A methylation marker combination for early pancreatic cancer diagnosis, product and application thereof

Abstract

The application discloses a methylation marker combination for early pancreatic cancer diagnosis, a product and application thereof, and relates to the technical field of early pancreatic cancer detection. A reagent for detecting the methylation marker combination is used for detecting the methylation level of the methylation marker combination, and the methylation marker combination comprises markers 1-15. The combination of the 15 markers has extremely high sensitivity and specificity for early pancreatic cancer diagnosis, and supports the detection of the methylation level of various sample types. The methylation marker combination provided by the application can quickly identify the risk of early pancreatic cancer, so that medical related examination can be prepared in advance, and early prevention can be realized.

Description

Technical Field

[0001] This invention relates to the field of early pancreatic cancer detection technology, and more specifically, to a combination of methylation markers for the diagnosis of early pancreatic cancer, products therein, and their applications. Background Technology

[0002] Pancreatic cancer is one of the most malignant digestive system tumors, characterized by insidious onset, rapid progression, and lack of specific early symptoms. Most patients are diagnosed at a locally advanced stage or with distant metastasis, resulting in a low overall five-year survival rate. Pancreatic cancer, especially pancreatic ductal adenocarcinoma, has become one of the deadliest malignant tumors in the digestive system due to its insidious onset, lack of specific early clinical symptoms, and limited detection rate on imaging. Since most patients are diagnosed at a locally advanced stage or with distant metastasis, they are no longer eligible for radical surgery. Therefore, establishing non-invasive detection technologies suitable for early screening and diagnosis is of significant clinical importance. Traditional serological markers such as CA19-9, while widely used in the auxiliary diagnosis of pancreatic cancer, have limited sensitivity and specificity in early-stage pancreatic cancer and are easily interfered with by benign diseases such as biliary obstruction and pancreatitis. In recent years, liquid biopsy technologies based on circulating cell-free DNA or circulating tumor DNA have developed rapidly. Among these, DNA methylation detection, which can reflect early epigenetic abnormalities in tumor development, is considered one of the important directions for early pancreatic cancer screening. Previous studies have shown that pancreatic cancer cells exhibit widespread abnormal methylation of CpG islands in the early stages of tumorigenesis. This includes hypermethylation of promoter regions leading to silencing of tumor suppressor genes, and hypomethylation of some gene regions causing genomic instability. Utilizing the methylation characteristics of peripheral blood cfDNA, non-invasive detection of abnormal signals related to pancreatic cancer can be achieved.

[0003] Currently, the main technical approaches for detecting early pancreatic cancer using blood methylation include methylation-specific PCR, real-time quantitative methylation PCR, digital PCR, targeted methylation next-generation sequencing, and whole-genome methylation sequencing. Among these, MSP and qMSP primarily target CpG sites of specific candidate genes, with common targets including ADAMTS1, BNC1, BMP3, NPTX2, SPARC, TFPI2, SFRP1, CD1D, and PENK. With the development of high-throughput sequencing technology, multi-site joint methylation models are gradually becoming the mainstream approach. By jointly analyzing multiple differentially methylated regions and combining them with machine learning algorithms to construct classification models, detection performance can be further improved. However, current technologies still have several shortcomings, such as: a lack of unified standards for methylation targets across different studies; abnormal methylation of some targets also occurring in benign diseases such as chronic pancreatitis; low early ctDNA levels leading to insufficient detection sensitivity; and the high cost and complex data analysis associated with high-throughput methylation sequencing. Therefore, developing blood methylation biomarkers and detection methods with higher specificity and sensitivity, suitable for early pancreatic cancer screening, remains an important research direction in this field.

[0004] DNA methylation abnormalities typically occur in the early stages of cancer and persist throughout the entire process of cancer development and progression. Once formed, the methylation state requires prolonged and continuous stimulation from the external environment to change. Therefore, DNA methylation serves as an important biomarker for early screening, auxiliary detection, and prognosis. Currently, common first-generation methylation detection methods using Q-PCR technology only provide a close-up, localized view of methylation, limited to individual genes and individual samples, offering only qualitative detection, and the sensitivity and specificity of detection need improvement. In contrast, next-generation sequencing (NGS) provides a panoramic, high-resolution view of methylation, enabling quantitative detection of methylation rates at specific methylation sites in specific genes. Moreover, it can simultaneously quantify multiple methylation sites across a large number of genes, providing more accurate results. Furthermore, NGS protocols can simultaneously perform large-scale parallel testing of hundreds or thousands of samples, significantly reducing experimental costs and testing cycles, making it an important early tumor methylation detection method now and in the future.

[0005] In view of this, the present invention is proposed. Summary of the Invention

[0006] The purpose of this invention is to provide a combination of methylation markers, products, and applications for the early diagnosis of pancreatic cancer, which can identify the risk of early pancreatic cancer with high sensitivity, high specificity, and rapidness, so as to conduct relevant medical examinations in advance and prevent it at an early stage.

[0007] This invention is implemented as follows: In a first aspect, the present invention provides the application of a reagent for detecting a combination of methylation markers in the preparation of diagnostic or prognostic products for early pancreatic cancer. The reagent is used to detect the methylation level of the combination of methylation markers, which includes markers 1-15. Using GRch37 as a reference genome, the positions of markers 1-15 on the chromosome are shown below: Marker 1: chr6:26271360-26271510, Marker 2: chr7:27135681-27135820, Marker 3: chr1:63789353-63789499, Marker 4: chr19:13617039-13617183, Marker 5: chr19:30016105-30016253, Marker 6: chr11:112833649-112833795, Marker 7: chr16:51185037-51185189, Marker 8: chr8:4116697 0-41167125, Marker 9: chr8:41166905-41167047, Marker 10: chr1:50886756-50886899, Marker 11: chr16:23847451-23847596, Marker 12: chr20:34189660-34189812, Marker 13: chr16:23847268-23847409, Marker 14: chr1:156611869-156612022 and Marker 15: chr20:34188981-34189119.

[0008] Secondly, the present invention also provides a primer combination for early diagnosis or prognosis of pancreatic cancer, comprising primers for detecting markers 1-15, wherein the primers include: a first primer pair-a fifteenth primer pair for detecting markers 1-15 respectively, and the nucleotide sequences of the first primer pair-a fifteenth primer pair are shown in SEQ ID NO.1-SEQ ID NO.30 respectively.

[0009] Thirdly, the present invention also provides a kit, biochip, or detection device for the diagnosis or prognosis of early pancreatic cancer, comprising: a primer combination for the diagnosis or prognosis of early pancreatic cancer.

[0010] Fourthly, the present invention also provides a method for early diagnosis or prognosis of pancreatic cancer, wherein all steps of the method are implemented by a computer, and the method includes the following steps: Obtain the methylation rate of each region in the combination of methylation biomarkers in the sample to be tested; obtain the early pancreatic cancer diagnosis result or prognosis result based on the early pancreatic cancer diagnosis model or prognosis model.

[0011] Fifthly, the present invention also provides a computer-readable storage medium storing an executable program instruction set, which is loaded by a processor and executed by the aforementioned method for early pancreatic cancer diagnosis or prognosis.

[0012] In a sixth aspect, the present invention also provides an apparatus for early diagnosis or prognosis of pancreatic cancer, the apparatus comprising a processor and a memory, the memory storing a set of executable program instructions, the set of executable program instructions being loaded and executed by the processor to realize the above-described method for early diagnosis or prognosis of pancreatic cancer.

[0013] The present invention has the following beneficial effects: This invention identified multiple methylation sites in 15 DNA regions through screening. The methylation rate of these 15 DNA regions is highly correlated with pancreatic cancer. Systematic validation was performed on leukocytes, tissue samples, and plasma samples. Combined with a machine learning model, a method for early diagnosis or prognosis of pancreatic cancer was obtained, exhibiting extremely high diagnostic sensitivity and specificity. Compared to traditional imaging, endoscopy, or serum biomarkers (such as CEA and SCC), this invention is more suitable for early screening, avoiding invasiveness and low sensitivity issues.

[0014] The methylation levels of each marker can be rapidly detected. When the sample is selected from plasma, whole blood, or other types, non-invasive early signal detection can be achieved, making it suitable for large-scale screening.

[0015] The biomarkers for early pancreatic cancer diagnosis provided by this invention support multiple sample types (such as tissue, urine, blood, and cerebrospinal fluid) and can be extended to prognostic assessment, helping patients to intervene early and improve survival rates.

[0016] The early pancreatic cancer diagnosis or prognosis method provided by this invention can use standard equipment and reagent kits, is low in cost, and has standardized operation, making it suitable for clinical application. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 The ROC curve of the training set in Model 1; Figure 2 The ROC curve of the training set in Model 2; Figure 3The ROC curve of the training set in Model 3; Figure 4 The ROC curve of the validation set in Model 1; Figure 5 The ROC curve of the validation set in Model 2; Figure 6 The ROC curve of the validation set in Model 3. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below. Where specific conditions are not specified in the embodiments, conventional conditions or conditions recommended by the manufacturer shall apply. Reagents or instruments whose manufacturers are not specified are all conventional products that can be purchased commercially.

[0020] The term “subject” as used in this article can be understood as anyone involved in early screening, pancreatic cancer screening, or pancreatic cancer prognostic assessment.

[0021] The term "threshold" refers to a threshold used in diagnosing the risk of early-stage pancreatic cancer. It compares the methylation rate of a combination of biomarkers in a subject's sample to a specific threshold, and outputs the subject's risk or pancreatic cancer prognosis based on the comparison result. Specifically, the threshold includes, but is not limited to, a positive cutoff value—a critical value used to delineate results and define the boundaries of judgment. The threshold must be set based on the performance requirements of the target detection scenario (such as sensitivity, specificity, and accuracy), and determined through clinical data validation, statistical model analysis (such as ROC curve analysis and Youden's index calculation), or industry standard calibration to ensure its validity and reliability in the detection method or diagnostic model.

[0022] Methylation rate at each site = number of methylated reads / (number of methylated reads + number of unmethylated reads).

[0023] The average methylation rate of each marker region is calculated based on its site.

[0024] The term "early-stage pancreatic cancer" includes stage I and stage II pancreatic cancer. Specifically, it includes stages IA, IB, IIA, and IIB.

[0025] Pancreatic cancer includes exocrine tumors and neuroendocrine tumors. Exocrine tumors include pancreatic ductal adenocarcinoma, adenosquamous carcinoma, colloid carcinoma, hepatoid carcinoma, medullary carcinoma, signet ring cell carcinoma, undifferentiated carcinoma, acinar cell carcinoma, intraductal papillary mucinous tumor (IPMN) with invasive carcinoma, etc.

[0026] Neuroendocrine tumors include types such as insulinoma, gastrinoma, and glucagonoma.

[0027] In a first aspect, the present invention provides the application of a reagent for detecting a combination of methylation markers in the preparation of diagnostic or prognostic products for early pancreatic cancer. The reagent is used to detect the methylation level of the combination of methylation markers, which includes markers 1-15. Using GRch37 as a reference genome, the positions of markers 1-15 on the chromosome are shown below: Marker 1: chr6:26271360-26271510, Marker 2: chr7:27135681-27135820, Marker 3: chr1:63789353-63789499, Marker 4: chr19:13617039-13617183, Marker 5: chr19:30016105-30016253, Marker 6: chr11:112833649-112833795, Marker 7: chr16:51185037-51185189, Marker 8: chr8:4116697 0-41167125, Marker 9: chr8:41166905-41167047, Marker 10: chr1:50886756-50886899, Marker 11: chr16:23847451-23847596, Marker 12: chr20:34189660-34189812, Marker 13: chr16:23847268-23847409, Marker 14: chr1:156611869-156612022 and Marker 15: chr20:34188981-34189119.

[0028] Furthermore, marker 1 includes methylation sites: chr6: 26271508, chr6: 26271492, chr6: 26271486, chr6: 26271481, chr6: 26271475, chr6: 26271465, chr6: 26271463, chr6: 26 271456, chr6: 26271453, chr6: 26271440, chr6: 26271423, chr6: 26271405, chr6: 26271396, chr6: 26271388, chr6: 26271363.

[0029] Marker 2 includes methylation sites: chr7:27135800、chr7:27135794、chr7:27135792、chr7:27135781、chr7:27135778、ch r7: 27135773, chr7: 27135770, chr7: 27135747, chr7: 27135731, chr7: 27135724, chr7: 27135685.

[0030] Marker 3 includes methylation sites: chr1:63789355, chr1:63789365, chr1:63789392, chr1:63789404, chr1:63789410, chr1:63789423, chr1:63789426, chr1:63789429, chr1:63789435, chr1:63789444, chr1:63789465, chr1:63789467, chr1:63789475, chr1:63789478, chr1:63789497.

[0031] Marker 4 includes methylation sites: chr19:13617060, chr19:13617067, chr19:13617072, chr19:13617075, chr19:13617079, chr19:13617084, chr19:13617091, chr19:13617094, chr19:13617097, chr19:13617100, chr19:136 17103, chr19: 13617106, chr19: 13617109, chr19: 13617112, chr19: 13617130, chr19: 13617142, chr1 9: 13617144, chr19: 13617147, chr19: 13617159, chr19: 13617165, chr19: 13617171, chr19: 13617178.

[0032] Marker 5 includes methylation sites: chr19:30016110, chr19:30016122, chr19:30016136, chr19:30016147, chr19:30016152, chr19:30016160, chr19:30016170, chr19:30016188, chr19:30016192, chr19:30016195, chr19:30016202, chr19:30016204, chr19:30016222, chr19:30016224, and chr19:30016244.

[0033] Marker 6 includes methylation sites: chr11:112833792, chr11:112833777, chr11:112833774, chr11:112833770, chr11:112833766, chr11:112833752, chr11:112833748, chr11:112833746, chr11:112833746, chr11:112833792, chr11:112833752, chr11:112833748, chr11:112833746, chr11:112833792 ...4, chr11:112833799, chr11:112833797, chr11:112833777, chr11:112833774, chr11:112833799, chr11:112833777, chr11:112833774, chr11:112833770, chr11:112833766, chr11:112833752, chr11:112833748, chr11:112833746, chr11:112833799, chr11 r11:112833737、chr11:112833716、chr11:112833711、chr11:112833707、chr11:112 833696, chr11: 112833689, chr11: 112833685, chr11: 112833667, chr11: 112833652.

[0034] Marker 7 includes methylation sites: chr16:51185039, chr16:51185055, chr16:51185058, chr16:51185060, chr16:51185065, chr16:51185082, chr16:51185085, chr16:51185093, chr16:51185104, chr16:51185110, chr16:51185132, chr16:51185145, and chr16:51185147.

[0035] Marker 8 includes methylation sites: chr8:41166990, chr8:41166992, chr8:41167003, chr8:41167007, chr8:41167015, chr8:41167021, chr8:41167026, chr8:41167028, chr8:41167045, chr8:41167048, chr8:41167 053, chr8: 41167058, chr8: 41167064, chr8: 41167072, chr8: 41167076, chr8: 41167082, chr8: 41 167085, chr8: 41167087, chr8: 41167098, chr8: 41167107, chr8: 41167109, chr8: 41167113, chr8: 41167119.

[0036] Marker 9 includes methylation sites: chr8:41167046, chr8:41167029, chr8:41167027, chr8:41167022, chr8:41167016, chr8:41167008, chr8:41167004, chr8:41166993, chr8:41166991, chr8:41166 969, chr8: 41166957, chr8: 41166955, chr8: 41166953, chr8: 41166949, chr8: 41166946, chr8:41166941、chr8:41166939、chr8:41166927、chr8:41166924、chr8:41166916、chr8: 41166909.

[0037] Marker 10 includes methylation sites: chr1:50886759, chr1:50886773, chr1:50886779, chr1:50886782, chr1:50886789, chr1:50886800, chr1:50886809, chr1:50886827, chr1:50886829, chr1:50886831, chr1:50886836, chr1:50886849, chr1:50886862, chr1:50886867, chr1:50886876, chr1:50886879, chr1:50886889, chr1:50886895.

[0038] Marker 11 includes methylation sites: chr16:23847456, chr16:23847462, chr16:23847472, chr16:23847479, chr16:23847487, chr16:23847489, chr16:23847491, chr16:23847507, chr16:23847513, chr16:23847 519, chr16: 23847522, chr16: 23847525, chr16: 23847529, chr16: 23847535, chr16: 23847547, c hr16:23847551、chr16:23847556、chr16:23847560、chr16:23847568、chr16:23847575、chr16: 23847586.

[0039] Marker 12 includes methylation sites: chr20:34189678, chr20:34189680, chr20:34189686, chr20:34189692, chr20:34189698, chr20:34189700, chr20:34189703, chr20:34189706, chr20:34189715, chr20:34189750, chr20:34189772, chr20:34189774, chr20:34189781, chr20:34189790, chr20:34189792.

[0040] Marker 13 includes methylation sites: chr16:23847272, chr16:23847287, chr16:23847297, chr16:23847307, chr16:23847311, chr16:23847325, chr16:23847338, chr16:23847344, chr16:23847346, chr16:23847348, c hr16:23847351、chr16:23847353、chr16:23847357、chr16:23847366、chr16:23847369、chr16:23 847380, chr16: 23847384, chr16: 23847386, chr16: 23847392, chr16: 23847404, chr16: 23847408.

[0041] Marker 14 includes methylation sites: chr1:156611890, chr1:156611893, chr1:156611902, chr1:156611905, chr1:156611915, chr1:156611926, chr1:156611945, chr1:156611949, chr1:156611951, chr1:156611960, chr1:156611963, chr1:156611994, chr1:156612002, chr1:156612015.

[0042] Marker 15 includes methylation sites: chr20:34188981, chr20:34189004, chr20:34189008, chr20:34189010, chr20:34189013, chr20:34189025, chr20:34189035, chr20:34189054, chr20:34189074, chr20:34189076, chr20:34189084, chr20:34189090, chr20:34189097, and chr20:34189117.

[0043] After detecting the methylation levels of methylation biomarker combinations in different samples (leukocytes, tissues, and plasma), methylation biomarkers with high background methylation levels in leukocytes were screened out, while combinations of methylation biomarkers with high average methylation rates and high contribution rates in various samples were retained. Furthermore, biomarker combinations with a specificity of ≥95% in normal human samples were screened out. Thus, 15 reliable biomarker combinations with high detection specificity and sensitivity were obtained, which can meet the detection requirements of various sample types. These biomarker combinations can identify the risk of early pancreatic cancer with high sensitivity, high specificity, and rapid, non-invasive detection, which is helpful for the early prevention of pancreatic cancer.

[0044] In a preferred embodiment of the present invention, the early pancreatic cancer diagnostic or prognostic product is selected from reagent kits, biochips, or detection devices.

[0045] The reagents for detecting combinations of methylation markers are selected from at least one of the following reagents used in methods for detecting the methylation level of combinations of methylation markers: Pyrosequencing, bisulfite conversion sequencing, methylation array, qPCR, digital PCR, second-generation sequencing, third-generation sequencing, whole-genome methylation sequencing, DNA enrichment detection, simplified bisulfite sequencing, HPLC, MassArray, and methylation-specific PCR.

[0046] In one embodiment, the kit includes: primers for detecting methylation markers, sulfite conversion reagents, PCR amplification reagents (including enzymes, dNTPs, and buffers), magnetic bead purification reagents, and NGS library construction reagents.

[0047] Biochips include, but are not limited to, microfluidic chips or microarray chips.

[0048] In a preferred embodiment of the present invention, the reagent for detecting the combination of methylation markers includes primers for detecting methylation markers. The primers include a first primer pair to a fifteenth primer pair for detecting markers 1 to 15, respectively, and the nucleotide sequences of the first primer pair to the fifteenth primer pair are shown in SEQ ID NO.1 to SEQ ID NO.30.

[0049] Early pancreatic cancer diagnostic or prognostic products target samples such as cells, tissues, plasma, urine, whole blood, or cerebrospinal fluid. The biomarker combination provided by this invention has good diagnostic or prognostic efficacy for a variety of samples, offering the technical advantage of broad sample applicability.

[0050] Secondly, the present invention also provides a primer combination for the early diagnosis or prognosis of pancreatic cancer, comprising primers for detecting biomarkers 1-15, wherein the primers include: a first primer pair-a fifteenth primer pair for detecting biomarkers 1-15 respectively, the nucleotide sequences of the first primer pair-a fifteenth primer pair being as shown in SEQ ID NO.1-SEQ ID NO.30 respectively. The primer sequences are shown in Table 1 below: Table 1 is a sequence list of each marker.

[0051] In Table 1, Y represents C / T and R represents A / G.

[0052] Thirdly, the present invention also provides a kit, biochip, or detection device for the diagnosis or prognosis of early pancreatic cancer, comprising: a primer combination for the diagnosis or prognosis of early pancreatic cancer.

[0053] In a preferred embodiment of the present invention, the kit, biochip, or detection device further includes primers for detecting the internal reference gene ACTB, the primer sequences of which are shown in SEQ ID NO.31-SEQ ID NO.32.

[0054] The location of the internal reference gene ACTB is: chr7:5571788-5571938.

[0055] The methylation sites of the internal reference gene ACTB include: chr7:5571938, chr7:5571933, chr7:5571932, chr7:5571926, chr7:5571921, chr7:5571917, chr7:5571916, chr7:5571915, chr7:5571912, chr7:5571911, chr7:5571910, chr7:5571909, chr7:5571902, chr7:5571891, chr7:5571887, and chr7:557187. 4. chr7:5571873, chr7:5571872, chr7:5571868, chr7:5571846, chr7:5571844, chr7:5571841, chr7:5571836, chr7:5571833, chr7: 5571828, chr7:5571821, chr7:5571820, chr7:5571814, chr7:5571811, chr7:5571810, chr7:5571808, chr7:5571801, chr7:5571789.

[0056] In all samples, the methylation conversion rate of multiple methylation sites on the internal reference gene marker ACTB was greater than 99.5%, indicating successful methylation conversion in this detection step. Conversion rate = 1 - (number of unconverted cytosines / total number of cytosines at the detection site) × 100%. A conversion rate of ACTB in the internal reference region > 99.5% is required; otherwise, the tested sample will be discarded.

[0057] Quality control is ensured by an internal control ACTB conversion rate of >99.5%, providing a panoramic high-resolution view that is superior to qualitative detection by first-generation Q-PCR.

[0058] Fourthly, the present invention also provides a method for early diagnosis or prognosis of pancreatic cancer, wherein all steps of the method are implemented by a computer, and the method includes the following steps: Obtain the methylation rate of each region in the methylation biomarker combination in the sample; based on an early pancreatic cancer diagnostic model or prognostic model, obtain early pancreatic cancer diagnostic results or prognostic results. The sample is selected from cells, tissues, plasma, urine, whole blood, or cerebrospinal fluid.

[0059] In a preferred embodiment of the present invention, an early pancreatic cancer diagnostic model or prognostic model is constructed by the following method: obtaining the methylation rate of the region where each marker in the combination of methylation markers in the training samples is located, and establishing a dataset; using the dataset and a machine learning model, an early pancreatic cancer diagnostic model or prognostic model is constructed.

[0060] Machine learning models include, but are not limited to, at least one of the following: logistic regression, random forest, K-nearest neighbors, Gaussian Bayes, support vector machine, AdaBoost, XGBoost, Naive Bayes, and DT decision tree.

[0061] When the machine learning model is logistic regression, the model prediction value for early pancreatic cancer diagnosis or prognosis is calculated based on the average methylation rate of all methylation sites in the region of each marker in the combination of methylation markers in the tested sample. Model predicted value = Σ (average methylation rate_i × weight coefficient_i), where i represents the marker number, and its value ranges from 1 to 15; the model predicted value is compared with the preset threshold. If the model predicted value is greater than the preset threshold, it is judged as early pancreatic cancer positive or poor prognosis.

[0062] In one implementation, the LogisticRegression model (solver='liblinear') from the Python sklearn library is used.

[0063] The method for calculating the weight coefficients (w) in logistic regression is as follows: Let the features be $x = (x_1, x_2, ..., x_n)$, the coefficients be $\boldsymbol{w} = (w_1, w_2, ..., w_n)$, and the bias be $w_0$.

[0064] (1) Linear part: $$z = w_0 + w_1x_1 + w_2x_2 + \dots + w_nx_n = \boldsymbol{w}^T\boldsymbol{x}$$; (2) Solve for the pre-weight coefficients of the above features: Core logic: The Sigmoid function maps linear predicted values ​​to the [0,1] interval, representing the probability that the dependent variable is 1. (i) Sigmoid function: $$\sigma(z) = \frac{1}{1+e^{-z}}$$, with a range of $(0,1)$, is used to map linear combinations to probabilities.

[0065] (ii) Predicted probabilities after mapping: $$\hat y = \sigma(\boldsymbol{w}^T\boldsymbol{x}) = \frac{1}{1+e^{-\boldsymbol{w}^T\boldsymbol{x}}}$$.

[0066] (iii) Binary label $y\in\{0,1\}$.

[0067] The solution for w in the predicted probability is not unique. To solve for w in this formula, we use the **Iterative Weighted Least Squares (IRLS) method**, which can be written in the form of: $$\boldsymbol{w}_{k+1} = \boldsymbol{w}_k - (X^TWX)^{-1}X^T(\hat y- y)$$; $W$ is a diagonal weight matrix composed of predicted probabilities.

[0068] (iv) Input the binary classification label y of the training set and the methylation rate x of the training set data into the formula, and iterate continuously according to the gradient convergence to finally obtain the weight coefficients.

[0069] Fifthly, the present invention also provides a computer-readable storage medium storing an executable program instruction set, which is loaded by a processor and executed by the aforementioned method for early pancreatic cancer diagnosis or prognosis.

[0070] The term "computer-readable storage medium" includes data storage media or cloud drives based on a single physical entity such as a CD, CD-ROM, hard disk, optical storage medium, or magnetic disk. Furthermore, the term also includes data storage media composed of physically separate entities that are effectively connected to each other, preferably in a query-and-search manner, to provide the aforementioned data set.

[0071] In a sixth aspect, the present invention also provides an apparatus for early diagnosis or prognosis of pancreatic cancer, the apparatus comprising a processor and a memory, the memory storing a set of executable program instructions, the set of executable program instructions being loaded and executed by the processor to realize the above-described method for early diagnosis or prognosis of pancreatic cancer.

[0072] Specifically, the electronic device may include a memory, a processor, a bus, and a communication interface, which are electrically connected directly or indirectly to each other to enable data transmission or interaction. For example, these components may be electrically connected to each other via one or more buses or signal lines. The processor may process information and / or data related to target identification to perform one or more functions described in this application.

[0073] The memory can be, but is not limited to, Random Access Memory (RAM), Read Only Memory (ROM), Programmable Read-Only Memory (PROM), Erasable Programmable Read-Only Memory (EPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), etc.

[0074] A processor can be an integrated circuit chip with signal processing capabilities. This processor can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0075] In a seventh aspect, the present invention also provides an apparatus for early diagnosis or prognosis of pancreatic cancer, comprising: an input module, a control module, and an output module; The input module is configured to: obtain the methylation rate of the region where each marker in the combination of methylation markers is located in the subject's biological sample, wherein the combination of methylation markers is the combination of methylation markers mentioned above; The control module includes: a results analysis module, which is configured to: generate early pancreatic cancer risk prediction results, prognostic results, or generate early pancreatic cancer negative or positive results for the subject based on the methylation rate of the combination of methylation markers contained in the subject's sample; The output module is configured to output the subject's early pancreatic cancer risk prediction results, prognostic results, or the subject's negative or positive pancreatic cancer results.

[0076] In one implementation, the results analysis module is configured to: calculate, based on the average methylation rate of all methylation sites in the region of each marker in the combination of methylation markers in the subject sample, the model prediction value of the early pancreatic cancer diagnostic model or the early pancreatic cancer prognostic model according to the following formula: Model predicted value = Σ (average methylation rate_i × weight coefficient_i), where i represents the biomarker number, ranging from 1 to 15. The model predicted value is compared with a threshold. If the model predicted value is greater than the threshold, it is judged as early pancreatic cancer positive. When the model is used for prognosis, a predicted value greater than the threshold indicates that the patient has a higher risk of adverse prognostic events (such as recurrence, metastasis, death, etc.), that is, the result is "positive prognosis" or "poor prognosis", requiring closer follow-up or more aggressive adjuvant therapy.

[0077] The features and performance of the present invention will be further described in detail below with reference to embodiments.

[0078] Example 1 This embodiment provides an experimental method for detecting the methylation level of candidate biomarkers in a sample to be tested.

[0079] The experimental method is as follows: (a) Collection of blood samples 1. Whole blood collection We used disposable vacuum blood collection tubes (Zhejiang Medical Device Registration Certificate No. 20232221417) to collect and preserve 10 ml of blood in the morning fasting state from 464 patients with early-stage (stage I and II) pancreatic cancer (mean age 63 years, females 44.40%, of which 351 were included in the training set and 113 in the validation set) and 255 healthy individuals (mean age 60 years, females 46.27%, of which 198 were included in the training set and 57 in the validation set).

[0080] 2. Preparation of plasma samples Centrifuge the blood collection tube containing whole blood for 12 minutes at a centrifugal force of 1500 rcf. Remove the blood collection tube from the centrifuge and transfer the supernatant plasma into a labeled 2 mL centrifuge tube to obtain the sample required for the experiment.

[0081] (ii) Pretreatment of cell-free DNA in plasma The methylation detection sample pretreatment reagent (plasma sample extraction and bisulfite conversion kit (magnetic bead method), manufacturer: Jiaxing Yunying Medical Laboratory Co., Ltd.; catalog number: C128) was used to lyse, convert, and purify the sample. The operation steps are as follows.

[0082] 1. Preparation of reagents in the kit: Prepare the reagents in the above kit according to the table below.

[0083] Table 2. Reagent Preparation

[0084] 2. cfDNA extraction: Take 1.8 ml of plasma, add 100 μl of 20% SDS, mix by inverting, centrifuge briefly, then add 30 μl of PK enzyme (20 mg / ml), vortex 4-5 times to mix; incubate at 60℃ for 30 min, place at room temperature for 5 min, and centrifuge briefly.

[0085] The above mixture was further extracted using an Auto-Pure-20B purifier. The program was run according to the settings. After the program was completed, the liquid was taken out from the last well and placed into a new PCR tube. The tube was labeled with the corresponding sample number, which is cfDNA. It can then be used for the next step of sulfite conversion.

[0086] (1) Sample addition: Add all the solution in the EP tube to column 1 of the purification kit (add different samples to different wells), and be sure to record the sample number; (2) After installing the magnetic rod sleeve, place the purification kit into the fully automated nucleic acid extractor and set it up and operate it according to the prescribed procedure: (3) After the program finishes running, the solution is transferred to a new 0.5 mL centrifuge tube. The collected solution is the cfDNA with magnetic beads and sulfite.

[0087] (III) Sulfite Conversion System preparation: Prepare sulfite reaction systems in 0.5 mL PCR tubes according to the table below for the extracted DNA samples.

[0088] Table 3. Conversion Table for Sulfite Reaction System

[0089] Sulfite Conversion After vortexing and mixing, and briefly centrifuging, place the PCR tube in a PCR instrument and incubate at 95℃ for 5 min and 85℃ for 60 min. After incubation, cool to room temperature and briefly centrifuge to obtain the transformation product.

[0090] (iv) Bis-DNA purification treatment The purified product was run using the Auto-Pure-32A purifier according to the set program. After the run was completed, the liquid was removed from the well and placed into a new PCR 8-strand strip, which was then labeled with the corresponding sample number. This is the purified Bis-DNA, which can be used for the next library enrichment reaction.

[0091] (1) Sample addition: Add all the solution from the PCR tube to column 1 of the purification kit (add different samples to different wells), and be sure to record the sample number; (2) After installing the magnetic rod sleeve, place the purification kit into the fully automated nucleic acid extractor and set it up and operate it according to the prescribed procedure: (3) After the program finishes running, aspirate the solution into a new 1.5 mL centrifuge tube. The collected solution is the DNA after sulfite conversion and should be used for library construction immediately.

[0092] (v) Library enrichment reaction The Bis DNA obtained after purification in the previous step was subjected to a library enrichment reaction.

[0093] Preparation of enrichment PCR reaction Prepare the enrichment reaction solution as needed according to the formula shown in Table 4. After preparation, aliquot the solution into clean eight-slot containers. The aliquoted enrichment reaction solutions are grouped according to the mixed primers and named MEC Enrichment PCR Reaction Strip-1 to MEC Enrichment PCR Reaction Strip-4. Divide the 15 cancer biomarker primers and 1 internal control biomarker primer from Table 1 into 4 combinations, each containing different biomarker primers. Use the 4 mixed enrichment reaction solutions for the enrichment of different biomarkers, and name them MEC Enrichment PCR Reaction Strip 1 to MEC Enrichment PCR Reaction Strip 4, respectively.

[0094] Table 4. Enrichment Reaction Table

[0095] The primer sequence list for each marker is shown in Table 1.

[0096] The enrichment reaction shall be carried out according to the following steps: (1) Gently open the caps of the MEC enrichment PCR reaction strip-1 to MEC enrichment PCR reaction strip-4 prepared in the previous step, and add 1 μL of Taq enzyme to MEC enrichment reaction solution-1 to MEC enrichment PCR reaction solution-11 respectively, ensuring that the pipette tip is deeply inserted into the reaction solution; (2) Add 20 μL of the Bis DNA of the sample to be tested after sulfite conversion to the four enrichment reaction strips in sequence along the tube wall of the PCR tube, and carefully close the tube caps. (3) After vortexing the PCR reaction strip to mix well, centrifuge, taking care to avoid the presence of air bubbles; (4) Place the above 4 enrichment PCR reaction strips into the PCR instrument; (5) Open the PCR instrument settings interface, set the amplification program according to Table 5 below, and perform PCR amplification.

[0097] Table 5. PCR reaction procedure for library enrichment

[0098] (vi) Library preparation reaction: Preparation for the connector connection reaction: Prepare the required number of adapters according to the formula shown in Table 6 (excluding Ace Taq enzyme and the PCR reaction product enriched in the previous step). Alternatively, prepare the adapters in advance and aliquot them into clean 8-strand strips. Store at -20°C for later use.

[0099] After removing the connector reaction solution from the refrigerator and allowing it to melt, centrifuge it briefly in a centrifuge before use.

[0100] Table 6. Reaction Components for Library Preparation

[0101] The connection reaction steps are as follows: (1) Gently open the cap of the connector to the reaction strip and add 0.5 μL of Taq enzyme into the connector to the reaction strip, ensuring that the pipette tip is fully inserted into the reaction solution; (2) Take 5 μL of the enriched PCR reaction amplification product from the previous step and add it sequentially along the tube wall of the PCR tube to the adapter connection reaction strip, and carefully close the tube cap. (3) After connecting the connector to the reaction strip, shake to mix and then centrifuge, taking care to avoid the presence of air bubbles; (4) Place the above-mentioned adapter-connected reaction strip into the PCR instrument; (5) Open the PCR instrument settings interface, set the amplification program according to the table below, and perform PCR amplification.

[0102] Table 7. PCR Reaction Procedure for Library Preparation

[0103] (vii) Sorting of Library Products 1. Remove the magnetic beads used for purification, shake to mix, and incubate at room temperature for at least 30 minutes; 2. Add 30 μL of MEC library-1 to MEC library-2, add 30 μL of MEC library-3 to MEC library-4, and add 10 μL of purified water to each sample reaction tube in sequence; 3. Take 126 μL of the incubated magnetic beads (shake well again before use) and add them sequentially to each sample reaction tube. Shake well and then incubate briefly to completely mix and resuspend the DNA and magnetic beads. 4. Incubate at room temperature for 5 minutes; 5. Incubate on a magnetic rack for 5 minutes until the solution is clear. Carefully aspirate and discard the supernatant, being careful not to disturb the magnetic beads. 6. Add 200µL of freshly prepared 80% ethanol solution (the ethanol solution should just cover the magnetic bead sample), place it on a magnetic rack, and incubate it on the magnetic rack for 30 seconds until the solution becomes clear. Discard the supernatant. 7. Repeat step 6 above for a second wash; 8. Ensure that the ethanol solution in the centrifuge tube has been completely discarded, and place the eight-tube or PCR tube on a magnetic rack to air dry at room temperature for 3-5 minutes; 9. Remove the eight-tube or PCR tube from the magnetic rack, add 70 μL of purified water to fully wet the magnetic beads, shake well to mix, and then quickly centrifuge to collect the liquid at the bottom of the tube. Let it stand at room temperature for 5 min. 10. Place the 8-tube or PCR tube on a magnetic rack and let it stand for 5 minutes until the solution is clear. Transfer 60 μL of the supernatant to a new 8-tube or 1.5 mL centrifuge tube to obtain the sequencing library. Use the Equalbit 1×dsDNA HS Assay Kit and its matching Qubit fluorescence instrument to detect the concentration. The library concentration should be ≥1 ng / μL. Otherwise, the library preparation sample does not meet the requirements and should be reconstructed.

[0104] 11. Perform deep sequencing on the library using an Illumina MiniSeq sequencer, with each sample being sequenced at 150M.

[0105] (viii) Methylation data processing workflow: The first step is data processing and quality control.

[0106] 1. Statistically analyze the methylation rate of methylation sites for 22 gene markers in plasma samples from pancreatic cancer-positive and healthy individuals. The methylation rate is calculated as: (Number of methylated reads at that site / (Number of methylated reads at that site + Number of unmethylated reads at that site)). Data from methylation sites located on primers are removed, and the methylation rate of methylation sites with excessively low sequencing depth (below 10000X) is further excluded. The average methylation rate of all methylation sites within each gene marker was calculated to represent the methylation rate of that gene marker. Gene marker 1 contains 15 methylation sites, gene marker 2 contains 11 methylation sites, gene marker 3 contains 15 methylation sites, gene marker 4 contains 22 methylation sites, gene marker 5 contains 15 methylation sites, gene marker 6 contains 17 methylation sites, gene marker 7 contains 13 methylation sites, gene marker 8 contains 23 methylation sites, gene marker 9 contains 21 methylation sites, gene marker 10 contains 18 methylation sites, gene marker 11 contains 21 methylation sites, gene marker 12 contains 15 methylation sites, gene marker 13 contains 21 methylation sites, gene marker 14 contains 14 methylation sites, and gene marker 15 contains 14 methylation sites.

[0107] Marker 16 contains 8 methylation sites: chr11:2322621-2322770; Marker 17 contains 15 methylation sites: chr19:58220331-58220480; Marker 18 contains 14 methylation sites: chr14:42075876-42076025; Marker 19 contains 7 methylation sites: chr19:2720672-2720821; Marker 20 contains 15 methylation sites: chr21:28217172-28217321; Marker 21 contains 15 methylation sites: chr2:177043236-177043385; Marker 22 contains 10 methylation sites: chr4:134071747-134071896.

[0108] 2. The methylation conversion rate of all 33 methylation sites on the internal reference gene marker ACTB in all samples was greater than 99.5%, indicating that the methylation conversion in this experiment was successful.

[0109] The second step is model training. The methylation rates of the 15 genes selected from the training set data are substituted into the logistic regression model for training. The coefficients and thresholds corresponding to each gene marker are obtained. The methylation level of each gene marker is multiplied by the corresponding coefficient and then added together. If the sum is greater than the threshold, the sample is judged as positive; otherwise, it is negative.

[0110] The third step is model tuning and evaluation. The weights of the dependent variable in the model are adjusted according to the ratio of positive and negative data to balance the training set data. The regularization parameters are tuned to find the optimal parameters. The threshold is adjusted to adapt to clinical testing. The sensitivity and specificity of different models are calculated. ROC curves are plotted to evaluate the excellence of different models and select the best model.

[0111] The fourth step is to perform model validation analysis on the validation set using the optimal model described above, calculate its sensitivity and specificity, and plot the ROC curve.

[0112] Based on the above method, and after multiple rounds of screening and validation, a method for detecting methylation biomarkers in early pancreatic cancer based on NGS sequencing technology was established. 255 methylation sites in 15 DNA regions were selected from the initial candidate biomarkers, and systematic validation was performed in leukocyte, tissue, and plasma samples. Finally, a logistic regression detection model was constructed.

[0113] Example 2 This embodiment provides a screening process for combinations of methylation biomarkers. From an initial pool of 22 candidate biomarkers, 255 methylation sites in 15 DNA regions were selected and systematically validated in leukocyte, tissue, and plasma samples. Finally, a logistic regression detection model was constructed. Screening was based on metrics such as methylation rate, sequencing depth, contribution rate, and specificity.

[0114] (a) Preliminary screening experiment for white blood cells 1. Collected white blood cell samples from 16 normal individuals; 2. PCR amplification was performed using candidate primer combinations; the primer sequences for markers 1-15 are shown in Table 1, and the primer sequences for markers 16-22 are shown in Table 8 below: Table 8. Primer sequence list for detecting markers 16-22.

[0115] 3. Perform NGS sequencing to analyze the average methylation rate and average sequencing depth of methylation sites; 4. Screening criteria: average methylation rate <10%, average sequencing depth >100×.

[0116] Experimental Results and Analysis: 1. All biomarkers showed good sequencing depth, meeting the requirement of sequencing depth >100×; 2. Among the 22 biomarkers, based on the average methylation rate and average sequencing depth of each biomarker, two biomarkers (chr4:134071747-134071896, chr2:177043236-177043385) had an average methylation rate >10%, which did not meet the screening criteria. 3. The 20 selected biomarkers entered the pancreatic cancer tissue detection and verification stage.

[0117] Conclusion: Twenty candidate biomarkers with low methylation rates and good sequencing depth in leukocytes were successfully screened, laying the foundation for subsequent tissue sample validation.

[0118] (II) Organization Sample Validation Experiment By validating tissue samples, the sensitivity and contribution rate of candidate biomarkers in pancreatic cancer tissues were evaluated, and biomarkers with high sensitivity and high contribution rate in tissues were screened out.

[0119] Experimental methods: 1. Collected 20 pancreatic cancer tissue samples; 2. Detection was performed using primers retained after leukocyte screening; 3. Analyze the methylation rate and contribution rate of each biomarker in tissue samples; 4. Screening criteria: Contribution rate ≥ 80%.

[0120] Experimental Results and Analysis: 1. A positive marker was defined as one with an average methylation rate greater than 15% for each marker, and those with a depth >500× were included in the statistical analysis; 2. Among the 20 biomarkers, based on the calculation of average methylation rate and contribution rate, Table 9 shows that 3 biomarkers (chr21:28217172-28217321, chr19:2720672-2720821, chr14:42075876-42076025) had a contribution rate of less than 80%, and did not meet the screening criteria. Table 9 Contribution rate of each biomarker in the organization

[0121] The 17 biomarkers selected have entered the validation phase for pancreatic cancer plasma testing.

[0122] After validation using tissue samples, 17 biomarkers with high sensitivity (contribution rate ≥ 80%) in pancreatic cancer tissues were selected, providing reliable candidate biomarkers for validation using plasma samples.

[0123] (III) Plasma sensitivity testing Validate tissue-screened biomarkers in plasma samples, assess the contribution rate of each biomarker in plasma, and screen biomarker combinations suitable for plasma testing.

[0124] Experimental methods: 1. Collect plasma samples from 20 pancreatic cancer patients; 2. Detection was performed using primer combinations selected from tissue screening; 3. Analyze the contribution rate of each biomarker in the plasma of pancreatic cancer patients; Experimental Results and Analysis: The 17 biomarkers screened in the early stage were tested. The average methylation of each biomarker was greater than 10% as a positive marker and the contribution rate was greater than 30%. Statistical calculations were performed, and one biomarker (chr19:58220331-58220480) had a contribution rate of less than 30%, which did not meet the screening criteria.

[0125] Sixteen biomarkers with good contribution rates in plasma were successfully screened.

[0126] Table 10 Contribution rate of plasma biomarkers

[0127] (iv) Plasma-specific detection In normal human plasma samples, biomarkers with high contribution rates from pancreatic cancer plasma were screened, and the specificity of each biomarker in normal human plasma was evaluated to screen out biomarkers with high specificity.

[0128] Experimental methods: 1. Collect plasma samples from 20 healthy individuals; 2. Detection was performed using primer combinations selected from pancreatic cancer plasma screening; 3. Analyze the specificity of each biomarker in normal human plasma; Experimental Results and Analysis: Sixteen biomarkers were screened from pancreatic cancer plasma. A positive marker was defined as an average methylation level greater than 10% for each biomarker, and a specificity of ≥ 95% was required for screening. The results in Table 11 show that one biomarker (chr11:2322621-2322770) had a specificity lower than 95% and did not meet the screening criteria. Fifteen biomarkers with high specificity were selected to provide reliable data for model construction.

[0129] Table 11 Specificity of plasma biomarkers

[0130] Through multiple rounds of screening, 15 biomarker combinations were finally identified, totaling 255 methylation sites, for subsequent model construction.

[0131] Example 3 This embodiment involves sample collection, data processing, and detection model construction.

[0132] 1. Sample Collection Training set: 351 plasma samples from early-stage pancreatic cancer (pathological examination included pancreatic ductal adenocarcinoma, adenosquamous carcinoma, intraductal papillary mucinous tumor (IPMN) with invasive carcinoma, etc.) and 198 negative plasma samples from healthy individuals were collected. Among the positive samples, pancreatic ductal adenocarcinoma accounted for the highest proportion; the negative samples were from healthy individuals.

[0133] Validation set: 113 positive plasma samples from pancreatic cancer patients and 57 negative plasma samples from healthy individuals.

[0134] 2. DNA extraction and sulfite conversion Cell-free DNA was extracted from plasma samples and transformed using a sulfite transformation kit. Transformation efficiency was assessed using the transformation rate of the internal control region ACTB, calculated as: Transformation rate = (Number of untransformed cytosines / Total number of cytosines at the detection site) × 100%. A transformation rate >99.5% was required; otherwise, the sample was discarded.

[0135] 3. Targeted amplification and NGS sequencing Targeted amplification was performed using the primer combination described above. After purification, the amplified products were deep sequenced on an Illumina MiniSeq sequencer (150M reads per sample). Sequencing data processing included removing primer sites and filtering sites with a sequencing depth <10000X. The methylation rate for each site was calculated as: number of methylated reads / (number of methylated reads + number of unmethylated reads).

[0136] The average methylation rate of each marker region is calculated based on its site: Marker 1: 15 loci; Marker 2: 11 loci; Marker 3: 15 loci; Marker 4: 22 loci; Marker 5: 15 loci; Marker 6: 17 loci; Marker 7: 13 loci; Marker 8: 23 loci; Marker 9: 21 loci; Marker 10: 18 loci; Marker 11: 21 loci; Marker 12: 15 loci; Marker 13: 21 loci; Marker 14: 14 loci; Marker 15: 14 loci.

[0137] 4. Model Training and Optimization Three models were trained using the LogisticRegression model (solver='liblinear') from the Python sklearn library, based on the average methylation rate of 15 markers in the training set. In the `class_weight` parameter, 0 represents negative samples and 1 represents positive samples, with the weight of positive samples being a multiple of that of negative samples.

[0138] Table 12. Model parameters for pancreatic cancer methylation project

[0139] Model predicted value = Σ (average methylation rate_i × weight coefficient_i). If the predicted value > threshold, it is considered positive.

[0140] The method for calculating the weight coefficients (w) in logistic regression is as follows: Let the features be $x = (x_1, x_2, ..., x_n)$, the coefficients be $\boldsymbol{w} = (w_1, w_2, ..., w_n)$, and the bias be $w_0$.

[0141] (1) Linear part: $$z = w_0 + w_1x_1 + w_2x_2 + \dots + w_nx_n = \boldsymbol{w}^T\boldsymbol{x}$$ (2) Solve for the pre-weight coefficients of the above features: Core logic: The Sigmoid function maps linear predicted values ​​to the [0,1] interval, representing the probability that the dependent variable is 1. (i) Sigmoid function: $$\sigma(z) = \frac{1}{1+e^{-z}}$$, with a range of $(0,1)$, is used to map linear combinations to probabilities.

[0142] (ii) Predicted probabilities after mapping: $$\hat y = \sigma(\boldsymbol{w}^T\boldsymbol{x}) = \frac{1}{1+e^{-\boldsymbol{w}^T\boldsymbol{x}}}$$.

[0143] (iii) Binary label $y\in\{0,1\}$.

[0144] The solution for w in the predicted probability is not unique. To solve for w in this formula, we use the **Iterative Weighted Least Squares (IRLS) method**, which can be written in the form of: $$\boldsymbol{w}_{k+1} = \boldsymbol{w}_k - (X^TWX)^{-1}X^T(\hat y- y)$$; The "$" in the formula is the end symbol of the LaTeX mathematical mode, indicating that the formula ends here. $W$ means "W is a diagonal weight matrix composed of predicted probabilities".

[0145] (iv) Input the binary classification label y of the training set and the methylation rate x of the training set data into the formula, and iterate continuously according to the gradient convergence to finally obtain the weight coefficients.

[0146] Table 13. Training set model for pancreatic cancer methylation projects

[0147] Combination Figure 1 , Figure 2 , Figure 3Based on the training set shown in Table 13, among the three training models, Model 1 identified 329 positive cases and 175 negative cases, with a sensitivity of 93.73% and a specificity of 88.38%, a threshold of 3.53, and an area under the ROC curve (AUC) of 0.969323, demonstrating the best identification ability. Model 2 identified 320 positive cases and 182 negative cases, with a sensitivity of 91.17% and a specificity of 91.92%, a threshold of 0.48, and an area under the ROC curve (AUC) of 0.966906. Model 3 identified 317 positive cases and 183 negative cases, with a sensitivity of 90.31% and a specificity of 92.42%, a threshold of 2.22, and an area under the ROC curve (AUC) of 0.967395. Training results: Model 1 was the best, with sensitivity of 93.73% (329 / 351), specificity of 88.38% (175 / 198), threshold of 3.53, and AUC of 0.969323.

[0148] Table 14. Validation set results of pancreatic cancer methylation projects

[0149] From Table 14 above and Figure 4 It can be seen that in the validation set (113 pancreatic cancer cases + 57 normal individuals), 106 cases in Model 1 were judged as positive, with a sensitivity of 93.81%, and 52 cases were judged as negative, with a specificity of 91.23%. The area under the ROC curve, AUC, was 0.979662.

[0150] Figure 5 and Figure 6 The ROC curves for the validation set in Model 2 and Model 3 are shown in the figures. The results show that the 15 biomarkers provided by this invention have good diagnostic efficacy and a specificity of over 91%.

[0151] Example 4 This embodiment provides a kit for the diagnosis of early pancreatic cancer. The kit includes: primer combinations (optimized concentrations) as shown in Table 1 above, sulfite conversion reagent, PCR amplification reagent (including enzymes, dNTPs, and buffer), magnetic bead purification reagent, and NGS library construction reagent.

[0152] This kit is used by extracting DNA from plasma, transforming, amplifying, and sequencing it, calculating the methylation rate, and inputting the data into a model for analysis. It is suitable for clinical screening, testing 10 mL plasma samples, with a sensitivity of 90% and a specificity of 91%.

[0153] Example 5 This embodiment provides a device for early pancreatic cancer diagnosis, which includes: an input module, a control module, and an output module; The input module is configured to: obtain the methylation rate of the region where each marker in the combination of methylation markers is located in the subject's biological sample, wherein the combination of methylation markers is the combination of methylation markers mentioned above; The control module includes: a results analysis module, which is configured to generate early pancreatic cancer risk prediction results for the subject based on the methylation rate of the combination of methylation markers contained in the subject's sample; The output module is configured to output the subject's early pancreatic cancer risk prediction results.

[0154] The results analysis module is configured to: calculate the model prediction value of the early pancreatic cancer diagnostic model based on the average methylation rate of all methylation sites in the region of each marker in the combination of methylation markers in the subject sample, according to the following formula: Model predicted value = Σ (average methylation rate_i × weight coefficient_i), where i represents the marker number, and its value ranges from 1 to 15; the model predicted value is compared with the threshold. If the model predicted value is greater than the threshold, it is judged as early pancreatic cancer positive.

[0155] The aforementioned device can be integrated into a cloud platform, supporting batch processing. Clinical application: In screening high-risk populations, positive samples can be further confirmed endoscopically, reducing unnecessary invasive examinations.

[0156] The above embodiments verify the effectiveness of the methylation biomarker combination, product, and device of the present invention in the early detection of pancreatic cancer, which can achieve non-invasive, high-sensitivity, high-specificity, and high-accuracy detection.

[0157] In summary, the model of this invention can identify the risk of early pancreatic cancer with a highly sensitive, specific, rapid, and non-invasive method, enabling early medical examinations and prevention.

[0158] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. The application of a reagent for detecting a combination of methylation markers in the preparation of diagnostic products for early pancreatic cancer, characterized in that, The reagent for detecting the methylation marker combination is used to detect the methylation level of the methylation marker combination, which includes markers 1-15. Using GRch37 as the reference genome, the positions of markers 1-15 on the chromosome are shown below: Marker 1: chr6:26271360-26271510, Marker 2: chr7:27135681-27135820, Marker 3: chr1:63789353-63789499, Marker 4: chr19:13617039-13617183, Marker 5: chr19:30016105-30016253, Marker 6: chr11:112833649-112833795, Marker 7: chr16:51185037-51185189, Marker 8: chr8:4116697 0-41167125, Marker 9: chr8:41166905-41167047, Marker 10: chr1:50886756-50886899, Marker 11: chr16:23847451-23847596, Marker 12: chr20:34189660-34189812, Marker 13: chr16:23847268-23847409, Marker 14: chr1:156611869-156612022 and Marker 15: chr20:34188981-34189119.

2. The application according to claim 1, characterized in that, The early pancreatic cancer diagnostic product is selected from reagent kits, biochips, or detection devices; The reagent used to detect the combination of methylation markers is selected from at least one of the reagents used in methods for detecting the methylation level of the combination of methylation markers: Pyrosequencing, bisulfite conversion sequencing, methylation array, qPCR, digital PCR, second-generation sequencing, third-generation sequencing, whole-genome methylation sequencing, DNA enrichment detection, simplified bisulfite sequencing, HPLC, MassArray, and methylation-specific PCR.

3. The application according to claim 2, characterized in that, The reagent for detecting the combination of methylation markers includes primers for detecting methylation markers, the primers comprising: a first primer pair to a fifteenth primer pair for detecting markers 1 to 15 respectively, the nucleotide sequences of the first primer pair to the fifteenth primer pair being as shown in SEQ ID NO.1 to SEQ ID NO.30 respectively; The early pancreatic cancer diagnostic product is designed for samples such as cells, tissues, plasma, urine, whole blood, or cerebrospinal fluid.

4. A primer combination for the early diagnosis of pancreatic cancer, characterized in that, It includes primers for detecting markers 1-15 as described in any one of claims 1-3, the primers comprising: a first primer pair-a fifteenth primer pair for detecting markers 1-15 respectively, the nucleotide sequences of the first primer pair-a fifteenth primer pair being as shown in SEQ ID NO.1-SEQ ID NO.30 respectively.

5. A reagent kit, biochip, or detection device for the early diagnosis of pancreatic cancer, characterized in that, It includes the primer combination for early pancreatic cancer diagnosis as described in claim 4.

6. The reagent kit, biochip, or detection device according to claim 5, characterized in that, The kit, biochip, or detection device also includes primers for detecting the internal reference gene ACTB, the primer sequences of which are shown in SEQ ID NO.31-SEQ ID NO.

32.

7. A method for constructing an early pancreatic cancer diagnostic model, characterized in that, The early pancreatic cancer diagnostic model is constructed by the following method: obtaining the methylation rate of the region where each marker in the combination of methylation markers described in any one of claims 1-3 is located in the training samples, and establishing a dataset; using the dataset, an early pancreatic cancer diagnostic model is constructed through a machine learning model; The machine learning model is selected from at least one of the following: logistic regression, random forest, K-nearest neighbors, support vector machine, AdaBoost, XGBoost, Naive Bayes, and DT decision tree; When the machine learning model is logistic regression, the model prediction value for early pancreatic cancer diagnosis is calculated based on the average methylation rate of all methylation sites in the region of each marker in the methylation marker combination in the tested sample. Model predicted value = Σ (average methylation rate_i × weight coefficient_i), where i represents the marker number, and its value ranges from 1 to 15; the model predicted value is compared with a preset threshold. If the model predicted value is greater than the preset threshold, it is judged as early pancreatic cancer positive. The samples to be tested are selected from cells, tissues, plasma, urine, whole blood, or cerebrospinal fluid.

8. A computer-readable storage medium storing an executable program instruction set, the executable program instruction set being loaded by a processor and executed by the method for constructing the early pancreatic cancer diagnostic model according to claim 7.

9. A device for the early diagnosis of pancreatic cancer, characterized in that, The device includes a processor and a memory, the memory storing a set of executable program instructions, which are loaded and executed by the processor to implement the method for constructing the early pancreatic cancer diagnostic model as described in claim 7, and to diagnose early pancreatic cancer.

Images