Method and system for predicting tumor differential gene expression profile combined with pathohistological features

Abstract

The application belongs to the technical field of computers and provides a tumor differential gene expression profile prediction method and system combined with pathological omics features. The model training method comprises the following steps: obtaining the nucleus feature spectrum of the specific tumor area of the digital pathology section according to the data pathology section data; obtaining the cohort gene expression matrix according to the transcriptomics data; determining the clinical diagnosis stage of the cancer tissue and normal tissue sample and the cancer sample according to the cohort gene expression matrix and the patient clinical information, and performing gene differential expression analysis to obtain the differential gene expression profile of the specific tumor and the differential gene expression profile of the specific tumor stage; and taking the nucleus feature spectrum of the specific tumor area of the digital pathology section as the input and taking the differential gene expression profile of the specific tumor and the differential gene expression profile of the specific tumor stage as the output to train a generative model. The application greatly reduces the sample sequencing cost and improves the subsequent diagnosis and prognosis efficiency.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of computer, and particularly relates to a tumor differential gene expression profile prediction method and system combined with pathological omics features. BACKGROUND

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

[0003] Gene expression data describes the state of cell activity under different conditions, and is closely related to the tumor development of cancer patients, and plays a key role in guiding clinical drug administration and anticancer drug design; however, gene expression data relies on transcriptome sequencing, and the sequencing cost is expensive, especially for emerging spatial transcriptome sequencing.

[0004] In clinical routine tests, pathological staining section production cost is low; and the histopathological changes reflect the tumor development and evolution of cancer patients, so the histopathological section is used as the gold standard for clinical diagnosis and prognosis. In recent years, many studies have analyzed the molecular level, such as gene mutation, from digital pathology section phenotype, which reveals the correlation between the digital pathology section phenotype characteristics of cancer patients and the gene molecular mode.

[0005] However, the prior art has the following problems:

[0006] 1. The prediction is basically at the molecular level of gene mutation, copy number variation, etc.; and the gene expression prediction is less;

[0007] 2. Gene mutation and copy number variation information are usually static and irreversible; and gene expression is a dynamic development process like section phenotype, so it is more valuable and meaningful to study the pathological section phenotype to analyze the gene expression profile;

[0008] 3. The existing molecular level prediction mainly predicts a small number of special targets, that is, only the key and important genes are selected for expression prediction; and the prediction of key targets is one by one, which has the problems of complex prediction process, long time consumption, incompleteness and incompleteness. SUMMARY

[0009] In order to solve the technical problems in the background art, the present application provides a tumor differential gene expression profile prediction method and system combined with pathological omics features, which accurately predicts the molecular level differential gene expression from the digital pathology section phenotype, and improves the subsequent diagnosis and prognosis efficiency.

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

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

[0012] The model training method comprises the following steps:

[0013] According to the data of pathological sections, the cell nucleus feature spectrum of the specific tumor area of the digital pathological section is obtained.

[0014] According to the transcriptomic data, a cohort gene expression matrix is obtained; according to the cohort gene expression matrix and the patient clinical information, the clinical diagnosis stage of the cancer tissue and normal tissue sample and the cancer sample is determined, and gene differential expression analysis is performed to obtain a differential gene expression spectrum of the specific tumor and a differential gene expression spectrum of the specific tumor stage.

[0015] The generative model is trained by taking the cell nucleus feature spectrum of the specific tumor area of the digital pathological section as input and taking the differential gene expression spectrum of the specific tumor and the differential gene expression spectrum of the specific tumor stage as output.

[0016] Further, the process of obtaining the cell nucleus feature spectrum of the specific tumor area of the digital pathological section comprises:

[0017] Obtaining digital pathological section data and extracting a tissue area;

[0018] According to the tissue area, a tumor area is extracted;

[0019] Performing cell nucleus instantiation segmentation on the tumor area to obtain all cell nuclei;

[0020] Based on all the cell nuclei, cell nucleus features are extracted to obtain the cell nucleus feature spectrum of the specific tumor area of the digital pathological section.

[0021] Further, the process of obtaining the cohort gene expression matrix comprises:

[0022] Performing quality control and preprocessing on the transcriptomic data to obtain first transcriptomic data;

[0023] Aligning the first transcriptomic data with a reference genome sequence, and performing sorting and indexing to obtain second transcriptomic data;

[0024] Performing read processing, expression quantification and standardization on the second transcriptomic data to obtain a cohort gene expression matrix.

[0025] Further, the process of training the generative model comprises:

[0026] The cell nucleus feature spectrum of the specific tumor area of the digital pathological section is first transformed into a feature embedding, and the output result is obtained from the feature embedding;

[0027] According to the output result, the differential gene expression spectrum of the specific tumor and the differential gene expression spectrum of the specific tumor stage, a loss function is constructed, and the generative model is optimized.

[0028] Further, the fusion of the nucleus feature spectrum and the differential gene expression profile of the specific tumor and the differential gene expression profile of the specific tumor staging is expressed as a feature embedding.

[0029] The second aspect of the application provides a model training system.

[0030] The model training system comprises:

[0031] The first data processing module is configured to obtain a nucleus feature spectrum of a specific tumor region of a digital pathology section according to data pathology section data.

[0032] The second data processing module is configured to obtain a cohort gene expression matrix according to transcriptomics data at the gene level, determine the clinical diagnosis staging of cancer and normal tissue samples and cancer samples according to the cohort gene expression matrix and patient clinical information, and perform gene differential expression analysis to obtain a differential gene expression profile of a specific tumor and a differential gene expression profile of specific tumor staging.

[0033] The training module is configured to train a generative model with the nucleus feature spectrum of the specific tumor region of the digital pathology section as input and the differential gene expression profile of the specific tumor and the differential gene expression profile of the specific tumor staging as output.

[0034] The third aspect of the application provides a tumor differential gene expression profile prediction method combined with pathological omics features.

[0035] The tumor differential gene expression profile prediction method combined with pathological omics features comprises:

[0036] According to the tumor region of the patient to be tested, a tumor region nucleus feature spectrum is extracted.

[0037] Based on the tumor region nucleus feature spectrum, a generative model is used to obtain a differential gene expression profile of a specific tumor and a differential gene expression profile of the specific tumor staging; the generative model is obtained by the model training method of the first aspect.

[0038] The fourth aspect of the application provides a tumor differential gene expression profile prediction system combined with pathological omics features.

[0039] The tumor differential gene expression profile prediction system combined with pathological omics features comprises:

[0040] The feature extraction module is configured to extract a tumor region nucleus feature spectrum according to the tumor region of the patient to be tested.

[0041] The prediction module is configured to: based on the tumor region cell nucleus feature spectrum, adopt a generative model to obtain a differential gene expression spectrum of a specific tumor and a differential gene expression spectrum of a specific tumor stage; and the generative model is obtained by the model training method in the first aspect.

[0042] The fifth aspect of the present application provides a computer readable storage medium.

[0043] A computer readable storage medium having a computer program stored thereon, the program being executed by a processor to implement the model training method of the first aspect, or to implement the steps in the joint pathology genomic feature tumor differential gene expression spectrum prediction method of the third aspect.

[0044] The sixth aspect of the present application provides a computer device.

[0045] A computer device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor implements the model training method of the first aspect, or implements the steps in the joint pathology genomic feature tumor differential gene expression spectrum prediction method of the third aspect when executing the program.

[0046] Compared with the prior art, the present application has the following beneficial effects:

[0047] (1) The present application can quickly and accurately locate the tumor-related gene set of a tumor patient by constructing a molecular level transcriptomic differential gene expression spectrum, and promote subsequent clinical individualized drug treatment.

[0048] (2) The present application realizes accurate prediction of molecular level differential gene expression from digital pathology section phenotypes, comprehensive prediction of gene expression, improves the efficiency of prediction, greatly reduces the cost of sample sequencing, and improves the efficiency of subsequent diagnosis and prognosis.

[0049] (3) The present application establishes the correlation between the digital pathology section phenotype cell nucleus feature spectrum and the molecular level differential gene expression spectrum, realizes multi-omics fusion representation; and can provide more possibilities for precision medicine based on multi-omics multi-modal biological information tumor analysis. BRIEF DESCRIPTION OF DRAWINGS

[0050] The drawings accompanying the specification of the present application serve to provide a further understanding of the present application, and the illustrative embodiments of the present application and their descriptions serve to explain the present application, and do not constitute an improper limitation on the present application.

[0051] Figure 1 is a differential gene expression spectrum construction diagram shown by the present application;

[0052] Figure 2is a digital pathology slice phenotype feature spectrum construction flowchart shown by the present application;

[0053] Figure 3 is a generative model structure diagram shown by the present application;

[0054] Figure 4 is a whole architecture diagram shown by the present application. DETAILED DESCRIPTION

[0055] The present application is further described below in conjunction with the accompanying drawings and examples.

[0056] It should be noted that the following detailed description is merely exemplary in nature and is intended to provide further description of the present application. Unless otherwise defined, 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 application belongs.

[0057] It is to be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments in accordance with the present application. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" and / or "comprising," when used in this specification, specify the presence of stated features, steps, operations, elements, components, and / or groups thereof, but do not preclude the presence or addition of one or more other features, steps, operations, elements, components, and / or groups thereof.

[0058] It should be noted that the flowchart and block diagrams in the drawings show the architecture, functionality, and operation of possible implementations of methods and systems according to various embodiments of the present disclosure. It should also be noted that each block in the flowchart and block diagrams can represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical functions (s). It should also be noted that in some alternative implementations, the functions noted in the blocks can occur out of the order noted in the figures. For example, two blocks shown in succession can in fact be executed substantially concurrently or the blocks can sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each of the blocks of the flowchart and / or block diagrams and combinations of blocks in the flowchart and / or block diagrams can be implemented by special purpose hardware-based systems that perform the specified functions or operations, or combinations of special purpose hardware and

[0059] Embodiment one

[0060] The embodiment provides a model training method. The embodiment takes the method applied to a server as an example for illustration. It can be understood that the method can also be applied to a terminal, and can also be applied to a system including a terminal and a server and is realized through interaction of the terminal and the server. The server can be a stand-alone physical server, a server cluster or a distributed system formed by multiple physical servers, or a cloud server providing basic cloud computing services such as cloud service, cloud database, cloud computing, cloud function, cloud storage, network server, cloud communication, middleware service, domain name service, security service CDN, and big data and artificial intelligence platform. The terminal can be a smart phone, a tablet computer, a notebook computer, a desktop computer, and the like, but is not limited thereto. The terminal and the server can be directly or indirectly connected through wired or wireless communication, which is not limited in the application. In the embodiment, the method includes the following steps:

[0061] According to the data of the pathological section, the cell nucleus feature spectrum of the specific tumor area of the digital pathological section is obtained.

[0062] According to the transcriptomic data, a cohort gene expression matrix is obtained. According to the cohort gene expression matrix and the patient clinical information, the clinical diagnosis stage of the cancer tissue and normal tissue sample and the cancer sample is determined, and gene differential expression analysis is performed to obtain a differential gene expression spectrum of the specific tumor and a differential gene expression spectrum of the specific tumor stage.

[0063] The generative model is trained by taking the cell nucleus feature spectrum of the specific tumor area of the digital pathological section as input and taking the differential gene expression spectrum of the specific tumor and the differential gene expression spectrum of the specific tumor stage as output.

[0064] Specifically, the specific scheme of the embodiment can be realized by referring to the following content:

[0065] As Figure 1As shown, in order to construct the differential gene expression profile, first, the patient's transcriptomic raw sequencing data is quality controlled and pre-processed, including de-ligation and low-quality base removal; then the above quality-controlled data is aligned with the reference genome sequence, and sorted and indexed; further, the above processed data is subjected to read counting, expression quantification and standardization, and then the gene expression matrix of the cohort data is obtained; considering the redundancy of gene data of each sample, the above expression matrix needs to be pre-processed by filtering low-expression genes; in addition, in the tumor analysis work, tumor-related genes have clinical diagnostic and prognostic value, therefore, according to the patient's clinical information, the patent determines the clinical diagnostic stage of the cancer tissue and normal tissue sample and the cancer sample patient, and performs gene differential expression analysis; here, the differential gene determination principle of expression analysis is the difference fold (log2FC>1) or the test p value (p-value<0.05); finally, we obtain the differential gene set of a specific tumor and the differential gene set of a specific tumor stage, which respectively constitute the tumor differential gene expression profile TDGEP (Tumor-associated Differential Genes Expression Profile) and the tumor stage differential gene expression profile TSDGEP (Tumor Stage-associated Differential Genes Expression Profile) as the prediction target for subsequent modeling work.

[0066] The digital pathology section phenotype feature profile is used to input into the generative model to predict the above-constructed differential gene expression profile, which is a prerequisite for establishing the correlation between digital section phenotype and molecular level gene expression. As shown, Figure 2 As shown, first, we use the Otsu threshold method to filter the background from the full-scan digital pathology section and extract the tissue region; then, a tumor region recognition model is developed and trained using a tumor region labeling data set, and is applied to the above extracted tissue region to further extract the tumor region; next, a cell nucleus instance segmentation model is developed and trained based on a multi-class cell nucleus annotation data set, and is applied to the above extracted tumor region for cell nucleus instance segmentation; further, all the segmented cell nuclei are subjected to feature extraction and statistical quantification; finally, the cell nucleus feature profile TsNFP (Tumor-specific Nuclei Features Profile) of the specific tumor region of the digital pathology section is obtained. Among them, the tumor region recognition model and the cell nucleus instance segmentation model can use deep learning neural network models such as CNN, RNN, etc., but are not limited to a specific network or structure.

[0067] After constructing the nucleus feature spectrum TsNFP of a specific tumor region on a digital pathology section, the differential gene expression spectrum TDGEP of a specific tumor, and the differential gene expression spectrum TSDGEP of a specific tumor stage, this embodiment will model to predict the differential gene expression spectrum, realize the multi-omics correlation modeling of tumor from the phenotype of digital pathology section to the gene expression at the molecular level. Here, the nucleus feature spectrum is input into a deep network model for learning, and finally the differential gene expression spectrum is predicted.

[0068] Considering the effectiveness of mining the correlation between multi-omics data, a generative model of deep learning is used for modeling learning. The generative model can be a variational autoencoder, a generative adversarial network, a flow model, or a diffusion model. The purpose of using the generative model is to regard the nucleus feature spectrum of the digital pathology section phenotype as a data sampling point of the prior distribution, and then transform it into a new sampling point that satisfies the new data distribution of the molecular level gene expression spectrum through the generative model.

[0069] As shown in Figure 3 , assuming that the nucleus feature spectrum of the digital pathology section is an input vector x, and the differential gene expression spectrum is an output vector y. Then the key of all generative model modeling is that the input x is first transformed into the feature embedding h = f(x) of the intermediate layer, and then the feature embedding h is transformed into the output vector y to construct the loss , and then adjust the parameters of f(·) and g(·) to optimize the model. After training the generative model, at this time the parameters of f(·) and g(·) are fixed, and for a new digital pathology section nucleus feature spectrum input x', y ′ =g(f(x')) is the predicted tumor differential gene expression spectrum.

[0070] Embodiment two

[0071] The embodiment provides a model training system.

[0072] The model training system comprises:

[0073] The first data processing module is configured to obtain the nucleus feature spectrum of a specific tumor region of a digital pathology section according to the digital pathology section data;

[0074] The second data processing module is configured to obtain the cohort gene expression matrix according to the transcriptomics data at the gene level; determine the clinical diagnosis stage of the cancer tissue and normal tissue sample and the cancer sample according to the cohort gene expression matrix and the patient clinical information, and perform gene differential expression analysis to obtain the differential gene expression spectrum of a specific tumor and the differential gene expression spectrum of a specific tumor stage;

[0075] a training module configured to train the generative model by taking the nucleus feature spectrum of a specific tumor region of a digital pathology section as input, and taking the differential gene expression spectrum of the specific tumor and the differential gene expression spectrum of the specific tumor stage as output.

[0076] It should be noted that the first data processing module, the second data processing module and the training module described above have the same examples and application scenarios as the steps in Embodiment One, but are not limited to the content disclosed in Embodiment One. It should be noted that the modules described above can be executed in a computer system such as a set of computer executable instructions as part of a system.

[0077] Embodiment Three

[0078] The embodiment provides a tumor differential gene expression spectrum prediction method combined with pathological omics features.

[0079] The tumor differential gene expression spectrum prediction method combined with pathological omics features comprises:

[0080] According to the tumor region of the patient to be tested, the nucleus feature spectrum of the tumor region is extracted.

[0081] Based on the nucleus feature spectrum of the tumor region, a generative model is used to obtain the differential gene expression spectrum of the specific tumor and the differential gene expression spectrum of the specific tumor stage. The generative model is obtained by the model training method described in Embodiment One.

[0082] The specific scheme of the embodiment can be implemented by the following scheme:

[0083] As shown in Figure 4 , first, the transcriptomics data is processed and analyzed to obtain the differential gene expression spectrum, then the nucleus feature spectrum of the tumor region of all digital pathology sections in the cohort is extracted, and finally the nucleus feature spectrum of the digital pathology section phenotype is input into the generative model to model and predict the gene expression spectrum at the molecular level, and the correlation between the digital pathology section phenotype and the molecular gene expression is mined to realize the fusion representation of multiple omics features.

[0084] The data used in the embodiment is complete and sample-matched multi-omics data, which contains the clinical information of the patient and the paired digital pathology section and transcriptomics data, to ensure the extraction of the nucleus feature spectrum and the gene expression spectrum and the subsequent prediction.

[0085] For the above-mentioned well-learned generative model, not only the predicted output can be obtained according to the input, but also the feature embedding h of the intermediate layer can be obtained through f(.). Here, the feature embedding h is transformed from the input, and the feature embedding h can be transformed into the predicted output, that is, the transformation f(x') of the input and the inverse transformation g'(y') of the output are both the feature embedding h. Therefore, here, the constructed generative model establishes the correlation between the digital pathology section phenotype cell nucleus feature spectrum and the molecular level gene expression spectrum, and the obtained feature embedding h is the fusion representation of both, which can be further used to design downstream multi-omics fusion tumor analysis work.

[0086] The present application constructs a molecular level transcriptomic differential gene expression spectrum, and predicts the molecular level differential gene expression based on the digital section level phenotype modeling, which can greatly reduce the cost of transcriptome sequencing and improve the efficiency of subsequent clinical diagnosis and prognosis. On the other hand, the correlation between the digital pathology section macroscopic phenotype and the molecular level gene expression is explored in the modeling process, and a multi-omics fusion representation is performed, which provides more possibilities for precision medicine based on multi-omics multi-modal biological information tumor analysis.

[0087] The present application constructs a molecular level transcriptomic differential gene expression spectrum, which can quickly and accurately locate the tumor related gene set of the tumor patient.

[0088] The present application predicts the molecular level differential gene expression from the digital pathology section phenotype, which improves the subsequent diagnosis and prognosis.

[0089] The present application establishes the correlation between the digital pathology section phenotype cell nucleus feature spectrum and the molecular level differential gene expression spectrum, and performs multi-omics fusion representation, which provides more possibilities for precision medicine based on multi-omics multi-modal biological information tumor analysis.

[0090] Example four

[0091] The present embodiment provides a tumor differential gene expression spectrum prediction system combined with pathological omics features.

[0092] The tumor differential gene expression spectrum prediction system combined with pathological omics features comprises:

[0093] The feature extraction module is configured to extract the tumor region cell nucleus feature spectrum according to the tumor region of the patient to be tested.

[0094] The prediction module is configured to obtain the differential gene expression spectrum of a specific tumor and the differential gene expression spectrum of the specific tumor stage of the specific tumor based on the tumor region cell nucleus feature spectrum by using a generative model, and the generative model is obtained by the model training method of example one.

[0095] It should be noted that the feature extraction module and the prediction module described above have the same examples and application scenarios as the steps in Embodiment Three, but are not limited to the content disclosed in Embodiment Three. It should be noted that the modules described above can be executed in a computer system such as a set of computer executable instructions as part of a system.

[0096] Embodiment Five

[0097] The embodiment provides a computer readable storage medium, which stores a computer program, and the program is executed by a processor to implement the model training method in Embodiment One, or implement the steps in the tumor differential gene expression spectrum prediction method combined with pathological features in Embodiment Three.

[0098] Embodiment Six

[0099] The embodiment provides a computer device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the model training method in Embodiment One, or implements the steps in the tumor differential gene expression spectrum prediction method combined with pathological features in Embodiment Three when executing the program.

[0100] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can be in the form of a hardware embodiment, a software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can be in 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 application is described with reference to flowcharts and / or block diagrams of the method, device (system), and computer program product according to the embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of the flows 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 general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to produce a machine, so that the instructions executed by the computer or other programmable data processing devices produce a device that implements the functions specified in the flowcharts and / or block diagrams. Figure 1 The system that implements the functions specified in one or more flows and / or blocks Figure 1 The system that implements the functions specified in one or more flows and / or blocks

[0102] These computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer readable memory produce an article of manufacture including instructions that implement the flow Figure 1 one or more flow or block Figure 1 one or more blocks or blocks specified in the flow.

[0103] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions executed on the computer or other programmable apparatus provide steps for implementing the flow Figure 1 one or more flow or block Figure 1 one or more blocks or blocks specified in the flow.

[0104] Those of ordinary skill in the art can understand that all or part of the flow of the above-mentioned embodiment method can be completed by computer program instructions to instruct related hardware, and the program can be stored in a computer readable storage medium. When the program is executed, it can include the flow of the above-mentioned embodiment method. The storage medium can be a magnetic disc, an optical disc, a read-only memory (ROM) or a random access memory (RAM), etc.

[0105] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. Those skilled in the art can make various modifications and changes to the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. Model training method, characterized in that, include: According to the digital pathology section data, the cell nuclear characteristic spectrum of the specific tumor area of ​​the digital pathology section is obtained; Based on the transcriptomic data, the cohort gene expression matrix was obtained; The process of obtaining the cohort gene expression matrix includes: Performing quality control and preprocessing on the transcriptomics data to obtain the first transcriptomics data; Comparing the first transcriptomic data with the reference genome sequence, sorting and indexing the data to obtain the second transcriptomic data; Perform read processing, expression quantification, and normalization on the second transcriptomics data to obtain the cohort gene expression matrix; Based on the cohort gene expression matrix and patient clinical information, the clinical diagnostic stages of cancer tissue and normal tissue samples and cancer samples were determined, and differential gene expression analysis was performed to obtain differential gene expression profiles for specific tumors and differential gene expression profiles for specific tumor stages; The generative model is trained using the nuclear characteristic spectrum of a specific tumor area in a digital pathology section as input and the differential gene expression spectrum of the specific tumor and the differential gene expression spectrum of the specific tumor stage as output. The constructed generative model establishes an association between the phenotypic nuclear characteristic spectrum of the digital pathology section and the gene expression spectrum at the molecular level.

2. The model training method according to claim 1, characterized in that The process of obtaining the cell nuclear characteristic spectrum of a specific tumor area in a digital pathology section includes: Acquire digital pathology slide data and extract tissue areas; Extract tumor regions based on tissue regions; Perform cell nucleus instance segmentation on the tumor area to obtain all cell nuclei; Based on all cell nuclei, cell nuclear features are extracted to obtain the cell nuclear feature spectrum of the specific tumor area in the digital pathology section.

3. The model training method according to claim 1, characterized in that The process of training a generative model includes: The cell nucleus feature spectrum of a specific tumor area in a digital pathology section is first transformed into feature embedding, and the output result is obtained from the feature embedding; Based on the output results, the differential gene expression profile of the specific tumor and the differential gene expression profile of the specific tumor stage, a loss function is constructed to optimize the generative model.

4. The model training method according to claim 3, characterized in that The fusion of the nuclear signature profile with the differential gene expression profile of a specific tumor and the differential gene expression profile of the specific tumor stage is represented as a signature embedding.

5. A model training system, characterized in that: include: A first data processing module is configured to: obtain a cell nuclear characteristic spectrum of a specific tumor area of ​​a digital pathology section according to the digital pathology section data; The second data processing module is configured to: obtain a cohort gene expression matrix based on the transcriptomics data at the gene layer; The process of obtaining the cohort gene expression matrix includes: Performing quality control and preprocessing on the transcriptomics data to obtain the first transcriptomics data; Comparing the first transcriptomic data with the reference genome sequence, sorting and indexing the data to obtain the second transcriptomic data; Perform read processing, expression quantification, and normalization on the second transcriptomics data to obtain the cohort gene expression matrix; Based on the cohort gene expression matrix and patient clinical information, the clinical diagnostic stages of cancer tissue and normal tissue samples and cancer samples were determined, and differential gene expression analysis was performed to obtain differential gene expression profiles for specific tumors and differential gene expression profiles for specific tumor stages; The training module is configured to: use the cell nuclear characteristic spectrum of a specific tumor area in a digital pathology section as input, and use the differential gene expression spectrum of the specific tumor and the differential gene expression spectrum of the specific tumor stage as output, to train a generative model, and the constructed generative model establishes an association between the phenotypic cell nuclear characteristic spectrum of the digital pathology section and the gene expression spectrum at the molecular level.

6. A method for predicting tumor differential gene expression profiles based on combined pathological omics features, characterized in that: include: According to the tumor area of ​​the patient to be tested, extract the characteristic spectrum of cell nuclei in the tumor area; Based on the characteristic spectrum of cell nuclei in the tumor area, a generative model is used to obtain the differential gene expression spectrum of a specific tumor and the differential gene expression spectrum of the specific tumor stage; the generative model is obtained by the model training method described in any one of claims 1-4.

7. A tumor differential gene expression profile prediction system combined with pathological omics features, characterized by: include: A feature extraction module is configured to: extract a characteristic spectrum of cell nuclei in the tumor region according to the tumor region of the patient to be tested; A prediction module is configured to: based on the nuclear characteristic spectrum of the tumor region, use a generative model to obtain the differential gene expression spectrum of a specific tumor and the differential gene expression spectrum of the specific tumor stage; the generative model is obtained by the model training method according to any one of claims 1 to 4.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the model training method according to any one of claims 1 to 4 is implemented, or the steps in the method for predicting tumor differential gene expression profiles based on combined pathological genomic features according to claim 6 are implemented.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, it implements the model training method according to any one of claims 1 to 4, or implements the steps in the method for predicting tumor differential gene expression profiles based on combined pathological genomic features according to claim 6.

Images