A gene methylation diagnostic model for differentiated thyroid cancer and its construction method

Abstract

This invention discloses a method for constructing a gene methylation diagnostic model for differentiated thyroid cancer (PTC). The method includes the following steps: S1, obtaining a test sample; S2, extracting and storing DNA from the test sample; S3, performing methylation analysis; S4, establishing a diagnostic model based on DNA methylation, where the model shows that patients with cg03596178, cg06033721, cg06688989, cg07209244, cg07485775, cg14484681, cg19979108, and cg20943461 are more likely to experience disease progression and are considered risk factors; S5, cross-validating and evaluating the performance to obtain the gene methylation diagnostic model. This model, by detecting and statistically analyzing specific methylation sites, can differentiate patients with recurrent or metastatic characteristics in differentiated thyroid cancer. The relevant model can assist in the clinical diagnosis and follow-up of thyroid patients.

Description

Technical Field

[0001] This invention belongs to the field of medical modeling, specifically a gene methylation diagnostic model for differentiated thyroid cancer. Background Technology

[0002] Thyroid cancer is the most common endocrine malignancy, and its incidence has been gradually increasing worldwide in recent years. Thyroid cancer can be classified into several types, including papillary thyroid cancer (PTC), medullary thyroid carcinoma (MTC), and anatomyeloid thyroid cancer (ATC). PTC is the most common type, also known as differentiated thyroid cancer, and has the best prognosis. Anatomyeloid thyroid cancer has the highest malignancy, but because it accounts for less than 2% of all thyroid cancers, it receives the most attention among thyroid cancers.

[0003] Currently, the diagnosis of thyroid cancer mainly relies on ultrasound-guided fine-needle aspiration biopsy (FNAB) to obtain cellular samples from thyroid nodules. The benign or malignant nature of these cells is then determined by microscopic observation of their pathological characteristics. FNAB is an invasive procedure, heavily dependent on the clinician's ultrasound diagnostic skills and the accuracy of the aspiration procedure. Inexperienced physicians may obtain the wrong tissue, leading to missed diagnoses. Therefore, FNAB is not only invasive but also carries a significant risk of misdiagnosis. Furthermore, the morphological characteristics of the samples collected by sonographers via FNAB are then assessed by pathologists under a microscope, further increasing the subjectivity of this diagnostic method. Moreover, approximately 20%–30% of the thyroid nodule cells in pathological smears containing FNAB samples are difficult to determine as benign or malignant using current clinical techniques. These 20% to 30% of patients face the possibility of being missed or misdiagnosed. Every patient's life is precious. For a healthy person, being missed means that the patient may have to pay the price with their life. Misdiagnosing a healthy person as a thyroid cancer patient will greatly increase the patient's psychological pressure and may lead to a series of problems.

[0004] Under current technological conditions, diagnosing thyroid diseases using other components such as urine and blood remains very difficult. Therefore, the diagnosis of thyroid cancer still relies on femtosecond biopsy (FNAB) to obtain samples. However, we can supplement existing diagnostic methods with other objective testing methods, thereby reducing errors caused by subjective factors. In conclusion, there is an urgent clinical need to establish a more accurate and sensitive objective diagnostic method to assist existing methods, enabling the testing of minute tissue samples obtained through biopsy to obtain more objective data, thereby reducing the possibility of misdiagnosis and missed diagnosis and improving patients' quality of life. Summary of the Invention

[0005] In response to the problems existing in the background art, this invention proposes a gene methylation diagnostic model for differentiated thyroid cancer and its construction method.

[0006] Technical solution:

[0007] A gene methylation diagnostic model for differentiated thyroid cancer is constructed through the following steps:

[0008] S1. Obtain test samples and manually grade them;

[0009] S2. Extract and store DNA from the test samples;

[0010] S3. DNA from thyroid cancer cells in the experimental group and thyroid nodules in the normal control group was used to construct an RRBS library and perform methylation analysis. RRBS sequencing was used to determine the methylation level of genes in 23,565 gene promoter regions (transcription start site ± 1K). The gene promoter regions detected were the hg38 version (Homo sapiens genome assembly GRCh38-NCBI-NLM (nih.gov)). The average DNA methylation value of each gene promoter region obtained in the normal group and the average DNA methylation value in the thyroid cancer group were used to calculate the statistical difference between the normal group and the thyroid cancer group using a T-test, and genes with P < 0.05 were retained.

[0011] S4. A diagnostic model based on DNA methylation was established. The model showed that PTC patients with cg03596178, cg06033721, cg06688989, cg07209244, cg07485775, cg14484681, cg19979108, and cg20943461 (gene methylation sequencing is an existing technology; after sequencing, data on the methylation level of all genes in the sample will be available). If it is necessary to specifically detect these sites, probes and methylation panels related to these sites can be designed to specifically detect the status of these sites. Patients with these sites have a higher likelihood of disease progression and are considered risk factors. They should be closely followed up. Each cg site has the same weight. A patient with one site is assigned 1 point, with two sites assigned 2 points, and so on. The higher the score, the higher the risk. Specifically: Y = x1 + x2 + ... + x8, where Y represents the diagnostic grade, x1, x2, ... x8 represent... The table lists eight differentially methylated sites (cg03596178, cg06033721, cg06688989, cg07209244, cg07485775, cg14484681, cg19979108, cg20943461) in thyroid cancer and normal thyroid tissue. If one of these differentially methylated sites is highly expressed in the patient sample, it is denoted as 1, then Y = 1; if two sites are highly expressed, it is denoted as 2, then Y = 2, and so on. Therefore, Y is in the range [0, 8]. Specifically, Y in [0, 1] represents the low-risk group for thyroid cancer; Y in [2, 3] represents the intermediate-risk group for thyroid cancer; and Y in [4, 8] represents the high-risk group for thyroid cancer.

[0012] S5. Download the dataset (GSE53051) containing methylation sequencing data of tumor samples and normal samples from the GEO database. Select the methylation data corresponding to thyroid tumor samples from GSE53051, and after the same preprocessing procedure as the original data from the TCGA database, obtain the filtered and normalized methylation data from GSE53051 for subsequent analysis. Select PTC and normal samples of thyroid tumors from the database to verify the methylation levels (box plots) and ROC analysis of the methylation sites included in the diagnostic model obtained in S4 in the GSE53051 data; then recalculate using the linear model obtained from SVM, and verify the performance of the diagnostic model through ROC analysis.

[0013] Preferably, in S1, a fine-needle aspiration biopsy (FNAC) under ultrasound guidance is used to obtain the test sample.

[0014] Preferably, in S2, the extracted DNA concentration is 30 ng / ul, the purity is OD260 / 280≥1.8, the volume is generally 30ul, and it is stored in a test tube at -80℃.

[0015] Preferably, in S3, the RRBS library is sequenced on Genome Analyzer II based on the established single-end sequencing operation steps; the raw sequencing data is filtered and evaluated; methylation-related information of cytosine is obtained, including coverage analysis, methylation analysis and DMRs analysis; the amount of methylated cytosine with a sequence depth coverage of at least 10 and covered by at least 4 reads is selected to determine the level of gene methylation.

[0016] Preferably, in S4, the specific steps are as follows:

[0017] S4-1. The raw data was processed using the `impute.knn` function of the `impute` package in R. After imputing the data and performing a dimensionality test, the differentially methylated genes between PTC and NC were calculated using the `limma` package in R. The screening threshold for differentially methylated genes (DMGs) was |logFC|>1&p.value<0.05. The differentially methylated probes obtained from these samples and the normal control were labeled as DMP-1, and the results of the differential analysis were presented as a volcano plot.

[0018] S4-2. Thyroid cancer-related data (methylation data, clinical data, and survival data) were downloaded from the TCGA database from UCSC Xena (https: / / xenabrowser.net / datapages / ). Methylation data from PTC samples with clinical information were retained. After selection, data from 562 samples were included in the analysis, including 498 tumor samples and 64 normal samples. The methylation data stored in TCGA represents the beta values ​​of methylation sites (probes), while RRBS sequencing yields gene methylation values. Therefore, different differential analysis methods and screening thresholds were used for different data. Missing values ​​were then filled using the `impute.knn` function of the `impute` package in R, and probes were filtered using the `ChAMP` package. The filtered data was further normalized, and the final methylation matrix was used for subsequent analysis. Differentially methylated probes (DMPs) analysis was performed using the ChAMP package. The threshold for significant DMPs was |logFC|>0.25 & adj.P.value<10⁻¹⁵. ChAMP not only performs differential analysis on DMPs but also annotates them to identify the genes corresponding to the methylation sites. Therefore, after identifying DMPs with significant differences between TumorVS.Control groups, the DMPs derived from TCGA data were labeled DMP-2. DMP-1 and DMP-2 were then superimposed to obtain DMP-3.

[0019] S4-3. Based on TCGA data, ROC analysis was performed on the methylation sites corresponding to DMG-3 to verify the classification power (diagnostic power) of differential DMPs on PTC samples and normal samples.

[0020] S4-4. Select DMPs with AUC>0.85 to construct diagnostic models, and use box plots in TCGA data to view the methylation status of these sites in PTC and Normal.

[0021] The DMP model with AUC > 0.85 was further trained using SVM (Support Vector Machine), and the performance of the diagnostic model was further evaluated using ROC analysis.

[0022] S4-5. Based on this, construct a diagnostic model.

[0023] Preferably, the constructed diagnostic model is trained using the linear method of the Biometric Research Branch (BDR) array tool v.4.4.0 from the National Cancer Institute.

[0024] Preferably, the linear methods include Support Vector Machine (SVM), diagonal linear discriminant analysis (DLDA), and compound covariate predictor analysis.

[0025] Preferably, in S5, performance is evaluated using leave-one-out cross-validation (LOOCV).

[0026] Preferably, in S4, the follow-up time is set according to the score: 1 point is followed up once a year, 2 points are followed up every six months, 3 points and 4 points are followed up every three months, 5 points and 6 points are followed up once every two months, 7 points are followed up once a month, and 8 points are followed up once every half month.

[0027] Preferably, ultrasound is used for each follow-up visit, as ultrasound does not involve radiation; the next follow-up and treatment measures are then determined based on whether the thyroid nodules have progressed.

[0028] Beneficial effects of the present invention

[0029] This invention targets thyroid cancer by detecting specific methylation sites of genes in ultrasound-guided biopsy samples and calculating the likelihood of a patient developing thyroid cancer, thus aiding in the diagnosis of differentiated thyroid carcinoma. This diagnostic model utilizes the gold standard of methylation sequencing, pyrosequencing, requiring a small sample size, offering high accuracy, and employing a simple method. It typically utilizes about half the samples obtained from ultrasound-guided biopsy in clinical practice, does not affect the pathological smear of the biopsy sample, eliminates the need for separate sampling, and causes minimal trauma to the patient. Therefore, this invention possesses multiple advantages, including being non-invasive, highly accurate, and objective. Applying this invention to assist in the clinical diagnosis and follow-up of thyroid patients can increase the objectivity and accuracy of diagnosis, demonstrating significant clinical application value.

[0030] This invention utilizes ultrasound-guided fine-needle aspiration biopsy to sample thyroid cancer tissue, followed by pyrosequencing. The results show that abnormal gene methylation is prevalent in thyroid cancer tissue compared to normal thyroid tissue, indicating a close association between abnormal gene methylation and thyroid cancer. This invention combines scientific research findings with practical clinical needs, leveraging advancements in science and technology to apply advanced pyrosequencing technology to the clinical diagnosis of thyroid cancer without causing additional trauma to patients. It offers numerous advantages, including minimal invasiveness, low cost, and high value. Future pathology reports will integrate morphological diagnosis and molecular information. The combined testing of molecular diagnostics and traditional pathology brings immense hope to patients and signifies more precise diagnosis and treatment for doctors, representing the future direction of medical research and clinical application. Attached Figure Description

[0031] Figure 1 This is a schematic diagram of the application process of the diagnostic model.

[0032] Figure 2 The bubble chart for enrichment analysis of PTC vs Normal shows a schematic diagram of the sequencing results.

[0033] Figure 3 A schematic diagram of thyroid-specific genes obtained from the analysis of DMG based on PaGenBase in the Metascape database.

[0034] Figure 4 A graph showing the difference between FNAC samples from thyroid cancer and normal tissue.

[0035] Figure 5 This is a graph showing the difference between thyroid cancer samples and normal controls in the TCGA database.

[0036] Figure 6 This is a schematic diagram of the intersection of DMG-1 and DMG-2.

[0037] Figure 7 The results of methylation level and ROC diagram for cg03596178, one of the components of the diagnostic model.

[0038] Figure 8 The results of methylation level and ROC diagram for cg06033721, the second component of the diagnostic model.

[0039] Figure 9 This is a schematic diagram of ROC results where the AUC exceeds that of a single DMP.

[0040] Figure 10 The figure shows the results of validating the diagnostic efficacy of the diagnostic model on an independent dataset from GEO.

[0041] Figure 11 The figure shows the results of validating the diagnostic efficacy of the diagnostic model on an independent dataset from GEO.

[0042] Figure 12 The figure shows the results of validating the overall diagnostic efficacy of the diagnostic model on the GEO dataset. Detailed Implementation

[0043] The present invention will be further described below with reference to embodiments, but the scope of protection of the present invention is not limited thereto.

[0044] In this article, the term "pyrosequencing" refers to an enzyme cascade chemiluminescent reaction catalyzed by four enzymes: DNA polymerase, ATP sulfurytase, luciferase, and apyrase.

[0045] Combination Figure 1 The entire diagnostic process mainly includes ultrasound-guided thyroid nodule biopsy, pyrosequencing, and comparison and processing of sequencing data, as detailed in the technical solution.

[0046] Example 1

[0047] 1. Experimental Methods

[0048] (1) A large number of FNAB samples were collected by senior clinicians in the ultrasound department of the hospital. DNA was extracted and the samples were frozen at -80°C for further follow-up and testing. The corresponding patients were followed up for many years to obtain and organize relevant clinical data. FNAB samples that were diagnosed as papillary thyroid carcinoma after surgery during the later follow-up and normal thyroid nodule samples that had not been found to have tumors during many years of follow-up were selected. At the same time, the concentration of DNA samples should reach 30ng / ul, the purity OD260 / 280 ≥ 1.8, and the volume should be 30ul or more to ensure the quality of testing and the accuracy of results.

[0049] (2) The samples selected through the above steps were subjected to methylation sequencing using the current gold standard for methylation sequencing, namely reduced representation bisulfite sequencing (RRBS). Strict sample selection requirements were met, and highly accurate methylation quantification analysis was performed using the Qiagen Q48 pyrosequencing system. The latest human genome version, hg38 (Homo sapiens genome assembly GRCh38-), was used for sequencing.

[0050] The degree of gene methylation in the promoter region (transcription start site ± 1K) of genes was detected using NCBI-NLM (nih.gov). The sequencing data were accurate and reliable. The screening threshold for differentially methylated genes was |logFC|>1&p.value<0.05. The DMGs obtained from the sequencing of experimental samples were labeled as DMG-1.

[0051] (3) Thyroid cancer-related data were downloaded from the TCGA database from UCSC Xena (https: / / xenabrowser.net / datapages / ). A total of 562 samples were included in the analysis, including 498 tumor samples and 64 normal samples.

[0052] For methylation data stored in TCGA, probes were filtered using the ChAMP package. The filtered data was then normalized to obtain the final methylation matrix. Differentially methylated sites were then analyzed, with significant DMPs selected based on |logFC|>0.25 & adj.P.value<10⁻¹⁵. Finally, while performing differential analysis on the DMPs using the ChAMP package, the DMPs were annotated to obtain the gene corresponding to each methylation site. DMG obtained from tumor samples in the TCGA database was labeled as DMG-2.

[0053] (4) DMG-1 and DMG-2 were compared, and the intersection of the DMG sets was labeled as DMG-3. The methylation sites corresponding to DMG-3 were analyzed using ROC analysis based on TCGA data to verify the classification power of differentially expressed DMPs on PTC samples and normal samples, i.e., the ability to accurately distinguish thyroid cancer from the samples (diagnostic power); those with AUC >

[0054] A diagnostic model was constructed using DMP of 0.85, and box plots were used in TCGA data to examine the methylation status of these sites in thyroid cancer and normal groups.

[0055] (5) The DMP with AUC>0.85 was further trained using SVM support vector machine, and the performance of the diagnostic model was further evaluated by ROC analysis.

[0056] (6) Download the dataset (GSE53051) from the GEO database, which contains methylation data of samples with thyroid tumors and methylation sequencing data of normal samples. Filter and normalize the dataset to obtain methylation data for validation of the diagnostic model. Specifically, perform methylation level (box plot) and ROC analysis on the methylation sites included in the diagnostic model obtained in the previous step using the GSE53051 data; then recalculate using the linear model obtained from SVM, and validate the performance of the diagnostic model through ROC analysis.

[0057] 2. Experimental Results

[0058] The level of DNA methylation in FNAB samples is associated with a variety of thyroid diseases, including thyroid cancer. Figure 2 Differential methylation sites are closely related to many thyroid functions. Figure 3 This study showed an association between differentially methylated sites and thyroid-related diseases, and that excessively low methylation levels in specific genes may be associated with thyroid cancer. Figure 4 In this study, differential analysis between FNAC samples from thyroid cancer and normal tissues revealed differences in gene methylation levels across the entire genome; the figure shows that gene methylation levels in the tumor group were lower than those in the normal control group. Furthermore, analysis of a large number of thyroid cancer and normal control samples in the TCGA database also showed that abnormal methylation of specific genes is closely related to the occurrence and development of thyroid cancer. Figure 5 In the study, the differential methylation levels of thyroid cancer samples and normal controls in the TCGA database were consistent with those of the sequencing samples, showing a decrease in methylation levels in the tumor group. Therefore, we used differentially methylated sites (DMP-1) in the sample data and differentially methylated sites (DMP-2) in the database as the basis for distinguishing between thyroid cancer and normal thyroid tissue. Figure 6 In the study, DMG-1 had 749 DMGs, and DMG-2 had 730 DMGs. The intersection of these two yielded 20 DMGs, which are the main DMGs capable of accurately distinguishing tumors from normal tissues. By comparing the sequencing results of experimental clinical data with data from the TCGA database and taking the intersection, this invention identified eight gene methylation sites that can distinguish thyroid cancer from normal thyroid tissue: cg07209244, cg20943461, cg07485775, cg03596178, cg14484681, cg19979108, cg06033721, and cg06688989. These eight differentially methylated sites demonstrate good diagnostic efficacy in individually distinguishing thyroid cancer from normal tissues. Figure 7 In the process of identifying DMPs using DMG, a total of 8 DMPs were ultimately used to construct the diagnostic model after screening based on other conditions. Due to space limitations, the methylation and ROC curves of all DMPs are not listed here. The figure shows the methylation level results and ROC diagram of cg03596178, one of the components of the diagnostic model. The methylation level box plot on the left shows that the methylation level of this DMP has significant differences between the tumor group and the normal group, and the ROC curve on the right shows its good diagnostic efficacy. Figure 8The image shows the methylation level results and ROC curve for cg06033721, the second component of the diagnostic model. Similarly, the left-hand box plot of methylation levels shows a significant difference in DMP methylation levels between the tumor and normal groups, while the right-hand ROC curve demonstrates its good diagnostic efficacy. The efficacy of combining all eight differentially methylated sites is even better. Figure 9 In the data, the AUC exceeded the ROC result of a single DMP, indicating that the diagnostic model composed of these 8 DMPs performed better in grouping tumors and normal tissues compared to a single DMP, further demonstrating the diagnostic efficacy of the model. In the GEO independent datasets on thyroid cancer and normal thyroid tissue, the differentially methylated sites in the diagnostic model also demonstrated good diagnostic power in distinguishing between thyroid cancer tissues and normal tissues. Figure 10 The diagnostic efficacy of the diagnostic model was validated on an independent dataset from GEO. The methylation level results (left) and ROC curve (right) of cg03596178 showed that it has the diagnostic efficacy to distinguish between tumor tissue and normal tissue, which is the same as the results obtained in TCGA data. Figure 11 The diagnostic efficacy of the diagnostic model was validated on an independent dataset from GEO. The methylation level results (left) and ROC curve (right) of cg06033721 indicate that it has diagnostic efficacy in distinguishing between tumor and normal tissues (same as the results obtained in TCGA data). Furthermore, the overall disputed efficacy AUC of the diagnostic model reached 1, demonstrating good diagnostic power. Figure 12 The overall diagnostic efficacy of the diagnostic model was validated on the GEO dataset, which has a larger data scale than the TCGA dataset. The SVM model test showed that the diagnostic model perfectly distinguished thyroid cancer from the large amount of data and diagnosed thyroid cancer.

[0059] Example 2

[0060] Based on the above research results, this invention uses machine learning algorithms in bioanalysis to provide a relatively objective and cost-effective diagnostic method for thyroid cancer by calculating the numerical values ​​of differential gene methylation sites. It can be used as an auxiliary diagnostic tool for differentiated thyroid cancer. Specifically, a diagnostic model for thyroid cancer based on ultrasound-guided fine-needle aspiration biopsy and pyrosequencing is proposed, with the following formula: Y = x1 + x2 + ... + x8, where Y represents the diagnostic grade, and x1, x2, ... x8 represent eight differentially methylated sites (cg03596178, cg06033721, cg06688989, cg07209244, cg07485775, cg14484681, cg19979108, cg20943461) in thyroid cancer and normal thyroid tissue. When one of these differentially methylated sites is highly expressed in the patient sample, it is denoted as 1, then Y = 1; when two sites are highly expressed, it is denoted as 2, then Y = 2, and so on. Therefore, Y is in the range [0, 8]. Among them, Y in [0, 1] represents the low-risk group for thyroid cancer; Y in [2, 3] represents the medium-risk group for thyroid cancer; and Y in [4, 8] represents the high-risk group for thyroid cancer.

[0061] Example 3

[0062] Example of thyroid cancer detection

[0063] Following the method described in Example 1, the patient's gene methylation data were obtained. The risk of developing thyroid cancer in the patient was calculated using the formula Y = x1 + x2 + ... + x8, as shown in the following example:

[0064] In Sample 1 (a normal sample from which no thyroid cancer was found during follow-up): x1 = 0; x2 = 0; x3 = 0; x4 = 0; x5 = 0; x6 = 0; x7 = 0; x8 = 0, therefore Y = x1 + x2 + ... + x8 = 0. The diagnostic model Y value for patients in Sample 1 is 0, which belongs to the interval [0, 1], indicating a low-risk prevalence group, meaning that the probability of these patients developing thyroid cancer is low, consistent with the results of later follow-up.

[0065] In Sample 2 (a thyroid nodule biopsy sample from a patient diagnosed with thyroid cancer before the onset of the disease): x1 = 0; x2 = 0; x3 = 1; x4 = 1; x5 = 0; x6 = 0; x7 = 0; x8 = 0, therefore Y = x1 + x2 + ... + x8 = 2. The diagnostic model Y value for the patients in Sample 2 is 2, which falls within the interval [2, 3], indicating a medium-risk prevalence group. This means the patients have a certain risk of developing thyroid cancer, and the result is accurate.

[0066] Table 1

[0067]

[0068]

[0069] The specific embodiments described herein are merely illustrative of the spirit of the invention. Those skilled in the art to which this invention pertains may make various modifications or additions to the described specific embodiments or use similar methods to substitute them, without departing from the spirit of the invention or exceeding the scope defined by the appended claims.

Claims

1. A method for constructing a gene methylation diagnostic model for differentiated thyroid cancer based on ultrasound-guided fine-needle aspiration biopsy and pyrosequencing, characterized in that... It includes the following steps: S1. Obtain test samples and manually grade them; S2. Extract and store DNA from the test samples; S3. DNA containing PTC and normal thyroid nodules was used for RRBS library construction and methylation analysis was performed. S4. A diagnostic model was established based on DNA methylation. The model showed that patients with cg03596178, cg06033721, cg06688989, cg07209244, cg07485775, cg14484681, cg19979108, and cg20943461 in PTC patients were more likely to have disease progression and were considered risk factors, requiring close follow-up. Each DMP had the same weight. If the values ​​of these DMPs were assigned x1, x2, ..., x8 in sequence, then the diagnostic model value of the patient would be Y = x1 + x2 + ... + x8. If the patient's test results showed high expression of one DMP in the diagnostic model, Y would be 1; if the patient had two DMPs in the diagnostic model, Y would be 2, and so on. The higher the Y score of the patient, the higher the risk. S5. Cross-validation is used to evaluate performance and obtain a gene methylation diagnostic model.

2. The method according to claim 1, characterized in that... In S1, a fine-needle aspiration biopsy (FNAC) under ultrasound guidance was used to obtain the test sample.

3. The method according to claim 1, characterized in that... In S2, the extracted DNA was stored in a test tube at -80°C.

4. The method according to claim 1, characterized in that... In S3, the RRBS library was sequenced on Genome Analyzer II based on the established single-end sequencing procedure; the raw sequencing data was filtered and evaluated; methylation-related information of cytosine was obtained, including coverage analysis, methylation analysis, and DMRs analysis; the amount of methylated cytosine with a sequence depth coverage of at least 10 and covered by at least four reads was used to determine the level of gene methylation.

5. The method according to claim 1, characterized in that... In S4, the specific steps are as follows: S4-1, The differential methylation probe between the sample and the normal control was labeled as DMP-1; S4-2. Obtain methylation data from the TCGA Cancer Genome Atlas database, including clinical features of thyroid cancer, including 498 thyroid cancer cases and 64 normal controls; label the DMPs derived from the TCGA data as DMP-2. DMP-1 and DMP-2 are superimposed to obtain DMP-3; S4-3. Receiver operating characteristic (ROC) curve analysis was performed on the DMP-3 data to identify DMPs with diagnostic value that could distinguish PTC from normal controls. S4-4. Gene probes with high area under the ROC curve (AUC) are further analyzed and filtered. S4-5. Based on this, construct a diagnostic model.

6. The method according to claim 5, characterized in that... The constructed diagnostic model was trained using the linear method of the Array Tool v. 4.4.0 from the National Cancer Institute's Center for Biological Research.

7. The method according to claim 6, characterized in that... The linear methods include Support Vector Machine (SVM), Diagonal Linear Discriminant Analysis (DLDA), and Composite Covariate Predictor Analysis.

8. The method according to claim 1, characterized in that... In S5, performance is evaluated using leave-one-out cross-validation (LOOCV).

9. A gene methylation diagnostic model for differentiated thyroid cancer, characterized in that... The model is constructed by the method described in any one of claims 1-8.

Images