A method and system for predicting DNA methylation levels based on combined pathological phenotypic features

Abstract

The present invention belongs to the field of computer technology and provides a method and system for predicting DNA methylation level profiles in conjunction with pathological phenotype characteristics. The training method includes obtaining a characteristic spectrum of cell nuclei in the tumor region of the digital pathology slide data based on digital pathology slide data; obtaining a methylation level matrix of all CpG sites based on the patient's original DNA methylation sequencing data; determining the clinical diagnostic stage of cancer tissue samples, normal tissue samples, and cancer patients based on the methylation level matrix of all CpG sites and the patient's clinical information, performing differential methylation CpG site analysis, and obtaining a tumor differential CpG site methylation level spectrum and a tumor stage differential CpG site methylation level spectrum; and training a generative model using the characteristic spectrum of cell nuclei in the tumor region of the digital pathology slide data as input and the differential CpG site methylation level spectrum of the tumor and the differential CpG site methylation level spectrum of the tumor as output.

Description

Technical Field

[0001] This invention belongs to the field of computer technology, and in particular relates to a method and system for predicting DNA methylation level profiles based on combined pathological phenotypic features. Background Technology

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] DNA methylation is one of the most common epigenetic modifications in cells, playing a crucial role in cell differentiation, cell identity maintenance, genetic imprinting, and tumorigenesis. It is involved in the regulation of expression of many cancer-related genes and can therefore be used to analyze and design cancer drugs. However, DNA methylation sequencing is expensive, especially for bisulfite whole-genome methylation sequencing.

[0004] In routine clinical testing, the preparation cost of pathological stained sections is low; moreover, changes in histopathology reflect the tumor development and evolution in cancer patients. Therefore, histopathological sections are considered the gold standard for clinical diagnosis and prognosis. In recent years, many studies have analyzed the molecular level, such as gene mutations, from the phenotypic analysis of digital pathological sections, revealing the correlation between the phenotypic characteristics of digital pathological sections in cancer patients and molecular patterns at the gene level.

[0005] However, current research has the following problems:

[0006] The focus is on molecular levels such as gene mutations, copy number variations, and RNA transcriptional expression; there is relatively little work on predicting from the molecular level of DNA methylation.

[0007] Gene mutation copy number variation information is usually static and irreversible; however, DNA methylation, like tissue section phenotype, is a dynamic development process. Therefore, studying pathological tissue section phenotypes to analyze the molecular level, especially the DNA methylation level profile, is more valuable and meaningful.

[0008] Existing molecular-level predictions mainly involve predicting a small number of specific targets, that is, only selecting key and important genes for prediction; moreover, the prediction of key targets is done one by one, which is a complex and time-consuming process. Summary of the Invention

[0009] To address the technical problems mentioned above, this invention provides a method and system for predicting DNA methylation levels based on combined pathological phenotypic features. This invention constructs a molecular-level differential CpG site methylation level spectrum and predicts the methylation status of molecular-level differential CpG sites based on slice-level phenotypic features. On the one hand, this can significantly reduce the cost of methylation sequencing, improve the efficiency of subsequent clinical diagnosis and prognosis, and provide assistance for further research on important carcinogenic pathways. On the other hand, during the modeling process, it explores the correlation between the macroscopic phenotype of digital pathological slices of tumors and molecular-level gene expression regulatory elements, and performs multi-omics fusion representation, providing more possibilities for precision medicine based on multi-omics and multimodal bioinformatics tumor analysis.

[0010] In order to achieve the above object, the present invention adopts the following technical solutions:

[0011] The first aspect of the present invention provides a model training method.

[0012] A model training method, comprising:

[0013] Based on the digital pathological slide data, the nuclear feature spectrum of the tumor region was obtained from the digital pathological slide data;

[0014] Based on the patient's raw DNA methylation sequencing data, a methylation level matrix of all CpG sites was obtained. Based on the methylation level matrix of all CpG sites and the patient's clinical information, the clinical diagnostic stage of cancer tissue samples, normal tissue samples, and cancer samples was determined. Differential methylation CpG site analysis was performed to obtain the tumor differential CpG site methylation level spectrum and the tumor stage differential CpG site methylation level spectrum.

[0015] The generative model is trained by taking the tumor region nuclear feature spectrum of digital pathological slide data as input and the differential CpG site methylation level spectrum of the tumor and the differential CpG site methylation level spectrum of the tumor stage as output.

[0016] Furthermore, the process of obtaining the tumor region nuclear feature spectrum of digital pathological slide data includes:

[0017] Background filtering is performed on digital pathology slide data to extract tissue regions;

[0018] Based on tissue regions, extract tumor regions;

[0019] The tumor region is segmented by cell nuclei to obtain all cell nuclei;

[0020] Based on all cell nuclei, local and global features of all cell nuclei are extracted to construct a tumor region cell nucleus feature spectrum from digital pathological slide data.

[0021] Furthermore, the process of obtaining the methylation level matrix of all CpG sites includes:

[0022] The patient’s raw DNA methylation sequencing data were quality controlled and preprocessed to obtain the first raw DNA methylation sequencing data.

[0023] The first DNA methylation raw sequencing data was compared with the reference genome sequence, and then sorted and indexed to obtain the second DNA methylation raw sequencing data.

[0024] The methylation levels of the second DNA methylation raw sequencing data were calculated to obtain a methylation level matrix for all CpG sites.

[0025] Furthermore, the principle for determining the differentially methylated CpG sites is as follows: the absolute value of the difference in average methylation level is greater than 0.2 and the p value is corrected. The resulting differentially methylated CpG sites include hypermethylated CpG sites and hypomethylated CpG sites.

[0026] Furthermore, the process of training the generative model includes:

[0027] The feature embedding is obtained by taking the cell nuclear feature spectrum of the tumor region from digital pathological slide data as input;

[0028] The feature embeddings are transformed to obtain the output result;

[0029] Based on the output results, the methylation level spectrum of differential CpG sites in the tumor, and the methylation level spectrum of differential CpG sites in the tumor stage, a loss function is constructed to optimize the parameters of the generative model.

[0030] A second aspect of the present invention provides a model training system.

[0031] A model training system, comprising:

[0032] The first data processing module is configured to: obtain the tumor region cell nuclear feature spectrum based on the digital pathological slide data;

[0033] The second data processing module is configured to: obtain a methylation level matrix of all CpG sites based on the patient's raw DNA methylation sequencing data; and, based on the methylation level matrix of all CpG sites and the patient's clinical information, determine the clinical diagnostic stage of cancer tissue samples, normal tissue samples, and cancer sample patients, perform differential methylation CpG site analysis, and obtain the tumor differential CpG site methylation level spectrum and the tumor stage differential CpG site methylation level spectrum.

[0034] The model training module is configured to take the tumor region cell nuclear feature spectrum of digital pathological slide data as input and the differential CpG site methylation level spectrum and the differential CpG site methylation level spectrum of tumor stage as output to train a generative model.

[0035] A third aspect of the present invention provides a method for predicting DNA methylation level profiles in conjunction with pathological phenotypic features.

[0036] A method for predicting DNA methylation level profiles based on combined pathological phenotypic features, comprising:

[0037] Obtain digital pathological slide data to obtain the tumor region cell nuclear feature spectrum of the digital pathological slide data;

[0038] A generative model was used to process the nuclear feature spectra of tumor regions in all digital pathological sections to predict the methylation level spectra of differential CpG sites in the tumor and the methylation level spectra of differential CpG sites in the tumor stage; the generative model was trained using the model training method described in the first aspect.

[0039] A fourth aspect of the present invention provides a DNA methylation level profiling prediction system that combines pathological phenotypic features.

[0040] A DNA methylation level profiling prediction system combining pathological phenotypic features includes:

[0041] The data acquisition module is configured to acquire digital pathological slide data and obtain the tumor region cell nuclear feature spectrum of the digital pathological slide data.

[0042] The prediction module is configured to: process the nuclear feature spectra of tumor regions in all digital pathological slices using a generative model to predict the methylation level spectra of differential CpG sites in the tumor and the methylation level spectra of differential CpG sites in the tumor stage; the generative model is trained using the model training method described in the first aspect.

[0043] A fifth aspect of the present invention provides a computer-readable storage medium.

[0044] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the model training method as described in the first aspect, or implements the steps in the DNA methylation level spectrum prediction method combining pathological phenotypic features as described in the third aspect.

[0045] A sixth aspect of the present invention provides a computer device.

[0046] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the model training method as described in the first aspect, or to implement the steps in the DNA methylation level spectrum prediction method combining pathological phenotypic features as described in the third aspect.

[0047] Compared with the prior art, the present invention has the following beneficial effects:

[0048] (1) By constructing a molecular-level DNA methylation differential CpG site methylation level spectrum, this invention can quickly and accurately locate the set of tumor-related CpG sites in tumor patients, and can further search for CpG regions to promote subsequent individualized clinical drug treatment.

[0049] (2) This invention enables accurate prediction of the methylation level of differential CpG sites at the molecular level from the phenotype of digital pathological slides, which simplifies the prediction process, greatly reduces the cost of sample sequencing, improves the efficiency of the prediction process, and improves the efficiency of subsequent diagnosis and prognosis.

[0050] (3) This invention establishes the correlation between the phenotypic cell nuclear feature spectrum of digital pathological slides and the molecular-level differential CpG site methylation level spectrum, realizing multi-omics fusion representation; it can provide more possibilities for precision medicine based on multi-omics multimodal biological information tumor analysis. Attached Figure Description

[0051] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0052] Figure 1 This is an overall architecture diagram shown in the present invention;

[0053] Figure 2 This is a diagram showing the results of the DNA methylation differential site analysis of this invention. Detailed Implementation

[0054] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0055] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0056] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0057] It should be noted that the flowcharts and block diagrams in the accompanying drawings illustrate the possible implementation architecture, functions and operations of the methods and systems according to the various embodiments of the present disclosure. It should be noted that each box in the flowchart or block diagram can represent a module, program segment, or a part of code, and the module, program segment, or a part of code can include one or more executable instructions for implementing the logical functions specified in the various embodiments. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, or they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the flowchart and / or block diagram, and the combination of the boxes in the flowchart and / or block diagram, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or can be implemented using a combination of dedicated hardware and computer instructions.

[0058] Example 1

[0059] like Figure 1 As shown, this embodiment provides a model training method. This embodiment uses the application of this method to a server as an example for illustration. It is understood that this method can also be applied to terminals, and can also be applied to systems including terminals, servers, and other components, and implemented through interaction between the terminal and the server. The server can be an independent physical server, a server cluster composed of multiple physical servers, or a distributed system. It can also be a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network servers, cloud communication, middleware services, domain name services, CDN security services, and big data and artificial intelligence platforms. The terminal can be a smartphone, tablet, laptop, desktop computer, etc., but is not limited to these. The terminal and server can be directly or indirectly connected via wired or wireless communication, which is not limited herein. In this embodiment, the method includes the following steps:

[0060] Based on the digital pathological slide data, the nuclear feature spectrum of the tumor region was obtained from the digital pathological slide data;

[0061] Based on the patient's raw DNA methylation sequencing data, a methylation level matrix of all CpG sites was obtained. Based on the methylation level matrix of all CpG sites and the patient's clinical information, the clinical diagnostic stage of cancer tissue samples, normal tissue samples, and cancer samples was determined. Differential methylation CpG site analysis was performed to obtain the tumor differential CpG site methylation level spectrum and the tumor stage differential CpG site methylation level spectrum.

[0062] The generative model is trained by taking the tumor region nuclear feature spectrum of digital pathological slide data as input and the differential CpG site methylation level spectrum of the tumor and the differential CpG site methylation level spectrum of the tumor stage as output.

[0063] The specific solution of this embodiment can be implemented using the following methods:

[0064] like Figure 1 As shown, for the construction of phenotypic features of digital pathological slides, firstly, background filtering is performed on the digital pathological slide data to extract tissue regions; then, a tumor region recognition model is developed based on a tumor region labeling dataset and applied to the extracted tissue regions to further extract tumor regions; next, a cell nucleus instance segmentation model is developed based on a multi-class cell nucleus labeling dataset and applied to the extracted tumor regions for cell nucleus instance segmentation; further, local feature extraction and global graph feature construction are performed on all the segmented cell nuclei; finally, the tumor region cell nucleus feature spectrum of the digital pathological slide data is obtained. Both the tumor region recognition model and the cell nucleus instance segmentation model can employ existing deep learning neural network models, including CNN networks, RNN networks, etc. It should be noted that this embodiment is not limited to a specific network or structure.

[0065] like Figure 1 As shown, to construct a differential CpG site methylation level profile, the raw DNA methylation sequencing data of patients were first subjected to quality control and preprocessing, including adapter removal and low-quality base removal. Then, the quality-controlled data was aligned with the reference genome sequence using BSMAP software, and sorted and indexed. Further, methylation levels were calculated on the processed data (in whole-genome methylation sequencing, methylation levels can be calculated based on the ratio of C reads not converted to T to C reads converted to T), thus obtaining the methylation level matrix of all CpG sites in the cohort data. Considering the data redundancy of the samples, the above matrix needed to be filtered for NA values.

[0066] Furthermore, in tumor analysis, tumor-related genes have clinical diagnostic and prognostic value. Therefore, this patent, based on patient clinical information, determined cancer tissue and normal tissue samples, as well as the clinical diagnostic stage of cancer patients, and performed differential methylation CpG site analysis. Here, the principle for determining differential methylation CpG sites is that the absolute value of the difference in average methylation level is greater than 0.2 and the adjusted p-value is <0.05. The obtained differential methylation CpG sites include hypermethylated CpG sites and hypomethylated CpG sites, such as... Figure 2 As shown.

[0067] Ultimately, we obtained differentially methylated CpG sites for specific tumors and differentially methylated CpG sites for specific tumor stages, which respectively constituted the tumor differential CpG site methylation level spectrum and the tumor stage differential CpG site methylation level spectrum as prediction targets for subsequent modeling work.

[0068] After constructing and obtaining the nuclear feature spectrum of tumor regions on digital pathological slides, the differential CpG site methylation level spectrum of specific tumors, and the differential CpG site methylation level spectrum of specific tumor stages, this embodiment will perform modeling and prediction of the differential CpG site methylation level spectrum to realize multi-omics association modeling of tumors from digital pathological slide phenotype to molecular level DNA methylation level.

[0069] Here, nuclear feature spectra are fed into a deep network model for learning, ultimately predicting the methylation level spectra of differentially expressed CpG sites. We employ generative model frameworks such as variational autoencoders, generative adversarial networks, flow models, or diffusion models for modeling and learning to further explore the effectiveness of associations between multi-omics data.

[0070] The purpose of employing a generative model is to treat the nuclear feature spectrum of digital pathological slide phenotypes as data sampling points of a priori distribution, and then transform it into new sampling points that satisfy the new data distribution of DNA methylation level spectrum at the molecular level through a generative model. Assuming the nuclear feature spectrum of digital pathological slides is the input vector i, and the differential CpG site methylation level spectrum is the output vector o, then the key to all generative modeling lies in the fact that the input i is first transformed into the intermediate layer's feature embedding e = s(i), and then the feature embedding s is further transformed into the output... Construct the loss with the output vector o The parameters of s(·) and t(·) are then adjusted to optimize the model. After the generative model is trained, the parameters of s(·) and t(·) are fixed. For the new digital pathological slide cell nuclear feature spectrum, the input i',o'=t(s(i')) is used as the prediction, and the obtained tumor differential CpG site methylation level spectrum is obtained.

[0071] For a well-learned generative model, not only can the predicted output be obtained from the input, but the feature embedding e of the intermediate layer can also be obtained through s(·). Here, the feature embedding e is derived from the input i, and the feature embedding e can be transformed into the predicted output. That is, the transformation s(i') of the input and the inverse transformation t'(o') of the output are both feature embedding e.

[0072] Therefore, the generative model constructed here establishes a correlation between the phenotypic nuclear feature spectrum of digital pathological slides and the DNA methylation level spectrum at the molecular level. The obtained feature embedding e is a fusion representation of the two, which can be further used to design downstream multi-omics fusion tumor analysis.

[0073] Example 2

[0074] This embodiment provides a model training system.

[0075] A model training system, comprising:

[0076] The first data processing module is configured to: obtain the tumor region cell nuclear feature spectrum based on the digital pathological slide data;

[0077] The second data processing module is configured to: obtain a methylation level matrix of all CpG sites based on the patient's raw DNA methylation sequencing data; and, based on the methylation level matrix of all CpG sites and the patient's clinical information, determine the clinical diagnostic stage of cancer tissue samples, normal tissue samples, and cancer sample patients, perform differential methylation CpG site analysis, and obtain the tumor differential CpG site methylation level spectrum and the tumor stage differential CpG site methylation level spectrum.

[0078] The model training module is configured to take the tumor region cell nuclear feature spectrum of digital pathological slide data as input and the differential CpG site methylation level spectrum and the differential CpG site methylation level spectrum of tumor stage as output to train a generative model.

[0079] It should be noted that the first data processing module, the second data processing module, and the model training module described above are the same examples and application scenarios implemented in the steps of Embodiment 1, but are not limited to the content disclosed in Embodiment 1. It should also be noted that the above modules, as part of the system, can be executed in a computer system such as a set of computer-executable instructions.

[0080] Example 3

[0081] This embodiment provides a method for predicting DNA methylation level profiles based on combined pathological phenotypic features.

[0082] A method for predicting DNA methylation level profiles based on combined pathological phenotypic features, comprising:

[0083] Obtain digital pathological slide data to obtain the tumor region cell nuclear feature spectrum of the digital pathological slide data;

[0084] A generative model was used to process the nuclear feature spectra of tumor regions in all digital pathological sections to predict the methylation level spectra of differential CpG sites in the tumor and the methylation level spectra of differential CpG sites in the tumor stage; the generative model was trained using the model training method described in Example 1.

[0085] like Figure 1 As shown, the specific scheme of this embodiment can be as follows: First, the DNA methylation data is processed and analyzed to obtain the methylation level spectrum of differential CpG sites. Then, the nuclear feature spectrum of tumor regions is extracted from all digital pathological slices in the queue. Finally, the nuclear feature spectrum of the digital pathological slice phenotype is sent into the generative model to model and predict the molecular level methylation spectrum, and the correlation between the digital pathological slice phenotype and the molecular level DNA methylation is mined to realize the multi-omics feature fusion representation.

[0086] This embodiment uses complete and sample-matched multi-omics data, including the patient's clinical information, paired digital pathological sections, and DNA methylation data, to ensure the extraction of cell nuclear feature profiles and the extraction and subsequent prediction of differential CpG site methylation level profiles.

[0087] This embodiment constructs a molecular-level DNA methylation differential CpG site methylation level spectrum to quickly and accurately locate the set of significantly different CpG sites in tumor patients.

[0088] This embodiment accurately predicts the methylation status of differentially expressed CpG sites at the molecular level from digital pathological slide phenotypes, thereby improving the efficiency of subsequent diagnosis and prognosis.

[0089] This embodiment establishes the correlation between the phenotypic cell nuclear feature spectrum of digital pathological sections and the molecular-level differential CpG site methylation level spectrum, and performs multi-omics fusion representation; providing more possibilities for precision medicine based on multi-omics multimodal bioinformatics tumor analysis.

[0090] Example 4

[0091] This embodiment provides a DNA methylation level prediction system that combines pathological phenotypic features.

[0092] A DNA methylation level profiling prediction system combining pathological phenotypic features includes:

[0093] The data acquisition module is configured to acquire digital pathological slide data and obtain the tumor region cell nuclear feature spectrum of the digital pathological slide data.

[0094] The prediction module is configured to: process the nuclear feature spectra of tumor regions in all digital pathological sections using a generative model to predict the methylation level spectra of differential CpG sites in the tumor and the methylation level spectra of differential CpG sites in the tumor stage; the generative model is trained using the model training method described in Example 1.

[0095] It should be noted that the data acquisition module and prediction module described above implement the same examples and application scenarios as those in Embodiment 3, but are not limited to the content disclosed in Embodiment 3. It should also be noted that the above modules, as part of the system, can be executed in a computer system such as a set of computer-executable instructions.

[0096] Example 5

[0097] This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the model training method as described in Embodiment 1, or implements the steps in the DNA methylation level spectrum prediction method based on combined pathological phenotypic features as described in Embodiment 3.

[0098] Example 6

[0099] This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in the model training method described in Embodiment 1, or the DNA methylation level spectrum prediction method based on combined pathological phenotypic features described in Embodiment 3.

[0100] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of hardware embodiments, software embodiments, or embodiments combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage and optical storage, etc.) containing computer-usable program code.

[0101] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0102] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0103] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0104] Those skilled in the art will appreciate that all or part of the processes in the above-described method embodiments can be implemented by instructing related hardware through a computer program. The program can be stored in a computer-readable storage medium, and when executed, the program can include the processes in the above-described method embodiments. The storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0105] 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. A model training method, characterized in that, include: Based on the digital pathological slide data, the nuclear feature spectrum of the tumor region was obtained from the digital pathological slide data; Based on the patient's raw DNA methylation sequencing data, a methylation level matrix of all CpG sites was obtained. Based on the methylation level matrix of all CpG sites and the patient's clinical information, the clinical diagnostic stage of cancer tissue samples, normal tissue samples, and cancer samples was determined. Differential methylation CpG site analysis was performed to obtain the tumor differential CpG site methylation level spectrum and the tumor stage differential CpG site methylation level spectrum, which served as prediction targets for subsequent modeling work. The generative model is trained by taking the tumor region nuclear feature spectrum of digital pathological slide data as input and the differential CpG site methylation level spectrum of the tumor and the differential CpG site methylation level spectrum of the tumor stage as output. The process of training the generative model includes: The feature embedding is obtained by taking the cell nuclear feature spectrum of the tumor region from digital pathological slide data as input; The feature embeddings are transformed to obtain the output result; Based on the output results, the methylation level spectrum of differential CpG sites in the tumor, and the methylation level spectrum of differential CpG sites in the tumor stage, a loss function is constructed to optimize the parameters of the generative model.

2. The model training method according to claim 1, characterized in that, The process of obtaining the tumor region nuclear feature spectrum of digital pathological slide data includes: Background filtering is performed on digital pathology slide data to extract tissue regions; Based on tissue regions, extract tumor regions; The tumor region is segmented by cell nuclei to obtain all cell nuclei; Based on all cell nuclei, local and global features of all cell nuclei are extracted to construct a tumor region cell nucleus feature spectrum from digital pathological slide data.

3. The model training method according to claim 1, characterized in that, The process of obtaining the methylation level matrix of all CpG sites includes: The patient’s raw DNA methylation sequencing data were quality controlled and preprocessed to obtain the first raw DNA methylation sequencing data. The first DNA methylation raw sequencing data was compared with the reference genome sequence, and then sorted and indexed to obtain the second DNA methylation raw sequencing data. The methylation levels of the second DNA methylation raw sequencing data were calculated to obtain a methylation level matrix for all CpG sites.

4. The model training method according to claim 1, characterized in that, The principle for determining the differentially methylated CpG sites is as follows: the absolute value of the difference in average methylation level is greater than 0.2 and the p value is corrected. The resulting differentially methylated CpG sites include hypermethylated CpG sites and hypomethylated CpG sites.

5. A model training system, employing a model training method as described in any one of claims 1-4, characterized in that, include: The first data processing module is configured to: obtain the tumor region cell nuclear feature spectrum based on the digital pathological slide data; The second data processing module is configured to: obtain a methylation level matrix of all CpG sites based on the patient's raw DNA methylation sequencing data; and, based on the methylation level matrix of all CpG sites and the patient's clinical information, determine the clinical diagnostic stage of cancer tissue samples, normal tissue samples, and cancer sample patients, perform differential methylation CpG site analysis, and obtain the tumor differential CpG site methylation level spectrum and the tumor stage differential CpG site methylation level spectrum. The model training module is configured to take the tumor region cell nuclear feature spectrum of digital pathological slide data as input and the differential CpG site methylation level spectrum and the differential CpG site methylation level spectrum of tumor stage as output to train a generative model.

6. A method for predicting DNA methylation levels based on combined pathological phenotypic features, characterized in that, include: Obtain digital pathological slide data to obtain the tumor region cell nuclear feature spectrum of the digital pathological slide data; Generative models were used to process the nuclear feature spectra of tumor regions in all digital pathological sections. The differential CpG site methylation level spectrum and the differential CpG site methylation level spectrum of the tumor stage are predicted; the generative model is trained using the model training method described in any one of claims 1-4.

7. A DNA methylation level profiling prediction system combining pathological phenotypic features, characterized in that, include: The data acquisition module is configured to acquire digital pathological slide data and obtain the tumor region cell nuclear feature spectrum of the digital pathological slide data. The prediction module is configured to: process the nuclear feature spectra of tumor regions in all digital pathological slices using a generative model to predict the methylation level spectra of differential CpG sites in the tumor and the methylation level spectra of differential CpG sites in the tumor stage; the generative model is trained using the model training method described in any one of claims 1-4.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the model training method as described in any one of claims 1-4, or the steps in the DNA methylation level spectrum prediction method based on combined pathological phenotypic features as described in claim 6.

9. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the model training method as described in any one of claims 1-4, or the steps in the DNA methylation level spectrum prediction method based on combined pathological phenotypic features as described in claim 6.

Images