A screening method and device for methylation diagnostic markers for early diagnosis of colon cancer
Patent Information
- Application Number
- CN202311544663.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-11-17
- Publication Date
- 2026-10-30
- Estimated Expiration
- 2043-11-17
AI Technical Summary
[0007]本发明是要解决现有筛选方法获得的结肠癌分子标志物在结肠癌早期诊断时准确性不足，产生假阴性的问题，提供一种用于结肠癌早期诊断甲基化诊断标志物的筛选方法及装置
[0038]本发明基于结肠癌临床和DNA甲基化数据，通过机器学习和统计学方法挖掘正常以及结肠癌前期与进展期之间的特异DNA甲基化基因标志物，并建立结肠癌分期诊断模型。本发明方法筛选的肠癌分子标志物在结肠癌早期诊断时具有较高的准确性。
Smart Images
Figure CN117524302B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of colorectal cancer diagnostic markers, and more specifically to a method and apparatus for screening methylated diagnostic markers for the early diagnosis of colorectal cancer. Background Technology
[0002] Colorectal cancer is one of the most common malignant tumors and the third leading cause of cancer death. Currently, treatment decisions for colorectal cancer primarily rely on pathological indicators such as TNM staging. The colorectal cancer staging system defines stage I as early-stage, stage II / III as advanced-stage, and stage IV as late-stage. Existing methods for colorectal cancer prevention and treatment are limited. Although many treatments, including surgery, chemotherapy, and antibody therapy, are used to treat this type of cancer, the five-year survival rate remains relatively poor. Late diagnosis, lack of reliable biomarkers and precise treatment targets, as well as the heterogeneity among colorectal cancer patients, all negatively impact patient prognosis and quality of life.
[0003] Previous clinical studies have shown that nearly 90% of patients diagnosed with colorectal cancer at an early stage can survive for 5 to 10 years, while only 12% of patients diagnosed at a late stage survive, highlighting the importance of early diagnosis. Several clinical methods for early detection of colorectal cancer have been developed, including fecal occult blood testing, computed tomography (CT) scans, colonoscopy, and molecular tumor markers. However, studies have demonstrated that these methods have limited value in colorectal cancer detection. For example, fecal immunohistochemistry, which has long been reported, has produced false negatives in cancer diagnosis, and CT scans have low sensitivity for diagnosing T1 or T2 tumors. Therefore, the diagnosis of colorectal cancer requires continuous expansion of new methods and technologies, including the possibility of using blood tumor markers for detection.
[0004] Recent research indicates that cancer is a heterogeneous malignant disease, exhibiting heterogeneity not only in its genetic makeup but also in its epigenetic makeup. Besides genetic variations, disordered and unbalanced epigenetic alterations are also significant factors in cancer development and progression. DNA methylation is one of the most important regulatory mechanisms in epigenetics, closely related to genome stability and gene transcription activity. Under the catalysis of DNA methyltransferases, the methyl group (-CH3) is selectively added to the 5th carbon atom of CpG dinucleotide cytosine in the DNA sequence, generating 5-methylcytosine. Numerous studies have found that DNA methylation participates in biological processes such as embryonic development, imprinting inactivation, and cell differentiation by regulating related genes. Abnormal DNA methylation can induce cancer development and progression by activating oncogenes and inactivating tumor suppressor genes. Approximately 70% of gene promoter regions have a CG content exceeding 50%. The vast majority of cellular reprogramming is related to tissue-specific methylation regulation along CpG islands. Abnormal methylation in gene promoter regions can lead to the silencing of tumor suppressor genes, thereby promoting tumorigenesis and development, and affecting gene expression and related biological functions. Currently, there is increasing evidence linking abnormal methylation to tumorigenesis. Studies on many cancers have demonstrated the ability of DNA methylation to analyze epigenetic heterogeneity in cancer and have also shown that DNA methylation status can be used as a marker for cancer molecular subtypes.
[0005] With the continuous improvement and development of next-generation sequencing technology, whole-genome sequencing of a large number of tumor samples, including DNA methylation, has increased. Numerous studies have identified key methylation genes in colorectal cancer and explored their application value in the disease. Extensive evidence indicates that the survival of colorectal cancer patients is influenced by both environmental and genetic factors. Abnormal DNA methylation plays a crucial role in the development of colorectal cancer; it can affect the expression of related genes, leading to abnormal gene products, dysregulation of related pathways, and ultimately impacting cancer progression.
[0006] Existing methods for screening molecular markers for colorectal cancer based on DNA methylation have insufficient accuracy in the early diagnosis of colorectal cancer and have produced false negatives. Therefore, this research will be of great significance for the early diagnosis of colorectal cancer. Summary of the Invention
[0007] The present invention aims to address the problem that existing screening methods for colorectal cancer molecular markers are not accurate enough in the early diagnosis of colorectal cancer, resulting in false negatives, and provides a screening method and apparatus for methylation diagnostic markers for the early diagnosis of colorectal cancer.
[0008] The present invention provides a method for screening methylation diagnostic biomarkers for early diagnosis of colorectal cancer, comprising the following steps:
[0009] Step 1: Using a database, obtain the DNA methylation profiles located in the promoter region of human methylation 450 microarray colorectal cancer samples. Preprocess the DNA methylation profiles, and then use the R language t.test() function to screen for differentially expressed DNA methylation sites in the two groups of samples, targeting the DNA methylation profiles of colorectal cancer tissue and adjacent normal tissue. Obtain DNA methylation sites with significant differential expression. Map the obtained differentially expressed DNA methylation sites to genes according to the 450 microarray annotation file to obtain differentially expressed DNA methylation genes.
[0010] Step 2: Based on the AJCC staging criteria, the differentially methylated DNA genes obtained in Step 1 are divided into three groups of differentially methylated DNA profiles: normal, early stage, and progressive stage. Then, new DNA methylated DNA profiles for the normal, early stage, and progressive stage are constructed.
[0011] Step 3: The three groups of DNA methylation profiles of the new normal, early stage, and progressive stage were randomly divided into training set and test set in a ratio of 7:3. Then, the differential DNA methylation genes were identified pairwise in the three training sets and three test sets by combining the t test with the Fold Change value, and a total of six groups of differential DNA methylation genes were obtained.
[0012] Six groups of differentially methylated DNA genes were intersected by taking the intersection of the upregulated and downregulated genomes, resulting in three groups of upregulated differentially methylated DNA genes and three groups of downregulated differentially methylated DNA genes. Then, the intersection of the three groups of upregulated differentially methylated DNA genes and the three groups of downregulated differentially methylated DNA genes was taken to obtain the differentially methylated DNA genes in normal, early, and advanced stages of colorectal cancer.
[0013] Step 4: Divide the differentially methylated DNA genes in normal, early-stage, and advanced-stage colorectal cancer obtained in Step 3 into a training set and a test set in a 7:3 ratio. Based on the mean methylation value of each gene in the training set in normal samples and the mean methylation value in advanced-stage samples, group the samples in the training set and the test set. Samples with methylation values greater than the mean of normal samples are classified into the first category, those between the mean of normal samples and the mean of advanced-stage samples are classified into the second category, and those with methylation values less than the mean of advanced-stage samples are classified into the third category. This yields the grouping data for the training set and the test set samples.
[0014] Step 5: Using the training set data grouped in Step 4, a prediction model is built using the random forest algorithm. The feature contribution is measured by the degree of change in the Gini index of the genes. Genes are sorted according to the magnitude of the change in the Gini index. The number of genes input into the model training data is gradually reduced to rebuild the model. The AUC value is used to characterize the model performance. By combining the model's AUC value with the number of genes contained in the model, the genes included in the final diagnostic model are determined, thus obtaining the methylation diagnostic biomarkers for colorectal cancer.
[0015] Furthermore, the sampling method for colon cancer samples in step one is as follows:
[0016] Random sampling was performed on cancer tissue samples, with the number of samples taken each time being the same as the number of adjacent normal tissue samples, and the number of sampling times was 1000. The samples obtained from each sampling were screened for differentially expressed DNA methylation sites, and the intersection of the differentially expressed sites was the final differentially expressed DNA methylation site.
[0017] Furthermore, the conditions for significantly differentially expressed DNA methylation sites in step one are as follows:
[0018] (1) The p-value was corrected using the Benjamini-Hochberg method. A corrected FDR < 0.05 was considered to be statistically significant.
[0019] (2) The mean difference of DNA methylation level β between the two groups of samples > 0.2 is considered to be a significant difference in DNA methylation sites between the samples.
[0020] In step (1), the P-value is obtained through a t-test. The t-test process involves using a t-test to screen for differentially methylated sites based on the DNA methylation site levels of different groups. The formula for the t-test is:
[0021]
[0022]
[0023] in, , is the average value of each gene in the sample, and μ is the average value of genes in another sample. This represents the standard deviation of the sample.
[0024] Furthermore, the method for constructing the three sets of DNA methylation profiles for the normal, early, and progressive phases in step two is as follows:
[0025] (1) Using QDMR software, the entropy value H0 of each differentially methylated DNA gene was calculated based on Shannon information entropy for the three groups of differential DNA methylation profiles in the normal, early and progressive stages;
[0026]
[0027] in, Representing the relative methylation probability, the sum of DNA methylation levels in region r of the sample is considered the total methylation value. The ratio of DNA methylation level within a sample region to the total value is defined as the relative methylation probability. ;
[0028] (2) After obtaining the entropy values of the three groups, DNA methylation is reconstructed by multiplying the entropy with the methylation value of each group, thereby obtaining three new DNA methylation spectra of normal, early and progressive phases.
[0029] Furthermore, the formula for calculating the Fold Change value in step three is as follows:
[0030]
[0031] in, , The values are the mean methylation values of the same gene in groups a and b, respectively. The p-values were corrected using the Benjamini-Hochberg method. A significant difference was considered to be FDR < 0.05 and Fc > 2 or Fc < 0.5. When Fc > 2, the gene was considered upregulated in group a, and when Fc < 0.5, the gene was considered downregulated in group a.
[0032] Furthermore, the Gini index of the gene in step five is calculated using the following formula:
[0033]
[0034] Where C represents the number of categories in the sample set at this time. This indicates the proportion of category c in this node q, with a value between 0 and 1.
[0035] A storage medium storing at least one instruction, which is loaded and executed by a processor to implement a method for screening methylation diagnostic biomarkers for early diagnosis of colorectal cancer.
[0036] An apparatus for screening methylation diagnostic biomarkers for early diagnosis of colorectal cancer, the apparatus comprising a processor and a memory, the memory storing at least one instruction, the at least one instruction being loaded and executed by the processor to implement a method for screening methylation diagnostic biomarkers for early diagnosis of colorectal cancer.
[0037] The beneficial effects of this invention are:
[0038] This invention, based on clinical and DNA methylation data of colorectal cancer, uses machine learning and statistical methods to discover specific DNA methylation gene markers between normal individuals and those in the precancerous and advanced stages of colorectal cancer, and establishes a colorectal cancer staging diagnostic model. The colorectal cancer molecular markers screened by this invention have high accuracy in the early diagnosis of colorectal cancer.
[0039] 1. Because DNA methylation values fluctuate significantly across different samples, the method of this invention reconstructs the methylation profile. The original DNA methylation values vary greatly among the normal, early-stage, and advanced-stage colon cancer groups. Specifically, the DNA methylation value ranges from 0 to 1 in each group, but values ranging from 0.01 to 0.99 are present in each group. Therefore, directly identifying differentially expressed genes across different groups has proven infeasible. The method provided in this invention, which reconstructs the DNA methylation profile, reduces the degree of confusion in DNA methylation values across different groups.
[0040] 2. Due to the significant difference in methylation values between cancer samples and adjacent normal samples, this invention employs random sampling within the cancer samples, with each sample containing the same number of adjacent normal samples, to screen for differentially methylated sites. This prevents errors in calculation results caused by excessive differences in sample size across different groups.
[0041] 3. In this invention, when identifying differentially expressed genes after the construction of the new DNA methylation profile, training and test sets are divided into different groups, and the differentially expressed genes at the intersection are used as the final result for identifying differentially expressed genes in different groups. This makes the acquisition of differentially expressed genes more accurate.
[0042] 4. This invention uses a combination of the gene Gini index of random forest and model AUC to screen biomarker genes, thereby determining the accuracy of biomarker genes. Attached Figure Description
[0043] Figure 1 The AUC values of the colorectal cancer diagnostic biomarkers obtained by the method of this invention in the TCGA colorectal cancer staging random forest model are shown. Detailed Implementation
[0044] Specific Implementation Method 1: This implementation method is used for screening methylation diagnostic biomarkers in the early diagnosis of colorectal cancer, and includes the following steps:
[0045] DNA methylation profiles of HM450K colon cancer samples were downloaded from the TCGA database, and methylation sites located in the promoter region were selected for analysis.
[0046] DNA methylation profiles were preprocessed, and then the R language t.test() function was used to screen for differentially expressed DNA methylation sites in the two groups of samples, targeting the DNA methylation profiles of colon cancer tissue and adjacent normal tissue.
[0047] The steps for preprocessing DNA methylation profiles are as follows:
[0048] (1) The region from 2kb upstream to 0.5kb downstream of the transcription start site is defined as the promoter region, and the DNA methylation sites in the promoter region are used for downstream analysis.
[0049] (2) In the DNA methylation sites of the promoter region, the DNA methylation sites belonging to SNPs, the DNA methylation sites on the X and Y chromosomes, and the DNA methylation sites corresponding to multiple genes are deleted one after another.
[0050] (3) Delete sites where more than 70% of the total samples lack DNA methylation values.
[0051] (4) The missing DNA methylation values are supplemented in the DNA methylation profiles filtered by steps (1), (2) and (3) using the k-nearest neighbor (KNN) method. The impute.knn() function in the 'impute' package is used for the supplementation, with parameters set to k=10 and maxp=3000.
[0052] After the above processing, step (1) first obtains the methylation sites located in the promoter region. DNA methylation in the promoter region is closely related to tumorigenesis, and studies have shown that promoter methylation affects diagnosis and treatment. Step (2) obtains methylation probes because methylation probes vary greatly on sex chromosomes and SNPs, and removing them can prevent interference with the recognition of differentially methylated probes. Step (3) removes sites that lack DNA methylation values in more than 70% of the total samples, thus removing probes from the low-quality methylation chip.
[0053] DNA methylation sites exhibiting significant differential expression should simultaneously meet the following conditions:
[0054] (1) Differential methylation sites were screened by two-tailed t-test for DNA methylation site levels between cancer samples and adjacent normal samples. The Benjamini-Hochberg method was used to correct the p-values, and a corrected FDR (false discovery rate) < 0.05 was considered statistically significant.
[0055] The p-value was obtained using a t-test, where the t-test procedure involves screening for differentially methylated sites by applying a t-test to the DNA methylation site levels of different groups. The formula for the t-test is:
[0056]
[0057]
[0058] in, , is the average value of each gene in the sample, and μ is the average value of genes in another sample. This represents the standard deviation of the sample.
[0059] (2) A mean difference of >0.2 in the DNA methylation level β value between the two groups is considered a significant difference in DNA methylation sites between the samples. The β value is the DNA methylation value.
[0060] Because of the significant difference in methylation values between cancer samples and adjacent normal samples, this invention employs random sampling from cancer samples, with each sampling number equal to the number of adjacent normal samples, to screen for differentially methylated sites. A total of 1000 sampling iterations were performed, and the intersection of these 1000 differentially methylated site results was considered the final differentially methylated sites. The obtained differentially methylated DNA sites were then mapped to genes according to the annotation file of the 450K microarray, thereby obtaining differentially methylated DNA genes and their DNA methylation levels.
[0061] Then, we explored the clinical differences between precancerous and advanced stages of colorectal cancer.
[0062] After identifying differentially methylated DNA genes in adjacent and cancerous colon cancer samples, and combining this with the AJCC staging criteria, three sets of differentially methylated DNA profiles were obtained for normal, early-stage, and advanced stages. DNA methylated genes that differed across all three groups were then identified. Since DNA methylation values fluctuate significantly across different samples, directly calculating the differences has proven infeasible. Therefore, this invention reconstructs the original methylation profiles using the following method:
[0063] (1) The QDMR (quantitative differentially methylated region) software was used to calculate the entropy value of each differentially methylated DNA gene in the normal, early, and progressive stages based on Shannon information entropy. The QDMR algorithm is mainly based on the following formula for calculating entropy:
[0064]
[0065] in, Representing the relative methylation probability, the sum of DNA methylation levels in region r of the sample is considered the total methylation value. The ratio of DNA methylation level within a sample region to the total value is defined as the relative methylation probability. .
[0066] (2) After obtaining the entropy of the three groups of genes, DNA methylation is reconstructed by multiplying the entropy with the methylation value of each group, thereby obtaining new DNA methylation profiles for the normal, early and progressive phases.
[0067] Next, this invention identified differentially methylated DNA genes in normal, early-stage, and advanced-stage colorectal cancer. After obtaining the new DNA methylation profiles for normal, early-stage, and advanced-stage cancer, the samples from these three stages were randomly divided into training and test sets in a 7:3 ratio. Subsequently, differentially methylated DNA genes were identified using a t-test combined with Fold change values. The formula for calculating Fold Change is as follows:
[0068]
[0069] in, , These represent the mean methylation values of the same gene in groups a and b, respectively. The Benjamini-Hochberg method was used to correct for P-values. A statistically significant difference was considered to be FDR < 0.05 and Fc > 2 or Fc < 0.5. An Fc > 2 indicated upregulation of the gene in group a, and an Fc < 0.5 indicated downregulation in group a. Groups a and b refer to different groups; for example, when calculating the normal and early-stage groups, groups a and b correspond to the normal and early-stage groups, respectively; when calculating the early-stage and progressive-stage groups, they correspond to the early-stage and progressive-stage groups, respectively; and when calculating the normal and progressive-stage groups, they correspond to the normal and progressive-stage groups, respectively.
[0070] Therefore, differentially methylated DNA genes were identified pairwise in three training sets and three test sets, resulting in six groups of differentially methylated DNA genes. First, the differentially methylated DNA genes identified in the training and test sets were intersected according to the upregulated and downregulated genomes, yielding three groups of upregulated and three groups of downregulated differentially methylated DNA genes. Next, the intersections of the upregulated and downregulated genomes were taken to obtain the final differentially methylated DNA genes for normal, early-stage, and advanced stages of colorectal cancer.
[0071] Finally, a colorectal cancer diagnostic model was constructed. The novel expression profiles of differentially expressed DNA methylated genes from the three sample groups were divided into training and testing sets in a 7:3 ratio. Based on the mean value of novel methylation for each gene in the training set in normal samples and the mean value of novel methylation in advanced-stage samples, the samples in the training and testing sets were grouped. Samples with values greater than the mean value in normal samples were classified as the first category; those between the mean values in normal samples and advanced-stage samples were classified as the second category; and those with values less than the mean value in advanced-stage samples were classified as the third category. This yielded the data after grouping the training and testing sets.
[0072] The mean value of neomethylation for each gene in the training set in normal samples is calculated by summing the neomethylation values of all normal samples in the training set and dividing by the number of normal samples in the training set. The mean value of neomethylation in progress phase samples is calculated by summing the neomethylation values of progress phase samples in the training set and dividing by the number of progress phase samples in the training set.
[0073] Next, using the previously grouped training set data (i.e., the new DNA methylation values corresponding to the samples in the training set), a prediction model was built using the random forest algorithm. The feature contribution was measured by the degree of change in the Gini index of genes. Genes were sorted according to the magnitude of the Gini value change, and the number of genes input into the model training data was gradually reduced to rebuild the model. The AUC value was used to characterize the model performance. The genes included in the final diagnostic model were determined by combining the model's AUC value with the number of genes included in the model. A final gene set was selected when the AUC value was greater than 0.9 and the model contained fewer than 5 genes. The final model was then used to perform 10-fold cross-validation on the sample data to determine the robustness of the diagnostic model, thus obtaining the methylation diagnostic biomarkers and diagnostic model for colorectal cancer.
[0074] The Gini index of a gene is calculated using the following formula:
[0075]
[0076] Where C represents the number of categories in the sample set at this time. This indicates the proportion of category c in this node q, with a value between 0 and 1.
[0077] Measuring features The importance of node q in the i-th tree (i.e., node q through attributes) The change in the Gini index before and after branching is calculated using the following formula:
[0078]
[0079] in, and This represents the Gini index of the two new nodes generated after this feature branch. Assume the feature... If all nodes appearing in the i-th tree form a set Q, then... The importance of decision tree i is calculated using the following formula:
[0080]
[0081] When there are n trees in a random forest with a total number of decisions, the importance VIM will be calculated by summing the VIM of each of the n trees:
[0082]
[0083] Finally, all calculated importance values are normalized to obtain the final indicator for measuring feature contribution:
[0084]
[0085] The average change in the Gini index (the impurity of node splitting across all decision trees in a random forest) of a given feature is used as the importance metric, VIM. A higher VIM indicates that even a small change in the feature can lead to a significant change in the average change, making this variable more important for correct classification.
[0086] This method successfully mined biomarkers for early diagnosis of colorectal cancer using DNA methylation data. The accuracy of these biomarkers for early diagnosis of colorectal cancer can be verified using DNA methylation data from different stages of colorectal cancer obtained from TCGA.
[0087] The AUC values of the colorectal cancer diagnostic biomarkers obtained by this method in the TCGA colorectal cancer staging random forest model are as follows: Figure 1 As shown, the AUC value exceeds 0.9, proving that the markers identified by this invention are highly accurate. Specific Implementation Method Two:
[0089] This embodiment is a storage medium storing at least one instruction, which is loaded and executed by a processor to implement a method for screening methylation diagnostic biomarkers for early diagnosis of colorectal cancer. Specific implementation method three:
[0091] This embodiment is an apparatus for screening methylation diagnostic biomarkers for early diagnosis of colorectal cancer. The apparatus includes a processor and a memory. The memory stores at least one instruction, which is loaded and executed by the processor to implement a method for screening methylation diagnostic biomarkers for early diagnosis of colorectal cancer.
[0092] This invention may have other embodiments. Without departing from the spirit and essence of this invention, those skilled in the art can make various corresponding changes and modifications according to this invention, but these corresponding changes and modifications should all fall within the protection scope of the appended claims.
Claims
1. A method for screening methylated diagnostic biomarkers for early diagnosis of colorectal cancer, characterized in that, The method includes the following steps: Step 1: Using a database, obtain the DNA methylation profiles located in the promoter region of human methylation 450 microarray colon cancer samples. Then, for the DNA methylation profiles of colon cancer tissue and adjacent normal tissue, use the R language t.test() function to screen for differentially expressed DNA methylation sites in the two groups of samples to obtain DNA methylation sites with significant differential expression. Map the obtained differentially expressed DNA methylation sites to genes according to the 450 microarray annotation file to obtain differentially expressed DNA methylation genes. Step 2: Based on the AJCC staging criteria, the differentially methylated DNA genes obtained in Step 1 are divided into three groups of differentially methylated DNA profiles: normal, early stage, and progressive stage. Then, new DNA methylated DNA profiles for the normal, early stage, and progressive stage are constructed. Step 3: The three groups of DNA methylation profiles of the new normal, early stage, and progressive stage were randomly divided into training set and test set in a ratio of 7:
3. Then, the differential DNA methylation genes were identified pairwise in the three training sets and three test sets by combining the t test with the Fold Change value, and a total of six groups of differential DNA methylation genes were obtained. Six groups of differentially methylated DNA genes were intersected by taking the intersection of the upregulated and downregulated genomes, resulting in three groups of upregulated differentially methylated DNA genes and three groups of downregulated differentially methylated DNA genes. Then, the intersection of the three groups of upregulated differentially methylated DNA genes and the three groups of downregulated differentially methylated DNA genes was taken to obtain the differentially methylated DNA genes in normal, early, and advanced stages of colorectal cancer. Step 4: Divide the differentially methylated DNA genes of colorectal cancer in normal, early and advanced stages obtained in Step 3 into a training set and a test set in a 7:3 ratio; Based on the mean methylation value of each gene in the training set in normal samples and the mean methylation value in advanced stage samples, group the samples in the training set and the test set into three categories: those with a value greater than the mean of normal samples are classified as category 1, those with a value between the mean of normal samples and the mean of advanced stage samples are classified as category 2, and those with a value less than the mean of advanced stage samples are classified as category 3. This gives us the grouping data for the training set and the test set. Step 5: Using the training set data grouped in Step 4, a prediction model is built using the random forest algorithm. The feature contribution is measured by the degree of change in the Gini index of genes. Genes are sorted according to the magnitude of the change in the Gini index. The number of genes input into the model training data is gradually reduced to rebuild the model. The AUC value is used to characterize the model performance. By combining the model's AUC value with the number of genes contained in the model, the gene set of the final diagnostic model is determined, thus obtaining the methylation diagnostic biomarkers for colorectal cancer.
2. The method for screening methylation diagnostic biomarkers for early diagnosis of colorectal cancer according to claim 1, characterized in that, The sampling method for colon cancer samples in step one is as follows: Random sampling was performed on cancer tissue samples, with the number of samples taken each time being the same as the number of adjacent normal tissue samples, and the number of sampling times was 1000. The samples obtained from each sampling were screened for differentially expressed DNA methylation sites, and the intersection of the differentially expressed sites was the final differentially expressed DNA methylation site.
3. The method for screening methylation diagnostic biomarkers for early diagnosis of colorectal cancer according to claim 1 or 2, characterized in that, The conditions for significantly differentially expressed DNA methylation sites in step one are as follows: (1) The p-value was corrected using the Benjamini-Hochberg method. The corrected FDR < 0.05 was considered to be statistically significant. (2) The mean difference of DNA methylation level β between the two groups of samples > 0.2 is considered to be a significantly different DNA methylation site between the samples.
4. The method for screening methylation diagnostic biomarkers for early diagnosis of colorectal cancer according to claim 3, characterized in that, In step (1), the P-value is obtained through a t-test. The t-test process is as follows: Differential methylation sites are screened using a t-test based on the DNA methylation site levels of different groups. The formula for the t-test is: ， ， in, , is the average value of each gene in the sample, and μ is the average value of genes in another sample. This represents the standard deviation of the sample.
5. The method for screening methylation diagnostic biomarkers for early diagnosis of colorectal cancer according to claim 1, characterized in that, The method for constructing the three sets of new DNA methylation profiles for the normal, early, and progressive phases in step two is as follows: (1) Using QDMR software, the entropy value H0 of each differentially methylated DNA gene was calculated based on Shannon information entropy for the three groups of differential DNA methylation profiles in the normal, early and progressive stages; ， in, Representing the relative methylation probability, the sum of DNA methylation levels in region r of the sample is considered the total methylation value. The ratio of DNA methylation level within a sample region to the total value is defined as the relative methylation probability. ; (2) After obtaining the entropy values of the three groups, DNA methylation is reconstructed by multiplying the entropy with the methylation value of each group, thereby obtaining three new DNA methylation spectra of normal, early and progressive phases.
6. The method for screening methylation diagnostic biomarkers for early diagnosis of colorectal cancer according to claim 1, characterized in that, The formula for calculating the Fold Change value in step three is as follows: ， in, , The values are the mean methylation values of the same gene in groups a and b, respectively. The p-values were corrected using the Benjamini-Hochberg method. A significant difference was considered to be FDR < 0.05 and Fc > 2 or Fc < 0.
5. When Fc > 2, the gene was considered upregulated in group a, and when Fc < 0.5, the gene was considered downregulated in group a.
7. The method for screening methylation diagnostic biomarkers for early diagnosis of colorectal cancer according to claim 1, characterized in that, The Gini index of the gene in step five is calculated using the following formula: ， Where C represents the number of categories in the sample set at this time. This indicates the proportion of category c in this node q, with a value between 0 and 1.
8. The method for screening methylation diagnostic biomarkers for early diagnosis of colorectal cancer according to claim 1, characterized in that, When the AUC value is greater than 0.9 and the model contains fewer than 5 genes, it is used as the gene set for the final diagnostic model.
9. A storage medium, characterized in that, The storage medium stores at least one instruction, which is loaded and executed by a processor to implement the screening method for methylation diagnostic biomarkers for early diagnosis of colorectal cancer as described in any one of claims 1 to 8.
10. A device for screening methylation diagnostic biomarkers in the early diagnosis of colorectal cancer, characterized in that, The device includes a processor and a memory, the memory storing at least one instruction, which is loaded and executed by the processor to implement the screening method for methylation diagnostic biomarkers for early diagnosis of colorectal cancer as described in any one of claims 1 to 8.