Predictive genomic biomarker for combination therapy

Abstract

A genomic biomarker comprising a set of 5-50 genes is provided for predicting responsiveness to an anti-cancer treatment that includes a combination of palbociclib and sunitinib. These genes are differentially expressed in tumors that respond to the combination therapy compared to those that do not. At least some of the genes in the biomarker are significantly overexpressed in non-responsive tumors relative to responsive ones. This genomic biomarker helps determine whether a subject diagnosed with a cancerous disease will benefit from the combination therapy prior to its administration, enabling the optimization of medical treatment and supporting precision medicine.

Description

[0001] PREDICTIVE GENOMIC BIOMARKER FOR COMBINATION THERAPY

[0002] FIELD OF THE INVENTION

[0003] The present disclosure relates to a predictive genomic biomarker for prognosticating the efficacy of combination therapy of cancer, particularly the efficacy of a combined anti-cancer treatment comprising palbociclib and sunitinib.

[0004] BACKGROUND

[0005] Precision medicine aims to optimize medical care by tailoring treatments to individual patients based on their genetic and molecular profiles. A key element of this approach is the development of predictive biomarkers, which are assessed before treatment to indicate a patient's likelihood of responding to a specific therapy. Predictive biomarkers enable personalized treatment plans, improving patient outcomes and satisfaction while potentially increasing the adoption and market share of new drugs.

[0006] By identifying patients most likely to benefit, predictive biomarkers enhance drug efficacy, improve clinical trial success rates, and streamline the trial process. This can reduce the number of participants needed, lower costs, and accelerate drug development.

[0007] Combination treatment with Palbociclib and sunitinib is disclosed in WO 2020136642 and Moskovits et al., 2022 (Moskovits et al., Cancer Letters 536 (2022) 215665: 1-11). A clinical trial assessing this combination is ongoing in Rabin Medical Center, Israel.

[0008] SUMMARY

[0009] The present disclosure addresses problems such as uncertainty and lack of means to determine a priori the efficacy of combining palbociclib (a CDK4 / 6 inhibitor) and sunitinib (a multi-targeted receptor tyrosine kinase inhibitor) in a combination therapy of various cancers. The need for a responsiveness predicting biomarker to a palbociclib + sunitinib combination therapy (herein interchangeably referred to as "combo therapy" or simply "combo") is met, according to the present disclosure, by the provision of a putative genetic predisposition signature comprising a set of 5-50 genes, which are differentially expressed in tumors that respond to the combined therapy versus tumors that are unresponsive to the treatment. At least some of the genes in this set or panel were significantly overexpressed in non-responsive tumors compared to responsive tumors. Importantly, the disclosed gene signature of the non- responsive tumors comprised genes that were not differentially expressed in tumors subjected to monotherapies with either palbociclib or sunitinib. Moreover, the disclosed gene signature of the non-responsive tumors comprised genes that could not have been derived or assumed from tumors known to be either responsive or non-responsive to monotherapies involving palbociclib or sunitinib. Furthermore, highly relevant genes such as CIITA, CCDC88B, CTA-126B4.7, ETV7, FERMT3, FGD2, FGR, HLA-DRA, RASGRP3, SP140 and TMEM140 in the disclosed genomic biomarker were not outstanding as responsiveness-related genes in tumors treated with combinations of palbociclib with other drugs, including other multi-targeted receptor tyrosine kinase inhibitor, nor in tumors treated by combining sunitinib with other drugs, including other CDK4 / 6 inhibitors.

[0010] In one aspect, the present disclosure relates to a genomic biomarker comprising a set of 5-50 genes for predicting responsiveness to an anti-cancer treatment with a combination of palbociclib and sunitinib. These genes are differentially expressed in a cancer responsive to a combination therapy of palbociclib and sunitinib versus a combination therapy unresponsive cancer. For example, the genomic biomarker genes are at last 2-fold overexpressed in a cancer which is not responsive to the combination therapy compared to a responsive cancer. The genes were identified based on a differential expression analysis of at least 5 subjects determined to be responsive, and at least 5 subjects determined to be non-responsive to the combo therapy and had only up to 10% overlap in values between combination therapy responders and non-responders across all cancerous samples analyzed.

[0011] A disclosed genomic biomarker may comprise, for example, 5-30 genes, at least 5 of which are selected from the genes SP140, FGR, HLA-DRA, LINC00528, GPR132, ABCA6, SELPLG, LINC00494, RASGRP3, CIITA, NCF4, PTPN22, FGD2, FERMT3, JAK3, POU2F2, IFI30, LAMP3, TMEM140, PSMB9, ETV7, CTA-126B4.7, TMC8, CCDC88B, PSMB8-AS1, TFEB, PSME2, or AC009133.17. The genomic biomarker may further comprise the transcript RP11-134L10.1.

[0012] For example, the genomic biomarker may comprise at least 5 genes selected from CIITA, TFEB, CCDC88B, CTA-126B4.7, ETV7, FERMT3, FGD2, FGR, HLA-DRA, RASGRP3, SP140, LINC00528, GPR132, IFI30, POU2F2 and TMEM140, and the transcript RP11-134L10.1.

[0013] The genomic biomarker is useful, in accordance with the present disclosure, for determining whether a subject diagnosed with a cancerous disease will benefit from a combination therapy with palbociclib and sunitinib nor not, before administering the therapy. Thus, in another aspect, the present disclosure relates to a method for predicting responsiveness to a combination anti-cancer therapy comprising palbociclib and sunitinib, the method comprising the steps:

[0014] (i) obtaining a sample of cancerous tissue from the subject;

[0015] (ii) subjecting the sample to gene expression profiling;

[0016] (iii) evaluating the gene expression level of each gene in a predetermined set of 5-50 genes;

[0017] (iv) comparing the expression level of each gene in the predetermined set of genes to a corresponding first reference value and / or to a corresponding second reference value; and

[0018] (v) determining that the subject is unlikely to respond to the combination therapy if the expression levels of at least some of the genes are higher by at least 2-fold compared to their corresponding first reference values; or

[0019] (vi) determining that the subject is likely to respond to the combination therapy if at least 5 of the genes are not expressed and / or the expression levels of at least some of the genes are lower by at least 2-fold compared to their corresponding second reference values. A corresponding first reference value represents the score of the expression levels of a predetermined gene across 5-10 different cancerous tissues from subjects identified as responsive to the anti-cancer therapy, and a corresponding second reference value represents the score of the of the expression levels of the same gene across 5-10 different cancerous tissues from subjects identified as non-responsive to the anti-cancer therapy.

[0020] In yet a further aspect, the present disclosure relates to a method for treating cancer in a subject intended for a combination therapy comprising palbociclib and sunitinib, the method comprising:

[0021] (i) obtaining a sample of cancerous tissue from the subject;

[0022] (ii) subjecting the sample to gene expression profiling;

[0023] (iii) evaluating the gene expression level of each gene in a predetermined set of 5-50 genes;

[0024] (iv) comparing the expression level of each gene in the set to a corresponding reference value; and

[0025] (v) administering to the subject a therapeutically effective amount of the combination therapy provided that least 5 of the predetermined genes are not expressed and / or the expression levels of at least some of the predetermined genes are lower by at least 2-fold compared to their corresponding reference values, wherein a corresponding reference value is a score of the expression levels of a predetermined gene across 5-10 different cancerous tissues from subjects identified as non-responsive to the anti-cancer therapy.

[0026] In some embodiments, the predetermined set of 5-50 genes is the genomic biomarker disclosed herein.

[0027] Cancer patients which may benefit from a disclosed prognosis and / or anti-cancer combination therapy include patients diagnosed with gastric cancer, colon cancer, cholangiocarcinoma, pancreas cancer, ovarian cancer, breast cancer, lung cancer, Ewing sarcoma, sarcoma of the uterus, neuroendocrine cancer, melanoma, hepatocellular carcinoma, adenoid cystic carcinoma, squamous cell carcinoma or carcinosarcoma.

[0028] BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Some embodiments are herein described, by way of example only, with reference to the accompanying drawings. With specific reference now to the drawings in detail, it is stressed that the particulars shown are by way of example and for purposes of illustrative discussion of embodiments. In this regard, the description taken with the drawings makes apparent to those skilled in the art how embodiments may be practiced.

[0030] In the drawings:

[0031] Figs. 1A-1B are schematic presentations of 196 differentially expressed genes (DEGs) evaluated by differential expression analysis (DEA) of a dataset generated by RNA sequencing (using RNA-Seq technique) of cancerous tissues of patient-derived xenografts (PDX) models that responded to a combination treatment comprising palbociclib and sunitinib (Responder) versus PDX models which did not respond to the combination treatment (Non-responder). 1A: a heatmap presenting the color-coded expression level of each of the 196 DEGs observed for responder models (recited in green) and non-responder models (recited in red). IB: a collection of 196 boxplots superimposed on principal component analysis (PCA) plots, each of which accounts for one differentially expressed gene. Each dot represents a PDX model. The dark (blue) dots are of the RA-441 model. In the boxplot, the lower and upper sides (faces) of the box represent the lower and upper quartiles of the data, respectively, and the line across the middle of the box is the overall median (the middle observation of a set of data). The lower quartile marks the median of the lower half of the observations, and the upper quartile is the median of the upper half. The lines protruding from the box are the "whiskers", which extend only to the last data point within the limit of 1.5 time the length of the box;

[0032] Figs. 2A-2B are schematic presentations of the expression levels of 29 differentially expressed genes (DEGs) constituting a genomic biomarker for predicting responsiveness to a combination therapy comprising palbociclib and sunitinib. Expression levels are presented for PDX models that responded to the combination therapy (Responder) versus PDX models which did not respond to the combination therapy (Non-responder). 2A: a heatmap presenting the color-coded expression level of each of the 29 DEGs observed for Responder models (recited in green) and non-responder models (recited in red). 2B: a collection of 29 boxplots superimposed on principal component analysis (PCA) plots, each of which accounts for one differentially expressed gene of the genomic biomarker. Each dot represents a PDX model. The dark dots are of the RA-441 model;

[0033] Fig. 3 is a collection of bar graphs showing the expression levels of 15 genes identified as differentially expressed in tumors which responded to a combination therapy comprising palbociclib and sunitinib (responders) versus tumors which did not respond (non-responders), and tumors which mildly responded (mild-resp.). Data were obtained by NanoString assay of 17 PDX models. The PDX models are indicated along the x axis, and each of the 15 genes has a color code;

[0034] Figs. 4A-4B are a collection of 29 boxplots superimposed on principal component analysis (PCA) plots, presenting expression levels of 29 differently expressed genes (DEGs) in PDX models treated with sunitinib monotherapy (4A) or palbociclib monotherapy (4B). Data are presented for PDX models of tumors that were either combo responders (Combo-R) or combo mid- or non- responders (Combo-NR). NR: PDX models which did not respond to the sunitinib / palbociclib monotherapy; R: PDX models that responded to sunitinib / palbociclib monotherapy. Square rames mark DEGs that presented opposite trends of expression level in monotherapy compared to combo therapy;

[0035] Figs. 5A-5B are bar graphs showing the expression levels of 29 differently expressed genes (DEGs) in 2 PDX models of tumors that did not respond to combination therapy comprising palbociclib and sunitinib (combo therapy), subjected to monotherapy with sunitinib (Sunitinib) or palbociclib (Palbociclib), or to a combo therapy (Combo), or not treated at all (Control). mRNA expression levels where detected using RNA-Seq analysis. 5A: RA-179J model; 5B: RA-441 model. All expression levels values were normalized (relative to control samples in each model); Fig. 6 is a bar graph showing the expression levels of 15 genes identified as differently expressed in tumors responsive to combination therapy comprising palbociclib and sunitinib (combo therapy) versus combo therapy non-responsive tumors. Expression levels are shown in 4 PDX models: a non-responder to combo therapy (Not Res.) model, 2 models of responders to combo therapy (Responders), and one model of mild responder to combo therapy (Mid-resp). Data were obtained by NanoString assay;

[0036] Figs. 7A-7D are graph showing changes in tumor volume as function of time (days) of 4 tumors derived from 4 PDX models: 2 breast cancer models RA-393 and RA-198 (7A, 7D), and 2 pancreas cancer models: RA-200 and RA-454 (7B, 7C), following combination therapy with palbociclib and sunitinib or treatment with vehicles. Resp.: tumors that responded to the combination therapy; Non-Resp.: tumors that did not respond to the combination therapy; Mid- Resp.: tumors that responded mildly to the combination therapy; and

[0037] Fig. 8 is a collection of bar graphs showing the expression levels of 11 genes and one transcript identified as differentially expressed in cancerous patients which responded to a combination therapy comprising palbociclib and sunitinib (Responders) and cancerous patients which did not respond (Non-responders). Data were obtained by NanoString assay for 8 patients enrolled in a clinical trial.

[0038] DETAILED DESCRIPTION

[0039] Many patients with advanced cancer develop resistance to treatment. Combination therapies targeting several signaling pathways simultaneously constitute a promising approach for overcoming drug resistance and providing long-term tumor control since cancer cells are characterized by multiple altered molecular pathways and multiple mechanisms of resistance. However, the effect of a combination therapy cannot be predicted a priori. Therefore, before combination therapies are applied, it is highly desirable to verify the patient's expected responsiveness to the combined therapy, i.e., whether the patient will indeed respond to the combined drug treatment, and if so what type of response should be expected, e.g., an additive effect, a synergistic effect, or an antagonistic effect. A biomarker for evaluating a patient's responsiveness to combination therapy is therefore highly desirable.

[0040] Palbociclib is an inhibitor of cyclin dependent kinases 4 and 6 (CDK4 / 6) that reduces proliferation by blocking progression of the cell from G1 into S phase of the cell cycle. Palbociclib is approved by the United States Food and Drug Administration (FDA) for the treatment, inter alia, of hormone receptor positive (HR)+and human epidermal growth factor receptor 2 negative (HER2)- advanced / metastatic breast cancer, in combination with an aromatase inhibitor or a hormone therapy such as antiestrogen fulvestrant, which acts as antagonist of the estrogen receptor (ER). Sunitinib is a multi-targeted receptor tyrosine kinase inhibitor (mtRTKI), which inhibits platelet-derived growth factor receptors (PDGFRa and PDGFRP), vascular endothelial growth factor receptors (VEGFR1, VEGFR2, and VEGFR3), stem cell factor receptor (KIT), Fms-like tyrosine kinase-3 (FLT3), colony stimulating factor receptor Type 1 (CSF-1R), as well as the glial cell-line derived neurotrophic factor receptor (RET). Sunitinib is FDA-approved for treating gastrointestinal stromal tumor, advanced / recurrent renal cell carcinoma (RCC), and progressive pancreatic neuroendocrine tumors. Palbociclib and sunitinib, like most CDK4 / 6 inhibitors and mRTKIs, respectively, do not have predictive markers. As palbociclib and sunitinib do not share a mechanism of action, combining them is not expected to have major pharmacodynamics interactions. However, these drugs do share adverse events such as hepatotoxicity, blood dyscrasias, and gastrointestinal adverse events, which could potentially lead to unacceptable toxicity of the combination.

[0041] A patient-derived xenograft (PDX) mouse model, also referred to herein as "PDX model", is a model of cancer obtained by implanting tissue and / or cells of a tumor of a human patient into an immune-deficient mouse (herein also referred to as "PDX mouse"). This implant preserves the human origin cell-cell interactions and tumor microenvironment. PDX models reflect tumor biology and drug responsiveness more accurately than human cancer cell lines and as such constitute effective translational pre-clinical models for investigating, inter alia, drug combinations.

[0042] It has been previously shown by the present inventors that treating PDX mice with a combination of palbociclib and sunitinib ("combo treatment", "combo therapy" or "combo"), demonstrated a synergistic inhibitory effect in various solid cancers (e.g., gastric, colon, cholangiocarcinoma, pancreas, ovarian, breast, lung, Ewing sarcoma and carcinosarcoma), as compared to PDX mice treated with vehicle (control) and PDX mice treated with palbociclib alone or with sunitinib alone (monotherapy), with no unexpected toxicities.

[0043] The present disclosure is based on a discovery by the present inventors that tumor tissues obtained from PDX mice unresponsive to the combo treatment exhibited a distinct gene profile, referred to herein as "genetic signature". In this profile, certain genes were statistically significantly differentially expressed compared to the gene profile of PDX mice responsive to the combo treatment. Surprisingly, this genetic signature included overexpressed genes that were not overexpressed in tumors unresponsive to treatment with either palbociclib or sunitinib as monotherapies.

[0044] For instance, when the inventors analyzed the expression levels of genes identified as overexpressed in tumors unresponsive to the combination therapy, they found that most of these genes were not overexpressed in tumors unresponsive to sunitinib monotherapy compared to tumors responsive to sunitinib monotherapy (see Example 3 and Fig. 4A). Interestingly, an opposite trend was observed for several genes (highlighted in square frames in Fig. 4A), where higher gene expression was associated with a positive response to sunitinib. Similarly, when evaluating the expression levels of these overexpressed genes in tumors resistant to palbociclib monotherapy, the inventors did not observe elevated expression of these genes in palbociclib-resistant tumors compared to palbociclib-responsive tumors (see Example 3 and Fig. 4B).

[0045] Genomic biomarker

[0046] Biomarkers are biological indicators used to predict various factors, such as disease risk, facilitate early detection, enhance treatment selection, and monitor the outcomes of therapeutic interventions. Predictive biomarkers specifically help identify which patients are most likely or unlikely to benefit from a particular treatment.

[0047] As defined herein, a genomic biomarker is a measurable DNA or RNA characteristic that reflects normal biological processes, disease-related processes, or responses to therapeutic or other interventions.

[0048] A genomic biomarker can comprise one or more DNA and / or RNA characteristics. DNA characteristics include, but are not limited to, single nucleotide polymorphisms (SNPs), variability of short sequence repeats, haplotypes, DNA modifications (e.g., methylation), deletions or insertions of single nucleotides, copy number variations, and cytogenetic rearrangements (e.g., translocations, duplications, deletions, or inversions). RNA characteristics include, but are not limited to, RNA sequences, RNA expression levels, RNA processing (e.g., splicing and editing), and microRNA levels. These characteristics can encompass measurements of gene expression (individual or sets of genes), gene function (individual or sets of genes), gene regulation (individual or sets of genes), or the presence and properties of specific genetic sequences. The term "pharmacogenomics" (PGx), as defined herein, refers to the study of variations in DNA and / or RNA characteristics in relation to drug response.

[0049] The terms "drug response" and "responsiveness to drug therapy" are interchangeable herein and refer to how a drug behaves in the body, including its absorption and distribution or disposition (pharmacokinetics (PK)), as well as its effects on the body, such as how it works (pharmacodynamics (PD)), its effectiveness, and any side effects it may cause.

[0050] In one aspect, the present disclosure relates to a genomic biomarker based on a set or a panel of 5-50 genes for predicting responsiveness to an anti-cancer treatment comprising a combination of palbociclib and sunitinib. This panel of genes is also referred to herein as "biomarker gene panel" or "biomarker gene set". The genes in a disclosed biomarker gene panel are selected based on their expression level, function and / or regulation characteristics.

[0051] In some embodiments, a disclosed genomic biomarker consists of RNA characteristics such as RNA expression levels and / or RNA sequences. The RNA may be, for example, messenger RNA (mRNA).

[0052] In some embodiments, the genes in the biomarker gene panel are determined based on a differential expression analysis of cancerous tissues from at least 5 subjects that responded to the combination therapy (herein termed "responders") and cancerous tissues from at least 5 subjects that did not respond to the combination therapy (herein termed "nonresponders"). In some embodiments, the differential expression analysis is performed using PDX models of responders and non-responders.

[0053] The present inventors conducted bioinformatic analyses that identified 196 genes with differential expression between responders and non-responders. To refine this list and establish a predictive gene profile for further validation, the biological pathways associated with these differentially expressed genes were analyzed. Specific protein families and complexes encoded by some of these genes were identified based on their biological activity or function. For each relevant protein family or complex linked to a particular pathway, one or two representative genes were selected. These genes are hypothesized to play key roles in biological processes influencing responsiveness or resistance to the combination therapy.

[0054] For instance, some genes in the disclosed genomic biomarker are involved in lysosomal function, regulation, and vesicle dynamics within cells (e.g., structural components, translocation). These genes are proposed to contribute to drug resistance by promoting drug exocytosis through the lysosome as a resistance mechanism. In some embodiments, the genes in the biomarker panel show more than a twofold difference in expression levels between responders and non-responders. For example, a gene's expression level in non-responders may be two times, 2.5 times, three times, 3.5 times, four times, or even higher compared to responders. Additionally, or alternatively, each of the genes in the panel exhibits minimal overlap -no more than 10% -in expression values between responders and non-responders across all analyzed cancerous tissues. That is, for each gene in the genomic biomarker, the range of expression values in patients who benefit from the treatment and patients who do not benefit from the treatment is largely distinct, with very little overlap between the two groups. This distinction makes it easier to use these genes as biomarkers because their expression levels clearly differentiate between responders and non- responders, improving the predictive accuracy of the biomarker panel.

[0055] In some embodiments, the genomic biomarker comprises a set of 5-50 genes, for example, 5-10, 5-15, 6-12, 7-15, 10-16, 10-20, 10-30, 12-20, 12-15, 12-18, 15-20, 15-25, 17-25, 20-25, 20-30, 20-35, 25-30, 25-40, 30-40, 30-50 or 40-40 genes. In some embodiments, a disclosed biomarker comprises 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29 or 30 genes.

[0056] A genomic biomarker disclosed herein comprises a panel of at least 5 genes selected from the genes identified by the gene symbols SP140, FGR, HLA-DRA, LINC00528, GPR132, ABCA6, SELPLG, LIN00494, RASGRP3, CIITA, NCF4, PTPN22, FGD2, FERMT3, JAK3, POU2F2, IFI30, LAMP3, TMEM140, PSMB9, ETV7, CTA-126B4.7, TMC8, CCDC88B, PSMB8-AS1, TFEB or PSME2, or AC009133.17. the genomic biomarker may further include the transcript RP11- 134L10.1. These as well as other genes disclosed herein have been determined as being particularly informative with respect to responsiveness to the combo treatment.

[0057] In some embodiments, the genomic biomarker comprises at least the genes designated as CIITA, TFEB, CCDC88B, CTA-126B4.7, ETV7, FERMT3, FGD2, FGR, HLA-DRA, RASGRP3, SP140, LINC00528, GPR132, IFI30, POU2F2 and TMEM140, and the transcript RP11- 134L10.1.

[0058] The Class II Major Histocompatibility Complex Transactivator (CIITA) gene is a key regulator of the immune system, playing a critical role in the expression of Class II Major Histocompatibility Complex (MHC) molecules, which are essential for antigen presentation and adaptive immune responses. MHC Class II molecules are responsible for presenting antigens to CD4+T cells, initiating immune responses. CIITA expression is tightly controlled and is primarily induced in antigen-presenting cells (APCs) such as dendritic cells, macrophages, and B cells. Although the CIITA gene is primarily associated with immune regulation, emerging evidence suggests it may have indirect or context-dependent roles in drug resistance, particularly in the context of immunotherapies, tumor microenvironment modulation, and stress response pathways.

[0059] The ETS Variant Transcription Factor 7 (ETV7) gene also known as TEL2, encodes a transcription factor that is part of the erythroblast transformation specific (ETS) family, which plays critical roles in regulating cellular processes such as proliferation, differentiation, and survival. The ETV7 gene plays a significant role in cancer biology, and its dysregulation is associated with drug resistance through pathways involving apoptosis evasion, drug efflux, and DNA repair. Targeting the ETV7 protein or its downstream pathways may overcome therapy resistance in cancers where it is overexpressed.

[0060] The Transcription Factor EB (TFEB) gene encodes a key regulator of lysosomal biogenesis, autophagy, and cellular metabolism, all of which are mechanisms that can contribute to drug resistance in cancer. As part of the MiT / TFE family of transcription factors TFEB maintains cellular homeostasis and supports cancer cell survival through mechanisms such as:

[0061] 1. Lysosomal drug sequestration and exocytosis. TFEB enhances lysosomal biogenesis and exocytosis, enabling cancer cells to sequester and expel chemotherapeutic drugs, reducing their effectiveness.

[0062] 2. Autophagy activation. TFEB promotes autophagy, a process that degrades damaged organelles and recycles cellular components, thereby helping cells survive under stress. Many cancer therapies rely on inducing stress or damage within tumor cells. Enhanced autophagy, driven by TFEB, can counteract these therapies.

[0063] 3. Metabolic adaptation. TFEB supports metabolic reprogramming in cancer cells, allowing them to adapt to harsh tumor microenvironments or therapeutic pressure. By regulating lipid and carbohydrate metabolism, TFEB helps maintain energy homeostasis, which may indirectly enhance drug resistance.

[0064] 4. Stress response pathways. TFEB activation is part of the cellular response to oxidative stress and other harmful conditions. In drug-resistant cells, TFEB-mediated stress responses can mitigate the cytotoxic effects of treatments. Elevated TFEB expression or nuclear localization could serve as a predictive biomarker for drug resistance in cancer patients.

[0065] The Coiled-Coil Domain Containing 88B (CCDC88B) gene encodes a protein involved in immune system regulation. It is primarily expressed in immune cells such as T cells, B cells, and macrophages. The protein contains coiled-coil domains, which are structural motifs facilitating protein-protein interactions. CCDC88B plays a role in regulating immune responses, including inflammation and cytokine signaling pathways. The CCDC88B protein has been implicated in certain cancers, where its expression might influence tumor immunity and responses to immunotherapy. Dysregulation of immune signaling pathways involving CCDC88B might contribute to resistance to immunomodulatory drugs.

[0066] The CTA-126B4.7 gene is affiliated with the non-coding RNA (ncRNA) class. It is not prominently discussed in scientific literature. Its function and significance are unclear.

[0067] The FERM Domain Containing Kindlin 3 (FERMT3) gene encodes a protein known as Kindlin-3, which plays a crucial role in integrin activation and signaling. Integrins are cell surface receptors involved in cell adhesion, migration, and communication with the extracellular matrix. Kindlin-3 is particularly important in hematopoietic cells (blood and immune cells), where it regulates integrin-mediated adhesion and signaling, essential for proper immune responses and platelet function. While FERMT3 is not directly classified as a "drug resistance" gene, its role in integrin signaling, cell adhesion, and immune function makes it relevant in contexts where these processes contribute to resistance mechanisms, particularly in cancer and immune-related therapies.

[0068] The FYVE, RhoGEF and PH Domain Containing 2 (FGD2) gene encodes a protein involved in cellular signaling, particularly the regulation of the actin cytoskeleton and membrane trafficking. FGD2 primarily functions as a guanine nucleotide exchange factor (GEF) for cell division cycle 42 (Cdc42) protein, activating it and facilitating processes like cytoskeletal organization and cell polarity. FGD2 is expressed in specific immune cells such as macrophages and dendritic cells, where it helps mediate responses to external stimuli, phagocytosis, and immune signaling. While FGD2 is not directly recognized as a drug resistance gene, its involvement in cytoskeletal dynamics, cell signaling, and immune responses makes it a candidate for influencing resistance mechanisms in certain contexts, particularly in cancer and immune-related therapies. For example, dysregulation alterations in cytoskeletal dynamics and cell adhesion can contribute to drug resistance by enhancing cell survival and escape mechanisms in tumors.

[0069] The FGR Proto-Oncogene, Src Family Tyrosine Kinase (FGR) gene encodes a member of the Src family of non-receptor tyrosine kinases. These kinases are involved in signaling pathways that regulate a wide range of cellular processes, including proliferation, differentiation, migration, and survival. For example, FGR influences survival pathways involving PI3K / AKT and MAPK. FGR protein plays a role in signaling pathways in macrophages, neutrophils, and other immune cells, affecting their activation, migration, and phagocytosis. It contributes to cytoskeletal remodeling, which is critical for cell adhesion and motility.

[0070] FGR's involvement in signaling pathways tied to survival, migration, and immune function may indirectly or directly influence drug resistance in specific contexts. For example, FGR can contribute to tumor progression (oncogenic signaling) through its role in cell survival, migration, and metastasis. Overexpression or dysregulation of FGR has been reported in certain cancers, including lung and colorectal cancers. Furthermore, by promoting pro-survival signaling pathways, FGR may enable cancer cells to evade the effects of chemotherapy or targeted therapies.

[0071] The HLA-DRA gene encodes the alpha subunit of the HLA-DR protein, which is a class II major histocompatibility complex (MHC) molecule, which pairs with HLA-DRB (its beta-chain counterpart) to form the functional HLA-DR heterodimer. This complex presents processed peptides from extracellular antigenic peptides to CD4+T-helper cells, thereby initiating and regulating immune responses.

[0072] The HLA-DRA gene is not directly involved in drug metabolism or efflux but has indirect roles in drug resistance due to its central involvement in immune regulation and disease pathology. In cancer, tumor cells can downregulate MHC class II molecules like HLA-DR to evade immune detection, reducing the efficacy of immunotherapies such as immune checkpoint inhibitors or cancer vaccines. Low expression of HLA-DRA in tumors is associated with poor prognosis and resistance to immune-based treatments.

[0073] The RAS Guanyl Releasing Protein 3 (RASGRP3) gene encodes a member of the RAS guanyl nucleotide exchange factor (GEF) family. RASGRP3 protein activates RAS family GTPases by promoting the exchange of GDP for GTP, which triggers downstream signaling pathways involved in cell proliferation, survival, and differentiation. It is particularly relevant in the context of the immune system, cancer, and vascular biology. RAS and RAP GTPases proteins are central regulators of signaling pathways like the MAPK / ERK and PI3K / AKT pathways. Dysregulation of RASGRP3 expression or activity has been implicated in the development and progression of some cancers, particularly through its impact on cell survival and proliferation pathways. The RASGRP3 gene is not traditionally categorized as a drugresistance gene, but its involvement in critical signaling pathways that regulate survival, proliferation, and angiogenesis links it to drug resistance mechanisms. For example, overactivation of the RAS-MAPK or RAS-PI3K-AKT pathways is a well-known mechanism of drug resistance in cancer. Since RASGRP3 activates RAS, its overexpression or dysregulation may contribute to resistance to therapies targeting these pathways, such as tyrosine kinase inhibitors or MEK inhibitors.

[0074] The SP140 gene encodes a nuclear body protein that plays a role in regulating transcription, chromatin remodeling, and immune responses. SP140 is predominantly expressed in immune cells, such as lymphocytes and macrophages, and has been implicated in the regulation of immune-related processes, including inflammation and pathogen response. It is also associated with certain autoimmune and inflammatory diseases. SP140 functions as a transcriptional regulator, controlling the expression of genes involved in immune responses, particularly in macrophages and B cells.

[0075] The SP140 gene is indirectly related to drug resistance through its role in transcriptional regulation, immune modulation, and chromatin remodeling. Its involvement in autoimmune diseases and cancers like chronic lymphocytic leukemia (CLL) suggests that dysregulation of SP140 could influence therapeutic outcomes and resistance mechanisms. For example, in CLL altered SP140 expression or mutations might contribute to resistance to standard chemotherapy or targeted therapies by influencing survival pathways or immune evasion. SP140 is part of chromatin-associated complexes that influence gene expression, thereby possibly affecting the adaptation of cells to drug-induced stress.

[0076] The Transmembrane Protein 140 (TMEM140) gene encodes a protein of unknown precise function. It is probably a multi-pass transmembrane protein expressed in various tissues, including the brain and some cancer cells. TMEM140 may play a role in cancer biology, cellular signaling, and possibly in neural functions. TMEM140 is overexpressed in certain cancers, such as gliomas and breast cancer. It may influence cancer progression through pathways that affect cell proliferation, invasion, and survival. TMEM140's expression has been linked to hypoxic conditions (low oxygen levels), a common feature in the tumor microenvironment.

[0077] Potential involvement of TMEM140 in cancer progression and cellular signaling could contribute indirectly to resistance mechanisms. For example, since TMEM140 is linked to hypoxia, its expression might contribute to drug resistance in tumors. Hypoxic tumor environments often lead to reduced drug efficacy and increased resistance to chemotherapy and radiation therapy. Overexpression of TMEM140 in cancer cells may activate survival pathways, helping cells evade the cytotoxic effects of chemotherapy or targeted therapies. As a transmembrane protein, TMEM140 might influence the function or expression of drug transporters, although direct evidence is lacking. Alterations in drug uptake or efflux could also contribute to resistance.

[0078] The LINC00528 gene represents a long non-coding RNA (IncRNA), a class of RNA molecules that do not encode proteins but play critical roles in regulating gene expression at the transcriptional, post-transcriptional, and epigenetic levels. Like other IncRNAs, LINC00528 functions as a regulatory RNA, modulating the expression of target genes via interactions with transcription factors, RNA-binding proteins, or chromatin-remodeling complexes.

[0079] LINC00528 has been implicated in various biological processes, particularly in cancer biology, through its ability to influence cell proliferation, apoptosis, and metastasis. LINC00528 is dysregulated in certain cancers, including breast cancer, lung cancer, and glioma. It is often associated with promoting tumor progression, invasion, and metastasis.

[0080] The LINC00528 gene may contribute to drug resistance in some ways including inhibiting apoptosis and promoting cell survival cancer cells, which could lead to resistance to chemotherapy or targeted therapies. Also, dysregulation of LINC00528 can activate oncogenic signaling pathways, such as the PI3K / AKT or MARK pathways, which are commonly associated with drug resistance. LINC00528 might influence the tumor microenvironment by regulating the expression of cytokines, chemokines, or extracellular matrix components, potentially affecting the efficacy of therapies.

[0081] The GPR132 gene (also known as G protein-coupled receptor 132) encodes a protein that is part of the G protein-coupled receptor (GPCR) family. These receptors are involved in a variety of cellular processes, including signal transduction and response to external signals. GPR132 is specifically involved in regulating immune responses and inflammation, and its expression has been found in different tissues, including the lungs and liver. In terms of drug resistance, GPR132 has been implicated in some contexts where drug resistance may be a factor. Some studies suggest that GPR132 may be involved in cancer cell survival and metastasis. Since the activity of GPR132 can affect tumor microenvironments, it could contribute to resistance to chemotherapy or other cancer treatments. Specifically, it may influence the immune system's response to tumors, which can impact the effectiveness of immune-based therapies.

[0082] The transcript RP11-134L10.1 relates to a gene of the RP11 series. It is a long noncoding RNA (IncRNA) that does not have clear specific functions but likely has regulatory roles in the cell, such as regulation of gene expression (at transcriptional or post-transcriptional levels), chromatin remodeling and interactions with other RNAs, proteins, or DNA. Many IncRNAs, including those in the RP11 family, have been implicated in modulating sensitivity to chemotherapy and targeted therapies. For example, they can affect drug efflux by regulating proteins like P-glycoprotein, affect Apoptosis by influencing cell survival pathways, and affect signaling pathways by modulating pathways such as Wnt / p-catenin, PI3K / AKT, or NF-KB.

[0083] The IFI30 gene, also known as gamma-interferon-inducible lysosomal thiol reductase (GILT), encodes a lysosomal enzyme involved in the reduction of protein disulfide bonds in acidic environments. It is primarily expressed in antigen-presenting cells such as dendritic cells, macrophages, and B cells, whereby it is associated with antigen processing and presentation, playing a crucial role in the immune response. IFI30 facilitates the unfolding of disulfide-bonded proteins within lysosomes, enabling their degradation into peptides that can be presented on MHC class II molecules. IFI30 expression is upregulated by interferon-gamma (IFN-y), a cytokine critical for immune activation.

[0084] IFI30 may have implications in drug resistance, particularly in the context of cancer. IFI30 is implicated in promoting the survival of cancer cells by influencing the tumor microenvironment, immune evasion, and oxidative stress responses. In certain cancers, IFI30 overexpression is linked to resistance to chemotherapy or immunotherapy. In addition, its role in lysosomal function and redox regulation can impact the degradation of chemotherapeutic agents or modulate apoptosis pathways, contributing to resistance.

[0085] The POU2F2 gene encodes the POU class 2 homeobox 2 transcription factor, also known as Oct-2. It plays a significant role in regulating gene expression in immune cells, particularly in B-cell development and differentiation. POU2F2 belongs to the POU gene family of transcription factors, characterized by their conserved DNA-binding POU domain. POU2F2 is also expressed in some non-immune cells and is implicated in various cellular processes such as proliferation, differentiation, and survival. It is also detected in cancers under specific conditions.

[0086] POU2F2 might play a role in drug resistance, particularly in the context of cancer. Aberrant expression of POU2F2 has been observed in certain cancers, where it promotes tumor cell proliferation, survival, and metastasis. PO(72F2-mediated activation of survival pathways can make tumor cells less susceptible to chemotherapy-induced apoptosis. For example, elevated POU2F2 expression in some solid tumors correlates with resistance to treatments, possibly through its role in transcriptional reprogramming and stress responses. Furthermore, POU2F2 might interact with oncogenic pathways (e.g., MYC or NF-KB), which are known to contribute to drug resistance. It can also regulate immune-related genes, potentially affecting the tumor microenvironment and responses to immunotherapy.

[0087] POU2F2 expression levels could serve as a biomarker for drug resistance or poor prognosis in specific cancers.

[0088] Methods of prognosis and treatment

[0089] In one aspect, the present disclosure relates to a method for predicting, in a subject in need thereof, responsiveness to a combination anti-cancer therapy comprising palbociclib and sunitinib, whereby the prediction is based on a genomic biomarker disclosed herein.

[0090] Embodiments of this aspect relate to a prognosis method comprising at least the following steps:

[0091] (i) obtaining a sample of cancerous tissue from the subject;

[0092] (ii) subjecting the sample to gene expression profiling;

[0093] (iii) evaluating the gene expression level of each gene in a predetermined set of 5-50 genes;

[0094] (iv) comparing the expression level of each gene in the set to a corresponding first reference value and / or to a corresponding second reference value; and

[0095] (v) determining that the subject is unlikely to respond to the combination therapy if the expression levels of at least some of the predetermined genes are higher by at least 2-fold compared to their corresponding first reference values; or

[0096] (vi) determining that the subject is likely to respond to the combination therapy if at least 5 of the predetermined genes are not expressed and / or if the expression levels of at least some of the predetermined genes are lower by at least 2-fold compared to their corresponding second reference values, wherein a corresponding first reference value is a score value of the expression levels of a predetermined gene across 5-10 different cancerous tissues of subjects determined to be responsive to the anti-cancer therapy, and a corresponding second reference value is a score value of the expression levels of a predetermined gene across 5-10 different cancerous tissues of subjects determined to be non- responsive to the anti-cancer therapy.

[0097] A disclosed method for early diagnosis of a subject's likelihood to respond to a combination therapy comprising co-administration of palbociclib and sunitinib is based on detection of differences in the expression levels of specific, predetermined genes (also referred to as "target genes") between a biological sample of the subject (herein also termed "test sample") and 5-10 reference samples. As such, the expression levels of these specific target genes in the test sample and the reference samples are compared and when a statistically significant difference in expression levels is observed, the likelihood of responding to the combo therapy is either diagnosed and confirmed or negated. A score of the expression levels of a predetermined gene in 5-10 reference samples is referred to herein as the "reference value". The specific predetermined genes constitute the genomic biomarker.

[0098] In the first step of the disclose method, a sample of cancerous tissue is taken from the subject. The sample taken may be cells, solid tissue or bodily fluids (e.g., ascites (buildup of excess fluid in the abdominal cavity) and pleural effusion containing tumor cells).

[0099] The sample may be a biopsy, i.e., a small sample of cells, tissue or fluid removed from a subject for examination. Additionally, or alternatively, in some cases of solid cancers, the whole cancerous mass may be the sample taken.

[0100] The term "biopsy", as used herein, refers both to the procedure of removing bodily mass as well as the actual mass removed (cells, tissues). Different types of biopsy procedures include, but are not limited to (1) a punch biopsy operated on the skin, whereby an instrument punches a small hole in the skin to obtain a skin sample; (2) a needle biopsy employing a special hollow needle guided by X-ray, ultrasound, CT scan or MRI to obtain tissue from an organ or from tissue underneath the skin; (3) an endoscopic biopsy in which an endoscope is used to remove tissue, such as from the stomach during a gastroscopy; (4) an excision biopsy which utilizes a surgery to remove a larger section of tissue; and (5) perioperative biopsy that is carried out during surgery. How a biopsy is carried out will depend on what tissue sample is taken and where it is taken from.

[0101] In step (ii) of the method, the sample is subjected to a gene profiling procedure for the purpose of obtaining the gene expression fingerprint (sometimes also referred to herein as the genetic or signature) of the sample.

[0102] Gene expression is the process by which the information encoded in a gene is used to produce a functional product, such as a protein or RNA molecule. It is the fundamental mechanism by which genetic information leads to cellular functions and organismal traits.

[0103] Gene expression involves two primary steps: (1) transcription: the DNA sequence of a gene is transcribed into messenger RNA (mRNA); and (2) translation: the mRNA is translated into a protein by ribosomes, or in the case of non-coding RNAs (like long non-coding RNA (IncRNA) or ribosomal RNA (rRNA)), the RNA itself functions as the end product. The determination of the genes expressed under specific circumstances in a cells or tissue sample provides a global picture of the cellular or tissue function.

[0104] The term "gene expression profile", as used herein, refers to the complete set of gene expression levels in a particular cell, tissue, or organism at a specific time or under certain conditions. It represents which genes are actively transcribed into RNA and at what levels. This "snapshot" of gene activity provides insights into the biological state of the system being studied.

[0105] Gene expression profile reflects the unique expression patterns for a given cell type, developmental stage, disease state, or environmental condition, and as such is applied, for example, in identifying biomarkers for diseases (e.g., cancer subtypes), understanding cellular responses to external stimuli (e.g., drug treatments or environmental stress) and differentiating cell types or developmental stages.

[0106] The term "gene expression profiling", herein interchangeable with the term "genomic profiling", refers to the experimental process conducted to measure and analyze the gene expression levels of specific genes or across the entire genome. It aims to quantify the expression of single, tens, hundreds or even thousands of genes simultaneously, often to compare different conditions (comparative analysis), such as but not limited to, healthy vs. diseased tissues or treated vs. untreated cells, and the like. Genomic profiling is further utilized for understanding which biological pathways are active or disrupted, for finding gene signatures associated with particular diseases or drug responses, and for precision medicine, namely tailoring therapies based on the gene expression profiles of individual patients. In the context of the present disclosure, gene expression profiling is used for obtaining the gene expression profile of a cancerous tissue sample obtained from a subject for finding gene signatures associated with responsiveness and non-responsiveness to combo therapy.

[0107] Gene expression is determined by measuring the amount of RNA or protein produced from a gene. Techniques to assess gene expression focus on either the RNA or protein levels. Embodiments disclosed herein pertain to evaluating gene expression by measuring RNA Levels.

[0108] RNA-based methods assess transcriptional activity by quantifying the mRNA produced from a gene. State of art techniques include:

[0109] (i) Quantitative PCR (qPCR): measures mRNA levels by reverse-transcribing RNA into complementary DNA (cDNA) and amplifying it. The fluorescence intensity is used to quantify expression. Often used for targeted profiling of selected genes rather than the whole transcriptome.

[0110] (ii) Northern Blotting: separates RNA samples by size using gel electrophoresis, followed by detection with a labeled probe specific to the target RNA.

[0111] (iii) Microarrays: detect and quantify the expression of thousands of genes simultaneously using pre-defined DNA probes of specific genes immobilized on a solid surface.

[0112] (iv) RNA Sequencing (RNA-Seq): a high-throughput method that sequences RNA to quantify and identify all expressed genes (i.e., the entire transcriptome).

[0113] (v) NanoString analysis: a highly sensitive and robust technique that uses a direct digital detection method to quantify RNA, DNA, protein and / or miRNAs molecules in the same platform without requiring reverse transcription or amplification. The NanoString platform utilizes a technology based on molecular barcodes and hybridization for detecting and quantifying specific nucleic acid sequences.

[0114] In some embodiments, gene expression profiling is conducted using the RNA-Seq technology. RNA-Seq detects low-abundance transcripts, quantifies transcripts with a large range of expression levels, identifies novel transcripts, isoforms, and non-coding RNAs and provides precise sequence information. This technique is suitable for various species, regardless of whether a reference genome is available. RNA-Seq is suitable for measuring differential expression between conditions (e.g., disease vs. normal), linking expression changes to biological pathways, and for analyzing gene expression at the single-cell level. The NanoString's technology is useful, inter alia, in identifying potential biomarkers for diseases or treatment response and offers insights into patient-specific molecular profiles for precision medicine. The key principles of RNA-Seq are described in the Material and Method section herein. Illumina, Nanopore sequencing, and PacBio are currently among the state-of-the-art high- throughput sequencing platforms. In some embodiments, sequencing is performed using the illumina RNA-Seq platform, which is based on sequencing-by-synthesis chemistry. The illumina RNA-Seq is further described in the Materials and Methods section herein.

[0115] In some embodiments, gene expression profiling is conducted using the NanoString technology. NanoString can eliminate biases associated with PCR-based methods like RNA-Seq or qPCR and is easy for use as it has simplified workflow with minimal sample preparation. Other advantages of NanoString analysis include high robustness, as it works well with degraded samples (e.g., FFPE tissues), direct digital quantification since it provides absolute counts rather than relative measurements. This technology is scalable (suitable for small or large-scale studies) and panels can be designed to include genes of interest. However, nnlike RNA-Seq, NanoString only detects genes included in the selected panel and it is less sensitive for detecting extremely low-abundance transcripts compared to RNA-Seq.

[0116] The principles of NanoString technique are described in the Material and Method section herein.

[0117] In step (iii) of a disclosed method, the identity and expression levels of genes of interest for example, the expression level and identity of at least some of the 5-50 target genes constituting the genomic biomarker contemplated herein, is assessed or evaluated. In step (iv) which follows, the expression level of each is compared to a corresponding first reference value and / or to a corresponding second reference value. In the final steps of a disclosed prognosis method, determination is made whether the subject is likely or unlikely to respond to the combination therapy.

[0118] After gene expression profile is determined by either gene expression profiling method, statistical approaches are employed to detect differences in expression levels across experimental groups. The statistical testing ensures that an observed change in expression levels of certain genes is due to an actual difference between the test sample and reference samples.

[0119] Each statistical test relies on a null hypothesis, which assumes that gene expression levels are the same across groups or samples. This assumption typically holds true for most genes. The p-value, which indicates the likelihood of a true difference between groups, represents the probability of observing a specific difference (or a more extreme one) under the null hypothesis. Small p-values challenge the validity of the null hypothesis. For example, a two-fold difference in expression levels between test and reference samples, combined with low p-values, is considered a reliable indicator, leading to the rejection of the null hypothesis for the tested genes. The commonly accepted threshold for rejecting a null hypothesis is a p-value less than 0.05, though this cutoff is arbitrary and may need adjustment depending on the noise level in the data.

[0120] In embodiments disclosed herein, the expression levels or counts of each gene within a predefined set of 5-50 genes in a sample taken from a cancer patient with a specific cancer type (the test sample) are compared against reference values. These reference values are defined as score representing the expression levels of the same genes in 5-10 reference samples. The reference samples typically include at least 5 cancerous tissue samples (e.g., 5, 6, 7, 8, 9, 10, or more), derived from human subjects diagnosed with the same type of cancer and / or from patient-derived xenograft (PDX) models, treated with combination therapy and determined to be responsive to the therapy based on pharmacokinetic (PK) and / or pharmacodynamic (PD) measures, such as drug absorption, distribution, efficacy, and adverse effects. If the gene expression profile of the test sample shows significantly higher expression levels for some of the predefined genes compared to the reference values (referred to as "first reference values"), it may predict that the patient will not respond to the combination therapy. In such cases, an alternative anti-cancer therapy should be considered.

[0121] Alternatively, the reference samples may consist of at least 5 cancerous tissue samples (e.g., 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 20, or more) as defined herein, derived from human subjects diagnosed with the same type of cancer and / or PDX models of such human subjects, treated with the combo therapy and determined to be non-responsive to it based on pharmacokinetics (PK) and / or pharmacodynamics (PD) measures. In this scenario, if the gene expression profile of the test sample demonstrates substantially and statistically significant lower expression levels for some of the predefined genes compared to the reference values (referred to as "second reference values"), or if a subset of target genes (e.g., 5-15 genes, depending on the genomic biomarker composition) is not expressed at all, this may predict that the patient will respond to the combination therapy. In such cases, administration of the combined anti-cancer therapy should be considered.

[0122] The term "at least some of the predetermined genes", in the context of embodiments described herein, means at least 80% of the predetermined set of genes. Thus, in a genomic biomarker comprising a set of 10 genes, the expression level of at least 8 genes should substantially differ from their corresponding reference values (either the first reference values or the second reference values), and when the genomic biomarker comprises a set of 15 genes, the expression level of at least 12 (for example, 12, 13, 14 or 15) genes should substantially differ from their corresponding reference values.

[0123] In some embodiments, expression levels values of at least 5 predetermined genes, which substantially differ from their corresponding reference values, serve for predicting responsiveness or non-responsiveness to combo therapy, in some embodiments, at least 10 genes that are substantially differently expressed are indicative to responsiveness or non- responsiveness to combo therapy.

[0124] The terms "substantially differs" and "substantially differently expressed", as used herein are interchangeable and refer to at least 1.5-fold difference in the expression level of a target gene in a test sample compared to its corresponding first or second reference value. For example, the expression level of a predetermined gene in the test sample of a lung cancer patient may be 1.5, 1.75, 2, 2.5, 2.75, 3, 3.2, 3.5, 3.75, 4 times or more higher in the test sample compared to a first reference value as defined herein, namely, a score of the expression levels of this predetermined gene across 5-10 different cancerous tissues of subjects diagnosed with the same lung cancer and determined to be responsive to the anti-cancer therapy.

[0125] Alternatively, the substantially differently expressed expression level of a predetermined gene in the test sample may be 1.5, 1.75, 2, 2.5, 2.75, 3, 3.2, 3.5, 3.75, 4 times (fold) or more lower in the test sample compared to a second reference value as defined herein, namely, a score of the expression levels of this predetermined gene across 5-10 different cancerous tissues of subjects having the same type of cancer and determined to be non-responsive to the anti-cancer therapy.

[0126] In some embodiments disclosed herein, a difference in expression levels that is considered informative for prognostic purposes involves at least a two-fold change in the expression levels of at least 80% of the predefined genes between the test sample and the reference samples. For instance, a 2-fold, 3-fold, 4-fold, or greater difference in the expression levels of at least 80% of the predetermined genes in the test sample compared to the reference samples serves as a strong and reliable indicator of the subject's responsiveness or non-responsiveness to the therapy.

[0127] Gene expression analysis in a test sample can provide absolute, relative, or normalized expression levels for each gene. For accurate measurements, normalization to a stable reference gene is recommended. An ideal reference gene should remain unaffected by experimental treatments and exhibit consistent expression across different tissues. For example, expression levels can be measured as the ratio of a target gene to an internal reference or housekeeping gene, such as (non-POU domain containing, octamer-binding), with its expression level set to a fixed value (e.g., 10,000).

[0128] Relative gene expression refers to the ratio of a target gene's expression level to that of a non-internal reference gene, such as the corresponding gene in a healthy subject or a cancer patient not treated with combination therapy.

[0129] In some embodiments, the overall expression levels of a set of target genes (e.g., 5-50 genes) are assessed, combined, and scored as a "combined score" (distinguished from the scores assigned as the first and second reference values). This combined score may be calculated as the sum, average, or weighted average of the target genes' expression levels, where weights reflect the importance of each gene. For a weighted average, the expression level of each gene is multiplied by its assigned weight, and the sum of these products is divided by the total weight.

[0130] In some embodiments, a combined score for the predetermined set of genes is evaluated in step (iii) of a disclosed method. Then, in step (iv) this combined score is compared to a first reference which, in accordance with these embodiments, is based on combined scores for these predetermined set of genes from 5-10 samples of cancerous tissues from subjects responsive to the combination therapy; and / or to a second reference value which is based on combined scores of this gene set from 5-10 samples of cancerous tissues from subjects non-responsive to the therapy.

[0131] Responsiveness to combination therapy in steps (iv) and (v) of the method is determined by comparing the combined score of the test sample to these reference values. If the combined score is at least 2-fold higher than the first reference value, the subject is unlikely to respond to therapy. If the combined score is at least 2-fold lower than the second reference value, the subject is likely to respond to therapy.

[0132] In some embodiments the predetermined set of genes is a genomic biomarker being a panel of at least 5 genes selected from the genes designated as SP140, FGR, HLA-DRA, LINC00528, GPR132, ABCA6, SELPLG, LIN00494, RASGRP3, CIITA, NCF4, PTPN22, FGD2, FERMT3, JAK3, POU2F2, IFI30, LAMP3, TMEM140, PSMB9, ETV7, CTA-126B4.7, TMC8, CCDC88B, PSMB8-AS1, TFEB, PSME2 or AC009133.17, and / or the transcript RP11-134L10.1. In some embodiments, a disclosed prognosis method is based on assessing the expression profile of at least 5 of the genes CIITA, TFEB, CCDC88B, CTA-126B4.7, ETV7, FERMT3, FGD2, FGR, HLA-DRA, RASGRP3, SP140, LINC00528, GPR132, IFI30, POU2F2 or TMEM140, and the transcript RP11-134L10.1.

[0133] The prognostic method described herein can be applied to customize personalized anti-cancer treatment. For instance, a combination therapy involving palbociclib and sunitinib may be administered to a cancer patient only if the method predicts a positive response to the treatment.

[0134] Thus, the present disclosure also pertains to a method for treating cancer in a subject intended or is a candidate for combination therapy with palbociclib and sunitinib, comprising:

[0135] (i) obtaining a sample of cancerous tissue from the subject;

[0136] (ii) subjecting the sample to gene expression profiling;

[0137] (iii) evaluating the gene expression level of each gene in a predetermined set of 5-50 genes;

[0138] (iv) comparing the expression level of each gene in the set to a corresponding reference value; and

[0139] (v) administering to the subject a therapeutically effective amount of the combination therapy provided that at least 5 of the genes in the predetermined set are not expressed and / or if the expression levels of at least some of the predetermined genes (i.e., at least 80%) are lower by at least 2-fold compared to their corresponding reference values, wherein a corresponding reference value is a score of the expression levels of a predetermined gene across 5-10 different cancerous tissues of subjects non-responsive to the anti-cancer therapy.

[0140] If, however, the gene expression levels in the subject's cancerous tissue are comparable to the reference values - e.g., similar or within a range of 1-20% below the corresponding reference values - an alternative anti-cancer treatment should be considered. This may include other anti-cancer drugs, such as the standard of care (SOC) treatments specific to the cancer type and stage, which offer a range of established therapeutic options.

[0141] Cancers which may be treated with combo therapy include, for example, various solid cancers such as, but not limited to, gastric cancer, colon cancer, cholangiocarcinoma, pancreatic cancer, ovarian cancer, breast cancer, lung cancer, Ewing sarcoma, neuroendocrine cancer, melanoma, hepatocellular carcinoma and carcinosarcoma. The terms "therapy", "treatment", "treating" and "treat", as used herein, are interchangeable and refer to the following: (a) inhibiting a cancerous disease by halting its progression, (b) relieving, alleviating, or improving the condition by causing regression of the disease, and (c) curing the cancerous disease.

[0142] These terms encompass promoting or facilitating beneficial changes in the recipient's condition, whether subjective or objective. Subjective improvements may include a patient's reported better overall feeling or reduced pain, indicating successful treatment. Objective improvements involve observable changes noted by a clinician, such as a reduction in tumor size or regression of abnormalities upon examination, which also indicate effective treatment. Preventing further deterioration of the recipient's condition is also included in the scope of these terms. A therapeutic benefit may encompass any subjective or objective indicators of a positive response to combination therapy.

[0143] The term "therapeutically effective amount", as used herein, refers to the dose or amount of a compound - such as a combination of palbociclib and sunitinib - that, when administered according to a disclosed method, is sufficient to achieve the intended anti-cancer effect. This amount may represent the lowest dose that provides therapeutic benefits to patients, either on average or for a specific proportion of patients. The therapeutically effective amount can vary based on factors such as the relative proportions of palbociclib and sunitinib, the type and severity of cancer, and the age, weight, and other characteristics of the individual receiving treatment.

[0144] The terms "comprise", "comprising", "includes", "including", "having" and their conjugates, mean "including but not limited to".

[0145] The term "consisting of" means "including and limited to"

[0146] As used herein, the singular form "a", "an" and "the" include plural references unless the context clearly dictates otherwise. For example, the term "a compound" or "at least one compound", may include a plurality of compounds, including mixtures thereof.

[0147] Whenever a numerical range is indicated herein it is meant to include any cited numeral (fractional or integral) within the indicated range. The phrases "ranging / ranges between" a first indicate number and a second indicate number and "ranging / ranges from" a first indicate number "to" a second indicate number are used herein interchangeably and are meant to include the first and second indicated numbers and all the fractional and integral numerals therebetween.

[0148] Various embodiments and aspects as delineated hereinabove and as claimed in the claims section below find experimental support in the following examples.

[0149] EXAMPLES

[0150] Reference is now made to the following examples, which together with the above descriptions illustrate some embodiments in a non-limiting fashion.

[0151] Materials and Methods

[0152] Preparation of patient-derived xenograft (PDX) models

[0153] Overall, 25 patient-derived xenograft (PDX) models were developed from different patients with various solid tumors, except for one case where 2 models were developed from sequential samples obtained from the same breast cancer patient (models 3-RA-179F and 18- RA-179J) (Table 1 in Example 2). Cancer specimens were obtained by core needle biopsy, tumor resection, or pleural effusion in patients diagnosed and treated at the Rabin Medical Center (RMC), Israel. The majority of the PDX models (around 65%) were derived from tumors that were resistant to at least one prior therapy (chemotherapy or hormonal therapy). In about 26% of the cases, the tumor samples used for the PDX models were collected before the patients received any treatment, and in 9% of the cases the samples were derived from patients who responded to therapy. The tissue collection was performed per institutional review board (IRB)-approved protocols with written informed consent from the patients.

[0154] All mice were maintained and treated in accordance with RMC guidelines for the care and use of experimental animals with approval from the Bar Ilan University and RMC Institutional Animal Care and Use Committee (IACUC).

[0155] Tumor materials were placed in cold DMEM medium supplemented with 10% fetal bovine serum (FBS) and 1:100 penicillin / streptomycin antibiotics and maintained on ice until processing. Within 0.5-2 hours, tumor fragments were cut into 2-3 mm pieces using sterile surgical instruments. Several pieces were used for implantation, and one was formally fixed for histological examination. Typically, several implantations were carried out (subcutaneously (SC) on the flanks, intraperitoneal (IP) implantation, and implantation directly into the mammary tissue). The recipients were 5-8 weeks old immunodeficient NRG mice (NOD-congenic mice harboring the Raglnull mutation (RagltmlMom) on chromosome 2, and the IL2rynull mutation Il2rgtmiwji) on the X chromosome (NOD.Cg-RagltmlMomH2rgtmlWjl / SzJ); Jackson Laboratories) or NSG™ mice (NOD.Cg-25 Prkdcscidl l2rgtmlWjl / SzJ I; Jackson Laboratories) that were female or male according to the patient's sex. For breast cancer PDX models, mice were also supplemented with 17p-estradiol (a steroid hormone; Tocris, Cas#50-28-2), using slow release by osmotic pumps implanted SC (28-day release, 1.08 mg / pellet, Alzet).

[0156] Before implantation, the tumor fragments were coated with Cultrex® Basement Membrane Extract (BME), Type 3 (Trevigen). The BME, Type 3 is a natural extracellular matrix hydrogel (extracted from Engelbreth-Holm-Swarm (EHS) tumor) that polymerizes at 37°C to form a reconstituted basement membrane.

[0157] Pleural fluid is a thin, lubricating fluid found in the pleural cavity, the space between the two layers of the pleura surrounding the lungs. The pleura are thin membranes that cover the lung (termed "visceral pleura") and line the chest wall and diaphragm (termed "parietal pleura"). Normal pleural fluid is clear or pale yellow and contains water, proteins, electrolytes, glucose and a small number of white blood cells. A pleural effusion is an abnormal build-up of fluid in the pleural space. Malignant pleural effusion (MPE) is the accumulation of fluid in the pleural cavity caused by the presence of cancer cells, either from a primary cancer in the pleura or as a result of metastasis from another site. It is a serious condition and often indicates advanced or metastatic disease. Patients with lung cancer, breast cancer, ovarian and lymphoma are most likely to develop MPE. Other causes of MPE include cancer that has spread from the stomach, kidney and colon. To establish PDX models from metastatic sources, patients' pleural fluid was centrifuged at 1500 RPM for 4 min, washed with phosphate-buffered saline (PBS) and cells were collected and counted. Cells (2-3 millions) derived from pleural fluid were resuspended in a volume of 0.05-0.1 mL Cultrex® BME:PBS (1:1) and were injected SC or IP to mice.

[0158] Mice were kept under pathogen-free conditions and received sterilized food and water ad libitum.

[0159] Drug efficacy experiments

[0160] When tumors reached a size of 60-200 mm3, the mice were randomized into 4 or 5 treatment groups, each comprising 5-6 mice. The treatment groups included: (i) vehicle (control); (ii) palbociclib (100 mg / kg); (iii) sunitinib (50 mg / kg); and (iv) palbociclib in combination with sunitinib (palbociclib 100 mg / kg, sunitinib 50 mg / kg). Standard-of-care (SOC) treatment serving as a positive control was used in most of the models. The drugs were administered orally, 5 days per week. The treatment continued until the model reached a tumor size of 1,500 mm3(the maximum allowed by the IACUC) or when clinical signs met those defined by the IACUC, after which the mice were euthanized according to the lACUC-approved protocol, and the tumors were harvested and examined histologically to confirm their human origin and morphological similarity to the corresponding engrafted tumor.

[0161] Patient-derived xenograft development was assessed by palpation and measurements by electronic caliper. Tumor volume was determined by the ellipsoidal formula (width x width x length) / 2. Mice body weight and clinical signs of toxicity were examined twice a week.

[0162] Tumor growth inhibition (TGI) was defined as: l-([mean volume of treated tumors] / [mean volume of control tumors]) and was expressed as a percentage.

[0163] When mice experienced a decrease in body weight, treatment was held for 1-2 days until weight gain.

[0164] RNA extraction from formalin-fixed paraffin-embedded (FFPE) tissue samples

[0165] Total RNA was extracted from 5 to 10 sections of 4 pm thick FFPE sections using the RNeasy® FFPE Kit (QIAGENE) according to the manufacturer's protocol. The concentrations of the extracted RNA were determined using the NanoDrop spectrophotometer, which can quantify and assess purity of RNA from only 1-2 pL of sample (Thermo Fisher Scientific). RNA was then subjected to NanoString analysis (Bruker Nano Group).

[0166] RNA sequencing (RNA-Seq)

[0167] RNA-Seq is a high-throughput sequencing technique used to analyze the transcriptome (the complete set of RNA transcripts, including coding and non-coding RNAs) of a cell or tissue. It provides quantitative and qualitative data on gene expression, transcript structure, and RNA modifications.

[0168] Key principles of RNA-Seq include:

[0169] (i) RNA extraction. RNA is isolated from cells or tissues, typically using chemical methods (e.g., TRIzol reagent) or column-based kits. RNA-Seq focuses on specific RNA types: mRNA sequencing or total RNA sequencing, namely capturing all RNA, including non-coding RNAs. Total or specific RNA species (e.g., mRNA) are selected based on the experiment's goals.

[0170] (ii) RNA selection or enrichment. mRNA can be selected based on polyadenylated (poly- A) tails using oligo-dT beads. Additionally, or alternatively, RNA enrichment may involve ribosomal RNA (rRNA) depletion which removes abundant rRNA to focus on other RNA species.

[0171] (iii) RNA Fragmentation. Long RNA molecules are fragmented into smaller pieces to facilitate sequencing. Fragmentation can be enzymatic, mechanical (e.g., sonication), or chemical.

[0172] (iv) cDNA Synthesis. RNA fragments are reverse transcribed into complementary DNA (cDNA) using reverse transcriptase. This step often involves the addition of primers and adapters for subsequent amplification and sequencing. In some RNA-Seq techniques, the cDNA is fragmented into smaller pieces instead of the original RAN molecules.

[0173] (v) Library Preparation. cDNA fragments are ligated to adapters containing unique sequences for indexing and sequencing. The libraries may be amplified using polymerase chain reaction (PCR) to create sufficient material for sequencing.

[0174] (vi) Sequencing. Libraries are loaded onto a high-throughput sequencing platform (e.g., Illumina, PacBio, or Oxford Nanopore). Each RNA fragment is sequenced to generate short (hundreds of base pairs) or long (thousands of base pairs), depending on the platform) reads, herein termed "RNA-Seq reads", corresponding to the cDNA fragment from which they were derived. Sequencing may follow either single-end or paired-end sequencing methods. Single-end sequencing sequences the cDNA fragments from just one end, whereas paired-end methods sequence from both ends. The depth to which the library is sequenced varies depending on the purpose of using the output data.

[0175] (vii) Read Alignment and Assembly. Sequenced reads are aligned to a reference genome or transcriptome. In cases where no reference is available, reads can be assembled de novo to reconstruct transcripts. Optionally, RNA-Seq reads can be aligned to curated databases of known transcripts such as RefSeq, UCSC genome browser, Ensembl or GENCODE. Non-limiting examples of known transcript references include the human genome assembly (GRCh38) and the mouse genome assembly (GRCm38) maintained by the Genome Reference Consortium (GRC).

[0176] (viii) Quantification of Gene Expression. The abundance of transcripts is quantified by counting the number of reads aligned to each gene or transcript. This data is normalized to account for differences in sequencing depth and transcript length. (ix) Data Analysis. Analysis involves identifying differentially expressed genes, alternative splicing events, transcript isoforms, or novel transcripts. Bioinformatics tools and pipelines, such as STAR, HISAT2, DESeq2, and edgeR, are commonly used.

[0177] For RNA sequencing described in the Example 2 herein, total RNA was extracted from tumor tissues using the TRIzol® method (BIO TRI RNA reagent, Biolab, cat#9010233100, Israel), according to the manufacturer's instructions. TRIzol® reagent is an acid-guanidinium-phenol based reagent designed for the extraction of RNA from various biological samples. The low pH of TRIzol® controls the separation of RNA from DNA and proteins. The guanidinium salt serves as a chaotropic agent to denature proteins and the phenol is used to extract nucleic acids and proteins. After solubilization and homogenization of samples in TRIzol®, the RNA, DNA and proteins were differentially extracted by the addition of a phase separation reagent (chloroform, l-bromo-3-chloropropane (BCP) or 4-bromoanisole (BAN)) solution. The solution separated the RNA away from the DNA and proteins into different layers. The upper, clear aqueous phase mainly contained RNA, and the middle interphase and lower organic phase contained DNA, proteins and lipids. Subsequently, the RNA in the upper aqueous phase was collected by alcohol- based precipitation.

[0178] The extracted RNA was shipped in dry ice to Macrogen Europe BV (Amsterdam, The Netherlands) for poly-A mRNA sequencing., using various RNA sequencing techniques.

[0179] In some embodiments, total extracted RNA was sequenced using the Illumina RNA-Seq protocol. Following RNA extraction and quality check (high RNA integrity is crucial for accurate results), mRNA enrichment was conducted by capturing the poly(A) tail of mRNA using oligo(dT) beads. Generally, the Illumina RNA-Seq protocol comprises the following steps:

[0180] (1) Purified RNA is reverse transcribed into complementary DNA (cDNA) using reverse transcriptase. First-strand synthesis is followed by second-strand synthesis to create doublestranded cDNA.

[0181] (2) For library preparation, the cDNA is fragmented into smaller pieces. Then, short DNA adapters necessary for Illumina sequencing are ligated to both ends of the cDNA fragments (adapter ligation). Unique barcodes (indices) may be added during this step to allow multiplexing of multiple samples in a single sequencing run.

[0182] The library is amplified through limited-cycle PCR to increase the quantity of DNA while introducing minimal bias. The final library is checked for size distribution, concentration, and purity using tools like qPCR, Bioanalyzer, or Tapestation. (3) For sequencing, the library is loaded onto an Illumina sequencing platform. Sequencing-by-synthesis (SBS) technology is used to read the bases of the cDNA fragments, producing short reads. For data analysis, the reads are aligned to a reference genome or transcriptome using tools like HISAT2 or STAR. Transcript abundance is estimated using dedicated software.

[0183] (4) Downstream Analysis: differential expression analysis (e.g., using DESeq2 or edgeR), is performed.

[0184] Specifically, RNA sequencing was performed as a paired end read on an Illumina TrueSeq™ platform. RNA-Seq reads libraries were prepared using a TrueSeq™ stranded Total RNA LT sample prep kit (Illumina®, San Diego, CA, USA), a robust, highly scalable whole- transcriptome analysis. Sequencing depth was ~30 M reads / sample. Raw reads were trimmed using fastp 0.20.0 (Chen et al., Bioinformatics 34(17):i884-i890, 2018) and aligned (mapped) to the Genome Reference Consortium Mouse Build 3 (GRCm38) assembly (a reference genome assembly) using STAR 2.6.0c (Dobin et al., Bioinformatics, 29(1):15-21, 2013).

[0185] NanoString Analysis

[0186] NanoString analysis is a method for gene expression profiling in biological samples, particularly when working with targeted gene panels. While it lacks the comprehensive coverage of RNA-Seq, it is advantageous for being simple, reproducible, and robust.

[0187] This technology enables the direct quantification of gene expression levels. It employs a unique digital barcoding system that allows the detection and counting of specific RNA molecules without requiring reverse transcription or amplification, which helps maintain data accuracy and reproducibility. The method is suitable for detecting the expression of up to 800 genes in a single reaction with high sensitivity and linearity across a broad range of expression levels.

[0188] The assay employs a pair of custom-designed probes, 30-50 bp each, which are complementary to each target RNA: (i) a capture probe, biotin-labeled to immobilize the target RNA on a surface for counting; and (ii) a reporter probe, carrying a unique fluorescent barcode specific to the RNA sequence. The reporter probe's barcode consists of fluorescently labeled segments arranged in a pattern unique to the gene of interest. The unique, fluorescently labeled molecules "tag" specific RNA targets, and serve as color-coded "barcodes" that can be read (identified) and counted (quantified) directly by the nCounter® system, providing an accurate representation of gene expression levels.

[0189] During the assay, these pairs of probes hybridize directly to the target RNAs in solution, eliminating the need for enzymatic processing. Direct hybridization avoids errors associated with PCR-based methods. Once hybridization is complete, the hybridized complexes are immobilized onto a streptavidin-coated surface of the cartridge. Excess probes are removed, and the cartridge is then placed in a digital analyzer for image acquisition and data processing. Hundreds of color codes designating mRNA targets of interest are directly imaged on the surface of the cartridge. Each barcode represents a single RNA molecule counted as a discrete event, thereby offering precise measurements. The assay outputs a digital readout that provides an absolute quantification of RNA abundance.

[0190] In embodiments described herein, NanoString nCounter® Gene Expression Assay (NanoString Technologies, Inc. Seattle, WA), was utilized, consisting of two parts, the nCounter® Prep Station or cartridge, and the digital analyzer. Probes design and construction were performed in collaboration with NanoString Technologies. nCounter Elements™ reagents that enabled analysis with color-coded molecular barcodes were purchased from NanoString Technologies, and probes were synthesized by Integrated DNA Technologies (IDT) (Coralville, IA). Code-Set assembly and RNA / Probe hybridization, RNA quality assessment and data analysis were performed as described in Chang et al., 2018 (J. Mol. Diagn., 20(l):63-77, 2018). In brief, a total of 200 ng RNA was mixed with the probes, the reagent nCounter Elements™ TagSet and hybridization buffer (NanoString Technologies), following the manufacturer's protocol. The mix was incubated for 20 hours at 67°C in a Bio-Rad C1000 Touch thermal cycler (Montreal, QC, Canada). Sample processing was performed using the nCounter® Prep Station, and RNA counting was performed using the nCounter® digital analyzer. The geometric means (average or mean) of the reference (housekeeping) transcripts DNAJC14, ERCC3, GUSB and MRPL19 were used to calculate a normalization factor and determine the RNA quality. Housekeeping genes encode proteins that are usually essential for the maintenance of cellular function and often remain constant under most experimental conditions. The bundle of selected housekeeping genes has low variance across a set of banked tumor samples from a variety of cancer types. Raw counts directly from the nCounter® digital analyzer were subjected to normalization using the internal positive spike-in controls by nSolver analysis software version 4.0 (NanoString Technologies), followed by probe-specific background correction. The internal positive spike-in controls in the NanoString nCounter® Gene Expression Assay are synthetic RNA molecules (not naturally occurring in the biological sample and distinct from the sample's RNA) that are added to the assay to serve as quality control and normalization references. The spike-ins are added to the biological sample during the hybridization step at precise, pre-determined concentrations, allowing accurate calibration and performance checks. Each spike-in RNA has a unique reporter barcode, similar to the target RNA molecules, ensuring consistent measurement. They are processed and read by the nCounter® system, just like the RNA from the sample, and their known quantities allowfor comparison with the observed counts. These controls are critical for ensuring the accuracy, reliability, and reproducibility of the assay results as they provide benchmarks for assay performance, error detection, and data normalization.

[0191] Differential expression analysis

[0192] Differential gene expression (DGE) analysis identifies genes whose expression levels differ significantly between experimental conditions, such as treated vs. untreated samples or healthy vs. diseased states. The analysis focuses on quantifying changes in RNA abundance across groups to discover biological insights. Key Steps in DGE Analysis include:

[0193] (i) Data normalization: adjusting raw data to account for technical variability (e.g., differences in sequencing depth or library size).

[0194] (ii) Statistical modeling: applying statistical tests or models to detect significant changes in gene expression while accounting for biological and technical variability.

[0195] (iii) Multiple testing correction: correcting p-values to control for false positives when testing thousands of genes simultaneously.

[0196] (iv) Interpretation: identifying significantly upregulated or downregulated genes and linking them to biological pathways or processes.

[0197] Software like DESeq2, edgeR, and NanoStringDiff are commonly used, with choices depending on the data type (RNA-Seq, NanoString, microarray) and research objectives.

[0198] The DGE technique was applied to uncover genes that were the most differentially expressed in patients who responded to combo treatment and in patients who did not. For this purpose, RNA-Seq reads were obtained from PDX samples of various human cancers. Differential analysis of count data, including normalization and statistical analysis, were performed using the DESeq2 package (DESeq2 1.30.1; Love et al. Genome Biology, 15:1-21, 2014). All data analysis and visualization were performed using R software (version 4.1.2) for data manipulation, calculation and graphical display.

[0199] DESeq2 works with raw read count data, presented in a table where rows represent genes (or transcripts) and columns represent samples. This table reports, for each sample, the number of reads that have been assigned to each gene. DESeq2 uses negative binomial generalized linear models to model count data, accounting for biological variability and technical noise, and estimates size factors for normalization and dispersion parameters for modeling gene-specific variability. DESeq2 corrects for differences in sequencing depth and library sizes using a median-of-ratios method and incorporates shrinkage estimators to improve accuracy for genes with low counts or high dispersion. Results may be visualized using e.g., heatmaps.

[0200] Workflow of DESeq2 included the following steps:

[0201] (i) Data preparation: Import of raw count data and experimental design metadata into R.

[0202] (ii) Creating DESeq Data Set object: organizing data into a structured object for analysis.

[0203] (iii) Counts normalization: raw counts are normalized for size factors and library differences.

[0204] (iv) Parameters estimation: dispersion is estimated, and the negative binomial model is fitted to the data.

[0205] (v) DGE analysis performance: statistical testing is applied to identify differentially expressed genes.

[0206] (vi) Visualization and Interpretation: plots (e.g., volcano plots, heatmaps) are generated to visualize results and identify biologically significant findings.

[0207] The quantitative analysis focused on the strength rather than the mere presence of differential expression.

[0208] For ensuring data quality and making informed decisions before proceeding with differential gene expression analysis, several key tools were employed by DESeq2, including variance stabilizing transformation (VST), regularized log transformation (rlog) and principal component analysis (PCA).

[0209] In raw count data, genes with high counts often have higher variance, making comparisons difficult. VST reduces this effect by making variance roughly constant across genes. It transforms raw count data to stabilize variance across the range of mean values and enables meaningful visualization methods, such as heatmaps or PCA, by normalizing data variability. The command rlog performs a Iog2 scale transformation in a way that compensates for differences between samples for genes with low read count and normalizes between samples for library size. Because some genes are not expressed (detected) in some samples, their count is 0. As Iog2(0) in R triggers errors by some functions, 1 is add to every count value to create 'pseudo counts.' The lowest value then is 1, or 0 on the Iog2 scale (Iog2(l) = 0).

[0210] A principal component analysis (PCA) reduces the dimensionality of data by identifying the major axes (principal components) that capture the most variation between samples. This enables to visualize how samples cluster based on their gene expression profiles. PCA was used to assess sample relationships and ascertain differences between samples. It identifies outliers or batch effects. PCA visualizes the primary sources of variability in the data.

[0211] Principal component analysis transforms a large set of variables into a smaller one that still contains most of the information in the large set. The grouping of the samples in PCA is not specified. However, if the experiment is well controlled and has worked well, then replicate samples should cluster closely, whilst the greatest sources of variation in the data should be between treatments / sample groups. In running the PCA, it is desired to first normalize the data for library size and transform to a log scale, and this is done with DESeq2 by the commands VST and rlog.

[0212] PCA coordinates were calculated by plotPCA function (a function that makes a PCA plot from an expression set or matrix; Deseq2 package) and drawn by ggplot2, an R plotting package dedicated to data visualization.

[0213] Boxplot (also termed "whiskers plot") is a method used for depicting groups of numerical data through their quartiles, graphically. This type of diagram shows the quartiles in a box and a line extending from the lowest to the highest value. These may also have some lines extending from the boxes (whiskers) which indicate the variability outside the lower and upper quartiles. Outliers are indicated as individual points. The five-number summary in the boxplot is minimum, maximum, median, first quartile, and third quartile. The box plot distribution explains how tightly the data is grouped, how the data is skewed, and what is the symmetry of data. Boxplots of 196 genes were drawn by ggplot2.

[0214] For differential expression analysis using DESeq2 1.30.1, adjusted p-values < 0.00001 and absolute fold change >2 were used. In addition, to find the genes that provide optimal separation, only up to 10% overlap in values between the groups was allowed. The statistical tests and this separation were made based on a comparison between responders and nonresponders.

[0215] Complex heatmaps are efficient graphical, color-coding-based tools for visualizing associations between different sources of data sets and reveal potential patterns. The heatmap of 196 genes originating from cancerous tissues of subjects that responded or were unresponsive to combo therapy was obtained using ComplexHeatmap package, with default parameters on transformed data. The default color scheme is "blue-white-red" which is mapped to the minimal- mean-maximal values, respectively, in the matrix.

[0216] Gene identification

[0217] The Ensembl genome database is a comprehensive, open-access platform that provides high-quality genomic data for a wide range of species, including vertebrates. Ensembl provides annotated reference genomes, including gene structures, regulatory elements, and variant information. Ensembl assigns unique alphanumeric identifiers to transcripts (e.g., ENST00000456328) to distinguish them. These identifiers, termed herein "Ensembl Gene IDs", represent specific RNA transcripts for genes.

[0218] The Ensembl Gene ID precisely identifies the gene, is species-specific and consistent within the Ensembl system. However, it does not directly convey the biological function or gene name. A gene symbol is a short, standardized abbreviation representing a gene by a string of letters and numbers, e.g., BRCA2 (for "Breast Cancer 2"). A gene symbol provides a human- readable shorthand for a gene, but it is not unique across species or even within a species (e.g., different organisms may use the same symbol for unrelated genes). A standard gene name is a descriptive full name for a gene, providing context about its function, associated protein, or role in biology. The Standard Gene Name is typically longer and more descriptive, e.g., Breast Cancer Type 2 Susceptibility Protein (for BRCA2) and offers a more detailed explanation of the gene's function or biological role. It may vary across species or research contexts but generally aligns with gene symbols.

[0219] An R package is a collection of R functions, data, and documentation bundled together in a standardized format to extend the functionality of R, a programming language widely used for statistical computing and data analysis. biomaRt is an R package that provides a convenient interface to access and retrieve data from the BioMart system, a data management platform commonly used for querying biological databases. BioMart is integrated into Ensembl, making biomaRt a particularly useful interactive tool that affords easy programmatic access to the Ensembl database for mapping transcript IDs to gene symbols and gene descriptions (i.e., performing gene annotation). The biomaRt package was used in embodiments described herein for converting Ensembl transcript IDs into more meaningful and human-readable gene symbols and descriptions.

[0220] Identification of enriched biological pathways

[0221] Database for Annotation, Visualization and Integrated Discovery (DAVID) is a widely used bioinformatics resource for functional annotation of gene lists. It helps to explore the biological meaning of large gene datasets by identifying enriched biological pathways, functions, and processes associated with the genes.

[0222] Biological pathways refer to collections of genes or proteins that work together to perform specific biological functions (e.g., apoptosis, metabolism, or signal transduction). Enrichment analysis determines whether certain biological pathways or gene sets are overrepresented among the differentially expressed genes (DEGs) compared to what would be expected by chance. For example, if many DEGs are involved in the "immune response" pathway, this pathway is considered enriched. Enriched pathways help to identify the molecular mechanisms underlying the observed differences and highlight potential targets for therapeutic intervention.

[0223] DAVID was used to annotate the genes identified as significantly differentially expressed with functional information (e.g., their role in pathways, cellular components, or molecular functions), to visualize enriched biological themes or clusters of functionally related genes, and to identify the key pathways or processes associated with these DEGs. For this purpose, a list of significantly upregulated and / or downregulated genes (e.g., in gene symbols or Ensembl IDs) was uploaded to DAVID, and the tool performed enrichment analysis using various databases (e.g., KEGG, GO, UniProt). DAVID identified enriched terms or pathways, such as "Immune response", "Apoptotic signaling pathway" and "Cell cycle regulation". It provided p-values, enrichment scores, and detailed pathway visualizations.

[0224] After certain enriched biological pathways were identified, the focus was made on 29 selected genes related to these pathways. To visualize the expression levels of these 29 selected genes across multiple samples, a heatmap of the genes was generated by ComplexHeatmap R package, which provides the addition of annotations, clustering, and other advanced features to enhance data interpretation. The heatmap was generated using the package's default settings, meaning no significant customizations were applied to clustering, color scales, or other options. The input data for the heatmap consisted of normalized expression counts (e.g., from RNA-Seq) or other high-throughput experiments). Normalization adjusts the raw counts to account for differences in sequencing depth or library size.

[0225] The data was scaled, meaning the values were transformed (e.g., Z-scores) so that the expression levels are comparable across genes or samples. Scaling ensures that patterns of expression differences, rather than absolute values, are emphasized.

[0226] Statistical analysis

[0227] Comparison of gene expression levels between PDX groups was performed on samples taken on the last day of treatment.

[0228] The Wilcoxon-Mann-Whitney test for independent samples was used for testing the statistical significance of the difference between treatment groups regarding the primary endpoint (tumor volume).

[0229] The Kaplan-Meier analysis was performed across PDX models when tumors reached a volume of 1,500 mm3(the maximum tumor volume allowed by the ethics permit). Survival analysis and Kaplan-Meier plots were made using R version 4.0.2. All tests were 2-tailed; p-value of <0.05 was considered statistically significant.

[0230] The significance of the difference between the monotherapies and the combination therapy was assessed using a log-rank test. Log-rank test is a statistical method used to compare survival curves of two or more groups. It determines whether the differences in survival times between the groups are statistically significant. The null hypothesis (Ho) for the log-rank test is that there is no difference in the survival distributions of the groups (monotherapy vs. combination therapy). If the test yields a low p-value (e.g., p<0.05), the null hypothesis is rejected, indicating a significant difference in survival outcomes.

[0231] EXAMPLE 1

[0232] Categorization of tumor models based on responsiveness to combination therapy

[0233] Twenty-five PDX models were divided into 4 treatment groups for assessing therapy efficacy: (i) vehicle (control); (ii) palbociclib (100 mg / kg); (iii) sunitinib (50 mg / kg); and (iv) combo treatment, i.e., palbociclib together with sunitinib (palbociclib 100 mg / kg, sunitinib 50 mg / kg). For each cancer, 2-9 tumors were used (grown in different PDX mice). Tumors were collected at the end of the experiments, and flash frozen. RNA was extracted from the frozen tumor tissues and sequenced using the RNA-Seq technique as described in Materials and Methods.

[0234] The PDX models were categorized based on their response to treatment. Response to combo treatment was evaluated in terms of "synergism" and "overall response". A synergistic response was determined if both of the following conditions were met: (i) detection of statistically significant superiority of the palbociclib plus sunitinib combination over palbociclib and sunitinib, each alone, as monotherapies; and (ii) detection of statistically significant superiority of the response to combo treatment over the average effect of palbociclib and sunitinib as monotherapies. If only the first condition was met, the effect was considered additive.

[0235] For overall response, tumors that received combo treatment were compared to control tumors (untreated ("naive") tumors, i.e., patient-derived xenografts originating from tumors that developed in the patients and have not been exposed to treatment in the mice). Categorization based on the overall response was conducted in order to identify biomarkers for a potential response to the combo treatment, not only in a synergistic manner. Table 1 lists all samples used for the analysis.

[0236] Nineteen (19) models showed synergistic response, and 6 models did not show synergism but responded to combo treatment to some extent (with p-value > 0.05): 3 models responded with percent tumor growth inhibition (%TGI) > 70 and were thus considered as "responder models" (models RA-373, RA-292C and RA-360). The other 3 models responded to combo with %TGI < 70 and were considered as "non-responder models" (models RA-374, RA-179J, RA-441).

[0237] The tumors exhibited varying durations of response: in responder models the mean (SEM) duration was 61 days, whereas a mean duration of 21 days was observed in non-responder models (RA-374, RA-179J, RA-441). Among the non-responders, model RA-441 had a relatively longer response duration (45 days) compared to RA-179J (7 days) and RA-374 (10 days), thus it was considered as "mid-responder". Table 1. List of PDX models and their responsiveness to combo therapy

[0238] * Responders, but not synergistically EXAMPLE 2

[0239] Differential expression analysis (DEA) based on DESeq2 technique

[0240] Differential expression analysis (DEA) aimed at identifying significant gene expression variations was conducted to uncover genes that were the most differentially expressed in PDX models who responded to combo treatment compared to non-responsive models. For the initial analysis, only the synergistically responder models (n=19) were included in the "responders" group. The non-synergistically responders (n=3) were not taken into consideration.

[0241] RNA was extracted from frozen tumor tissues and sequenced using the RNA-Seq technique. Gene expression analysis and visualization was conducted using the DESeq2 package as described in Materials and Methods. Differential expression analysis of responder samples versus non-responder samples resulted in 196 differentially expressed genes (DEGs), wherein 150 genes were upregulated, and 46 genes were downregulated in non-responders compared to responders. The results are shown in Figs. 1A-1B.

[0242] The pathway enrichment analysis of these DEGs highlighted several significant pathways relating to lysosome, phagosome, endosome, transcription, cell surface, immunity and antigen processing and presentation.

[0243] For clinical assay, a genomic biomarker was established based on 29 genes, which were selected based on two principles: (1) they showed a very high separation between responders and on-responders across all samples; and (2) they were representative of groups of genes that coded for protein families or protein complexes that function in one or more of the above- mentioned significant biological pathways.

[0244] The heatmap and boxplots presented in Figs. 2A and 2B, respectively, show the differential expression between responder models and non-responder models with respect to the 29 selected genes. Interestingly, model RA-441, which was considered as "mid-responder" based on its efficacy performances, showed "mid phenotype" also with regards to the expression of these 29 selected genes. This correlation between expression and function strengthens the hypothesis that these genes are indeed related to the tumor's inclination to respond to the combo treatment.

[0245] A complete list of 196 DEGs is listed in Table 2 (R: responsive; NR: non-responsive). The first 29 genes listed in Table 2 were selected for establishing the genomic biomarker.

[0246] As noted above, the DEA was based on a comparison between responders and non- responders, excluding the group of non-synergistic responder models (RA-373, RA-292C and RA- 360). After the significantly differentially expressed set of 29 genes was selected, their expression levels were evaluated in the non-synergistic responder models. These genes presented the same expression pattern as responders, meaning that these genes predicted overall response, regardless of whether it is in a synergistic manner or not. This fact is of high practical importance for making the decision whether a patient may benefit from the treatment or not.

[0247] Table 2. List of 196 differentially expressed genes (DEGs)

[0248] EXAMPLE 3

[0249] Differential expression analysis (DEA) based on NanoString technique

[0250] To validate the potential of the genomic biomarker genes determined based on DEA using the RNA-Seq / DESeq2 platforms, DEA was conducted using the NanoString analysis platform (see Material and Methods supra). This method enabled the measurement of subtle changes from small and degraded samples such as formalin-fixed paraffin-embedded (FFPE) tissue specimens. The expression levels assessment was focused on 15 genes out of the 29 genes identified in DEA conducted using the RNA-Seq / DESeq2 technique. The results obtained for 17 PDX samples are shown in Fig. 3.

[0251] As shown in Fig. 3, the expression of the majority of these 15 genes (at least 10 genes) was significantly higher in the non-responders compared to the responder models. Of note, as in the RNA-Seq / DESeq2 studies, the RA-441R model presented gene expression levels which were between responders and non-responders, supporting the hypothesis that this set of genes is relevant to establish responsiveness.

[0252] These results imply that by testing the expression levels of at least 15 genes, one can predict the response to combo treatment for a given patient, wherein statistically significantly high expression level of at least 10 DEGs accounts for a non-responder, or statistically significantly low expression level of at least 10 DEG accounts for a responder.

[0253] EXAMPLE 4

[0254] Responsiveness to monotherapies in predicting responsiveness to combo therapy

[0255] The expression levels of the 29 DEGs across 22 PDX models, including combo therapy responder, mid- and non-responder models, were assessed following treatment with sunitinib alone (sunitinib monotherapy) or palbociclib alone (palbociclib monotherapy). The results are shown in Figs. 4A-4B. The expression levels of these genes in sunitinib monotherapy responsiveness (R) and non-responsiveness (NR) in the combo therapy responder models were compared to their expression levels in the combo non-responder models (Fig. 4A). Likewise, the expression levels of the genes in palbociclib monotherapy responsiveness (R) and non- responsiveness (NR) in the combo therapy responder models (Combo-R) were compared to their expression levels in the combo non-responder models (Combo-NR; Fig. 4B).

[0256] As seen in Fig. 4A, the expression of these 29 genes was not significantly different for sunitinib R and NR in the Combmo-R models. In fact, a trend was observed for several genes (marked in square frames in Fig. 4A), whereby responsiveness to sunitinib monotherapy in Combo-R was correlated with high expression levels compared to sunitinib non-responsiveness, which is opposite to the trend of expression of these genes in the combo responder models. Likewise, as seen in Fig. 4B, the expression levels of the 29 genes in palbociclib monotherapy non-responsiveness were not different than in palbociclib monotherapy responsiveness. Unexpectedly, the gene PSME2 showed opposite trend in palbociclib monotherapy compared to combo therapy (marked in red in Fig. 4B), in which response to palbociclib was correlated with high expression level, while combo responsiveness was correlated with low expression of these genes. Other genes presenting similar opposite expression patterns are also marked in red frames in Fig. 4B.

[0257] There was no significant "trend" in the expression levels of specific genes nor of the whole set of 29 genes in monotherapy treated cancers, which could predict responsiveness to combination therapy. For example, it is shown in Figs. 5A-5B herein that in two combo non- responsive cancer models (RA-179J and RA-441), only 2 genes in a set of the 29 genes were upregulated up to 2.5-fold following sunitinib monotherapy: AC009133.17 ir\ RA-441R model and GPR132 in RA-179J model. Interestingly, following palbociclib monotherapy in these two combo non-responsive models, the expression of some of the 29 genes was reduced. This is opposite to the trend wherein high gene expression indicated non-responsiveness to combo therapy.

[0258] Thus, it is unexpected, and certainly unique, that tumors which respond to monotherapies present high expression levels of genes that are under expressed or downregulated in tumors that are responsive to combo treatment.

[0259] EXAMPLE 5

[0260] Genomic biomarker feasibility evaluation

[0261] To confirm the predictivity potential of a set 15 genes identified as significantly differentially expressed in combo responders versus non-responders in Example 3 above, the mRNA expression levels of these 15 genes were evaluated in 4 new PDX models (2 of breast cancer and 2 of pancreas cancer patients) by NanoString analysis. These PDX models, termed herein "test models", are to be distinguished from the models described in Examples 1-3 that served for identifying the relevant differentially expressed genes of interest, herein the "training models". The expression levels of 15 genes in these new models are shown in Fig 6. The models were categorized as "responders" (RA-200 and RA-393), "non-responder" (RA-198), and "midresponder" (RA-454), based on the similarity of their gene expression signature to that observed in the "training models". Next, the ability of the set of 15 genes to serve as a genomic biomarker that reliably predicts the efficacy of combo treatment was evaluated in these 4 PDX models. Mice were treated with a combination of palbociclib and sunitinib, and tumors were handled as described in Materials and Methods. The results are shown in Figs. 7A-7D and Table 3.

[0262] As shown, models RA-393 and RA-200 (Figs. 7A and 7B, respectively) responded to the combo treatment (%TGI = 91 and 82), model RA-198 did not respond (%TGI = 56) (Fig. 7D), and model RA-454 showed mild responsiveness (%TGI = 79 with duration of response shorter than other responder models) (Fig. 7C), in line and in strong correlation with the gene expression signature of the 15-gene set in these 4 models.

[0263] Interestingly, the two breast cancers, as well as the two pancreatic cancers, obtained from different patients, reacted differently to the combo treatment: one breast cancer model satisfactorily responded to the combo treatment (RA-393), whereas the other breast cancer model (RA-198) was not responsive. These findings clearly show that responsiveness to anticancer treatment in general, and to combo treatment in particular, cannot and should not be a priori determined based on speculations and prior knowledge. Instead, the genetic signature of each individual in each circumstance should serve as the most reliable and confident biomarker for predicting treatment responsiveness. The genomic biomarker disclosed herein may serve as an important tool for tailoring the proper personalized treatment protocol.

[0264] Table 3. Tumor growth inhibition (TGI) and treatment duration in PDX models treated with a combination of palbociclib and sunitinib (combo treatment) EXAMPLE 6

[0265] Biomarker expression in tumors derived from patients in clinical trial

[0266] The phase 1 dose-escalation clinical trial investigating the palbociclib plus sunitinib combination therapy enrolled 35 patients with diverse tumor diagnoses who had failed at least one prior treatment. Tumor samples were collected from 8 patients before treatment initiation. RNA from these tumor samples was isolated and subjected to NanoString analysis of the 29 genes previously determined to be differentially expressed in responders to combo therapy versus non-responders (see Example 2). The 8 patients included 4 who responded to treatment and 4 who did not. Table 4 describes the patient characteristics, treatments received, and responses for these 8 patients. Notably, the patients who responded had salivary gland and nasal cavity adenoid cystic carcinoma (ACC) or skin squamous cell carcinoma (SCC), whereas those who did not respond had other cancers (colon, pancreas, sarcoma).

[0267] The NanoString analysis revealed that the expression of 11 genes correlated with treatment response, with lower expression levels in responders, and higher levels in non- responders (Fig. 8). These 11 DEGs were: TMEM140, FGR, CTA-126B4.7, HLA-DRA, FGD2, LINC00528, FERMT3, GPR132, RASGRP3, IFI30 and POU2F2. Further evaluated is the transcript RP11-134L10.1 of a non-coding RNA.

[0268] A predetermined threshold for these 11 genes and the transcript, predicted patient responsiveness with 75% accuracy (6 correct predictions out of 8), and 5 of the genes demonstrated 100% accuracy in predicting non-responsiveness when expression exceeded a specific threshold. The threshold was set according to the highest expression value of genes in Responders. However, if the variability in a specific gene expression level between Responders sample was high, then the threshold was set such that in at least 3 out of 4 Responders samples the gene expression was lower than the threshold. The remaining 17 genes of the 29 DEGs that underwent NanoString analysis included 14 genes that showed no correlation with treatment response and 3 genes for which the expression levels were below the threshold.

[0269] These results suggest that these 12 DEGs constitute promising biomarkers for predicting responsiveness to the combination therapy approach. Table 4. Patient characteristics, treatment, and response aAt screening.

[0270] Abbreviations: ACC: adenoid cystic carcinoma; PD: progressive disease; PO: per oz; PET-CT: positron emission tomography-computed tomography; PR: partial response; SCC: squamous cell carcinoma; SD: stable disease.

Claims

WHAT IS CLAIMED IS:

1. A genomic biomarker comprising a set of 5-50 genes, for predicting responsiveness to an anti-cancer treatment comprising a combination of palbociclib and sunitinib.

2. The genomic biomarker of claim 1, comprising a set of 5-30 genes, preferably 5-15 genes.

3. The genomic biomarker of claim 1 or 2, wherein the genes are differentially expressed in a cancer which is responsive to a combination therapy of palbociclib and sunitinib versus a cancer which is not responsive to the combination therapy.

4. The genomic biomarker of any one of claims 1 to 3, wherein the gene set comprises at least 5 genes selected from the genes designated as SP140, FGR, HLA-DRA, LINC00528, GPR132, ABCA6, SELPLG, LIN00494, RASGRP3, CIITA, NCF4, PTPN22, FGD2, FERMT3, JAK3, POU2F2, IFI30, LAMP3, TMEM140, PSMB9, ETV7, CTA-126B4.7, TMC8, CCDC88B, PSMB8-AS1, TFEB, PSME2 or AC009133.17, and optionally further comprises the transcript RP11-134L10.1.

5. The genomic biomarker of claim 4, comprising at least 5 genes selected from CIITA, TFEB, CCDC88B, CTA-126B4.7, ETV7, FERMT3, FGD2, FGR, HLA-DRA, RASGRP3, SP140, LINC00528, GPR132, IFI30, POU2F2 and TMEM140, and the transcript RP11-134L10.1.

6. The genomic biomarker of any one of claims 1 to 5, wherein the genes are identified based on a differential expression analysis of cancerous tissues from at least 5 subjects identified as responsive to the combination therapy, and cancerous tissues from at least 5 subjects identified as non-responsive to the combination therapy, with each gene presenting at least 2-fold change in expression between combination therapy responders and nonresponders across all cancerous tissues analyzed.

7. A method for predicting, in a subject in need thereof, responsiveness to a combination anti-cancer therapy comprising palbociclib and sunitinib, comprising:(i) obtaining a sample of cancerous tissue from the subject;(ii) subjecting the sample to gene expression profiling;(iii) evaluating the gene expression level of each gene in a predetermined set of 5-50 genes;(iv) comparing the expression level of each gene in the predetermined set to a corresponding first reference value and / or to a corresponding second reference value; and(v) determining that the subject is unlikely to respond to the combination therapy if the expression levels of at least some of the genes are higher by at least 2-fold compared to their corresponding first reference values; or(vi) determining that the subject is likely to respond to the combination therapy if at least 5 of the genes are not expressed and / or the expression levels of at least some of the genes are lower by at least 2-fold compared to their corresponding second reference values, wherein a corresponding first reference value represents the score of the expression levels of a predetermined gene across 5-10 different cancerous tissues from subjects identified as responsive to the anti-cancer therapy, and a corresponding second reference value represents the score of the of the expression levels of the same gene across 5-10 different cancerous tissues from subjects identified as non-responsive to the anti-cancer therapy.

8. A method for treating a cancer in a subject intended for a combination therapy comprising palbociclib and sunitinib, comprising:(i) obtaining a sample of cancerous tissue from the subject;(ii) subjecting the sample to gene expression profiling;(iii) evaluating the gene expression level of each gene in a predetermined set of 5-50 genes;(iv) comparing the expression level of each gene in the set to a corresponding reference value; and(v) administering to the subject a therapeutically effective amount of the combination therapy provided that least 5 of the predetermined genes are not expressed and / or the expression levels of at least some of the predetermined genes are lower by at least 2-fold compared to their corresponding reference values, wherein a corresponding reference value is a score of the expression levels of a predetermined gene across 5-10 different cancerous tissues from subjects identified as non-responsive to the anti-cancer therapy.

9. The method of claim 7 or 8, wherein the set of 5-50 genes is determined based on a differential expression analysis of cancerous tissues from 5-10 subjects identified as responsive the combination therapy, and cancerous tissues from 5-10 subjects identified as non-responsive to the combination therapy, with each gene presenting at least 2-fold change in expression between combination therapy responders and non-responders across all cancerous tissues analyzed.

10. The method of any one of claims 7 to 9, wherein at least some of the predetermined genes is at least 80% of the genes.

11. The method of any one of claims 7 to 10, wherein the set of predetermined genes comprises 5-30 genes, preferably 5-15 genes.

12. The method of claim 11, wherein the set of predetermined genes comprises as least 5 genes selected from the genes designated as SP140, FGR, HLA-DRA, LINC00528, GPR132, ABCA6, SELPLG, LIN00494, RASGRP3, CIITA, NCF4, PTPN22, FGD2, FERMT3, JAK3, POU2F2, IFI30, LAMP3, TMEM140, PSMB9, ETV7, CTA-126B4.7, TMC8, CCDC88B, PSMB8-AS1, TFEB, PSME2 or AC009133.17, and optionally further comprises the transcript RP11-134L10.1.

13. The method of claim 12, wherein the set of predetermined genes comprises at least the genes CIITA, TFEB, CCDC88B, CTA-126B4.7, ETV7, FERMT3, FGD2, FGR, HLA-DRA, RASGRP3, SP140, LINC00528, GPR132, IFI30, POU2F2 and TMEM140, and the transcript RP11- 134L10.1.

14. The method of any one of claims 7 to 13, wherein the cancerous tissue is of gastric cancer, colon cancer, cholangiocarcinoma, pancreas cancer, ovarian cancer, breast cancer, lung cancer, Ewing sarcoma, neuroendocrine cancer, melanoma, hepatocellular carcinoma or carcinosarcoma.

Images