TWI872797B - Diagnostic method of endometrial cancer - Google Patents

Abstract

The present invention provides a method for evaluating whether an individual suffers from endometrial cancer, which comprises analyzing the methylation level of one or more target genes to determine whether the individual suffers from endometrial cancer.

Description

Diagnosis of endometrial cancer

The present invention relates to a method for diagnosing endometrial cancer, and in particular, to evaluating whether a patient has endometrial cancer by using the methylation status of specific genes on cervical mucus.

Endometrial cancer (EC) is the most common gynecological malignancy, and its incidence and mortality are increasing worldwide. Early detection and treatment of EC are associated with a good prognosis, and in the early stages of the disease, it can usually be cured by surgery alone. However, patients with advanced EC usually require multiple types of treatment, and the treatment results are less than ideal. EC is usually divided into type I and type II based on clinical pathological characteristics and genetic abnormalities. The genomic data from the Cancer Genome Atlas identified four molecular classifications of EC, namely POLE mutation status, mismatch repair deficiency, p53 mutation, and no specific molecular profile (NSMP), each with different prevalence, diagnostic tests, molecular features, histological characteristics, and clinical outcomes. Unfortunately, such traditional and molecular classifications are not suitable for early detection or screening of EC.

The endometrium is susceptible to genetic and epigenetic changes. In addition to genetic abnormalities, epigenetic changes are also associated with the development of cancer. A well-established indicator of cancer development is DNA methylation aberrations, which include gene-specific hypermethylation and global hypomethylation, and DNA methylation aberrations occurring in cancer are very important for diagnostic development. In the past few years, the use of DNA methylation to identify cancer detection biomarkers has been very promising.

This study investigated the potential of cellular DNA methylation in cervical mucus for screening EC and evaluated the efficacy of candidate genes using physician-collected samples (MPap 2.0) and self-collected (MPap 3.0) samples.

The present invention uses 450K DNA methylation chips to generate methylation profiles of endometrial tissue and cervical mucus from patients with EC or leiomyoma. Candidate genes were screened by targeted sulfite sequencing for 121 physician-collected samples (MPap 2.0) and 113 tampon samples (MPap 3.0) and validated using qMS-PCR. Among 78 genes with 95 differentially methylation positions (DMPs), 18 genes with diagnostic potential were found; and 6 genes (ADCY8, EPHA10, HPSE2, TBX5, CDO1, and BHLHE22) were further screened by qMS-PCR. These six genes were used in logical regression and machine learning to design a prediction model, and the model was used in samples collected by doctors (area under the curve (AUC): 0.94-0.96) and self-collected samples (AUC: 0.91-0.95) and showed high accuracy. In addition, the AUC of the three gene combinations (TBX5, CDO1, BHLHE22) on samples obtained by different sampling methods were 0.97 (MPap 2.0) and 0.96 (MPap 3.0), respectively, and maintained 100% sensitivity and 95% specificity.

Therefore, the present invention discovered that the methylation status of genes on cervical mucus can be used to diagnose EC.

The terms "a" or "an" used herein describe elements and components of the present invention. The terms are only for the convenience of the description and basic concepts of the present invention. The description should be understood to include one or at least one, and unless the context indicates otherwise, singular terms include plural forms and plural terms include singular forms. When the word "comprising" is used in a claim, the terms "a" or "an" may mean one or more than one.

As used herein, the term "or" may mean "and/or".

A method for evaluating whether an individual suffers from endometrial cancer, comprising the following steps: (a) providing a sample obtained from the individual; (b) determining the methylation status of at least one target gene in the sample, wherein the at least one target gene comprises at least one of the following: EPHA10, ADCY8 and HSPE2; (c) determining whether the at least one target gene is highly methylated; and (d) evaluating whether the individual suffers from endometrial cancer based on the result of step (c), wherein the high methylation of the at least one target gene indicates that the individual suffers from endometrial cancer.

In a specific embodiment, the at least one target gene further comprises BHLHE22, CDO1 and TBX5, and the step (b) further comprises determining the methylation status of at least one of BHLHE22, CDO1 and TBX5.

According to certain optional embodiments, the at least one target gene comprises EPHA10, ADCY8, HSPE2, and BHLHE22. In some embodiments, the at least one target gene comprises EPHA10, ADCY8, HSPE2, BHLHE22, and CDO1. In other embodiments, the at least one target gene comprises EPHA10, ADCY8, HSPE2, BHLHE22, CDO1, and TBX5. In certain embodiments, the at least one target gene comprises any one, two, or three of EPHA10, ADCY8, HSPE2, BHLHE22, CDO1, and TBX5.

The present invention also provides a method for evaluating whether an individual suffers from endometrial cancer, comprising at least the following steps: (a) providing a sample obtained from the individual; (b) determining the methylation status of a gene combination in the sample, wherein the gene combination comprises BHLHE22, CDO1, TBX5, EPHA10, ADCY8 and HSPE2; (c) determining whether the gene combination is highly methylated; and (d) evaluating whether the individual suffers from endometrial cancer based on the result of step (c), wherein the high methylation of the gene combination indicates that the individual suffers from endometrial cancer.

As used herein, the term "diagnosis" or "assessment" refers to the identification of pathological conditions, diseases or symptoms, such as tumors of various gynecological tissue origin, including cervix, ovary, endometrium/uterus, vagina, vulva, uterus and peritoneum lining the uterus. In some embodiments, the term "diagnosis" or "assessment" refers to the identification of malignant tumors and benign tumors. In other embodiments, the term "diagnosis" or "assessment" refers to the identification of malignant tumors and normal tissues.

In one embodiment, the endometrial cancer comprises type 1 endometrial cancer or type 2 endometrial cancer.

As used herein, the term "individual" refers to animals including humans. Therefore, the term "individual" includes any mammal that can benefit from the methods of the present invention. The term "mammal" refers to all members of the class Mammalia. In one embodiment, the individual is a human.

In the present invention, the sample is a sample taken from an individual (preferably a human) or a sample originally existing in an individual (preferably a human). For example, the sample is an endometrial mucus, tissue or cell sample taken from the individual. In a specific embodiment, the sample is from endometrial mucus or tissue, cervical smear cells, uterine or vaginal secretions, or menstrual blood.

In another specific embodiment, the sample is obtained by collecting using sanitary napkins, tampons, sanitary pads, cotton swabs, cotton swabs, vaginal lavage, cervical brushes or cotton balls. The sampler can obtain the sample by gently rubbing the above tools or items on the cervix or endometrium.

As used herein, the term "methylation" refers to the covalent bonding of a methyl group at the C5 position of cytosine in a CpG dinucleotide in the core promoter region of a gene. The term "methylation state" refers to the presence or absence of 5-methyl-cytosine (5-mCyt) on one or more CpG dinucleotides in a gene or nucleic acid sequence of interest. As used herein, the term "methylation level" refers to the amount of methylation in one or more copies of a gene or nucleic acid sequence of interest. The methylation level can be calculated as an absolute value of methylation in the gene or nucleic acid sequence of interest. Furthermore, the "relative methylation level" can be expressed as the amount of methylated DNA relative to the total amount of DNA, or as the number of methylated copies of a gene or nucleic acid sequence of interest relative to the total number of copies of the gene or nucleic acid sequence. In addition, "methylation level" can refer to the percentage of methylated CpG sites in the DNA fragment of interest.

As used herein, the term "methylation profile" refers to a set of data on the methylation level of one or more target genes in a sample. In certain embodiments, the methylation profile is a profile compared to a methylation reference profile obtained from a sample of unknown type, such as a cancerous sample or a non-cancerous sample or a sample of different cancer stages.

In this article, "differential methylation" refers to the difference between the methylation level of one or more target genes in a sample or population and the methylation level of one or more target genes in other samples or populations. Differential methylation can be divided into increased methylation level (hypermethylation) or decreased methylation level (hypomethylation). Here, "hypermethylation" means that the methylation level of the target gene in the test sample is at least 10% higher than the average methylation level of the target gene in the reference sample. According to various different embodiments of the present invention, the increase in methylation level refers to an increase of at least 15, 20, 25, 30, 35, 40, 45 or 50%.

In a specific embodiment, the method for determining the methylation status of the gene in the sample in step (b) of the above method comprises methylation-specific polymerase chain reaction (MSP), quantitative methylation-specific polymerase chain reaction (qMSP), bisulfite sequencing (BS), bisulfite pyrosequencing, microarrays, mass spectrometry, denaturing high-performance liquid chromatography (DHPLC), pyrosequencing, methylated DNA immunoprecipitation (MeDIP or mDIP) coupled with quantitative polymerase chain reaction (MSP), ... reaction), methylated DNA immunoprecipitation sequencing (MeDIP-seq), nanopore sequencing, fiber optic biosensor or optical biosensor.

The method proposed by the present invention can be used to assess whether an individual has endometrial cancer in the early stage (stage I or II). Therefore, the method of the present invention can diagnose cancer early before it progresses to a more malignant stage. At the same time, the present invention also proves that whether the sample is collected by a doctor or collected by oneself, similar experimental results can be obtained by applying the method of the present invention. Therefore, the present invention has the characteristics of self-collection, which is convenient and highly private. The above are all advantages of the present invention in clinical application.

Early detection is the best way to improve EC treatment outcomes. In addition, it is also necessary to screen EC using a simple, non-invasive, painless, convenient, and accurate test method. Therefore, the present invention adopts a non-invasive cervicovaginal sampling method and screens EC through gene methylation analysis on cervical mucus.

In the present invention, the combination of target genes found in cervical mucus has a high degree of discrimination ability, whether it is used in hospitals or at home for cervical and vaginal sampling to perform DNA methylation analysis to detect EC, which is conducive to the early detection of gynecological tumors.

FIG1 is the discovery of candidate genes. FIG1A is a flowchart of the workflow of the present invention. FIG1B shows the DNA methylation phenotypes of 18 genes with 97 CpG sites from 16 non-EC cases and 16 EC cases (type I EC: 7 cases, type II EC: 9 cases) according to the sulfited target sequence sites. Each column represents a sample. FIG1C shows the average methylation status of the 18 genes obtained from vaginal samples. FIG1D shows the AUC of the 18 genes in the samples shown in FIG1C.

Figure 2 shows the prediction performance of physician-collected cervicovaginal samples. Figure 2A shows the methylation status of 6 genes in vaginal samples (physician-collected) from 81 non-EC cases and 42 EC cases using qMS-PCR analysis. ***p<0.001, ****p<0.0001. Figure 2B shows the differences in DNA methylation status of 6 genes as a heat map. The green and orange columns represent non-EC and EC samples, respectively. The degree of methylation ranges from low (blue) to high (red). Figure 2C shows the AUC of the best model of the 5 machine learning algorithms in the training set. Figure 2D shows the AUC of the best model of the 5 machine learning algorithms in the test set. Figure 2E shows the prediction performance of each model in terms of sensitivity, specificity, PPV, and NPV in the training set. Figure 2F shows the prediction performance of sensitivity, specificity, PPV, and NPV of each model in the test set.

Figure 3 shows the prediction performance of self-collected cervicovaginal samples. Figure 3A shows the methylation status of 6 genes in vaginal samples (self-collected by intravaginal tampons) of 59 non-EC cases and 54 EC cases analyzed by qMS-PCR. ***p<0.001, ****p<0.0001. Figure 3B shows the difference in DNA methylation status of 6 genes as shown in the heat map. The samples were stratified into training and test sets to establish the machine learning algorithm. Figure 3C shows the AUC of the best model of the 5 machine learning algorithms in the training set. Figure 3D shows the AUC of the best model of the 5 machine learning algorithms in the test set. Figure 3E shows the prediction performance of sensitivity, specificity, PPV and NPV of each model in the training set. Figure 3F shows the prediction performance of sensitivity, specificity, PPV, and NPV of each model in the test set.

Figure 4 shows the performance of MPap 2.0 and MPap 3.0. Figure 4A shows the feature selection of 6 genes analyzed in samples collected by doctors using the Boruta algorithm. Figure 4B shows the feature selection of 6 genes analyzed in self-collected samples using the Boruta algorithm. Figure 4C shows the prediction performance (AUC, accuracy, sensitivity, specificity, PPV, and NPV) of the combination of 3 genes (BHLHE22, CDO1, and TBX5) of each machine learning model in MPap 2.0. Figure 4D shows the prediction performance (AUC, accuracy, sensitivity, specificity, PPV, and NPV) of the combination of 3 genes (BHLHE22, CDO1, and TBX5) of each machine learning model in MPap 3.0.

The following embodiments are non-limiting and merely represent various aspects and features of the present invention.

method

Patient collection

Female patients aged 40 years or older with abnormal uterine bleeding were included. These patients were further excluded using the following criteria: (1) a history of gynecological or breast cancer or their cancer treatment; (2) hysterectomy; (3) current pregnancy, postpartum, or breastfeeding status; or (4) atypical squamous/glandular cells of uncertain significance or worse diagnosed in the cervix. Patients with incomplete clinical and pathological results were also excluded from data analysis.

DNA extraction, sulfite conversion, and DNA methylation measurement

Cervical secretions collected by cotton balls were placed in 50 mL centrifuge tubes with inner cups and stored at 4°C. The cotton balls were rinsed with 1 mL of phosphate buffered saline (PBS) and then centrifuged at 1,000 × g for 10 minutes to collect the elution. Genomic DNA was extracted from cervical secretions using the QIAmp DNA Mini Kit (QIAGEN, Hilden, Germany). The sulfite-converted DNA was treated with the EZ DNA Methylation Kit (Zymo Research Corp., Irvine, CA, USA). PCR products were amplified using the LightCycler 480 SYBR Green I Master (Roche, Basel, Switzerland) and the LightCycler 480. The 20μL reaction volume contained 2μL of sulfite-converted DNA, primers (250nmol/L each), and 10μL of Master Mix. Each gene on all samples was tested more than twice. In each independent methylation assay, one or more reference genes were used for relative quantitative calculations. The methylation degree was calculated as follows: dCp=(Cp of the target gene)-(Cp of the reference gene). Commonly used reference genes include GAPDH, β-actin, COL2A1, ACTB, etc. In a specific embodiment, COL2A1 is used as the reference gene.

The smaller the dCp, the higher the degree of methylation. A COL2A1 Cp value > 38 is defined as a detection failure.

Multiplex bisulfite PCR sequencing (MBPS) assay for targeted DNA methylation analysis

Genomic DNA was extracted from cervical secretions and sulfite converted. In order to perform multiple bisulfite PCR sequencing (MBPS) in the specified region of interest, the present invention designs relevant primer sets. In order to ensure that the primers can effectively amplify the bisulfite converted DNA, the PCR conditions are also optimized. Multiple bisulfite PCR is performed on the samples for post-sequencing Method evaluation. Conditions for multiple bisulfite PCR in clinical samples are constructed. After the PCR procedure, the generated amplicon is purified and prepared for sequencing. The library is established using the Illumina library preparation kit. The prepared library is then sequenced using the Illumina platform. The large amount of sequencing data thus obtained was used to accurately align with the reference human genome (GRCh38/hg38).

Statistical analysis and machine learning

The continuous data of patient characteristics and methylation levels were grouped and analyzed using the T test. Statistical significance was set at a two-tailed p value <0.05. The above analysis and drawing were performed using the statistical packages in R (version 3.6.3) and Prism (version 9.5.1; GraphPad software). The present invention uses the R package Boruta to determine the importance of candidate genes. The Tidymodels package was used to test the performance of the combination of candidate gene methylation in 5 machine learning models. The 5 models are logical regression, multilayer perceptron, random forest, support vector machine, and XGBoost. The samples were divided into a training set to establish a machine learning model (5 cross-validations and 5 repetitions), and a test set to verify the effectiveness of the model. MPAp 2.0: training set (non-cancer: 57, cancer: 29) and test set (non-cancer: 20, cancer: 15); MPap 3.0: training set (non-cancer: 39, cancer: 39) and test set (non-cancer: 20, cancer: 15). Predictive performance metrics including area under the curve (AUC), accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated for the best configuration of each model.

result

A total of 234 female patients participated in this study, of whom 98 (41.9%) were histologically diagnosed with endometrial cancer (EC) and 136 (58.1%) had no evidence of cancer (Table 1). Their median age was 52.7 years (interquartile range (IQR) 59.9-47.1). Women with EC were older (median age 54.2 years (IQR 62.0-58.7) vs. 45 years (IQR 53.7-48.3), p<0.0001). Most cases were low-grade (77.1%, grade I/II), early-stage (82.7%, FIGO stage I/II), and endometrioid histological subtype EC (79.6%). The control group was mainly symptomatic non-cancer patients or patients with leiomyoma (56.6%), although in 43.4% of cases, no underlying pathology was found.

Table 1. Demographic characteristics of all clinical samples.

Discovery and validation of candidate genes

The workflow for candidate gene development and evaluation is shown in Figure 1A. Two methylomics datasets of endometrial tissue and cervical mucus were analyzed to select candidate genes that are highly methylated in EC compared to non-EC. A total of 78 genes with 95 differentially methylated positions were found. Benign tumors, such as leiomyoma, show similarities to certain phenotypes of their malignant counterparts, such as uncontrolled growth, induction of angiogenesis, and invasion of surrounding tissues. Further complicating matters, benign tumors share molecular features with malignant tumors, which may hinder accurate differentiation for molecular cancer diagnosis. To improve diagnostic accuracy between these tumor types, the present invention included patients with both benign and malignant tumors. The present invention then used sulfiting targeted sequencing to validate 95 DMPs. These 78 genes were tested using sulfiting targeted sequencing. Discovery analysis identified 18 genes with diagnostic potential (ADAMTS16, ADCY8, C12orf42, CDH7, CYTL1, EPHA10, FGF11, HPSE2, HSAP1L, HTR1B, KCNA4, NRXN1, PAX6, RSPH9, RUAP1L, HTR1B, KCNA4, NRXN1, PAX6, RSPH9, RUSC1, SLITRK1, TBX5, and ZNF135) (Figure 1B). The average methylation status and area under the receiver operating characteristic curve (AUC) of the 18 genes are shown in Figures 1C and 1D.

The first six genes were used to design qMS-PCR primers, and four genes (ADCY8, EPHA10, HPSE2, and TBX5) were selected as potential biomarkers. The combination of these four genes and the other two genes (CDO1 and BHLHE22) was validated using qMS-PCR.

The ability to differentiate between cervicovaginal and uterine samples collected by physicians

Genes CDO1 and BHLHE22 were added to the experiment to validate a 6-gene panel (ADCY8, EPHA10, HPSE2, TBX5, CDO1, and BHLHE22). The invention also evaluated the efficacy of two independent external groups (including physician- and self-collected vaginal samples from 136 patients without EC and 98 patients with EC).

The methylation status of the six candidate genes in 121 samples collected by doctors (including 77 non-EC and 44 EC) is shown in Figure 2A. All of these genes showed significant differences in methylation status between non-EC and EC samples. The methylation status of each gene in patients and controls is shown in a heat map (Figure 2B).

In order to test the clinical performance of these 6 candidate genes, the present invention divided the samples into a training set (57 non-ECs and 29 ECs) and a test set (20 non-ECs and 15 ECs). The present invention uses logical regression and four machine learning models to evaluate the performance of these discriminative candidate genes. In the training set, the AUC of these 6 genes predicting EC was 0.96 in logical regression, 0.958 in multi-layer perceptron, 1.0 in random forest, 0.932 in support vector machine, and 1.0 in XGBoost machine learning model (Figure 2C). In the subsequent test set, these six genes also showed excellent ability to distinguish EC from non-EC: AUC was 0.957 in logical regression, AUC was 0.947 in multi-layer perceptron, AUC was 0.953 in random forest, AUC was 0.943 in support vector machine, and AUC was 0.947 in XGBoost machine learning model (Figure 2D). The sensitivity, specificity, negative predictive value (NPV) and positive predictive value (PPV) of the training set and test set of the present invention in different machine learning algorithms are shown in Figures 2E and 2F and Table 2. Overall, this analysis shows that the six candidate genes combined with the machine learning algorithm improve the model's ability to accurately predict disease groups.

Table 2. Performance of the six gene panels in physician- and self-collected cervicovaginal specimens

Validation of self-collected cervicovaginal specimens

To test the feasibility of DNA methylation assays using self-collected vaginal samples, the present invention analyzed these 6 genes in self-collected vaginal samples in a training set and a test set containing 59 non-EC samples and 54 EC samples. The methylation status and heat map of the 6 candidate genes are shown in Figures 3A and 3B.

The samples were divided into training and test sets to establish the machine learning algorithm. In the training set (non-EC: 39 cases, EC: 29 cases), 6 gene combinations contained important discriminative factors with AUC values of 0.888 (logical regression), 0.873 (multi-layer perceptron), 1.00 (random forest), 0.874 (support vector machine) and 0.997 (XGBoost) (Figure 3C). Therefore, the present invention evaluated 6 gene combinations to distinguish EC in the test set (non-EC: 20 cases, EC: 15 cases); their AUCs reached 0.91 (logical regression), 0.95 (multi-layer perceptron), 0.927 (random forest), 0.917 (support vector machine) and 0.937 (XGBoost) (Figure 3D and Table 2). The sensitivity and specificity of the training set versus the test set in physician-collected samples were 79-100% versus 93-100%, and 79-96% versus 93-100%, respectively. Table 2 shows the ability of the training set versus the test set to distinguish between EC and non-EC in self-collected cervicovaginal samples, with a sensitivity range of 85-100% versus 80-93% and a specificity range of 72-100% versus 75-90% (Figures 3E-3F and Table 2). In summary, these findings suggest that the six genomes have potential for EC detection.

The efficacy of MPap 2.0 (doctor-collected) and MPap 3.0 (self-collected) tests

To improve the clinical feasibility of the system, the present invention attempts to reduce the number of genes required in the test. Using the feature selection function of the Boruta algorithm, 3 gene combinations (TBX5, CDO1, and BHLHE22) produced better predictive biomarkers for detecting EC in physician-collected samples (MPap 2.0) and self-collected samples (MPap 3.0) (Figures 4A and 4B). For each machine learning model, the predictive performance (AUC, accuracy, sensitivity, specificity, PPV, and NPV) of the 3 gene combinations used in MPap 2.0 and MPap 3.0 are shown in Figures 4C and 4D and Table 3. On the training set of MPap 2.0, the present invention found an AUC of 0.95-0.98, a sensitivity of 79-97%, and a specificity of 82-91%. The discriminatory potential of MPap 2.0 in the test set was further analyzed, and the AUC was 0.94-0.97, the sensitivity was 93-100%, and the specificity was 65-85%. In addition, the training set of MPap3.0 predicted EC with an AUC of 0.86-0.96, a sensitivity of 79-92%, and a specificity of 77-87%. The execution prediction ability model (MPap 3.0) of the test set showed an AUC of 0.93-0.96, a sensitivity of 87-93%, and a specificity of 75-95% (Table 3). The 3 gene combinations (TBX5, CDO1, and BHLHE22) had high accuracy in detecting EC.

Table 3. Performance of MPap 2.0 and MPap 3.0

Although the above embodiments disclose specific embodiments of the present invention, they are not intended to limit the present invention. Those with ordinary knowledge in the technical field to which the present invention belongs may make various changes and modifications without departing from the principles and spirit of the present invention. Therefore, the scope of protection of the present invention shall be based on the scope defined in the patent application.

Claims (7)

A method for evaluating whether an individual suffers from endometrial cancer, comprising the following steps: (a) providing a sample obtained from the individual; (b) determining the methylation level of at least one target gene in the sample, wherein the at least one target gene comprises at least one of the following: EPHA10 and HSPE2, and the methylation level refers to the percentage of methylated CpG sites on the at least one target gene; (c) determining whether the at least one target gene is highly methylated, wherein the high methylation refers to the methylation level of the at least one target gene in the sample being at least 10% higher than the methylation level of the at least one target gene in a sample of an individual not suffering from endometrial cancer; and (d) evaluating whether the individual suffers from endometrial cancer based on the result of step (c), wherein the high methylation of the at least one target gene indicates that the individual suffers from endometrial cancer. The method as claimed in claim 1, wherein the at least one target gene is selected from at least one of EPHA10 and HSPE2, and further comprises at least one of ADCY8, BHLHE22, CDO1 and TBX5. A method for evaluating whether an individual suffers from endometrial cancer comprises at least the following steps: (a) providing a sample obtained from the individual; (b) determining the methylation level of a gene combination in the sample, wherein the gene combination comprises BHLHE22, CDO1, TBX5, EPHA10, ADCY8 and HSPE2, and the methylation level refers to the percentage of methylated CpG sites on the gene; (c) determining whether the gene combination is highly methylated, wherein the high methylation means that the methylation level of the gene combination in the sample is at least 10% higher than the methylation level of the gene combination in a sample of an individual who does not suffer from endometrial cancer; and (d) evaluating whether the individual suffers from endometrial cancer based on the result of step (c), wherein the high methylation of the gene combination represents that the individual suffers from endometrial cancer. A method as described in any one of claim 1 or 3, wherein the endometrial cancer comprises type 1 endometrial cancer or type 2 endometrial cancer. A method as described in any one of claim 1 or 3, wherein the sample is from endometrial mucus or tissue, cervical smear cells, uterine or vaginal secretions, or menstrual blood. A method as claimed in any one of claims 1 or 3, wherein the sample is obtained by collecting using a sanitary napkin, a tampon, a sanitary pad, a cotton swab, a cotton swab, a vaginal lavage, a cervical brush or a cotton ball. A method as described in any one of claim 1 or 3, wherein the method for determining the methylation status of the gene in the sample in step (b) comprises methylation-specific polymerase chain reaction, quantitative methylation-specific polymerase chain reaction, sulfite sequencing, sulfite pyrophosphate sequencing, microarray, mass spectrometry, denaturing high-performance liquid chromatography, pyrophosphate sequencing, methylated DNA immunoprecipitation and quantitative polymerase chain reaction, methylated DNA immunoprecipitation sequencing, nanopore sequencing, fiber optic biosensor or optical biosensor.