1339 (top) and NCIH3255 (bottom) cells at 6, 8 and 24 hours after treatment with 1 μM and 10 μM trametinib, as detected by Luminex.

[0047] Figure 4NFκB associates MAPK pathway mutations with cancer cell intrinsic PTI expression. (A) Box plot showing the association between cancer cell intrinsic PTI transcriptional programs and KRAS mutations in pan-cancer cell lines. (B) RELA gene copy number estimates (0 = diploid) for cancer cell lines with low, intermediate, and high cancer cell intrinsic PTI expression. (C) p65 / RELA pS536 levels from RPPA data of KRAS-mutant cancer cell lines stratified according to cancer intrinsic PTI status. (D) Table showing the top ten most significantly enriched motifs in the expanded list of cancer cell intrinsic PTI genes and the predicted transcription factor binders. (E) Sequence identifiers from multiple sequence alignments of NFKB1 and NFKB2 binding motifs from chromatin immunoprecipitation sequencing (ChIPseq) experiments, showing the common nucleotide sequences of the related motifs. (F) Odds ratios and 95% confidence intervals from multivariate linear regression models using mutation status to predict PTI scores in the pan-cancer TCGA dataset, after adjusting for cancer type and TMB. (G) Proportion of significant mutations in different oncogenic signaling pathways associated with PTI. Decomposition of significant MAPK pathway mutations associated with PTI in single cancers. (H) Comparison of intratumoral PTI characteristic scores between pan-cancer TCGA tumors with diploid copies of EGFR and RELA (WT) and pan-cancer TCGA tumors with chromosomal amplification. Wilcoxon test p-values ​​are shown.

[0048] Figure 5 Pro-tumorigenic inflammatory mediators originate from heterogeneous cell types and influence the transcriptional phenotype of tumor-infiltrating T cells. (A) Dimensionality reduction of single-cell transcriptomic data from primary HNSC tumors, colored according to (A) predominant cell type or (B) PTI feature module score. (C) PTI module score of a single cancer cell from primary HNSC. Patients defined as having high cancer cell-inherent PTI expression are shown in light gray, and patients with low cancer cell-inherent PTI expression are shown in dark gray. An asterisk indicates a corrected p-value from a nonparametric T-test comparing PTI expression. 高 (Light gray) and PTI 低 (Dark gray) group of cancer cells ( (D) Kruskal-Wallis test p-value < 0.001. Heatmap where circle size and color represent the mean expression level of individual PTI genes in various clusters of immune and stromal cells from the HNSC scRNAseq dataset. Gray squares highlight selected groups of co-expressed genes. (E) Mean proportion of immune cell populations inferred by CIBERSORT in “IFNγ-dominant” subtype tumors with high, intermediate, and low PTI levels. (F) Correlation between PTI feature scores and T cell unresponsiveness features, including contours to illustrate point density. Spearman correlation coefficients and p-values ​​are shown. (G) Individual CD8 from MEL and NSCLC tumors separated by PTI status. + CD8 ICB response profile score in T cells. Corrected p-values ​​from multiple nonparametric T-tests are shown. ).

[0049] Figure 6 PTI predicts a lack of benefit from checkpoint inhibitors and superiority over other resistance characteristics. (A) shows a pie chart of the proportion of samples (n=1032) from the pan-cancer immunotherapy metadata set for five cancer types. (B) shows a table of individual groups, sample numbers, and treatments in the pan-cancer immunotherapy metadata set. (C) shows the results based on PTI, TIS, IFNγ signaling, CD8... + Bar graph showing the proportion of responders (PR / CR) in patient groups stratified by median levels of T cell and CD274 expression. (D) Proportion of responders in four patient groups stratified simultaneously based on biomarkers of PTI and “hot” tumors, which from left to right include TIS, IFNγ signaling, CD8+, and IFNγ. + T cell and CD274 expression. Selected χ² values ​​are shown. 2 Test P-value ( (E) Table of OS and associated risks for patients in each PTI / TIS quadrant from the pan-cancer ICB dataset. Wald test p-values ​​are shown. (F) HR, 95% confidence intervals, and -log values ​​for resistance characteristics from a multivariate CoxPH model and randomly generated characteristics, using median cutoff values, with TIS status and dataset as covariates. 10P-values. This figure only shows those features associated with poor prognosis (HR>1). (G) Comparison of PTI scores among patients with PD or PR / CR after ICB in the CPI1000 dataset. P-values ​​refer to nonparametric T-tests. (H) Effect size and standard error of meta-analysis of PTI features and other putative biomarkers from seven independent ICB datasets from the CPI1000 cohort. For visualization purposes, the upper limit of the error bars is an odds ratio of 5. CYT = cytolysis score. (I) OS of CPI1000 patients stratified according to PTI features, with high and low levels of clonal TMB (top figure). P-values ​​for log-rank tests are shown.

[0050] Figure 7 Pro-tumorigenic inflammation is a putative predictive biomarker in clear cell renal cell carcinoma. (A) Heatmap of pairwise Spearman correlation coefficients between individual PTI genes in the internal Nanostring dataset. (B) Heatmap showing immune, matrix, and TME gene feature scores in samples sorted from left to right according to increasing PTI scores. The asterisks on the right side of the heatmap indicate the Spearman correlation p-value between the feature and the PTI feature. (C) PTI characteristics and CD8 + Correlation between T cell and IFNγ signaling gene signatures. Dots are colored according to PTI scores. Spearman Rho values ​​(R) and P values ​​are shown. (D) Heatmap of individual PTI gene expression levels in the Manchester Nanostring cohort for patients who can be used to evaluate prognosis. High and low PTI groups are indicated by top annotations (light gray = high, dark gray = low). (E) OS and associated risk table for Manchester cohort patients stratified by PTI. P values ​​refer to the log-rank test. (F) HR and 95% confidence intervals from univariate Cox regression models predicting OS for patients with selected gene signatures. Patients are stratified according to median cutoff values. (G) KM plot and associated risk table showing OS of Checkmate 009, 010, and 025 patients stratified by median PTI expression levels treated with nivolumab (left) or everolimus (right). (H) Comparison of PTI characteristic scores between PD and PR / CR patients treated with atezolizumab, atezolizumab plus bevacizumab, or sunitinib in the IMMotion150 trial. P-values ​​from nonparametric t-tests comparing PD and PR / CR patients are shown.

[0051] Figure 8COX-2 expression and PGE2 signaling are correlated with the expression of PTI mediators in multiple cell types in vitro. (A) Gene sequencing lists obtained from cancers with high and low PTGS2 levels, with normalized enrichment scores of GSEA using Hallmark gene features. (B) Log2FC values ​​and associated FDRs of PTI genes obtained from differential gene expression analysis of human cancer cell lines with high and low PTGS2 levels. (C) Mouse BMDM, human fibroblasts, and CD4 cells treated in vitro with synthetic PGE2. + Log2FC of the PTI gene was obtained from differential gene expression (DGE) analysis of T cells. Genes that were significantly upregulated (FDR < 0.05, log2FC > 1) were light gray, and genes that were downregulated were dark gray (FDR < 0.05, log2FC < -1).

[0052] Figure 9 PTI is a potent and independent prognostic indicator associated with carcinogenic transformation and progressive disease stages. (A) shows the PTI and CD8 levels in pan-cancer immune subtypes as described by Thorsson et al. (2018). + Box plots of scores for T-cell, T-cell inflammation (TIS), and IFNγ signaling characteristics. Results of nonparametric T-tests are shown. (B) OS of “IFNγ-dominant” patients stratified clockwise based on PTI, TIS and CD8 characteristics, TIL scores, and CD274 expression. Individual FDR values ​​from the Wald test are shown for all survival analyses. (C) HR, 95% confidence intervals, and Wald test p-values ​​for multivariate CoxPH models of OS in KIRK / ccRCC (n=531), LIHC (n=374), PAAD (n=176), glioblastoma (GBM, n=168), CESC (n=306), adrenocortical carcinoma (ACC, n=79), UCEC (n=277), THCA (n=513), and HNSC (n=240). (D) Mean PTI characteristic scores (±SEM) for samples from different stages of LUSC. (E) Correlation between disease stage and PTI score (mean ± SEM) in the TCGA datasets of LUSC (n=498), LUAD (n=509), PAAD (n=176), ccRCC (n=531), and TGCT (n=139).

[0053] Figure 10Intrinsic PTI expression in cancer cells is associated with MEK activity. (A) shows a heatmap of stratified cancer cell lines with high and low PTI from the Garnett et al. (2012) drug screening dataset, which contains microarray gene expression data for over 600 cell lines. These cell lines partially overlap with those in the CCLE dataset. (B) Ten PTI-positive cancer cell lines... 高 and ten PTI 低 Dose-response curves of NSCLC cell lines treated with refatinib. Shapes represent different cell lines. (C) Changes in confluence percentage over time relative to baseline for cell lines NCIH1339 (top) and NCIH3255 (bottom) treated with two concentrations of trametinib in 96-well plates. Lines represent time points for supernatant analysis. (D) Representative images from confluence readouts after treatment with 10 μM trametinib at 8, 24, and 72 hours. The confluence percentage of the entire well is shown in the lower right of each image. (E) Concentrations of four proteins in the supernatant of NCIH1339 lung cancer cells treated with 1 or 10 μM selumetinib at 6, 8, and 24 hours.

[0054] Figure 11 Cancer cell lines with high PTI mediator expression exhibit different copy number profiles and protein expression. (A) Copy number estimates of four different genes separated according to the intrinsic PTI status of cancer cells (0 = diploid). FDR values ​​from a nonparametric T-test are shown. (B) shows a volcano plot of differential expression of proteins and phospholipids as measured by RPPA, comparing PTI. 高 and PTI 低 KRAS mutant cell lines.

[0055] Figure 12 PTI mediators are derived from heterogeneous cell types. (A) Heatmap of Spearman correlations between PTI features and cell types inferred by the MCPcounter algorithm in the TCGA dataset. Significant (p < 0.05) correlations are indicated by asterisks. (B) t-SNE clustering of single cells from two independent scRNAseq patient datasets from NSCLC and MEL, with the top row colored by cell type and the bottom row colored by PTI module score. (C) and (D) Heatmaps showing the average expression of individual PTI genes in various clusters of immune and stromal cells in the NSCLC and MEL scRNAseq datasets, respectively, by size and color. Gray squares highlight co-expressed genes. (E) Scatter plot where the outline shows the PTI feature score versus CD8 in the “IFNγ-dominant” subtype of tumors. +Point density of correlations in gene set scores for T cell-related gene features (Singer et al., 2016; Guo et al., 2018). Spearman correlation coefficients (E) and correlation p-values ​​are shown.

[0056] Figure 13 PTI adds predictive and prognostic value to current biomarkers of ICB response. (A) AUC values ​​from a multivariate logistic regression model of a TIS-based prediction of patient ICB response (PD vs. PR / CR) model in 16 datasets. Asterisks indicate p-values ​​from nonparametric T-tests. (A) Gray lines represent the mean and upper and lower confidence intervals from 1000 iterations. (B) Heatmap of pairwise Spearman correlation coefficients for continuous immunobiomarkers, including PTI features. Shapes indicate the direction and strength of the correlation. SDI = Shannon diversity index. Boxes are placed around the four main clusters using hierarchical clustering. (C) Box plot showing the association between PTI scores and binary clinical and genomic biomarkers in the CPI1000 dataset. Wilcoxon test p-values ​​are shown. B1501 is the HLA type, and B44 and B62 refer to HLA supertypes. DDR = DNA damage response pathway mutation. RTK = receptor tyrosine kinase mutation. Max het = maximum HLA heterozygosity. HLA-LOH = loss of HLA locus heterozygosity. (D) Overall survival (OS) of patients with high, moderate, and low CPI1000 PTI (Mariathasan dataset n=331, Riaz n=38, VanAllen n=34, Hugon n=27, Snyder PLOS Med n=25, Snyder NEJM n=7). Asterisks indicate the p-value of the Wald test (…). (E) shows bar graphs of bias estimates from univariate and multivariate logistic regression models predicting patient response to ICB (PD v PR / CR) in the CPI1000 cohort. Green dots below the x-axis represent variables included in each model. Lines between green dots represent interaction terms. Statistical significance between nested models was determined based on likelihood ratio tests. (F) shows a KM plot of OS for patients in the CPI1000 cohort stratified by PTI status with high and low CXCL9 (top) or CD274 (bottom) expression. The log-rank test p-value is shown. (G) shows clonal TMB stratified by PTI characteristic status. 高 CXCL9 高 OS and log-rank P-value of patients with CPI1000.

[0057] Figure 14 Experimental workflow and clinical characteristics of the Manchester RCC cohort. (A) An example of the workflow applied to the internal ccRCC cohort is provided. (B) A table summarizing the clinical characteristics of patients, including total sample size and percentage. aPD(L)1 refers to treatment with either a PD1 or PDL1 blocking antibody. TKI = Tyrosine kinase inhibitor. Here, aVEGF refers to bevacizumab.

[0058] Figure 15 The receiver operating characteristic curves of tumor type-specific logistic regression model A (Table 3) show the false positive rate (x-axis) compared to the true positive rate (y-axis).

[0059] Figure 16 The training (x-axis) and testing (y-axis) AUC values ​​of an elastic network classification model trained with the PTI gene, clone_TMB, and CXCL9 on the CPI1000 dataset, where gene expression levels (A) are normalized according to tumor type (tumor type-specific dataset model) or (B) are normalized regardless of tumor type (model B). Points represent models trained with different values ​​of α hyperparameters between 0 and 1. (C) Training and testing AUC values ​​from a model trained to predict immunotherapy responses on the CPI1000 dataset, where only the PTI gene is used as a feature. The best model has an α hyperparameter of 0.2 and seven important features (AREG, CXCL1, CXCL2, EREG, HAS2, IL1A, and PTGS2). The value of α is shown next to the corresponding point (trained model).

[0060] Figure 17 A protocol for RNA extraction and gene expression profiling analysis of formalin-fixed paraffin-embedded (FFPE) surgically removed samples. Detailed Implementation

[0061] Some aspects and embodiments of the invention will now be discussed with reference to the accompanying drawings. Other aspects and embodiments will be apparent to those skilled in the art. All references herein are incorporated by way of citation.

[0062] In describing this invention, the following terminology will be used, and it is intended to be defined as follows.

[0063] The features and advantages of this disclosure will be more readily understood by those skilled in the art upon reading the following detailed description. It should be understood that certain features of this disclosure described in the context of individual embodiments for clarity may also be provided as single elements in combination. Conversely, for brevity, multiple features of this disclosure described in the context of individual embodiments may also be provided individually or in any sub-combination. It will be understood that embodiments relating to one broad aspect of the invention described below are equally applicable to each of the other broad aspects of the invention described above. It will also be understood that the following embodiments may be combined unless the context otherwise requires.

[0064] As used herein, “and / or” should be regarded as specifically disclosing each of the two indicated features or components, with or without the other. For example, “A and / or B” should be regarded as specifically disclosing (i) A, (ii) B, and (iii) each of A and B, as if each were listed separately herein.

[0065] sample

[0066] As used herein, “test sample” can be a cell or tissue sample (e.g., a biopsy), a biological fluid, or an extract (e.g., a protein or DNA extract obtained from the subject). In particular, the sample can be a tumor sample, such as a solid tumor, such as a gastroesophageal tumor, melanoma, bladder tumor, or kidney tumor. The sample can be a fresh sample obtained from the subject, or it can be a sample that has been processed and / or stored prior to determination (e.g., frozen, fixed, or subjected to one or more purification, enrichment, or extraction steps).

[0067] In some embodiments, the sample is a formalin-fixed paraffin-embedded (FFPE) sample. In some embodiments, the sample is a surgically removed sample. In some embodiments, the sample is stained with H&E (hematoxylin and eosin). In some preferred embodiments, the sample is a surgically removed FFPE sample. Preferably, the FFPE sample (e.g., a surgically removed FFPE sample) is stained with H&E.

[0068] Gene expression

[0069] The phrase "determining expression levels" refers to determining the expression level of the gene's expression product. Expression levels can be determined at the nucleic acid level and / or the protein level.

[0070] Determined gene expression levels can be considered to provide an “expression profile.” In this article, “expression profile” is understood to refer to a dataset relating to the expression levels of one or more relevant genes in an individual, presented in a form that allows comparison with comparable expression profiles (e.g., from individuals whose prognosis is already known) to help determine prognosis and select appropriate treatment for individual patients.

[0071] Determining gene expression levels may involve determining the presence or amount of mRNA in a cancer cell sample. Methods used for this practice are well known to those skilled in the art. Gene expression levels can be determined in cancer cell samples using any conventional method, such as nucleic acid microarrays or nucleic acid synthesis (e.g., quantitative PCR). For example, gene expression levels can be determined using a NanoStringnCounter analysis system (see, for example, US 7,473,767).

[0072] In some implementations, the gene expression levels of at least eight cancer-promoting genes are obtained by measuring the gene expression levels of at least eight cancer-promoting genes in samples obtained from tumors of mammalian cancer patients.

[0073] As an alternative or supplement, determining gene expression levels may involve identifying the levels of proteins expressed from the genes in a sample containing cancer cells obtained from an individual. Protein expression levels can be determined by any available means, including the use of immunoassays. For example, expression levels can be determined by immunohistochemistry (IHC), Western blotting, ELISA, immunoelectrophoresis, immunoprecipitation, flow cytometry, mass cytometry, and immunostaining. Using any of these methods, the relative expression levels of the proteins expressed from the genes listed in Table 1 can be determined.

[0074] In some implementations, the gene expression level of each of the said cancer-promoting genes is a normalized gene expression level.

[0075] Gene expression levels, as detailed herein, and the PTI scores derived therefrom, can be compared with the expression levels and corresponding scores of the same genes in cancers from patient groups with known survival time and / or treatment response. The patients being compared can be referred to as the “control group.” Therefore, determined gene expression levels and PTI scores can be compared with expression levels in a control group of an individual with cancer. Comparisons can be made with expression levels determined in cancer cells in the control group. Comparisons can be made with expression levels determined in cancer cell samples from the control group. The cancer in the control group can be the same type as the cancer in the individual. For example, if expression is determined in an individual with kidney cancer, the expression levels and PTI scores can be compared with the expression levels and PTI scores in cancer cells of a patient who also has kidney cancer.

[0076] Other factors may also match the control group with the tested individuals and the cancer. For example, the cancer stage may be the same, the cancer immune subtype (as determined by Thorsson et al. 2018) may be the same, and the subjects and control groups may be age-matched and / or sex-matched.

[0077] In addition, the control group may have been treated with the same form of surgery and / or the same anticancer treatment (e.g., chemotherapy).

[0078] Cloned TMB

[0079] Clonal tumor mutational burden (TMB) is a measure of the number of somatic mutations present in each tumor cell. Clonal TMB is advocated as a pan-cancer predictor of checkpoint blockade outcomes. Clonal TMB measurements are obtained from Litchfield et al., 2021.

[0080] PTI characteristics

[0081] Using bioinformatics workflows to discover PTI gene signatures Figure 1 B). TCGA (Cancer Genome Atlas) gene expression data in raw count form were downloaded from the SAGE synapse database (syn4311114) and clinical data (syn12026747). Raw counts were TMM-normalized using calcNormFactors in edgeR (3.24.3). Genes failing to reach 0.25 counts per million (CPM) in 10% of samples were removed from the analysis. Patients belonging to a single TCGA study were stratified into high and low PTGS2 subgroups using the upper and lower quartiles of expression levels normalized to log2. Non-solid tumor types such as LAML (acute myeloid leukemia) and DLBC (diffuse large B-cell lymphoma) were excluded from further analysis. A model matrix was designed to compare patients with high and low PTGS2, adjusted for cancer type. This was performed based on median PTGS2 levels using the top ten cancer types (LUAD, BLCA, LUSC, HNSC, PRAD, PAAD, ESCA, STAD, CESC, KICH). DGE analysis was performed using limma-trend (P < 0.05, FC > 1.2) (Ritchie et al., 2015). Fifty-four upregulated genes were filtered according to a list of ligand:receptor interactions (Ramilowski et al., 2015), and the resulting 16 genes were used as the final PTI features (Table 1).

[0082] Gene ID (as of January 31, 2023). Available at https: / / www.ncbi.nlm.nih.gov / gene. The nucleotides of each gene published with this NCBI Gene ID number as of January 31, 2023 are explicitly incorporated herein by reference.

[0083] PTI score

[0084] The terms “gene expression score,” “score,” and “PTI score” used in this article are used interchangeably. The terms “cancer-promoting gene,” “tumorigenic inflammatory gene,” and “PTI gene” used in this article are also used interchangeably.

[0085] Cancer-promoting genes or pro-tumorigenic inflammatory (PTI) genes as described in Table 1 were found to have tumor gene expression positively correlated with PTGS2 expression, tumor growth, and poor response to immunotherapy. The inventors found that, under the specific conditions evaluated, the scores for calculating the gene expression of these cancer-promoting genes demonstrated enhanced predictive power compared to previous predictive features (e.g., see...). Figure 6 F), the previously predicted features include the COX-IS ratio disclosed earlier, such as in WO 2019 / 243567. The term “score” as used herein, such as “PTI score”, is intended to have a broad meaning, encompassing not only the additive value of each individual gene expression value, but also any relationship between features, such as combining the gene expression values ​​of the cancer-promoting genes.

[0086] In some embodiments, at least eight cancer-promoting genes include HAS2 and / or EREG. In some embodiments, at least eight cancer-promoting genes include HAS2, EREG, CXCL8 (IL8), and / or VCAN. In some embodiments, at least eight cancer-promoting genes include HAS2, EREG, CXCL1, CXCL2, and / or AREG.

[0087] In some implementations, the gene expression of at least 8, at least 9, at least 10, at least 11, at least 12, at least 13, at least 14, at least 15, and / or all (i.e., all 16) cancer-promoting genes listed in Table 1 can be measured in samples obtained from the patient's tumor. The gene expression levels of at least 8 PTI genes listed in Table 1 can be used to calculate a PTI score based on gene expression in samples obtained from the patient's tumor. Based on the gene expression levels of at least 8 PTI genes, the PTI score can be used to predict the patient's treatment response and / or post-treatment prognosis.

[0088] PTI scores calculated using gene expression data from a reduced subset of the PTI gene set can have comparable or even stronger predictive power compared to PTI scores calculated using gene expression data from all 16 PTI genes. The optimal reduced PTI gene set may vary depending on the clinical context.

[0089] In some implementations, the PTI score (also known as the “score”) is calculated by z-normalizing the individual gene expression values ​​of at least eight PTI genes and then calculating the average of the normalized measurements.

[0090] In some implementations, the PTI score is calculated according to the following formula:

[0091] PTI Score =

[0092] Where n p It is the number of the aforementioned cancer-promoting genes. is a positively correlated gene, within the unit interval value of (i), and (e) represents the gene expression value, expressed as log2 per million counts (CPM).

[0093] In some implementation schemes, It is the average expression per million counts (per million kb readings (FPKM)) of positive gene log2 transformations.

[0094] In other implementations, the PTI score is calculated as a weighted sum of z-score normalized values ​​of gene expression values ​​for at least eight PTI genes, plus an intercept. The probability of response to immunotherapy can then be calculated using a sigmoid function according to the following equation, where β0 is the intercept and β... 1-n It refers to the coefficients of the at least 8 PTI genes.

[0095] PTI Score =

[0096] Pr (response to ICB) = 1 / 1 + e -PTI评分

[0097] The expression value can be represented in any of the following ways: RPKM (reads per kilobase million), FPKM (fragments per kilobase million), CPM (counts per million), and / or nanostring counts.

[0098] Stratification based on gene expression

[0099] In some embodiments, the present invention provides methods for classifying, predicting, or monitoring cancer in subjects. Specifically, one or more pattern recognition algorithms can be used to evaluate data obtained from gene expression analysis. Such analytical methods can be used to form predictive models that can be used to classify test data. For example, a convenient and particularly effective classification method employs multivariate statistical analysis modeling, first forming a model (“predictive mathematical model”) using data from samples from known subgroups (e.g., from subjects known to have specific cancer prognostic subgroups: high-risk and low-risk), and then classifying unknown samples (e.g., “test samples”) according to the subgroups.

[0100] Pattern recognition methods have been widely used to characterize many different types of problems, spanning linguistics, fingerprinting, chemistry, and psychology. In the case of the methods described in this paper, pattern recognition uses multivariate statistics (both parametric and nonparametric) to analyze data and thereby classify samples based on a series of observed measurements and predict the values ​​of some dependent variable. There are two main approaches.

[0101] The first approach to pattern recognition was called "unsupervised"; these simply reduced data complexity in a reasonable way and also produced display images that could be interpreted by the human eye. However, this type of approach may not be suitable for developing clinical assays that can be used to classify samples derived from objects independent of the initial sample set used to train the prediction algorithm.

[0102] The second approach to pattern recognition is called “supervised,” where a mathematical model is generated using a training set of samples with known categories or outcomes, and then evaluated using an independent validation dataset. Here, the “training set” of gene expression data is used to build a statistical model that correctly predicts the “subgroup” for each sample. This training set is then tested with independent data (called the “test set” or “validation set”) to determine the robustness of the computer-based model. These models are sometimes called “expert systems,” but can be based on a range of different mathematical procedures, such as support vector machines, decision trees, k-nearest neighbors, and naïve Bayes. Supervised methods can use datasets with dimensionality reduction (e.g., the first few principal components), but typically use unreduced data with all dimensions. In all cases, these methods allow for a quantitative description of the multivariate boundaries characterizing and separating each subtype according to its inherent gene expression profile. Confidence limits for any prediction can also be obtained, such as the probability level at which a sample is well-fitted. The robustness of the prediction model can also be checked using cross-validation by omitting selected samples from the analysis.

[0103] As used herein, the term "training set" refers to a portion of a dataset used to fit or train a model to predict or classify values ​​known in the training set but unknown in (future) test data. In some embodiments, the training set contains gene expression levels of at least eight cancer-promoting genes from at least 50, 100, 150, 200, 250, 300, 350, 400, 500, 600, 700, 800, or 900 samples from mammalian cancer patients. In some embodiments, the training set may be a dataset from the Cancer Genome Atlas (TCGA). In some embodiments, the training set may be the CPI1000 dataset (Litchfield et al., 2021), which comprises a collection of datasets from patients treated with ICBs (anti-PD(L)1, anti-CTLA4) spanning multiple cancer types with whole exome sequencing (WES) and RNA sequencing (RNAseq) data.

[0104] After stratifying the training samples according to subtypes, centroid-based prediction algorithms can be used to construct centroids based on the expression profiles of the gene sets described in Table 1.

[0105] "Translation" of the descriptor axes can be useful. Examples of such translations include normalization and mean-centering. "Normalization" can be used to remove differences between samples. Some common methods for calculating normalization factors include: (i) global normalization using all genes on a microarray or nanostring codeset; (ii) housekeeping gene normalization using housekeeping / invariant genes that are continuously expressed; and (iii) internal control normalization using a known amount of exogenous control genes added during hybridization.

[0106] In one implementation, the expression levels of each PTI cancer-promoting gene listed in Table 1 may be normalized, for example, relative to one or more control housekeeping genes. Exemplary housekeeping genes include ACTB (60), GAPDH (2597), and TBP (6908), with the number in parentheses after each gene name being the NCBI gene ID number for that gene; the nucleotide sequence of each gene disclosed with that NCBI gene ID number as of January 31, 2023, is expressly incorporated herein by reference. Those skilled in the art will understand that the methods disclosed herein are not limited to normalization relative to any particular housekeeping gene, and any suitable housekeeping gene known in the art may be used.

[0107] Many normalization methods are feasible, and they can generally be applied to any of several points in the analysis. In one implementation, the LOWESS method is used to normalize the microarray data; this is a globally locally weighted scatter plot smoothing normalization function. In another implementation, the qPCR and NanoString nCounter analysis data are normalized relative to the geometric mean of groups of multiple housekeeping genes. Furthermore, fold change analysis can be used for qPCR.

[0108] Mean centering can also be used to simplify the interpretation of data visualizations and calculations. Typically, for each descriptor, the mean of that descriptor across all samples is subtracted. In this way, the descriptor mean coincides with the origin, and all descriptors are "centralized" to zero. In unit variance scaling, the data can be scaled to an equal variance. Typically, the value of each descriptor is scaled by 1 / StDev, where "StDev" is the standard deviation of that descriptor across all samples. In another implementation, z-score normalization can be used to normalize each value in the dataset such that the z-score has a mean of 0 and a standard deviation of 1. In a sense, Pareto scaling lies between mean centering and unit variance scaling. In Pareto scaling, the value of each descriptor is scaled by l / sqrt(StDev), where "StDev" is the standard deviation of that descriptor across all samples, and "sqrt" represents the square root. In this way, each descriptor has a variance numerically equal to its initial standard deviation. Pareto scaling can be applied, for example, to raw data or mean-centered data.

[0109] Logarithmic scaling can be used to aid interpretation when data is positively skewed and / or when data spans a large range (e.g., several orders of magnitude). Typically, for each descriptor, the value is replaced with its logarithm. In equal range scaling, each descriptor is divided by its range for all samples. In this way, all descriptors have the same range, i.e., 1. However, this method is sensitive to the presence of outliers. In autoscaling, each data vector is mean-centered and unit-variance scaled. This technique is very useful because each descriptor is then equally weighted, and large and small values ​​are given equal importance. This can be important for genes expressed at very low but still detectable levels.

[0110] When comparing data from multiple analyses (e.g., comparing the expression profiles of one or more test samples with centroids constructed from samples collected and analyzed in independent studies), it is necessary to normalize the data from these datasets. In one implementation, these datasets are combined using distance-weighted discrimination (DWD) (Benito et al. (2004) Bioinformatics 20(1): 105-114, incorporated herein by reference in its entirety). DWD is a multivariate analysis tool that identifies systematic biases present in individual datasets and then makes global adjustments to compensate for these biases; essentially, each individual dataset is a multidimensional point cloud of data, and DWD takes two point clouds and moves one point cloud to better overlap with the other.

[0111] In some implementations described herein, Cox proportional hazards (CoxPH) model analysis can be used to assess the prognostic performance of gene expression scores (PTI scores). Cox proportional hazards model analysis is a regression method for survival data that provides estimates of hazard ratios and their confidence intervals. Cox models are a recognized statistical technique for exploring the relationship between patient survival and specific variables. This statistical method allows for the estimation of an individual's risk (i.e., hazard) given their prognostic variables, such as gene expression ratios as described herein. A "hazard ratio" is a measure of a patient's risk of death at any given time point for a specific prognostic variable. CoxPH regression analysis can be performed using PTI scores as input values ​​and disease stage and / or cancer type as layers.

[0112] The PTI score can be used as input to calculate a logistic regression model to predict patient treatment response outcomes. This logistic regression model can be calculated for both tumor type-independent and tumor type-specific data.

[0113] In some implementations, for both tumor type-independent and tumor type-specific approaches, the logistic regression model predicting a binary patient treatment response outcome is determined by: measuring the individual gene expression values ​​of at least eight PTI genes, adding a pseudo-count of 1 before normalizing each value to log2, and then z-score normalizing all samples in the dataset (“tumor type-independent”), or grouping samples according to cancer type (“Tum_Type”) and z-score normalizing according to cancer type (“tumor type-specific”), and then taking the average of scaled expression data for at least eight PTI genes in the “tumor type-independent” or “tumor type-specific” group to calculate a reference gene expression score. In some implementations, patients who achieved disease stabilization as the best RECIST response (metadata feature: Response_RECIST) are removed from the PTI score calculation. In some implementations, “under treatment” samples (metadata feature: Prior_IO) are removed from the PTI score calculation.

[0114] The PTI score has better predictive power than known resistance characteristics. The protective PTI score is better at predicting poor overall survival in patients treated with ICBs than known resistance characteristics (e.g., the early COX-IS ratio disclosed in WO 2019 / 243567). Figure 6 As illustrated in F and discussed in Example 11, pan-cancer survival analysis with PTI score (HR 1.38 [1.15-1.65], P=0.0004) was superior to pan-cancer survival analysis with COX-IS ratio (HR 1.18 [0.96-1.45], P=0.12).

[0115] Predicting treatment response

[0116] Individuals grouped into the good prognosis group can be identified as having cancer that is sensitive to immunotherapy, such as immune checkpoint blockade (ICB) therapy. They may also be referred to as responders, individuals who respond well to immunotherapy, or individuals who have a partial response (PR) or a complete response (CR) to immunotherapy, such as immune checkpoint blockade (ICB) therapy (collectively referred to as PR / CR).

[0117] Individuals grouped into the poor prognosis group can be identified as having cancer resistant to immunotherapy (such as immune checkpoint blockade (ICB) therapy). They can be referred to as nonresponders, individuals who do not respond well to immunotherapy, individuals who do not benefit from immunotherapy, or individuals who have disease progression (PD) after immunotherapy (such as immune checkpoint blockade (ICB) therapy).

[0118] When an individual is grouped into a good prognosis group, they may be treated with appropriate immunotherapy (e.g., immune checkpoint blockade therapy), as described in further detail below. When an individual is grouped into a poor prognosis group, the option to treat them with the aforementioned immunotherapy may be cancelled, and they may receive, for example, surgery, radiation therapy, and / or another form of anticancer agent (e.g., one or more non-immunochemotherapeutic agents or anti-angiogenic agents).

[0119] In some embodiments, a computer-implemented method for predicting the treatment response to anticancer immunotherapy in mammalian cancer patients includes: a) providing gene expression data comprising gene expression levels of at least 8, at least 9, at least 10, at least 11, at least 12 or more (e.g., all) of the following cancer-promoting genes previously measured in samples obtained from tumors of mammalian cancer patients: PTGS2, CXCL1, CXCL2, CXCL5, CXCL6, IL1A, HAS2, IL11, IL1B, EREG, IL6, CXCL8 (IL8), MMP12, NRG1, VCAN, AREG; b) calculating a gene expression score (PTI score) of the at least 8 or more cancer-promoting genes provided in a); and c) predicting the treatment response of the mammalian cancer patient based on the PTI score calculated in step b). In some embodiments, the PTI score is calculated based on the gene expression levels of all 16 cancer-promoting genes.

[0120] In some implementations, the computer-implemented method also includes assessing tumor burden and / or neoantigen prevalence in cancer patients.

[0121] In some implementations, predicting treatment response in mammalian cancer patients includes comparing a calculated PTI score with a reference median gene expression score from a sample cohort of responders and a reference median gene expression score from a sample cohort of non-responders, wherein the sample cohorts of responders and non-responders are i) cancer patients with the same type of cancer as the said cancer patient or ii) cancer patients with tumors of an unrelated type.

[0122] In some implementations, where the PTI score indicates that cancer is predicted in response to anticancer immunotherapy, mammalian cancer patients are selected for anticancer immunotherapy.

[0123] In some embodiments, a method for predicting the treatment response to anticancer immunotherapy in mammalian cancer patients includes: a) measuring the gene expression of at least eight cancer-promoting genes selected from the following in a sample obtained from the patient's tumor: PTGS2, CXCL1, CXCL2, CXCL5, CXCL6, IL1A, HAS2, IL11, IL1B, EREG, IL6, CXCL8 (IL8), MMP12, NRG1, VCAN, and AREG; b) determining a PTI score for the gene expression of the at least eight cancer-promoting genes based on the gene expression score measured in a); and c) predicting the patient's treatment response based on the gene expression levels determined in step b).

[0124] In some implementations, determining a PTI score to predict a subject's treatment response includes: inputting data on gene expression levels of at least eight cancer-promoting genes representing the subject into a machine learning classifier, wherein the machine learning classifier has been trained with a training dataset containing gene expression levels of at least the same cancer-promoting genes that form the subject's PTI score, and wherein the training dataset is derived from at least two sample cohorts; one cohort containing multiple samples from multiple mammalian cancer patients (responders) known to respond to anticancer immunotherapy, and the other cohort containing multiple samples from mammalian cancer patients (non-responders) known not to respond to anticancer immunotherapy; and having the machine learning classifier classify the subject's PTI score as belonging to the responder or non-responder group. In some implementations, the training dataset includes gene expression levels of at least eight cancer-promoting genes from at least 50, at least 100, at least 150, at least 200, at least 250, and at least 300 samples from mammalian cancer patients known to respond to anticancer immunotherapy (referred to as responders), and gene expression levels of at least eight cancer-promoting genes from at least 50, at least 100, at least 150, at least 200, at least 250, at least 300, at least 350, at least 400, at least 450, at least 500, and at least 550 samples from mammalian cancer patients known to not respond to anticancer immunotherapy (referred to as nonresponders).

[0125] In some implementations, the machine learning classifier has been trained with multiple PTI scores from at least three sample groups, one group containing multiple samples from multiple mammalian cancer patients known to have fully responded to anticancer immunotherapy (referred to as full responders), one group containing multiple samples from multiple mammalian cancer patients known to have partially responded to anticancer immunotherapy (referred to as partial responders), and another group containing multiple samples from multiple mammalian cancer patients known to have not responded to anticancer immunotherapy (referred to as nonresponders).

[0126] A method for classifying samples containing nucleic acids and / or proteins from mammalian cancer patients includes: analyzing the sample to obtain gene expression levels of at least eight cancer-promoting genes in the mammalian cancer patient sample; analyzing the gene expression levels of the at least eight cancer-promoting genes to determine a PTI score; and analyzing the sample PTI score to classify the sample into a PTI score that matches an anticancer immunotherapy response score or an anticancer immunotherapy non-response score, wherein if the PTI score is classified as matching an anticancer immunotherapy response score, the sample is characterized as originating from an anticancer immunotherapy responsive subject; and if the PTI score is classified as matching an anticancer immunotherapy non-response score, the sample is characterized as originating from an anticancer immunotherapy non-response subject. Optionally, the sample may be derived from one or more cancer cells from a mammalian cancer patient.

[0127] A system for predicting treatment response to anticancer immunotherapy in mammalian cancer patients is also provided. The system includes at least one processor and at least one non-transitory computer-readable medium containing instructions that, when executed by the at least one processor, cause the at least one processor to: predict treatment response to anticancer immunotherapy in mammalian cancer patients and / or classify samples obtained from mammalian cancer subjects as PTI scores of responders or non-responders to anticancer immunotherapy. In some embodiments, the invention includes one or more computer-readable media containing instructions that, when executed by one or more processors, cause the one or more processors to predict treatment response to anticancer immunotherapy in mammalian cancer patients and / or classify samples obtained from mammalian cancer subjects as PTI scores of responders or non-responders to anticancer immunotherapy.

[0128] Whether a prognosis is considered good or bad can vary depending on the cancer and the stage of the disease. Generally, a good prognosis is one where overall survival (OS) and / or progression-free survival (PFS) is longer than the mean for that stage and cancer type. If PFS and / or OS are lower than the mean for that stage and cancer type, the prognosis is considered poor. The mean can be median survival OS or PFS. The term “overall survival” (OS) as used herein can be interpreted as the length of time a patient diagnosed with a disease (e.g., cancer) remains alive from the date of diagnosis or the start of treatment. The term “progression-free survival” (PFS) as used herein can be understood as the length of time a patient lives without progression of the disease (e.g., cancer) during and after treatment.

[0129] In some implementations, the machine learning classifier is a logistic regression model.

[0130] For example, a good prognosis can be considered if PFS > 6 months and / or OS > 18 months. Similarly, a poor prognosis can be considered if PFS < 6 months or OS < 18 months. In particular, for advanced cancer, a good prognosis can be considered if PFS > 6 months and / or OS > 18 months.

[0131] Generally speaking, a “good prognosis” is an individual patient’s survival (OS and / or PFS) that is favorable compared to the expected outcome in a comparable disease setting (e.g., similar stage and cancer type) of the patient population. This can be defined as being better than the median survival (i.e., surviving more than 50% of the patients in the population).

[0132] Generally speaking, a “poor prognosis” is an individual patient’s survival (OS and / or PFS) that is unfavorable compared to the expected outcome in a patient population with a comparable disease background (e.g., similar stage and cancer type). This can be defined as inferior to median survival (i.e., survival not exceeding 50% of the patients in the population).

[0133] "Predicting a cancer patient's response to a chosen treatment" aims to assess the likelihood that a patient will experience a positive or negative outcome under a given treatment.

[0134] As used in this article, “indicating positive treatment outcome” means an increased likelihood that a patient will obtain a beneficial outcome from the chosen treatment (e.g., tumor size reduction, “good” prognostic outcome, improvement in disease-related symptoms and / or quality of life).

[0135] "Indicative negative treatment outcome" means that the patient is more likely to not receive the benefits of the aforementioned positive treatment outcomes (e.g., no reduction in tumor size, "poor" prognostic outcome, no improvement in disease-related symptoms and / or quality of life).

[0136] In some implementations, the PTI score of a cancer patient may be used as a reference or compared with the reference median gene expression score derived from a sample group of cancer patients with the same type of cancer as the said cancer patient (and optionally age-matched, time-from-diagnosis-matched, and / or disease-stage-matched). This may be used to predict the treatment response and / or prognosis of the said cancer patient (“tumor type-specific” prediction).

[0137] In other implementations, the PTI score of a cancer patient can be compared with a reference median gene expression score from a sample group of cancer patients with the same and different types of cancer as the patient (“tumor type-independent” prediction) (and optionally age-matched, time-matched from diagnosis, and / or disease stage-matched).

[0138] treat

[0139] Methods for treating cancer in mammalian cancer patients may include: predicting the therapeutic response of the mammalian cancer patient to anticancer immunotherapy according to the method of the invention, determining that the mammalian cancer patient is predicted to respond to the anticancer immunotherapy, and administering a therapeutically effective amount of anticancer immunotherapy to a patient in need. Some methods for treating mammalian cancer patients may also include analyzing one or more additional biomarkers of anticancer immunotherapy response derived from the mammalian cancer patient to supplement and / or corroborate the prediction of anticancer response or non-response.

[0140] A pharmaceutical composition is provided comprising an anticancer immunotherapy for treating mammalian cancer patients, wherein the patients have been predicted to respond to the anticancer immunotherapy.

[0141] The mammalian cancer patient may have, or the cancer may be, any of the following cancer types: renal cell carcinoma, such as renal cell carcinoma; gastroesophageal tumors; melanoma; bladder cancer; gastric cancer; lung cancer, such as lung adenocarcinoma (LUAD); pancreatic cancer (PAAD); liver cancer (LIHC); cervical cancer (CESC); head and neck cancer (HNSC); thyroid cancer (THCA); uterine cancer (UCEC); and giant cell tenosynovitis (TGCT). The renal cell carcinoma may be clear cell renal cell carcinoma (ccRCC). The mammalian cancer may have, or the cancer may be, any of the following immune subtypes as defined by Thorsson et al. (2018): wound-healing, IFNγ-dominant, TGFβ-dominant, inflammatory and / or lymphocyte-depleted subtypes.

[0142] The anticancer immunotherapy may comprise IL-2 immunotherapy alone or in combination with antiVEGF therapy, or immune checkpoint blockade therapy. Optionally, the antiVEGF therapy may be bevacizumab. The immune checkpoint blockade therapy may comprise any of the following: programmed death-1 (PD-1) blockade, programmed death-ligand 1 (PD-L1) blockade, and / or cytotoxic T-lymphocyte-associated protein 4 (CTLA-4) blockade. Immune checkpoint blockade therapy may include treatment with nivolumab (anti-PD-1 blockade), pembrolizumab (anti-PD-1 blockade), atezolizumab (anti-PD-L1 blockade), and / or ipilimumab (anti-CTLA-4 blockade).

[0143] Features expressed in their particular form or in relation to the manner of performing the disclosed function or the method or process for obtaining the disclosed result, as disclosed in the foregoing description, the appended claims, or the drawings, may suitably be used alone or in any combination of such features to implement the invention in its various forms.

[0144] Although the invention has been described in conjunction with the foregoing exemplary embodiments, many equivalent modifications and variations will be apparent to those skilled in the art when this disclosure is given. Therefore, the exemplary embodiments of the invention set forth above are to be considered illustrative rather than restrictive. Various changes may be made to the described embodiments without departing from the spirit and scope of the invention.

[0145] To avoid any doubt, any theoretical explanations provided herein are for the purpose of improving the reader's understanding. The inventor does not wish to be bound by any of these theoretical explanations.

[0146] Any section headings used in this document are for organizational purposes only and should not be construed as limiting the subject matter described.

[0147] Throughout this specification (including the appended claims), unless the context otherwise requires, the words “comprising” and “including” and variations thereof shall be understood to imply the inclusion of the indicated integer or step or group of integers or steps, but not to exclude any other integer or step or group of integers or steps.

[0148] It is important to note that, unless the context clearly indicates otherwise, nouns without quantifiers as used in the specification and appended claims include plural pronouns. A range herein may be expressed as “about” a particular value, and / or “about” another particular value. When such a range is expressed, another embodiment includes from one particular value and / or to another particular value. Similarly, when a value is expressed as an approximation using the antecedent “about,” it will be understood that the particular value forms another embodiment. The term “about” in relation to numerical values ​​is optional and means, for example, + / - 10%.

[0149] The following content is presented by way of example and should not be construed as limiting the scope of the claims.

[0150] Example

[0151] Materials and methods

[0152] Data and code availability

[0153] Will be via GEO ( This provides transcriptome data from mouse bone marrow-derived macrophages treated in vitro with synthetic PGE2. (The data is available for...) Obtain the appropriate code for analysis.

[0154] Statistical methods

[0155] For all computer-based analyses (in silico analysis), R version 4.0.3 was used. All statistical tests were performed using R, or in some cases, within GraphPad Prism (7.0d). Unless otherwise specified, values ​​less than 0.05 ( ), 0.01 ( ) and 0.001 ( The p-values ​​for all cases were labeled. The Benjamini-Hochberg method was used to correct the p-values ​​for multiple hypothesis tests. These are referred to as FDR values ​​throughout the text. Two-tailed nonparametric tests were used when the assumption of a normal distribution was deemed unreasonable. Tests are used to compare categorical variables. In cases with more than two groups, the Kruskal-Wallis test and the Dunn multiple comparison test are used, including p-value correction for multiple hypothesis testing using FDR. For univariate survival analysis, Wald's test p-value and log-rank test p-value are used to assess statistical significance. For multivariate survival analysis, Wald's test p-values ​​for each covariate are obtained using the coxph function in the survival R package (3.1-12).

[0156] TGCA differential gene expression and network analysis

[0157] Raw counts of TCGA (Cancer Genome Atlas) gene expression data, along with clinical data (syn12026747), were downloaded from the SAGE synapse database (syn4311114). Raw counts were TMM-normalized using calcNormFactors in edgeR (3.24.3). Genes not reaching 0.25 counts per million (CPM) in 10% of samples were removed from the analysis. Patients belonging to each TCGA study were stratified into high and low PTGS2 subgroups using the upper and lower quartiles of expression levels normalized using log2. Non-solid tumor types, such as LAML (acute myeloid leukemia) and DLBC (diffuse large B-cell lymphoma), were excluded from further analysis. A model matrix was designed to compare patients with high and low PTGS2 levels, adjusted for cancer type. This was performed using the top ten cancer types by median PTGS2 level. DGE analysis was performed using limma-trend (Ritchie et al., 2015). Upregulated genes were filtered according to a ligand list (Ramilowski et al., 2015) (P < 0.05, FC > 1.2), and the resulting 16 genes were used as the final PTI features (Table 1).

[0158] For gene set enrichment analysis (GSEA), based on directional log... 10 The p-values ​​were used to rank the genes. The ranking was then used as input to the fgsea (1.8.0) package (500 iterations). The genes were derived from a molecular feature database ( Download Hallmark gene signatures. Import upregulated genes from DGE analysis into the STRING database. A protein-protein interaction network was constructed using STRING. Nodes without any edges were removed. A table of biological processes with significantly enriched gene ontology was derived from STRING. The network was also exported and analyzed in R using the igraph package (1.2.7).

[0159] TCGA Survival Analysis

[0160] In all cases, the characteristic score of the target gene characteristic was calculated using the average expression of the characteristic gene with Z-score normalized expression data. Pairwise Spearman correlation coefficients were calculated using the rcorr function of the Hmisc package (4.3.0). Survival analysis was performed in R using the survival (3.1-12) package, and KM plots were plotted using survminer (0.4.6). For survival analysis, OS days and OS status information from the pan-TCGA clinical data table downloaded from SAGE synapse (syn12026747) were utilized. For pan-cancer survival analysis, continuous PTI scores were used, and stratification was performed within each pan-cancer immune subtype according to quartiles or median values ​​(Thorsson et al., 2018). In the CoxPH model, HR was adjusted by using cancer type and disease stage as strata. Detailed immunological and genomic annotations of TCGA patients were obtained using the supplemental file (Supp1a.txt) from Thorsson et al. (2018). For cancer type-specific survival analyses, continuous feature scores were used in a multivariate CoxPH model adjusted for available clinical parameters. Patient stages were converted to continuous variables (e.g., Ia and Ib = 1). To infer TMB in the TCGA data, the “non-silent mutation rate” annotated by Thorsson et al. (2018) was used. For HNSC, the final HPV status of patients was extracted from supplementary data from the TCGA study (Lawrence et al., 2015).

[0161] Analysis of cells treated with PGE2

[0162] Mouse BMDM was generated as previously described (Zelenay et al., 2015). Macrophages were treated with synthetic PGE2 for 24 hours, and then RNA was isolated and sequenced. FastQC was used (…). Reads from RNA sequencing experiments were quality checked and bases were trimmed. Next, the reads were aligned to the mouse genome (mm10) using STAR (Dobin et al., 2013) with default parameters. Then, the featureCounts function from the R package Rsubread (Liao, Smyth, and Shi, 2014) was used with Ensembl GTF annotations. Gene-level counting data were generated. DGE analysis was performed using the DESeq2 (1.24.0) package, and DGE was defined as those with FDR < 0.05 and 1 < log2FC < -1. Data for CD4 cells treated with PGE2 were downloaded from the gene expression omnibus (GEO). + Publicly available microarray datasets of T cells (GSE17200) and human fibroblasts (GSE63659) were used. DGE analysis was performed using the limma package to fit a linear model with array weights. Significant DEG was defined as FDR < 0.05 and 1 < log2FC < -1.

[0163] TCGA mutation analysis

[0164] Download the pan-cancer mutation annotation file from SAGE synapse (syn7214402). The R package maftools (2.2.10) was used for plotting and data exploration. Mutation data were filtered to exclude non-solid tumor types and variants classified as RNA, Silent, 3' Flank, or 5' Flank. For each patient sample, the total number of mutations was calculated and normalized to log2. Multivariate linear regression with the following formula was used to identify pan-cancer predictors of PTI expression: lm(PTI ~ mutation + log2-total mutations + cancer type). The same formula was applied to each cancer type for selected known cancer driver genes (Lawrence et al., 2014). The p-values ​​of the linear models were corrected using the Benjamini-Hochberg method. For the KRAS gene, data were obtained from cBioPortal (…). Downloaded pan-cancer TCGA mutation, structural variant, and copy number estimates.

[0165] CCLE PTI Expression Analysis

[0166] From the Cancer Cell Line Encyclopedia (CCLE) Download gene expression data ( ) and cell line annotations ( Transcript IDs were converted to gene symbols using the EnsDb.Hsapiens.v79 and ensembldb R packages. Cell lines named “FIBROBLAST” or “HAEM” were discarded. TPM values ​​were converted to a log2 scale with a pseudo-count of 1. Pairwise Spearman correlations between PTI genes were calculated, and the average Spearman value of the pairwise correlation coefficients of PTI genes was calculated, along with the average Spearman value of 1000 random gene features (n=16) of equal length to the PTI feature. MMP12 was not present in the CCLE dataset. To compare cell lines with high and low cancer cell-inherent PTGS2 expression, upper and lower quartiles were compared for DGE analysis using edgeR (3.28.1). Genes with a CPM greater than 1 in the 10% of samples were retained for analysis. Cancer types defined by the TCGA code variables in the cell line annotation file were adjusted in the modeling. Upregulated DEG was defined as those with a corrected P-value < 0.05 and log2FC > 2. Using the same procedure as PTGS2, PTI was compared. 高 and PTI 低 Cell lines. Highly upregulated genes (FC>5, FDR<0.05) from PTI comparisons were filtered to remove long non-coding RNAs and pseudogenes. For the final list of upregulated genes (n=500), motif enrichment analysis was performed using the Rcistarget (1.6.0) package. (The last part, "using data from...", appears to be incomplete and requires further context.) Downloaded radix sorting file We search for conserved motifs in the 500 base pairs immediately upstream of the TSS of the target gene.

[0167] CCLE mutation and somatic copy number analysis

[0168] From the DepMap portal ( Download mutation call () ) and copy number estimated by gene ( Information on cell line aneuploidy and chromosome arm copy number was extracted from Supplementary Table 1 of a recent study (Cohen-Sharir et al., 2021). For cell line mutation analysis, the following variant categories were discarded: RNA, silent, 3' flanking, and 5' flanking, and filtering was performed to include 568 published cancer driver genes (Martínez-Jiménez et al., 2020). Feature scores for PTI characteristics were modeled using multivariate linear regression with driver gene mutation status, mutation rate, and cancer type as covariates. For cancer type, each level of the TCGA code categorical variable was converted into a series of binary variables (e.g., HNSC = 1, others = 0). To map the KRAS gene ( Mutations in the gene sequence were analyzed using maftools (2.2.10). For gene-level SCNA estimation, a multivariate logistic regression model was fitted using covariates of gene copy number, aneuploidy, and cancer type to predict high and low cancer cell intrinsic PTI scores. Significant predictors were defined as those with an FDR < 0.25.

[0169] CCLE RPPA Data Analysis

[0170] Download RPPA data from the CCLE database ( Cell lines were filtered to include only those with KRAS mutations. Cell lines with high PTI (n=44) were then compared to those with low PTI (n=18) using the limma package. Upregulated proteins and phospholipids were defined as those with FC > 1.5 and FDR < 0.05, while downregulated proteins were defined as those with FC < -1.5 and FDR < 0.05.

[0171] Drug screening data analysis

[0172] Download the R data file containing gene expression data from cell lines screened by Garnett et al. (2012). The microarray data was normalized using the rma function from the affy (1.62.0) package, and the ID () was normalized using biomaRt (2.40.5). Convert the data to gene symbols. For observations mapped to a single gene, select the row with the highest variance. Calculate the intrinsic PTI characteristic score for each cell line, as in CCLE. Define the top 25% of cell lines with the highest expression as PTIs. 高Download the cell line drug screening data from the supplementary documentation of Garnett et al. (2012). As in Jerby-Arnon et al. (2018), the cell lines sensitive to each drug were defined as the top 25% with the lowest IC50. Drug cell line sensitivity was modeled as a function of cell line PTI status (high / low), while adjusting for cancer type as a covariate within a linear mixed-effects model using the lme4 package (1.1-23). ​​The following formula was used: Preferential killing of PTIs was inferred by filtering models in which PTI status was a significant predictor of cell line sensitivity (coefficient > 0 and P < 0.05). 高 Drugs in cell lines. The IC25, IC50, IC75, and IC90 concentrations of the drug refatinib against the target cell lines were extracted from the supplemental data of Garnett et al (2012).

[0173] TCGA Immunodeconvolution

[0174] The MCPcounter (Becht et al., 2016) immune deconvolution algorithm was applied to the pan-cancer TCGA expression dataset. The obtained immune cell scores were correlated with PTI features in each cancer type dataset, and Spearman correlation coefficients for PTI features were plotted using corrplot(0.9).

[0175] Single-cell RNA sequencing data analysis

[0176] All analyses of the scRNAseq data were performed using the Seurat package (3.1.5). Unwanted variant sources were regressed using the sctransform package (0.2.1) and the SCTransform function. Feature scores for individual cells were calculated using the AddModuleScore function. Original cluster annotations were used where possible. The FindMarkers function was used for DGE analyses between clusters or groups, with a cut-off of 25% cell expression in the target cluster. To plot the expression of individual PTI genes within a specific cluster, the mean log2-normalized TPM or CPM expression level was calculated for each cluster. Hierarchical clustering was performed on the mean expression matrix to cluster PTI genes by cell type-specific expression.

[0177] Lambrechts et al (2018) NSCLC:

[0178] The R data file containing log2CPM data and t-SNE coordinates was downloaded from ArrayExpress, accession code E-MTAB-6149. Total cell counts and mitochondrial content were regressed using sctransform (0.2.1). Clusters from the original publication were re-annotated to broad cell types. Dendritic cells were combinations of cells from clusters annotated as “cross-presenting dendritic cells (14)”, “Langerhans cells (3)”, “plasma-like dendritic cells (24)”, and “monocyte-derived dendritic cells (19)”. Clusters containing “fibroblasts” were grouped together as fibroblasts, and the same principle applied to endothelial cells and cancer cells. Naturally occurring lung cell types were discarded from the analysis. Tumors from patient 4 (PTI) were grouped together. 高 CD8 + Cells in the T cell cluster are compared to those from all other tumors.

[0179] Puram et al (2017) HNSC:

[0180] Using data compiled from recent research, from Zenodo ( Download the expression data and annotations as R data files. Discard cells from lymph node samples (HN10, HN12, HN13, HN24, HN7, HN8, HN23). The same cell filtering steps as in the original study (Puram et al., 2017) were used. sctransform regression was used to extract the cell counts for each cell and variables describing the enzymes used for reverse transcription. Six PCs were used for graph-based clustering, and the resulting clusters were annotated with biomarker genes and annotations from the original study. CD8 expression was compared between tumors with high and low intrinsic PTI expression in cancer cells using the FindMarkers function for DGE analysis. + T cells.

[0181] Jerby-Arnon et al (2018) melanoma:

[0182] Download gene expression from GEO (GSE115978) ) and cell annotation ( Cells with at least 500 genes were retained, and genes expressed by two or more cells were filtered out. Twenty PCs were selected for graph-based clustering. Twenty-five new clusters were defined and annotated using known biomarker genes and annotations from the original study. Clusters were grouped into broad cell categories to reflect those in other scRNAseq datasets. Sub-clustering was performed on all T cells to define CD4. +T cells, naïve-like T cells, and CD8 + T cells enable comparison of CD8 between tumors with high and low intrinsic PTI expression in cancer cells. + T cells.

[0183] TCGA CIBERSORT and CD8 + T cell characterization analysis

[0184] Using the pan-solid tumor TCGA expression dataset, so-called "hot" tumors were first selected by filtering to include only patients with tumors of the "IFNγ-dominant" immune subtype (Thorsson et al., 2018). Then, PTI features and their correlation with CD8 were calculated. + Several T-cell phenotype-related features include dysfunction (Singer et al., 2016), unresponsiveness (Azizi et al., 2018), and exhaustion (Guo et al., 2018), with the unresponsiveness stemming from tolerance to autoantigen-specific CD8. + Gene set from preliminary research on T cells (Schietinger et al., 2012). Calculation of CD8. + Spearman correlation coefficient between T cell phenotypic characteristics and intratumoral PTI scores. CIBERSORT data from Thorsson et al. (2018) were used to compare the proportions of different immune populations among total immune cells in tumors with high, intermediate, and low PTI. The mean proportions of each population were compared between tumors with high and low PTI using a nonparametric t-test, with p-values ​​adjusted for multiple hypothesis testing.

[0185] Analysis of Immune Checkpoint Blockade Dataset

[0186] Transcriptome data from patients treated with ICB from 12 publications were compiled. For datasets requiring normalization from counts to TPM values, genes expressing higher than 0.25 CPM in less than 10% of samples across the entire population were first filtered out. For three datasets (Kim et al., 2018; Mcdermott et al., 2018; Gide et al., 2019), FASTQ files were downloaded. FastQC software was used ( The RNA-seq reads were quality checked and bases were trimmed. Then, the reads were aligned to the human genome assembly using the STAR alignment tool (Dobin et al., 2013) with default parameter settings. Using featureCounts (Liao, Smyth, and Shi, 2014) and Ensemble GTF annotations ( The mapped data is converted into integer counts of transcript levels. Transcript counts are then converted into TPMs.

[0187] For each cohort, patients were stratified into high and low trait score groups based on the median cut-off value. Then, the following methods were used: Tests were used to compare the proportion of responders (PR / CR) versus non-responders (PD) between stratified patient groups. SD patients were excluded from this analysis. To model the response to ICB in each dataset, a multivariate logistic regression model was fitted using TIS features alone as baseline predictive models. In the multivariate logistic regression model, each resistance feature was combined with the TIS feature, and the model included a linear multiplicative interaction term of the two variables in the model (e.g., glm(response ~ TIS + PTI + TIS:PTI)). The AUC value for each model was calculated using the pROC (1.17.0.1) R package, and the AUC of the 16 datasets was compared by a nonparametric t-test. OS was also compared for patients in each quadrant using baseline and pre-ICB datasets where OS data was available. For the pan-cancer immunotherapy meta-dataset, a multivariate Cox regression model was fitted, which included the dataset and TIS status as covariates. The same analysis was performed on 1000 random gene traits and published ICB resistance traits. The Wald test p-values ​​from each resistance / random trait in each multivariate CoxPH model were corrected using the Benjamini-Hochberg method.

[0188] Analysis of the developmental stages of lung squamous cell carcinoma

[0189] Microarray data for different developmental stages of LUSC were downloaded from the GEO database (GSE33479) (Mascaux et al., 2019). When multiple probes mapped to the same gene, the probe with the highest variance in the sample cohort was selected. PTI eigenvalues ​​were calculated as in TCGA analysis.

[0190] In vitro assay of lung cancer cell lines

[0191] Human lung cancer cell lines NCIH1339 and NCIH3255 were cultured in Nunc T75 cell culture flasks (Life Technologies: 156499) in RPMI 1640 medium (Lonza: LZBE12-702F) containing L-glutamine and 10% fetal bovine serum (Merck: F7524) and 5% penicillin / streptomycin (Thermo: 11548876) at 37°C. Cells were washed with Dulbecco phosphate-buffered saline (Fisher: 12559069), resuspended in trypsin-EDTA, and cultured with 1×10⁻⁶... 5 1 cell / well plate seeding in 12-well plates, or 2 × 10⁻⁶ cells / well plate. 4 Cells were seeded per well in 96-well plates (Starlab: CC7682-7596). After 24 hours, the medium was removed from the plates and replaced with fresh medium containing 0.1% dimethyl sulfoxide as a control, or with 1 μM or 10 μM trametinib or selumetinib. At subsequent time points of 1, 2, 4, 6, 8, and 24 hours, the supernatant was collected and stored at -20°C. Cells were resuspended in trypsin-EDTA and centrifuged at 1500 rpm for 5 minutes. The supernatant was discarded, and the cell pellet was stored at -80°C. To monitor and infer cell proliferation, the drug-treated 96-well plates were imaged for three days at two-hour intervals using an Incucyte instrument. The percentage of confluence and changes in confluence relative to baseline were used as surrogate markers of cell proliferation.

[0192] To measure soluble cytokines and chemokines in cell culture supernatant, the Luminex assay protocol was used. A 96-well flat-bottom Luminex plate was washed with purified H2O. 50 μl of lyophilized protein standards were prepared with cell culture medium, centrifuged at 15,000 rpm for 10 min, and placed on ice for 10 min. Serial dilutions of the protein standards were prepared using cell culture medium. Beads were diluted 1:100 in assay wash buffer and added to each well with 50 μl. The plate was then placed on a magnet for two minutes, the liquid was flicked off, and 50 μl of the sample / standard diluted 1:2 in culture medium was plated. The plate was shaken for two hours and then washed three times with assay wash buffer. Antibodies were diluted 1:200 in detection antibody dilution and added to each well with 25 μl. The plate was then shaken at room temperature for 30 min and washed three more times with wash buffer. Add 30 μl of streptavidin-PE to each well and place the plate on a shaker for 30 minutes. Then wash the plate four more times in washing buffer and add 75 μl of reading buffer to each well. The plate was then read using a Luminex MAGPIX instrument. Protein concentrations were extrapolated from a standard curve fitted to the net MFI level within a GraphPad Prism v8.

[0193] CPI1000 Dataset Analysis

[0194] The inventors were granted access to single PTI gene expression, clinical, and genomic biomarker data from the CPI1000 dataset (Litchfield et al., 2021). Expression data from 523 patients were available. Log2-normalized TPM values ​​(log2(TPM+1)) for each gene were Z-score normalized, and PTI features were calculated as previously. MMP12 was omitted from the data, leaving 15 / 16 PTI genes present. Continuous variables were co-correlated, and hierarchical clustering was performed using Spearman Rho values ​​with the default parameters in the corrplot function (0.84) package. SD patients and samples under treatment were again discarded. Univariate logistic models were fitted for each biomarker to predict PD versus PR / CR patients. Effect sizes, standard errors, and Wilcoxon p-values ​​from six studies were extracted and submitted for meta-analysis using the R meta package (4.18-0) and the metagen function. To compare the effects of measuring PTI in combination with other biomarkers, a multivariate logistic regression model with four variables (tumor type, clonal TMB, CXCL9, and PTI) was fitted, including linear multiplication interaction terms for numerical variables.

[0195] The nested models were then compared using a likelihood ratio test. The variance of the patient response explained by each model was estimated by calculating the deviation using the anova function. For the survival analysis in the CPI1000 dataset, patients were stratified according to the 75% cutoff value of PTIs and the median values ​​of clonal TMB, CXCL9, and CD274. To determine the optimal combination of PTI genes for predicting ICB response in the CPI1000 dataset, several resilient network regression models were trained using the glmnet R package. Two scaled expression matrices were created using log2-normalized TPM values ​​from the CPI1000 dataset. For the first, the data for all samples was scaled (tumor type independent); for the second, the data was grouped by cancer type and subsequently scaled (tumor type specific). Patients with SD and those with samples in treatment were removed. The corresponding matrices were split into 70% training data and 30% test data. Resilient network regression models were trained using 10-fold cross-validation to optimize... Hyperparameters were configured such that the value of the α hyperparameter increased in increments of 0.1 between 0 and 1. For each trained model, AUC was used as a metric to select the best model. Each trained model was then used to predict patient ICB responses in the test dataset.

[0196] Manchester ccRCC group patients

[0197] Ethical approval was obtained under the Manchester Cancer Research Centre Biobank Research Tissue Bank Ethics for a retrospective analysis of tumor FFPE and plasma from patients with RCC (ref: 18 / NW / 0092). Forty-eight surgical blocks were identified from patients who received immunotherapy at some stage of their treatment course, including those treated with high-dose IL-2. No tumor tissue was observed after pathological evaluation of H&E-stained sections from three tumors. Two blocks came from the same patient who underwent two surgeries to remove brain metastases. Another patient had different blocks from surgeries involving the removal of adrenal and lung metastases. Pathologists marked tumor areas and healthy tissue on the H&E sections and defined histological grades and morphology.

[0198] Formalin-fixed paraffin-embedded (FFPE) tissue processing, RNA extraction, and gene expression profiling analysis

[0199] Tissue excised by FFPE surgery is the most common clinically available sample and can be used as input material for molecular inflammatory profiling analysis, such as using the NanoString nCounter platform. H&E (hematoxylin and eosin) stained FFPE ccRCC blocks are sectioned and placed on slides. The samples are macro-dissected to exclude unwanted tissues, such as predominantly necrotic or normal tissue, blood clots, and cartilage, and then scraped into 1.5 mL Eppendorf. Tissue sections of an average size of 15 × 5 μm are used. Smaller samples that do not require macro-dissection are directly excised from the block into 1.5 mL Eppendorf. RNA is extracted from up to 10 slides using the commercially available Qiagen RNeasy FFPE kit (RNeasy FFPE kit for RNA extraction; RSP-54). Figure 17First, total RNA concentration was determined using a NanoDrop instrument (Thermo Fisher). RBA quantification was also performed using a Qubit 2.0 fluorometer (Invitrogen), and quality estimates were calculated using the Bioanalyzer (Agilent) platform. RNA RIN values ​​ranged from 1 to 6.4, with an average of 2.2. The extracted RNA was then used as input material for a molecular inflammatory profiling system, such as the NanostringnCounter platform, which is compatible with samples exhibiting low RNA quality and high fragmentation, such as those derived from FFPE tissue.

[0200] Nanostring gene expression analysis

[0201] Gene expression profiling was performed using the Nanostring nCounter protocol and a commercial pan-immune codeset including an additional 30 custom genes. For the nCounter workflow, suitable housekeeping genes were selected and added to the PTI gene coding set. To quantify gene expression, 150 ng of total RNA was hybridized overnight at 65°C with capture and reporter probes specific to PTIs and housekeeping genes. Subsequent reactions, including the hybridized sample and master mix, were then fed to the Nanostring nCounter preparation station, which aligned the samples for imaging. The samples were then scanned using the nCounter digital analyzer, and the fluorescent barcodes were counted. This involved quantifying the target RNA transcripts by directly counting the fluorescence signal from the reporter-capture probe complex retained on the cartridge surface. Nanostring output data (RCC file) was analyzed in R using the NanoStringQC function in the nanostringr package (0.1.4.9000). This analysis involved exporting the resulting gene expression counts to NanoString nSolver 4.0 software for quality control analysis and normalization (for potential internal control and / or housekeeping genes). NanoString nSolver 4.0 software enables rapid and direct analysis and correction of batch effects using reference signals without requiring advanced computational skills. The analysis flagged samples that failed to meet the necessary thresholds for data quality (signal-to-noise ratio > 50, percentage of detected genes > 50%). A total of three samples failed the quality control steps due to low signal-to-noise ratios. The published workflow for normalizing NanoString count data was followed (Bhattacharya et al., 2020). First, a negative binomial model was fitted for each housekeeping gene expression using the MASS package (7.3–53) with binary variables as predictors (primary renal cell carcinoma and others). Nine housekeepers were excluded based on P < 0.05. Then, the RUVseq package (1.24.0) was used to identify and remove unwanted variant sources with parameter k=1. The Limma (3.46.0) and DESeq2 (1.30.0) packages were used for DGE analysis. Batch-normalized expression data were then used for downstream correlation analysis. Gene signature scores were then calculated as the average expression of batch-corrected expression data normalized to Z-scores.

[0202] Example 1: PTI: Characterized by human-derived COX-2-centered pro-tumorigenic inflammation

[0203] Given the extensive literature suggesting that the COX-2 / PGE2 pathway promotes multiple cancer markers, including the inventors’ previous work establishing this inflammatory axis as central to immune escape (Zelenay et al., 2015; Bonavita et al., 2020; Pelly et al., 2021), it is hypothesized that transcripts associated with high COX-2 mRNA levels will capture pan-cancer conserved features of pro-tumorigenic inflammatory signaling networks.

[0204] Mining TCGA data revealed that the transcript levels of PTGS2, which encodes COX-2, are heterogeneous both between and within cancer types. Figure 1 A). Therefore, the inventors designed a computational process that compares high and low PTGS2 levels in tumors from the top ten cancer types expressing PTGS2. Figure 1 A, Figure 1 B). Among the 54 genes overexpressed in PTGS2-high cancers, in addition to ECM modifiers HAS2 and MMP12, there are also classic secretory factors associated with local and systemic inflammation, such as IL6, CXCL8, IL1B and CXCL2, cell surface receptors (CSF3R, IL7R, CD69), neutrophil marker (GOS2), and EGFR ligands and growth factors such as AREG and EREG. Figure 1 B). Gene set enrichment analysis (GSEA) using Hallmark pathway characteristics showed that the TNF-NFκB pathway and other inflammatory pathways (IL6-STAT3) were involved in COX-2. 高 Upregulation in cancer ( Figure 8 A). Interestingly, it was also overexpressed in epithelial-mesenchymal transition (EMT), complement, hypoxia, KRAS signaling, and apoptosis pathways (FDR < 0.05, NES > 1). Figure 8 A). Therefore, COX-2 expression is a pan-cancer node associated with inflammatory tumors that express several other cancer markers.

[0205] Next, genes encoding extracellular molecules (including secreted factors) are inferred to be strong candidate guiding regulators of the cellular and molecular immune TME profile. This is achieved by filtering Ptgs2 using a “predicted ligand” list (Ramilowski et al., 2015). 高Genes upregulated in tumors. The genes obtained are cytokines (IL1A, IL1B, IL6), chemokines (CXCL1, CXCL2, CXCL5, CXCL6, CXCL8), and growth factor (NRG1). These, along with PTGS2, are named the 15-gene signature, namely the protumourigenic inflammatory (PTI) signature.

[0206] Human cancer cell lines with high PTGS2 expression had significantly elevated expression of evaluable PTI genes in 13 / 15 of them. Figure 8 B). Two of the most abundant non-cancerous cell types in the TME are fibroblasts and macrophages. To investigate whether PGE2 signaling to fibroblasts and macrophages could explain the association of these ligands with COX-2 in tumors, transcriptomic data from cells treated with synthetic PGE2 in vitro were analyzed. Notably, in both fibroblasts and macrophages, each ligand was regulated by PGE2 ( Figure 1 D). Many ligands are also affected by human CD4. + Induction of T cells by PGE2 treatment ( Figure 1 D). These findings validate the inventors' bioinformatics process and demonstrate that the PTI signature of the ligand represents a hierarchical downstream node of the COX-2 / PGE2 axis, which functions as a "hub" in the network.

[0207] To investigate COX-2 高 To determine whether upregulated genes in tumors are members of biologically significant and cohesive signaling cascades, a protein-protein interaction network was constructed using a compiled STRING database (Szklarczyk et al., 2019). Figure 1 C). Consistent with their shared participation in signal transduction pathways, COX-2 高 The upregulated genes in tumors encoded proteins with significantly more interactions than randomly expected (P < 1E-16), averaging 9.7 edges per node. These proteins were overexpressed in gene ontology biological processes such as inflammatory responses, myeloid cell chemotaxis, responses to cell death, and stress. Figure 1 D). This data highlights how the genes identified in this process participate in the sharing process and perform signal transduction cascades, with COX-2 at the center of the network.

[0208] Example 2: PTI characteristics as a powerful indicator of poor prognosis

[0209] Next, we questioned whether PTI characteristics were associated with poor prognosis. Using data from The Cancer Genome Atlas (TCGA), we performed pan-cancer Cox Proportional Hazard (CoxPH) regression analysis, using disease stage and cancer type as strata. PTI characteristics were strongly associated with shortened survival in cancer patients, both as a continuous variable (HR = 1.23, P = 1E-05) and based on quartile stratification (PTI). 高 Compared to PTI 低 (HR = 1.25, P = 0.02) Figure 1 E).

[0210] Example 3: PTI characteristics are a powerful indicator of poor prognosis in various cancer immune subtypes.

[0211] The interaction between PTI and cancer-suppressive inflammation features in predicting overall survival (OS) in cancer patients was investigated using multivariate CoxPH models including interaction terms (Ayers et al., 2017; Danaher et al., 2017). By including linear multiplicative interaction terms (e.g., the product of PTI and TIS features) in the model, the impact of PTI on survival was estimated to depend on the degree of IFNγ signaling and T cell inflammation. As expected, PTI was generally associated with increased mortality (HR = 1.25, P < 0.0001). Figure 2 A). Therefore, for a given patient with an average IFNγ signaling level, an increase in PTI by one value was associated with a 25% increased risk. In contrast, for patients with an average PTI level, an incremental increase in IFNγ signaling corresponded to a 16% reduction in the risk of death (HR = 0.84, P < 0.0001). Figure 2 A). A significant interaction term exists between PTI and IFNγ features (HR=1.13, P=0.004). To illustrate this interaction, the hazard ratio at high and low expression levels of IFNγ signaling was simulated ( Figure 2 B). In patients with low levels of IFNγ signaling, the risk rate remains elevated regardless of PTI levels, while in patients with high IFNγ signaling, the risk of death is lower and significantly increases with PTI expression. Figure 2 B). In fact, at high levels of PTI (e.g., PTI score = 1.5), Figure 2 In case B), patients with high IFNγ signaling in the TME (which is generally a marker of better prognostic tumors) will have a similar risk of death as patients with non-T-cell inflammatory tumors. Therefore, in patients with simultaneously low PTI levels, high intratumoral IFNγ signaling is associated with better outcomes. (The text abruptly shifts to a seemingly unrelated topic about PTI and CD8 signaling.) +Similar interaction effects were observed between T cells and TIS signatures (Table 2). Overall, these findings reveal a close cross-talk between pro-tumorigenic and anti-tumorigenic inflammatory phenotypes, highlighting the significant antagonistic effect of PTI signatures on T cell-induced inflammatory tumors.

[0212] Immune infiltration and TME vary considerably across tumor types, but previous pan-cancer analyses have defined six broad immune subtypes: “wound healing (C1)”, “IFNγ-dominant (C2)”, “inflammatory (C3)”, “lymphocyte depletion (C4)”, “immune quiescence (C5)”, and “TGFβ-dominant (C6)” (Thorsson et al., 2018). These subtypes have been found to possess unique genomic, transcriptomic, and clinical characteristics (Thorsson et al., 2018). It is hypothesized that PTI will preserve prognostic potential and its antagonistic effect against tumor immunity across different immune contexts. PTI was highest in the “wound healing”, “IFNγ-dominant”, and “TGFβ-dominant” subtypes. Figure 9 A), and has a significant prognostic effect in all three subtypes as well as the "inflammatory" and "lymphocyte depletion" subtypes. Figure 2 C). On the other hand, TIS, CD8 + T cell and IFNγ signaling signatures are more heterogeneous, and are highest in the "IFNγ-dominant" subtype. Figure 8 B). Furthermore, biomarkers for these so-called “hot” T-cell inflammatory tumors, as well as antigenic biomarkers such as non-silent mutation rate (referred to in this article as tumor mutational burden (TMB)) and single nucleotide variant neoantigen burden (SNV NAB), show inconsistent and uncommon prognostic outcomes across subtypes. Figure 2 (C) (Thorsson et al., 2018). Therefore, PTI can be a prognostic indicator in various tumor immune subtypes with completely opposite characteristics, while typical markers of so-called "hot" tumors, such as IFNγ signaling, are associated with prolonged OS in a single immune subtype.

[0213] Example 4: PTI characteristics are a powerful indicator of poor prognosis in each IFNγ-dominant subtype.

[0214] To expand this analysis, the inventors focused on the "IFNγ-dominant" subtype, which encompasses patients with highly inflammatory T-cell tumors, and for these patients, CD8... +T cells have shown a clear association with better outcomes, which is generally true for patients treated with immunotherapy. In these highly invasive tumors, PTIs are strongly associated with poor survival at both high and low IFNγ levels. Figure 2 D). In fact, despite IFNγ and CD8 + T cell infiltration levels were also high, but PTI 高 IFNγ 高 The median overall survival (OS) of patients with PTI was approximately 5 years, compared to which PTI patients had a median overall survival (OS) of approximately 5 years. 低 IFNγ 高 Median OS in the group > 15 years ( Figure 2 D). These results indicate that the benefits of high levels of IFNγ signaling are strongly antagonized by the accompanying PTI expression. Similar findings were observed when different established immunomodulatory signatures indicative of T-cell inflammation (TIS, IFNγ, CD8, PDL1) or when the proportion of tumor-infiltrating lymphocytes (TILs) determined by immunohistochemistry (Thorsson et al., 2018) and PD-L1 (CD274) mRNA expression were used instead of IFNγ signaling signatures, all of which are reported indicators of highly immunogenic, ICB-responsive tumors. Figure 9 B). Therefore, COX-2-centered PTIs have antagonistic effects on multiple markers of more immunogenic tumors.

[0215] To account for potential confounding effects of organ-specific immune signatures (Salmon et al., 2019), a multivariate CoxPH model was used to examine the interaction between PTI and IFNγ signaling in various cancer types. Across nine cancer types, PTI signatures were found to be statistically significant prognostic factors, both as a single variable and through their interaction with IFNγ signaling, after adjusting for clinical factors such as age, sex, and disease stage. Figure 2 E (ccRCC (n=531), LIHC (n=374), PAAD (n=176), glioblastoma (GBM, n=168), CESC (n=306), adrenocortical carcinoma (ACC, n=79), UCEC (n=277), THCA (n=513), and HNSC (n=240)). In pancreatic cancer (PAAD), ccRCC, lung adenocarcinoma (LUAD), liver cancer (LIHC), and cervical cancer (CESC), regardless of IFNγ signaling levels, elevated PTI was associated with higher mortality. Figure 2E). Significant interactions between PTI and IFNγ signaling were observed in thyroid cancer (THCA), head and neck cancer (HNSC), and uterine cancer (UCEC), thus mimicking observations in pan-cancer datasets. Figure 2 E). In summary, the prognostic potential of PTI extends to various malignancies, and crucially, it is usually related to its inhibitory effect on IFNγ signaling.

[0216] Example 5: Prognostic ability of each PTI gene in PTI characteristics

[0217] Survival analysis of each PTI gene was performed to assess whether there were dominant prognostic factors that could explain the characteristics of PTIs. The COX-2 gene (PTGS2) itself indicates poor survival in ccRCC patients. Figure 9 C). EGFR ligands EREG and AREG are associated with poor survival in cancers with prevalent KRAS mutations (such as PAAD) or cancers with extensive alterations in the receptor tyrosine kinase-RAS pathway (such as GBM and THCA). Cytokines IL11 and IL1A, as well as chemokines CXCL5 and CXCL1, have prognostic roles in CESC and ccRCC. This data highlights how inflammatory ligands heterogeneously promote the association between PTI and poor survival in a variety of malignancies, and that no single gene has such a consistent prognostic role as the trait itself.

[0218] Example 6: Association between PTI levels and cancer stage

[0219] This study questions the association between PTI levels and cancer stages. To this end, PTI levels were compared in healthy, precancerous, and malignant disease states using a lung squamous cell carcinoma (LUSC) dataset, which contains gene expression data from nine progressive cancer development stages (Mascaux et al., 2019). PTI levels were lowest in normal lung tissue and gradually increased. Figure 9 D). Increased expression of PTI features was first significant in moderate developmental dysplasia lesions and peaked in cancer samples ( Figure 9 D). These findings suggest that highly proliferating precancerous lesions begin to utilize (co-opt) inflammatory transcriptional networks before becoming fully malignant.

[0220] Finally, the correlation between disease progression stage and PTI levels was investigated using the TNM cancer staging criteria from TCGA (which takes into account the size of the primary tumor, lymph node spread, and the presence of metastases in other organs). For pan-cancer, a modest but statistically significant correlation was found between PTI expression and tumor stage, regardless of cancer type (P=0.005). In the selected cancer types, PTI levels significantly increased with disease progression. Figure 9 E). In most malignancies, PTI increases in pre-metastatic tumor stages (I to III), while in LUSC, PTI levels are comparable between stages I and III and elevated in stage IV tumors with distant metastases. Figure 9 E). In summary, PTI is associated with oncogenic transformation, tumor formation, and progression to more advanced and metastatic stages of disease. Overall, these data support the view that inflammation is indispensable for tumor progression and establish PTI characteristics as a broad and consistent indicator of poor prognosis.

[0221] Example 7: MEK inhibition weakens the secretion of pro-tumorigenic inflammatory mediators in cancer cells.

[0222] Considering previous observations that many cancer cell lines express COX-2 and secrete PGE2 (Zelenay et al., 2015; Bell et al. 2022), this study evaluated whether COX-2 is co-expressed with other PTI genes in cancer cells. Transcriptome data from over 800 human cancer cell lines in the Cancer Cell Line Encyclopedia (CCLE) were analyzed, revealing heterogeneous expression of PTI genes among cell lines, but with positive correlations. Figure 3 A).

[0223] To identify methods for targeting cancer cells expressing the PTI factor, data from a pan-cancer drug screening study were analyzed. This study involved treating over 600 human cancer cell lines with matched transcriptional data using a suite of chemotherapy and targeted agents (Garnett et al., 2012). Cell lines were grouped by PTI expression, and the association between PTI status and drug sensitivity was screened while adjusting for cancer type. This revealed the potential for targeting PTI-specific cancer cells. 高 The line is selectively sensitive to four different MEK inhibitors (refatinib, CI1040, PD0325901, and AZD6244). Figure 3 B, Figure 10 A). In comparison, PTI 低 The cell lines are sensitive to receptor tyrosine kinase (RTK) inhibitors and PARP1 / 2 inhibitors that target VEGFR, PDGFR, and FGFR. Figure 3 B). Therefore, PTI 高 Cell lines are not easily subjected to comprehensive inhibition of RTKs, most of which are thought to transmit signaling through ERKs. In fact, more targeted inhibition of the MAPK cascade with MEK1 / 2 inhibitors leads to better control of PTIs. 高 Instead of PTI 低Effective growth inhibition of cell lines. For example, for cell lines treated with the MEK1 / 2 inhibitor refatinib, there was a significant inverse correlation between PTI and the half-maximal inhibitory concentration (IC50) of refatinib. Figure 3 C, upper quadrant). Top left quadrant ( Figure 3 C, above image) shows refatinib-sensitive PTIs 高 The group encompasses several cancer types, including 27% NSCLC, 18% esophageal cancer, 8% pancreatic cancer, and 8% bladder cancer. Compared to all other groups, refatinib-sensitive PTIs... 高 The group contained significantly more NSCLC (21% vs 7%, FDR = 0.006). Focusing solely on NSCLC cell lines, an even more significant inverse correlation was found between refatinib IC50 and PTI (…). Figure 3 C, as shown in the image below).

[0224] To determine whether pharmacological inhibition of MEK inhibits PTI production in cancer cells, two human lung cancer cell lines (NCIH1339 and NCIH3255) with high PTI levels based on CCLE transcriptional data were investigated. The approved selective MEK1 / 2 inhibitor trametinib inhibited the proliferation of these cells 24 hours after treatment. Figure 10 C, Figure 10 D). Consistent with their transcriptomes, analysis of their supernatants by multiplex cytokine / chemokine assays showed that both cell lines secreted PTI factors, such as IL-1α, CXCL1, IL-6, and IL-8. Figure 3 C). Notably, trametinib inhibited KRAS as early as 6 hours and as late as 24 hours after treatment, before having any significant effect on cell growth. G12R (NCIH1339) and KRAS WT EGFR-amplified cell line (NCIH3255) produces all these inflammatory mediators. Figure 3 C). Similar results were obtained with the second MEK1 / 2 inhibitor, selumetinib ( Figure 10 E). Notably, selumetinib did not alter the production of osteopontin (an inflammatory mediator that is not part of the PTI), indicating that MEK inhibition does not indiscriminately inhibit cytokine production. Figure 10 E). In summary, these data indicate that inhibiting MEK kinase activity can rapidly suppress the production of PTI factors by cancer cells.

[0225] Example 8: NFκB links MAPK pathway mutations with intrinsic PTI expression in cancer cells

[0226] PTI 高The varying sensitivities of cell lines to MEK inhibitors suggest that oncogenic alterations leading to increased MAPK activity may be associated with PTI levels in cancer cell lines and tumors. Consistent with this hypothesis, KRAS-mutant cell lines exhibit higher PTI expression (…). Figure 4 A). Next, a screening of cancer driver gene copy number estimates was performed (Martínez-Jiménez et al., 2020), which predicted high intrinsic PTI expression in cancer cells independent of cancer type and aneuploidy. Genes positively correlated with PTI levels included HRAS, RASA1 (RAS GTPase activating protein (GAP)), RTK PDGFRB and FLT4 (which transmit signals via MAPK signaling), and RELA (a gene encoding the NFκB subunit p65, which can be transactivated by MEK). Figure 4 B, Figure 11 (A)(Chang et al., 2003). In summary, these results indicate that KRAS mutations and / or amplifications in MAPK-NFκB signaling typically drive PTI expression in cancer cells.

[0227] Given the crossover of the MAPK and NFκB pathways, it is inferred that the aggregation of NFκB activation may be the main driver of PTI levels. Indeed, analysis of reverse phase protein array (RPPA) data showed that KRAS mutant cell lines with high PTI had significantly increased phosphorylation levels of activated p65 (RELA) serine 536, and also higher levels of EGFR, ECM component fibronectin, transglutaminase, and annexin-1, which regulate NFκB activity (Moraes, Ampomah, and Lim, 2018). Figure 11 C). The phosphorylation level of p65 serine 536 gradually increased in correlation with PTI expression ( Figure 4 C). Furthermore, motif enrichment analysis showed that NFκB-binding motifs were significantly enriched in the promoter regions of the expanded PTI gene set (see Methods). Among the top ten motifs significantly enriched in the PTI gene promoter regions, three were NFκB transcription factors, including a highly conserved motif combining both the classical form (NFKB1) and the non-classical form (NFKB2) of NFκB. Figure 4 D, Figure 4 E). In summary, these findings at the gene and protein levels suggest that NFκB activity promotes the expression of the PTI transcriptional program in cancer cells carrying KRAS mutations.

[0228] To test whether intratumoral PTI levels are associated with genomic alterations leading to overactivation of the MAPK and NFκB pathways, mutational calls from the pan-cancer TCGA dataset were analyzed. Consistent with cancer cell line data, mutations in MAPK pathway components (MAP2K1, NF1, and BRAF encoding MEK1) were significant predictors of pan-cancer PTI expression, independent of mutational burden and cancer type (FDR < 0.25). Notably, TP53 mutations were also strongly positively correlated with PTI levels. Figure 4 F), while mutations in the oncogene NFE2L2 or tumor suppressor gene mutations in the NRF2 (AXIN2) and WNT (KEAP1) pathways are associated with PTI. 低 Enriched in tumors ( Figure 4 F). The same analysis across all cancer types yielded >30 significant mutations associated with high PTI expression (FDR < 0.25). Most of these mutations were MAPK pathway genes, including KRAS in LUAD, ccRCC, and PAAD, NF1 in LUSC and GBM, and BRAF in THCA. Figure 4 G). Therefore, MAPK activation is a PTI in many tissue types. 高 Tumor markers. Furthermore, EGFR and RELA amplification in solid tumors were found to be positively correlated with intratumoral PTI characteristic scores. Figure 4 H). Therefore, analysis of cancer cell lines and bulk tumor genomic and transcriptomic data showed that both mutations in the MAPK / NFκB pathway and SCNA drive the expression of PTI factors.

[0229] Example 9: High PTI signature expression underscores in different malignant and non-malignant cell types

[0230] To study PTI 高 With PTI 低 To analyze the infiltration composition, researchers used the MCPcounter algorithm to infer the bulk tumor transcriptome, calculating pairwise Spearman correlations between PTI levels and various immune and non-immune cell types (Becht et al., 2016). In most cancer types, PTI signature levels were significantly positively correlated with multiple cell types, including endothelial cells (26 / 33), fibroblasts (30 / 33), neutrophils (31 / 33), and monocyte lineage cells (28 / 33). In contrast, cytotoxic cells, such as CD8... + The correlation between T cells (X / 33) or NK cells (X / 33) and PTI expression was poor and the consistency was much lower. Figure 12 A). This data indicates that PTI 高The main tumor-infiltrating cell components of the tumor include various myeloid cell subsets and mesenchymal cells such as fibroblasts, indicating that the intrinsic PTI expression of cancer cells leads to the expansion and / or recruitment of different cell types.

[0231] To determine the cellular origin of PTI factors, three independent scRNAseq datasets from three different cancer types (HNSC, MEL, NSCLC) were queried (Puram et al., 2017; Jerby-Arnon et al., 2018; Lambrechts et al., 2018). Single cells from tumor samples were re-clustered using unsupervised, graph-based clustering, with state assignment guided by metadata from the original studies. t-distributed stochastic neighbor embedding (t-SNE) identified multiple and distinct cell types as PTI sources. Figure 5 A, Figure 12 B). Measured mean expression of each PTI gene in each primitive cluster showed high levels of expression in neutrophils, macrophages, fibroblasts, mast cells, and endothelial cells. Figure 5 B, Figure 12 C). Some PTI genes are expressed at higher levels in specific cell types. For example, IL6 and HAS2 are mainly expressed by fibroblasts, while CXCL5 is specific to macrophages. Figure 5 B, Figure 12 C). Other PTI genes, including VCAN, AREG, PTGS2, and CXCL2, are expressed in various immune and non-immune cell types. Figure 5 B, Figure 12 C). In summary, this data indicates that, in addition to being directly expressed by cancer cells, PTI factors can also be heterogeneously produced by various immune and non-immune cell types.

[0232] Next, PTI levels within individual cancer cells isolated from ten independent HNSC patients were compared. There was heterogeneity in PTI expression among the patients. Figure 5 D). Researchers compared PTI levels within cancer cells from different patients to define PTI based on intrinsic expression in cancer cells. 高 and PTI 低 Tumor ( Figure 5 D). In summary, these findings support a model in which inflammatory mediators are also expressed by many cell types within the TME, many of which are co-expressed by cancer cells and can be induced by PGE2.

[0233] Example 10: PTI characteristics affect the proportion and transcriptotype of tumor-infiltrating cytotoxic lymphocytes.

[0234] To investigate whether high PTI levels are associated with a unique leukocyte infiltration component in "hot" tumors, CIBERSORT estimates of the proportion of immune cells in the "IFNγ-dominant" subtype of tumors were analyzed (Newman et al., 2015; Thorsson et al., 2018). Comparing the mean cell type proportions among patients with different PTI levels revealed that the proportion of leukocytes was generally consistent across groups. Figure 5 E). Lymphocytes in PTI 高 The proportion of tumor cells among total immune cells was significantly lower (FDR = 0.00003). Figure 5 E). PTI 高 Tumors have a statistically significantly lower proportion of CD8. + T cells (FDR = 0.002) and activated NK cells (FDR = 8.6E-08), but with a high proportion of activated memory CD4 cells. + T cells (FDR = 1.2E-19) and resting NK cells (FDR = 1.2E-30, Figure 5 E). These results indicate that the overall immune infiltration of “hot” tumors with high levels of PTI is subtly altered, and the effectors of antitumor immunity are reduced.

[0235] CD8 infiltrating tumors + T cells exhibit a range of transcriptional states, from naive to dysfunctional and cytotoxic (van der Leun, Thommen, and Schumacher, 2020). Based on previous analyses, it is hypothesized that PTI expression promotes CD8 expression. + T cell dysfunction, and therefore will be associated with suboptimal CD8. + T cell transcriptional signatures are positively correlated. Based on this hypothesis, PTIs were found to be associated with dysfunctional self-antigen-specific CD8. + The characteristics of T cells show a strong positive correlation (Schietinger et al., 2012), which is called the T cell non-responsiveness characteristic (Azizi et al., 2018). Figure 5 F). Therefore, PTI and CD8 + The presence of T cells is closely related to the CD8. + T cells are transcriptionally similar to cells that cannot produce an effective anti-tumor immune response. Furthermore, PTIs and CD8... + T cell dysfunction and exhaustion are positively correlated (Singer et al., 2016; Guo et al., 2018) Figure 12 D). Previous analysis of CD8 in tumors that responded well to ICB in MEL. +Studies of T cells have revealed single-cell, primary CD8+ cells associated with ICB response. + T cell characteristics are composed of genes involved in memory, survival, and activation (Sade-Feldman et al., 2018). Infiltrating PTIs were discovered. 低 CD8 of MEL and NSCLC + The T cell ratio comes from its PTI. 高 The expression of those counterparts exhibiting significantly higher CD8 ICB response characteristics. Therefore, PTI 高 CD8 in tumors + T cells are not similar to the primary / memory-like cells in MEL that promote ICB responses. Regarding tumor-derived CD8... + Analysis of T cell single-cell transcriptomes supported the inventors' bulk tumor discovery and established a strong link between the expression of pro-tumorigenic inflammatory mediators and suboptimal T cell transcriptional status.

[0236] Example 11: PTI trait predicts the lack of benefit from checkpoint inhibitors and is superior to other resistance traits.

[0237] To date, strong predictors of ICB resistance remain elusive. Given the strong contrasting prognostic ability of PTIs as biomarkers of T-cell inflammation (many of which are associated with ICB response), this study investigated the utility of PTIs as biomarkers of ICB resistance. Sixteen transcriptome datasets containing >1000 pre-treatment samples (n=1032) from patients treated with ICB were compiled. Figure 6 A, Figure 6 B). This pan-cancer ICB metadata dataset includes patients with five cancer types (ccRCC, mUC, MEL, GCA, and NSCLC) who were treated with anti-CTLA-4, anti-PD1, or anti-PD-L1 antibodies as monotherapy or in combination. 25% of the samples came from patients who responded to treatment (partial or complete response (PR / CR)), while 43% came from patients who did not benefit (progressive disease (PD)). Patients were stratified according to the median pre-treatment PTI level in each dataset. 低 Significant enrichment of PR / CR was found in patients ( (P = 0.004). Conversely, consistent with reported values ​​of T-cell inflammatory markers, low TIS (… p = 4 × 10 -7 CD8 + T cells ( p = 4 × 10 -6 ), IFNγ signal transduction ( p = 3 × 10 -6) characteristics or PD-L1 gene expression (CD274, Patients with p=0.003 were enriched in PD patients ( Figure 6 (C) (Tumeh et al., 2014; Ayers et al., 2017). Crucially, when combined with all these currently "hot" T-cell inflammatory tumor biomarkers (TIS, CD8, IFNγ, and / or CD274), PTI exhibits combined predictive power, making the CR / PR score more predictive of tumor repercussions. 高 TIS 低 The lowest level among patients, gradually increasing, in PTI 低 TIS 高 The highest among patients ( Figure 6 D). Similarly, in these ICB-treated patients, the combination of PTI and TIS also showed improved predictive value for OS (D). Figure 6 E). PTI 高 TIS 高 The patient's overall survival (OS) was significantly worse than that of PTI. 低 TIS 高 The median survival of patients differed by approximately one year. It is noteworthy that PTI... 高 TIS 高 and PTI 低 TIS 低 Patient survival was comparable, consistent with the inventors' earlier TCGA data analysis, indicating an antagonistic effect between PTIs and antitumor immune mediators. Figure 6 E). In summary, this analysis suggests that PTI characteristics can help differentiate between ICB responders and non-responders, including in patients with equivalent levels of intratumoral T-cell inflammation.

[0238] Next, the researchers attempted to benchmark the PTI signature against other reported signatures associated with ICB resistance. These included WNT / β-linkin signaling signatures previously shown to be associated with “cold” malnourished T-cell inflammatory tumors (Gajewski et al., 2019), gene signatures associated with immunosuppressive TGFβ signaling via fibroblasts (F-TBRS, Chakravarthy et al., 2018) and ECM remodeling (TGFβ-ECM, Mariathasan et al., 2018), EMT signatures associated with decreased response rates in T-cell inflammatory mUCs (stromal-EMT, Wang et al., 2019), and a signature of angiogenesis-related factors highly expressed during the inflammation-to-cancer (I2C) transition associated with melanoma ICB resistance (Chen et al., 2019). To evaluate the additional predictive value of each of these features in combination with T-cell inflammatory features, the TIS feature was used as the baseline predictive model, which is one of the most robust and widely validated gene expression biomarkers for ICB (Ayers et al., 2017; Cristescu et al., 2018). In the combined predictive model, each continuous feature score was combined with the TIS, where the area under the curve (AUC) was calculated as the model performance metric for each of the 16 ICB-treated patient cohorts. Notably, the PTI / TIS model had the highest mean AUC compared to other resistance feature models and TIS alone. Figure 13 A). Importantly, the mean AUC (0.79) of the PTI plus TIS model was significantly higher than that of TIS alone (mean AUC = 0.69, P = 0.02), highlighting the predictive utility of combining PTI and TIS values ​​in the pan-cancer ICB dataset. The researchers also compared the performance of all resistance features in predicting patient OS after ICB treatment (which does not necessarily follow response rates). Patients were stratified according to the median cutoff value for each feature and fitted to a multivariate CoxPH model adjusted for TIS status and the ICB dataset. 1000 randomly generated features of equal length to PTI were also included as controls. Consistent with objective response data, PTI is by far the most powerful resistance feature for predicting poor OS in ICB-treated patients. Figure 6 In summary, this data further demonstrates the unique and superior utility of combining PTI with immune biomarkers of “hot” tumors to predict patient response and survival after ICCB.

[0239] The currently claimed PTI feature has improved predictive ability compared to the COX-IS ratio disclosed in the inventor's earlier application WO 2019 / 243567. Figure 6 F shows the predictive power of the resistance trait in predicting poor overall survival in patients treated with ICB. The PTI trait (HR 1.38 [1.15-1.65], P=0.0004) is superior to the inventors' previous COX-IS ratio (HR 1.18 [0.96-1.45], P=0.12), thus illustrating the improved predictive power.

[0240] Next, the CPI1000 dataset (Litchfield et al., 2021) was examined. This dataset comprises a group of datasets that substantially overlap with the datasets in the inventors' aforementioned metadata set. These datasets are from patients treated with ICBs (anti-PD(L)1, anti-CTLA4) across multiple cancer types and contain whole-exome sequencing (WES) and RNA sequencing (RNAseq) data. This unique dataset underwent extensive harmonization of WES, RNAseq, and clinical data across several cohorts. First, the analysis focused on 523 patients from seven studies covering three cancer types (MEL, BLCA, ccRCC), annotating multiple genomic and transcriptomic tumor features previously associated with ICB response (Snyder et al., 2014, 2017; Van Allen et al., 2015; Hugo et al., 2017; Riaz et al., 2017; Mariathasan et al., 2018; Mcdermott et al., 2018). This analysis also included features associated with APOBEC (Spearman = 0.26, P = 9 × 10⁻⁶). -7 TP53 mutation status (Spearman = 0.22, P = 0.0004) and CD274 expression (Spearman = 0.22, P = 3 × 10⁻⁶). -7 Apart from a moderate positive correlation, PTI levels are not strongly correlated with any single parameter. Figure 13 (B, 13C), which supports the view that PTIs capture a unique and independent feature of tumor inflammation, and their potential use as an independent immune biomarker.

[0241] To evaluate the predictive power of PTI in this coordinated dataset, benchmark comparisons were performed against multiple features and parameters of ICB resistance (e.g., F-TBRS and matrix-EMT (Mariathasan et al., 2018; Wang et al., 2019) or multiple features and parameters of response (e.g., IMPRES algorithm, TIS score, CXCL13 expression, TLS features, CYT score, and different TMB measures) (Rizvi et al., 2015; Rooney et al., 2015; McGranahan et al., 2016; Ayers et al., 2017; Auslander et al., 2018; Mariathasan et al., 2018; Cabrita et al., 2020; Litchfield, Reading, et al., 2020). Notably, PTI levels in pre-treatment samples from PD patients were higher than in PR / CR patients (n=552, Figure 6 G). Next, the inventors fitted univariate logistic regression models for PTI and other biomarkers on seven independent datasets constituting a broader pan-cancer cohort (VanAllen et al., 2015; Hugo et al., 2017; Riaz et al., 2017; Snyder et al., 2017; Mariathasan et al., 2018; Mcdermott et al., 2018; Liu et al., 2019). To assess the consistency of each putative biomarker, a meta-analysis was performed to calculate the overall effect size. PTI was the best predictor of ICB resistance (odds ratio = 0.67, P = 0.03). Figure 6 H). PTI in the entire CPI1000 cohort (including CR / PR, PD, and stable disease (SD) patients). 高 Patients with poor overall survival (OS) were significantly worse than those with moderate and low PTI levels (HR = 1.41 [1.1–1.8], P = 0.007). Figure 13 D). Furthermore, after adjusting for cancer type and prior exposure to anti-CTLA-4 therapy, PTI remained a significant prognostic factor (HR = 1.29 [1.006–1.65], P = 0.04), further demonstrating the independent predictive value of PTI signature as an immune biomarker.

[0242] In the CPI1000 dataset, the model combining clonal TMB with CXCL9 expression was shown to be the best model for predicting patient response to ICB (Litchfield et al., 2021). By fitting predictive models covering cancer type, clonal TMB, CXCL9, PTI, and their combinations, we tested whether PTI levels significantly increased the predictive utility of the clonal TMB and CXCL9 combination. Notably, PTI significantly increased predictive value beyond cancer type, clonal TMB, and CXCL9. Figure 13 E, likelihood ratio test, P = 0.01). The clone TMB containing the highest proportion of ICB responders... hi or CXCL9 hi or TMB hi CXCL9 hi Patients (Litchifield et al) were further stratified according to PTI characteristics. Figure 6 I, Figure 13 F, Figure 13 (G). These findings further demonstrate the value of measuring PTIs in pre-treatment samples in conjunction with other immune biomarkers. Furthermore, this data positions PTIs as a useful tool for identifying patients who should receive more aggressive combination therapy (e.g., using COX-2 inhibitors or other drugs that inhibit PTIs) to maximize their benefit from ICBs.

[0243] Example 12: Using tumor type-independent and tumor type-specific methods to predict binary patient response outcomes model

[0244] The logistic regression model was trained on the CPI1000 dataset. First, TPM values ​​for the PTI gene were obtained from the CPI10000 dataset. A pseudo-count of 1 was added before normalizing these values ​​to log2. Then, the data was either scaled across all samples in the dataset (z-score normalization) (tumor type independent) or b) grouped by cancer type (metadata feature: "Tum_Type") and the expression values ​​were scaled by tumor type (tumor type specific). The mean of the PTI gene expression data was then determined to calculate the PTI score. Patients who achieved disease stabilization as the best RECIST response (metadata feature: Response_RECIST) were removed along with "under treatment" samples (metadata feature: Prior_IO).

[0245] For both tumor type-independent and tumor type-specific approaches, three different logistic regression models (discussed in more detail below) were trained to predict binary patient response outcomes (metadata feature: Response_Responder).

[0246] The three logistic regression models are shown in Table 3. The results for logistic regression models A through C are shown in Table 4. For the tumor-specific model A, the results were extracted according to Table 5. The coefficients of clone_TMB and CXCL9 were calculated. The area under the curve (AUC) was used as an indicator of model performance; a larger AUC indicates better predictive utility. The receiver operating characteristic curve of this logistic regression model A (tumor type specific) is shown in... Figure 15 As shown in the image.

[0247] Model C, which combines [the following] in the absence of TMB or CXCL9. The expressions demonstrate the prediction of binary patient response outcomes for both tumor type-independent and tumor type-specific methods, as shown in Table 4 below.

[0248] Table 3: Summary of logistic regression models A, B, and C for both tumor type-independent and tumor type-specific methods for predicting binary patient response outcomes.

[0249] Table 4: Results of logistic regression models A, B, and C for tumor type-independent and tumor type-specific methods.

[0250]

[0251] Table 5: Coefficients extracted from logistic regression model A (tumor type specific).

[0252] Example 13: Clinical value of PTI characteristics as an immunobiomarker for ICB outcome

[0253] Finally, the predictive utility of PTI features and its potential clinical value as an immunobiomarker for analyzing formalin-fixed paraffin-embedded (FFPE) tissues from patients treated with immunotherapy were evaluated. ccRCC was considered, a type of cancer where ICB has become a mainstream treatment but current immunobiomarkers fail to predict outcomes. As a proof-of-concept, archived FFPE samples were obtained from 42 patients (91% ccRCC) who had received treatment for advanced disease with high-dose IL-2 immunotherapy (n=23) or ICB as monotherapy (n=11) or in combination with anti-CTLA-4 (n=4) or anti-VEGF (n=1). Overall survival data for these patients were matched ( Figure 14 A). This diverse real-world clinical consortium includes nephrectomy samples, as well as samples from surgical procedures involving the removal of brain and lung metastases (A). Figure 14B). RNA was extracted and multiplexed using a pan-cancer immunoprofiling dataset containing 800 immune-related genes (including most PTI signature genes (12 / 16)) on the Nanostring nCounter platform. In over 50% of the samples, 83% of the present PTI genes were expressed above the detection limit and were generally correlated with each other (mean Spearman = 0.3). Figure 7 A). Therefore, regardless of the quantification method, PTI represents a conserved TME transcriptional program. PTI characteristics are associated with the presence of different myeloid cell subsets (neutrophils, dendritic cells, macrophages) and the characteristics of the stroma and complement. Figure 7 B). Therefore, consistent with earlier bulk RNAseq findings, PTI signatures are associated with the presence of multiple immune populations and the molecular characterization of disease. Crucially, PTI levels are correlated with CD8... + T cell signatures (CD8A, CD8B, FLT3LG, GZMM, PRF1) and IFNγ signaling signatures (CXCL9, CXCL10, IDO1, STAT1, IFNG, HLA-DRA) are negatively correlated. Figure 7 C).

[0254] Of the 42 RCC tumor biopsies, 25 were pre-treatment samples from patients receiving IL-2 immunotherapy, anti-PD(L)1, or anti-PD1 / anti-CTLA-4 blockade. Patients were stratified based on PTI levels in these samples to identify those with significantly different survival outcomes. 高 The patient had a significantly shortened overall survival (OS). Figure 7 D, Figure 7 E). Crucially, in this patient cohort, the immune characteristics of "hot" T-cell inflammatory tumors did not show a significant association with patient survival (E). Figure 7 F). Previously proposed “myeloid cell characteristics” for predicting ICB outcomes in ccRCC patients (Mcdermott et al., 2018) also failed to show significant correlation. These unexpected findings, obtained using clinically compatible assays in a small and heterogeneous cohort of real-world ccRCC patients treated with immunotherapy, further highlight the potential of monitoring PTIs for predicting ICB outcomes and guiding patient treatment.

[0255] Example 14: PTI as a predictor of outcomes in ccRCC patients following immunotherapy

[0256] Finally, to further validate the specific utility of PTIs in predicting patient outcomes following immunotherapy in ccRCC, transcriptomic data from Checkmate trials 009, 010, and 025 were analyzed. In all these trials, nivolumab (an anti-PD-1 blocker) demonstrated clinical benefit in previously treated advanced ccRCC patients, including in the phase III study (Checkmate 025) compared to the mTOR inhibitor everolimus (Braun et al., 2020). Notably, patient stratification based on pre-treatment PTI expression levels was negatively correlated with survival in patients treated with nivolumab (HR = 1.46 [1.02–2.08], P = 0.04), but not in patients treated with everolimus (HR = 1.03 [0.71–1.50], P = 0.87). Figure 7 G). To test the predictive ability of PTI in patients treated with ICB as first-line therapy, data from the phase II IMMotion150 study were analyzed. In this study, previously untreated patients with advanced ccRCC were placed in one of three groups to receive atezolizumab (anti-PD-L1) and bevacizumab (anti-VEGF), atezolizumab monotherapy, or sunitinib (Mcdermott et al., 2018). Consistent with the inventors' analysis in patients treated with second-line ICB, non-responder patients had higher baseline PTI levels compared to responders after first-line atezolizumab or atezolizumab plus bevacizumab combination therapy. Figure 7 H). In contrast, there was no difference in baseline PTI expression between responders and non-responders to sunitinib (a multi-target tyrosine kinase inhibitor). Figure 7 H). In summary, these findings suggest that PTIs selectively identify ccRCC patients resistant to current immunotherapy regimens and demonstrate that their predictive utility for ICBs outweighs their prognostic value.

[0257] Example 15: Minimum number of PTI genes required to calculate scores for predicting treatment response and / or prognosis

[0258] Elastic regression is a statistical model that helps reduce the feature space and select the most predictive features in classification tasks such as predicting ICB responses. To illustrate this, we first train an elastic regression model using the expression levels of the PTI gene, along with clones TMB and CXCL9, as input features. This is achieved using a normalized expression matrix. In the first of these (Model D), the expression levels are normalized for each tumor type (tumor type specific), and for the second (Model E), the expression levels are normalized in a tumor-independent manner (tumor type independent). Additionally, in a tumor-independent manner (Model F), without clones TMB and CXCL9 as covariates, the elastic regression model is trained on members of the PTI gene set. The elastic regression model is trained with different α hyperparameter values ​​between 0 (ridge regression) and 1 (lasso regression). The key difference is that increasing α shrinks the coefficients of less important features towards 0, thus allowing for more stringent selection of important features.

[0259] method

[0260] First, the tumor type-independent (model E) and tumor type-specific (model D) datasets were divided into training (70%) and test (30%) sets. A resilient network classifier model was trained to predict responses (CR / PR) using 10-fold cross-validation and AUC as the performance metric. The model was trained with different α hyperparameter values ​​between 0 and 1, in increments of 0.1. The λ value corresponding to the best training performance was selected as the optimal model for each training iteration. Then, each optimized model was used to predict patient responses to ICB on the test dataset, and the AUC was calculated. The AUC values ​​from the training and test sets were then plotted. Figure 16 A to B correspond to models D and E, respectively. An α value of 0 indicates that all available features were used (all 16 PTI genes, clone _TMB, and CXCL9).

[0261] Even without clones _TMB and CXCL9 as covariates, a resilient network regression model (model F) was trained on members of the PTI gene set. Figure 16 C).

[0262] result

[0263] Figure 16Figures A and 16B show the test AUC and training AUC for tumor type-specific (Model D) and tumor type-independent (Model E) models, respectively, where AUC is calculated using clonal TMB and gene expression levels of all 16 PTI genes and CXCL9. Points represent models trained with different α hyperparameter values ​​between 0 and 1. The corresponding α value is shown next to each point. Generally, higher α values ​​are closer to lasso regression, and the fewer features the model uses. Figure 16 It is clear that models trained with an α value greater than 0 tend to have improved training and testing performance. Notably, many models show a reduction in the number of important features compared to the number of features used in training. Important features are defined as any feature that has a non-zero regression coefficient after regularization is applied.

[0264] Table 6 illustrates the key features and coefficients of the optimal tumor type-independent dataset model (Model E) with an α value of 0.3. This optimal model E has a reduced feature set. Four genes (EREG, HAS2, CXCL8, and VCAN) from the PTI features, along with clones _TMB and CXCL9, were identified as key features.

[0265] Table 6: Key features and corresponding coefficients of the optimal model E (tumor type independent) with α hyperparameter of 0.3.

[0266] Figure 16 C illustrates the training and testing AUC values ​​from a tumor type-independent model (model F) featuring only PTI genes as features (excluding CXCL9 and clone_TMB as covariates). The value of α is shown next to the corresponding point (training model). The optimal model has an α hyperparameter of 0.2 and seven important features (AREG, CXCL1, CXCL2, EREG, HAS2, IL1A, and PTGS2), corresponding to seven of the 16 PTI genes. The coefficients for each of these seven features in the optimal model F (α=0.2) are illustrated in Table 7.

[0267]

[0268] Table 7: Key features of the optimal model F with α hyperparameter of 0.2 and its corresponding coefficients.

[0269] Although the elastic regression model trained on only PTI gene set members (in the absence of clones _TMB and CXCL9) did not perform as well as those trained on PTI genes in the presence of clones _TMB and CXCL9, it still explained a smaller proportion of the variance in the patient's ICB response.

[0270] discuss

[0271] Determining the likelihood of a patient benefiting from ICB is a significant challenge in this field (Hegde and Chen, 2020). The inventors inferred that the biological mechanisms beyond current biomarkers must support the differences in patient outcomes following ICB. Subsequently, the inventors showed that pro-tumorigenic inflammatory ligands centered on COX-2 typically control such differences, including even in patients with high clonal TMB and CXCL9 expression (who are more likely to benefit from ICB) (Litchfield et al., 2021). These findings suggest a need for multi-combination biomarkers, requiring both WES and gene expression assays, crucially including PTI genes in addition to those measuring T-cell inflammation. A key consideration in the development of gene-based biomarkers is intratumoral heterogeneity (ITH). The extent to which sampling bias might confound PTI-based patient classifiers remains unclear. However, one way to circumvent these problems is to select PTI genes that have a lower overall ITH and are less susceptible to sampling bias, as previously done in NSCLC for prognostic characteristics (Biswas et al., 2019).

[0272] Analysis of cell line drug screening data using publicly available methods (Jerby-Arnon et al., 2018) (Garnett et al., 2012) showed that PTI 高 Cell lines are sensitive to MEK inhibitors. However, this dataset does not include next-generation inhibitors of previously challenging targets such as KRAS (Canon et al., 2019). Furthermore, identifying core targetable nodes disrupting the PTI signaling network will require in vivo experiments with a full complement of interacting cell types. Nevertheless, this work suggests that the COX-2 / PGE2 axis is a strong candidate (Pelly et al., 2021). Numerous studies have reported the favorable immunomodulatory properties of MEK inhibitors (Hu-Lieskovan et al., 2015; Ebert et al., 2016; Bommareddy et al., 2018; Baumann et al., 2020). While further work is needed to test whether MEK inhibitors specifically inhibit the PTI transcriptional program in vivo, the current data further support the view that they have a broader range of applications than just as cell inhibitors (Templeton and Musib, 2015; Hamarsheh et al., 2020).

[0273] Using batch tumor RNAseq, immune deconvolution algorithms, and scRNAseq data (Becht et al., 2016; Puram et al., 2017; Jerby-Arnon et al., 2018; Lambrechts et al., 2018), it was shown that in addition to cancer cells themselves, various tumor cell types also express the PTI gene and infiltrate PTI. 高 Tumors, particularly neutrophils, have shown a high correlation with PTI levels and, consistent with working models, have been shown to be elevated in KRAS-mutant environments (Hamarsheh et al., 2020; Polidoro et al., 2020). However, it remains unclear whether these cells play an active role in maintaining pro-cancer inflammation. Furthermore, it is unclear whether the production of inflammatory mediators inherent in cancer cells propagates PTI signatures to neighboring non-cancer cells, or vice versa. Future mechanistic studies utilizing spatially resolved transcriptomic data from human tumors should help address these questions.

[0274] The potential basis for the association between PTI ligands and ICB resistance was investigated. PTI factors have multiple roles in TME and contribute to all cancer markers.

[0275] This study revealed a profound interaction between PTIs and T cell infiltration and IFNγ signaling. In fact, directly regulating the activation state of intratumoral lymphocytes may contribute to the ICB resistance phenotype. (The text also mentions CD8 activation, but this seems unrelated to the previous sentence and is likely a separate, incomplete thought.) + Patients with T-cell inflammatory tumors or high levels of IFNγ signaling have different outcomes depending on PTI levels; this trend has also been observed in patients with high plasma levels of the pro-tumorigenic chemokine IL-8 (Yuen et al., 2020). The COX-2 / PGE2 axis (a central node in the PTI phenotype) has previously been associated with T-cell exhaustion (Harris et al., 2002; Chen et al., 2015). Using CD8 ICB response characteristics (Sade-Feldman et al., 2018), and CD8... + T cell dysfunction features (Azizi et al., 2018; Guo et al., 2018) indicate PTI 高 CD8 in tumors + T cells are in a suboptimal activated state.

[0276] In summary, a COX-2-centric inflammatory gene signature of 16 genes was defined, which is associated with molecularly distinct T-cell infiltration and adverse outcomes, including those following treatment with immune checkpoint inhibitors.

[0277] References

[0278] The foregoing has cited numerous publications to more fully describe and disclose the invention and the prior art to which it pertains. Full citations of these references are provided below. The entire contents of each of these references are incorporated herein by reference.

[0279] For standard molecular biology techniques, see Sambrook, J., Russell, DW Molecular Cloning, A Laboratory Manual. 3 ed. 2001, Cold Spring Harbor, New York: Cold Spring Harbor Laboratory Press.

Claims

1. A computer-implemented method for predicting the therapeutic response to anticancer immunotherapy in mammalian cancer patients, the method comprising: a) Provides multiple features comprising gene expression levels of at least 8, at least 9, at least 10, at least 11, at least 12 or more (e.g., all) of the following cancer-promoting genes previously measured in samples obtained from tumors of mammalian cancer patients: PTGS2, CXCL1, CXCL2, CXCL5, CXCL6, IL1A, HAS2, IL11, IL1B, EREG, IL6, CXCL8 (IL8), MMP12, NRG1, VCAN, AREG; b) Calculate the score using the features provided in a); and c) Predict the treatment response of the mammalian cancer patient based on the score calculated in step b).

2. The computer-implemented method of claim 1, wherein the feature further comprises cloning_TMB and CXCL9.

3. The computer-implemented method of claim 1 or claim 2, wherein the feature comprises all of the following gene expression levels: PTGS2, CXCL1, CXCL2, CXCL5, CXCL6, IL1A, HAS2, IL11, IL1B, EREG, IL6, CXCL8 (IL8), MMP12, NRG1, VCAN, AREG.

4. The computer-implemented method according to any one of the preceding claims, wherein the gene expression level is a normalized gene expression level.

5. The computer-implemented method of any of the preceding claims, wherein the score is calculated by: normalizing the individual gene expression values ​​of the feature using z-scores and calculating the average of the normalized measurements.

6. The computer-implemented method according to any one of the preceding claims, wherein the score is calculated according to the following formula: Rating = Where n p It is the number of gene expression characteristics. Genes are positively correlated, within the unit interval (i), and (e) represents gene expression value, expressed as log2 per million counts (CPM); optionally... in (e) is the average expression per million counts (per million kb readings (FPKM)) of positive gene log2 transformations.

7. The method of any one of claims 1 to 5, wherein the score is a probability score calculated using the following formula: Probability Score Where β0 is the intercept weight, β is the weight vector for each k variable, and x is the sample-related variable vector, where the variables include the multiple features.

8. The method of any one of the preceding claims, wherein the computer-implemented method further comprises assessing the tumor burden and / or neoantigen prevalence of the cancer patient.

9. The method of any one of the preceding claims, wherein predicting the treatment response of the mammalian cancer patient comprises: The calculated score from step b) is compared with the reference median score from the sample group of respondents and the reference median score from the sample group of non-responders, wherein the sample groups of respondents and non-responders are i) cancer patients with the same type of cancer as the said cancer patient or ii) cancer patients with a tumor type unrelated to the said cancer patient.

10. The method of any of the preceding claims, wherein the score calculated in step b) indicates that cancer is predicted in response to anticancer immunotherapy, and the method further comprises selecting mammalian cancer patients for anticancer immunotherapy.

11. The method of any one of the preceding claims, wherein the gene expression levels of the at least eight cancer-promoting genes are obtained by measuring the gene expression levels of at least eight, at least nine, at least ten, at least eleven, at least twelve or more (e.g., all) of the following cancer-promoting genes in a sample obtained from a tumor of a mammalian cancer patient: PTGS2, CXCL1, CXCL2, CXCL5, CXCL6, IL1A, HAS2, IL11, IL1B, EREG, IL6, CXCL8 (IL8), MMP12, NRG1, VCAN, AREG.

12. A method for predicting the therapeutic response to anticancer immunotherapy in mammalian cancer patients, the method comprising: a) Measure multiple characteristics, including measuring the expression of at least 8, at least 9, at least 10, at least 11, at least 12 or more (e.g., all) of the following cancer-promoting genes in samples obtained from the patient's tumor: PTGS2, CXCL1, CXCL2, CXCL5, CXCL6, IL1A, HAS2, IL11, IL1B, EREG, IL6, CXCL8 (IL8), MMP12, NRG1, VCAN, AREG; b) Use the features measured in a) to determine the score; and c) Predict the patient's treatment response based on the score determined in step b).

13. The method of claim 12, wherein the feature further comprises clone_TMB and CXCL9.

14. The method of any one of the preceding claims, wherein determining the score to predict the subject's treatment response comprises: Data representing the gene expression levels of at least eight cancer-promoting genes containing the object are input into a machine learning classifier, wherein the machine learning classifier has been trained with a training dataset containing at least the gene expression levels of the same cancer-promoting genes as those forming the score of the object, and wherein the training dataset is derived from at least two sample groups. One cohort contains multiple samples from multiple mammalian cancer patients (responders) who are known to respond to anticancer immunotherapy, and the other cohort contains multiple samples from multiple mammalian cancer patients (non-responders) who are known not to respond to anticancer immunotherapy. as well as The machine learning classifier categorizes the object's score into either the response group or the non-response group.

15. The method of claim 14, wherein the training dataset comprises gene expression levels of at least eight cancer-promoting genes from at least 50, at least 100, at least 150, at least 200, at least 250, and at least 300 samples from known mammalian cancer patients (responders) who respond to anticancer immunotherapy, and gene expression levels of at least eight cancer-promoting genes from at least 50, at least 100, at least 150, at least 200, at least 250, at least 300, at least 350, at least 400, at least 450, at least 500, and at least 550 samples from known mammalian cancer patients (non-responders) who do not respond to anticancer immunotherapy.

16. The method of claims 14 and 15, wherein the machine learning classifier has been trained with multiple scores derived from at least three sample groups. One cohort contains multiple samples from multiple mammalian cancer patients known to have fully responded to anticancer immunotherapy (full responders), one cohort contains multiple samples from multiple mammalian cancer patients known to have partially responded to anticancer immunotherapy (partial responders), and another cohort contains multiple samples from multiple mammalian cancer patients known to have not responded to anticancer immunotherapy (non-responders).

17. The method of claims 14 to 16, wherein the machine learning classifier is a logistic regression model, an elastic network regression model, or a tree-based model, such as a random forest classifier.

18. A method for classifying samples obtained from mammalian cancer subjects, said samples containing nucleic acids and / or proteins. The method includes: a) Analyze the samples to obtain gene expression levels of at least eight cancer-promoting genes selected from the following: PTGS2, CXCL1, CXCL2, CXCL5, CXCL6, IL1A, HAS2, IL11, IL1B, EREG, IL6, CXCL8 (IL8), MMP12, NRG1, VCAN, and AREG; and b) Using the gene expression level obtained in step a) to carry out the method of any one of claims 1 to 17.

19. The method of claim 18, wherein the sample is derived from one or more cancer cells of the object.

20. A method for treating cancer in mammalian cancer patients, the method comprising: a) To predict the treatment response to anticancer immunotherapy in the mammalian cancer patient by implementing the method of any one of the preceding claims; b) Determine that the mammalian cancer patient is predicted to respond to the anticancer immunotherapy in step a); as well as c) Administer a therapeutically effective dose of the anticancer immunotherapy to the recipient.

21. A method for treating cancer in mammalian cancer patients, the method comprising: a) Implementing the method according to any one of claims 1 to 20; b) Determine the score of the mammalian cancer patient to indicate that the mammalian cancer patient is predicted to respond to anticancer immunotherapy; and c) Administer immunotherapy to mammalian cancer patients who require it.

22. The method of any one of the preceding claims, wherein the method further comprises analyzing one or more additional biomarkers of anticancer immunotherapy response derived from the mammalian cancer patient, thereby supplementing and / or confirming the prediction of antitherapy response or non-response.

23. The method according to any one of the preceding claims, wherein the sample obtained from the patient's tumor is a formalin-fixed paraffin-embedded (FFPE) sample.

24. A system for predicting the therapeutic response to anticancer immunotherapy in mammalian cancer patients, the system comprising: At least one processor; and At least one non-transitory computer-readable medium containing instructions that, when executed by the at least one processor, cause the at least one processor to perform the method of any one of claims 1 to 17.

25. One or more computer-readable media comprising instructions that, when executed by one or more processors, cause the one or more processors to perform the steps of the method of any one of claims 1 to 17.

26. A pharmaceutical composition for treating a mammalian cancer patient, the pharmaceutical composition comprising anticancer immunotherapy, wherein the patient has been predicted to respond to the anticancer immunotherapy by the method of any one of claims 1 to 19.

27. The method, system, and / or pharmaceutical composition according to any one of the preceding claims, wherein the mammalian cancer patient suffers from the following and / or the cancer is: kidney cancer, such as renal cell carcinoma or renal cell carcinoma of the kidney (KIRC); gastroesophageal tumors; esophageal cancer (ECA); melanoma; bladder cancer (BLCA); gastric cancer; lung cancer, such as lung adenocarcinoma (LUAD), lung squamous cell carcinoma (LUSC), or non-small cell lung cancer (NSCLC); pancreatic cancer (PAAD); liver cancer (LIHC); glioblastoma (GBM); adrenocortical carcinoma (ACC); cervical cancer (CESC); head and neck cancer, such as head and neck squamous cell carcinoma (HNSC); thyroid cancer (THCA); uterine cancer (UCEC); giant cell tumor of the tendon sheath (TGCT); melanoma; BRCA; sarcoma and pan-carcinoma.

28. The method, system, and / or pharmaceutical composition of claim 27, wherein the renal cell carcinoma is clear cell renal cell carcinoma (ccRCC).

29. The method, system and / or pharmaceutical composition according to any one of the preceding claims, wherein the immune subtype of said cancer is one of the following: wound healing, IFNγ dominant, TGFβ dominant, inflammatory and / or lymphocyte depleted subtypes.

30. The method, system, and / or pharmaceutical composition according to any one of the preceding claims, wherein the anticancer immunotherapy comprises IL-2 immunotherapy alone or in combination with antiVEGF therapy or immune checkpoint blockade therapy.

31. The method, system, and / or pharmaceutical composition of claim 30, wherein the immune checkpoint blockade therapy comprises programmed death-1 (PD-1) blockade, programmed death-ligand 1 (PD-L1) blockade, and / or cytotoxic T-lymphocyte-associated protein 4 (CTLA-4) blockade.

32. The method, system, and / or pharmaceutical composition of claim 30 or claim 31, wherein the immune checkpoint blockade therapy comprises treatment with nivolumab, pembrolizumab, atezolizumab, and / or ipilimumab.

33. The method, system, and / or pharmaceutical composition of claim 30, wherein the anti-VEGF treatment is bevacizumab.

Images