Erbb-2 and erbb3 targeting agents for use in the treatment of malignant cells or cancers that have elevated NRG1 expression

Abstract

The invention relates to the field of ErbB-2 and / or ErbB3 targeting agents, such as antibodies binding ErbB-2 and / or ErbB3. In particular it relates to the field of therapeutic (human) antibodies for the treatment of a ErbB-2 / ErbB-3 positive cell, tumor or cancer. More in particular it relates to treating malignant cells, tumors or cancers comprising high NRG1 expression levels.

Description

[0001]Title: ErbB-2 and ErbB3 targeting agents for use in the treatment of malignant cells or cancers that have elevated NRG1 expression The invention relates to the field of ErbB-2 and / or ErbB3 targeting agents, such as antibodies binding ErbB-2 and / or ErbB3. In particular it relates to the field of therapeutic (human) antibodies for the treatment of a ErbB-2 / ErbB-3 positive cell, tumor or cancer. More in particular it relates to treating malignant cells, tumors or cancers comprising high neuregulin-1 (NRG1) expression levels. The receptors for all NRG1 isoforms are the ErbB family of tyrosine kinase transmembrane receptors. The family is also referred to as the human epidermal growth factor (EGF) receptor family (HER). The family has four members: ErbB (Erythroblastoma)-1, ErbB-2, ErbB-3 and ErbB-4. Epidermal growth factor (EGF) receptor (EGFR, ErbB1, or HER1). The receptors (reviewed in Yarden Y, Pines G (2012). The ERBB network: at last, cancer therapy meets systems biology. Nat Rev Cancer 12(8):553–63) are widely expressed on epithelial cells. Upregulation of HER receptors or their ligands, such as heregulin (HRG) or epidermal growth factor (EGF), is a frequent event in human cancer (Wilson TR at al. (2012). Widespread potential for growth-factor-driven resistance to anticancer kinase inhibitors. Nature 487(7408):505–9). Overexpression of ErbB-1 and ErbB-2 in particular occurs in epithelial tumors and is associated with tumor invasion, metastasis, resistance to chemotherapy, and poor prognosis (Zhang H et al. (2007). ErbB receptors: from oncogenes to targeted cancer therapies. J Clin Invest. 117(8):2051– 8). In the normal breast, ErbB-3 has been shown to be important in the growth and differentiation of luminal epithelium. For instance, loss / inhibition of ErbB-3 results in selective expansion of the basal over the luminal epithelium (Balko JM et al. (2012). The receptor tyrosine kinase ErbB3 maintains the balance between luminal and basal breast epithelium. Proc Natl Acad Sci U S A.109(1):221–6). Binding of ligand to the extracellular domain of the receptor tyrosine kinase (RTK) induces receptor dimerization, both between the same (homodimerization) and different (heterodimerization) receptor subtypes. Dimerization can activate the intracellular tyrosine kinase domains, which undergo autophosphorylation and, in turn, can activate a number of downstream pro-proliferative signaling pathways, including those mediated by mitogen-activated protein kinases (MAPK) and the prosurvival pathway Akt (reviewed in Yarden Y and Pines G, 2012). No specific endogenous ligand has been identified for ErbB-2, which is therefore assumed to normally signal through heterodimerization (Sergina NV et al. (2007). Escape from HER- family tyrosine kinase inhibitor therapy by the kinase-inactive HER3. Nature. 445(7126):437–41). ErbB-3 can be activated by engagement of its ligands. These ligands include but are not limited to neuregulin (NRG) / heregulin (HRG). High NRG1 expression arising from autocrine signaling or gene amplification is associated with poor prognosis in certain cancers, and resistance to standard therapies. There is thus an unmet need for treatment that addresses cancers with elevated NRG1 levels. To date, there are at least 37 expressed mRNA isoforms known from the NRG1 gene. Proteolytic processing of the extra cellular domain of the transmembrane NRG1 isoforms release soluble factors. HRG1-β1 is one of the proteins encoded by the gene. It contains an Ig domain and an EGF-like domain that binds to receptor tyrosine kinases ErbB-3 and ErbB-4. The NRG1 gene and the isoforms are known under a number of different aliases such as: Neuregulin 1; Pro-NRG1; HRGA; SMDF; HGL; GGF; NDF; NRG1 Intronic Transcript 2 (Non- Protein Coding); Heregulin, Alpha (45kD, ERBB2 P185-Activator); Acetylcholine Receptor-Inducing Activity; Pro-Neuregulin-1, Membrane-Bound Isoform; Sensory And Motor Neuron Derived Factor; Neu Differentiation Factor; Glial Growth Factor 2; NRG1-IT2; MSTP131; MST131; ARIA; GGF2; HRG1; and HRG. External Ids for NRG1 Gene are HGNC: 7997; Entrez Gene: 3084; Ensembl: ENSG00000157168; OMIM: 142445 and UniProtKB: Q02297. Isoforms of NRG1 can form by alternative splicing, and include forms that are transmembrane, externally membrane bound, shed, secreted or intracellular (Falls DL (2003). Neuregulins: functions, forms, and signaling strategies. Exp Cell Res 284: 14–30;Hayes NV, Gullick WJ (2008). The neuregulin family of genes and their multiple splice variants in breast cancer. J Mammary Gland Biol Neoplasia. (2):205-14. doi: 10.1007 / s10911-008-9078-4). They bind to ErbB-3 or ErbB-4, which probably signal as heterodimers with ErbB-2 (HER2). Although the NRG1-encoded proteins are usually thought of as mitogens, they can also be powerfully proapoptotic: in particular, expressing NRG1 in cells can cause apoptosis of the expressing cell (Weinstein EJ et al. (1998). The oncogene heregulin induces apoptosis in breast epithelial cells and tumors. Oncogene 17: 2107–2113). In the present disclosure, the inventors demonstrate that Patient-Derived Xenograft (PDX) models with elevated NRG1 expression levels exhibit significant growth inhibition in response to exposure to a bispecific antibody that comprises a first antigen-binding site that can bind an extracellular part of ErbB-2 and a second antigen-binding site that can bind an extracellular part of ErbB-3. Furthermore, the inventors show that human subjects having an ErbB-2, ErbB-3 or ErbB-2 / ErbB-3 positive cancer comprising elevated NRG1 expression, experience tumor reduction, or cessation of tumor growth from treatment with such an antibody. Meulendijks D et al., (Phase Ib Study of Lumretuzumab Plus Cetuximab or Erlotinib in Solid Tumor Patients and Evaluation of HER3 and Heregulin as Potential Biomarkers of Clinical Activity. Clin. Cancer Res; 23(18), 2017, 5406- 5415) describes a study investigating the safety, clinical activity and target- associated biomarkers of lumretuzumab, an anti-HER3 antibody, combined with either erlotinib or cetuximab. The authors concluded that, in the squamous non– small cell lung cancer (sqNSCLC) cohort, there was no evidence of meaningful clinical benefit despite enriching for tumors with higher HRG mRNA expression levels. Furthermore, in the investigated cohort, it was concluded that even higher HRG or HER3 mRNA expression levels were not associated with increased objective overall response rate (ORR) or clinically meaningful durations of response. SUMMARY OF THE INVENTION The present disclosure relates to classifying a subject having an ErbB-2, ErbB-3 or ErbB-2 / ErbB-3 positive cancer as being likely to respond to treatment with an ErbB2 and / or ErbB3 targeting agent, wherein the targeting agent may be a binding agent or may be an inhibitor thereof. The present disclosure relates to a method for identifying a subject having an ErbB-2, ErbB-3 or ErbB-2 / ErbB-3 positive cancer for treatment with an ErbB2 and / or ErbB3 targeting agent, comprising determining NRG1 expression level in a cancer cell containing sample from the subject. The present disclosure relates to a method for diagnosing a subject having an ErbB-2, ErbB-3 or ErbB-2 / ErbB-3 positive cancer and selecting for treatment with an ErbB2 and / or ErbB3 targeting agent, comprising determining NRG1 expression level in a cancer cell containing sample from the subject. The present disclosure relates to a method for selecting a subject having an ErbB-2, ErbB-3 or ErbB-2 / ErbB-3 positive cancer for treatment with an ErbB-2 and / or ErbB-3 targeting agent, the method comprising determining NRG1 expression level in a cancer cell containing sample from said subject. In certain aspects, said agent is zenocutuzumab. In certain aspects, said agent comprises an anti-ErbB2 specific antibody, such as a monospecific, bivalent antibody, which includes trastuzumab or pertuzumab. In certain aspect, said agent comprises an anti-ErbB-3 specific antibody, such as MM-121 (Merrimack Pharmaceuticals; also referred to as #Ab6), HMBD-001 (Hummingbird) or RG7116 (Lumretuzumab, Roche). In certain aspect, said agent comprises an anti-ErbB- 2 / anti-ErbB-3 specific antibody, such as MM-111 from Merrimack Pharmaceuticals. In certain aspect, said agent comprises a small molecule ErbB2 tyrosine kinase inhibitor (TKI), such as lapatinib (Tyverb / Tykerb®), neratinib, afatinib, tucatinib or AZD8931. In certain aspect, said agent comprises an antibody drug conjugate, for example trastuzumab emtansine or DS-8201. The present disclosure also relates to an ErbB2 and / or ErbB3 targeting agent for use in a method of treatment of a subject having an ErbB-2, ErbB-3 or ErbB-2 / ErbB-3 positive cancer, wherein said subject is classified as being likely to respond to treatment with an ErbB2 and / or ErbB3 targeting agent selected for treatment. The present disclosure also relates to an ErbB2 and / or ErbB3 targeting agent for use in a method of treatment of a subject having an ErbB-2, ErbB-3 or ErbB-2 / ErbB-3 positive cancer, wherein said subject is identified for treatment with an ErbB2 and / or ErbB3 targeting agent selected for treatment. The present disclosure also relates to an ErbB2 and / or ErbB3 targeting agent for use in a method of diagnosing a subject having an ErbB-2, ErbB-3 or ErbB-2 / ErbB-3 positive cancer, wherein said subject is diagnosed for treatment with an ErbB2 and / or ErbB3 targeting agent if having elevated NRG1 expression levels. It was found by the present inventors that by determining the 95 percentile of NRG1 expression level for samples of a particular reference dataset described further herein, a threshold value for NRG1 expression could be identified that can serve in classifying and selecting subjects having cancer for treatment with an ErbB2 and / or ErbB3 targeting agent. The present disclosure also relates to an ErbB2 and / or ErbB3 targeting agent for use in a method of treatment of a subject having an ErbB-2, ErbB-3 or ErbB-2 / ErbB-3 positive cancer, wherein said cancer expresses NRG1 levels equal to or higher than the 95 percentile of a reference dataset of cancer cells comprising samples. The present disclosure also relates to an ErbB2 and / or ErbB3 targeting agent for use in a method of treatment of a subject having an ErbB-2, ErbB-3 or ErbB-2 / ErbB-3 positive cancer, wherein the cancer of the subject comprises a threshold expression level of NRG1 which is increased as compared to that of a reference dataset, the threshold expression level of NRG1 having been determined according to a threshold determination analysis of a reference dataset of samples comprising cancer cells. The present disclosure also relates to an ErbB2 and / or ErbB3 targeting agent for use in a method of treatment of a subject having an ErbB-2, ErbB-3 or ErbB-2 / ErbB-3 positive cancer, wherein a cancer comprising sample obtained from the subject comprises an amount of NRG1 expression which is elevated in comparison to that of a sample free of cancer cells. The present disclosure also relates to an ErbB2 and / or ErbB3 targeting agent for use in a method of treatment of a subject having an ErbB-2, ErbB-3 or ErbB-2 / ErbB-3 positive cancer, said subject, or a sample obtained from said subject, having an amount of neuregulin-1 (NRG1) expression which is equal to or exceeds the 95 percentile for NRG1 expression of a reference dataset, optionally said subject, or a sample obtained from said subject, having an amount of phosphorylated HER3 which is elevated in comparison to that of a healthy subject; or said subject, or a sample obtained form said subject, having an amount of FAM83E expression which is elevated in comparison to that of a healthy subject and said subject or a sample obtained from said subject, having an amount of OLFML2B expression which is reduced in comparison to that of a sample free of cancer cells. The present disclosure also relates to a method of treatment of a subject having an ErbB-2, ErbB-3 or ErbB-2 / ErbB-3 positive cancer, the method comprising determining that the amount of neuregulin-1 (NRG1) expression in a sample obtained from the subject is elevated in comparison to a sample from a healthy subject, and administering an effective amount of an ErbB2 and / or ErbB3 targeting agent to the subject from which the sample having elevated NRG1 expression levels is obtained, optionally further determining that the amount of phosphorylated HER3 is elevated in comparison to a sample from a healthy subject, or determining that the amount of FAM83E expression is elevated and determining that the amount of OLFML2B expression is reduced in comparison to a sample from a healthy subject. The present disclosure also relates to a method of treatment of a subject having an ErbB-2, ErbB-3 or ErbB-2 / ErbB-3 positive cancer, the method comprising determining the amount of neuregulin-1 (NRG1) expression as mentioned herein, and administering a therapeutically effective amount of an ErbB2 and / or ErbB3 targeting agent to said subject if the amount of NRG1 in the sample is equal to or exceeds the 95 percentile for NRG1 expression of a reference dataset, optionally further determining that the amount of phosphorylated HER3 is elevated in comparison to a sample from a healthy subject, or determining that the amount of FAM83E expression is elevated and determining that the amount of OLFML2B expression is reduced in comparison to a sample from a healthy subject. In certain aspects, the cancer comprising sample is positive for pHER3 status. In certain aspects, the cancer comprising sample comprises an amount of FAM83E expression which is elevated in comparison to a non-cancer cells comprising sample and comprises an amount of OLFML2B expression which is reduced in comparison to a sample from a sample free of cancer cells. In certain aspects, a method of the present disclosure further comprises determining the presence or amount of pHER3 in a sample from the subject, or determining FAM83E expression level and OLFML2B expression level in a sample from the subject. In certain aspects, a subject of the present disclosure is selected for treatment or classified as being likely to respond to treatment with an ErbB2 and / or ErbB3 targeting agent when having an elevated NRG1 expression level. In certain aspects, a subject of the present disclosure is selected for treatment or classified as being likely to respond to treatment with an ErbB2 and / or ErbB3 targeting agent when having an elevated NRG1 expression level and when having a positive pHER3 status; or when having an elevated NRG1 expression level, having an elevated FAM83E expression and having a decreased OLFML2B expression level. In certain aspects, a method of the present disclosure comprises: - determining an expression level of NRG1 in a sample comprising cancer cells obtained from the subject; - comparing the expression level of the NRG1 in the sample with a threshold expression level of NRG1, the threshold expression level of NRG1 having been determined according to a threshold determination analysis of a reference dataset of samples comprising cancer cells; and - determining by the comparison that the expression level of the NRG1 in the sample is higher than the threshold expression level of NRG1 of the reference dataset, thereby classifying that the cancer is like to respond to treatment with the ErbB2 and / or ErbB3 targeting agent, or thereby selecting the subject for the treatment. In certain aspects, a method of the present disclosure comprises - determining NRG1 expression level by determining NRG1 sequence read abundance per individual sample of a reference dataset of samples comprising cancer cells, - calculating the 95 percentile for the NRG1 expression of the reference dataset, - determining the amount of NRG1 expression level in a cancer cells comprising sample obtained from the subject, and - selecting the subject for treatment, or classifying the subject is likely to respond to treatment with an ErbB2 and / or ErbB3 targeting agent, if the NRG1 expression level in the sample obtained from the subject is equal to or higher than the 95 percentile of the reference dataset. In certain aspects, a method of the present disclosure further comprises determining the presence or amount of pHER3 in a sample from the subject, and selecting the subject for treatment, or classifying the subject is likely to respond to the treatment, if additionally, the sample from the subject is positive for pHER3. In certain aspects, a method of the present disclosure further comprises determining FAM83E expression level and OLFML2B expression level in a sample from the subject, and selecting the subject for treatment, or classifying the subject as likely to respond to the treatment, if additionally, the sample from the subject has an elevated FAM83E expression level and has a decreased OLFML2B expression level. The cell may be a malignant cell or a cancer cell. The cancer cell may be a cancer cell associated with elevated NRG1 gene expression, such as a cancer cell driven by elevated NRG1 gene expression. In certain aspects, the subject of the present disclosure having an ErbB-2 and / or ErbB-3 positive tumor may be a subject that is in remission. In certain aspects, the cancer is brain cancer, breast cancer, pancreatic cancer, lung cancer, esophageal cancer, liver cancer, cholangiocarcinoma or gastric cancer. In certain aspects, the reference dataset of samples comprising cancer cells comprises a cancer which is brain cancer, breast cancer, pancreatic cancer, lung cancer, esophageal cancer, liver cancer, cholangiocarcinoma or gastric cancer. In certain aspects, the cancer cells comprising sample comprises brain cancer cells, breast cancer cells, pancreatic cancer cells, lung cancer cells, esophageal cancer cells, liver cancer cells, cholangiocarcinoma cancer cells, prostate cancer cells or gastric cancer cells. In certain aspects, the malignant cell is a cancer cell. In certain aspects, the malignant or cancer cell is an epithelial cell. In certain aspects, the cell is a lung cancer cell, such as a non-small cell lung cancer, an esophageal cancer cell, a liver cancer cell, a cholangiocarcinoma cancer cell, a breast cancer cell or gastric cancer cell. In certain aspects, the tumor is of an epithelial origin. In certain aspects, the tumor is a pancreatic cancer, breast cancer, lung cancer, esophageal cancer, liver cancer, cholangiocarcinoma or gastric cancer or a metastasis thereof. The malignant cell may be a cancer cell. In certain aspects, said cancer is a pancreatic cancer comprising elevated NRG1 expression and in certain aspects, said cancer is positive for pHER3. In certain aspects, said cancer is a pancreatic cancer comprising elevated NRG1 levels and is positive for FAM83E expression and has reduced OLFML2B expression levels. In certain aspects, said cancer is a lung cancer comprising elevated NRG1 expression and in certain aspects, said cancer is positive for pHER3. In certain aspects, said cancer is a lung cancer comprising elevated NRG1 levels and is positive for FAM83E expression and has reduced OLFML2B expression levels. In certain aspects, said cancer is an esophageal cancer comprising elevated NRG1 expression and in certain aspects, said cancer is positive for pHER3. In certain aspects, said cancer is an esophageal cancer comprising elevated NRG1 levels and is positive for FAM83E expression and has reduced OLFML2B expression levels. In certain aspects, said cancer is a liver cancer cell comprising elevated expression and in certain aspects, said cancer is positive for pHER3. In certain aspects, said cancer is a liver cancer comprising elevated NRG1 levels and is positive for FAM83E expression and has reduced OLFML2B expression levels. In certain aspects, said cancer is a breast cancer comprising elevated expression and in certain aspects, said cancer is positive for pHER3. In certain aspects, said cancer is a breast cancer comprising elevated NRG1 levels and is positive for FAM83E expression and has reduced OLFML2B expression levels. In certain aspects, said cancer is a cholangiocarcinoma comprising elevated NRG1 expression and in certain aspects, said cancer is positive for pHER3. In certain aspects, said cancer is a cholangiocarcinoma comprising elevated NRG1 levels and is positive for FAM83E expression and has reduced OLFML2B expression levels. In certain aspects, said cancer is a gastric cancer comprising elevated NRG1 expression and in certain aspects, said cancer is positive for pHER3. In certain aspects, said cancer is a gastric cancer comprising elevated NRG1 levels and is positive for FAM83E expression and has reduced OLFML2B expression levels. In certain aspects, the subject selected for treatment has undergone a therapy that targeted towards EGFR inhibition, such as with an EGFR binding antibody, which may be cetuximab. As is clear to a skilled person the bispecific antibodies disclosed herein are also for the use in the preparation of a medicament and for the use in therapy, as disclosed herein. The present disclosure thus also relates to a use of an ErbB2 and / or ErbB3 targeting agent in the manufacture of a medicament for the treatment of a subject having an ErbB-2, ErbB-3 or ErbB-2 / ErbB-3 positive cancer, wherein said subject is identified for treatment with an ErbB2 and / or ErbB3 targeting agent and / or is classified as being likely to respond to treatment with an ErbB2 and / or ErbB3 targeting agent selected for treatment. The present disclosure also relates to a use of an ErbB2 and / or ErbB3 targeting agent in the manufacture of a medicament for the treatment of an ErbB- 2, ErbB-3 or ErbB-2 / ErbB-3 positive cancer, wherein said cancer expresses NRG1 levels equal to or higher than the 95 percentile for NRG1 expression of a reference dataset of cancer cells comprising samples. The present disclosure also relates to a use of an ErbB2 and / or ErbB3 targeting agent in the manufacture of a medicament for the treatment of an ErbB- 2, ErbB-3 or ErbB-2 / ErbB-3 positive cancer, wherein the cancer of the subject comprises a threshold expression level of NRG1 which is increased as compared to that of a reference dataset, the threshold expression level of NRG1 having been determined according to a threshold determination analysis of a reference dataset of samples comprising cancer cells. The present disclosure also relates to a use of an ErbB2 and / or ErbB3 targeting agent in the manufacture of a medicament for the treatment of an ErbB- 2, ErbB-3 or ErbB-2 / ErbB-3 positive cancer, wherein a cancer comprising sample obtained from the subject comprises an amount of NRG1 expression which is elevated in comparison to that of a sample free of cancer cells. The present disclosure also relates to a use of an ErbB2 and / or ErbB3 targeting agent in the manufacture of a medicament for the treatment of an ErbB- 2, ErbB-3 or ErbB-2 / ErbB-3 positive cancer, said subject, or a sample obtained from said subject, having an amount of neuregulin-1 (NRG1) expression which is equal to or exceeds the 95 percentile for NRG1 expression of a reference dataset, optionally said subject, or a sample obtained from said subject, having an amount of phosphorylated HER3 which is elevated in comparison to that of a healthy subject; or said subject, or a sample obtained from said subject, having an amount of FAM83E expression which is elevated in comparison to that of a healthy subject and said subject or a sample obtained from said subject, having an amount of OLFML2B expression which is reduced in comparison to that of a sample free of cancer cells. In certain aspects, a bispecific antibody that comprises a first antigen- binding site that can bind an extracellular part of ErbB-2 and a second antigen- binding site that can bind an extracellular part of ErbB-3 is provided for use in the treatment of an individual that has an ErbB-2, ErbB-3 or ErbB-2 / ErbB-3 positive cell or cancer, which cell or cancer comprises elevated NRG1 expression. The cell of the present disclosure may be a malignant cell having elevated NRG1 expression or a cancer cell having elevated NRG1 expression. The present disclosure also provides an ErbB2 and / or ErbB3 targeting agent for use in a method of treatment of a subject having an ErbB-2, ErbB-3 or ErbB-2 / ErbB-3 positive cancer, said method comprising determining that the amount of neuregulin-1 (NRG1) expression is elevated in comparison to a sample from a healthy subject. In certain aspects, said method further comprises determining that the amount of phosphorylated HER3 is elevated in comparison to a sample from a healthy subject. In certain alternative aspects, said method further comprises determining that the amount of FAM83E expression is elevated and determining that the amount of OLFML2B expression is reduced in comparison to a sample from a healthy subject. In certain aspects, said method comprises administering the antibody to the subject having an elevated NRG1 expression level. In certain aspects, said method comprises administering the antibody to the subject if it has been determined that the subject has an elevated NRG1 expression level. The present disclosure also provides an ErbB2 and / or ErbB3 targeting agent for use in a method of treatment of a subject having an ErbB-2, ErbB-3 or ErbB-2 / ErbB-3 positive cancer, said method comprising determining the amount of neuregulin-1 (NRG1) expression and administering the ErbB2 / ErbB3 targeting agent to the subject if the amount of NRG1 is equal to or exceeds the 95 percentile for NRG1 expression of a reference dataset in the sample. In certain aspects, said method further comprises determining that the amount of phosphorylated HER3 is elevated in comparison to a sample from a healthy subject. In certain alternative aspects, said method further comprises determining that the amount of FAM83E expression is elevated and determining that the amount of OLFML2B expression is reduced in comparison to a sample from a healthy subject. In certain aspects, said method comprises administering the antibody to the subject having an elevated NRG1 expression level. The present disclosure also provides a method of treatment of a subject having an ErbB-2, ErbB-3 or ErbB-2 / ErbB-3 positive cancer, the method comprising determining that the amount of neuregulin-1 (NRG1) expression in a sample obtained from the subject is elevated in comparison to a sample from a healthy subject, and administering an effective amount of an ErbB2 and / or ErbB3 targeting agent to the subject from which the sample having elevated NRG1 expression levels is obtained, optionally further determining that the amount of phosphorylated HER3 is elevated in comparison to a sample from a healthy subject, or determining that the amount of FAM83E expression is elevated and determining that the amount of OLFML2B expression is reduced in comparison to a sample from a healthy subject. The present disclosure also provides a method of treatment of a subject having an ErbB-2, ErbB-3 or ErbB-2 / ErbB-3 positive cancer, the method comprising determining the amount of neuregulin-1 (NRG1) expression and administering a therapeutically effective amount of an ErbB2 and / or ErbB3 targeting agent to said subject if the amount of NRG1 in the sample is equal to or exceeds the 95 percentile for NRG1 expression of a reference dataset. In certain aspects, said method further comprises determining that the amount of phosphorylated HER3 is elevated in comparison to a sample from a healthy subject. In certain aspects, said method further comprises determining that the amount of FAM83E expression is elevated and determining that the amount of OLFML2B expression is reduced in comparison to a sample from a healthy subject. In certain aspects, the amount of NRG1 exceeds the 95 percentile. In certain aspects, in the methods and uses disclosed herein, a bispecific antibody is used which comprises a first antigen-binding site that can bind domain I of ERBB2 and a second antigen-binding site that can bind domain III of ERBB3. In certain aspects, said antibody comprises at least i) the CDR1, CDR2 and CDR3 sequences of an ErbB-2 specific heavy chain variable region selected from the group consisting of MF2973, MF3004, MF3958, MF2971, MF3025, MF2916, MF3991, MF3031 and MF3003; and / or ii) at least the CDR1, CDR2 and CDR3 sequences of an ErbB-3 specific heavy chain variable region selected from the group consisting of MF3178; MF3176; MF3163; MF6055; MF6056; MF6057; MF6058; MF6059; MF6060; MF6061; MF6062; MF6063; MF6064; MF 6065; MF6066; MF6067; MF6068; MF6069; MF6070; MF6071; MF6072; MF6073 and MF6074. In certain aspects, said antibody comprises i) an ErbB-2 specific heavy chain variable region sequence selected from the group consisting of the heavy chain variable region sequences of MF2973, MF3004, MF3958, MF2971, MF3025, MF2916, MF3991, MF3031, and MF3003 and / or ii) an ErbB-3 specific heavy chain variable region sequence selected from the group consisting of the heavy chain variable region sequences of MF3178; MF3176; MF3163; MF6055; MF6056; MF6057; MF6058; MF6059; MF6060; MF6061; MF6062; MF6063; MF6064; MF 6065; MF6066; MF6067; MF6068; MF6069; MF6070; MF6071; MF6072; MF6073 and MF6074. In certain aspects, the antibody comprises at least the CDR1, CDR2 and CDR3 sequences of the ErbB-2 specific heavy chain variable region MF3958 and the antibody comprises at least the CDR1, CDR2 and CDR3 sequences of the ErbB-3 specific heavy chain variable region MF3178. In certain aspects, the bispecific antibody comprises the ‘heavy chain for erbB-2 binding’ as depicted in Figure 2E and the ‘heavy chain for erbB-3 binding’ as depicted in Figure 2F. In certain aspects, the first antigen binding site and said second antigen binding site comprise a light chain variable region comprising the IgVKl-39 gene segment, such as the rearranged germline human kappa light chain IgVKl- 39*01 / IGJKl*01 or IgVκ1-39*01 / IGJκ5*01. In certain aspects, the light chain variable region comprises a CDR1 having the sequence (QSISSY), a CDR2 having the sequence (AAS), and a CDR3 having the sequence (QQSYSTPPT). In certain aspects, the bispecific antibody binding ErbB2 and ErbB3 is or comprises zenocutuzumab (International Nonproprietary Name). DETAILED DESCRIPTION OF THE INVENTION The present disclosure relates to classifying a subject or a selecting a subject for treatment with an ErbB2 and / or ErbB3 targeting agent, wherein the targeting agent may be a binding agent or may be an inhibitor thereof. In a certain aspect, said agent is zenocutuzumab. In certain aspect, said agent comprises an anti-ErbB2 specific antibody, such as a monospecific, bivalent antibody, which includes trastuzumab or pertuzumab. In certain aspect, said agent comprises an anti-ErbB-3 specific antibody, such as MM-121 (Merrimack Pharmaceuticals; also referred to as #Ab6), HMBD-001 (Hummingbird) or RG7116 (Lumretuzumab, Roche). In certain aspect, said agent comprises an anti-ErbB-2 / anti-ErbB-3 specific antibody, such as MM-111 from Merrimack Pharmaceuticals. In certain aspect, said agent comprises a small molecule ErbB2 tyrosine kinase inhibitor (TKI), such as lapatinib (Tyverb / Tykerb®), neratinib, afatinib, tucatinib or AZD8931. In certain aspect, said agent comprises an antibody drug conjugate, for example trastuzumab emtansine or DS-8201. The present disclosure provides a method for determining that an ErbB- 2, ErbB-3 or ErbB-2 / ErbB-3 positive cancer, or a subject suffering therefrom, is likely to respond to treatment with an ErbB2 and / or ErbB3 targeting agent, and subsequently selecting a subject for treatment of the cancer, the method comprising: - determining an expression level of NRG1 in a sample comprising cancer cells obtained from the subject; - comparing the expression level of the NRG1 in the sample with a threshold expression level of NRG1, the threshold expression level of NRG1 having been determined according to a threshold determination analysis of a reference dataset of samples comprising cancer cells; and - determining by the comparison that the expression level of the NRG1 in the sample is higher than the threshold expression level of NRG1 of the reference dataset, thereby classifying that the cancer is like to respond to treatment with the ErbB2 and / or ErbB3 targeting agent, or thereby selecting the subject for the treatment. The present disclosure provides a method for classifying a subject having an ErbB-2, ErbB-3 or ErbB-2 / ErbB-3 positive cancer as likely to respond to treatment with an ErbB2 and / or ErbB3 targeting agent, and subsequently selecting a subject for treatment of the cancer, the method comprising: - determining NRG1 expression level by determining NRG1 sequence read abundance per individual sample of a reference dataset of samples comprising cancer cells, - calculating the 95 percentile for the NRG1 expression of the reference dataset, - determining the amount of NRG1 expression level in a cancer cells comprising sample obtained from the subject, and - selecting the subject for treatment, or classifying the subject is likely to respond to treatment with an ErbB2 and / or ErbB3 targeting agent, if the NRG1 expression level in the sample obtained from the subject is equal to or higher than the 95 percentile of the reference dataset. The present disclosure provides a method for selecting a subject having an ErbB-2, ErbB-3 or ErbB-2 / ErbB-3 positive cancer for treatment with an ErbB2 and / or ErbB3 targeting agent, the method comprising: - determining NRG1 expression level by determining NRG1 sequence read abundance per individual sample of a reference dataset of samples comprising cancer cells, - calculating the 95 percentile for the NRG1 expression of the reference dataset, - determining the amount of NRG1 expression level in a cancer cells comprising sample obtained from the subject, and - selecting the subject for treatment, or classifying the subject is likely to respond to treatment with an ErbB2 and / or ErbB3 targeting agent, if the NRG1 expression level in the sample obtained from the subject is equal to or higher than the 95 percentile of the reference dataset. In certain aspects, a method of the present disclosure further comprises determining the presence or amount of pHER3 in a sample from the subject, and selecting the subject for treatment, or classifying the subject is likely to respond to the treatment, if additionally, the sample from the subject is positive for pHER3. In certain aspects, the subject is selected for treatment or classified as being likely to respond to treatment with an ErbB2 and / or ErbB3 targeting agent when a sample obtained from the subject has an elevated expression level as compared to a reference. In certain aspects, the subject is selected for treatment or classified as being likely to respond to treatment with an ErbB2 and / or ErbB3 targeting agent when the NRG1 expression level is determined to be elevated in a sample obtained from the subject as compared to a reference. In certain aspects, the subject is selected for treatment or classified as being likely to respond to treatment with an ErbB2 and / or ErbB3 targeting agent when the NRG1 expression level is determined to be elevated in a sample obtained from the subject as compared to a reference and when the pHER status is determined to be positive in a sample obtained from the subject. In certain aspects, the subject is selected for treatment or classified as being likely to respond to treatment with an ErbB2 and / or ErbB3 targeting agent when the NRG1 expression level is determined to be elevated in a sample obtained from the subject as compared to a reference, when the FAM83E expression level is determined to be elevated in the sample obtained from the subject as compared to a reference and when the OLFML2B expression level is determined to be decreased in the sample obtained from the subject as compared to a reference. The reference may be selected from, e.g., a healthy patient, a healthy sample, a threshold level of NRG1, a threshold level of FAM83E, a threshold level of OLFML2B or pHER3, the 95 percentile of a reference dataset of cancer cells comprising samples ordered by NRG1 expression level, a sample free from cancer cells or a sample comprising cancer cells. In certain aspects, the reference is a sample obtained from a subject suffering from cancer. In certain aspects, the reference is a sample comprising cancer cells. In certain aspects, the reference is a sample comprising cancer cells obtained from a subject suffering from cancer. In certain aspects, a method of the present disclosure further comprises determining the presence or amount of phosphorylated HER3 (pHER3) in a sample from the subject, or determining FAM83E expression level and OLFML2B expression level in a sample from the subject. In certain aspects, a method of the present disclosure further comprises selecting a subject for treatment or classifying as being likely to respond to treatment with an ErbB2 and / or ErbB3 targeting agent when having an elevated NRG1 expression level. In certain aspects, a method of the present disclosure further comprises selecting a subject for treatment or classifying as being likely to respond to treatment with an ErbB2 and / or ErbB3 targeting agent, when having an elevated NRG1 expression level and when having a positive pHER3 status; or when having an elevated NRG1 expression level, having an elevated FAM83E expression and having a decreased OLFML2B expression level. In certain aspects, a method of the present disclosure further comprises: - measuring a pHER3 level of in a sample comprising cancer cells from the subject; - comparing the pHER3 level in the sample comprising cancer cells with a threshold level of pHER3, the threshold level of pHER3 having been determined according to a threshold determination analysis of a first dataset of samples comprising cancer cells; and - determining by combining the comparison that the pHER3 level in the sample comprising cancer cells is higher than the threshold pHER3 level and that the expression level of the NRG1 in the sample comprising cancer cells is higher than the threshold expression level of NRG1, thereby determining that the cancer is like to respond to treatment with the ErbB2 and / or ErbB3 targeting agent. In certain aspects, said method further comprises: - measuring an expression level of FAM83E and an expression level of OLFML2B in a sample comprising cancer cells from the subject; - comparing both the expression level of FAM83E and the expression level of OLFML2B in the sample comprising cancer cells with a threshold level of FAM83E and OLFML2B, the threshold expression level of FAM83E and OLFML2B levels having been determined according to a threshold determination analysis of a first dataset of samples comprising cancer cells; and - determining by combining the comparison that 1) the FAM83E expression level in the sample comprising cancer cells is higher than the threshold FAM83E expression level, 2) the OLFML2B expression level in the sample comprising cancer cells is lower than the threshold OLFML2B expression level, and 3) that the expression level of the NRG1 in the sample comprising cancer cells is higher than the threshold expression level of NRG1, thereby determining that the cancer is like to respond to treatment with the ErbB2 and / or ErbB3 targeting agent. In certain aspects, said method further comprises, following the determination that the cancer is like to respond to treatment with the ErbB2 and / or ErbB3 targeting agent, administering the ErbB2 and / or ErbB3 targeting agent to the subject. In certain aspects, the presence or amount of pHER3 is determined in a sample from the subject, and the subject is selected for treatment, or classified as being likely to respond to the treatment, if additionally, the sample from the subject is positive for pHER3. In certain aspects, the expression level of FAM83E and OLFML2B is determined in a sample from the subject, and the subject is selected for treatment, or classified as being likely to respond to the treatment, if additionally, the sample from the subject has an elevated FAM83E expression level and has a decreased OLFML2B expression level. In certain aspects, determining NRG1, FAM83E and / or OLFML2B expression levels comprises aligning sequencing reads of NRG1, FAM83E and / or OLFML2B, respectively, to the GRCh37.66 or GRCh38 / hg38 genome build as reference genome. In certain aspects, said genome build is GRCh38.p13. In certain aspects, said genome build is GRCh38.p13 update 106 (dated April 2022). In certain aspects, the cancer is brain cancer (glioblastoma multiforme, GBM), pancreatic cancer (pancreatic adenocarcinoma, PAAD), lung cancer ((lung adenocarcinoma (LUAD) or lung squamous cell carcinoma (LUSC), or non-small cell lung cancer (NSCLC)), esophageal cancer (esophageal carcinoma, ESCA), liver cancer (liver hepatocellular carcinoma, LIHC), cholangiocarcinoma (CHOL) or gastric cancer (stomach adenocarcinoma, STAD). In certain aspects, the cancer is glioblastoma multiforme, pancreatic adenocarcinoma, lung adenocarcinoma, lung squamous cell carcinoma, non-small cell lung cancer (NSCLC), esophageal carcinoma, liver hepatocellular carcinoma, cholangiocarcinoma, prostate adenocarcinoma or gastric adenocarcinoma. In certain aspects, the sample is from a brain cancer, breast cancer, pancreatic cancer, lung cancer, esophageal cancer, liver cancer, cholangiocarcinoma or gastric cancer. In certain aspects, the reference dataset of samples comprising cancer cells comprises a cancer which is brain cancer, breast cancer, pancreatic cancer, lung cancer, esophageal cancer, liver cancer, cholangiocarcinoma or gastric cancer. In certain aspects, the cancer cells comprising sample comprises brain cancer cells, breast cancer cells, pancreatic cancer cells, lung cancer cells, esophageal cancer cells, liver cancer cells, cholangiocarcinoma cancer cells or gastric cancer cells. The malignant cell may be a cancer cell. The cancer cell may be a cancer cell associated with elevated NRG1 gene expression, such as a cancer cell driven by elevated NRG1 gene expression. In certain aspects, the malignant cell is a cancer cell. In certain aspects, the cell is an epithelial cell. In certain aspects, the cell is a lung cancer cell, such as a non-small cell lung cancer, a pancreatic cancer cell, a breast cancer cell, an esophageal cancer cell, a liver cancer cell, a cholangiocarcinoma cancer cell or gastric cancer cell. In certain aspects, the tumor is of an epithelial origin. In certain aspects, the tumor is a pancreatic cancer, a breast cancer, a lung cancer (including non- small cell lung cancer), esophageal cancer, liver cancer, cholangiocarcinoma or gastric cancer or a metastasis thereof. The malignant cell may be a cancer cell. In certain aspects, said cancer is a pancreatic cancer cell comprising elevated NRG1 expression and in certain aspects, said cancer cell is positive for pHER3. In certain aspects, said cancer is a pancreatic cancer cell comprising elevated NRG1 levels and is positive for FAM83E expression and has reduced OLFML2B expression levels. In certain aspects, said cancer cell is a lung cancer cell comprising elevated NRG1 expression and in certain aspects, said cancer cell is positive for pHER3. In certain aspects, said cancer is a lung cancer cell comprising elevated NRG1 levels and is positive for FAM83E expression and has reduced OLFML2B expression levels. In certain aspects, said cancer cell is an esophageal cancer cell comprising elevated NRG1 expression and in certain aspects, said cancer cell is positive for pHER3. In certain aspects, said cancer is an esophageal cancer cell comprising elevated NRG1 levels and is positive for FAM83E expression and has reduced OLFML2B expression levels. In certain aspects, said cancer cell is a liver cancer cell comprising elevated expression and in certain aspects, said cancer cell is positive for pHER3. In certain aspects, said cancer is a liver cancer cell comprising elevated NRG1 levels and is positive for FAM83E expression and has reduced OLFML2B expression levels. In certain aspects, said cancer cell is a cholangiocarcinoma cell comprising elevated NRG1 expression and in certain aspects, said cancer cell is positive for pHER3. In certain aspects, said cancer is a cholangiocarcinoma cell comprising elevated NRG1 levels and is positive for FAM83E expression and has reduced OLFML2B expression levels. In certain aspects, said cancer cell is a gastric cancer cell comprising elevated NRG1 expression and in certain aspects, said cancer cell is positive for pHER3. In certain aspects, said cancer is a gastric cancer cell comprising elevated NRG1 levels and is positive for FAM83E expression and has reduced OLFML2B expression levels. In certain aspects, the reference dataset of samples comprising cancer cells comprises at least 50 or at least 200 samples comprising cancer cells obtained from human subjects suffering from cancer. The reference dataset may be constructed from samples taken from subjects enrolled in a particular clinical trial. In certain aspects, the reference dataset comprises samples from a publicly known and accessible source, such as the Cancer Genome Atlas Program (TCGA). One advantage using said source is its public availability, reproducibility and large number of samples which provides statistically suitable and meaningful results. In certain aspects, the reference dataset comprises a TCGA dataset, such as dataset TCGA-CHOL, TCGA-GBM, TCGA-LIHC, TCGA-LUAD, TCGA-LUSC, TCGA- PAAD, TCGA-PRAD, TCGA-ESCA, or TCGA-STAD. In certain aspects, the cancer samples of the reference dataset correspond to the cancer type the subject of the present disclosure suffers from. In certain aspects, the TCGA dataset used is the entire dataset for a specific cancer indication (e.g. TCGA-CHOL). In certain aspects, the reference dataset comprises the TCGA ‘legacy’ release, in particular the ‘legacy’ release as downloaded per February 17th2020. As is clear to a skilled person, a sample free of cancer cells is the same in terms of origin or tissue as the sample from the subject to be classified or selected for treatment as mentioned herein to allow optimal comparison of results. Such a sample may also be termed a ‘healthy sample’ and may be obtained from a healthy part of an organ or tissue from the subject suffering from cancer or from a subject not suffering from cancer. Herein, a ‘threshold determination analysis’ is to be understood to include a determination wherein the 95 percentile for NRG1 expression of a reference dataset of samples comprising cancer cells is calculated based on sorting of samples according to NRG1 expression levels, calculating the 95 percentile for samples having highest NRG1 expression levels, and comparing a sample obtained from a subject having cancer to the 95 percentile to determine that the level of NRG1 of said sample is equal to or above the 95 percentile. The reference dataset of samples comprising cancer cells can be a dataset from the public domain but also from a clinical trial directed at treating cancer with any therapeutic agent of interest, including any ErbB2 and / or ErbB3 targeting agent. Typically, such a reference dataset of samples has at least 50 or at least 200 samples for reasons related to statistical significance. In certain aspects, the NRG1 expression, FAM83E and / or OLFML2B is based on sequence reads obtained by RNA sequencing. In certain aspects, the 95 percentile as established for the reference dataset of samples comprising cancer cells is based on sequence reads such as those obtained by RNA sequencing. In certain aspects, the 95 percentile as established for the reference dataset is based on Transcripts Per Million (TPM) or Fragments Per Kilobase Million (FPKM). In certain aspects, establishing or calculating the 95 percentile for a reference dataset includes an in silico alignment of sequence reads to a transcriptome or reference genome for each individual sample of the dataset to the NRG1 gene or exons thereof, and determining NRG1 transcript abundance or NRG1 expression level. Next, the 95 percentile is calculated, for instance using equation 2 and 3 as mentioned herein or equation 1 in program ‘R’ as mentioned herein as well. In certain aspects, establishing or calculating the median expression level for a reference dataset, includes an in silico alignment of sequence reads to a transcriptome or reference genome for each individual sample of the dataset to either the FAM83E and OLFML2B genes or exons thereof, and determining FAM83E or OLFML2B transcript abundance or expression level of the respective gene. Next, the median expression level is calculated for FAM83E or OLFML2B. In certain aspects, the in silico alignment to a transcriptome or reference genome for NRG1 includes the NRG1 gene of the GRCh37.66 or GRCh38 / hg38 genome build. In certain aspects, expression of NRG1 includes the DNA sequence of NRG1 isoforms of the GRCh37.66 or GRCh38 / hg38 genome build. In certain aspects, expression of NRG1 includes the protein coding DNA sequences of ENSG00000157168. In certain aspects, expression of NRG1 includes SEQ ID NOs: 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, or any allelic variant thereof. In certain aspects, the in silico alignment to a transcriptome or reference genome for FAM83E includes the FAM83E gene of the GRCh37.66 or GRCh38 / hg38 genome build. In certain aspects, expression of FAM83E includes the DNA sequence of FAM83E isoforms of the GRCh37.66 or GRCh38 / hg38 genome build. In certain aspects, expression of FAM83E includes the protein coding DNA sequences of ENSG00000105523. In certain aspects, expression of FAM83E includes SEQ ID NOs: 195 or 197, or any allelic variant thereof. In certain aspects, the in silico alignment to a transcriptome or reference genome for OLFML2B includes the OLFML2B gene of the GRCh37.66 or GRCh38 / hg38 genome build. In certain aspects, expression of OLFML2B includes the DNA sequence of OLFML2B isoforms of the GRCh37.66 or GRCh38 / hg38 genome build. In certain aspects, expression of OLFML2B includes the protein coding DNA sequences of ENSG00000162745. In certain aspects, expression of OLFML2B includes SEQ ID NOs: 199, 201 and / or 203 or any allelic variant thereof. In certain aspects, the 95 percentile as determined for NRG1 expression, is determined per cancer type. In certain aspects, the 95 percentile is determined using below depicted equation 1, which originates from equation 7 from the ‘Project for Statistical Computing’. The calculation can be done using software termed ‘R’ using its quantile function, see https: / / stat.ethz.ch / R-manual / R- devel / library / stats / html / quantile.html). Equation 1 Only by way of example and using imaginary values, in a first step the samples are ordered according to NRG1 expression (using appropriate TPM values): {12, 23, 34, 45, 54, 56, 67, 78, 88, 90}, after which the index is calculated: (95 / 100) * 10 = 9.5. Since in this case the index is not an integer (as it may happen for other examples), the values at positions 9 and 10 are taken. The values at the 9th and 10th positions are retrieved: 88, 90. Next, the average is calculated: (88 + 90) / 2 = 89. In this case, the 95th percentile of the list is 89. Alternatively, the estimate for the p-quantile (the k-th q-quantile, where p = k / q) from a sample of size N by computing a real valued index h using the following equation (see Hyndman RJ and Fan Y (1996). Sample Quantiles in Statistical Packages. The American Statistician, 50(4): 361–365). H is calculated according to equation 2: h = (N − 1)*p + 1, where N is the number of observations (i.e. the number of TPM values herein) and p the chosen quantile (probability, herein 0.95). The equation for calculating the quantile Qp is then via equation 3 : Qp = x⌊h⌋ + (h − ⌊h⌋) (x⌈h⌉ − x⌊h⌋), where h is calculated as above, and x are the observations (i.e. TPM values herein), wherein ⌊h⌋ means the 'floor' of h, ⌈h⌉ means the 'ceiling' of h. Only by way of example, if h is 9.55, ⌊h⌋ = 9 and ⌈h⌉ = 10. In this example, in the equation's notation x⌊h⌋, then means the 9th number in x (when ordered). If p = 0.95 and x of 10 TPM values: x = [0.1, 0.3, 5.2, 10.1, 6.8, 20.5, 3.7, 9.5, 4.2, 0.3]: Step 1. Ordering x: x = [0.1, 0.3, 0.3, 3.7, 4.2, 5.2, 6.8, 9.5, 10.1, 20.5] Step 2. Calculate h: h = (N - 1)*p + 1 turns into (10 - 1) * 0.95 + 1 , which is 9.55. Step 3. Calculate the quantile (Qp): x⌊h⌋ + (h − ⌊h⌋) (x⌈h⌉ − x⌊h⌋) turns into x[position 9] + (9.55 - 9)*(x[position 10] -x[position 9]) , which turns into: 10.1 + (9.55 - 9)*(20.5 -10.1), which is 15.82. Herein the term ‘RNA sequencing’ is to be understood as to also include sequencing of DNA reverse transcribed from RNA, which is typically referred to as cDNA. In certain aspects, elevated NRG1 expression is established using commercially known RNA sequencing methods employed by commercial vendors, including Caris (Caris Life Sciences, USA) Tempus (USA) and the like. Whole Transcriptome Sequencing (WTS) uses the capabilities of high-throughput sequencing to gain insight into the RNA profiles of patient tumors, including detailed gene expression data. In certain aspects, quantifying gene expression by RNA sequencing is by counting the number of reads that map (or align) to a gene of interest (read count) using a program such as HTSeq-count. In certain aspects, the 95 percentile for NRG1 is 1.202 log10 TPM. In certain aspects, the FAM83E expression is considered elevated when higher than a log10 TPM level of at least 0.30 Log10 TPM or at least 0.45 Log10 TPM or at least 0.66 Log10 TPM or at least 0.85 Log10 TPM or at least 1.20 Log10 TPM, or at least 1.40 Log10 TPM, or at least 1.70 Log10 TPM, or at least 2.00 Log10 TPM. In certain aspects, the FAM83E expression is considered elevated when higher than a log10 TPM level of at least 2.00 Log10 TPM. In certain aspects, the OLFML2B expression is considered decreased when lower than a log10 TPM level of less than 1.40 Log10 TPM or less than 1.18 Log10 TPM, or less than 0.92 Log10 TPM, or less than 0.72 Log10 TPM, or less than 0.70 Log10 TPM, or less than 0.50 Log10 TPM, or less than 0.20 Log10 TPM, or less than 0.02 Log10 TPM or 0.00 Log10 TPM. In certain aspects, the OLFML2B expression is considered decreased when lower than a log10 TPM level of less than 0.02 Log10 TPM or 0.00 Log10 TPM. In certain aspects, said cancer is gastric cancer, such as gastric or stomach adenocarcinoma. In certain aspects, the FAM83E expression is considered elevated when higher than a log10 TPM level of at least 1.8 or 2.0 log10 TPM. In certain aspects, the OLFML2B expression is considered decreased when lower than a log10 TPM level below 0.5 or 0.2 log10 TPM. In certain aspects, the cancer is gastric cancer, such as gastric or stomach adenocarcinoma (STAD) and the 95 percentile for NRG1 is 1.002 log10 TPM, such as determined for TCGA-STAD. In certain aspects, the cancer is gastric cancer and FAM83E expression is considered elevated when higher than a median log10 TPM level of a reference dataset, such as 1.325 log10 TPM as determined for TCGA- STAD. In certain aspects, the cancer is gastric cancer and OLFML2B expression is considered decreased when lower than a median log10 TPM level of a reference dataset, such as below 1.326 log10 TPM as determined for TCGA-STAD. In certain aspects, the cancer is cholangiocarcinoma and the 95 percentile for NRG1 is 1.069 log10 TPM, such as determined for TCGA-CHOL. In certain aspects, the cancer is cholangiocarcinoma and FAM83E expression is considered elevated when higher than a median log10 TPM level of a reference dataset, such as -0.513 log10 TPM as determined for TCGA-CHOL. In certain aspects, the cancer is cholangiocarcinoma and OLFML2B expression is considered decreased when lower than a median log10 TPM level of a reference dataset, such as below 0.989 log10 TPM as determined for TCGA-CHOL. In certain aspects, the cancer is brain cancer, such as glioblastoma multiforme (GBM) and the 95 percentile for NRG1 is 0.882 log10 TPM, as determined for TCGA-GBM. In certain aspects, the cancer is brain cancer, such as glioblastoma multiforme (GBM) and FAM83E expression is considered elevated when higher than a median log10 TPM level of a reference dataset, such as -1.046 log10 TPM as determined for TCGA-GBM. In certain aspects, the cancer is brain cancer, such as glioblastoma multiforme (GBM) and OLFML2B expression is considered decreased when lower than a median log10 TPM level of a reference dataset, such as below 1.36 log10 TPM as determined for TCGA-GBM. In certain aspects, the cancer is liver cancer, such as liver hepatocellular carcinoma (LIHC) and the 95 percentile for NRG1 is 1.057 log10 TPM, as determined for TCGA-LIHC. In certain aspects, the cancer is liver cancer, such as liver hepatocellular carcinoma (LIHC) and FAM83E expression is considered elevated when higher than a median log10 TPM level of a reference dataset, such as -1.952 log10 TPM as determined for TCGA-LIHC. In certain aspects, the cancer is liver cancer, such as liver hepatocellular carcinoma (LIHC) and OLFML2B expression is considered decreased when lower than a median log10 TPM level of a reference dataset, such as below 0.580 log10 TPM as determined for TCGA-LIHC. In certain aspects, the cancer is lung cancer, such as lung adenocarcinoma (LUAD) and the 95 percentile for NRG1 is 0.798 log10 TPM, as determined for TCGA-LUAD. In certain aspects, the cancer is lung cancer, such as lung adenocarcinoma (LUAD) and FAM83E expression is considered elevated when higher than a median log10 TPM level of a reference dataset, such as 1.258 log10 TPM as determined for TCGA-LUAD. In certain aspects, the cancer is lung cancer, such as lung adenocarcinoma (LUAD) and OLFML2B expression is considered decreased when lower than a median log10 TPM level of a reference dataset, such as below 1.250 log10 TPM as determined for TCGA-LUAD. In certain aspects, the cancer is lung cancer, such as lung squamous cell carcinoma (LUSC) and the 95 percentile for NRG1 is 1.452 log10 TPM, as determined for TCGA-LUSC. In certain aspects, the cancer is lung cancer, such as lung squamous cell carcinoma (LUSC) and FAM83E expression is considered elevated when higher than a median log10 TPM level of a reference dataset, such as 0.811 log10 TPM as determined for TCGA-LUSC. In certain aspects, the cancer is lung cancer, such as lung squamous cell carcinoma (LUSC) and OLFML2B expression is considered decreased when lower than a median log10 TPM level of a reference dataset, such as below 1.250 log10 TPM as determined for TCGA-LUSC. In certain aspects, the cancer is pancreatic cancer, such as pancreatic adenocarcinoma (PAAD) and the 95 percentile for NRG1 is 0.469 log10 TPM, as determined for TCGA-PAAD. In certain aspects, the cancer is pancreatic cancer, such as pancreatic adenocarcinoma (PAAD) and FAM83E expression is considered elevated when higher than a median log10 TPM level of a reference dataset, such as 1.359 log10 TPM as determined for TCGA-PAAD. In certain aspects, the cancer is pancreatic cancer, such as pancreatic adenocarcinoma (PAAD) and OLFML2B expression is considered decreased when lower than a median log10 TPM level of a reference dataset, such as below 1.659 log10 TPM as determined for TCGA-PAAD. In certain aspects, the cancer is esophageal cancer, such as esophageal adenocarcinoma (ESCA) and the 95 percentile for NRG1 is 1.733 log10 TPM, as determined for TCGA-ESCA. In certain aspects, the cancer is esophageal cancer, such as esophageal adenocarcinoma (ESCA) and FAM83E expression is considered elevated when higher than a median log10 TPM level of a reference dataset, such as 1.071 log10 TPM as determined for TCGA-ESCA. In certain aspects, the cancer is esophageal cancer, such as esophageal adenocarcinoma (ESCA) and OLFML2B expression is considered decreased when lower than a median log10 TPM level of a reference dataset, such as below 1.073 log10 TPM as determined for TCGA-ESCA. In certain aspects, methods of the present disclosure comprise determining the presence or amount of pHER3 from a sample obtained from the subject. Such a sample is obtained from a subject and contains ErbB-2, ErbB-3 or ErbB-2 / ErbB-3 positive cancer cells or cancer tissue from which said subject is suffering. In certain aspects, the presence or amount of pHER3 is determined.by immunohistochemistry (IHC), such as by using an antibody that specifically binds HER3 phosphorylated at position Tyr1289, such as monoclonal anti-rabbit antibody clone D1B5. In certain aspects, the sample is a formalin-fixed, paraffin-embedded human tissue of a solid tumor. In certain aspects, the sample is positive for pHER3, as established by IHC. In certain aspects, pHER3 is considered positive if the sample has an H-score of at least 20, or at least 40, or at least 120, or at least 150, or at least 170, as established by IHC. In certain aspects, the presence or amount of pHER3 is at least 20, or at least 40, or at least 120, or at least 150, or at least 170, as established by IHC and expressed as an H-score. In certain aspects, an H-score equals (0 x % [cat. 0]) + (1 x % [cat. 1+]) + (2 x % [cat. 2+]) + (3 x % [cat. 3+]). In certain aspects, the presence or amount of pHER3 is determined by in- situ hybridization (ISH). In certain aspects, the NRG1 of the present disclosure comprises the NRG1 gene identified by Entrez Gene ID No. 3084, or any allelic variant thereof. In certain aspects, the NRG1 comprises any one of SEQ ID No: 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, or any allelic variant thereof, wherein the allelic variant is defined as having at least 95% sequence identity to the NRG1 sequence. In certain aspects, the allelic variant is defined as having at least 98% sequence identity to the NRG1 sequence. the allelic variant is defined as having at least 99% sequence identity to the NRG1 sequence. In certain aspects, the NRG1 of the present disclosure comprises the NRG1 gene identified by Entrez Gene ID No. 3084, or any allelic variant thereof. In certain aspects, the NRG1 comprises any one of SEQ ID No: 129, 131, 133, 134, 136, 137, 139, 140, 141, 142, 143, 144, 145, 147, 149, 151, 152, 154, 155, 157, 159, 161, 163, 164, 165, 166, 167, 169, 171, 173, 175, 177, 178, 179, 181, 182, 183, or any allelic variant thereof, wherein the allelic variant is defined as having at least 95% sequence identity to the NRG1 sequence. In certain aspects, the allelic variant is defined as having at least 98% sequence identity to the NRG1 sequence. the allelic variant is defined as having at least 99% sequence identity to the NRG1 sequence. In certain aspects, an allelic variant is a variant that is at least 95% identical to the sequence of SEQ ID No: 129 or SEQ ID No: 130. In certain aspects, an allelic variant is a variant that is at least 98% identical to the sequence of SEQ ID No: 129 or SEQ ID No: 130. In certain aspects, an allelic variant is a variant that is at least 99% identical to the sequence of SEQ ID No: 129 or SEQ ID No: 130. In certain aspects, an allelic variant is a variant that is at least 95% identical to the sequence of SEQ ID No: 131 or SEQ ID No: 132. In certain aspects, an allelic variant is a variant that is at least 98% identical to the sequence of SEQ ID No: 131 or SEQ ID No: 132. In certain aspects, an allelic variant is a variant that is at least 99% identical to the sequence of SEQ ID No: 131 or SEQ ID No: 132. In certain aspects, an allelic variant is a variant that is at least 95% identical to the sequence of SEQ ID No: 133. In certain aspects, an allelic variant is a variant that is at least 98% identical to the sequence of SEQ ID No: 133. In certain aspects, an allelic variant is a variant that is at least 99% identical to the sequence of SEQ ID No: 133. In certain aspects, an allelic variant is a variant that is at least 95% identical to the sequence of SEQ ID No: 134 or SEQ ID No: 135. In certain aspects, an allelic variant is a variant that is at least 98% identical to the sequence of SEQ ID No: 134 or SEQ ID No: 135. In certain aspects, an allelic variant is a variant that is at least 99% identical to the sequence of SEQ ID No: 134 or SEQ ID No: 135. In certain aspects, an allelic variant is a variant that is at least 95% identical to the sequence of SEQ ID No: 136. In certain aspects, an allelic variant is a variant that is at least 98% identical to the sequence of SEQ ID No: 136. In certain aspects, an allelic variant is a variant that is at least 99% identical to the sequence of SEQ ID No: 136. In certain aspects, an allelic variant is a variant that is at least 95% identical to the sequence of SEQ ID No: 137 or SEQ ID No: 138. In certain aspects, an allelic variant is a variant that is at least 98% identical to the sequence of SEQ ID No: 137 or SEQ ID No: 138. In certain aspects, an allelic variant is a variant that is at least 99% identical to the sequence of SEQ ID No: 137 or SEQ ID No: 138. In certain aspects, an allelic variant is a variant that is at least 95% identical to the sequence of SEQ ID No: 139. In certain aspects, an allelic variant is a variant that is at least 98% identical to the sequence of SEQ ID No: 139. In certain aspects, an allelic variant is a variant that is at least 99% identical to the sequence of SEQ ID No: 139. In certain aspects, an allelic variant is a variant that is at least 95% identical to the sequence of SEQ ID No: 140. In certain aspects, an allelic variant is a variant that is at least 98% identical to the sequence of SEQ ID No: 140. In certain aspects, an allelic variant is a variant that is at least 99% identical to the sequence of SEQ ID No: 140. In certain aspects, an allelic variant is a variant that is at least 95% identical to the sequence of SEQ ID No: 141. In certain aspects, an allelic variant is a variant that is at least 98% identical to the sequence of SEQ ID No: 141. In certain aspects, an allelic variant is a variant that is at least 99% identical to the sequence of SEQ ID No: 141. In certain aspects, an allelic variant is a variant that is at least 95% identical to the sequence of SEQ ID No: 142. In certain aspects, an allelic variant is a variant that is at least 98% identical to the sequence of SEQ ID No: 142. In certain aspects, an allelic variant is a variant that is at least 99% identical to the sequence of SEQ ID No: 142. In certain aspects, an allelic variant is a variant that is at least 95% identical to the sequence of SEQ ID No: 143. In certain aspects, an allelic variant is a variant that is at least 98% identical to the sequence of SEQ ID No: 143. In certain aspects, an allelic variant is a variant that is at least 99% identical to the sequence of SEQ ID No: 143. In certain aspects, an allelic variant is a variant that is at least 95% identical to the sequence of SEQ ID No: 144. In certain aspects, an allelic variant is a variant that is at least 98% identical to the sequence of SEQ ID No: 144. In certain aspects, an allelic variant is a variant that is at least 99% identical to the sequence of SEQ ID No: 144. In certain aspects, an allelic variant is a variant that is at least 95% identical to the sequence of SEQ ID No: 145 or SEQ ID No: 146. In certain aspects, an allelic variant is a variant that is at least 98% identical to the sequence of SEQ ID No: 145 or SEQ ID No: 146. In certain aspects, an allelic variant is a variant that is at least 99% identical to the sequence of SEQ ID No: 145 or SEQ ID No: 146. In certain aspects, an allelic variant is a variant that is at least 95% identical to the sequence of SEQ ID No: 147 or SEQ ID No: 148. In certain aspects, an allelic variant is a variant that is at least 98% identical to the sequence of SEQ ID No: 147 or SEQ ID No: 148. In certain aspects, an allelic variant is a variant that is at least 99% identical to the sequence of SEQ ID No: 147 or SEQ ID No: 148. In certain aspects, an allelic variant is a variant that is at least 95% identical to the sequence of SEQ ID No: 149 or SEQ ID No: 150. In certain aspects, an allelic variant is a variant that is at least 98% identical to the sequence of SEQ ID No: 149 or SEQ ID No: 150. In certain aspects, an allelic variant is a variant that is at least 99% identical to the sequence of SEQ ID No: 149 or SEQ ID No: 150. In certain aspects, an allelic variant is a variant that is at least 95% identical to the sequence of SEQ ID No: 151 or SEQ ID No: 152. In certain aspects, an allelic variant is a variant that is at least 98% identical to the sequence of SEQ ID No: 151 or SEQ ID No: 152. In certain aspects, an allelic variant is a variant that is at least 99% identical to the sequence of SEQ ID No: 151 or SEQ ID No: 152. In certain aspects, an allelic variant is a variant that is at least 95% identical to the sequence of SEQ ID No: 153 or SEQ ID No: 154. In certain aspects, an allelic variant is a variant that is at least 98% identical to the sequence of SEQ ID No: 153 or SEQ ID No: 154. In certain aspects, an allelic variant is a variant that is at least 99% identical to the sequence of SEQ ID No: 153 or SEQ ID No: 154. In certain aspects, an allelic variant is a variant that is at least 95% identical to the sequence of SEQ ID No: 155 or SEQ ID No: 156. In certain aspects, an allelic variant is a variant that is at least 98% identical to the sequence of SEQ ID No: 155 or SEQ ID No: 156. In certain aspects, an allelic variant is a variant that is at least 99% identical to the sequence of SEQ ID No: 155 or SEQ ID No: 156. In certain aspects, an allelic variant is a variant that is at least 95% identical to the sequence of SEQ ID No: 157 or SEQ ID No: 158. In certain aspects, an allelic variant is a variant that is at least 98% identical to the sequence of SEQ ID No: 157 or SEQ ID No: 158. In certain aspects, an allelic variant is a variant that is at least 99% identical to the sequence of SEQ ID No: 157 or SEQ ID No: 158. In certain aspects, an allelic variant is a variant that is at least 95% identical to the sequence of SEQ ID No: 159 or SEQ ID No: 160. In certain aspects, an allelic variant is a variant that is at least 98% identical to the sequence of SEQ ID No: 159 or SEQ ID No: 160. In certain aspects, an allelic variant is a variant that is at least 99% identical to the sequence of SEQ ID No: 159 or SEQ ID No: 160. In certain aspects, an allelic variant is a variant that is at least 95% identical to the sequence of SEQ ID No: 161 or SEQ ID No: 162. In certain aspects, an allelic variant is a variant that is at least 98% identical to the sequence of SEQ ID No: 161 or SEQ ID No: 162. In certain aspects, an allelic variant is a variant that is at least 99% identical to the sequence of SEQ ID No: 161 or SEQ ID No: 162. In certain aspects, an allelic variant is a variant that is at least 95% identical to the sequence of SEQ ID No: 163. In certain aspects, an allelic variant is a variant that is at least 98% identical to the sequence of SEQ ID No: 163. In certain aspects, an allelic variant is a variant that is at least 99% identical to the sequence of SEQ ID No: 163. In certain aspects, an allelic variant is a variant that is at least 95% identical to the sequence of SEQ ID No: 164. In certain aspects, an allelic variant is a variant that is at least 98% identical to the sequence of SEQ ID No: 164. In certain aspects, an allelic variant is a variant that is at least 99% identical to the sequence of SEQ ID No: 164. In certain aspects, an allelic variant is a variant that is at least 95% identical to the sequence of SEQ ID No: 165. In certain aspects, an allelic variant is a variant that is at least 98% identical to the sequence of SEQ ID No: 165. In certain aspects, an allelic variant is a variant that is at least 99% identical to the sequence of SEQ ID No: 165. In certain aspects, an allelic variant is a variant that is at least 95% identical to the sequence of SEQ ID No: 166. In certain aspects, an allelic variant is a variant that is at least 98% identical to the sequence of SEQ ID No: 166. In certain aspects, an allelic variant is a variant that is at least 99% identical to the sequence of SEQ ID No: 166. In certain aspects, an allelic variant is a variant that is at least 95% identical to the sequence of SEQ ID No: 167 or SEQ ID No: 168. In certain aspects, an allelic variant is a variant that is at least 98% identical to the sequence of SEQ ID No: 167 or SEQ ID No: 168. In certain aspects, an allelic variant is a variant that is at least 99% identical to the sequence of SEQ ID No: 167 or SEQ ID No: 168. In certain aspects, an allelic variant is a variant that is at least 95% identical to the sequence of SEQ ID No: 169 or SEQ ID No: 170. In certain aspects, an allelic variant is a variant that is at least 98% identical to the sequence of SEQ ID No: 169 or SEQ ID No: 170. In certain aspects, an allelic variant is a variant that is at least 99% identical to the sequence of SEQ ID No: 169 or SEQ ID No: 170. In certain aspects, an allelic variant is a variant that is at least 95% identical to the sequence of SEQ ID No: 171 or SEQ ID No: 172. In certain aspects, an allelic variant is a variant that is at least 98% identical to the sequence of SEQ ID No: 171 or SEQ ID No: 172. In certain aspects, an allelic variant is a variant that is at least 99% identical to the sequence of SEQ ID No: 171 or SEQ ID No: 172. In certain aspects, an allelic variant is a variant that is at least 95% identical to the sequence of SEQ ID No: 173 or SEQ ID No: 174. In certain aspects, an allelic variant is a variant that is at least 98% identical to the sequence of SEQ ID No: 173 or SEQ ID No: 174. In certain aspects, an allelic variant is a variant that is at least 99% identical to the sequence of SEQ ID No: 173 or SEQ ID No: 174. In certain aspects, an allelic variant is a variant that is at least 95% identical to the sequence of SEQ ID No: 175 or SEQ ID No: 176. In certain aspects, an allelic variant is a variant that is at least 98% identical to the sequence of SEQ ID No: 175 or SEQ ID No: 176. In certain aspects, an allelic variant is a variant that is at least 99% identical to the sequence of SEQ ID No: 175 or SEQ ID No: 176. In certain aspects, an allelic variant is a variant that is at least 95% identical to the sequence of SEQ ID No: 177. In certain aspects, an allelic variant is a variant that is at least 98% identical to the sequence of SEQ ID No: 177. In certain aspects, an allelic variant is a variant that is at least 99% identical to the sequence of SEQ ID No: 177. In certain aspects, an allelic variant is a variant that is at least 95% identical to the sequence of SEQ ID No: 178. In certain aspects, an allelic variant is a variant that is at least 98% identical to the sequence of SEQ ID No: 178. In certain aspects, an allelic variant is a variant that is at least 99% identical to the sequence of SEQ ID No: 178. In certain aspects, an allelic variant is a variant that is at least 95% identical to the sequence of SEQ ID No: 179 or SEQ ID No: 180. In certain aspects, an allelic variant is a variant that is at least 98% identical to the sequence of SEQ ID No: 179 or SEQ ID No: 180. In certain aspects, an allelic variant is a variant that is at least 99% identical to the sequence of SEQ ID No: 179 or SEQ ID No: 180. In certain aspects, an allelic variant is a variant that is at least 95% identical to the sequence of SEQ ID No: 181. In certain aspects, an allelic variant is a variant that is at least 98% identical to the sequence of SEQ ID No: 181. In certain aspects, an allelic variant is a variant that is at least 99% identical to the sequence of SEQ ID No: 181. In certain aspects, an allelic variant is a variant that is at least 95% identical to the sequence of SEQ ID No: 182. In certain aspects, an allelic variant is a variant that is at least 98% identical to the sequence of SEQ ID No: 182. In certain aspects, an allelic variant is a variant that is at least 99% identical to the sequence of SEQ ID No: 182. In certain aspects, an allelic variant is a variant that is at least 95% identical to the sequence of SEQ ID No: 183. In certain aspects, an allelic variant is a variant that is at least 98% identical to the sequence of SEQ ID No: 183. In certain aspects, an allelic variant is a variant that is at least 99% identical to the sequence of SEQ ID No: 183. In certain aspects, FAM83E expression comprises FAM83E mRNA expression or mRNA sequence reads and OLFML2B expression comprises OLFML2B mRNA expression or mRNA sequence reads. In certain aspects, FAM83E comprises HGNC: 25972, also known as Ensembl: ENSG00000105523. In certain aspects, FAM83E mRNA sequence comprises SEQ ID No 195, or a sequence having at least 98% sequence identity thereto. In certain aspects, FAM83E mRNA sequence comprises SEQ ID No 197, or a sequence having at least 98% sequence identity thereto. In certain aspects, OLFML2B comprises HGNC: 24558, also known as Ensembl: ENSG00000162745. In certain aspects, OLFML2B mRNA sequence comprises SEQ ID No: 199, or a sequence having at least 98% sequence identity thereto. In certain aspects, OLFML2B mRNA sequence comprises SEQ ID No: 201, or a sequence having at least 98% sequence identity thereto. In certain aspects, OLFML2B mRNA sequence comprises SEQ ID No: 203, or a sequence having at least 98% sequence identity thereto. A comparison of sequences and determination of percentage of sequence identity between two sequences can be accomplished using a mathematical algorithm. The skilled person will be aware of the fact that several different computer programs are available to align two sequences and determine the identity between two sequences (Kruskal JB (1983) An overview of sequence comparison: Time warps, string edits and macromolecules. Society for Industrial and Applied Mathematics 25(2):201-237.). The percent sequence identity between two amino acid sequences or nucleic acid sequences may be determined using the Needleman and Wunsch algorithm for the alignment of two sequences (Needleman SB and Wunsch CD (1970). A general method applicable to the search for similarities in the amino acid sequence of two proteins. J. Mol. Biol. 48(3), 443-453). The Needleman-Wunsch algorithm has been implemented in the computer program NEEDLE. For the purpose of this disclosure, the NEEDLE program from the EMBOSS package is used to determine percent identity of amino acid and nucleic acid sequences (version 2.8.0, Rice P, Longden I, and Bleasby A (2000). EMBOSS: The European Molecular Biology Open Software Suite . Trends in Genetics 16(6):276— 277, http: / / emboss.bioinformatics.nl / ). For protein sequences, EBLOSUM62 is used for the substitution matrix. For DNA sequences, DNAFULL is used. The parameters used are a gap-open penalty of 10 and a gap extension penalty of 0.5. After alignment by the program NEEDLE as described above the percentage of sequence identity between a query sequence and a sequence of this disclosure is calculated as follows: Number of corresponding positions in the alignment showing an identical amino acid or identical nucleotide in both sequences divided by the total length of the alignment after subtraction of the total number of gaps in the alignment. The NRG1-gene codes for the various isoforms of NRG1. Various isoforms and their expected function are described in Adélaïde J et al. (2003). A recurrent chromosome translocation breakpoint in breast and pancreatic cancer cell lines targets the neuregulin / NRG1 gene. Genes Chromosomes Cancer. 37(4):333-45. GGF and GGF2 isoforms contain a kringle-like sequence plus Ig and EGF-like domains; and the SMDF isoform shares only the EGF-like domain with other isoforms. The EGF-like domain is encoded by the 3’ end of the gene. In certain aspects, the EGF-like domain is present in NRG1 isoforms of the present disclosure. In certain aspects, the cancer or any sample of the present disclosure is not NRG1 fusion positive. Herein, the term ‘NRG1 fusion positive’ is understood to mean that the subject or sample obtained therefrom comprises a cancer cell, or cells, that comprises an NRG1 fusion gene comprising at least a portion of the NRG1-gene fused to a sequence from a different chromosomal location. Typically, the cell comprises an NRG1 fusion gene comprising at least the 3’ end of the NRG1-gene fused to a 5’ sequence from a different chromosomal location. The said cancer cell may be a cancer cell associated with an NRG1-fusion gene, such as a cancer cell driven by an NRG1-fusion. Such an NRG1-fusion gene expresses a protein that comprises an NRG1 EGF-like domain. In certain aspects, the NRG1 fusion gene is a fusion of the 3’ end of the NRG1-gene with the 5’ sequence of another gene. Reference herein is made to for instance Duruisseaux et al. (2016) NRG1 fusion in a French cohort of invasive mucinous lung adenocarcinoma. Cancer Med 5(12): 3579–3585, for examples on NRG1-fusion genes. ErbB-1 is known under various synonyms, the most common of which is EGFR. EGFR has an extracellular domain (ECD) that is composed of four sub- domains, two of which are involved in ligand binding and two of which are involved in homo-dimerisation and hetero-dimerisation. EGFR integrates extracellular signals from a variety of ligands to yield diverse intracellular responses. The major signal transduction pathway activated by EGFR is composed of the Ras-mitogen- activated protein kinase (MAPK) mitogenic signalling cascade. Activation of this pathway is initiated by the recruitment of Grb2 to tyrosine phosphorylated EGFR. This leads to activation of Ras through the Grb2-bound Ras-guanine nucleotide exchange factor Son of Sevenless (SOS). In addition, the PI3-kinase-Akt signal transduction pathway is also activated by EGFR, although this activation is much stronger in case there is co-expression of ErbB-3 (HER3). EGFR is implicated in several human epithelial malignancies, notably cancers of the breast, bladder, non- small cell lung cancer lung, colon, ovarian head and neck and brain. Activating mutations in the gene have been found, as well as over-expression of the receptor and of its ligands, giving rise to autocrine activation loops. This RTK has therefore been extensively used as target for cancer therapy. Both small-molecule inhibitors targeting the RTK and monoclonal antibodies (mAbs) directed to the extracellular ligand-binding domains have been developed and have shown hitherto several clinical successes, albeit mostly for a select group of patients. The database accession number for the human EGFR protein and the gene encoding it is (GenBank NM_005228.3). This accession number is primarily given to provide a further method of identification of EGFR protein as a target, the actual sequence of the EGFR protein bound by an antibody may vary, for instance because of a mutation in the encoding gene such as those occurring in some cancers or the like. Where reference herein is made to EGFR, the reference refers to human EGFR unless otherwise stated. The antigen-binding site that binds EGFR, binds EGFR and a variety of variants thereof such as those expressed on some EGFR positive tumors. The term ‘ErbB-1’ as used herein refers to the protein that in humans is encoded by the ERBB-1 gene. Alternative names for the gene or protein include EGFR, ERBB, HER1, Erb-B2 receptor tyrosine kinase 1. Where reference is made herein to ErbB-1, the reference refers to human ErbB-1. The term ‘ErbB-2’ as used herein refers to the protein that in humans is encoded by the ERBB-2 gene. Alternative names for the gene or protein include CD340; HER-2; HER-2 / neu; MLN 19; NEU; NGL; TKR1. The ERBB-2 gene is frequently called HER2 (from human epidermal growth factor receptor 2). Where reference is made herein to ErbB-2, the reference refers to human ErbB-2. An antibody comprising an antigen-binding site that binds ErbB-2, binds human ErbB-2. The ErbB-2 antigen-binding site may, due to sequence and tertiary structure similarity between human and other mammalian orthologs, also bind such an ortholog but not necessarily so. Database accession numbers for the human ErbB-2 protein and the gene encoding it are (NP_001005862.1, NP_004439.2 NC_000017.10 NT_010783.15 NC_018928.2). The accession numbers are primarily given to provide a further method of identification of ErbB-2 as a target, the actual sequence of the ErbB-2 protein bound the antibody may vary, for instance because of a mutation in the encoding gene such as those occurring in some cancers or the like. The ErbB-2 antigen binding site binds ErbB-2 and a variety of variants thereof, such as those expressed by some ErbB-2 positive cells or tumor cells. In certain aspects, the antigen-binding site that binds ErbB-2 binds domain I of ErbB-2. The term ‘ErbB-3’ as used herein refers to the protein that in humans is encoded by the ERBB3 gene. Alternative names for the gene or protein are HER3; LCCS2; MDA-BF-1; c-ErbB-3; c-ErbB3; ErbB3-S; p180-ErbB3; p45-sErbB3; and p85-sErbB3. Where reference is made herein to ErbB-3, the reference refers to human ErbB-3. An antibody comprising an antigen-binding site that binds ErbB-3, binds human ErbB-3. The ErbB-3 antigen-binding site may, due to sequence and tertiary structure similarity between human and other mammalian orthologs, also bind such an ortholog but not necessarily so. Database accession numbers for the human ErbB-3 protein and the gene encoding it are (NP_001005915.1, NP_001973.2, NC_000012.11, NC_018923.2, NT_029419.12). The accession numbers are primarily given to provide a further method of identification of ErbB-3 as a target, the actual sequence of the ErbB-3 protein bound by an antibody may vary, for instance because of a mutation in the encoding gene such as those occurring in some cancers or the like. The ErbB-3 antigen binding site binds ErbB- 3 and a variety of variants thereof, such as those expressed by some ErbB-3 positive tumor cells. In certain aspects, the antigen-binding site that binds ErbB-3 binds domain III of ErbB-3. Herein, the terms ‘HER3’ and ‘ErbB3’ are used interchangeably. The term ‘ErbB-4’ as used herein refers to the protein that in humans is encoded by the ERBB-4 gene. Alternative names for the gene or protein include HER4, Erb-B2 receptor tyrosine kinase 4, and Human Epidermal Growth Factor Receptor 4. Where reference is made herein to ErbB-4, the reference refers to human ErbB-4. In certain aspects, the cancer comprises an oncogenic mutation in KRAS. In certain aspects, the cancer comprises an oncogenic missense mutation in KRAS, such as mutation G12D. Mutation G12D confers a structural conformation that can influence downstream signaling and may lead to oncogenic activity. The words cancer and tumor are used herein typically both refer to cancer, unless otherwise specifically stated. Antigen-binding sites in an antibody are typically present in the variable domains. The variable domains comprise a heavy chain variable region and a light chain variable region. In certain aspects, said positive cancer is a gastric cancer. However, the invention can be applied to a wide range of ErbB-2, ErbB-3 or ErbB-2 / ErbB-3 positive cancers, like brain cancer, breast cancer, pancreatic cancer, lung cancer, esophageal cancer, liver cancer and cholangiocarcinoma, the like. In certain aspects, the malignant cell is an epithelial cell. Alternatively, the cell or tumor is a cell or tumor of an epithelial origin. In certain aspects, the cell or tumor is a lung cancer, or a metastasis thereof. In certain aspects, the tumor is of epithelial origin. In certain aspects, the cells of the ErbB-2 / ErbB-3 positive cell or tumor have elevated levels of heregulin expression. Heregulin is a growth factor that is involved in growth of ErbB-3 positive cell or tumor cells. Typically, when the cell or tumor cells express high levels of heregulin , currently known therapies like trastuzumab, pertuzumab and lapatinib are no longer capable of inhibiting cell or tumor growth. This phenomenon is called heregulin resistance. Elevated heregulin levels are typically present during the formation of metastases (i.e. the migration, invasion, growth and / or differentiation of a cell or tumor cells or tumor initiating cells). Typically, tumor initiating cells are identified based on stem cell markers such as for instance CD44, CD24, CD133 and / or ALDH1. These processes can therefore barely be counteracted with currently known therapies like trastuzumab and pertuzumab. The bispecific antibodies disclosed herein are capable of counteracting the formation of metastases in subjects having cells tumors that comprise elevated NRG1 expression levels. In certain aspects, the subject is a human subject. In certain aspects, the subject is a subject eligible for monoclonal antibody therapy using an ErbB-2 specific antibody such as trastuzumab. In certain aspects, the subject has received prior treatment with an ErbB-2 and / or ErbB-3 targeting agent. In certain aspects, the subject or cancer has progressed after having received a prior treatment with a monospecific bivalent antibody comprising antigen-binding sites that bind an extracellular part of ErbB-2 or an extracellular part of ErbB-3, or a prior treatment with a tyrosine kinase inhibitor (TKI) of ErbB-2 or chemotherapy or with a combination thereof. In certain aspects, the TKI is one or more of lapatinib, canertinib, neratinib, tucatinib, CP-724714, tarloxitinib, mubritinib, afatinib, varlitinib, and dacomitinib. In certain aspects, the TKI is afatinib. In certain aspects, the monospecific bivalent antibody comprising antigen-binding sites that bind an extracellular part of ErbB-3 comprises patritumab, seribantumab, lumretuzumab, elgemtumab, GSK2849330, KTN3379 or AV-203. In certain aspects, the subject has received prior treatment with trastuzumab or pertuzumab. The amount of ErbB-2 and / or ErbB-3 targeting agent to be administered to a patient is typically in the therapeutic window, meaning that a sufficient quantity is used for obtaining a therapeutic effect, while the amount does not exceed a threshold value leading to an unacceptable extent of side-effects. The lower the amount of antibody needed for obtaining a desired therapeutic effect, the larger the therapeutic window will typically be. The selected dosage level will depend upon a variety of factors including the route of administration, the time of administration, the rate of excretion of the particular compound being employed, the duration of the treatment, other drugs, compounds and / or materials used in combination, the age, sex, weight, condition, general health and prior medical history of the patient being treated, and like factors well known in the medical arts. The dosage can be in the range of the dosing regime for trastuzumab or lower. The ErbB2 and / or ErbB3 targeting agent can be formulated as a pharmaceutical composition comprising a pharmaceutically acceptable carrier, diluent, or excipient, and additional, optional, active agents. The antibodies and compositions comprising the antibodies can be administered by any route including parenteral, enteral, and topical administration. Parenteral administration is usually by injection, and includes, e.g., intravenous, intramuscular, intraarterial, intrathecal, intraventricular, intracapsular, intraorbital, intracardiac, intradermal, intraperitoneal, transtracheal, subcutaneous, subcuticular, intraarticular, subcapsular, subarachnoid, intraspinal, intracerebrospinal, intratumoral, and intrasternal injection and infusion. In certain aspects, the bispecific antibody of the present disclosure is administered or used in an amount of 750 mg once every two weeks, such as a flat dose. In certain aspects, the bispecific antibody is zenocutuzumab. The disclosure provides bispecific antibodies for use in the methods and treatments described herein. Suitable bispecific antibodies comprise a first antigen- binding site that binds ErbB-2 and a second antigen-binding site that binds ErbB- 3, wherein the bispecific antibody reduces or can reduce a ligand-induced receptor function of ErbB-3 on an ErbB-2 and ErbB-3 positive cell. Antibodies and their preparation are disclosed in WO 2015 / 130173, which is hereby incorporated by reference. The examples in WO 2015 / 130173 further describe a number of properties of the antibodies, such as ligand binding and epitope mapping. As used herein, the term ‘antigen-binding site’ refers to a site derived from a bispecific antibody which is capable of binding to antigen. An unmodified antigen- binding site is typically formed by and present in the variable domain of the antibody. The variable domain contains said antigen-binding site. A variable domain that binds an antigen is a variable domain comprising an antigen-binding site that binds the antigen. In certain aspects, an antibody variable domain comprises a heavy chain variable region (VH) and a light chain variable region (VL). The antigen-binding site can be present in the combined VH / VL variable domain, or in only the VH region or only the VL region. When the antigen-binding site is present in only one of the two regions of the variable domain, the counterpart variable region can contribute to the folding and / or stability of the binding variable region, but does not significantly contribute to the binding of the antigen itself. As used herein, antigen-binding refers to the typical binding capacity of an antibody to its antigen. An antibody comprising an antigen-binding site that binds to ErbB-2, binds to ErbB-2 and, under otherwise identical conditions, at least 100- fold lower to the homologous receptors ErbB-1 and ErbB-4 of the same species. An antibody comprising an antigen-binding site that binds to ErbB-3, binds to ErbB-3 and, under otherwise identical conditions, not to the homologous receptors ErbB-1 and ErbB-4 of the same species. Considering that the ErbB-family is a family of cell surface receptors, the binding is typically assessed on cells that express the receptor(s). Binding of an antibody to an antigen can be assessed in various ways. One way is to incubate the antibody with the antigen (e.g. cells expressing the antigen), removing unbound antibody (e.g. by a wash step) and detecting bound antibody by means of a labeled antibody that binds to the bound antibody. Antigen binding by an antibody is typically mediated through the complementarity regions of the antibody and the specific three-dimensional structure of both the antigen and the variable domain allowing these two structures to bind together with precision (an interaction similar to a lock and key), as opposed to random, non-specific sticking of antibodies. As an antibody typically recognizes an epitope of an antigen, and as such epitope may be present in other compounds as well, antibodies according to the present invention that bind ErbB-2 and / or ErbB-3 may recognize other proteins as well, if such other compounds contain the same epitope. Hence, the term ‘binding’ does not exclude binding of the antibodies to another protein or protein(s) that contain the same epitope. In certain aspects, such other protein(s) is not a human protein. An ErbB-2 antigen-binding site and an ErbB-3 antigen-binding site as defined herein typically do not bind to other proteins on the membrane of cells in a post-natal, typically adult human. A bispecific antibody as disclosed herein is typically capable of binding ErbB-2 and ErbB-3 with a binding affinity of at least 1x10e-6 M, as outlined in more detail below. The term ‘interferes with binding’ as used herein means that the antibody is directed to an epitope on ErbB-3 and the antibody competes with ligand for binding to ErbB-3. The antibody may diminish ligand binding, displace ligand when this is already bound to ErbB-3 or it may, for instance through steric hindrance, at least partially prevent that ligand can bind to ErbB-3. The term ‘antibody’ as used herein means a proteinaceous molecule which is certain aspects belongs to the immunoglobulin class of proteins, containing one or more variable domains that bind an epitope on an antigen, where such domains are derived from or share sequence homology with the variable domain of an antibody. In certain aspects, antibodies for therapeutic use are as close to natural antibodies of the subject to be treated as possible (for instance human antibodies for human subjects). Antibody binding can be expressed in terms of specificity and affinity. The specificity determines which antigen or epitope thereof is specifically bound by the binding domain. The affinity is a measure for the strength of binding to a particular antigen or epitope. Specific binding is defined as binding with affinities (KD) of at least 1x10e-6 M, or 1x10e-7 M, or higher than 1x10e-9 M. Typically, antibodies for therapeutic applications have affinities of up to 1x10e-10 M or higher. Antibodies such the bispecific antibodies of the present invention comprise the constant domains (Fc part) of a natural antibody. In certain aspects, an antibody of the invention is typically a bispecific full length antibody, e.g. of the human IgG subclass. In certain aspects, an antibody as disclosed herein is of the human IgG1 subclass. Such antibodies have good ADCC properties, have favorable half-life upon in vivo administration to humans and CH3 engineering technology exists that can provide for modified heavy chains that preferentially form heterodimers over homodimers upon co-expression in clonal cells. In certain aspects, an antibody as disclosed herein is a ‘‘full length’’ antibody. The term ‘full length’ is defined as comprising an essentially complete antibody, which however does not necessarily have all functions of an intact antibody. For the avoidance of doubt, a full length antibody contains two heavy and two light chains. Each chain contains constant (C) and variable (V) regions, which can be broken down into domains designated CH1, CH2, CH3, VH, and CL, VL. An antibody binds to antigen via the variable domains contained in the Fab portion, and after binding can interact with molecules and cells of the immune system through the constant domains, mostly through the Fc portion. The terms ‘variable domain’, ‘VH / VL pair’, ‘VH / VL’ are used herein interchangeably. Full length antibodies according to the invention encompass antibodies wherein mutations may be present that provide desired characteristics. Such mutations should not be deletions of substantial portions of any of the regions. However, antibodies wherein one or several amino acid residues are deleted, without essentially altering the binding characteristics of the resulting antibody are embraced within the term ‘full length antibody’. For instance, an IgG antibody can have 1-20 amino acid residue insertions, deletions or a combination thereof in the constant region. For instance, ADCC activity of an antibody can be improved when the antibody itself has a low ADCC activity, by slightly modifying the constant region of the antibody (Junttila TT, Parsons K et al. (2010). Superior In vivo Efficacy of Afucosylated Trastuzumab in the Treatment of HER2-Amplified Breast Cancer. Cancer Research 70(11): 4481-4489) In certain aspects, antibodies are full length IgG antibodies because of their favourable half life and the need to stay as close to fully autologous (human) molecules for reasons of immunogenicity. In certain aspects, an antibody as disclosed herein is a bispecific IgG antibody, such as a bispecific full length IgG1 antibody. IgG1 is favoured based on its long circulatory half-life in man. In order to prevent any immunogenicity in humans, the bispecific IgG antibody may be a human IgG1. The term ‘bispecific’ (bs) means that one part of the antibody (as defined above) binds to one epitope on an antigen whereas a second part binds to a different epitope. The different epitope is typically present on a different antigen. The first and second antigens are in fact two different proteins. In certain aspects, a bispecific antibody is an antibody that comprises parts of two different monoclonal antibodies and consequently binds to two different types of antigen. One arm of the bispecific antibody typically contains the variable domain of one antibody and the other arm contains the variable domain of another antibody. The heavy chain variable regions of the bispecific antibody are typically different from each other, whereas the light chain variable regions are typically the same. A bispecific antibody wherein the different heavy chain variable regions are associated with the same, or a common, light chain is also referred to as a bispecific antibody with a common light chain. In certain aspects, bispecific antibodies are obtained by co-expression of two different heavy chains and a common light chain in a single cell. When wildtype CH3 domains are used, co-expression of two different heavy chains and a common light chain will result in three different species, AA, AB and BB. To increase the percentage of the desired bispecific product (AB) CH3 engineering can be employed, or in other words, one can use heavy chains with compatible heterodimerization domains, as defined hereunder. The term ‘compatible heterodimerization domains’ as used herein refers to protein domains that are engineered such that engineered domain A’ will typically form heterodimers with engineered domain B’ and vice versa, whereas homodimerization between A’-A’ and B’-B’ is diminished. The term ‘common light chain’ refers to light chains which may be identical or have some amino acid sequence differences while the binding specificity of the full length antibody is not affected. It is for instance possible, to prepare or find light chains that are not identical but still functionally equivalent, e.g., by introducing and testing conservative amino acid changes, changes of amino acids in regions that do not or only partly contribute to binding specificity when paired with the heavy chain, and the like. The terms ‘common light chain’, ‘common VL’, ‘single light chain’, ‘single VL’, with or without the addition of the term ‘rearranged’ are all used herein interchangeably. In certain aspects, a common light chain (variable region) has a germline sequence. In certain aspects, germline sequence is a light chain variable region that is frequently used in the human repertoire and has good thermodynamic stability, yield and solubility. In certain aspects, the light chain comprises a light chain region comprising the amino acid sequence of an O12 / IgVκ1-39*01 gene segment as depicted in the common light chain IGKV1-39 / jk1 sequences with 0-10, typically 0-5 amino acid insertions, deletions, substitutions, additions or a combination thereof. IgVκ1-39 is short for Immunoglobulin Variable Kappa 1-39 Gene. The gene is also known as Immunoglobulin Kappa Variable 1-39; IGKV139; IGKV1-39; O12a or O12. External Ids for the gene are HGNC: 5740; Entrez Gene: 28930; Ensembl: ENSG00000242371. The variable region of IGKV1-39 is listed in the common light chain sequences (Figure 2C and 2D). The V-region can be combined with one of five J-regions. The common light chain sequences herein (Figure 2C and 2D) describe two possible sequences for IgVκ1-39 in combination with a J-region. The joined sequences are indicated as IGKV1-39 / jk1 and IGKV1- 39 / jk5; alternative names are IgVκ1-39*01 / IGJκ1*01 or IgVκ1-39*01 / IGJκ5*01 (nomenclature according to the IMGT database worldwide web at imgt.org). In certain aspects, the O12 / IgVκ1-39*01 comprising light chain variable region is a germline sequence. In certain aspects, the IGJκ1*01 or / IGJκ5*01 comprising light chain variable region is a germline sequence. In certain aspects, the IGKV1-39 / jk1 or IGKV1-39 / jk5 light chain variable regions are germline sequences. In certain aspects, the light chain variable region comprises a germline O12 / IgVκ1-39*01. In certain aspects, the light chain variable region comprises the kappa light chain IgVκ1-39*01 / IGJκ1*01 or IgVκ1-39*01 / IGJκ5*01, in certain aspects, a IgVκ1-39*01 / IGJκ1*01. In certain aspects, the light chain variable region comprises a germline kappa light chain IgVκ1-39*01 / IGJκ1*01 or germline kappa light chain IgVκ1-39*01 / IGJκ5*01, or typically a germline IgVκ1- 39*01 / IGJκ1*01. Those of skill in the art will recognize that ‘common’ also refers to functional equivalents of the light chain of which the amino acid sequence is not identical. Many variants of said light chain exist wherein mutations (deletions, substitutions, additions) are present that do not materially influence the formation of functional binding regions. The light chain can also be a light chain as specified herein above, having 1-5 amino acid insertions, deletions, substitutions or a combination thereof. In certain aspects, both the first antigen binding site and said second antigen binding site comprise a light chain variable region comprising a CDR1 having the sequence (QSISSY), a CDR2 having the sequence (AAS), and a CDR3 having the sequence (QQSYSTPPT). The antibodies disclosed herein can reduce a ligand-induced receptor function of ErbB-3 on an ErbB-2 and ErbB-3 positive cell. In the presence of excess ErbB-2, ErbB-2 / ErbB-3 heterodimers may provide a growth signal to the expressing cell in the absence of detectable ligand for the ErbB-3 chain in the heterodimer. This ErbB-3 receptor function is herein referred as a ligand- independent receptor function of ErbB-3. The ErbB-2 / ErbB-3 heterodimer also provide a growth signal to the expressing cell in the presence of an ErbB-3 ligand. This ErbB-3 receptor function is herein referred to as a ligand-induced receptor function of ErbB-3. The term ‘ErbB-3 ligand’ as used herein refers to polypeptides which bind and activate ErbB-3. Examples of ErbB-3 ligands include, but are not limited to neuregulin 1 (NRG) and neuregulin 2, betacellulin, heparin-binding epidermal growth factor, and epiregulin. The term includes biologically active fragments and / or variants of a naturally occurring polypeptide. The ErbB-2 protein contains several domains (see for reference figure 1 of Landgraf, R. Breast Cancer Res. 2007; 9(1): 202-). The extracellular domains are referred to as domains I-IV. The place of binding to the respective domains of antigen-binding sites of antibodies described herein has been mapped. A bispecific antibody with an antigen-binding site (first antigen-binding site) that binds domain I or domain IV of ErbB-2 (first antigen-binding site) comprises a heavy chain variable region that maintains significant binding specificity and affinity for ErbB-2 when combined with various light chains. Bispecific antibodies with an antigen-binding site (first antigen-binding site) that binds domain I or domain IV of ErbB-2 (first antigen-binding site) and an antigen-binding site for ErbB-3 (second antigen-binding site) are more effective in reducing a ligand-induced receptor function of ErbB-3 when compared to a bispecific antibody comprising an antigen-binding site (first antigen-binding site) that binds to another extra-cellular domain of ErbB-2. In certain aspects, a bispecific antibody comprising an antigen- binding site (first antigen-binding site) that binds ErbB-2, wherein said antigen- binding site binds to domain I of ErbB-2. In certain aspects, said bispecific antibodies comprise a first antigen-binding site that binds domain I of ErbB-2 and a second antigen-binding site that binds domain III of ErbB-3. In certain aspects, said antibody comprises an antigen-binding site that binds at least one amino acid of domain I of ErbB-2 selected from the group consisting of T144, T164, R166, P172, G179, S180 and R181, and surface-exposed amino acid residues that are located within about 5 amino acid positions from T144, T164, R166, P172, G179, S180 or R181. In certain aspects, said antibody comprises an antigen-binding site that binds at least one amino acid of domain III of ErbB-3 selected from the group consisting of R426 and surface-exposed amino acid residues that are located within 11.2 Å from R426 in the native ErbB-3 protein. A bispecific antibody with an antigen-binding site (first antigen-binding site) that binds ErbB-2, and that further comprises ADCC are more effective than other ErbB-2 binding antibodies that did not have significant ADCC activity, particularly in vivo. In certain aspects, a bispecific antibody thus exhibits ADCC. It was found that antibodies wherein said first antigen-binding site binds to domain IV of ErbB-2 had intrinsic ADCC activity. A domain I binding ErbB-2 binding antibody that has low intrinsic ADCC activity can be engineered to enhance the ADCC activity Fc regions mediate antibody function by binding to different receptors on immune effector cells such as macrophages, natural killer cells, B-cells and neutrophils. Some of these receptors, such as CD16A (FcγRIIIA) and CD32A (FcγRIIA), activate the cells to build a response against antigens. Other receptors, such as CD32B, inhibit the activation of immune cells. By engineering Fc regions (through introducing amino acid substitutions) that bind to activating receptors with greater selectivity, antibodies can be created that have greater capability to mediate cytotoxic activities desired by an anti-cancer Mab. One technique for enhancing ADCC of an antibody is afucosylation. (See for instance Junttila TT, Parsons K et al. (2010). Superior In vivo Efficacy of Afucosylated Trastuzumab in the Treatment of HER2-Amplified Breast Cancer. Cancer Research 70(11): 4481-4489). Further provided is therefore a bispecific antibody as disclosed herein, which is afucosylated. Alternatively, or additionally, multiple other strategies can be used to achieve ADCC enhancement, for instance including glycoengineering (Kyowa Hakko / Biowa, GlycArt (Roche) and Eureka Therapeutics) and mutagenesis (Xencor and Macrogenics), all of which seek to improve Fc binding to low-affinity activating FcγRIIIa, and / or to reduce binding to the low affinity inhibitory FcγRIIb. Several in vitro methods exist for determining the efficacy of antibodies or effector cells in eliciting ADCC. Among these are chromium-51 [Cr51] release assays, europium [Eu] release assays, and sulfur-35 [S35] release assays. Usually, a labeled target cell line expressing a certain surface-exposed antigen is incubated with antibody specific for that antigen. After washing, effector cells expressing Fc receptor CD16 are typically co-incubated with the antibody-labeled target cells. Target cell lysis is subsequently typically measured by release of intracellular label, for instance by a scintillation counter or spectrophotometry. In certain aspects, the bispecific antibodies used in the disclosed methods do not significantly affect the survival of cardiomyocytes. Cardiotoxicity is a known risk factor in ErbB-2 targeting therapies and the frequency of complications is increased when trastuzumab is used in conjunction with anthracyclines thereby inducing cardiac stress. In certain aspects, the bispecific antibodies disclosed herein are used in humans. Thus in certain aspects, antibodies are human or humanized antibodies. Tolerance of a human to a polypeptide is governed by many different aspects. Immunity, be it T-cell mediated, B-cell mediated or other is one of the variables that are encompassed in tolerance of the human for a polypeptide. In certain aspects, the constant region of a bispecific antibody is a human constant region. The constant region may contain one or more, typically not more than 10, such as not more than 5 amino-acid differences with the constant region of a naturally occurring human antibody. In certain aspects, the constant part is entirely derived from a naturally occurring human antibody. Various antibodies produced herein are derived from a human antibody variable domain library. As such these variable domains are human. The unique CDR regions may be derived from humans, be synthetic or derived from another organism. The variable region is considered a human variable region when it has an amino acid sequence that is identical to an amino acid sequence of the variable region of a naturally occurring human antibody, but for the CDR region. The variable region of an ErbB-2 binding VH, an ErbB-3 binding VH, or a light chain in an antibody may contain one or more, typically not more than 10, such as not more than 5 amino-acid differences with the variable region of a naturally occurring human antibody, not counting possible differences in the amino acid sequence of the CDR regions. Such mutations occur also in nature in the context of somatic hypermutation. Antibodies may be derived from various animal species, at least with regard to the heavy chain variable region. It is common practice to humanize such e.g. murine heavy chain variable regions. There are various ways in which this can be achieved among which there are CDR-grafting into a human heavy chain variable region with a 3D-structure that matches the 3-D structure of the murine heavy chain variable region; deimmunization of the murine heavy chain variable region, such as done by removing known or suspected T- or B- cell epitopes from the murine heavy chain variable region. The removal is typically by substituting one or more of the amino acids in the epitope for another (typically conservative) amino acid, such that the sequence of the epitope is modified such that it is no longer a T- or B-cell epitope. Such deimmunized murine heavy chain variable regions are less immunogenic in humans than the original murine heavy chain variable region. In certain aspects, a variable region or domain is further humanized, such as for instance veneered. By using veneering techniques, exterior residues which are readily encountered by the immune system are selectively replaced with human residues to provide a hybrid molecule that comprises either a weakly immunogenic or substantially non-immunogenic veneered surface. In certain aspects, an animal as used in the invention is a mammal, such as a primate or a human. In certain aspects, a bispecific antibody disclosed herein comprises a constant region of a human antibody. According to differences in their heavy chain constant domains, antibodies are grouped into five classes, or isotypes: IgG, IgA, IgM, IgD, and IgE. These classes or isotypes comprise at least one of said heavy chains that is named with a corresponding Greek letter. In certain aspects, the constant region comprises an IgG constant region, such as an IgG1 constant region, or a mutated IgG1 constant region. Some variation in the constant region of IgG1 occurs in nature, such as for instance the allotypes G1m1, 17 and G1m3, and / or is allowed without changing the immunological properties of the resulting antibody. Typically between about 1-10 amino acid insertions, deletions, substitutions or a combination thereof are allowed in the constant region. In certain aspects, bispecific antibodies as disclosed herein comprise: - at least the CDR1, CDR2 and CDR3 sequences of an ErbB-2 specific heavy chain variable region selected from the group consisting of MF2973, MF3004, MF3958, MF2971, MF3025, MF2916, MF3991, MF3031 and MF3003 ; and / or - at least the CDR1, CDR2 and CDR3 sequences, or at least the heavy chain variable region sequence, of an ErbB-3 specific heavy chain variable region selected from the group consisting of MF3178; MF3176; MF3163; MF3099; MF3307; MF6055; MF6056; MF6057; MF6058; MF6059; MF6060; MF6061; MF6062; MF6063; MF6064; MF 6065; MF6066; MF6067; MF6068; MF6069; MF6070; MF6071; MF6072; MF6073 and MF6074. CDR sequences are for instance varied for optimization purposes, e.g. in order to improve binding efficacy or the stability of the antibody. Optimization is for instance performed by mutagenesis procedures where after the stability and / or binding affinity of the resulting antibodies are tested and an improved ErbB-2 or ErbB-3 -specific CDR sequence is selected. A skilled person is well capable of generating antibody variants comprising at least one altered CDR sequence. For instance, conservative amino acid substitution is applied. Examples of conservative amino acid substitution include the substitution of one hydrophobic residue such as isoleucine, valine, leucine or methionine for another hydrophobic residue, and the substitution of one polar residue for another polar residue, such as the substitution of arginine for lysine, glutamic acid for aspartic acid, or glutamine for asparagine. In certain aspects, antibodies comprise a variable domain that binds ErbB- 2, wherein the VH chain of said variable domain comprises the amino acid sequence of VH chain MF2973; MF3004; MF3958 (is humanized MF2971); MF2971; MF3025; MF2916; MF3991 (is humanized MF3004); MF3031 or MF3003; or comprises the amino acid sequence of VH chain MF2973; MF3004; MF3958 (is humanized MF2971); MF2971; MF3025; MF2916; MF3991 (is humanized MF3004); MF3031; or MF3003 as having at most 15, typically 1, 2, 3, 4, 5, 6, 7, 8, 9 or 10, such as at most 1, 2, 3, 4 or 5, amino acid insertions, deletions, substitutions or a combination thereof with respect to the above mentioned VH chain sequence. In certain aspects, the VH chain of the variable domain that binds ErbB-3 comprises the amino acid sequence of VH chain MF3178; MF3176; MF3163; MF6055; MF6056; MF6057; MF6058; MF6059; MF6060; MF6061; MF6062; MF6063; MF6064; MF 6065; MF6066; MF6067; MF6068; MF6069; MF6070; MF6071; MF6072; MF6073 or MF6074. In certain aspects, the VH chain of the variable domain that binds Erb-B3 comprises the amino acid sequence of MF3178, MF3176, MF3163, MF6058, MF6061 or MF6065; or in certain aspects comprises the amino acid sequence of MF3178, MF3176, MF3163, MF6058, MF6061 or MF6065 having at most 15, or typically 1, 2, 3, 4, 5, 6, 7, 8, 9 or 10, or such as in at most 1, 2, 3, 4 or 5, amino acid insertions, deletions, substitutions or a combination thereof with respect to the respective VH chain sequence. In certain aspects, the VH chain of the variable domain that binds ErbB-3 comprises the amino acid sequence of MF3178; or comprises the amino acid sequence of MF3178 having at most 15, or typically 1, 2, 3, 4, 5, 6, 7, 8, 9 or 10, or such as at most 1, 2, 3, 4 or 5, amino acid insertions, deletions, substitutions or a combination thereof with respect to the VH chain sequence. In certain aspects, the above-mentioned amino acid insertions, deletions and substitutions are not present in the CDR1, CDR2 and CDR3 regions. In certain aspects, the above-mentioned amino acid insertions, deletions and substitutions are also not present in the FR4 region. In certain aspects, the antibody comprises at least the CDR1, CDR2 and CDR3 sequences of MF2971, MF3958, MF3004 or MF3991 or at least the CDR1, CDR2 and CDR3 sequences of MF3958. In certain aspects, said antibody comprises at least the CDR1, CDR2 and CDR3 sequences of MF3178, MF3176, MF3163, MF6058, MF6061 or MF6065, or at least the CDR1, CDR2 and CDR3 sequence of MF3178. In certain aspects, the ErbB-2 specific heavy chain variable region comprises the amino acid sequence of the VH chain MF3958 having at most 15, or typically 1, 2, 3, 4, 5, 6, 7, 8, 9 or 10, or such as at most 1, 2, 3, 4 or 5, amino acid insertions, deletions, substitutions or a combination thereof with respect said VH (wherein said insertions, deletions, substitutions are not in CDR1, CDR2, or CDR3). In certain aspects, the ErbB-3 specific heavy chain variable region comprises the amino acid sequence of the VH chain MF3178 having at most 15, typically 1, 2, 3, 4, 5, 6, 7, 8, 9 or 10, or such as at most 1, 2, 3, 4 or 5, amino acid insertions, deletions, substitutions or a combination thereof with respect said VH. The one or more amino acid insertions, deletions, substitutions or a combination thereof are typically not in the CDR1, CDR2 and CDR3 region of the VH chain. They are also typically not present in the FR4 region. In certain aspects, an amino acid substitution is a conservative amino acid substitution. In certain aspects,, the ErbB-2 specific heavy chain variable region comprises the amino acid sequence of the VH chain MF3991 having at most 15, typically 1, 2, 3, 4, 5, 6, 7, 8, 9 or 10, such as at most 1, 2, 3, 4 or 5, amino acid insertions, deletions, substitutions or a combination thereof with respect said VH (in certain aspects, said insertions, deletions, substitutions are not in CDR1, CDR2, or CDR3). In certain aspects, the ErbB-3 specific heavy chain variable region comprises the amino acid sequence of the VH chain MF3178 having at most 15, typically 1, 2, 3, 4, 5, 6, 7, 8, 9 or 10, such as at most 1, 2, 3, 4 or 5, amino acid insertions, deletions, substitutions or a combination thereof with respect said VH. In certain aspects, the one or more amino acid insertions, deletions, substitutions or a combination thereof are not in the CDR1, CDR2 and CDR3 region of the VH chain. In certain aspects, they are also not present in the FR4 region. In certain aspects, an amino acid substitution is a conservative amino acid substitution. In certain aspects, the first antigen-binding site of the antibody comprises at least the CDR1, CDR2 and CDR3 sequences of MF3958, or CDR1, CDR2 and CDR3 sequences that differ in at most three, in at most two, or in at most one amino acid from the CDR1, CDR2 and CDR3 sequences of MF3958, and wherein said second antigen-binding site comprises at least the CDR1, CDR2 and CDR3 sequence of MF3178, or CDR1, CDR2 and CDR3 sequences that differ in at most three, in at most two, or in at most one amino acid from the CDR1, CDR2 and CDR3 sequences of MF3178. In certain aspects, the bispecific antibody comprises i) a first antigen binding site comprising an ErbB-2 specific heavy chain variable region comprising the CDR1, CDR2, and CDR3 sequence of MF3958 and a light chain variable region and ii) a second antigen binding site comprising an ErbB-3 specific heavy chain variable region comprising the CDR1, CDR2, and CDR3 sequence of MF3178 and a light chain variable region. In certain aspects, the ErbB-2 specific heavy chain variable region has the MF3958 sequence and the ErbB-3 specific heavy chain variable region has the MF3178 sequence. This combination is also referred to as the PB4188 antibody. In certain aspects, the PB4188 antibody is afucosylated. In certain aspects, the bispecific antibody comprises the ‘heavy chain for erbB-2 binding’ as depicted in Figure 2E and the ‘heavy chain for erbB-3 binding’ as depicted in Figure 2F. In certain aspects, the antigen binding sites of the bispecific antibody comprise a germline light chain O12, such as the rearranged germline human kappa light chain IgVκ1-39*01 / IGJκ1*01 or a fragment or a functional derivative thereof (nomenclature according to the IMGT database worldwide web at imgt.org). The terms rearranged germline human kappa light chain IgVκ1-39*01 / IGJκ1*01, IGKV1-39 / IGKJ1, huVκ1-39 light chain or in short huVκ1-39 are used. The light chain can have 1, 2, 3, 4 or 5 amino acid insertions, deletions, substitutions or a combination thereof. In certain aspects, the mentioned 1, 2, 3, 4 or 5 amino acid substitutions are conservative amino acid substitutions, and the insertions, deletions, substitutions or a combination thereof are typically not in the CDR1, CDR2 or CDR3 region or FR4 region of the VL chain. In certain aspects, the first antigen binding site and the second antigen binding site comprise the same light chain variable region, or rather, a common light chain. In certain aspects, the light chain variable region comprises a CDR1 having the sequence (QSISSY), a CDR2 having the sequence (AAS), and a CDR3 having the sequence (QQSYSTPPT). In certain aspects, the light chain variable region comprises the common light chain sequence depicted in Figure 2C. Various methods are available to produce bispecific antibodies and are discussed in WO 2015 / 130173. One method involves the expression of two different heavy chains and two different light chains in a cell and collecting antibody that is produced by the cell. Antibody produced in this way will typically contain a collection of antibodies with different combinations of heavy and light chains, some of which are the desired bispecific antibody. The bispecific antibody can subsequently be purified from the collection. For the purpose of clarity a concise description features are described herein as part of the same or separate embodiments, however, it will be appreciated that the scope of the invention may include embodiments having combinations of all or some of the features described. BRIEF DESCRIPTION OF THE DRAWINGS Figure 1 : Amino acid alignment of MF3178 variants. Dots indicate the same amino acid as in MF3178 at that position. The CDR1, CDR2 and CDR3 sequences of MF3178 are in bold and underlined. Figure 2 : Amino acid sequences of relevant antibodies and parts thereof. Figure 2A: Heavy chain variable region sequence of erbB-3 binding antibodies, including heavy chain CDR sequences annotated using KABAT. Figure 2B: Heavy chain variable region sequence of erbB-2 binding antibodies, including heavy chain CDR sequences annotated using KABAT. Figure 2C: Light chain variable region sequence of erbB2 / erbB-3 binding antibodies. Figure 2D: VK1-39 / jk5 common light chain variable domain (CDR regions according to IMGT underlined). Figure 2E: Heavy chain for erbB-2 binding. Figure 2F: Heavy chain for erbB-3 binding. Figure 2G: DNA sequences from expressed NRG1 isoforms and proteins translated therefrom. CLAUSES 1. A method for classifying a subject having an ErbB-2, ErbB-3 or ErbB- 2 / ErbB-3 positive cancer as being likely to respond to treatment with an ErbB-2 and / or ErbB-3 targeting agent, comprising determining NRG1 expression level in a cancer cell containing sample from the subject. 2. A method for identifying a subject having an ErbB-2, ErbB-3 or ErbB- 2 / ErbB-3 positive cancer for treatment with an ErbB2 and / or ErbB3 targeting agent, comprising determining NRG1 expression level in a cancer cell containing sample from the subject. 3. A method for diagnosing a subject having an ErbB-2, ErbB-3 or ErbB- 2 / ErbB-3 positive cancer and selecting for treatment with an ErbB2 and / or ErbB3 targeting agent, comprising determining NRG1 expression level in a cancer cell containing sample from the subject. 4. A method for selecting a subject having an ErbB-2, ErbB-3 or ErbB- 2 / ErbB-3 positive cancer for treatment with an ErbB-2 and / or ErbB-3 targeting agent, the method comprising determining NRG1 expression level in a cancer cell containing sample from said subject. 5. The method of any one of the preceding clauses, further comprising determining the presence or amount of phosphorylated HER3 (pHER3) in a sample from the subject, or determining FAM83E expression level and OLFML2B expression level in a sample from the subject. 6. The method of any one of the preceding clauses, wherein the subject is selected for treatment or classified as being likely to respond to treatment with an ErbB2 and / or ErbB3 targeting agent when said NRG1 expression level is determined to be elevated. 7. The method of any one of the preceding clauses, wherein the subject is selected for treatment or classified as being likely to respond to treatment with an ErbB2 and / or ErbB3 targeting agent when said NRG1 expression level is determined to be elevated and when said pHER3 status is elevated or determined to be positive; or when said NRG1 expression level is determined to be elevated, when FAM83E expression is determined to be elevated and when OLFML2B expression level is determined to be decreased. 8. The method of any one of the preceding clauses, the method comprising: - determining an expression level of NRG1 in a sample comprising cancer cells obtained from the subject; - comparing the expression level of the NRG1 in the sample with a threshold expression level of NRG1, the threshold expression level of NRG1 having been determined according to a threshold determination analysis of a reference dataset of samples comprising cancer cells; and - determining by the comparison that the expression level of the NRG1 in the sample is higher than the threshold expression level of NRG1 of the reference dataset, thereby classifying that the cancer is like to respond to treatment with the ErbB2 and / or ErbB3 targeting agent, or thereby selecting the subject for the treatment. 9. The method of any one of the preceding clauses, wherein said method comprises, - determining NRG1 expression levels from samples comprising cancer cells of a reference dataset of samples, - calculating the 95 percentile for NRG1 expression levels of the reference dataset, - comparing the NRG1 expression level determined from the sample of said subject to the 95 percentile of the reference dataset, wherein the sample has an elevated NRG1 expression level when the NRG1 expression level is equal to or higher than the 95 percentile of the reference dataset and - selecting the subject for treatment, or classifying the subject is likely to respond to treatment with an ErbB2 and / or ErbB3 targeting agent, if the NRG1 expression level in the sample obtained from the subject is equal to or higher than the 95 percentile of the reference dataset. 10. The method of clause 8 or 9, further comprising determining the presence or amount of pHER3 in a sample from the subject, and selecting the subject for treatment, or classifying the subject is likely to respond to the treatment, if additionally, the sample from the subject is determined to be positive for the presence of pHER3. 11. The method of clause 8 or 9, further comprising determining FAM83E expression level and OLFML2B expression level in a sample from the subject, and selecting the subject for treatment, or classifying the subject is likely to respond to the treatment, if additionally, the sample from the subject is determined to elevated for FAM83E expression level and is determined to decreased for OLFML2B expression level. 12. The method according to any one of the preceding clauses, wherein the NRG1 expression, FAM83E and / or OLFML2B is based on sequence reads obtained by RNA sequencing. 13. The method according to any one of the preceding clauses, wherein the 95 percentile as established for the reference dataset is based on sequence reads obtained by RNA sequencing. 14. The method according to any one of the preceding clauses, wherein expression levels are expressed as Transcripts Per Million (TPM). 15. The method according to any one of the preceding clauses, wherein the 95 percentile, as calculated for the reference dataset is based on Transcripts Per Million (TPM). 16. The method according to any one of the preceding clauses, wherein determining NRG1, FAM83E and / or OLFML2B expression levels comprises aligning sequencing reads of NRG1, FAM83E and / or OLFML2B, respectively, to the GRCh37.66 or GRCh38 / hg38 genome build as reference genome. 17. The method according to any one of the preceding clauses, wherein the reference dataset comprises at least 50 or at least 200 samples comprising cancer cells obtained from human subjects. 18. The method according to any one of the preceding clauses, wherein the reference dataset comprises a TCGA dataset, such as dataset TCGA-CHOL, TCGA- GBM, TCGA-LIHC, TCGA-LUAD, TCGA-LUSC, TCGA-PAAD, TCGA-ESCA or TCGA-STAD or at least 50 or at least 200 samples obtained therefrom. 19. The method according to any one of the preceding clauses, wherein the cancer is gastric cancer and the 95 percentile for NRG1 is 1.002 log10 TPM. 20. The method according to any one of the preceding clauses, wherein the cancer is gastric cancer and FAM83E expression is considered elevated when higher than a median log10 TPM level of a reference dataset, such as 1.325 log10 TPM as determined for TCGA-STAD. 21. The method according to any one of the preceding clauses, wherein the cancer is gastric cancer and OLFML2B expression is considered decreased when lower than a median log10 TPM level of a reference dataset, such as below 1.326 log10 TPM as determined for TCGA-STAD. 22. The method according to any one of the preceding clauses, wherein the presence or amount of pHER3 is determined by immunohistochemistry (IHC), such as by using an antibody that specifically binds HER3 phosphorylated at position Tyr1289, such as monoclonal anti-rabbit antibody clone D1B5. 23. The method according to any one of the preceding clauses, wherein the amount of pHER3 or being positive for pHER3 comprises an H-score for pHER3 of at least 20, or at least 40, or at least 120, or at least 150, as established by IHC. 24. The method according to any one of the preceding clauses, wherein the cancer is brain cancer, breast cancer, pancreatic cancer, lung cancer (including non-small cell lung cancer), esophageal cancer, liver cancer, cholangiocarcinoma or gastric cancer and / or wherein the sample is from a brain cancer, breast cancer, pancreatic cancer, lung cancer, esophageal cancer, liver cancer, cholangiocarcinoma or gastric cancer. 25. The method according to any one of the preceding clauses, wherein the sample is a formalin-fixed, paraffin-embedded human tissue of a solid tumor or a liquid biopsy sample. 26. The method according to any one of the preceding clauses, wherein NRG1 comprises the NRG1 gene identified by HGNC: 7997 or any allelic variant thereof. 27. The method according to any one of the preceding clauses, wherein NRG1 comprises any one of SEQ ID No: 129 through SEQ ID No: 183, or any allelic variant thereof. 28. The method according to any one of the preceding clauses, wherein FAM83E comprises SEQ ID NO: 195 or 197; or the gene identified by HGNC: 25972, or any allelic variant thereof. 29. The method according to any one of the preceding clauses, wherein OLFML2B comprises SEQ ID NO: 199, 201 or 203; or the gene identified by HGNC: 24558; or allelic variants thereof. 30. An ErbB2 and / or ErbB3 targeting agent for use in a method of treatment of a subject having an ErbB-2, ErbB-3 or ErbB-2 / ErbB-3 positive cancer, wherein said subject is determined to respond to treatment with an ErbB2 and / or ErbB3 targeting agent, diagnosed or selected for treatment according to any one of the preceding clauses. 31. An ErbB2 and / or ErbB3 targeting agent for use in a method of treatment of a subject having an ErbB-2, ErbB-3 or ErbB-2 / ErbB-3 positive cancer, wherein said cancer expresses NRG1 levels equal to or higher than the 95 percentile of a reference dataset of cancer cells comprising samples. 32. The ErbB2 and / or ErbB3 targeting agent of clause 31, wherein the NRG1 level, the 95 percentile and / or the reference dataset are according to any one of clauses 1-29. 33. An ErbB2 and / or ErbB3 targeting agent for use in a method of treatment of a subject having an ErbB-2, ErbB-3 or ErbB-2 / ErbB-3 positive cancer, wherein the cancer of the subject comprises a threshold expression level of NRG1 which is increased as compared to that of a reference dataset, the threshold expression level of NRG1 having been determined according to a threshold determination analysis of a reference dataset of samples comprising cancer cells. 34. An ErbB2 and / or ErbB3 targeting agent for use in a method of treatment of a subject having an ErbB-2, ErbB-3 or ErbB-2 / ErbB-3 positive cancer, wherein a cancer comprising sample obtained from the subject comprises an amount of NRG1 expression which is elevated in comparison to that of a sample free of cancer cells or said cancer has an elevated NRG1 expression level. 35. The ErbB2 and / or ErbB3 targeting agent for use according to any one of clauses 30-34, wherein a cancer comprising sample obtained from said subject is positive for pHER3. 36. The ErbB2 and / or ErbB3 targeting agent for use according to any one of clauses 30-35, wherein the cancer comprising sample comprises an amount of FAM83E expression which is elevated in comparison to a non-cancer cells comprising sample and comprises an amount of OLFML2B expression which is reduced in comparison to a sample from a sample free of cancer cells. 37. An ErbB2 and / or ErbB3 targeting agent for use in a method of treatment of a subject having an ErbB-2, ErbB-3 or ErbB-2 / ErbB-3 positive cancer, said cancer or a sample obtained from said subject, having an amount of neuregulin-1 (NRG1) expression which is equal to or exceeds the 95 percentile for NRG1 expression of a reference dataset, optionally said cancer, or a sample obtained from said subject, having an amount of phosphorylated HER3 which is elevated in comparison to that of a healthy subject; or said cancer, or a sample obtained form said subject, having an amount of FAM83E expression which is elevated in comparison to that of a healthy subject and said cancer or a sample obtained from said subject, having an amount of OLFML2B expression which is reduced in comparison to that of a sample free of cancer cells. 38. A method of treating a subject having an ErbB-2, ErbB-3 or ErbB- 2 / ErbB-3 positive cancer, the method comprising determining that the amount of neuregulin-1 (NRG1) expression in a sample obtained from the subject is elevated as compared to a reference, and administering an effective amount of an ErbB2 and / or ErbB3 targeting agent to said subject. 39. A method of treating a subject having an ErbB-2, ErbB-3 or ErbB- 2 / ErbB-3 positive cancer, the method comprising determining that the amount of neuregulin-1 (NRG1) expression in a sample obtained from the subject is elevated in comparison to a sample from a healthy subject, and administering an effective amount of an ErbB2 and / or ErbB3 targeting agent to the subject from which the sample having elevated NRG1 expression levels is obtained, optionally further determining that the amount of phosphorylated HER3 in a sample obtained from said subject is elevated in comparison to a sample from a healthy subject, or that said sample is positive for pHER3; or determining that the amount of FAM83E expression is elevated and determining that the amount of OLFML2B expression is reduced in a sample obtained from said subject in comparison to a sample from a healthy subject. 40. A method of treating a subject having an ErbB-2, ErbB-3 or ErbB- 2 / ErbB-3 positive cancer, the method comprising determining the amount of neuregulin-1 (NRG1) expression according to any one of clauses 1-23, and administering a therapeutically effective amount of an ErbB2 and / or ErbB3 targeting agent to said subject if the amount of NRG1 in the sample exceeds the 95 percentile for NRG1 expression of a reference dataset, optionally further determining that the amount of phosphorylated HER3 is elevated in comparison to a sample from a healthy subject, or determining that the amount of FAM83E expression is elevated and determining that the amount of OLFML2B expression is reduced in comparison to a sample from a healthy subject. 41. The method of any one of clauses 38-40, comprising a step of diagnosing a subject for elevated NRG1 expression level by determining NRG1 expression level in a cancer cell containing sample from the subject. 42. The ErbB2 and / or ErbB3 targeting agent for use or the method according to any one of the preceding clauses, wherein the method of treatment comprises administering zenocutuzumab in an amount of 750 mg once every two weeks. 43. The ErbB2 and / or ErbB3 targeting agent for use or the method according to any one of the preceding clauses, wherein the ErbB-2, ErbB-3 or ErbB- 2 / ErbB-3 positive cancer or sample is not NRG1 fusion positive. 44. The ErbB2 and / or ErbB3 targeting agent for use or the method according to any one of the preceding clauses, wherein the targeting agent is a bispecific antibody comprising a first antigen-binding site that can bind domain I of ERBB2 and a second antigen-binding site that can bind domain III of ERBB3. 45. The ErbB2 and / or ErbB3 targeting agent for use or the method according to any one of the preceding clauses, wherein the targeting agent is a bispecific antibody comprising i) at least the CDR1, CDR2 and CDR3 sequences of an ERBB2 specific heavy chain variable region selected from the group consisting of MF2973, MF3004, MF3958, MF2971, MF3025, MF2916, MF3991, MF3031 and MF3003 or wherein said antibody comprises CDR sequences that differ in at most 3 amino acids from the CDR1, CDR2 and CDR3 sequences of MF2973, MF3004, MF3958, MF2971, MF3025, MF2916, MF3991, MF3031 or MF3003; and / or ii) at least the CDR1, CDR2 and CDR3 sequences of an ERBB3 specific heavy chain variable region selected from the group consisting of MF3178; MF3176; MF3163; MF6055; MF6056; MF6057; MF6058; MF6059; MF6060; MF6061; MF6062; MF6063; MF6064; MF6065; MF6066; MF6067; MF6068; MF6069; MF6070; MF6071; MF6072; MF6073 and MF6074, or wherein said antibody comprises CDR sequences that differ in at most 3 amino acids from the CDR1, CDR2 and CDR3 sequences of MF3178; MF3176; MF3163; MF6055; MF6056; MF6057; MF6058; MF6059; MF6060; MF6061; MF6062; MF6063; MF6064; MF6065; MF6066; MF6067; MF6068; MF6069; MF6070; MF6071; MF6072; MF6073 or MF6074. 46. The ErbB2 and / or ErbB3 targeting agent for use or the method according to any one of the preceding clauses, wherein the targeting agent is a bispecific antibody comprising i) an ERBB2 specific heavy chain variable region sequence selected from the group consisting of the heavy chain variable region sequences of MF2973, MF3004, MF3958, MF2971, MF3025, MF2916, MF3991, MF3031 and MF3003, or wherein said antibody comprises a heavy chain variable region sequence that differs in at most 15 amino acids from the heavy chain variable region sequence of MF2973, MF3004, MF3958, MF2971, MF3025, MF2916, MF3991, MF3031 or MF3003; and / or ii) an ERBB3 specific heavy chain variable region sequence selected from the group consisting of the heavy chain variable region sequences of MF3178; MF3176; MF3163; MF6055; MF6056; MF6057; MF6058; MF6059; MF6060; MF6061; MF6062; MF6063; MF6064; MF 6065; MF6066; MF6067; MF6068; MF6069; MF6070; MF6071; MF6072; MF6073 and MF6074, or wherein said antibody comprises a heavy chain variable region sequence that differs in at most 15 amino acids from the heavy chain variable region sequence of MF3178; MF3176; MF3163; MF6055; MF6056; MF6057; MF6058; MF6059; MF6060; MF6061; MF6062; MF6063; MF6064; MF 6065; MF6066; MF6067; MF6068; MF6069; MF6070; MF6071; MF6072; MF6073 or MF6074. 47. The ErbB2 and / or ErbB3 targeting agent for use or the method according to any one of the preceding clauses, wherein the targeting agent is a bispecific antibody comprising the heavy chain variable regions sequences of MF3958 and MF3178. 48. The ErbB2 and / or ErbB3 targeting agent for use or the method according to any one of the preceding clauses, wherein the targeting agent is a bispecific antibody comprising a variable domain that comprises said first antigen binding site and a variable domain that comprises said second antigen binding site, and wherein the first and the second antigen binding sites comprise a light chain variable region which comprises a CDR1 which comprises sequence QSISSY, a CDR2 which comprises sequence AAS and a CDR3 which comprises sequence QQSYSTPPT. 49. The ErbB2 and / or ErbB3 targeting agent for use or the method according to any one of the preceding clauses, wherein the subject is a human subject. 50. The ErbB2 and / or ErbB3 targeting agent for use or method according to any one of the preceding clauses, wherein the ErbB2 / ErbB3 targeting agent comprises an anti-ErbB2 specific antibody, an anti-ErbB-3 specific antibody, an anti-ErbB-2 / anti-ErbB-3 specific antibody, a small molecule ErbB2 tyrosine kinase inhibitor or an antibody drug conjugate. 51. The ErbB2 and / or ErbB3 targeting agent for use or the method according to any one of the preceding clauses, wherein the ErbB2 / ErbB3 targeting agent comprises a monospecific, bivalent antibody, which includes trastuzumab or pertuzumab. 52. The ErbB2 and / or ErbB3 targeting agent for use or the method according to any one of the preceding clauses, wherein the ErbB2 / ErbB3 targeting agent comprises MM-121 (Merrimack Pharmaceuticals; also referred to as #Ab6), HMBD-001 (Hummingbird) or RG7116 (Lumretuzumab, Roche). 53. The ErbB-2 and / or ErbB-3 targeting agent for use or the method according to any one of the preceding clauses, wherein the ErbB2 / ErbB3 targeting agent comprises MM 111 (Merrimack Pharmaceuticals). 54. The ErbB-2 and / or ErbB-3 targeting agent for use or the method according to any one of the preceding clauses, wherein the ErbB-2 / ErbB-3 targeting agent comprises lapatinib (Tyverb / Tykerb®), neratinib, afatinib, tucatinib or AZD8931 55. The ErbB-2 and / or ErbB-3 targeting agent for use or the method according to any one of the preceding clauses, wherein the ErbB-2 / ErbB-3 targeting agent comprises trastuzumab emtansine or DS-8201. EXAMPLES As used herein ‘MFXXXX’ wherein X is independently a numeral 0-9, refers to a Fab comprising a variable domain wherein the VH has the amino acid sequence identified by the 4 digits depicted in Figure 1, Figure 2A for heavy chain sequences of ErbB3 binding Fabs or Figure 2B for heavy chain sequences of ErbB2 binding Fabs. Unless otherwise indicated the light chain variable region typically has a sequence of Figure 2C. The light chain constant region in the examples has a sequence as depicted in Figure 2C. ‘MFXXXX VH’ refers to the amino acid sequence of the VH identified by the 4 digits. The MF further comprises a constant region of a light chain and a constant region of a heavy chain that normally interacts with a constant region of a light chain. The VH / variable region of the heavy chains differs and typically also the CH3 region, wherein one of the heavy chains has a KK mutation of its CH3 domain and the other has the complementing DE mutation of its CH3 domain (see for reference PCT / NL2013 / 050294 (published as WO2013 / 157954)) and Figure 2E and Figure 2F. Bispecific antibodies in the examples have an Fc tail with a KK / DE CH3 heterodimerization domain, a CH2 domain and a CH1 domain as indicated in Figure 2E and 2F, a common light chain as indicated in Figure 2C and a VH as specified by the MF numbers. Example 1: Multispecific antibodies binding domain I of ERBB2 and domain III of ERBB3. Bispecific antibodies comprising heavy chain variable regions as mentioned in Figure 2A and 2B were obtained as described in WO 2015 / 130173. The indicated ERBB3 binding domains may also be used to provide monospecific antibodies. MF2973 MF3004 MF3958 MF2971 MF3025 MF2916 MF3991 MF3031 MF3003 MF3178 X X X X X X X X X MF3176 X X X X X X X X X MF3163 X X X X X X X X X MF6055 X X X X X X X X X MF6056 X X X X X X X X X MF6057 X X X X X X X X X MF6058 X X X X X X X X X MF6059 X X X X X X X X X MF6060 X X X X X X X X X MF6061 X X X X X X X X X MF6062 X X X X X X X X X MF6063 X X X X X X X X X MF6064 X X X X X X X X X MF6065 X X X X X X X X X MF6066 X X X X X X X X X MF6067 X X X X X X X X X MF6068 X X X X X X X X X MF6069 X X X X X X X X X MF6070 X X X X X X X X X MF6071 X X X X X X X X X MF6072 X X X X X X X X X MF6073 X X X X X X X X X MF6074 X X X X X X X X X Table 1. Any of the heavy chain variable regions binding ERBB2 can be combined with any of the heavy chain variable regions binding ERBB3 in the bispecific antibodies of the present disclosure. Variants of the variable heavy chain regions may also be combined in the bispecific antibodies of the present disclosure. Example 2: RNA sequencing from material obtained from PDX models RNA was extracted using RNeasy Mini Kit (Qiagen, #74106). The tissue from the PDX models was disrupted by mortar in liquid nitrogen and roughly 30 mg of crushed tissue was used for RNA extraction. The tissue was placed in 1.5 ml microcentrifuge tubes containing 1 stainless steel bead (5 mm mean diameter) and 600 μl Buffer RLT. The tubes were placed in the TissueLyser adapter set and operated for 5 min at 30Hz. The lysate is centrifuged for 3 min at maximum speed. The supernatant was carefully removed by pipetting and transferred to a new clean tube. One volume of 70% ethanol was added to the lysate and mixed well by pipetting. Up to 700 µl of the sample was transferred, including any precipitate, to an RNeasy Mini spin column placed in a 2 ml collection tube. It is then centrifuged for 15 s at ≥8000 x g. The flow-through was discarded.350 µl Buffer RW1 was added to the RNeasy column and centrifuged for 15 s at 8000 x g. The flow-through was discarded.10 µl DNaseI stock solution was added to 70 µl Buffer RDD, mixed gently, and centrifuged briefly. 80 µl of DNase I incubation mix was added directly to the RNeasy column membrane and placed on the bench for 15 min at 20-30°C. 350 µl Buffer RW1 was added to the RNeasy column, and centrifuged for 15 s at 8000 x g. The flow-through was discarded. 500 µl Buffer RPE was added to the spin column and centrifuged for 2 min at 8000 x g to wash the spin column membrane. The flow- through was discarded. Then 500 µl Buffer RPE was added to the spin column and centrifuged for 2 min at 8000 x g to wash the spin column membrane. The RNeasy spin column was then placed in a new 1.5 ml collection tube. 30-50 µl RNase-free water was added directly to the center of the spin column membrane and centrifuged for 1 min at full speed to elute the RNA. The integrity of the total RNA was determined by 2100 Bioanalyser (Agilent) and quantified using the NanoDrop (Thermo Scientific). Only high-quality RNA sample (OD260 / 280=1.8~2.2, OD260 / 230≥2.0, RIN≥7, >500 ng) was used to construct sequencing library. The sequencing was performed either by Illumina or MGISeq2000. Illumina The mRNA-focused sequencing libraries from total RNA. PolyA mRNA was purified from total RNA using oligo-dT-attached magnetic beads and then fragmented by fragmentation buffer. Taking these short fragments as templates, first stranded cDNA was synthesized using reverse transcriptase and random primers, followed by second stranded cDNA synthesis. Then the synthesized cDNA was subjected to end- repair, phosphorylation and 'A' base addition according to library construction protocol. Then sequencing adapters were added to both size of the cDNA fragments. After PCR amplification for cDNA fragments, the targets of 250-350 bp were clean up. After library construction, Qubit 2.0 fluorometer dsDNA HS Assay (Thermo Fisher Scientific) was used to quantify concentration of the resulting sequencing libraries, while the size distribution was analyzed using Agilent BioAnalyzer 2100 (Agilent). After library validation, Illumina cBOT cluster generation system with HiSeq PE Cluster Kits (Illumina) was used to generate clusters. Paired-end sequencing was performed using an Illumina system following Illumina-provided protocols for 2 x150 paired-end sequencing. MGIseq2000 RNA enrichment was done with rRNA Depletion Kit (MGI, #1000005953). The RNA elution product from mRNA enrichment process was further fragmented by Fragmentation Buffer and mixed with Reverse Transcription Mixture and placed into the thermocycler to run the program and synthesize the first cDNA strand. second strand was then synthesized by the Second Strand Enzyme Mix and Second Strand Buffer and another reaction program. The second strand synthesis product was cleaned up by DNA clean Beads and Magnetic Separation Rack. The purified second strand synthesis product was then subject to end repair and A-tailing based on the library construction protocol. Adapters were added to cDNA fragments. After cleanup of adapter ligation product, cDNA fragments were PCR amplified followed by the cleanup of PCR product. Qubit 2.0 fluorometer dsDNA HS Assay (Thermo Fisher Scientific) was used to quantify the concentration of the purified PCR product, while the size distribution was analyzed using Agilent BioAnalyzer 2100 (Agilent). Denaturation of the PCR product was followed by the single strand circularization, enzymatic digestion, and cleanup. Quality control on enzymatic digestion product was performed followed by paired-end sequencing using a MGI system following MGI-provided protocols for 2 x150 paired-end sequencing. RNAseq raw data analysis After Illumina or MGIse2000, the quality of RNAseq raw data was checked by FastQC software. The adapter and lowquality sequences were trimmed by Trimmomatic software. The reads were mapped to human(hg19) and mouse genome(mm10) by STAR software, the reads that prefer to map to the mouse genome were removed. The clean data after trimming and mouse read cleaning were used for following analysis. The reads were mapped to reference genes (ENSEMBL GRCh37.66, see Table 2) by Bowtie software, and the gene expression was calculated by MMSEQ software. The expression values are log_2(FPKM). The reads were mapped on the hg19 genome by STAR software, and the variants were called by the HaplotypeCaller command of GATK software, and were annotated by VEP software. The mutations with AF >0.01 in public database or AF>0.2 in Crownbio tumor models were taken as germline ones. The driver status of mutations were predicted following the method of CancerGenomeInterpreter database. The gene fusion was detected by arriba2, SOAPfuse, and Fusioncatcher software. SEQ ID NOs Source SEQ ID NOs Source129 ENST00000519301.6 131 ENST00000652698.1133 ENST00000651149.1 134 ENST00000650866.1136 ENST00000650856.1 137 ENST00000520407.5139 ENST00000614767.1 140 ENST00000523534.5141 ENST00000651335.1 142 ENST00000517967.1143 ENST00000523320.1 144 ENST00000631040.2145 ENST00000523079.5 147 ENST00000650919.1149 ENST00000356819.7 151 ENST00000651807.1152 ENST00000650967.1 154 ENST00000652588.1155 ENST00000521670.5 157 ENST00000287842.7159 ENST00000650980.1 161 ENST00000405005.7163 ENST00000651175.1 164 ENST00000650964.1165 ENST00000518206.5 166 ENST00000522569.1167 ENST00000651333.1 169 ENST00000652592.1171 ENST00000520502.7 173 ENST00000523041.2175 ENST00000650819.1 177 ENST00000651696.1178 ENST00000518084.5 179 ENST00000522402.5181 ENST00000539990.3 182 ENST00000519240.5183 ENST00000523681.1Table 2. Sequence identifiers related to NRG1 isoforms obtained from the public domain as indicated. Example 3: RNA sequencing of patient samples Cancer cell comprising samples were obtained from patients suffering from gastric cancer and processed as follows. Sections of 10-20 µm thick were cut from formalin-fixed paraffin-embedded (FFPE) patient sample blocks and placed in a 1.5 ml Safe-Lock microcentrifuge tube. Deparaffinization can be performed by using Deparaffinization Solution, heptane, or xylene. When using Deparaffinization Solution, add 320 µl Deparaffinization solution. Vortex vigorously for 10 s, and centrifuge briefly to bring the sample to the bottom of the tube. Incubate at 56°C for 3 min, then allow to cool at room temperature (15–25°C), and centrifuge at full speed for 2 min. Carefully remove the supernatant by pipetting without disturbing the pellet and remove any residual Deparaffinization Solution using a fine pipet tip. Keep the lid open, and incubate for 10 min at 37°C to dry the pellet. When using heptane, add 500 µl heptane, vortex for 10 s, and incubate for 10 min at room temperature. Then add 25 μl methanol, vortex vigorously for 10 s, and centrifuge for 2 min at 9000 x g. Carefully remove the supernatant by pipetting without disturbing the pellet and remove any residual heptane / methanol using a fine pipet tip. Add 1 ml ethanol (96–100%, Fisher Scientific, #BP2818) to the pellet, mix by vortexing, and centrifuge at full speed for 2 min. Carefully remove the supernatant by pipetting without disturbing the pellet and remove any residual ethanol using a fine pipet tip. Keep the lid open, and incubate at room temperature or at up to 37°C for 10 min or until all residual ethanol has evaporated. When using xylene, add 1 ml xylene (99-100%, Fisher Scientific, #10467270), vortex vigorously for 10 s, and centrifuge at full speed for 2 min. Carefully remove the supernatant by pipetting without disturbing the pellet. Add 1 ml ethanol (96–100%) to the pellet, mix by vortexing, and centrifuge at full speed for 2 min. The ethanol extracts residual xylene from the sample. Carefully remove the supernatant by pipetting without disturbing the pellet. Carefully remove any residual ethanol using a fine pipet tip. Keep the lid open, and incubate at room temperature or at up to 37°C. Incubate for 10 min or until all residual ethanol has evaporated. After deparaffinization, RNA was extracted from FFPE slides by using the AllPrep® DNA / RNA FFPE extraction kit (Qiagen, #80234). The pellet was resuspended by adding 150 μl Buffer PKD and flicking the tube to loosen the pellet. 10 μl proteinase K was added and mixed by vortexing, followed by incubation at 56°C for 15 min. The sample was incubated on ice for 3 min, and centrifuged for 15 min at 20,000 x g. The supernatant was carefully transferred, without disturbing the pellet, to a new 1.5 ml or 2 ml Safe-Lock microcentrifuge tube for RNA purification. In order to purify total RNA, the supernatant was incubated at 80°C for 15 min. The tube was briefly centrifuged to remove drops from the inside of the lid. 320 µl of Buffer RLT was added to adjust binding conditions, and mixed by vortexing or pipetting.720 µl or 1120 µl ethanol (96–100%) was added and mixed well by vortexing or pipetting.700 µl of the sample was transferred, including any precipitate that might have formed, to an RNeasy MinElute spin column placed in a 2 ml collection tube. The lid was closed gently, and the sample was centrifuged for 15 s at ≥8000 x g (≥10,000 rpm), followed by discarding the flow-through. The step was repeated until the entire sample had passed through the RNeasy MinElute spin column.350µl of Buffer FRN was added to the RNeasy MinElute spin column. The lid was closed gently, and the sample was centrifuged for 15 s at ≥8000 x g (≥10,000 rpm), followed by discarding the flow-through. 10 µl of DNase I stock solution was added to 70 µl Buffer RDD, mixed gently by inverting the tube, and the sample was centrifuged briefly to collect residual liquid from the sides of the tube. The DNase I incubation mix (80 μl) was added directly to the RNeasy MinElute spin column membrane, and placed on the benchtop (20–30°C) for 15 min.500 µl of Buffer FRN was added to the RNeasy MinElute spin column. The lid was closed gently, and the sample was centrifuged for 15 s at ≥8000 x g (≥10,000 rpm). The flow-through was saved. An RNeasy MinElute spin column was placed in a new 2 ml collection tube, and the flow-through was applied to the spin column. The lid was closed gently, and centrifuged for 15 s at ≥8000 x g (≥10,000 rpm). The flow-through was discarded.500 µl of Buffer RPE was added to the RNeasy MinElute spin column. The lid was closed gently, and the sample was centrifuged for 15 s at ≥8000 x g (≥10,000 rpm) to wash the spin column membrane. The flow- through was discarded. Then, 500 µl Buffer RPE was added to the RNeasy MinElute spin column. The lid was closed gently, and the sample was centrifuged for 15 s at ≥8000 x g (≥10,000 rpm) to wash the spin column membrane. The collection tube with the flow-through was discarded. The RNeasy MinElute spin column was placed in a new 2 ml collection tube. The lid was opened, and the sample was centrifuged at full speed for 5 min. The collection tube with the flow- through was discarded. The RNeasy MinElute spin column was placed in a new 1.5 ml collection tube. 14–30 μl RNase-free water was added directly to the spin column membrane. The lid was closed gently, and the sample was incubated for 1 min at room temperature, followed by centrifuging at full speed for 1 min to elute the RNA. To generate RNA sequencing libraries, extracted RNA was processed with the Illumina® TruSeq® RNA Exome workflow (Illumina, #20020189). Briefly, RNA was fragmented using divalent cations under elevated temperature. cDNA was generated from the cleaved RNA fragments using random priming during first and second strand synthesis. Sequencing adapters were ligated to the resulting double-stranded cDNA fragments. Finally, the libraries were captured with an Illumina Exome Panel (Illumina, #20020183) prior to be sequenced. Prior to the capture, the quality of each library was checked using the Agilent 4150 TapeStation, using High sensitivity D1000 Reagents (Agilent, #5067-5583) and Screentapes (Agilent, #5067-5582). Library concentrations were measured by using the Invitrogen Qubit 3.0 Fluorometer with the DNA High Sensitivity Kit (Thermofisher, #Q32854). Libraries were sequenced (75 bp paired end) using a Mid Output cartridge on an Illumina NextSeq 500. RNAseq data processing started with FASTQ files from paired-end sequencing. Read trimming was done using CutAdapt, and read quality was assessed before and after read trimming using FastQC. Reads were aligned to the genome (GRCh38 / hg38) using STAR, using the Gencode v36 gene annotation. Gene-level expression values were subsequently derived using featureCounts. Example 4: PDX models and tumor growth inhibition analysis RNA sequencing was performed on patient-derived xenograft (PDX) models at Crown Bioscience as described in Example 3. The expression levels of NRG1 were analyzed and depicted in Table 3. Phospho-HER3 / ErbB3 (Tyr1289) (D1B5) Rabbit mAb (#2842, Cell Signalling Technology) and Rabbit (DA1E) mAb IgG XP® Isotype control (#3900S, Cell Signalling Technology) were used for IHC staining. Slides were scanned on the Aperio platform (40x magnification). H scores (between 0-300) were determined. The percentage values of positively stained tumor cells per intensity group (0 – negative, 1+ – weak, 2+ – moderate, 3+ – high) were documented for membranous staining. At the end the H-score could be determined using the following formula: H-Score = (0 x % [cat. 0]) + (1 x % [cat. 1+]) + (2 x % [cat.2+]) + (3 x % [cat. 3+]). Anti-tumor efficacy of zenocutuzumab was assessed in PDX models at Crown Bioscience. The PDX models had relatively high NRG1 expression compared with other samples of the same indication, the 95 percentile of NRG1 expression within the corresponding indication was taken as cutoff. Mice were inoculated with patient-derived tumors from a wide range of indications (see Table 4). Approximately 16 recipient BALB / c nude or NOD / SCID mice (Beijing Anikeeper Biotech Co., Ltd and The Jackson Laboratory) per PDX model received subcutaneous implants consisting of either tumor pieces or cell suspensions, depending on the specific model. All animals were females that weighed 16-20 g at study initiation; estimated age was 6-8 weeks at the time of tumor inoculation. For the PDX models involving inoculation with tumor fragments, these tumors were first expanded s.c. in donor BALB / c nude or NOD / SCID mice, depending on the model. For the PDX models involving inoculation with a cell suspension, the cells were thawed before use and suspended in 50% Matrigel in PBS. Tumor growth was followed by caliper measurement until tumors reach an average size of 100-200 mm3. Animals with the closest tumor volume were then randomized and treatment started the next day (day 0), with 3 mice receiving an intraperitoneal injection of 25 mg / kg zenocutuzumab in injectable 0.9% NaCl, and 3 mice receiving 0.9% NaCl only, once weekly for 3-5 weeks. Mouse body weight and tumor volume were determined twice a week. After tumor inoculation, the animals were checked daily for morbidity and mortality. During routine monitoring, the animals were checked for any effects of tumor growth or treatment on behavior such as changes in mobility, food and water consumption, body weight gain / loss (body weights were measured twice a week after randomization), eye / hair matting and any other abnormalities. The overall condition of the animals was acceptable, with mice showing only mild clinical signs (tumor scabbing, tumor ulceration, rough hair coat, excessive grooming, or hunched posture). Depending on tumor growth, mice were sacrificed 48 h after the fourth, fifth or sixth dose, or upon reaching a humane endpoint. Tumors were processed to formalin-fixed paraffin-embedded (FFPE) blocks and stored. Tumor growth inhibition (TGI) was calculated for the PDX models as follows: TGI = (1-Ti / Vi) × 100%, where Ti = mean tumor volume of treated group on final day when all mice were measured, and Vi = mean tumor volume of vehicle group on final day when all mice were measured. TGI of >60% was used as cutoff value. To compare tumor volumes between two groups at a pre- specified day, Bartlett's test was used to check the assumption of homogeneity of variance for a pair of groups. If the p-value of Bartlett's test was ≥0.05, the Welch’s t-test was used, otherwise a Mann-Whitney U test was used to obtain nominal p- values. All statistical analyses were done in SPSS (version 18.0). All tests were two-sided unless otherwise specified, and p-values of <0.05 were regarded as statistically significant. NRG1 Membrane (log2 p-HER3 (H- TGI 95% Model Indication FPKM) score) (%) threshold BN0769 BN 3.6 48.3 98.00 3.6*PA0787 PA 5.6 151.7 85.40 1.6LU6432 LU 6.2 151.7 77.60 4.6ES3431 ES 8.0 47.7 76.70 5.5GL0720 GL 4.1 25.7 68.00 3.7*LI6662 LI 3.9 178.3 66.90 3.1Table 3. Indications: BN: Brain; PA: Pancreatic; LU; lung; ES; Esophageal; GL: Gall Bladder; LI: Liver. NRG1 expression levels determined at Log2 FPMK. pHER3 scores are in H-score as determined via IHC. TGI (%): tumor growth inhibition. Percentile: NRG1 expression value of the respective sample under investigation identified as its percentile. *small sample size for accurately calculating 95% threshold. The results of Table 3 indicate that samples with NRG1 expression levels equal to or above the 95 percentile for NRG1 expression levels positively correlate with response to zenocutuzumab, which served as a model agent targeting ErbB2 and / or ErbB3. The insight that the 95 percentile can serve as a marker that positively correlates to zenocutuzumab response was next tested in a clinical setting in patients suffering from gastric cancer. Example 5: Determining 95 percentile scores for publicly available databases RNA sequence read information for NRG1, FAM83E and OLFML2B were obtained from the TCGA-CHOL, TCGA-GBM, TCGA-LIHC, TCGA-LUAD, TCGA-LUSC, TCGA-PAAD, TCGA-PRAD, TCGA-ESCA, and TCGA-STAD datasets of the Cancer Genome Atlas Program (TCGA, Data Release 39.0 - December 04, 2023). First, the p-quantile was determined following the respective sample size N of the selected dataset by computing a real valued index h using equation: h = (N − 1)*p + 1, where N is the number of observations (i.e. the number of TPM values from the selected project shown in Table 4) and p the chosen quantile (probability, herein 0.95). The equation for calculating the quantile Qp is then Qp = x⌊h⌋ + (h − ⌊h⌋) (x⌈h⌉ − x⌊h⌋), where h is calculated as above, and x are the observations (i.e. TPM values herein), wherein ⌊h⌋ means the 'floor' of h, ⌈h⌉ means the 'ceiling' of h. Ordering of x was performed, next h was calculated [h = (N - 1)*p + 1], after which the quantile was calculated by (Qp): x⌊h⌋ + (h − ⌊h⌋) (x⌈h⌉ − x⌊h⌋). project indication log10log10 TPM median log10 TPM TPM 95 FAM83E median percentile OLFML2B NRG1 TCGA- Cholangiocarcinoma 1.069 -0.513 0.989 CHOL TCGA- Glioblastoma 0.882 -1.047 1.326 GBM multiforme TCGA- Liver Hepatocellular 1.057 -1.95 0.581 LIHC Carcinoma TCGA- Lung 0.798 1.258 1.250 LUAD adenocarcinoma TCGA- Lung Squamous Cell 1.452 0.811 1.250 LUSC Carcinoma TCGA- Pancreatic 0.4697 1.359 1.659 PAAD Adenocarcinoma TCGA- Stomach 1.002 1.325 1.326 STAD Adenocarcinoma Table 4. Values calculated from the datasets / projects and cancer indications obtained from the Cancer Genome Atlas Program (TCGA, https: / / www.cancer.gov / ccg / research / genome-sequencing / tcga). Log10 TPM for NRG1 is the 95 percentile as determined, whereas log10 TPM median is the median expression for FAM83E and OLFML2B as determined. Example 6: A phase I / II study of zenocutuzumab, a full length IgG1 bispecific antibody targeting HER2 and HER3, in patients with solid tumors Study Duration: Accrual to the dose escalation part of the study (Part 1, first patient dosed on February 3, 2015), has been completed after recruiting 28 patients. The first patient in Part 2 of the study, the dose expansion phase, was dosed on 15 January 2016 in Europe. The total duration of Part 2 is approximately 25-32 months; however, the actual duration of Part 2 will be influenced by several variables, notably overall subject recruitment rate. Number of Sites: Up to 8 sites are estimated to be involved during the study. Additional sites may be added to ensure there is an acceptable enrollment rate or to replace non-enrolling / withdrawn sites. Number of Patients: Twenty-eight (28) patients were enrolled in Part 1. At least 20 evaluable patients may be enrolled in Part 2 for the following group: advanced / metastatic gastric or gastroesophageal junction cancer; GC or GEC. Patients who do not complete at least two cycles of study treatment due to other reasons than disease progression, are not evaluable for efficacy and are replaced in the respective group. This Example describes Part 2. While the example describes the administration of zenocutuzumab , an ErbB-2, ErbB-3 binding bispecific antibody, the example is not intended to be limiting to the use of this specific embodiment and applies to other bispecific antibodies disclosed herein. Study Medication Zenocutuzumab, a bispecific humanized full length IgG1 antibody, is formulated at 20 mg / mL. Zenocutuzumab Eligible patients were enrolled and received sequential treatment cycles, with a treatment cycle being 4 weeks (28 days). All patients received zenocutuzumab with a flat dose of 750 mg IV once every two weeks (Q2W). Zenocutuzumab was administered as a 2-hour IV infusion on Day 1, then Q2W of every 28-day cycle. Study Objectives: Part 1, primary objective: Determination of the maximum tolerated dose (MTD) and / or maximum recommended dose (MRD) of zenocutuzumab. Evaluation of adverse events (AEs) and dose limiting toxicities (DLT). Secondary objectives: To characterize the safety and tolerability of zenocutuzumab. Frequency and nature of AEs / serious adverse events (SAEs). PK profile of zenocutuzumab. Assessment of PK variables, including total exposure, maximum concentration (Cmax) clearance, volume of distribution (V), volume of distribution at steady state (Vss), half-life (t1 / 2), AUC0-t (area under the concentration versus time curve from time zero to time t), AUC0-∞ (area under the concentration versus time curve), tmax (time to reach maximum concentration). Immunogenicity of zenocutuzumab. Incidence and serum titers of anti-drug antibodies against zenocutuzumab. Evaluation of anti-tumor response and clinical benefit rate (CBR). Anti-tumor activity and clinical benefit assessed by RECIST v1.1 determining objective overall response rate (ORR), duration of response (DOR), progression-free survival (PFS) and survival; CBR is defined as the proportion of patients in whom a complete response (CR) or partial response (PR) or stable disease (SD) is observed (where SD duration is a minimum of 12 weeks). Part 2. Primary objective (safety): To characterize the safety and tolerability of zenocutuzumab. Frequency and nature of AEs. Primary objective (efficacy): To explore the relationships between the anti-tumor activity of zenocutuzumab and disease-related biomarkers. Overall response rate (ORR), DOR, CBR (defined as the proportion of patients in whom a CR or PR is observed, or SD of a minimum duration of 12 weeks) per RECIST 1.1 as per local investigator’s assessment. The relationship between anti-tumor activity and biomarkers including expression of HER2, HER3, and heregulin are explored, and serum biomarkers such as CA-125 (ovarian, endometrial). Secondary objective: To characterize the PK profile of zenocutuzumab. Assessment of PK variables, including total exposure, Cmax, V, Vss, t1 / 2, AUC0-t, AUC0-∞, tmax. Immunogenicity of zenocutuzumab. Incidence and serum titers of anti-drug antibodies against zenocutuzumab. Evaluation of PFS and overall survival, duration of response. Study Design: This is a Phase I / II, open-label, multi-center, multi-national, dose escalation, single group assignment study to assess the safety, tolerability, PK, PD, immunogenicity and anti-tumor activity of zenocutuzumab. The study is designed in 2 parts. Part 1 Accrual to Part 1 of the study was completed on 24 November 2015. During Part 1 of the study nine dose levels were investigated: 40 mg, 80 mg, 160 mg in cohorts of 1 patient; and 240 mg, 360 mg, 480 mg, 600 mg, 750 mg, and 900 mg in cohorts of 3 patients. Zenocutuzumab was initially given over approximately 60 minutes on Day 1 of a 3-week treatment cycle. No dose limiting toxicities (DLTs) were experienced at any of the dose levels. Three additional patients were dosed in each of the 600 mg and 750 mg cohorts in order to have sufficient PK information. As an MTD was not reached at the dose level of 900 mg, the Data Review Committee (DRC) for CL01 decided to assign the dose level of 750 mg as recommended phase 2 dose (RP2D) of the study, based on the cumulative safety, available PK data and PK simulations. Part 2 Part 2 includes a further characterization of the safety and tolerability of the selected dose level of zenocutuzumab, as well as assessment of CBR, defined as the proportion of patients with a CR, PR or durable SD (SD for at least 12 weeks in duration), in expansion groups of selected patient populations. Part 2 has commenced in Europe. No within-patient dose escalation will be permitted in Part 2. Patient populations of interest to be assessed in Part 2 of the study are HER2 amplified in advanced / metastatic in GC or GEC. At least 20 and up to approximately 40 patients may be enrolled. Duration of Treatment Patients in both Part 1 and 2 of the study may remain on treatment until disease progression, death, unacceptable toxicity or discontinuation for any other reason. Tumor Assessment Tumor assessment was evaluated according to RECIST version 1.1 per local investigator. Imaging was obtained at Screening and at the end of every 2 cycles of treatment. Example 7 Testing a response signature in patient data Response data from patients that were administered zenocutuzumab at 750 mg Q2W from the clinical trial as described in Example 6 and corresponding RNAseq data from patient samples as described in Example 2 were analyzed. Firstly, in order to determine the predictive power of NRG1 expression and pHER3, an RNAseq based proxy for pHER3 was used. LASSO regression identified OLFML2B and FAM83E from a predetermined gene set as having optimal predictive power for baseline membrane p-HER3 as calculated by IHC. When correlating RNAseq based gene expression with IHC calculated membrane pHER3 levels in the PDX samples as described in Example 3, FAM83E showed a positive correlation (R=0.9, p=4.1e-11) with these p-HER3 levels in PDX models, with OLFML2B the same but a negative correlate (R=-0.8, p=3.1e-07). The strength of the linear relationships indicate their potential for use as a proxy for pHER3. The RNAseq expression of these two genes was thus used to estimate pHER3 in clinical patient data obtained from a sufficient number of patients treated according to the clinical trial protocol of Example 6. Patients with high FAM83E expression and low or absent OLFML2B levels were estimated to be positive for p-HER3. The complete responder (CR) suffering from gastric cancer from the dataset was identified as having elevated levels of NRG1 expression across the cohort (1.202 log10 TPM), along with high expression of the positive predictor (FAM83E, 4.696 log10 TPM) and absent expression of the negative predictor (OLFML2B, 0 log10 TPM). It is noted the log10TPM for NRG1 for the complete responder is higher than the 95 percentile as established for both the TCGA-STAD reference dataset as well as for the clinical trial data set, which is 0.96 log10 TPM. Taken together, the results obtained from the experiments done on the PDX models indicate that the correlation in expression of NRG1 and pHER3 could be applied to a human clinical trial setting as well with a similar outcome. Also, data obtained from breast cancer and non-small cell lung cancer cell lines showed that cell lines expressing elevated levels of NRG1 and which are positive for pHER, correlate well with being responsive to treatment with zenocutuzumab.

Claims

Claims 1. A method for identifying a subject having an ErbB-2, ErbB-3 or ErbB- 2 / ErbB-3 positive cancer for treatment with an ErbB2 and / or ErbB3 targeting agent, comprising determining NRG1 expression level in a cancer cell containing sample from the subject.

2. A method for selecting a subject having an ErbB-2, ErbB-3 or ErbB- 2 / ErbB-3 positive cancer for treatment with an ErbB-2 and / or ErbB-3 targeting agent, the method comprising determining NRG1 expression level in a cancer cell containing sample from said subject.

3. The method of claim 1 or 2, further comprising determining the presence or amount of phosphorylated HER3 (pHER3) in a sample from the subject, or determining FAM83E expression level and OLFML2B expression level in a sample from the subject.

4. The method of any one of the preceding claims, wherein the subject is selected for treatment or classified as being likely to respond to treatment with an ErbB2 and / or ErbB3 targeting agent when said NRG1 expression level is determined to be elevated.

5. The method of any one of the preceding claims, wherein the subject is selected for treatment or classified as being likely to respond to treatment with an ErbB2 and / or ErbB3 targeting agent when said NRG1 expression level is determined to be elevated and when said pHER3 status is elevated or determined to be positive; or when said NRG1 expression level is determined to be elevated, when FAM83E expression level is determined to be elevated and when OLFML2B expression level is determined to be decreased.

6. The method of any one of the preceding claims, the method comprising: - determining an expression level of NRG1 in a sample comprising cancer cells obtained from the subject;- comparing the expression level of the NRG1 in the sample with a threshold expression level of NRG1, the threshold expression level of NRG1 having been determined according to a threshold determination analysis of a reference dataset of samples comprising cancer cells; and - determining by the comparison that the expression level of the NRG1 in the sample is higher than the threshold expression level of NRG1 of the reference dataset, thereby classifying that the cancer is like to respond to treatment with the ErbB2 and / or ErbB3 targeting agent, or thereby selecting the subject for the treatment.

7. The method of any one of claims 1-5, wherein said method comprises, - determining NRG1 expression levels from samples comprising cancer cells of a reference dataset of samples, - calculating the 95 percentile for NRG1 expression levels of the reference dataset, - comparing the NRG1 expression level determined from the sample of said subject to the 95 percentile of the reference dataset, wherein the sample has an elevated NRG1 expression level when the NRG1 expression level is equal to or higher than the 95 percentile of the reference dataset and - selecting the subject for treatment, or classifying the subject is likely to respond to treatment with an ErbB2 and / or ErbB3 targeting agent, if the NRG1 expression level in the sample obtained from the subject is equal to or higher than the 95 percentile of the reference dataset.

8. The method of claim 6 or 7, further comprising determining the presence or amount of pHER3 in a sample from the subject, and selecting the subject for treatment, or classifying the subject is likely to respond to the treatment, if additionally, the sample from the subject is determined to be positive for the presence of pHER3.

9. The method of claim 6 or 7, further comprising determining FAM83E expression level and OLFML2B expression level in a sample from the subject, and selecting the subject for treatment, or classifying the subject is likely to respond to the treatment, if additionally, the sample from the subject is determined to beelevated for FAM83E expression level and is determined to be decreased for OLFML2B expression level.

10. The method according to any one of the preceding claims, wherein the NRG1 expression, FAM83E and / or OLFML2B is based on sequence reads obtained by RNA sequencing.

11. The method according to any one of the preceding claims, wherein the 95 percentile as established for the reference dataset is based on sequence reads obtained by RNA sequencing.

12. The method according to any one of the preceding claims, wherein expression levels are expressed as Transcripts Per Million (TPM).

13. The method according to any one of the preceding claims, wherein the 95 percentile, as calculated for the reference dataset is based on Transcripts Per Million (TPM).

14. The method according to any one of the preceding claims, wherein determining NRG1, FAM83E and / or OLFML2B expression levels comprises aligning sequencing reads of NRG1, FAM83E and / or OLFML2B, respectively, to the GRCh37.66 or GRCh38 / hg38 genome build as reference genome.

15. The method according to any one of the preceding claims, wherein the reference dataset comprises at least 50 or at least 200 samples comprising cancer cells obtained from human subjects.

16. The method according to any one of the preceding claims, wherein the reference dataset comprises a TCGA dataset, such as dataset TCGA-CHOL, TCGA- GBM, TCGA-LIHC, TCGA-LUAD, TCGA-LUSC, TCGA-PAAD, TCGA-ESCA or TCGA-STAD or at least 50 or at least 200 samples obtained therefrom.

17. The method according to any one of the preceding claims, wherein the cancer is gastric cancer and the 95 percentile for NRG1 is 1.002 log10 TPM.

18. The method according to any one of the preceding claims, wherein the cancer is gastric cancer and FAM83E expression is considered elevated when higher than a median log10 TPM level of a reference dataset, such as 1.325 log10 TPM as determined for TCGA-STAD.

19. The method according to any one of the preceding claims, wherein the cancer is gastric cancer and OLFML2B expression is considered decreased when lower than a median log10 TPM level of a reference dataset, such as below 1.326 log10 TPM as determined for TCGA-STAD.

20. The method according to any one of the preceding claims, wherein the presence or amount of pHER3 is determined by immunohistochemistry (IHC), such as by using an antibody that specifically binds HER3 phosphorylated at position Tyr1289, such as monoclonal anti-rabbit antibody clone D1B5.

21. The method according to any one of the preceding claims, wherein the amount of pHER3 or being positive for pHER3 comprises an H-score for pHER3 of at least 20, or at least 40, or at least 120, or at least 150, as established by IHC.

22. The method according to any one of the preceding claims, wherein the cancer is brain cancer, breast cancer, pancreatic cancer, lung cancer (including non-small cell lung cancer), esophageal cancer, liver cancer, cholangiocarcinoma or gastric cancer and / or wherein the sample is from a brain cancer, breast cancer, pancreatic cancer, lung cancer, esophageal cancer, liver cancer, cholangiocarcinoma or gastric cancer.

23. The method according to any one of the preceding claims, wherein the sample is a formalin-fixed, paraffin-embedded human tissue of a solid tumor or a liquid biopsy sample.

24. An ErbB2 and / or ErbB3 targeting agent for use in a method of treatment of a subject having an ErbB-2, ErbB-3 or ErbB-2 / ErbB-3 positive cancer, wherein said subject is classified as being likely to respond to treatment with an ErbB2and / or ErbB3 targeting agent according to the method of any one of the preceding claims or selected for treatment according to any one of the preceding claims.

25. An ErbB2 and / or ErbB3 targeting agent for use in a method of treatment of a subject having an ErbB-2, ErbB-3 or ErbB-2 / ErbB-3 positive cancer, wherein said cancer expresses NRG1 levels equal to or higher than the 95 percentile of a reference dataset of cancer cells comprising samples.

26. The ErbB2 and / or ErbB3 targeting agent of claim 25, wherein the NRG1 level, the 95 percentile and / or the reference dataset are according to any one of claims 1-24.

27. An ErbB2 and / or ErbB3 targeting agent for use in a method of treatment of a subject having an ErbB-2, ErbB-3 or ErbB-2 / ErbB-3 positive cancer, wherein the cancer of the subject comprises a threshold expression level of NRG1 which is increased as compared to that of a reference dataset, the threshold expression level of NRG1 having been determined according to a threshold determination analysis of a reference dataset of samples comprising cancer cells.

28. An ErbB2 and / or ErbB3 targeting agent for use in a method of treatment of a subject having an ErbB-2, ErbB-3 or ErbB-2 / ErbB-3 positive cancer, wherein a cancer comprising sample obtained from the subject comprises an amount of NRG1 expression which is elevated in comparison to that of a sample free of cancer cells or said cancer has an elevated NRG1 expression level.

29. The ErbB2 and / or ErbB3 targeting agent for use according to any one of claims 24-28, wherein a cancer comprising sample obtained from said subject is positive for pHER3.

30. The ErbB2 and / or ErbB3 targeting agent for use according to any one of claims 24-29, wherein the cancer comprising sample comprises an amount of FAM83E expression which is elevated in comparison to a non-cancer cells comprising sample and comprises an amount of OLFML2B expression which is reduced in comparison to a sample free of cancer cells.

31. An ErbB2 and / or ErbB3 targeting agent for use in a method of treatment of a subject having an ErbB-2, ErbB-3 or ErbB-2 / ErbB-3 positive cancer, said cancer or a sample obtained from said subject, having an amount of neuregulin-1 (NRG1) expression which is equal to or exceeds the 95 percentile for NRG1 expression of a reference dataset, optionally said cancer, or a sample obtained from said subject, having an amount of phosphorylated HER3 which is elevated in comparison to that of a healthy subject; or said cancer, or a sample obtained form said subject, having an amount of FAM83E expression which is elevated in comparison to that of a healthy subject and said cancer or a sample obtained from said subject, having an amount of OLFML2B expression which is reduced in comparison to that of a sample free of cancer cells.

32. A method of treating a subject having an ErbB-2, ErbB-3 or ErbB- 2 / ErbB-3 positive cancer, the method comprising determining that the amount of neuregulin-1 (NRG1) expression in a sample obtained from the subject is elevated as compared to a reference, and administering an effective amount of an ErbB2 and / or ErbB3 targeting agent to said subject.

33. A method of treating a subject having an ErbB-2, ErbB-3 or ErbB- 2 / ErbB-3 positive cancer, the method comprising determining that the amount of neuregulin-1 (NRG1) expression in a sample obtained from the subject is elevated in comparison to a sample from a healthy subject, and administering an effective amount of an ErbB2 and / or ErbB3 targeting agent to the subject from which the sample having elevated NRG1 expression levels is obtained, optionally further determining that the amount of phosphorylated HER3 in a sample obtained from said subject is elevated in comparison to a sample from a healthy subject, or that said sample is positive for pHER3; or determining that the amount of FAM83E expression is elevated and determining that the amount of OLFML2B expression is reduced in a sample obtained from said subject in comparison to a sample from a healthy subject.

34. A method of treating a subject having an ErbB-2, ErbB-3 or ErbB- 2 / ErbB-3 positive cancer, the method comprising determining the amount of neuregulin-1 (NRG1) expression according to any one of claims 1-23, andadministering a therapeutically effective amount of an ErbB2 and / or ErbB3 targeting agent to said subject if the amount of NRG1 in the sample exceeds the 95 percentile for NRG1 expression of a reference dataset, optionally further determining that the amount of phosphorylated HER3 is elevated in comparison to a sample from a healthy subject, or determining that the amount of FAM83E expression is elevated and determining that the amount of OLFML2B expression is reduced in comparison to a sample from a healthy subject.

35. The ErbB2 and / or ErbB3 targeting agent for use or the method according to any one of the preceding claims, wherein the method of treatment comprises administering the bispecific antibody in an amount of 750 mg once every two weeks.

36. The ErbB2 and / or ErbB3 targeting agent for use or the method according to any one of the preceding claims, wherein the ErbB -2, ErbB-3 or ErbB-2 / ErbB-3 positive cancer or sample is not NRG1 fusion positive.

37. The ErbB2 and / or ErbB3 targeting agent for use or the method according to any one of the preceding claims, wherein the targeting agent is a bispecific antibody comprising a first antigen-binding site that can bind domain I of ERBB2 and a second antigen-binding site that can bind domain III of ERBB3.

38. The ErbB2 and / or ErbB3 targeting agent for use or the method according to any one of the preceding claims, wherein the targeting agent is a bispecific antibody comprising i) at least the CDR1, CDR2 and CDR3 sequences of an ERBB2 specific heavy chain variable region selected from the group consisting of MF2973, MF3004, MF3958, MF2971, MF3025, MF2916, MF3991, MF3031 and MF3003 or wherein said antibody comprises CDR sequences that differ in at most 3 amino acids from the CDR1, CDR2 and CDR3 sequences of MF2973, MF3004, MF3958, MF2971, MF3025, MF2916, MF3991, MF3031 or MF3003; and / or ii) at least the CDR1, CDR2 and CDR3 sequences of an ERBB3 specific heavy chain variable region selected from the group consisting of MF3178; MF3176; MF3163; MF6055; MF6056; MF6057; MF6058; MF6059; MF6060; MF6061; MF6062;MF6063; MF6064; MF6065; MF6066; MF6067; MF6068; MF6069; MF6070; MF6071; MF6072; MF6073 and MF6074, or wherein said antibody comprises CDR sequences that differ in at most 3 amino acids from the CDR1, CDR2 and CDR3 sequences of MF3178; MF3176; MF3163; MF6055; MF6056; MF6057; MF6058; MF6059; MF6060; MF6061; MF6062; MF6063; MF6064; MF6065; MF6066; MF6067; MF6068; MF6069; MF6070; MF6071; MF6072; MF6073 or MF6074.

39. The ErbB2 and / or ErbB3 targeting agent for use or the method according to any one of the preceding claims, wherein the targeting agent is a bispecific antibody comprising i) an ERBB2 specific heavy chain variable region sequence selected from the group consisting of the heavy chain variable region sequences of MF2973, MF3004, MF3958, MF2971, MF3025, MF2916, MF3991, MF3031 and MF3003, or wherein said antibody comprises a heavy chain variable region sequence that differs in at most 15 amino acids from the heavy chain variable region sequence of MF2973, MF3004, MF3958, MF2971, MF3025, MF2916, MF3991, MF3031 or MF3003; and / or ii) an ERBB3 specific heavy chain variable region sequence selected from the group consisting of the heavy chain variable region sequences of MF3178; MF3176; MF3163; MF6055; MF6056; MF6057; MF6058; MF6059; MF6060; MF6061; MF6062; MF6063; MF6064; MF 6065; MF6066; MF6067; MF6068; MF6069; MF6070; MF6071; MF6072; MF6073 and MF6074, or wherein said antibody comprises a heavy chain variable region sequence that differs in at most 15 amino acids from the heavy chain variable region sequence of MF3178; MF3176; MF3163; MF6055; MF6056; MF6057; MF6058; MF6059; MF6060; MF6061; MF6062; MF6063; MF6064; MF 6065; MF6066; MF6067; MF6068; MF6069; MF6070; MF6071; MF6072; MF6073 or MF6074.

40. The ErbB2 and / or ErbB3 targeting agent for use or the method according to any one of the preceding claims, wherein the targeting agent is a bispecific antibody comprising the heavy chain variable regions sequences of MF3958 and MF3178.

41. The ErbB2 and / or ErbB3 targeting agent for use or the method according to any one of the preceding claims, wherein the targeting agent is a bispecific antibody comprising a variable domain that comprises said first antigen binding site and a variable domain that comprises said second antigen binding site, and wherein the first and the second antigen binding sites comprise a light chain variable region which comprises a CDR1 which comprises sequence QSISSY, a CDR2 which comprises sequence AAS and a CDR3 which comprises sequence QQSYSTPPT.

42. The ErbB2 and / or ErbB3 targeting agent for use or the method according to any one of the preceding claims, wherein the subject is a human subject.

43. The ErbB2 and / or ErbB3 targeting agent for use or method according to any one of the preceding claims, wherein the ErbB2 / ErbB3 targeting agent comprises an anti-ErbB2 specific antibody, an anti-ErbB-3 specific antibody, an anti-ErbB-2 / anti-ErbB-3 specific antibody, a small molecule ErbB2 tyrosine kinase inhibitor or an antibody drug conjugate.

44. The ErbB2 and / or ErbB3 targeting agent for use or the method according to any one of the preceding claims, wherein the ErbB2 / ErbB3 targeting agent comprises a monospecific, bivalent antibody, which includes trastuzumab or pertuzumab.

45. The ErbB2 and / or ErbB3 targeting agent for use or the method according to any one of the preceding claims, wherein the ErbB2 / ErbB3 targeting agent comprises MM-121 (Merrimack Pharmaceuticals; also referred to as #Ab6), HMBD- 001 (Hummingbird) or RG7116 (Lumretuzumab, Roche).

46. The ErbB2 and / or ErbB3 targeting agent for use or the method according to any one of the preceding claims, wherein the ErbB2 / ErbB3 targeting agent comprises MM-111 (Merrimack Pharmaceuticals).

47. The ErbB2 and / or ErbB3 targeting agent for use or the method according to any one of the preceding claims, wherein the ErbB2 / ErbB3 targeting agent comprises lapatinib (Tyverb / Tykerb®), neratinib, afatinib, tucatinib or AZD8931.

48. The ErbB2 and / or ErbB3 targeting agent for use or the method according to any one of the preceding claims, wherein the ErbB2 / ErbB3 targeting agent comprises trastuzumab emtansine or DS-8201.