Omics data difference evaluation system based on multi-group comparison

Abstract

The invention provides an omics data difference evaluation system based on multi-group comparison, and relates to the technical field of biological information analysis, and the system comprises a data processing unit which carries out the preprocessing of the omics data of three or more spatial transcriptomes, and obtains a first spatial region gene expression profile; the splitting unit is used for splitting the first spatial region gene expression profile according to a spatial domain to obtain a plurality of second spatial region gene expression profiles; the difference calculation unit is used for establishing a multi-classification prediction model based on the multi-layer perceptron and calculating a state difference score of each spatial structure domain by the multi-classification prediction model; and the evaluation unit performs sorting according to the state difference scores from high to low, and the higher the state difference scores are, the larger the difference degree of the corresponding spatial structure domains is. The method is not limited to differences of single features (such as genes and proteins) and is not limited by specific omics, spatial structure domain differences among multiple groups can be evaluated from the overall level, and comparability among omics data is improved.

Description

Technical Field

[0001] The present application relates to the technical field of bioinformatics analysis, and in particular to a system for evaluating omics data differences based on multi-group comparison. Background Art

[0002] With the development of science and technology, current omics technologies can obtain the molecular composition and dynamic changes of the research objects through high-throughput technology platforms. Each sample can usually obtain the numerical values ​​of hundreds or even tens of thousands of features. For example, a sample of transcriptomics can obtain the expression levels of tens of thousands of genes, and a sample of proteomics can obtain the abundance of hundreds to thousands of proteins. Some omics can even obtain multi-dimensional, multi-level, and high-resolution information, such as single-cell transcriptomes, which can not only distinguish the expression levels of genes but also the expression of each cell in a sample; or spatial transcriptomes, which can obtain the number of transcripts at each site on a tissue section, that is, omics data has the characteristics of high dimensionality and complexity.

[0003] In omics analysis, intergroup comparisons are crucial for achieving research objectives. Their significance lies in revealing underlying biological mechanisms or functional associations by comparing multidimensional features across populations or experimental conditions. Multi-group comparisons are crucial. For example, in disease research, comparing various biological functions (such as biological signaling, metabolism, and cell types) across healthy, diseased, and treated groups can reveal which vital activities or cellular components exhibit significant differences during disease progression and treatment. Currently, methods such as fold difference and significance are commonly used to assess the magnitude of differences. However, fold difference only applies to comparisons between two groups, while significance essentially reflects the reliability of differences and does not directly assess the magnitude of differences. Furthermore, in the multi-level spatial transcriptome, conventional difference detection methods are primarily used to compare functional differences between individual genes or gene sets and cannot be used to assess global differences in expression profiles. Summary of the Invention

[0004] The present application provides an omics data difference assessment system based on multi-group comparison to at least solve the above technical problems existing in the prior art.

[0005] According to a first aspect of the present application, a system for evaluating omics data differences based on multi-group comparison is provided, comprising: The data processing unit pre-processes the omics data of three or more spatial transcriptomes and annotates the spatial domain to which each spatial site belongs to obtain the gene expression profile of the first spatial region; a splitting unit, splitting the first spatial region gene expression profile according to spatial structural domains to obtain a plurality of second spatial region gene expression profiles; A difference calculation unit, which establishes a multi-classification prediction model based on a multi-layer perceptron, and calculates the state difference score of each spatial structure domain by the multi-classification prediction model; The evaluation unit sorts the data from high to low according to the state difference score. The higher the state difference score, the greater the degree of difference in the corresponding spatial structure domain.

[0006] In certain embodiments of the first aspect of the present application, the preprocessing includes data normalization.

[0007] In certain embodiments of the first aspect of the present application, after splitting the first spatial region gene expression profile to obtain several second spatial region gene expression profiles, PCA dimensionality reduction is performed on the second spatial region gene expression profile to obtain a PC matrix corresponding to each spatial domain.

[0008] In certain embodiments of the first aspect of the present application, the state difference score is calculated as follows: Randomly splitting the PC matrix into a second number of sub-matrices according to spatial sites; One of the second number of sub-matrices is used as a validation set, and the remaining sub-matrices are used as training sets; PC is used as a feature, the PC value is used as the eigenvalue, and the grouping is used as classification information, and a classification prediction model is established using a multi-layer perceptron and trained; the classification prediction model is used to predict the grouping of each spatial site in the validation set; based on the predicted grouping and the actual grouping of the validation set, the AUC value is calculated as the state difference score of the spatial structure domain corresponding to the PC matrix.

[0009] In certain embodiments of the first aspect of the present application, the state difference score is calculated as follows: Randomly splitting the PC matrix into a second number of sub-matrices according to spatial sites; Using one of the second number of sub-matrices as a validation set and the remaining sub-matrices as a training set; using PC as a feature, PC value as a eigenvalue, and grouping as classification information, using a multi-layer perceptron to establish a classification prediction model and perform training; using the classification prediction model to predict the grouping of each spatial site in the validation set; and calculating the AUC value based on the predicted grouping and the actual grouping of the validation set, as the AUC value of the sub-matrix of the validation set; The remaining submatrices of the second number of submatrices are used as validation sets in turn, and the AUC value of each submatrix is ​​calculated; The average of the AUC values ​​of all sub-matrices is calculated as the state difference score of the spatial structure domain corresponding to the PC matrix.

[0010] In certain embodiments of the first aspect of the present application, the second number is 10.

[0011] According to a second aspect of the present application, a system for evaluating omics data differences based on multi-group comparison is provided, comprising: The data processing unit pre-processes the omics data of three or more single-cell transcriptomes and annotates the cell type of each cell to obtain the first single-cell gene expression profile; a splitting unit, splitting the first single-cell gene expression profile according to cell type to obtain a plurality of second single-cell gene expression profiles; a difference calculation unit, which establishes a multi-classification prediction model based on a multi-layer perceptron, and calculates the state difference score of each cell type using the multi-classification prediction model; The evaluation unit sorts the cells according to the state difference scores from high to low, wherein the higher the state difference score is, the greater the degree of difference of the corresponding cell types is.

[0012] In certain embodiments of the second aspect of the present application, the state difference score of each cell type is calculated as follows: Perform PCA dimensionality reduction on the second single-cell gene expression profile to obtain the PC matrix corresponding to each cell type; randomly splitting the PC matrix into a second number of sub-matrices by cells; Using one of the second number of sub-matrices as a validation set and the remaining sub-matrices as a training set; using PCs as features, PC values ​​as eigenvalues, and grouping as classification information, using a multi-layer perceptron to establish a classification prediction model and perform training; using the classification prediction model to predict the grouping of each cell in the validation set; and calculating the AUC value based on the predicted grouping and the actual grouping of the validation set, as the AUC value of the sub-matrix of the validation set; The remaining submatrices of the second number of submatrices are used as validation sets in turn, and the AUC value of each submatrix is ​​calculated; The average of the AUC values ​​of all sub-matrices is calculated as the state difference score of the cell type corresponding to the PC matrix.

[0013] In certain embodiments of the second aspect of the present application, the preprocessing includes data normalization.

[0014] In certain embodiments of the second aspect of the present application, the second number is 10.

[0015] According to a third aspect of the present application, a system for evaluating omics data differences based on multi-group comparison is provided, comprising: The data processing unit pre-processes the omics data of three or more plasma proteomes, annotates the functions to be compared, and the proteins corresponding to the functions, and obtains the first protein abundance spectrum; a splitting unit, splitting the first protein abundance profile according to function to obtain a plurality of second protein abundance profiles; a difference calculation unit, which establishes a multi-classification prediction model based on a multi-layer perceptron and calculates a state difference score of each function using the multi-classification prediction model; The evaluation unit sorts the status difference scores from high to low, wherein the higher the status difference score is, the greater the degree of difference in the corresponding function is.

[0016] In certain embodiments of the third aspect of the present application, the state difference score of each function is calculated as follows: Perform PCA dimensionality reduction on the second protein abundance spectrum to obtain the PC matrix corresponding to each function; Randomly splitting the PC matrix into a second number of sub-matrices according to samples; Using one of the second number of sub-matrices as a validation set and the remaining sub-matrices as a training set; using PCs as features, PC values ​​as eigenvalues, and grouping as classification information, establishing a classification prediction model using a multi-layer perceptron and performing training; using the classification prediction model to predict the grouping of each sample in the validation set; and calculating the AUC value based on the predicted grouping and the actual grouping of the validation set, as the AUC value of the sub-matrix of the validation set; The remaining submatrices of the second number of submatrices are used as validation sets in turn, and the AUC value of each submatrix is ​​calculated; The average of the AUC values ​​of all sub-matrices is calculated as the state difference score of the function corresponding to the PC matrix.

[0017] In certain embodiments of the third aspect of the present application, the preprocessing includes data normalization.

[0018] In certain embodiments of the third aspect of the present application, the second number is 10.

[0019] Compared with the prior art, this application has the following beneficial effects: This application uses a multi-layer perceptron to build a multi-classification model for multiple sets of spatial transcriptome data. Using model prediction accuracy as an evaluation metric for the level of differentiation, this application obtains results on the overall magnitude of differences across different spatial domains. Compared to existing technologies, the technical solution provided by this application is not limited to differences in individual features (such as genes or proteins) or to specific omics. It can assess spatial domain differences across multiple groups at a holistic level, increasing the comparability of omics data.

[0020] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present application, nor is it intended to limit the scope of the present application. Other features of the present application will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The above and other objects, features and advantages of the exemplary embodiments of the present application will become readily understood by reading the detailed description below with reference to the accompanying drawings. In the accompanying drawings, several embodiments of the present application are shown in an illustrative and non-limiting manner, in which: In the drawings, the same or corresponding reference numerals denote the same or corresponding parts.

[0022] Figure 1 A schematic diagram of the system structure of this application is shown. DETAILED DESCRIPTION

[0023] In order to make the purpose, features, and advantages of this application more obvious and easy to understand, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative efforts shall fall within the scope of protection of this application.

[0024] Example 1: Please refer to Figure 1 , this embodiment 1 provides an omics data difference evaluation system based on multi-group comparison, including a data processing unit, a splitting unit, a difference calculation unit and an evaluation unit.

[0025] The data processing unit preprocesses the omics data of three or more spatial transcriptomes, including data normalization, to reduce the influence of the numerical range of each feature, and adds necessary annotations to the data.

[0026] This method targets the spatial transcriptome and requires normalization of the gene expression profiles using Seurat software. To compare the magnitude of overall state changes across different spatial domains, the spatial domain to which each spatial locus belongs must be identified and annotated to obtain the gene expression profile for the first spatial region. Table 1 below shows the gene expression profile for a specific spatial region after normalization and standardization.

[0027] Table 1: Spatial regional gene expression profiles

[0028] The splitting unit splits the first spatial region gene expression profile according to the spatial structure domain to obtain a plurality of second spatial region gene expression profiles.

[0029] And PCA dimensionality reduction is performed on the gene expression spectrum of the second spatial region to obtain a PC matrix corresponding to each spatial structural domain. Specifically, after PCA dimensionality reduction, only the first first number of PCs is retained to generate a PC matrix corresponding to a single cell type, and the first first number is preferably 30.

[0030] Principal Component Analysis (PCA) is an unsupervised dimensionality reduction method that projects high-dimensional data into a low-dimensional orthogonal space (principal components, or PCs) through linear transformation, preserving information with the highest variance. Its core is the eigenvalue decomposition of the covariance matrix, which achieves data compression by selecting the top principal components (those with the highest contribution). The PCA algorithm itself has advantages such as removing noise, reducing redundant features, alleviating the curse of dimensionality, and eliminating inter-feature correlation.

[0031] Since the gene expression profiles of organisms may interfere with each other (i.e., have linear correlation), PCA dimensionality reduction can effectively remove the interference between eigenvalues ​​during data analysis, and at the same time remove the noise effects caused by experimental techniques, sample status, system errors, etc., thereby reducing data complexity.

[0032] The difference calculation unit establishes a multi-classification prediction model based on a multi-layer perceptron, and calculates the state difference score of each spatial structure domain by the multi-classification prediction model.

[0033] The state difference score is calculated using the following two methods: First method: The PC matrix is ​​randomly split into a second number of sub-matrices according to the spatial sites; it is worth mentioning that the second number is preferably 10, that is, it is split into 10 sub-matrices, which are respectively recorded as M1, M2, ..., M 10 .

[0034] One of the second number of sub-matrices is used as a validation set, and the remaining sub-matrices are used as training sets. For example, M1 is used as a validation set, and M2~M 10 As a training set; M2 can also be used as a validation set, and M1, M3~M 10 As a training set.

[0035] Using PCs as features, PC values ​​as eigenvalues, and grouping as classification information, a multilayer perceptron (MLP) was used to build and train a classification prediction model. The MLP is a classic feedforward artificial neural network consisting of an input layer, one or more hidden layers, and an output layer. It is one of the foundational models of deep learning. By simulating the connectivity of biological neurons, utilizing nonlinear activation functions and a multilayer structure, it is capable of learning complex input-output mappings without being restricted by linear separability. MLPs can be used to build multi-classification models. They can predict the classification based on the differences in input data, and the accuracy of their predictions also reflects the discriminability of the input data.

[0036] The classification prediction model is used to predict the grouping of each spatial site in the validation set; based on the predicted grouping and the actual grouping of the validation set, the AUC value is calculated as the state difference score of the spatial structure domain corresponding to the PC matrix.

[0037] The AUC value is the area under the ROC curve, that is, the area under the receiver operating characteristic curve, which is used to quantify the model's ability to sort positive and negative samples.

[0038] Second method: Randomly splitting the PC matrix into a second number of sub-matrices according to spatial sites; Using one of the second number of sub-matrices as a validation set and the remaining sub-matrices as a training set; using PC as a feature, PC value as a eigenvalue, and grouping as classification information, using a multi-layer perceptron to establish a classification prediction model and perform training; using the classification prediction model to predict the grouping of each spatial site in the validation set; and calculating the AUC value based on the predicted grouping and the actual grouping of the validation set, as the AUC value of the sub-matrix of the validation set; The remaining sub-matrices of the second number of sub-matrices are used as validation sets in turn, and the AUC value of each sub-matrix is ​​obtained. For example, when sub-matrix M1 is used as the validation set, M2~M 10 When used as a training set, the calculated AUC value is the AUC value of the sub-matrix M1, recorded as A1; similarly, M2 is used as a validation set, M1, M3~M 10 When used as a training set, the calculated AUC value is the AUC value of submatrix M2, denoted as A2.

[0039] Calculate the AUC values ​​of all sub-matrices A1~A 10 The average value of is taken as the state difference score of the spatial structure domain corresponding to the PC matrix.

[0040] Unlike the first method, the second method uses each submatrix as a validation set and calculates the corresponding AUC value, which can eliminate the impact of individual abnormal data on the state difference score. For example, when the first method contains abnormal data in submatrix M1 and uses it as the validation set, the AUC value calculated based on the predicted grouping and the actual grouping of the validation set will also show obvious abnormalities and cannot reflect the actual difference state.

[0041] The evaluation unit sorts the data from high to low according to the state difference score. The higher the state difference score, the greater the degree of difference in the corresponding spatial structure domain.

[0042] Example 2: The second embodiment provides an omics data difference assessment system based on multi-group comparison, specifically a single-cell transcriptome data difference assessment system based on multi-group comparison, including a data processing unit, a splitting unit, a difference calculation unit and an assessment unit.

[0043] The data processing unit preprocesses the omics data from three or more single-cell transcriptomes. This method, specific to single-cell transcriptomes, requires normalization and standardization of cell expression profiles using Seurat software. If comparing the magnitude of overall state changes across different cell types, each cell's cell type (e.g., epithelial, B, T, etc.) must be annotated to obtain the first single-cell gene expression profile. Table 2 below shows an example of the first single-cell gene expression profile after normalization and standardization.

[0044] Table 2: First single-cell gene expression profile

[0045] A splitting unit is configured to split the first single-cell gene expression profile according to cell type to obtain a plurality of second single-cell gene expression profiles.

[0046] The difference calculation unit establishes a multi-classification prediction model based on a multi-layer perceptron and calculates the state difference score of each cell type using the multi-classification prediction model. The specific calculation method is as follows: PCA dimensionality reduction was performed on the second single-cell gene expression profile, retaining only the top 30 PCs after PCA dimensionality reduction to obtain a PC matrix corresponding to each cell type. Table 3 below shows an example of a PC matrix for a B cell, retaining the top 30 PCs after PCA dimensionality reduction.

[0047] Table 3: PC matrix after PCA dimensionality reduction

[0048] The PC matrix is ​​randomly split into a second number of sub-matrices according to cells. It is worth mentioning that the second number is preferably 10, that is, it is split into 10 sub-matrices, which are respectively recorded as M1, M2, ..., M 10 .

[0049] One of the second number of sub-matrices is used as a validation set, and the remaining sub-matrices are used as training sets. Table 4 below shows an example of the sub-matrix M1 used as a validation set.

[0050] Table 4: Submatrix M1 as validation set

[0051] Using PC as a feature, PC value as a eigenvalue, and grouping as classification information, a multi-layer perceptron is used to establish a classification prediction model and train it; using the classification prediction model to predict the grouping of each cell in the validation set; based on the predicted grouping and the actual grouping of the validation set, the AUC value is calculated as the AUC value of the submatrix of the validation set; The remaining submatrices of the second number of submatrices are used as validation sets in turn, and the AUC value of each submatrix is ​​calculated; The average of the AUC values ​​of all sub-matrices is calculated as the state difference score of the cell type corresponding to the PC matrix. The detailed calculation process and principle are referred to the second calculation method of the state difference score in Example 1, which will not be repeated here.

[0052] The evaluation unit sorts the cells according to the state difference scores from high to low, wherein the higher the state difference score is, the greater the degree of difference of the corresponding cell types is.

[0053] Example 3: This embodiment 3 provides a system for evaluating omics data differences based on multiple group comparisons, specifically a system for evaluating plasma proteomics data differences based on multiple group comparisons, including: The data processing unit preprocesses the omics data for three or more plasma proteomes. This method, specific to plasma proteomes, requires logarithmic processing of protein abundance data and normalization of the values. To compare overall differences across functions, the functions to be compared and the corresponding proteins are annotated to generate a first protein abundance profile. Table 5 below shows a normalized protein abundance profile.

[0054] Table 5: Protein abundance profile

[0055] A splitting unit is configured to split the first protein abundance spectrum according to function to obtain a plurality of second protein abundance spectrums.

[0056] The difference calculation unit establishes a multi-classification prediction model based on a multi-layer perceptron, and calculates the state difference score of each function using the multi-classification prediction model; the specific calculation method is as follows: Perform PCA dimensionality reduction on the second protein abundance spectrum to obtain the PC matrix corresponding to each function; Randomly splitting the PC matrix into a second number of sub-matrices according to samples; Using one of the second number of sub-matrices as a validation set and the remaining sub-matrices as a training set; using PCs as features, PC values ​​as eigenvalues, and grouping as classification information, establishing a classification prediction model using a multi-layer perceptron and performing training; using the classification prediction model to predict the grouping of each sample in the validation set; and calculating the AUC value based on the predicted grouping and the actual grouping of the validation set, as the AUC value of the sub-matrix of the validation set; The remaining submatrices of the second number of submatrices are used as validation sets in turn, and the AUC value of each submatrix is ​​calculated; Calculate the average of the AUC values ​​of all sub-matrices as the state difference score of the function corresponding to the PC matrix. For the detailed calculation process and principle, please refer to the second calculation method of the state difference score in Example 1, which will not be repeated here.

[0057] The evaluation unit sorts the status difference scores from high to low, wherein the higher the status difference score is, the greater the degree of difference in the corresponding function is.

[0058] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this application can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solution of this application can be achieved. This is not limited herein.

[0059] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features being referred to. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one such feature. Throughout the description of this application, "plurality" means two or more, unless otherwise specifically defined.

[0060] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A system for evaluating omics data differences based on multi-group comparison, characterized in that: include: The data processing unit pre-processes the omics data of three or more spatial transcriptomes and annotates the spatial domain to which each spatial site belongs to obtain the gene expression profile of the first spatial region; a splitting unit, splitting the first spatial region gene expression profile according to spatial structural domains to obtain a plurality of second spatial region gene expression profiles; A difference calculation unit, which establishes a multi-classification prediction model based on a multi-layer perceptron, and calculates the state difference score of each spatial structure domain by the multi-classification prediction model; The evaluation unit sorts the data from high to low according to the state difference score. The higher the state difference score, the greater the degree of difference in the corresponding spatial structure domain.

2. The omics data difference evaluation system based on multi-group comparison according to claim 1, characterized in that: The preprocessing includes data normalization.

3. The omics data difference assessment system based on multi-group comparison according to claim 1, characterized in that: After splitting the first spatial region gene expression profile to obtain a plurality of second spatial region gene expression profiles, PCA dimensionality reduction is performed on the second spatial region gene expression profile to obtain a PC matrix corresponding to each spatial domain.

4. The omics data difference assessment system based on multi-group comparison according to claim 3, characterized in that: The state difference score is calculated as follows: Randomly splitting the PC matrix into a second number of sub-matrices according to spatial sites; One of the second number of sub-matrices is used as a validation set, and the remaining sub-matrices are used as training sets; PC is used as a feature, the PC value is used as the eigenvalue, and the grouping is used as classification information, and a classification prediction model is established using a multi-layer perceptron and trained; the classification prediction model is used to predict the grouping of each spatial site in the validation set; based on the predicted grouping and the actual grouping of the validation set, the AUC value is calculated as the state difference score of the spatial structure domain corresponding to the PC matrix.

5. The omics data difference assessment system based on multi-group comparison according to claim 3, characterized in that: The state difference score is calculated as follows: Randomly splitting the PC matrix into a second number of sub-matrices according to spatial sites; Using one of the second number of sub-matrices as a validation set and the remaining sub-matrices as a training set; using PC as a feature, PC value as a eigenvalue, and grouping as classification information, using a multi-layer perceptron to establish a classification prediction model and perform training; using the classification prediction model to predict the grouping of each spatial site in the validation set; and calculating the AUC value based on the predicted grouping and the actual grouping of the validation set, as the AUC value of the sub-matrix of the validation set; The remaining submatrices of the second number of submatrices are used as validation sets in turn, and the AUC value of each submatrix is ​​calculated; The average of the AUC values ​​of all sub-matrices is calculated as the state difference score of the spatial structure domain corresponding to the PC matrix.

6. The omics data difference assessment system based on multi-group comparison according to claim 5, characterized in that: The second number is 10.

7. A system for evaluating omics data differences based on multi-group comparison, characterized in that: include: The data processing unit pre-processes the omics data of three or more single-cell transcriptomes and annotates the cell type of each cell to obtain the first single-cell gene expression profile; a splitting unit, splitting the first single-cell gene expression profile according to cell type to obtain a plurality of second single-cell gene expression profiles; a difference calculation unit, which establishes a multi-classification prediction model based on a multi-layer perceptron, and calculates the state difference score of each cell type using the multi-classification prediction model; The evaluation unit sorts the cells according to the state difference scores from high to low, wherein the higher the state difference score is, the greater the degree of difference of the corresponding cell types is.

8. The omics data difference assessment system based on multi-group comparison according to claim 7, characterized in that: The state difference score for each cell type is calculated as follows: Perform PCA dimensionality reduction on the second single-cell gene expression profile to obtain the PC matrix corresponding to each cell type; randomly splitting the PC matrix into a second number of sub-matrices by cells; Using one of the second number of sub-matrices as a validation set and the remaining sub-matrices as a training set; using PCs as features, PC values ​​as eigenvalues, and grouping as classification information, using a multi-layer perceptron to establish a classification prediction model and perform training; using the classification prediction model to predict the grouping of each cell in the validation set; and calculating the AUC value based on the predicted grouping and the actual grouping of the validation set, as the AUC value of the sub-matrix of the validation set; The remaining submatrices of the second number of submatrices are used as validation sets in turn, and the AUC value of each submatrix is ​​calculated; The average of the AUC values ​​of all sub-matrices is calculated as the state difference score of the cell type corresponding to the PC matrix.

9. A system for evaluating omics data differences based on multi-group comparison, characterized in that: include: The data processing unit pre-processes the omics data of three or more plasma proteomes, annotates the functions to be compared, and the proteins corresponding to the functions, and obtains the first protein abundance spectrum; a splitting unit, splitting the first protein abundance profile according to function to obtain a plurality of second protein abundance profiles; a difference calculation unit, which establishes a multi-classification prediction model based on a multi-layer perceptron and calculates a state difference score of each function using the multi-classification prediction model; The evaluation unit sorts the status difference scores from high to low, wherein the higher the status difference score is, the greater the degree of difference in the corresponding function is.

10. The omics data difference assessment system based on multi-group comparison according to claim 9, characterized in that: The state difference score for each feature is calculated as follows: Perform PCA dimensionality reduction on the second protein abundance spectrum to obtain the PC matrix corresponding to each function; Randomly splitting the PC matrix into a second number of sub-matrices according to samples; Using one of the second number of sub-matrices as a validation set and the remaining sub-matrices as a training set; using PCs as features, PC values ​​as eigenvalues, and grouping as classification information, establishing a classification prediction model using a multi-layer perceptron and performing training; using the classification prediction model to predict the grouping of each sample in the validation set; and calculating the AUC value based on the predicted grouping and the actual grouping of the validation set, as the AUC value of the sub-matrix of the validation set; The remaining submatrices of the second number of submatrices are used as validation sets in turn, and the AUC value of each submatrix is ​​calculated; The average of the AUC values ​​of all sub-matrices is calculated as the state difference score of the function corresponding to the PC matrix.

Images