Gene signatures for predicting melanoma metastasis and patient prognosis

Abstract

The present invention relates to a gene signature for predicting melanoma metastasis and patient prognosis. [Solution] The present disclosure provides gene signatures for classifying individuals with cutaneous melanoma. The "SLN gene signatures" provided herein classify individuals based on prognosis and / or as having metastasis-positive or -negative sentinel lymph nodes (SLNs). The "N-SLN gene signatures" provided herein classify individuals as having metastasis-positive or -negative non-sentinel lymph nodes (N-SLNs). [Selection diagram] None

Description

[Technical field]

[0001] FIELD OF THEINVENTION The present disclosure provides gene signatures that classify individuals suffering from cutaneous melanoma. The "SLN gene signatures" provided herein classify individuals based on prognosis and / or as having metastasis-positive or -negative sentinel lymph nodes (SLNs). The "N-SLN gene signatures" provided herein classify individuals as having metastasis-positive or -negative non-sentinel lymph nodes (N-SLNs). [Background technology]

[0002] 2. Background of the Invention Cutaneous melanoma is a malignant tumor arising from pigment-producing melanocyte lineage cells located primarily in the basal layer of the epidermis of the skin. Most lesions present with signs described by the ABCDE rule: asymmetry, irregular borders, uneven pigmentation, diameter greater than 6 mm, and evolution (i.e. tendency to change rapidly) (Abbasi NR, Shaw HM, Rigel DS, et al. Early Diagnosis of Cutaneous Melanoma Revisiting the ABCD Criteria. JAMA. 2004;292(22):2771-2776). Sites are typically asymptomatic but may cause itching and / or bleeding (especially in later stages). Detection of suspicious lesions is usually performed by skin self-examination, which is recommended to be performed routinely based on the ABCDE criteria or the "ugly duckling sign" (Grob J. The'Ugly Duckling'Sign: Identification of the Common Characteristics of Nevi in ​​an Individual as a Basis for Melanoma Screening. Arch Dermatol. 1998;134:103-104). Since there are several known clinical and pathological variants, when melanoma is subsequently formally diagnosed by a physician, it is important to determine the specific subtype. The most common form is cutaneous melanoma (superficial spreading melanoma, which accounts for about 70% of cases), which is particularly frequent in fair-skinned people. The severity of the condition depends largely on the ability of melanoma cells to migrate from the primary area. For this reason, it is a high priority to evaluate whether the tumor is localized or has spread to lymph nodes or organs.

[0003] Staging of melanoma is crucial for determining the patient's prognosis and further monitoring and treatment strategies. This is reflected in the striking contrasts in 5-year overall survival numbers reported in a systematic literature review from nine European countries: 95%-100% (stage I), 65%-92.8% (stage II), 41%-71% (stage III) and 9%-28% (stage IV). These differences depend largely on the metastatic potential of melanoma versus localized melanoma lesions. Accurate discrimination between different stage classes is important and is most commonly based on the TNM system, which refers to the thickness of the primary tumor (T), the presence of tumor cells and / or spread to lymph nodes (N) and the presence of distant metastases to other organs (M). When assessing the extent of the primary tumor stage, physicians take into account not only the thickness of the tumor, but also additional characteristics such as the presence of ulcers and the mitotic rate of the primary tumor cells. Usually, lymph node and metastatic spread will be evaluated only in patients with high tumor stage, based on thick melanoma and / or other variables such as ulceration.As understood in the art, prognosis refers to prediction of a patient's medical outcome.For example, an individual can be classified as having a poor prognosis or a good prognosis.The prognosis of a patient suffering from melanoma indicates, for example, long-term survival, overall survival, progression-free survival, recurrence vs. disease remission prediction, and the likelihood of disease progression.

[0004] Currently, a widely applied procedure for accurate stratification of patients and prediction of prognosis is the determination of the presence of metastases in SLNs by the SLNB (sentinel lymph node biopsy) technique. The management of patients with cutaneous melanoma has progressed significantly since the inclusion of lymphatic mapping by SLNB procedure in the early 1990s (Morton DL, Wen DR, Wong JH, et al. Technical details of intraoperative lymphatic mapping for early stage melanoma. Arch Surg. 1992;127(4):392-399). Surgical techniques have been improved by the realization of a dual-modality intraoperative approach using blue dye and radiotracer with gamma probe detection. In addition, pathological evaluation has been upgraded by the adoption of serial sectioning of SLNs and immunohistochemistry. This has resulted in a better identification of the first regional lymph node or group of nodes (i.e., SLNs) located in close proximity to the tumor and therefore the likely site of metastatic disease. This procedure is also called "sentinel lymph node mapping".

[0005] The significant impact of SLN positivity was also shown in the MSLT-1 study, suggesting a difference in 5-year survival rates of 72.3% and 90.2% for patients with tumor-positive and tumor-negative SLNs, respectively (Morton DL, Thompson JF, Cochran AJ, et al. Sentinel-Node Biopsy or Nodal Observation in Melanoma. new engl J Med. 2006;355(13):1307-1317). According to the American Joint Committee on Cancer (AJCC) Melanoma Guidelines 8th Edition, the SLNB procedure is recommended for patients with cutaneous melanomas ≥ 0.8 mm (Gershenwald JE, Scolyer RA, Hess KR, et al. Melanoma Staging: Evidence-Based Changes in the American Joint Committee on Cancer Eighth Edition Cancer Staging Manual. CA Cancer J Clin. 2017;67(6):472-492). In this patient group, SLNB procedures are generally performed, and furthermore, treatment depends on the level of metastasis. Within the borderline group, SLNB may be considered, especially if the melanoma appears with additional adverse prognostic parameters. In patients with a melanoma thickness of <0.8 mm, standard treatment is generally considered sufficient, and SLNB is not recommended. Standard treatment includes local excision of the primary melanoma with wide margins, i.e., surgical removal of the tumor. As used herein, "resection" is understood to mean the surgical removal of melanoma-specific malignant tissue from a human patient. According to one embodiment, resection is understood to mean the removal of malignant tissue such that the presence of residual malignant tissue in said patient is undetectable by available methods.

[0006] The percentage of SLNBs classified as positive is quite variable and depends to a large extent on the known prognostic factors of the primary tumor. In patients with clinical stage I or II, the percentage of SLN metastasis is shown to be 15-30%, whereas in thin melanomas it is 5.2%. The current version of the Melanoma Expert Panel Staging Guidelines noted the clinical significance of subcategorizing T1 melanomas at 0.8 mm. This is based on the trend detected in some survival studies of T1 melanomas that there is a potential clinical intercept in the 0.7-0.8 mm region. However, based on long-term follow-up of patients after SLNB, it was shown that local nodal recurrence occurs in patients with sentinel nodes that were initially tumor-free. This information allowed the calculation of SLBN test performance, which was established to have an overall false-negative rate of 12.5%. More recently, Morton et al. reported that in intermediate-thickness melanomas with 16.0% positive SLNBs, 4.8% had a false-negative test result based on recurrence within a 10-year follow-up period. With SLNB being positive in 32.9% of thick melanomas, the false-negative rate was 10.3%.

[0007] SLNB is not only a method to potentially determine the stage of cutaneous melanoma, but also a part of treatment that may or may not be necessary depending on the metastatic classification of the SLN. The SLNB procedure can cause complications for the patient and is costly. Therefore, the procedure is performed only in a select group of patients considered to be at higher risk of metastatic spread (from the majority of low-risk lesions). Metastatic risk can be assessed by evaluation of clinicopathological factors including tumor invasion depth, known as Breslow depth, and tumor surface ulceration. Tumors growing deeper vertically into the skin and melanoma ulceration are associated with a higher risk of adverse outcomes. For example, SLN biopsy is not recommended for T1a thin melanomas, "may be recommended" for patients with T1b thin melanomas, recommended for patients with T2 and T3 intermediate thickness melanomas, and "may be recommended" for patients with T4 thick melanomas. Summary of the Invention [Problem to be solved by the invention]

[0008] Although this concept using clinicopathological variables generally allows the identification of high-risk patients at the extreme ends of the tumor spectrum, the method is not accurate for intermediate lesions. Furthermore, exceptions exist within the high or low risk groups. For example, 5% of "thin" melanomas (<0.8 mm depth of invasion) are known to metastasize locally, despite the fact that they are usually classified as having a low-risk profile based on standard clinicopathological variables. To better distinguish higher-risk lesions from the 95% of biologically indolent lesions, the mitotic rate of the tumor (mitoses / mm 2 Additional histological variables such as stenosis, stenosis, and molecular methods such as fluorescence in situ hybridization (FISH) have been introduced. Unfortunately, these techniques have been only partially successful, resulting in the fact that up to 95% of SLNBs are negative. As a result, overtreatment is seen with current diagnostic criteria, with the majority of patients undergoing unnecessary SLNB procedures and developing associated side effects that could have been avoided. Conversely, some patients who are currently deemed not to be SLNB candidates based on clinicopathological factors may nevertheless present with SLN metastasis or distant metastasis at a later stage. Thus, although the SLNB procedure accurately identifies SLN-positive nodes in the resected tissue, the selection of eligible patients for this SLNB procedure remains a challenge. One objective of the present invention is to stratify individuals with regard to the risk of having metastatic positive SLNs and therefore the need for a SLNB procedure. A further objective of the present invention is to predict the prognosis of individuals suffering from primary cutaneous melanoma. This information is useful, for example, in determining optimal treatment strategies. [Means for solving the problem]

[0009] Summary of the Invention The present invention provides a method of classifying an individual suffering from primary cutaneous melanoma, comprising determining a gene expression signature in a sample from said individual, the gene expression signature comprising three or more of the following genes: ITGB3, PLAT, SPP1, GDF15 and IL8. Preferably, the gene expression signature comprises three or more of the following genes: ITGB3, PLAT, GDF15 and IL8, more preferably the gene expression signature comprises ITGB3, PLAT, GDF15 and IL8. Preferably, the gene expression signature comprises three or more of the following genes: GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, SERPINE2 and TGFBR1, more preferably, the gene expression signature comprises three or more of the following genes: GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, ADIPOQ, SERPINE2 and TGFBR1, more preferably, the gene expression signature comprises three or more of the following genes: GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, ADIPOQ, PRKCB, SERPINE2, ADAM12, LGALS1 and TGFBR1. Also preferred is a gene signature comprising three or more of the following genes: GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, PRKCB, SERPINE2, ADAM12, LGALS1 and TGFBR1.

[0010] Also provided is a method of determining a therapeutic and / or diagnostic work-up schedule for an individual suffering from cutaneous melanoma, comprising determining in a sample from said individual the expression levels of three or more of the following genes: ITGB3, PLAT, SPP1, GDF15 and IL8, and determining the therapeutic and / or diagnostic work-up schedule based on the expression levels. Preferably, the gene expression signature comprises three or more of the following genes: ITGB3, PLAT, GDF15 and IL8, more preferably the gene expression signature comprises ITGB3, PLAT, GDF15 and IL8. Preferably, the gene expression signature comprises three or more of the following genes: GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, SERPINE2 and TGFBR1, more preferably, the gene expression signature comprises three or more of the following genes: GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, ADIPOQ, SERPINE2 and TGFBR1, more preferably, the gene expression signature comprises three or more of the following genes: GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, ADIPOQ, PRKCB, SERPINE2, ADAM12, LGALS1 and TGFBR1. Also preferred is a gene signature comprising three or more of the following genes: GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, PRKCB, SERPINE2, ADAM12, LGALS1 and TGFBR1.

[0011] Also provided is a method of predicting prognosis of an individual suffering from primary cutaneous melanoma, comprising determining a gene expression signature in a sample from the individual, wherein the gene expression signature comprises three or more of the following genes: ITGB3, PLAT, SPP1, GDF15 and IL8. Preferably, the gene expression signature comprises three or more of the following genes: GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, SERPINE2 and TGFBR1, more preferably, the gene expression signature comprises three or more of the following genes: GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, ADIPOQ, SERPINE2 and TGFBR1, more preferably, the gene expression signature comprises three or more of the following genes: GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, ADIPOQ, PRKCB, SERPINE2, ADAM12, LGALS1 and TGFBR1. Also preferred is a gene signature comprising three or more of the following genes: GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, PRKCB, SERPINE2, ADAM12, LGALS1 and TGFBR1.

[0012] In one embodiment, the individual is classified as having metastasis-positive SLN or as having metastasis-negative SLN. In one embodiment, the prognosis of the individual is determined based on gene expression level. Preferably, the individual is classified as having poor prognosis or good prognosis. The individual can be selected for SLNB based on the classification and / or expression level. The individual classified as having metastasis-positive SLN or otherwise having poor prognosis is treated by performing SLNB and / or adjuvant therapy.

[0013] The invention also provides a method of classifying an individual suffering from primary cutaneous melanoma, comprising determining a gene expression signature in a sample from said individual, wherein the gene expression signature comprises at least one of the following genes: KRT14, SPP1, FN1 and LOXL3.

[0014] Also provided is a method of treating an individual suffering from primary cutaneous melanoma, comprising: - determining a gene expression signature in a sample from the individual, the gene expression signature comprising three or more of the following genes: ITGB3, PLAT, SPP1, GDF15 and IL8; - classifying the individual as having a metastatic positive SLN and / or a poor prognosis based on the gene expression signature; - treating said individual by performing SLNB and / or providing cancer treatment to said individual. The method includes:

[0015] Preferably, the gene expression signature comprises three or more of the following genes: ITGB3, PLAT, GDF15 and IL8, more preferably, the gene expression signature comprises ITGB3, PLAT, GDF15 and IL8. Preferably, the gene expression signature comprises three or more of the following genes: GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, SERPINE2 and TGFBR1, more preferably, the gene expression signature comprises three or more of the following genes: GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, ADIPOQ, SERPINE2 and TGFBR1, more preferably, the gene expression signature comprises three or more of the following genes: GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, ADIPOQ, PRKCB, SERPINE2, ADAM12, LGALS1 and TGFBR1. Also preferred is a gene signature that includes three or more of the following genes: GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, PRKCB, SERPINE2, ADAM12, LGALS1, and TGFBR1.

[0016] The present invention provides a method of treating an individual suffering from primary cutaneous melanoma, comprising: - determining a gene expression signature in a sample from the individual, the gene expression signature comprising at least one of the following genes: KRT14, SPP1, FN1 and LOXL3; - classifying said individual as at high risk of having a metastatic positive N-SLN based on the gene expression signature; - treating said individual by performing a complete lymphadenectomy and / or providing said individual with cancer treatment. Also provided is a method comprising:

[0017] 1. A method for analyzing a gene signature in an individual suffering from primary cutaneous melanoma, comprising: - extracting RNA from a primary cutaneous melanoma lesion from said individual; - reverse transcribing RNA transcripts of at least three of the following genes: ITGB3, PLAT, SPP1, GDF15, and IL8, to generate cDNA of the RNA transcripts; - amplifying the cDNA to generate an amplicon from the cDNA for determining the expression level of the RNA transcript; A method is also provided that includes:

[0018] The present invention provides a method for analyzing a gene signature in an individual suffering from primary cutaneous melanoma, comprising: - extracting RNA from a primary cutaneous melanoma lesion from said individual; - reverse transcribing an RNA transcript of at least one of the following genes: KRT14, SPP1, FN1, and LOXL3 to generate a cDNA of the RNA transcript; - amplifying the cDNA to generate an amplicon from the cDNA for determining the expression level of the RNA transcript; Also provided is a method comprising:

[0019] Further provided is a kit for use in classifying an individual as having primary cutaneous melanoma, comprising: a) three or more of the following genes: ITGB3, PLAT, SPP1, GDF15, and IL8; and / or b) at least one of the following genes: KRT14, SPP1, FN1, and LOXL3, and optionally, c) at least one reference gene A kit comprising a primer pair for amplifying

[0020] Preferably, the kit comprises primer pairs for amplifying three or more of the following genes: ITGB3, PLAT, GDF15 and IL8, more preferably, the kit comprises primer pairs for amplifying ITGB3, PLAT, GDF15 and IL8. Preferably, the gene expression signature comprises three or more of the following genes: GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, SERPINE2 and TGFBR1, more preferably, the gene expression signature comprises three or more of the following genes: GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, ADIPOQ, SERPINE2 and TGFBR1, more preferably, the gene expression signature comprises three or more of the following genes: GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, ADIPOQ, PRKCB, SERPINE2, ADAM12, LGALS1 and TGFBR1. Also preferred is a gene signature that includes three or more of the following genes: GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, PRKCB, SERPINE2, ADAM12, LGALS1, and TGFBR1. [Brief description of the drawings]

[0021] [Figure 1]Average ROC curves of a logistic regression classifier trained on DLCV for 1) ITGB3, PLAT, SPP1, GDF15 and IL8 gene signature (molecular model), 2) clinicopathological variables (age and Breslow depth), and 3) combinations of ITGB3, PLAT, SPP1, GDF15 and IL8 gene signature with clinicopathological variables. The x-axis represents the false positive discovery rate (i.e., 1-specificity) and the y-axis represents the true discovery rate (i.e., sensitivity). [Diagram 2] Figure 1 shows the ROC curve for the ITLP score and SLN gene signature (referred to as the "logistic regression model") in the total 770 patient cohort. The x-axis represents the false positive discovery rate (i.e., 1-specificity) and the y-axis represents the true discovery rate (i.e., sensitivity). [Diagram 3] Box plots of the area under the ROC curve for the different subsets of ITGB3, PLAT, SPP1, GDF15 and IL8 genes, the full set of five ITGB3, PLAT, SPP1, GDF15 and IL8 genes and the ITLP signature. [Figure 4] Average ROC curves of logistic regression classifiers trained on DLCV for 1) gene expression, 2) clinicopathological variables, and 3) the combination of gene expression and clinicopathological variables. The x-axis represents the false positive discovery rate (i.e., 1-specificity) and the y-axis represents the true discovery rate (i.e., sensitivity). [Diagram 5] Box plots of the area under the ROC curve for the different gene subsets and the full set of four genes. [Figure 6] Overall performance comparison: CL vs. GE vs. GECL. ROC curves of logistic regression classifiers trained on DLCV for 1) gene expression, 2) clinicopathological variables, and 3) the combination of gene expression and clinicopathological variables. [Figure 7] NPV vs. SLNBRR. Negative predictive value (NPV) vs. sentinel lymph node reduction rate (SLNB RR) of a logistic regression classifier trained on DLCV for: 1) gene expression, 2) clinicopathological variables, and 3) the combination of gene expression and clinicopathological variables. [Figure 8] Gene subset-AUC boxplots. Boxplots of the area under the curve (AUC) of the ROC curves of logistic regression classifiers using subsets of 2, 3, 4, 5, 6, 7, and 8 genes trained on the full cohort. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0022] DETAILED DESCRIPTION OF THE DISCLOSED EMBODIMENTS The present disclosure provides, in part, methods, kits, genetic signatures, and means for detecting such genetic signatures for analysis of primary cutaneous melanoma tumor tissue samples. In one aspect, the present disclosure provides a "SLN genetic signature". The SLN genetic signature classifies individuals suffering from primary cutaneous melanoma, and in particular, the genetic signature classifies the risk of individuals having metastatic positive SLN and / or poor prognosis. This risk assessment is useful to physicians and patients when deciding whether a SLNB procedure and / or alternative treatment strategies are required. This assessment is also useful when selecting patients to be included in clinical trials.

[0023] As used herein, a SLN is the first lymph node (or first set of lymph nodes) to receive lymphatic drainage from a tumor and is the first lymph node (or first set of lymph nodes) to which cancer is likely to spread. An N-SLN is a lymph node that is not the first lymph node to receive lymphatic drainage from a tumor. Such N-SLNs are often nodes that are in the same nodal region or in close proximity to an SLN.

[0024] In some embodiments, the gene signature classifies the risk of metastasis-positive SLN. In some embodiments, the methods disclosed herein classify an individual as having metastasis-positive SLN or having metastasis-negative SLN. In some embodiments, the gene signature classifies the prognosis of an individual. As used herein, prognosis refers to the prediction of medical outcome and can be based on measures such as overall survival, melanoma-specific survival, recurrence-free survival, relapse-free survival, distant relapse-free survival, etc.

[0025] One advantage of SLN gene signature is that it can reduce the number of surgical procedures for patients classified as metastasis-negative SLN (and / or classified as having good prognosis). In particular, patients with intermediate lesions will likely have undergone SLNB procedures, even though the SLN may well be metastasis-negative in nature. Accurate classification of such patients using SLN gene signatures can avoid the need for SLNB procedures and replace SLNB as the current standard of care for intermediate lesions. Reduction of unnecessary SLNB reduces overall health care costs and reduces the number of patients who suffer complications due to removal of SLN. In addition, classification of individuals as SLN metastasis-positive or negative can also provide prognostic information and can be used to determine treatment or diagnostic work-up schedules.

[0026] Surprisingly, SLN gene signature can predict prognosis more accurately than standard care SLN biopsy (see Example 7). Without wishing to be bound by theory, one possible explanation for the improved prognostic ability of gene signature over SLN biopsy is related to the technical limitations of performing such biopsy (e.g., identifying the correct lymph node for biopsy, limitations of tumor cell detection, human error in sample processing / classification). Additionally or alternatively, the gene signature of the present disclosure may be able to predict SLN metastasis at a stage before it is detectable by biopsy (e.g., tumor has metastasized and tumor cells are on the way to SLN). In this regard, SLN signature can be used to replace SLNB as a criterion for inclusion in clinical trials and / or additional treatments.

[0027] An additional advantage of the SLN gene signature is that it identifies patients with thin melanomas who may not be eligible for SLNB based on clinical parameters under the current standard of care, but who are at high risk for metastasis-positive SLNs based on the gene signature. In particular, the gene signature will greatly increase the identification of metastasis-positive SLNs in patients with thin (<0.8 mm) melanomas who are currently not eligible for SLNB procedures according to guidelines. Early detection and treatment of such patients will increase progression-free and overall survival for this patient subpopulation.

[0028] The examples disclosed herein demonstrate that a gene expression signature (i.e., a SLN gene signature) comprising one or more of the following genes: ITGB3, PLAT, SPP1, GDF15 and IL8 is useful for classifying individuals and predicting prognosis, particularly for classifying metastasis-positive or -negative SLNs. Thus, in one aspect, the present disclosure provides a gene signature comprising one or more, preferably two or more, more preferably three or more of the following genes: ITGB3, PLAT, SPP1, GDF15 and IL8. Suitable gene signatures include the following combinations: ITGB3 and PLAT, ITGB3 and SPP1, ITGB3 and GDF15, ITGB3 and IL8, PLAT and SPP1, PLAT and GDF15, PLAT and IL8, SPP1 and GDF15, SPP1 and IL8, GDF15 and IL8. In some embodiments, the gene signature comprises one or more of ITGB3, PLAT, and SPP1, GDF15, and IL8. In some embodiments, the SLN gene signature comprises three or more of the following genes: ITGB3, PLAT, SPP1, GDF15, and IL8. In some embodiments, the SLN gene signature comprises four or more of the following genes: ITGB3, PLAT, SPP1, GDF15, and IL8. In some embodiments, the SLN gene signature comprises all of the following genes: ITGB3, PLAT, SPP1, GDF15, and IL8. In some embodiments, the gene expression signature comprises three or more of the following genes: ITGB3, PLAT, GDF15, and IL8, more preferably, the gene expression signature comprises ITGB3, PLAT, GDF15, and IL8. Preferably, the gene expression signature comprises GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, SERPINE2 and TGFBR1, more preferably, the gene expression signature comprises GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, ADIPOQ, SERPINE2 and TGFBR1, more preferably, the gene expression signature comprises GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, ADIPOQ, PRKCB, SERPINE2, ADAM12, LGALS1 and TGFBR1.

[0029] In some embodiments, the gene signature comprises at least three, at least four, or at least five of the following: ITGB3, PLAT, GDF15, SPP1, and IL8. Preferably, the gene signature comprises ITGB3, PLAT, GDF15, and IL8.

[0030] In some embodiments, the gene signature comprises at least three, at least four, at least five, at least six, at least seven, at least eight, or all of the following: GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, ADIPOQ, SERPINE2, and TGFBR1. The inventors also demonstrate that a gene signature lacking ADIPOQ performs similarly. Thus, in some embodiments, the gene signature comprises at least three, at least four, at least five, at least six, at least seven, or all of the following: GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, ADIPOQ, SERPINE2, and TGFBR1. In some embodiments, the gene signature comprises at least three, at least four, at least five, at least six, at least seven, or all of the following: GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, ADIPOQ, SERPINE2, and TGFBR1. - ITGB3, PLAT, GDF15, IL8, MLANA, and / or LOXL4 and SERPINE2, - ITGB3, PLAT, GDF15, IL8, MLANA, and / or SERPINE2 and TGFBR1, - ITGB3, PLAT, GDF15, IL8 and / or MLANA and TGFBR1, - ITGB3, PLAT, GDF15, IL8 and / or TGFBR1 and SERPINE2, - ITGB3, PLAT, GDF15, IL8, SERPINE2, and / or LOXL4 and TGFBR1, - ITGB3, PLAT, GDF15, IL8, LOXL4, - ITGB3, PLAT, GDF15, IL8, SERPINE2, - ITGB3, PLAT, GDF15, IL8, TGFBR1, or - ITGB3, PLAT, GDF15, IL8, MLANA Includes.

[0031] In some embodiments, the gene signature includes at least three, at least four, at least five, at least six, at least seven, at least eight, at least nine, at least ten, at least eleven, or all of the following: GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, ADIPOQ, PRKCB, SERPINE2, ADAM12, LGALS1, and TGFBR1. In some embodiments, the gene signature includes at least three, at least four, at least five, at least six, at least seven, at least eight, at least nine, at least ten, or all of the following: GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, PRKCB, SERPINE2, ADAM12, LGALS1, and TGFBR1.

[0032] In some embodiments, the gene signature is composed of the above-mentioned genes.As will be understood by those skilled in the art, when the SLN gene signature is composed of the above-mentioned genes, the method of performing the analysis can include measuring the expression of additional genes (e.g., for normalization), but only the gene signature is used to classify individuals.

[0033] The ITGB3 gene encodes integrin beta-3. An exemplary Homo sapiens mRNA sequence can be found in the NCBI database under NM_000212.2 (June 17, 2018).

[0034] The PLAT gene encodes plasminogen activator (tissue type). An exemplary Homo sapiens mRNA sequence can be found in the NCBI database under NM_001319189.1 (July 1, 2018).

[0035] The SPP1 gene encodes secreted phosphoprotein 1. An exemplary Homo sapiens mRNA sequence can be found in the NCBI database under NM_001040058.1 (June 24, 2018).

[0036] The GDF15 gene encodes growth differentiation factor 15. An exemplary Homo sapiens mRNA sequence can be found in the NCBI database under NM_004864.3 (June 17, 2018).

[0037] The IL8 gene encodes interleukin 8. An exemplary homo sapiens mRNA sequence can be found in the NCBI database under AF043337.1 (February 1, 2001).

[0038] The MLANA gene encodes Melan-A. An exemplary homo sapiens mRNA sequence can be found in the NCBI database under NM_005511 (October 20, 2018).

[0039] LOXL4 gene, encoding lysyl oxidase-like 4. An exemplary homo sapiens mRNA sequence can be found in the NCBI database under NM_032211 (November 22, 2018).

[0040] The ADIPOQ gene encodes adiponectin (C1Q and collagen domain-containing). An exemplary homo sapiens mRNA sequence can be found in the NCBI database under NM_004797 (December 2, 2018).

[0041] The PRKCB gene encodes protein kinase C β. An exemplary homo sapiens mRNA sequence can be found in the NCBI database under NM_212535 (November 12, 2018).

[0042] The SERPINE2 gene encodes serpin family E member 2. An exemplary homo sapiens mRNA sequence can be found in the NCBI database under NM_006216 (November 17, 2018).

[0043] The ADAM12 gene encodes the ADAM metallopeptidase domain 12. An exemplary homo sapiens mRNA sequence can be found in the NCBI database under NM_003474 (August 5, 2018).

[0044] The LGALS1 gene encodes galectin 1. An exemplary homo sapiens mRNA sequence can be found in the NCBI database under NM_002305 (November 22, 2018).

[0045] The TGFBR1 gene encodes transforming growth factor beta receptor 1. An exemplary homo sapiens mRNA sequence can be found in the NCBI database under NM_004612 (October 28, 2018).

[0046] The present disclosure further provides a method of classifying an individual, comprising determining a SLN gene signature in a sample. In some embodiments, an individual may be classified as having a metastasis-positive SLN or a metastasis-negative SLN. In another embodiment, an individual may be classified as having a good or bad prognosis. A gene signature associated with SLN metastasis was previously reported (Meves et al. J Clinical Oncology 2015 33:2509-2516). The algorithm predicts SLN metastasis by using the clinicopathological variables age, Breslow depth and ulceration in combination with the primary melanoma gene expression of four genes ITGB3, LAMB1, PLAT and TP53. As shown in FIG. 3, the present SLN gene signature performs better than previously reported signatures.

[0047] In one aspect, the present disclosure provides an "N-SLN gene signature." The N-SLN gene signature classifies individuals suffering from primary cutaneous melanoma, and in particular, the gene signature classifies the risk of individuals having metastatic positive non-sentinel lymph nodes (N-SLNs). This risk assessment is useful to physicians and patients in determining treatment options and in determining patient prognosis.

[0048] In some embodiments, the N-SLN gene signature classifies the risk of metastasis-positive N-SLN. Individuals can be classified as having metastasis-positive or metastasis-negative N-SLN. Tumor cell invasion into distant lymph nodes is an indicator of poor prognosis and suggests the use of more aggressive forms of treatment. Early detection and treatment are expected to improve patient outcomes.

[0049] The examples disclosed herein demonstrate that a gene expression signature (i.e., an N-SLN gene signature) comprising one or more of the following genes: KRT14, SPP1, FN1, LOXL3 is useful for classifying individuals, particularly for determining the risk of metastasis-positive N-SLN. Thus, in one aspect, the present disclosure provides a gene signature comprising at least one of the following genes: KRT14, SPP1, FN1, LOXL3. In some embodiments, the gene signature comprises at least two or at least three of the following genes: KRT14, SPP1, FN1, LOXL3. In some embodiments, the N-SLN gene signature comprises or consists of KRT14, SPP1, FN1, LOXL3. In some embodiments, the gene signature consists of the above-mentioned genes. As will be appreciated by those skilled in the art, when the N-SLN gene signature consists of the above-mentioned genes, the method of performing the analysis may include measuring the expression of additional genes (e.g., for normalization), but only the gene signature is used to classify the individual. In some embodiments, the N-SLN gene signature is determined in individuals who suffer from cutaneous melanoma recurrence / relapse and / or have already undergone a SLN biopsy.

[0050] The KRT14 gene encodes keratin 14. An exemplary Homo sapiens mRNA sequence can be found in the NCBI database under NM_000526.4 (June 17, 2018).

[0051] The FN1 gene encodes fibronectin 1. An exemplary Homo sapiens mRNA sequence can be found in the NCBI database under NM_001306129.1 (June 3, 2018).

[0052] The LOXL3 gene encodes lysyl oxidase-like 3. An exemplary homo sapiens mRNA sequence can be found in the NCBI database under NM_001289165.1 (June 30, 2018).

[0053] The present disclosure further provides a method of classifying an individual, comprising determining an N-SLN gene signature in a sample. In some embodiments, an individual can be classified as having metastasis-positive N-SLN or metastasis-negative N-SLN. In some embodiments, a method is provided for determining both the SLN gene signature and the N-SLN gene signature.

[0054] Analysis of the gene signatures disclosed herein may be performed on any individual, including mammals and humans, but preferably humans. In some embodiments, the individual has been diagnosed with T1-T3 cutaneous melanoma. In some embodiments, the individual has not yet had a biopsy of the SLN of the primary melanoma, particularly when the gene signature is an SLN gene signature. The gene signature is associated with early age, high mitotic rate (e.g., 2 / mm 2 It is particularly useful in classifying individuals with a history of disease, a family history of poor outcome disease, and / or lymphovascular invasion.

[0055] Gene expression signatures are useful for predicting the risk or likelihood of tumor cells metastasizing to SLN or N-SLN. As will be clear to those skilled in the art, classification of an individual does not mean that 100% of all patients predicted to be at risk will in fact have detectable metastases (also referred to as sensitivity or positive percent agreement), nor does it mean that 0% of all patients predicted not to have metastases will in fact have obvious metastases (also referred to as specificity or negative percent agreement), but rather that probability or "risk". As disclosed in the examples, gene expression signatures of SLN and N-SLN exhibit high performance levels of both sensitivity and specificity. As disclosed in the examples, SLN gene signatures can predict the prognosis of individuals suffering from melanoma better than standard care SLN biopsy. Thus, the present disclosure demonstrates that gene expression signatures are useful for predicting individual prognosis.

[0056] As known to those skilled in the art, metastatic burden, as measured by the volume of metastatic disease, can vary from individual to individual. In some embodiments, metastasis refers to the presence of tumor cell clusters and does not include lymph nodes that contain only isolated or rare tumor cells. In some embodiments, metastasis refers to the presence of cell clusters at least 0.1 mm in diameter, either with or without extra-capsular extension.

[0057] Obtaining suitable samples for determining gene expression is within the capabilities of those skilled in the art. Suitable samples include primary cutaneous melanoma lesion biopsies. Such biopsies include excision lesions (e.g., extensive excision removal of tumors). Samples can be processed or preserved by any means known in the art to be compatible with gene expression profiling. For example, samples can be formalin-fixed paraffin-embedded primary cutaneous melanoma lesion biopsies and frozen samples.

[0058] Preferably, the sample is an RNA-containing sample. General methods for mRNA extraction are well known in the art and are disclosed in standard textbooks of molecular biology, including Ausubel et al. (1997) Current Protocols of Molecular Biology, John Wiley and Sons. Methods for RNA extraction from paraffin-embedded tissues are disclosed, for example, in Rupp & Locker (1987) Lab Invest. 56:A67 and De Andres et al., BioTechniques 18:42044 (1995). In particular, RNA isolation can be performed using purification kits, buffer sets and proteases from commercial manufacturers such as Qiagen, following the manufacturer's instructions (QIAGEN Inc., Valencia, Calif.). For example, total RNA can be isolated from cells in culture using Qiagen RNeasy mini-columns. Many RNA isolation kits are commercially available and can be used in the method of the present invention.

[0059] The methods disclosed herein include determining a gene expression signature. In particular, the methods include determining a gene expression level. The gene expression level may be determined by measuring the expression level of a nucleic acid or a protein. Preferably, the mRNA expression level is determined. In some embodiments, the nucleic acid or protein is purified from the sample and the gene expression is measured by nucleic acid or protein expression analysis. The protein expression level may be determined by any method known in the art, including ELISA, immunocytochemistry, flow cytometry, Western blotting, proteomics and mass spectrometry.

[0060] Preferably, nucleic acid expression level is determined.Nucleic acid expression level can be determined by any method known in the art, including RT-PCR, quantitative PCR, Northern blotting, gene sequencing, particularly RNA sequencing and gene expression profiling technology.The representative method of sequencing-based gene expression analysis includes gene expression analysis by serial analysis of gene expression (SAGE) and massively parallel signature sequencing (MPSS).

[0061] Preferably, the nucleic acid is RNA, such as mRNA or pre-mRNA. As will be understood by those skilled in the art, the RNA expression level determined can be detected directly or can be determined indirectly, such as by first performing cDNA generation and / or RNA / cDNA amplification. In some embodiments, a primary melanoma sample is obtained, RNA is extracted from the tissue sample, and then RNA transcripts of genes of interest (e.g., biomarkers and housekeeping genes) are reverse transcribed to generate cDNA of the RNA transcripts, and the cDNA is amplified to generate amplicons from the cDNA for determining the expression level of the RNA transcripts.

[0062] In some embodiments, gene expression may be determined by NanoString gene expression analysis. NanoString is a multiplexed method of detecting gene expression, providing a direct measurement of mRNA without transcription or amplification. NanoString and its aspects are described in Geiss et al., "Direct multiplexed measurement of gene expression with color-coded probe pairs" Nature Biotechnology 26, 317-325 (2008).

[0063] The expression levels do not have to be absolute values, but rather can be normalized expression values ​​or relative values. For example, the expression levels can be normalized to housekeeping or reference gene expression. Such genes include ABCF1, ACTB, ALAS1, CLTC, G6PD, GAPDH, GUSB, HPRT1, LDHA, PGK1, POLR1B, POLR2A, RPL19, RPLPO, SDHA, TBP, and TUBB.

[0064] Normalization is also useful when expression is determined based on microarray data. Normalization allows for correction of intra-microarray and across-sample variations so that simultaneous analysis of data from different chips is possible. Robust multi-array analysis (RMA) algorithms can be used to preprocess probe set data into gene expression levels for all samples. (Irizarry RA, et al., Biostatistics (2003) and Irizarry RA, et al., Nucleic Acids Res. (2003)). In addition, Affymetrix default preprocessing algorithms (MAS 5.0) can also be employed. Additional methods for normalizing expression data are described in US Patent Application Publication No. 20060136145.

[0065] In some embodiments, expression levels are determined using real-time PCR (i.e., quantitative PCR or qPCR). In real-time PCR (qPCR), reactions are characterized by the time point during cycling at which target amplification is first detected, rather than the amount of target accumulated after a fixed number of cycles. This point at which signal is first detected is referred to as the threshold cycle (Ct).

[0066] In some embodiments, gene signature expression is quantified against each other by normalization to housekeeping gene expression, subtracting the Ct of the signature gene from the average Ct of the housekeeping genes. In some embodiments, these ΔCt values ​​are then combined with the patient's age and the Breslow depth of the melanoma lesion in an algorithm that calculates a prediction of SLN metastasis. In some embodiments, the housekeeping genes used for normalization are ACTB, RPLP0 and RPL8. However, other housekeeping genes can also be used. The ratio of gene expression signals is then combined with clinical variables in an algorithm that calculates a prediction of the patient's SLNB outcome. The results are expressed as a binary classification (negative or positive). A "negative" result suggests that the individual has a low risk of metastasis-positive SLN or that the individual has an otherwise good prognosis, while a "positive" result suggests that the individual has a high risk of metastasis-positive SLN or that the individual has an otherwise poor prognosis.

[0067] The methods described herein classify individuals based on gene expression signatures. In some embodiments, differential expression of one or more signature genes in an individual indicates that the individual is at risk for metastasis or is otherwise indicative of the individual's prognosis. As used herein, "differential expression" means that the expression level measured in a subject is significantly different from a reference. The reference may be a single value or a range of values. It is within the ability of a person skilled in the art to determine an appropriate reference value. In some embodiments, the reference value is a pre-determined value. In some embodiments, the reference value is the average of expression values ​​of a particular patient class. For example, the reference value may be the average of expression values ​​in a class of patients with clinically confirmed SLN metastases (or in the N-SLN signature, patients with clinically confirmed N-SLN metastases). The reference value may also take the form of a formula or be derived from a formula. It is within the ability of a person skilled in the art to determine whether an expression level in a patient is "significantly" different from a reference.

[0068] In an exemplary embodiment, the reference value is determined from a cohort of melanoma patients who underwent SLNB as described in the Examples. It will be apparent to those skilled in the art that data from similar studies may also be used.

[0069] The strength of the correlation between the expression levels of the differentially expressed genes and the specific patient response classes can be determined by a statistical significance test. For example, χ 2 Using the χ 2 A value can be assigned to each differentially expressed marker to indicate the correlation strength of the expression of that marker with a specific patient response class.Similarly, both the t-statistics metric and the Wilkins metric provide a value or score that is indicative of the correlation strength between the expression of a marker and its specific patient response class.In addition, SAM or PAM analysis tools can be used to determine the correlation strength.

[0070] In some embodiments, the gene expression signature from an individual is compared with a reference expression signature to determine whether the gene expression signature from an individual is sufficiently similar to the reference profile.Alternatively, the gene expression signature from an individual is compared with a plurality of reference expression signatures to select the reference expression profile that is most similar to the gene expression profile from the individual.Any method known in the art for comparing two or more data sets and detecting the similarity between them can be used to compare the gene expression signature from an individual with a reference expression profile.

[0071] In machine learning and statistics, classification means identifying which category set a new observation belongs to based on a training data set containing observations (or instances) whose category membership is known. An algorithm that realizes classification in a specific implementation, among others, is known as a classifier. Many classifiers with linear or non-linear classifier boundaries are known in the art, including but not limited to ClaNC, nearest neighbor average classifier, weighted voting, naive Bayes classifier, linear discriminant analysis (LDA), quadratic discriminant analysis (QDA), support vector machine (SVM), k-nearest neighbor (k-nn) classifier. In a preferred embodiment, a logistic regression classifier is used. An exemplary embodiment that realizes a logistic regression classifier is described in the examples.

[0072] As will be understood by those skilled in the art, the training of gene expression signatures can be carried out to favor sensitivity or specificity. Sensitivity refers to the proportion of actual positives that are correctly identified, and high sensitivity is desired to avoid false negatives (e.g., patients classified as metastasis negative who are actually positive). Specificity refers to the proportion of actual negatives that are correctly identified, and high specificity is desired to avoid false positives (e.g., patients classified as metastasis positive who are actually negative). Preferably, the classifier is trained to be highly sensitive to identify individuals with metastasis.

[0073] In some embodiments, the method of classifying an individual further utilizes the individual's age and / or the Breslow depth of the tumor. Optionally, ulceration and / or mitotic rate may be determined. Breslow depth is measured from the top of the granular layer of the epidermis (or from the bottom of the ulcer if the surface is ulcerated) to the deepest invading cells across the broad base of the tumor (dermis / subcutaneous). Ulceration refers to the shedding of dead tissue and is believed to result in the death of cells in the center of the melanoma, reflecting rapid tumor growth. Mitotic rate may be measured by examining excised tumors and counting the number of cells exhibiting mitosis. The higher the number of mitoses, the more likely the tumor is metastatic. In certain embodiments, a combined model is used that includes a gene signature including GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, SERPINE2, and TGFBR1, and clinical variables of age and Breslow depth. In some embodiments, the gene signature additionally includes AIDPOQ.

[0074] The present disclosure also provides kits for determining the gene expression signatures disclosed herein. In some embodiments, the kits include primer pairs for performing qPCR on the gene signatures disclosed herein. In some embodiments, the kits include primer pairs for performing qPCR on two or more, preferably three or more of the following genes: ITGB3, PLAT, SPP1, GDF15, and IL8, and / or one or more of the following genes: KRT14, SPP1, FN1, LOXL3. In some embodiments, the kits include primer pairs for housekeeping genes such as ACTB, RPLP0, RPL8. In some embodiments, the kits include the following: DNA polymerase, deoxynucleoside triphosphates, buffer, and Mg. 2+In some embodiments, the kit further comprises one or more of the following genes: ITGB3, PLAT, GDF15 and IL8. In some embodiments, the kit comprises primer pairs for amplifying three or more of the following genes: ITGB3, PLAT, GDF15 and IL8, more preferably, the kit comprises primer pairs for amplifying ITGB3, PLAT, GDF15 and IL8. Preferably, the gene expression signature comprises GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, SERPINE2 and TGFBR1, more preferably, the gene expression signature comprises GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, ADIPOQ, SERPINE2 and TGFBR1, more preferably, the gene expression signature comprises GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, ADIPOQ, PRKCB, SERPINE2, ADAM12, LGALS1 and TGFBR1. In some embodiments, the kit comprises a control nucleic acid for one or more primer pairs, preferably for each primer. Preferably, the control nucleic acid is a cDNA, more preferably the cDNA corresponds to a sequence spanning at least one intron / exon boundary of each gene. Such cDNA is useful for distinguishing gene expression from genomic contamination. In some embodiments, one or more primers of the primer pair are chemically modified. Such modified primers include fluorescently or radioactively labeled primers.

[0075] The results of gene expression analysis disclosed herein are useful for determining diagnostic work-up schedules. For example, individuals classified as having metastasis-positive SLN or poor prognosis can undergo SLNB. In some embodiments, individuals predicted to have SLN-positive or otherwise poor prognosis are administered immunotherapy. Subsequent SLNB readouts can be a measure of response to immunotherapy.

[0076] Based on the gene expression signature, an appropriate treatment regime can be determined. As used herein, the terms "treatment", "treat" and "treating" refer to reversing, reducing, delaying the occurrence of melanoma or one or more symptoms thereof, or inhibiting its progression. In some embodiments, individuals who have metastatic positive SLNs or are otherwise classified as having a poor prognosis can be treated by SLNB. The location of SLNs can be determined based on the location of melanoma and / or using methods such as "SLN mapping", as known to those skilled in the art and described herein.

[0077] Metastatic positive N-SLNs can be treated by performing a surgical procedure, such as a surgical lymphadenectomy. Metastatic N-SLNs can be treated by a complete lymphadenectomy and / or other therapies to treat melanoma. In some embodiments, a cancer treatment is administered to an individual. In some embodiments, an "adjuvant treatment" is administered to an individual. As used herein, adjuvant treatment refers to the administration of one or more agents to a patient after surgical removal of one or more cancerous tumors, where all detectable and resectable disease (e.g., cancer) has been removed from the patient, but there remains a statistical risk of relapse. Adjuvant treatment is useful for reducing the likelihood or severity of disease recurrence.

[0078] Known melanoma therapies that can be suggested based on gene expression signatures include: - Chemotherapy: for example, Dacarbazine (DTIC), Temozolomide (Temodal), Carboplatin (Paraplatin, Paraplatin AQ), Paclitaxel (Taxol), Cisplatin (Platinol AQ) and Vinblastine (Velbe); - targeted therapy agents: for example, BRAF inhibitors (vemurafenib (Zelboraf) and dabrafenib (Tafinlar)) and MEK inhibitors (cobimetinib (Cotellic) and trametinib (Mekinist)); - Radiation therapy, - Immunotherapy: for example cytokines (e.g. interferon alpha-2b or interleukin-2), immune checkpoint inhibitors (e.g. ipilimumab (Yervoy), nivolumab (Opdivo), pembrolizumab (Keytruda)) or oncolytic immunotherapy Examples include:

[0079] Administration of a suitable drug therapy may be by any suitable route, including oral, rectal, nasal, topical (including buccal and sublingual), vaginal and parenteral (including subcutaneous, intramuscular, intravenous, intradermal, intrathecal and epidural).

[0080] As used herein, the word "comprise" and its conjugations are used in their open-ended sense, meaning to include the items following the word, but not to exclude items not specifically listed. In addition, the verb "consisting of" may be replaced with "consisting essentially of." That is, a compound or ancillary compound as defined herein may contain additional ingredients other than those specifically stated, provided that said additional ingredients do not alter the unique characteristics of the invention.

[0081] The articles "a" and "an" are used herein to refer to one or to more than one (i.e. to at least one) of the grammatical object of the article. By way of example, "an element" means one element or more than one element.

[0082] The words "approximately" or "about" when used in connection with a numerical value (such as approximately 10, about 10) preferably mean that the value may be 1% greater or less than the given value of 10.

[0083] The present invention is further illustrated in the following examples, which do not limit the scope of the invention but merely serve to clarify it. EXAMPLES

[0084] Example 1: Gene signature predictive of SLN metastatic status We recruited a cohort of 813 consecutive melanoma patients who underwent SLN biopsy at a tertiary care center. The outcomes of interest were histologically identified SLNs and metastases in SLN biopsies. Expression of 29 pro-metastasis stromal reaction genes in primary melanoma diagnostic biopsy tissue was determined by polymerase chain reaction. Regularized logistic regression was applied to clinicopathological variables and molecular data with a double-loop cross-validation (DLCV) training-validation scheme.

[0085] Clinical Data Sets Only six of the above clinical variables show significant differences between SLNB-positive and SLNB-negative. ·age Breslow Depth ·Ulceration ·Mitosis rate Clark Level Vascular and lymphatic invasion

[0086] Of the six variables, we decide to exclude Clarke level and angiolymphatic invasion and only consider age, Breslow depth, ulceration and mitotic rate, since the latter variables are not always available and their quality may vary depending on the healthcare provider performing the SLNB.

[0087] We quantified gene expression by ΔCt, not by copy number as in Meves et al. 2015, because we did not find any significant difference using copy number in our signature and it would simply increase the experimental load. Compared to Meves et al. 2015, we also omitted the KRT14 background correction. We also omitted the so-called ITLP normalization, because it is based on ad-hoc overruling of the measured expression and requires many arbitrary parameters, mainly in the form of thresholds.

[0088] [Table 1]

[0089] It is controversial whether samples with cell clusters less than 0.1 mm in diameter should be considered positive for metastasis or are generally considered negative from a clinical point of view. Therefore, we decided to exclude 43 patients out of 813 with a volume of metastatic disease equal to 1 or 2 from the training set of the classifier and evaluated the performance of the classifier separately on this set. The following 29 genes were measured: KRT14, MLANA, MITF, ITGB3, PLAT, LAMB1, TP53, AGRN, THBS2, PTK2, SPP1, COL4A1, CDKN1A, CDKN2A, PLOD3, GDF15, FN1, TNC, THBS1, CTGF, LOXL1, LOXL3, ITGA5, ITGA3, ITGA2, CSRC, CXCL1, IL8, LAMB.

[0090] performance measure To evaluate the performance of the classifier, contingency tables were constructed. From these contingency tables, two criteria were derived: PPA (positive percent agreement) and NPA (negative percent agreement). These are defined as follows:

number

[0091] To optimize the training of a classifier, we need a single performance criterion to be maximized (or minimized in the case of the error criterion). For this purpose, we use the average of PPA and NPA.

number

[0092] Other performance measures: ·Negative predictive value ·Positive predictive value Accuracy Balance accuracy Log likelihood ratio for negative outcomes Log likelihood ratio for positive outcomes ·Area under the ROC curve.

[0093] classifier Classifiers are trained on the expression levels of all 29 genes, clinicopathological variables and both.

[0094] Classifier: Penalized Maximum Likelihood Logistic Regression We used a logistic regression classifier implemented in the R package glmnet. Parameters are estimated by maximum likelihood estimation with an L1-norm penalty term (LASSO regularization) to obtain a persimmonous representation.

[0095] Double-loop cross-validation of classifiers Wessels et al. [Wessels et al. Bioinformatics, Volume 21, Issue 19, 2005, Pages 3755-3762] described a broadly applicable framework for building diagnostic classifiers from high-throughput data using double-loop cross-validation (DLCV). The DLCV exercise allows developers to estimate / predict the performance (via generalization error) of a classifier for future application on data independent of the training dataset. This method was employed in combination with forward filtering as a feature selector, t-statistics as a criterion for evaluating individual genes, and various classifiers. The training and validation procedure was performed using 3-fold cross-validation in the outer (validation) loop and 10-fold cross-validation in the inner loop with 100 iterations. In the inner loop, the algorithm learns the optimal parameter λ for LASSO regularization. At every point, the data split was stratified with respect to the class prior probabilities.

[0096] The double-loop cross-validation method can be described in a few steps: 1. For each iteration, the data is split (stratified) into three parts (different splits for each iteration). 2. For each split, two parts are used for the inner loop (training set) and the third part is used for the outer loop for validation (validation set). 3. On the training set data, 10-fold cross-validation is performed to estimate the optimal λ to be used for the LASSO penalty term (constructing the learning curve). 4. A classifier is then trained on the full training set using the top optimal λ. 5. Finally, the performance of the classifier is evaluated on the validation set. 6. After all iterations are finished, a final classifier is created with the average optimal λ using all samples. The resulting average number of n is used to train a classifier. This classifier will then be applied on the external validation set.

[0097] Typically, datasets are unbalanced with respect to class prior probabilities. Therefore, balanced accuracy is a better classification performance than accuracy because it takes into account class prior probabilities. We used Brier score as the performance criterion in the inner loop for each iteration.

[0098] [Table 2]

[0099] Table 1 presents the performance of the final classifier trained on the full 770 patient cohort classifier for four different operating points: 1) max bACC: maximum balanced accuracy, 2) SEeqSP, sensitivity equal to specificity, 3) NPV97, NPV set at 0.97 in training, 4) LRNn025, log-likelihood ratio of a negative test result set at 0.25 in training. If the coefficient is positive, a higher value suggests an increased risk. If the coefficient is negative, a reduced value suggests a decreased risk. Variables with larger (absolute) coefficients have a larger contribution.

[0100] Table 2 presents the performance of the classifier trained on DLCV averaged over 100 iterations for four different operating points: 1) max bACC: maximum balanced accuracy; 2) SEeqSP, sensitivity equal to specificity; 3) NPV97, NPV set to 0.97 in training; 4) LRNn025, log-likelihood ratio for negative test results set to 0.25 in training.

[0101] [Table 3]

[0102] Table 3. The parameters "age" are introduced in years and "Breslow depth" in millimeters. Ulceration is a Boolean variable (yes / no). The table presents the performance of the final classifier trained on the full 770 patient cohort classifier for four different operating points: 1) max bACC: maximum balanced accuracy, 2) SEeqSP, sensitivity equal to specificity, 3) NPV97, NPV set to 0.97 in training, 4) LRNn025, log likelihood ratio of a negative test result set to 0.25 in training. Table 4 presents the performance of the classifier trained on DLCV averaged over 100 iterations for four different operating points: 1) max bACC: maximum balanced accuracy, 2) SEeqSP, sensitivity equal to specificity, 3) NPV97, NPV set to 0.97 in training, 4) LRNn025, log likelihood ratio of a negative test result set to 0.25 in training.

[0103] [Table 4]

[0104] Table 5 presents the performance of the final classifier trained on the full 770 patient cohort classifier for four different operating points: 1) max bACC: maximum balanced accuracy; 2) SEeqSP, sensitivity equal to specificity; 3) NPV97, NPV set at 0.97 in training; 4) LRNn025, log-likelihood ratio of a negative test result set at 0.25 in training.

[0105] Table 6 presents the performance of the classifier trained on DLCV averaged over 100 iterations for four different operating points: 1) max bACC: maximum balanced accuracy; 2) SEeqSP, sensitivity equal to specificity; 3) NPV97, NPV set to 0.97 in training; 4) LRNn025, log-likelihood ratio for negative test results set to 0.25 in training.

[0106] FIG. 1 shows the ROC curves of logistic regression classifiers trained on DLCV for 1) gene expression, 2) clinicopathological variables, and 3) the combination of gene expression and clinicopathological variables.

[0107] Table 7 presents the average performance of classifiers trained on DLCV for 1) gene expression ("GE", i.e., ITGB3, PLAT, SPP1, GDF15 and IL8 gene signatures), 2) clinicopathological variables ("CL", i.e., age and Breslow depth), and 3) the combination of gene expression and clinicopathological variables ("GECL"). Three different operating points were considered: 1) max bACC: maximum balanced accuracy, 2) SEeqSP, sensitivity equal to specificity, and 3) NPV97, where NPV was set to 0.97 in training.

[0108] Example 2 Comparative Example Classifier based on ITLP scores (ITGB3, LAMB1, PLAT and TP53) Table 8 presents the performance of the ITLP score for the entire cohort of 770 patients.

[0109] [Table 5]

[0110] Table 9 presents the performance of the final classifier trained on the full 770 patient cohort classifier for four different operating points: 1) max bACC: maximum balanced accuracy; 2) SEeqSP, sensitivity equal to specificity; 3) NPV97, NPV set at 0.97 in training; 4) LRNn025, log-likelihood ratio of a negative test result set at 0.25 in training.

[0111] Table 10 presents the performance of the classifier trained on DLCV averaged over 100 iterations for four different operating points: 1) max bACC: maximum balanced accuracy; 2) SEeqSP, sensitivity equal to specificity; 3) NPV97, NPV set to 0.97 in training; 4) LRNn025, log-likelihood ratio for negative test results set to 0.25 in training.

[0112] Example 3 Comparative Analysis ITLP vs. ITGB3, PLAT, GDF15, SPP1 and IL8 gene signature Figure 2 shows the ROC curves for the ITLP score and the ITGB3, PLAT, GDF15, SPP1 and IL8 gene signature (referred to as "logistic regression" in the figure). The ITGB3, PLAT, GDF15, SPP1 and IL8 gene signature clearly perform better than the ITLP signature.

[0113] Example 4: Performance of gene subsets In the previous examples, five genes were used in the gene signature: ITGB3, PLAT, GDF15, SPP1 and IL8. We investigated the performance of all possible subsets of two, three and four genes. For the following signatures, the number of subsets of a particular dimension that can be selected from the total number of genes is as follows: 10 subsets from the two gene signature, 10 subsets from the three gene signature, 5 subsets from the four gene signature and 1 signature including all five. We evaluated the performance in terms of area under the ROC curve compared to the ITLP signature. The AUC (or its range) was 0.68 for ITLP, 0.72-0.75 for all subsets of 2, 0.74-0.77 for all subsets of 3, 0.76-0.77 for all subsets of 4 and 0.77 for the five gene signature. This is also shown in Figure 3. Therefore, all gene signatures that include at least two of the following genes: ITGB3, PLAT, GDF15, SPP1 and IL8 perform better than the ITLP signature.

[0114] Example 5: Performance in 43 samples with small volume metastatic disease In the first instance, patients with small volumes of metastatic disease (volumes 1 and 2) were excluded from the cohort used to train the classifier. It is controversial whether samples with cell clusters less than 0.1 mm in diameter should be considered positive for metastasis and generally negative from a clinical point of view. In this study, 43 patients were initially excluded from the analysis because they had volumes 1 or 2. Application of the ITGB3, PLAT, GDF15, SPP1 and IL8 classifiers to such patients resulted in the classification of 29 as positive and 14 as negative.

[0115] Example 6: Misclassification analysis False negatives. The misclassified positive samples are mainly from patients with thin melanomas (<2 mm) without ulceration and without angiolymphatic invasion. In other words, these are patients who present a very low a priori risk of developing metastases. A small number of samples are misclassified after all 100 iterations of the algorithm.

[0116] False positives. The misclassified negative samples are mainly from patients with thick melanomas (>2 mm) with ulceration and angiolymphatic invasion. In other words, these are patients who present a high a priori risk of developing metastases. A small number of samples are misclassified after all 100 iterations of the algorithm.

[0117] Classifier output distribution The expected probability distribution is not Gaussian, but unimodal and has a long right tail. The threshold used to select the operating point is near the mean of the distribution. The estimated probability does not exceed 0.6.

[0118] Example 7: Prognostic associations of SLN gene signatures Kaplan-Meier survival estimates for the three types of survival were generated with the SLN classifier including the genes ITGB3, PLAT, SPP1, GDF15 and IL8 (referred to as "GECL" in the examples) (Tables 19-21(a)), SLNB status (Tables 19-21(a)) and a combination of the two of them (Tables 22-24(a)). As is known in the literature, positive SLNB status is associated with poorer survival. Notably, the GECL model provided a much stronger separation in survival estimates, as evidenced by the larger hazard ratio (Table 25(a)). In addition, in multivariate analysis, the GECL classifier also has a much larger hazard ratio and a more significant p-value (see Table 26(a)). Of particular note is the very good melanoma-specific survival of the GECL-negative group, with a survival estimate of 0.966 at 160 months (Table 19(a)).

[0119] Similarly, Kaplan-Meier survival estimates for the three types of survival were generated with a SLN classifier that included the genes GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, ADIPOQ, PRKCB, SERPINE2, ADAM12, LGALS1, and TGFBR1 (referred to as "GECL" in the Examples) (Tables 19-21), SLNB status (Tables 19-21), and combinations of the two (Tables 22-24). Tables 19-26(b) list the results with the NPV set to 0.97 in training, while Tables 19-26(c) list the results with the NPV set to 0.98 in training. See Example 8 for further discussion of this classifier.

[0120] Results from the GECL classifier output were combined with SLNB status to provide four groups: true positive (TP), false positive (FP), false negative (FN) and true negative (TN). As previously suggested, there are very few FN cases and a substantial percentage of FP cases. Notably, the false positive group (GECL positive and negative SLNB status) has very similar survival estimates to the true negative group. This suggests that the GECL classifier output is a better means of providing prognosis to melanoma patients than SLNB status.

[0121] The Kaplan-Meier method is used to estimate survival probabilities at several time intervals. The log-rank test is a nonparametric test used to compare survival curves between two or more groups.

[0122] Hazard ratio (HR) is defined as the ratio of (risk of outcome in one group) / (risk of outcome in the other group) occurring in a given time interval. A hazard ratio of 1 indicates a lack of association, a hazard ratio greater than 1 suggests an increased risk, and a hazard ratio less than 1 suggests a smaller risk. Hazard ratios are used to express the relative difference between two groups only.

[0123] Example 8: Refinement of the SLN classifier A four-gene panel was measured in the entire 855 cohort for a total of 109 unique genes. However, gene expression could not be measured in some samples because samples could not be retrieved or sufficient RNA was not present at the time of analysis. Therefore, discovery was performed in a cohort of 754 patients instead of 770. The discovery cohort did not include patients with small volume metastatic disease.

[0124] For each classifier we report the following: Performance of the final classifier trained on the entire cohort · Average performance of cross-validation (100 iterations of double-loop cross-validation with 3 folds in the outer loop and 10 folds in the inner loop). The performance of the 3 folds are concatenated to cover the whole cohort.

[0125] For each classifier, performance is calculated at four different operating points. MaxbACC: Maximum balance accuracy · SEeqSP: Sensitivity equal to specificity NPV97: Negative predictive value equal to 97% in training NPV98: Negative predictive value equal to 98% in training

[0126] Clinicopathological Model (CL) The parameters of the logistic classifier model based on clinicopathological variables are as follows: Features Parameter (intercept) -2.0547083 Age -0.0112913 Breslow depth 0.6116335 With vascular and lymphatic invasion 0.1205238

[0127] Table 27 presents the performance of the final classifier trained on the full 754 patient cohort for four different operating points: 1) max bACC: maximum balanced accuracy, 2) SEeqSP, sensitivity equal to specificity, 3) NPV97, NPV set to 0.97 in training, 4) NPV98, NPV set to 0.98 in training. If the coefficient is positive, a higher value suggests an increased risk. If the coefficient is negative, a reduced value suggests a decreased risk. Variables with larger (absolute) coefficients have a larger contribution.

[0128] Table 28 presents the performance of the classifier trained on DLCV averaged over 100 iterations for four different operating points: 1) max bACC: maximum balanced accuracy; 2) SEeqSP, sensitivity equal to specificity; 3) NPV97, NPV set to 0.97 in training; and 4) NPV98, NPV set to 0.98 in training.

[0129] [Table 6]

[0130] Table 29 presents the performance of the final classifier trained on the full 754 patient cohort for four different operating points: 1) max bACC: maximum balanced accuracy, 2) SEeqSP, sensitivity equal to specificity, 3) NPV97, NPV set to 0.97 in training, 4) NPV98, NPV set to 0.98 in training. If the coefficient is positive, a higher value suggests an increased risk. If the coefficient is negative, a reduced value suggests a decreased risk. Variables with larger (absolute) coefficients have a larger contribution.

[0131] Table 30 presents the performance of the classifier trained on DLCV averaged over 100 iterations for four different operating points: 1) max bACC: maximum balanced accuracy; 2) SEeqSP, sensitivity equal to specificity; 3) NPV97, NPV set to 0.97 in training; and 4) NPV98, NPV set to 0.98 in training.

[0132] [Table 7]

[0133] Table 31 presents the performance of the final classifier trained on the full 754 patient cohort for four different operating points: 1) max bACC: maximum balanced accuracy; 2) SEeqSP, sensitivity equal to specificity; 3) NPV97, NPV set to 0.97 in training; 4) NPV98, NPV set to 0.98 in training. If the coefficient is positive, a higher value suggests an increased risk. If the coefficient is negative, a reduced value suggests a decreased risk. Variables with larger (absolute) coefficients have a larger contribution.

[0134] Table 32 presents the performance of the classifier trained on DLCV averaged over 100 iterations for four different operating points: 1) max bACC: maximum balanced accuracy; 2) SEeqSP, sensitivity equal to specificity; 3) NPV97, NPV set to 0.97 in training; and 4) NPV98, NPV set to 0.98 in training.

[0135] FIG. 6 shows the ROC curves of the logistic regression classifier trained on DLCV for 1) gene expression, 2) clinicopathological variables, and 3) the combination of gene expression and clinicopathological variables, and FIG. 7 shows the negative predictive value (NPV) versus sentinel lymph node reduction rate (SLNB RR) of the logistic regression classifier trained on DLCV for 1) gene expression, 2) clinicopathological variables, and 3) the combination of gene expression and clinicopathological variables.

[0136] Comparison at different operating points (OP): CL vs. CE vs. GECL Table 33 shows the average performance of classifiers trained on DLCV for 1) gene expression ("GE", i.e., GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, ADIPOQ, PRKCB, SERPINE2, ADAM12, LGALS1, and TGFBR1 gene signatures), 2) clinicopathological variables ("CL", i.e., age, Breslow depth, and presence of angiolymphatic invasion), and 3) combinations of gene expression and clinicopathological variables ("GECL", i.e., age, Breslow depth, GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, ADIPOQ, SERPINE2, and TGFBR1). For operating points, maximum balanced accuracy (max bACC).

[0137] Table 34 shows the average performance of classifiers trained on DLCV for 1) gene expression ("GE", i.e., GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, ADIPOQ, PRKCB, SERPINE2, ADAM12, LGALS1, and TGFBR1 gene signatures), 2) clinicopathological variables ("CL", i.e., age, Breslow depth, and presence of angiolymphatic invasion), and 3) a combination of gene expression and clinicopathological variables ("GECL", i.e., age, Breslow depth, GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, ADIPOQ, SERPINE2, and TGFBR1). For the operating point, we use sensitivity equal to specificity (SEeqSP).

[0138] Table 35 shows the average performance of classifiers trained on DLCV for 1) gene expression ("GE", i.e., GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, ADIPOQ, PRKCB, SERPINE2, ADAM12, LGALS1, and TGFBR1 gene signatures), 2) clinicopathological variables ("CL", i.e., age, Breslow depth, and presence of angiolymphatic invasion), and 3) a combination of gene expression and clinicopathological variables ("GECL", i.e., age, Breslow depth, GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, ADIPOQ, SERPINE2, and TGFBR1). For the operating point, we use the NPV (NPV97) set to 0.97 in training.

[0139] Table 36 shows the average performance of classifiers trained on DLCV for 1) gene expression ("GE", i.e., GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, ADIPOQ, PRKCB, SERPINE2, ADAM12, LGALS1, and TGFBR1 gene signatures), 2) clinicopathological variables ("CL", i.e., age, Breslow depth, and presence of angiolymphatic invasion), and 3) a combination of gene expression and clinicopathological variables ("GECL", i.e., age, Breslow depth, GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, ADIPOQ, SERPINE2, and TGFBR1). For the operating point, we use the NPV (NPV98) set to 0.98 in training.

[0140] Performance stratified by T stage Table 37 presents the average performance stratified by T stage of classifiers trained on DLCV for clinicopathological variables ("CL", i.e., age, Breslow depth and presence of angiolymphatic invasion). For the operating point, we use the NPV (NPV97) set to 0.97 in training.

[0141] Table 38 shows the average performance stratified by T stage of classifiers trained on DLCV for gene expression ("GE", i.e., GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, ADIPOQ, PRKCB, SERPINE2, ADAM12, LGALS1, and TGFBR1). For the operating point, we use NPV (NPV97) set to 0.97 in training.

[0142] Table 39 shows the average performance stratified by T stage of classifiers trained on DLCV for gene expression and clinicopathological variable combinations ("GECL" i.e. age, Breslow depth, GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, ADIPOQ, SERPINE2 and TGFBR1). For operating points, use NPV set to 0.97 in training (NPV97).

[0143] Table 40 presents the average performance stratified by T stage of classifiers trained on DLCV for clinicopathological variables ("CL", i.e., age, Breslow depth and presence of angiolymphatic invasion). For the operating point, we use the NPV (NPV98) set to 0.98 in training.

[0144] Table 41 shows the average performance stratified by T stage of classifiers trained on DLCV for gene expression ("GE", i.e., GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, ADIPOQ, PRKCB, SERPINE2, ADAM12, LGALS1, and TGFBR1). For the operating point, we use NPV (NPV98) set to 0.98 in training.

[0145] Table 42 shows the average performance stratified by T stage of classifiers trained on DLCV for gene expression and clinicopathological variable combinations ("GECL" i.e. age, Breslow depth, GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, ADIPOQ, SERPINE2 and TGFBR1). For operating points, use NPV set to 0.98 in training (NPV98).

[0146] Performance stratified by clinical staging Table 43 presents the average performance stratified by clinical stage for classifiers trained on DLCV for clinicopathological variables ("CL", i.e., age, Breslow depth and presence of angiolymphatic invasion). For the operating point, we use the NPV (NPV97) set to 0.97 in training.

[0147] Table 44 shows the average performance of classifiers trained on DLCV for gene expression ("GE", i.e., GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, ADIPOQ, PRKCB, SERPINE2, ADAM12, LGALS1, and TGFBR1), stratified by clinical stage. For operating points, use NPV (NPV97) set to 0.97 in training.

[0148] Table 45 shows the average performance stratified by clinical stage of classifiers trained on DLCV for combinations of gene expression and clinicopathological variables ("GECL" i.e. age, Breslow depth, GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, ADIPOQ, SERPINE2 and TGFBR1). For the operating point, we use NPV (NPV97) set to 0.97 in training.

[0149] Table 46 presents the average performance stratified by clinical stage for classifiers trained on DLCV for clinicopathological variables ("CL", i.e., age, Breslow depth and presence of angiolymphatic invasion). For the operating point, we use the NPV (NPV98) set to 0.98 in training.

[0150] Table 47 shows the average performance of classifiers trained on DLCV for gene expression ("GE", i.e., GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, ADIPOQ, PRKCB, SERPINE2, ADAM12, LGALS1, and TGFBR1), stratified by clinical stage. For operating points, use NPV (NPV98) set to 0.98 in training.

[0151] Table 48 shows the average performance stratified by clinical stage of classifiers trained on DLCV for combinations of gene expression and clinicopathological variables ("GECL" i.e. age, Breslow depth, GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, ADIPOQ, SERPINE2 and TGFBR1). For the operating point, we use NPV (NPV98) set to 0.98 in training.

[0152] Gene subsets FIG. 8 depicts box plots of the area under the curve (AUC) of the ROC curves for logistic regression classifiers trained on the entire cohort using subsets of 2, 3, 4, 5, 6, 7, 8 genes selected from GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, ADIPOQ, SERPINE2, and TGFBR1.

[0153] Table 49 presents the number of subsets of a particular dimension that can be selected from the total number of genes in each signature, as well as the performance in terms of minimum and maximum areas under the ROC curve.

[0154] Example 9: Non-Sentinel Lymph Node (N-SLN) Profiler. Complete lymphadenectomy (CLND) with removal of non-sentinel lymph nodes (N-SLNs) has been the standard for patients with clinical node-negative melanoma with positive sentinel lymph nodes (SLNs). SLN biopsy followed immediately by CLND improves local disease control, and randomized clinical trials have shown that early surgery for smaller volumes of SLN-positive disease results in fewer long-term sequelae (e.g., lymphedema) than surgery at nodal relapse. Furthermore, SLN and N-SLN metastases are poor prognostic factors used to select patients for adjuvant therapy. However, in patients enrolled in MSLT-II, CLND does not confer a survival benefit and has a higher complication rate than SLN surgery alone. To improve patient selection for CLND, new methods are needed to identify patients likely to respond to CLND, i.e., those at risk of local metastasis to N-SLNs. Here, to predict the status of N-SLN, i.e., the presence or absence of metastasis, we design three classifiers based on gene expression (KRT14, SPP1, FN1 and LOXL3), clinicopathological variables (age, number of positive SLNs, maximum dimension of SLNs, maximum Breslow depth and maximum mitotic rate) and both. These classifiers can be used to select which patients should undergo CLND procedure.

[0155] The methodology used was nearly identical to that shown in Example 1. The same 29 genes were evaluated. The parameters of the gene expression-based logistic regression classifier model were as follows:

[0156] [Table 8]

[0157] Model performance The total number of patients in the cohort was 140.

[0158] Table 11 presents the performance of the classifier trained on DLCV averaged over 100 iterations for four different operating points: 1) max bACC: maximum balanced accuracy; 2) SEeqSP, sensitivity equal to specificity; 3) NPV97, NPV set to 0.97 in training; 4) LRNn025, log-likelihood ratio for negative test results set to 0.25 in training.

[0159] Table 12 presents the performance of the classifier trained on DLCV averaged over 100 iterations for four different operating points: 1) max bACC: maximum balanced accuracy; 2) SEeqSP, sensitivity equal to specificity; 3) NPV97, NPV set to 0.97 in training; 4) LRNn025, log-likelihood ratio for negative test results set to 0.25 in training.

[0160] [Table 9]

[0161] Table 13 presents the performance of the final classifier trained on the full 140 patient cohort classifier for four different operating points: 1) max bACC: maximum balanced accuracy; 2) SEeqSP, sensitivity equal to specificity; 3) NPV97, NPV set at 0.97 in training; 4) LRNn025, log-likelihood ratio of a negative test result set at 0.25 in training.

[0162] Table 14 presents the performance of the classifier trained on DLCV averaged over 100 iterations for four different operating points: 1) max bACC: maximum balanced accuracy; 2) SEeqSP, sensitivity equal to specificity; 3) NPV97, NPV set to 0.97 in training; 4) LRNn025, log-likelihood ratio for negative test results set to 0.25 in training.

[0163] [Table 10]

[0164] Parameters of gene expression-based logistic classifier model. Table 15 presents the performance of the final classifier trained on the full 140 patient cohort classifier for four different operating points: 1) max bACC: maximum balanced accuracy; 2) SEeqSP, sensitivity equal to specificity; 3) NPV97, NPV set at 0.97 in training; 4) LRNn025, log-likelihood ratio of a negative test result set at 0.25 in training.

[0165] Table 16 presents the performance of the classifier trained on DLCV averaged over 100 iterations for four different operating points: 1) max bACC: maximum balanced accuracy; 2) SEeqSP, sensitivity equal to specificity; 3) NPV97, NPV set to 0.97 in training; 4) LRNn025, log-likelihood ratio for negative test results set to 0.25 in training.

[0166] Figure 4 shows the average ROC curves of logistic regression classifiers trained on DLCV for 1) gene expression, 2) clinicopathological variables, and 3) the combination of gene expression and clinicopathological variables. The x-axis represents the false positive discovery rate (i.e., 1-specificity) and the y-axis represents the true discovery rate (i.e., sensitivity).

[0167] Figure 5: Box plots of the area under the ROC curve for the different gene subsets and the full set of four genes.

[0168] Table 17 presents the average performance of classifiers trained on DLCV for 1) gene expression, 2) clinicopathological variables, and 3) the combination of gene expression and clinicopathological variables. Three different operating points were considered: 1) max bACC, maximum balanced accuracy; 2) SEeqSP, sensitivity equal to specificity; and 3) NPV97, NPV set to 0.97 in training.

[0169] The N-SLN profiler signature contains four genes: KRT14, SPP1, FN1 and LOXL3. We also examined all possible subsets of two, three and four genes.

[0170] [Table 11]

[0171] We evaluated the performance in terms of the area under the ROC curve (see tables and figures).

[0172] [Table 12]

[0173] FIG. 5 shows box plots of the area under the ROC curve for the different gene subsets and the full set of four genes.

[0174] We describe the discovery, design and development of the N-SLN Profiler. We have shown that the N-SLN Profiler can be used to select which patients should undergo a CLND procedure.

[0175] The performance of gene expression-based classifiers is of interest because currently (i) there are no methods available to select patients who are likely to benefit from the CLND procedure and (ii) the clinicopathological variables used by the classifiers are not always available.

[0176] Example 10 Preselection based on Breslow depth (BD≦2 vs. BD>2) Breslow depth (BD) is an important clinicopathological variable characterizing primary cutaneous melanomas, and it is known that thin melanomas (BD ≦ 2 mm) may have molecular and physiologically different characteristics compared to thick melanomas (BD > 2 mm). Therefore, it would be reasonable to select two different operating points for thin and thick melanomas. For thin melanomas with Breslow depth ≦ 2 mm (cohort of 561 samples), we continued to use the classifier described in the previous section (7.8) with an operating point selected such that the NPV was 0.97 in training.

[0177] Table 18a shows the average performance of classifiers trained on DLCV for 1) gene expression, 2) clinicopathological variables, and 3) the combination of gene expression and clinicopathological variables. The operating point chosen is NPV97, i.e., NPV was set to 0.97 in training.

[0178] For thick melanomas with BD>2 mm (cohort of 209 samples), we use the same classifier but with a different operating point, i.e. the point chosen to maximize the balanced accuracy.

[0179] Table 18b shows the average performance of classifiers trained on DLCV for 1) gene expression, 2) clinicopathological variables, and 3) the combination of gene expression and clinicopathological variables. The selected operating points were chosen to maximize the balance accuracy.

[0180] The performance for Breslow depth >2 is inferior to that for Breslow depth ≤2 mm, but is still acceptable because the classifier achieves an NPV of 90% in the subpopulation with a prior probability of SLNB positivity of 35%, and when undergoing testing with a negative outcome, the probability of having a positive SLN drops from 35% to 10%.

[0181] Methods for determining whether a subject is classified as lymph node positive or lymph node negative As an example of a classification method, we use hypothetical data (see table) and for simplicity, two genes are used to predict by a classifier whether a sample will be labeled as node positive or node negative (the method / model is identical for both SLN Profiler and N-SLN Profiler, just the parameters and gene identities and clinical variables differ). The table shows the model parameters β0, β1, β2 for the two gene (x,y) toy model, the hypothetical gene expression data ΔCt, and the estimated log odds ratios.

number

[0182] [Table 13]

[0183] Table. Model parameters β0, β1, β2 and gene expression data for genes x and y ΔC t , estimated log odds ratio

number

number

[0184] Example 11 Similar to the analysis performed in Example 8, utilizing eight genes and two clinicopathological variables (age and Breslow depth), a preferred embodiment of the present invention was found to include the following gene sets: GDF15, MLANA, PLAT, IL8, ITGB3, LOXL4, SERPINE2, and TGFBR1. The parameters of the logistic regression model are provided below.

[0185] [Table 14]

[0186] Various operating points were evaluated and several were found to provide clinically relevant classifiers with high NPV and significant reduction in the number of SLNB procedures (SLNB.RR = SLNB reduction rate). The 975 operating point with an NPV of 0.116 was particularly favorable (see table below).

[0187] [Table 15]

[0188] Performance comparison of the GECL model using eight genes and two clinicopathological variables with those in Examples 1 and 8 suggests that it outperforms the expression-only (GE) or clinicopathological-only (CL) models (NPV and SLNB.RR) and performs very similarly to the GECL model in Example 1.

[0189] Kaplan-Meier analysis was performed to evaluate the GECL model with SLNB in ​​terms of RFS, DRFS and MSS. The 5-year / 60-month survival rates are reported in the table below. Both GECL and SLNB individually have a large separation between the positive / negative groups. More importantly, patients with negative SLNB (which is itself an oversight) identified as positive with GECL have a very poor survival. This demonstrates the clinical relevance of the GECL model as a prognostic marker.

[0190]

Table 16

[0191]

Table 17

[0192]

Table 18

[0193]

Table 19

[0194]

Table 20

[0195]

Table 21

[0196]

Table 22

[0197]

Table 23

[0198]

Table 24

[0199]

Table 25

[0200]

Table 26

[0201]

Table 27

[0202]

Table 28

[0203]

Table 29

[0204]

Table 30

[0205]

Table 31

[0206]

Table 32

[0207]

Table 33

[0208]

Table 34

[0209]

Table 35

[0210]

Table 36

[0211]

Table 37

[0212]

Table 38

[0213]

Table 39

[0214]

Table 40

[0215]

Table 41

[0216]

Table 42

[0217]

Table 43

[0218]

Table 44

[0219]

Table 45

[0220]

Table 46

[0221]

Table 47

[0222]

Table 48

[0223]

Table 49

[0224]

Table 50

[0225]

Table 51

[0226]

Table 52

[0227]

Table 53

[0228]

Table 54

[0229]

Table 55

[0230]

Table 56

[0231]

Table 57

[0232]

Table 58

[0233]

Table 59

[0234]

Table 60

[0235]

Table 61

[0236]

Table 62

[0237]

Table 63

[0238]

Table 64

[0239]

Table 65

[0240]

Table 66

Claims

1. 1. A method of classifying an individual suffering from primary cutaneous melanoma, comprising determining a gene expression signature in a sample from the individual, the gene expression signature comprising ITGB3, PLAT, GDF15 and IL8, and wherein the individual is classified as having metastasis-positive SLNs or as having metastasis-negative SLNs; The method, wherein the sample is a biopsy from a primary cutaneous melanoma lesion.

2. A method for determining a therapeutic and / or diagnostic work-up schedule for an individual suffering from cutaneous melanoma, comprising determining expression levels of ITGB3, PLAT, GDF15 and IL8 in a sample from the individual, and determining a therapeutic and / or diagnostic work-up schedule based on the expression levels.

3. 1. A method for predicting the prognosis of an individual suffering from primary cutaneous melanoma, comprising determining a gene expression signature in a sample from the individual, the gene expression signature comprising ITGB3, PLAT, GDF15 and IL8, and wherein the prognosis of the individual is determined based on the gene expression levels.

4. The method of claim 2 or 3, wherein the individual is classified as having a metastasis-positive SLN or as having a metastasis-negative SLN.

5. The method of any one of claims 1 to 4, wherein the individual is selected for SLNB based on the classification and / or expression level.

6. The method of any one of claims 1 to 5, wherein individuals classified as having metastatic positive SLNs are treated by performing SLNB.

7. The method of any one of claims 1 to 6, wherein the prognosis of the individual is determined based on the gene expression levels, and the individual is classified as having a poor prognosis or a good prognosis.

8. The method of any one of claims 1 to 7, wherein the gene expression signature further comprises MLANA, LOXL4, SERPINE2 and TGFBR1.

9. 9. The method of claim 8, wherein the gene expression signature further comprises ADIPOQ.

10. The method of claim 8 or 9, wherein the gene expression signature further comprises PRKCB, ADAM12, LGALS1.

11. 11. The method of any one of claims 1 to 10, further comprising classifying an individual as suffering from primary cutaneous melanoma, comprising determining a gene expression signature in a sample from said individual, said gene expression signature comprising at least one of the following genes: KRT14, SPP1, FN1, LOXL3.

12. 12. The method of any one of claims 1 to 11, wherein the method of classification further comprises determining the age and / or Breslow depth of the individual, optionally also determining ulceration of melanoma lesions.

13. The method of any one of claims 1 to 12, wherein the expression is determined by detection of RNA.

14. 1. A kit for use in classifying an individual suffering from primary cutaneous melanoma, comprising: a) ITGB3, PLAT, GDF15 and IL8, and / or b) at least one of the following genes: KRT14, SPP1, FN1, and LOXL3; A kit comprising a primer pair for amplifying

15. The kit of claim 14, further comprising a primer pair for amplifying at least one reference gene.

16. The kit of claim 15 , wherein the reference genes are ACTB, RPLP0 and / or RPL8.

17. The kit of claim 14, comprising primer pairs for amplifying MLANA, LOXL4, SERPINE2 and TGFBR1.

18. 18. The kit of claim 17, further comprising a primer pair for amplifying ADIPOQ.

19. The kit of claim 17 or 18, further comprising a primer pair for amplifying PRKCB, ADAM12, or LGALS1.

Images