A method, apparatus, device, and storage medium for detecting methylation signals based on cell-free DNA sequencing fragments in plasma.

Abstract

This invention discloses a method, apparatus, device, and storage medium for detecting methylation signals based on cell-free DNA sequencing fragments in plasma, relating to the field of non-invasive disease detection technology. The method includes calculating the average methylation value of all CpG sites contained in each cell-free DNA methylation sequencing fragment and discretizing it into a preset methylation level; normalizing the methylation levels corresponding to methylation sequencing fragments within a preset genomic region to obtain methylation status distribution characteristics; screening out significantly differential methylation status distribution characteristics and inputting them into an XGBoost classifier to output a plasma cell-free DNA methylation feature score. This invention employs fragment hierarchical encoding, discretizing the continuous methylation values ​​of a single fragment into a finite number of levels and statistically analyzing their distribution within the genomic region, effectively amplifying weak but specific disease signals, enabling their detection from a large normal background.

Description

Technical Field

[0001] This invention relates to a method, apparatus, device, and storage medium for detecting methylation signals based on cell-free DNA sequencing fragments in plasma, belonging to the field of non-invasive disease detection technology. Background Technology

[0002] Liquid biopsies enable non-invasive disease detection by analyzing biomarkers in bodily fluids such as blood. Among these, cell-free DNA (cfDNA) originates from cell apoptosis or necrosis in the body and carries genetic, fragment omics, and epigenetic information. In particular, DNA methylation is considered one of the most sensitive early cancer detection signals currently available.

[0003] Abnormal alterations in DNA methylation patterns are characteristic of many diseases, including cancer, making cfDNA methylation a highly promising biomarker for sensitive, early, non-invasive diagnosis. However, in early-stage cancer patients, tumor-derived cell-free DNA constitutes only a small fraction of all cell-free DNA. This makes accurate and sensitive detection of cancer-specific methylation signals against a backdrop of abundant non-tumor cell-free DNA a significant technical challenge.

[0004] To date, numerous studies have attempted to address this issue using DNA methylation profiling. For example, CancerDetector employs a probabilistic framework to infer the proportion of tumor-derived cell-free DNA through the co-methylation patterns of adjacent CpG sites; DISMIR and MethylBERT introduce deep learning techniques to integrate fragment methylation and sequence features to improve the accuracy of inferring the proportion of tumor-derived cell-free DNA. However, these existing technologies suffer from the following significant drawbacks and limitations:

[0005] 1. High tissue dependence, limiting clinical applicability: Methods such as CancerDetector, DISMIR, and MethylBERT heavily rely on matched tumor tissue samples and normal plasma samples to identify cancer-specific methylation markers. This requirement is often difficult to meet in real-world clinical diagnostic settings, limiting their large-scale application and making their models susceptible to biases related to specific datasets, resulting in poor generalization ability.

[0006] 2. Lack of universality and high scalability: Existing methods are mostly developed for specific cancer types and lack universality across cancer types. When applying DISMIR or CancerDetector to cancers other than liver cancer, complex and time-consuming marker re-screening or model parameter readjustment and training are required, which severely limits the scalability and reproducibility of the methods.

[0007] 3. High computational resource requirements and high application threshold: In particular, deep learning-based methods, such as DISMIR and MethylBERT, require a large amount of computational resources and time for model training, which sets a high technical threshold and economic cost for their widespread deployment and application in clinical institutions.

[0008] 4. Failure to adequately capture tumor heterogeneity: Existing methods are insufficient in effectively capturing and reflecting the complex methylation heterogeneity within tumor cells, which limits a deeper understanding of the biological complexity of cancer and the sensitivity of its detection.

[0009] Therefore, there is an urgent need in this field for a new cfDNA detection method that is independent of tissue samples, has broad applicability to a wide range of cancer types and non-cancer diseases, is computationally efficient, and can sensitively capture disease-specific methylation signals. Summary of the Invention

[0010] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method, device, equipment and storage medium for detecting methylation signals based on cell-free DNA sequencing fragments in plasma. It adopts fragment hierarchical encoding, which discretizes the continuous methylation values ​​of a single fragment into a limited number of levels and statistically analyzes their distribution in the genomic region, effectively amplifying weak but specific disease signals so that they can be detected from a large amount of normal background.

[0011] To achieve the above objectives, the present invention is implemented using the following technical solution:

[0012] In a first aspect, the present invention provides a method for detecting methylation signals based on cell-free DNA sequencing fragments in plasma, comprising:

[0013] S1. Obtain methylation sequencing data of cell-free DNA in plasma;

[0014] S2. For each methylated sequencing fragment in the methylated sequencing data of cell-free DNA in plasma, calculate the average methylation value of all CpG sites contained therein, and discretize the average methylation value to a preset methylation level.

[0015] S3. Count the number of methylated sequencing fragments at each preset methylation level within the preset genomic region, and normalize them to obtain the methylation status distribution characteristics of the genomic region;

[0016] S4, repeating S3, integrate the methylation status distribution characteristics of all genomic regions, and screen out the significantly different methylation status distribution characteristics;

[0017] S5. Using the distribution characteristics of significantly differential methylation states as input, and based on the XGBoost classifier, the output is the plasma free DNA methylation feature score.

[0018] Furthermore, the average methylation value is discretized to the closest preset methylation level.

[0019] Furthermore, the expression for the preset methylation level is as follows: .

[0020] Furthermore, the methylation level corresponding to the methylated sequencing fragments within the preset genomic region is statistically analyzed and normalized to obtain the methylation status distribution characteristics of that genomic region, including:

[0021] The number of methylated sequencing fragments corresponding to preset methylation levels within a preset genomic region is counted.

[0022] The distribution characteristics of methylation status in this genomic region were obtained by normalizing the number of methylated sequencing fragments.

[0023] Furthermore, the expression for the methylation state distribution characteristics is as follows: ,in, This represents the normalized value of the number of values ​​attributed to methylation level 0. This represents the normalized number of values ​​attributed to a methylation level of 0.25. This represents the normalized number of values ​​attributed to a methylation level of 0.5. This represents the normalized number of values ​​attributed to a methylation level of 0.75. This represents the normalized number of values ​​attributed to methylation level 1.

[0024] Furthermore, the method of synthesizing the methylation status distribution characteristics of all genomic regions and screening out significantly differential methylation status distribution characteristics includes:

[0025] D1. Obtain disease group data and healthy control group data. The disease group data includes methylation sequencing data samples of cell-free DNA in the plasma of multiple disease patients, and the healthy control group data includes methylation sequencing data of cell-free DNA in the plasma of multiple healthy test subjects.

[0026] D2. For each methylated sequencing fragment in the methylation sequencing data of cell-free DNA in plasma, calculate the average methylation value of all CpG sites contained therein, and discretize the average methylation value to a preset methylation level.

[0027] D3. Statistically analyze the methylation levels of methylated sequencing fragments within a predefined genomic region and normalize them to obtain the methylation status distribution characteristics of that genomic region.

[0028] D4. Repeat D3 to synthesize the methylation status distribution characteristics of all genomic regions and obtain the methylation status distribution characteristics of the disease group and the healthy control group.

[0029] D5. Compare the methylation status distribution characteristics of the disease group and the healthy control group, and use statistical test methods to preliminarily screen and obtain the differential methylation status distribution characteristics.

[0030] D6. Based on the Boruta algorithm, the distribution features of differential methylation states are screened to obtain the distribution features of significantly differential methylation states.

[0031] Furthermore, it also includes pre-training the XGBoost classifier, the pre-training method comprising:

[0032] Multiple data samples were acquired, including significantly different methylation state distribution features and their corresponding category labels.

[0033] The distribution features of significantly differentially methylated states in the data samples are used as input, and the XGBoost classifier is trained using leave-one-out cross-validation to obtain a pre-trained XGBoost classifier.

[0034] Secondly, the present invention also provides a methylation signal detection device based on cell-free DNA sequencing fragments in plasma, comprising:

[0035] The methylation sequencing data acquisition module is configured to acquire methylation sequencing data of cell-free DNA in plasma;

[0036] The methylation level discretization module is configured to calculate the average methylation value of all CpG sites contained in each methylated sequencing fragment in the methylated sequencing data of cell-free DNA in plasma, and discretize the average methylation value to a preset methylation level.

[0037] The methylation state distribution feature processing module is configured to statistically analyze the methylation level of methylated sequencing fragments within a preset genomic region and normalize them to obtain the methylation state distribution feature of that genomic region.

[0038] The module for screening significantly differentially methylated state distribution features is configured to integrate the methylated state distribution features of all genomic regions and screen out significantly differentially methylated state features from them.

[0039] The plasma free DNA methylation feature score acquisition module is configured to take the distribution features of significantly differential methylation states as input, and output the plasma free DNA methylation feature score based on the XGBoost classifier.

[0040] Thirdly, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for detecting multilevel methylation signals based on cell-free DNA sequencing fragments in plasma as described in any of the first aspects.

[0041] Fourthly, the present invention also provides an apparatus comprising:

[0042] Memory, used to store instructions;

[0043] A processor is configured to execute the instructions, causing the device to perform operations implementing the multilevel methylation signal detection method based on cell-free DNA sequencing fragments in plasma as described in any of the first aspects.

[0044] Compared with the prior art, the beneficial effects achieved by the present invention are as follows:

[0045] This invention employs fragment hierarchical encoding, which discretizes the continuous methylation values ​​of a single fragment into a limited number of levels and statistically analyzes their distribution in genomic regions. This effectively amplifies weak but specific disease signals, enabling them to be detected from a large amount of normal background. This is a fundamental improvement over all methods that rely on average methylation levels.

[0046] The entire testing process of this invention only requires cell-free DNA methylation sequencing data from the patient's plasma, completely eliminating the dependence on difficult-to-obtain matching tissue samples. It does not pre-set specific disease biomarkers and can be seamlessly applied to different types of cancer as well as non-cancerous diseases such as liver disease and amyotrophic lateral sclerosis. Furthermore, the computational cost has been effectively reduced. Attached Figure Description

[0047] Figure 1 This is a schematic flowchart of a method for detecting methylation signals based on cell-free DNA sequencing fragments in plasma, according to one embodiment of the present invention. Detailed Implementation

[0048] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and should not be used to limit the scope of protection of the present invention.

[0049] Example 1

[0050] like Figure 1 As shown, this embodiment of the invention provides a method for detecting methylation signals based on cell-free DNA sequencing fragments in plasma, comprising the following steps:

[0051] Methylation sequencing data is obtained from cell-free DNA (cfDNA, which refers to extracellular DNA fragments circulating in blood and other body fluids). The methylation sequencing data consists of a large number of methylated sequencing fragments.

[0052] For each methylated sequencing fragment, the average methylation value of all CpG sites it contains is calculated, and the average methylation value is discretized to the nearest preset methylation level. In this embodiment, the methylation level is preset to five levels, and its expression is as follows: .

[0053] Within a pre-defined genomic region (e.g., every 500 bp is considered a window), the number of sequencing fragments belonging to each methylation level is counted and normalized to obtain the methylation status distribution characteristics of that genomic region, expressed as: ,in, This represents the normalized value of the number of values ​​attributed to methylation level 0. This represents the normalized number of values ​​attributed to a methylation level of 0.25. This represents the normalized number of values ​​attributed to a methylation level of 0.5. This represents the normalized number of values ​​attributed to a methylation level of 0.75. This represents the normalized number of values ​​attributed to methylation level 1.

[0054] Based on the methylation status distribution characteristics of all genomic regions, significantly differential methylation status distribution characteristics were screened out, including:

[0055] D1. Obtain data from the disease group and the healthy control group. The disease group data includes methylation sequencing data of cell-free DNA in the plasma of 24 liver cancer patients, and the healthy control group data includes methylation sequencing data of cell-free DNA in the plasma of 30 healthy test subjects.

[0056] D2. For each methylated sequencing fragment in the methylation sequencing data of cell-free DNA in plasma, calculate the average methylation value of all CpG sites contained therein, and discretize the average methylation value to a preset methylation level.

[0057] D3. Statistically analyze the methylation levels of methylated sequencing fragments within the preset genomic region and normalize them to obtain the methylation status distribution characteristics of the genomic region.

[0058] D4. Repeat D3 to synthesize the methylation status distribution characteristics of all genomic regions and obtain the methylation status distribution characteristics of the disease group and the healthy control group, as shown in Table 1.

[0059] Table 1: Distribution characteristics of methylation status for each sample

[0060]

[0061] D5. Compare the methylation status distribution characteristics of the disease group and the healthy control group, and use statistical test methods (such as t test) to preliminarily screen and obtain the differential methylation status distribution characteristics.

[0062] D6. Based on the Boruta algorithm, the distribution features of differential methylation states were screened to obtain the distribution features of significantly differential methylation states. The distribution features of significantly differential methylation states refer to feature signals with high confidence and strong correlation with the disease, as shown in Table 2:

[0063] Table 2: Significant differences in methylation status distribution characteristics among samples

[0064]

[0065] Construct a pre-trained XGBoost classifier. The pre-training method includes: acquiring multiple data samples, wherein the data samples include significantly differential methylation state distribution features and their corresponding labels;

[0066] The significantly differentially methylation state distribution features in the data samples are used as input, and leave-one-out cross-validation is used to train the XGBoost classifier, resulting in a pre-trained XGBoost classifier. In this embodiment, leave-one-out cross-validation is used to verify the stability and generalization ability of the model in order to evaluate the performance of the classifier.

[0067] It should be noted that, not limited to the leave-one-out cross-validation method in this embodiment, any other method that can achieve effective pre-training of the XGBoost classifier should also be acceptable.

[0068] Based on a pre-trained XGBoost classifier, the methylation status distribution features of the validation set samples are input, and the corresponding plasma free DNA methylation feature scores are output.

[0069] In this embodiment, the pre-trained XGBoost classifier demonstrated excellent classification performance in the validation, achieving an area under the receiver operating characteristic curve (AUROC) of 0.943, an accuracy of 0.893, a precision of 0.909, and a sensitivity of 0.958 at 90% specificity.

[0070] Plasma free DNA methylation characterization score can be used to assist in disease detection. The higher the plasma free DNA methylation characterization score, the greater the probability of disease. Among them, the disease refers to cancer and methylation-related diseases such as amyotrophic lateral sclerosis (ALS).

[0071] Example 2

[0072] This embodiment also provides a methylation signal detection device based on cell-free DNA sequencing fragments in plasma, used to implement the methylation signal detection method based on cell-free DNA sequencing fragments in plasma described in Embodiment 1, including:

[0073] The methylation sequencing data acquisition module is configured to acquire methylation sequencing data of cell-free DNA in plasma.

[0074] The methylation level discretization module is configured to calculate the average methylation value of all CpG sites contained in each methylated sequencing fragment in the methylated sequencing data of cell-free DNA in plasma, and discretize the average methylation value to a preset methylation level.

[0075] The methylation state distribution feature processing module is configured to statistically analyze the methylation level of methylated sequencing fragments within a preset genomic region and normalize them to obtain the methylation state distribution feature of that genomic region.

[0076] The module for screening significantly differentially methylated state distribution features is configured to synthesize the methylated state distribution features of all genomic regions and screen out significantly differentially methylated state distribution features.

[0077] The plasma free DNA methylation feature score acquisition module is configured to take the distribution features of significantly differential methylation states as input, and output the plasma free DNA methylation feature score based on the XGBoost classifier.

[0078] Example 3

[0079] This embodiment also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the methylation signal detection method based on cell-free DNA sequencing fragments in plasma as described in any one of Embodiment 1.

[0080] Example 4

[0081] This embodiment also provides a device, including:

[0082] Memory, used to store instructions;

[0083] A processor is configured to execute the instructions, causing the device to perform operations that implement the methylation signal detection method based on cell-free DNA sequencing fragments in plasma as described in any of Embodiment 1.

[0084] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for detecting methylation signals based on cell-free DNA sequencing fragments in plasma, characterized in that, The detection method is based on non-disease diagnostic purposes and includes: S1. Obtain methylation sequencing data of cell-free DNA in plasma; S2. For each methylated sequencing fragment in the methylated sequencing data of cell-free DNA in plasma, calculate the average methylation value of all CpG sites contained therein, and discretize the average methylation value to a preset methylation level. S3. Count the number of methylated sequencing fragments at each preset methylation level within the preset genomic region, and normalize them to obtain the methylation status distribution characteristics of the genomic region; S4, repeating S3, synthesizes the methylation status distribution characteristics of all genomic regions and screens out significantly differential methylation status distribution characteristics, including: D1. Obtain disease group data and healthy control group data. The disease group data includes methylation sequencing data samples of cell-free DNA in the plasma of multiple disease patients, and the healthy control group data includes methylation sequencing data of cell-free DNA in the plasma of multiple healthy test subjects. D2. For each methylated sequencing fragment in the methylation sequencing data of cell-free DNA in plasma, calculate the average methylation value of all CpG sites contained therein, and discretize the average methylation value to a preset methylation level. D3. Statistically analyze the methylation levels of methylated sequencing fragments within a predefined genomic region and normalize them to obtain the methylation status distribution characteristics of that genomic region. D4. Repeat D3 to synthesize the methylation status distribution characteristics of all genomic regions and obtain the methylation status distribution characteristics of the disease group and the healthy control group. D5. Compare the methylation status distribution characteristics of the disease group and the healthy control group, and use statistical test methods to preliminarily screen and obtain the differential methylation status distribution characteristics. D6. Based on the Boruta algorithm, the distribution features of differential methylation states were screened to obtain the distribution features of significantly differential methylation states; S5. Using the distribution characteristics of significantly differential methylation states as input, and based on the XGBoost classifier, the output is the plasma free DNA methylation feature score.

2. The method for detecting methylation signals based on cell-free DNA sequencing fragments in plasma according to claim 1, characterized in that, The average methylation value is discretized to the closest preset methylation level.

3. The method for detecting methylation signals based on cell-free DNA sequencing fragments in plasma according to claim 2, characterized in that, The expression for the preset methylation level is: .

4. The method for detecting methylation signals based on cell-free DNA sequencing fragments in plasma according to claim 1, characterized in that, The methylation levels corresponding to methylated sequencing fragments within a predefined genomic region are statistically analyzed and normalized to obtain the methylation status distribution characteristics of that genomic region, including: The number of methylated sequencing fragments corresponding to preset methylation levels within a preset genomic region is counted. The distribution characteristics of methylation status in this genomic region were obtained by normalizing the number of methylated sequencing fragments.

5. The method for detecting methylation signals based on cell-free DNA sequencing fragments in plasma according to claim 3, characterized in that, The expression for the methylation state distribution characteristics is as follows: ,in, This represents the normalized value of the number of values ​​attributed to methylation level 0. This represents the normalized number of values ​​attributed to a methylation level of 0.

25. This represents the normalized number of values ​​attributed to a methylation level of 0.

5. This represents the normalized number of values ​​attributed to a methylation level of 0.

75. This represents the normalized number of values ​​attributed to methylation level 1.

6. The method for detecting methylation signals based on cell-free DNA sequencing fragments in plasma according to claim 1, characterized in that, It also includes pre-training the XGBoost classifier, the pre-training method comprising: Multiple data samples were acquired, including significantly different methylation state distribution features and their corresponding category labels. The distribution features of significantly differentially methylated states in the data samples are used as input, and the XGBoost classifier is trained using leave-one-out cross-validation to obtain a pre-trained XGBoost classifier.

7. A device for detecting methylation signals based on cell-free DNA sequencing fragments in plasma, characterized in that, The method for detecting methylation signals based on cell-free DNA sequencing fragments in plasma as described in any one of claims 1 to 6 includes: The methylation sequencing data acquisition module is configured to acquire methylation sequencing data of cell-free DNA in plasma; The methylation level discretization module is configured to calculate the average methylation value of all CpG sites contained in each methylated sequencing fragment in the methylated sequencing data of cell-free DNA in plasma, and discretize the average methylation value to a preset methylation level. The methylation state distribution feature processing module is configured to statistically analyze the methylation level of methylated sequencing fragments within a preset genomic region and normalize them to obtain the methylation state distribution feature of that genomic region. The module for screening significantly differentially methylation state distribution features is configured to synthesize the methylation state distribution features of all genomic regions and screen for significantly differentially methylation state distribution features, including: D1. Obtain disease group data and healthy control group data. The disease group data includes methylation sequencing data samples of cell-free DNA in the plasma of multiple disease patients, and the healthy control group data includes methylation sequencing data of cell-free DNA in the plasma of multiple healthy test subjects. D2. For each methylated sequencing fragment in the methylation sequencing data of cell-free DNA in plasma, calculate the average methylation value of all CpG sites contained therein, and discretize the average methylation value to a preset methylation level. D3. Statistically analyze the methylation levels of methylated sequencing fragments within a predefined genomic region and normalize them to obtain the methylation status distribution characteristics of that genomic region. D4. Repeat D3 to synthesize the methylation status distribution characteristics of all genomic regions and obtain the methylation status distribution characteristics of the disease group and the healthy control group. D5. Compare the methylation status distribution characteristics of the disease group and the healthy control group, and use statistical test methods to preliminarily screen and obtain the differential methylation status distribution characteristics. D6. Based on the Boruta algorithm, the distribution features of differential methylation states were screened to obtain the distribution features of significantly differential methylation states; The plasma free DNA methylation feature score acquisition module is configured to take the distribution features of significantly differential methylation states as input, and output the plasma free DNA methylation feature score based on the XGBoost classifier.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the method for detecting methylation signals based on cell-free DNA sequencing fragments in plasma as described in any one of claims 1 to 6.

9. A device, characterized in that, include: Memory, used to store instructions; A processor is configured to execute the instructions, causing the device to perform operations that implement the method for detecting methylation signals based on cell-free DNA sequencing fragments in plasma as described in any one of claims 1 to 6.

Images