A diabetes classification method and system based on big data

Abstract

The present invention discloses a diabetes typing method and system based on big data, which relates to the technical field of medical data processing; the method comprises collecting multi-dimensional data of patients and performing standardized processing; fusing the standardized multi-source data and the extracted features to generate a multi-dimensional feature vector; based on omics technology, deeply analyzing the individualized biomarkers of patients and extracting specific omics features; combining the multi-dimensional feature vector and the omics features to construct a diabetes typing model; after the diabetes typing model is trained and verified, the diabetes typing model is used to input the data of new patients to obtain personalized typing results, and personalized treatment suggestions are proposed in combination with the specific characteristics of the patients; the present invention can not only improve the accuracy of typing, but also provide personalized treatment suggestions for patients.

Description

Technical Field

[0001] The present invention relates to the technical field of medical data processing, and in particular to a diabetes typing method and system based on big data. Background Art

[0002] The management and treatment of diabetes has undergone significant changes with the development of medicine and bioinformatics. Traditional diabetes classification methods mainly rely on the analysis of clinical indicators and biochemical data, which often ignore the diversity and complexity of individual patients, resulting in insufficient accuracy of classification.

[0003] In addition, the lack of standardized processes in the feature extraction and fusion process leads to poor comparability of results between different studies, thus limiting the promotion and application of the technology. Summary of the invention

[0004] In view of the fact that the existing diabetes typing technology has many problems in terms of analysis methods, etc., the present invention proposes a diabetes typing method and system based on big data.

[0005] Therefore, the problem to be solved by the present invention is how to improve the accuracy of diabetes typing.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In the first aspect, an embodiment of the present invention provides a diabetes typing method based on big data, which includes collecting multi-dimensional data of patients and performing standardization processing; fusing the standardized multi-source data and the extracted features to generate a multi-dimensional feature vector; based on omics technology, conducting in-depth analysis of the patient's individualized biomarkers to extract specific omics features; combining the multi-dimensional feature vector and the omics features to construct a diabetes typing model; after the diabetes typing model is trained and verified, the diabetes typing model is used to input the data of new patients to obtain personalized typing results, and personalized treatment recommendations are made based on the patient's specific characteristics.

[0008] As a preferred embodiment of the diabetes typing method based on big data described in the present invention, the multi-dimensional data of the patient includes electronic medical records, genomic data, metabolic indicators, diet and exercise records.

[0009] As a preferred embodiment of the diabetes typing method based on big data described in the present invention, the in-depth analysis of the patient's personalized biomarkers based on omics technology and the extraction of specific omics features include the following steps: selecting metabolomics, genomics and proteomics data as omics data sources based on the patient's feature vector after data fusion to ensure that each type of omics data corresponds to a unique identifier; performing time series dynamic change analysis on the metabolomics data to analyze the changing trends of specific metabolic pathways at different time points and identify abnormal pathways in metabolic activities; performing specific site screening on the genomic data and screening out high-risk gene variation sites known to be associated with diabetes by querying a pre-set diabetes-related gene library; at the proteomics level, identifying key proteins related to diabetes based on the expression levels of specific proteins; and generating a personalized omics feature vector for the patient based on the key features of metabolism, genes and proteomics.

[0010] As a preferred embodiment of the diabetes typing method based on big data of the present invention, the selection of metabolomics, genomics and proteomics data as omics data sources includes: for each omics data item, calculating the correlation with diabetes risk and the patient's current health status: assuming that the diabetes risk assessment score is , the patient's health status feature vector is H, then a certain group feature The correlation calculation formula is:

[0011] ,

[0012] in, Indicates the correlation of a certain omics feature with diabetes risk and health status; Representation characteristics and diabetes risk and health status The covariance of and are the standard deviations of omics characteristics and diabetes risk and health status, respectively.

[0013] As a preferred solution of the diabetes typing method based on big data described in the present invention, wherein: the time series dynamic change analysis of the metabolomics data, the change trend of a specific metabolic pathway at different time points, and the identification of abnormal pathways in metabolic activities include the following steps: according to a known metabolic pathway database, the metabolomics features To map metabolic pathways, each metabolic feature Belong to one or more metabolic pathways, let the metabolic pathway be , and construct the corresponding relationship between metabolic characteristics and pathways; for each metabolic characteristic , analyze the dynamic change trend by calculating the change rate at different time points; calculate the change rate of all metabolic characteristics to form a complete time series dynamic change matrix ; For each metabolic pathway , calculate the total rate of change of the included metabolic features:

[0014] ,

[0015] in, The time series dynamic change matrix If the change rate of a metabolic pathway exceeds the preset abnormal threshold τ, the pathway is marked as an abnormal pathway. For the metabolic pathway marked as abnormal, its correlation with the progression of diabetes is further analyzed.

[0016] As a preferred embodiment of the diabetes typing method based on big data described in the present invention, the screening of high-risk gene variation sites known to be associated with diabetes includes: assuming that the genomic data consists of multiple single nucleotide polymorphism (SNP) sites, each SNP site corresponds to a gene variation; defining a genomic data matrix G, in which each row corresponds to a SNP site, and each column corresponds to a patient's sample; first establishing a diabetes-related gene library, each gene variation site in the diabetes-related gene library has a known association with diabetes risk; in the patient's genomic data matrix G, by comparing with the diabetes-related gene library, screening out high-risk SNP sites associated with diabetes; the screening condition is: if a certain SNP of the patient exists in the diabetes-related gene library, then the SNP is marked as a high-risk SNP associated with diabetes, and the screening result is a subset of a group of high-risk SNP sites. ; Perform statistics on the high-risk SNP sites screened out, and calculate the number of high-risk SNPs contained in each patient sample:

[0017] ,

[0018] in, It is an indicator function, indicating whether the SNP is a high-risk site; for the screened high-risk SNP sites, its specific mechanism of action in diabetes is further marked.

[0019] As a preferred embodiment of the diabetes typing method based on big data of the present invention, the generation of the patient's personalized omics feature vector is performed by weighted merging of key features of metabolism, genes and proteomics.

[0020] In the second aspect, an embodiment of the present invention provides a diabetes typing system based on big data, which includes: a data acquisition module, which is used to collect multi-dimensional data of patients and perform standardized processing; a feature fusion module, which is used to fuse the standardized multi-source data and the extracted features to generate a multi-dimensional feature vector; an omics analysis module, which is used to perform in-depth analysis of the patient's individualized biomarkers based on omics technology and extract specific omics features; a model construction module, which is used to combine the multi-dimensional feature vector and the omics data to construct a diabetes typing model; a personalized treatment recommendation module, which is used to use the diabetes typing model to input the data of new patients after the diabetes typing model training is completed and verified, to obtain personalized typing results, and to make personalized treatment recommendations based on the patient's specific characteristics.

[0021] In a third aspect, an embodiment of the present invention provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program instructions are executed by the processor, the steps of the diabetes classification method based on big data as described in the first aspect of the present invention are implemented.

[0022] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program instructions are executed by a processor, the steps of the diabetes typing method based on big data as described in the first aspect of the present invention are implemented.

[0023] The beneficial effects of the present invention are as follows: by collecting multi-dimensional data of patients and performing standardized processing, and combining omics technology with in-depth analysis of individualized biomarkers, the present invention can generate accurate multi-dimensional feature vectors, thereby constructing a more accurate diabetes typing model, which can not only improve the accuracy of typing, but also provide patients with personalized treatment recommendations, fully reflecting the importance of individual differences in diabetes management. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0025] Figure 1 This is a flow chart of the diabetes classification method based on big data.

[0026] Figure 2 This is the structural diagram of the diabetes classification system based on big data. DETAILED DESCRIPTION

[0027] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0028] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0029] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0030] Example 1, reference Figure 1~Figure 2 , which is the first embodiment of the present invention, and provides a diabetes typing method based on big data, including:

[0031] S1: Collect multi-dimensional data of patients, including electronic medical records, genomic data, metabolic indicators, diet and exercise records, and standardize them.

[0032] Specifically, the patient's medical history, diagnosis record, treatment plan, drug usage, laboratory test results, etc. are extracted from the electronic medical record system of the medical institution; genetic variation data related to diabetes (such as SNPs, CNVs, etc.) are obtained through whole genome sequencing or specific gene testing platforms, including structured data of gene sequences and variation sites; metabolic-related data are extracted from the patient's biochemical test results, such as blood glucose level, insulin resistance index (HOMA-IR), liver function, kidney function and other metabolic indicators; diet and exercise conditions are recorded through wearable devices, mobile applications or patient self-reports, including daily calorie intake, nutrients, exercise frequency, exercise intensity, etc., and the collected data are stored in a unified database.

[0033] The collected data comes from various formats and sources, so it needs to be standardized to ensure that all data is measured uniformly. This step usually includes operations such as data cleaning, normalization, and classification. For example, data in different units are processed in a unified dimension, outliers and noise are removed, and missing data is processed. On this basis, data mining or machine learning algorithms are used to extract key features.

[0034] S2: The standardized multi-source data and extracted features are integrated to generate a multi-dimensional feature vector and extract the most representative patient features.

[0035] In order to process these heterogeneous data, data fusion technology is usually used. First, based on the data type, the appropriate fusion level is selected, such as fusion at the data layer, feature layer, or decision layer. For this solution, feature layer fusion is more appropriate. Data from different dimensions, such as genomic data, metabolic indicators, and features in electronic medical records, are uniformly encoded to generate a unified multi-dimensional feature vector.

[0036] After generating the multidimensional feature vector, the most representative features are extracted using a data fusion algorithm. In combination with the complexity and multidimensional characteristics of medical data, the present invention uses principal component analysis for processing.

[0037] S3: Based on omics technologies (such as metabolomics, genomics or proteomics), in-depth analysis of patients' personalized biomarkers is performed to extract specific omics features.

[0038] S3.1 Based on the patient feature vector after data fusion, select metabolomics, genomics, and proteomics data as the omics data source, ensuring that each omics data corresponds to a unique identifier.

[0039] The selection of omics data is based on the patient's specific health status and the strength of correlation in diabetes risk assessment. The specific operations are as follows:

[0040] For each omics data item, calculate its correlation with diabetes risk and the patient's current health status. Suppose the diabetes risk assessment score is , the patient's health status feature vector is H, then a certain group feature ( Can be metabolomics data , genomics data or proteomics data ) is calculated as:

[0041] ,

[0042] in, Indicates the correlation of a certain omics feature with diabetes risk and health status; Representation characteristics and diabetes risk and health status The covariance of and are the standard deviations of omics characteristics and diabetes risk and health status, respectively.

[0043] Furthermore, to ensure that the omics data are sufficiently representative, a correlation threshold is set. , only the correlation Only omics features with values ​​greater than this threshold were selected for subsequent analysis.

[0044] S3.2 Perform time-series dynamic change analysis on metabolomics data to analyze the changing trends of specific metabolic pathways at different time points and identify abnormal pathways in metabolic activities.

[0045] Specifically, metabolomics features were analyzed based on known metabolic pathway databases (such as KEGG or HMDB). To map metabolic pathways, each metabolic feature Belong to one or more metabolic pathways, let the metabolic pathway be , the correspondence between metabolic characteristics and pathways can be constructed. Through this mapping, the patient's metabolic characteristics can be classified into specific metabolic pathways.

[0046] For each metabolic feature , by calculating its change rate at different time points to analyze its dynamic change trend, assuming a certain metabolic characteristic The rate of change at time point t is:

[0047] ,

[0048] in, For the current moment, For the next moment.

[0049] The rate of change reflects the speed of change of metabolic characteristics between adjacent time points and is used to assess whether there are significant fluctuations in metabolic activity.

[0050] Similar change rate calculations are performed on all metabolic features to form a complete time series dynamic change matrix:

[0051] .

[0052] Using the data in the dynamic change matrix, abnormal pathways in metabolic activity are identified. The specific operations are as follows:

[0053] For each metabolic pathway , calculate the total rate of change of the metabolic features it contains:

[0054] ,

[0055] If the rate of change of a metabolic pathway exceeds the preset abnormal threshold τ, the pathway is marked as an abnormal pathway. For metabolic pathways marked as abnormal, further analysis is conducted on their relevance to the progression of diabetes: by comparing existing diabetes metabolism research literature, it is confirmed whether the abnormal pathway is related to the occurrence, development or complications of diabetes; for metabolic pathways that are indeed related, they are marked as metabolic imbalance pathways related to the progression of diabetes, and used as the basis for subsequent feature extraction.

[0056] S3.3 Perform specific site screening on genomic data and screen out high-risk gene variation sites (SNPs) known to be associated with diabetes by querying a pre-set diabetes-related gene library.

[0057] Extract feature vectors from the genomic data in step S3.1. Assume that the genomic data consists of multiple single nucleotide polymorphism (SNP) sites, each SNP site corresponds to a gene variation. Define the genomic data matrix G, where each row corresponds to a SNP site and each column corresponds to a patient's sample.

[0058] Furthermore, a diabetes-related gene library is pre-established, and each gene variation site in the diabetes-related gene library has a known association with diabetes risk.

[0059] In the patient's genomic data matrix G, high-risk SNP sites associated with diabetes are screened out by comparing with the diabetes-related gene library.

[0060] The screening condition is: if a certain SNP of the patient exists in the diabetes-related gene library, then the SNP is marked as a high-risk SNP related to diabetes, and the screening result is a subset of a group of high-risk SNP sites. .

[0061] Statistics were performed on the high-risk SNP sites screened out, and the number of high-risk SNPs contained in each patient sample was calculated:

[0062] ,

[0063] in, It is an indicator function, indicating whether the SNP is a high-risk site.

[0064] The high-risk SNP sites screened out are further marked for their specific mechanisms of action in diabetes. Based on the results of literature and database queries, each high-risk SNP site is annotated to mark the type of variation that may be associated with the type of diabetes (such as type 1 diabetes or type 2 diabetes). For example, some SNPs may be associated with insulin resistance, while others may be associated with insulin secretion function.

[0065] It can be seen that the presence of these sites can be used as genomic features and directly used as input for subsequent diabetes typing models without the need for complex calculations.

[0066] S3.4 At the proteomic level, based on the expression levels of specific proteins, identify key proteins associated with diabetes, especially proteins related to insulin secretion and glucose metabolism.

[0067] Extract feature vectors from the proteomics data in step S3.1, assuming that the protein expression level of each sample is represented by the vector , indicating that each element Represents the expression level of the kth protein, and defines the proteomics data matrix P, where each row corresponds to a patient sample and each column corresponds to the expression level of a protein.

[0068] Furthermore, a protein database related to diabetes was established and a key protein library was defined, which contained proteins related to insulin secretion and glucose metabolism, such as insulin, glucagon, glucose transporters, etc.

[0069] These key proteins were screened out from the proteomics data matrix P. The expression of proteins related to diabetes in each sample was extracted by comparing the protein library. The screening condition was: if a protein belonged to the protein library, the protein was marked as a key protein related to diabetes.

[0070] For the key proteins screened out, the differences in protein expression levels in patients were compared with those in normal physiological states, with special attention paid to proteins related to insulin secretion and glucose metabolism. The screened proteins were compared with the insulin-related protein library to identify proteins closely related to insulin secretion or glucose metabolism, and further analyze whether changes in the expression of these proteins may lead to metabolic disorders in the course of diabetes.

[0071] S3.5 The key features of metabolism, genes and proteomics are combined to generate a personalized omics feature vector for the patient using a simple weighted combination method. The feature vector contains metabolic pathways, key gene loci and protein expression levels to ensure that the patient's biomarker characteristics are fully reflected.

[0072] S4: Combine multidimensional feature vectors and omics data to construct a diabetes classification model.

[0073] The diabetes classification model trains multiple feature inputs with the patient's existing diabetes type labels and automatically classifies patients using a data-driven approach.

[0074] First, since different omics data have different dimensions (such as metabolites, gene loci, and protein expression levels), the omics feature vectors need to be standardized to ensure that data of different dimensions are comparable.

[0075] If the diabetes label data is a categorical label (such as Type 1, Type 2), these categorical labels need to be encoded and converted into numerical labels.

[0076] Furthermore, due to the task requirements and data characteristics of the present invention, random forest is selected for diabetes classification. Random forest is an integrated learning method that can use multiple decision trees for classification. It is suitable for processing high-dimensional feature data (such as omics data) and has strong robustness.

[0077] After selecting the model, train the model, which includes the following steps:

[0078] For classification tasks, the commonly used loss function is cross entropy loss;

[0079] Optimize model weights using a gradient descent algorithm (such as Adam), iteratively updating parameters to minimize the loss function;

[0080] The training set is used to train the model, while the validation set is used to monitor the performance of the model on unseen data to prevent overfitting. The termination condition of training can be the convergence of the loss function or reaching the maximum number of iterations.

[0081] After training, the model is evaluated by accuracy, precision, and recall.

[0082] Finally, the trained and verified diabetes classification model can be used to predict the classification of new patients. The multidimensional feature vector of the new patient is input into the trained model, and the model will automatically output the diabetes type of the patient.

[0083] S5: After the diabetes classification model is trained and verified, the diabetes classification model is used to input the data of new patients to obtain personalized classification results. The diabetes type is automatically output based on the patient's specific data, and personalized treatment recommendations are made based on the patient's specific characteristics.

[0084] Specifically, the multidimensional omics data (metabolic pathways, gene mutation sites, protein expression levels) of the new patient are converted into feature vectors; the feature vectors of the new patient are input into the trained and verified diabetes classification model, and the diabetes classification model gives a predicted value of the diabetes type based on the input features.

[0085] The diabetes type output by the model is combined with the patient's metabolic, genetic and protein characteristics to further explain the biological mechanisms behind their diabetes type.

[0086] For example, if a patient's metabolic pathway shows abnormalities in metabolites related to insulin metabolism, it can be inferred that the patient's diabetes type may be more inclined to the type related to insulin dysfunction (such as Type 1 diabetes).

[0087] If a specific gene variation site (such as SNP) is screened out with high frequency in genetic data, it may mean that the patient has a genetic risk and is prone to a certain type of diabetes.

[0088] Provide personalized treatment plans based on the patient's diabetes type and characteristics, combined with medical guidelines and existing clinical research results. Specific recommendations may include:

[0089] Drug selection: Select appropriate drugs based on patient classification results and specific biomarkers. For example, if the patient's metabolic imbalance is related to insulin metabolism, it is recommended to use insulin therapy first.

[0090] Lifestyle recommendations: If metabolic data indicate that the patient's glucose metabolism is abnormal, it is recommended that the patient adjust his diet and exercise plan to lower blood sugar levels.

[0091] Gene therapy prospects: For patients whose genomics shows high-risk variant sites, regular genetic testing or exploration of potential gene therapy may be recommended.

[0092] Furthermore, the results predicted by the model can be used to dynamically monitor and adjust the patient's treatment plan over time. For example, new omics data can be re-input regularly and the treatment strategy can be adjusted through the updated classification model.

[0093] It can be seen that, during the application process, the model is adaptively adjusted through the patient's health change data (such as follow-up data, treatment response, blood glucose monitoring data, etc.), and the accuracy and personalization ability of the typing model are iteratively optimized. The present invention uses feedback data for continuous learning to improve the prediction ability of the typing model for future patients.

[0094] Furthermore, this embodiment also provides a diabetes classification system based on big data, including:

[0095] Data collection module, used to collect multi-dimensional data of patients and perform standardized processing;

[0096] The feature fusion module is used to fuse the standardized multi-source data and the extracted features to generate a multi-dimensional feature vector;

[0097] Omics analysis module, which is used to conduct in-depth analysis of patients’ individualized biomarkers based on omics technology and extract specific omics features;

[0098] A model building module, used to combine the multidimensional feature vector and the omics data to build a diabetes typing model;

[0099] The personalized treatment recommendation module is used to input the data of new patients using the diabetes typing model after the diabetes typing model training is completed and passed the verification, obtain personalized typing results, and make personalized treatment recommendations based on the patient's specific characteristics.

[0100] This embodiment also provides a computer device, which is suitable for the case of a diabetes classification method based on big data, including a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the diabetes classification method based on big data proposed in the above embodiment.

[0101] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covered on the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0102] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for implementing diabetes typing based on big data as proposed in the above embodiment is implemented.

[0103] In summary, the present invention can generate accurate multidimensional feature vectors by collecting multidimensional data of patients and performing standardized processing, and combining omics technology with in-depth analysis of individualized biomarkers, thereby constructing a more accurate diabetes classification model, which can not only improve the accuracy of classification, but also provide patients with personalized treatment recommendations, fully reflecting the importance of individual differences in diabetes management.

[0104] Example 2, referring to Table 1 and Table 2, is the second example of the present invention. This example provides a diabetes typing method based on big data. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through simulation experiments.

[0105] First, multi-dimensional data of 100 diabetic patients were collected from a tertiary hospital and physical examination center, covering four major data dimensions: electronic medical records, genomic data, metabolic indicators, diet and exercise records. All patients underwent gene sequencing, regular blood tests, and were equipped with wearable devices to record daily exercise and eating habits.

[0106] The patient's medical history, diagnosis record, treatment plan, laboratory test results, etc. are extracted from the hospital's electronic medical record system. These data involve metabolic indicators such as blood sugar, HOMA-IR index, liver and kidney function. In addition, the patient's genomic data is obtained through the genetic testing platform to extract SNP (single nucleotide polymorphism) and CNV (copy number variation) information. Metabolic data covers key indicators such as the patient's blood sugar level, insulin resistance index and renal function. The patient's exercise intensity and daily calorie intake are recorded through wearable devices. Due to different data formats, standardization processing is required, mainly including cleaning missing data, removing outliers, and normalizing data in different units to ensure that all data are unified into comparable dimensions.

[0107] The feature layer fusion method was used to generate multidimensional feature vectors for genomic, metabolic and medical record data. First, principal component analysis (PCA) was used to extract representative features, key SNP sites were extracted from genomic data, and features such as insulin resistance and blood sugar fluctuations were extracted from metabolic data.

[0108] The generated feature vectors were subjected to in-depth metabolomics, genomics, and proteomics analysis. Metabolomics analysis identified abnormal metabolic pathways by performing a time series analysis of the dynamic changes in the patient's blood glucose levels. The genomics part screened out high-risk SNP sites that were closely related to diabetes, such as rs7903146 ​​and rs12255372. At the proteomics level, key proteins related to insulin, such as insulin and glucose transporters, were screened out.

[0109] Using the above-extracted omics features, a diabetes classification model was constructed to classify patients. The model was trained using a support vector machine (SVM) through a machine learning algorithm, and the model accuracy was evaluated through cross-validation. The following is part of the data of the test subjects, recorded in two tables:

[0110] Table 1 Multi-dimensional data of patients

[0111] Test subjects Age (years) BMI (kg / m²) Fasting blood glucose (mmol / L) Glycated hemoglobin (%) HOMA-IR Exercise frequency (times / week) Patient A 45 28.5 7.1 6.8 2.1 3 Patient B 56 30.2 8.5 7.5 3.0 2 Patient C 37 26.4 6.9 6.7 2.2 4 Patient D 62 29.8 8.1 7.2 2.9 2 Patient E 50 31.1 9.0 8.0 3.5 1 Patient F 43 27.5 7.3 6.9 2.4 3

[0112] Table 2 Metabolomic and proteomic characteristics of patients

[0113] Omic characteristics Relevance score (%) Patient A Patient B Patient C Patient D Patient E Patient F Metabolic Pathway 1 85 1 1 0 1 1 0 Metabolic Pathway 2 78 0 1 1 0 1 1 Metabolic Pathway 3 82 1 0 1 1 0 1 Locus A 91 1 1 0 1 1 0 Locus B 87 0 1 1 1 1 1 Protein X 76 1 0 1 0 1 0

[0114] It can be seen from the above two tables that the genomic data in Table 1 reveals the distribution of different SNP sites, especially the variation types of rs7903146 ​​and rs12255372 sites in different patients. Combined with metabolic data (such as blood sugar, HOMA-IR index) for analysis, it can effectively distinguish different types of diabetic patients. In the prior art, it often relies only on single-dimensional data such as medical records or genomic information, while the present invention significantly improves the accuracy and comprehensiveness of diabetes typing through multi-dimensional data fusion.

[0115] In the metabolomics and proteomics data in Table 2, it can be seen that the secretion levels of insulin and glucagon in different patients are significantly different, especially the protein expression level directly reflects the differences in the metabolic functions of patients.

[0116] The present invention uses omics technology to dynamically analyze metabolic pathways and can identify the characteristics of metabolic disorders. For example, the metabolic pathway change rate of P02 patients reached 0.22, exceeding the normal threshold, which means that their metabolic activity is abnormal, indicating a worsening trend of diabetes.

[0117] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A diabetes typing method based on big data, characterized by: include: Collect multi-dimensional data of patients and perform standardized processing; The standardized multi-source data and extracted features are integrated to generate a multi-dimensional feature vector; Based on omics technology, we conduct in-depth analysis of patients’ individualized biomarkers and extract specific omics features; Combining the multidimensional feature vector and the omics feature to construct a diabetes typing model; After the diabetes typing model is trained and verified, the diabetes typing model is used to input the data of new patients to obtain personalized typing results, and personalized treatment suggestions are made in combination with the patient's specific characteristics; The patient's multi-dimensional data includes electronic medical records, genomic data, metabolic indicators, diet and exercise records; The in-depth analysis of the patient's individualized biomarkers based on omics technology and the extraction of specific omics features include the following steps: According to the patient feature vector after data fusion, metabolomics, genomics and proteomics data are selected as the omics data source, ensuring that each omics data corresponds to a unique identifier; Perform time series dynamic change analysis on metabolomics data to analyze the changing trends of specific metabolic pathways at different time points and identify abnormal pathways in metabolic activities; Specific site screening of genomic data, by querying a pre-set diabetes-related gene library, screen out high-risk gene variation sites known to be associated with diabetes; At the proteomic level, key proteins associated with diabetes were identified based on the expression levels of specific proteins; Integrate key features of metabolism, genes and proteomics to generate a personalized omics feature vector for the patient; The selection of metabolomics, genomics and proteomics data as omics data sources includes: For each omics data item, calculate the correlation with diabetes risk and the patient's current health status: Assume that the diabetes risk assessment score is , the patient's health status feature vector is H, then a certain group feature The correlation calculation formula is: ， in, Indicates the correlation of a certain omics feature with diabetes risk and health status; Representation characteristics and diabetes risk and health status The covariance of and are the standard deviations of omics characteristics and diabetes risk and health status, respectively; The method of performing time series dynamic change analysis on metabolomics data, analyzing the change trend of specific metabolic pathways at different time points, and identifying abnormal pathways in metabolic activities includes the following steps: Based on the known metabolic pathway database, metabolomics features were analyzed. To map metabolic pathways, each metabolic feature Belong to one or more metabolic pathways, let the metabolic pathway be , and construct the correspondence between metabolic characteristics and pathways; For each metabolic feature ,Analyze the dynamic change trend by calculating the rate of change at different time points; Calculate the change rate of all metabolic characteristics to form a complete time series dynamic change matrix ; For each metabolic pathway , calculate the total rate of change of the included metabolic features: ， in, The time series dynamic change matrix The value in ; If the change rate of a metabolic pathway exceeds the preset abnormal threshold τ, the metabolic pathway is marked as an abnormal pathway, and the metabolic pathway marked as abnormal is further analyzed for its correlation with the progression of diabetes; The high-risk gene variation sites known to be associated with diabetes are screened out including: Assume that the genomic data consists of multiple single nucleotide polymorphism (SNP) sites, and each SNP site corresponds to a gene variation; Define the genomic data matrix G, where each row corresponds to a SNP site and each column corresponds to a patient's sample; First, a diabetes-related gene library is established. Each gene variation site in the diabetes-related gene library has a known association with diabetes risk. In the patient's genomic data matrix G, high-risk SNP sites associated with diabetes are screened out by comparing with the diabetes-related gene library; The filter criteria are: If a patient's SNP site exists in the diabetes-related gene library, the SNP site is marked as a high-risk SNP site associated with diabetes, and the screening result is a subset of a group of high-risk SNP sites. ; Statistics were performed on the high-risk SNP sites screened out, and the number of high-risk SNPs contained in each patient sample was calculated: ， in, It is an indicator function, indicating whether the SNP site is a high-risk site; The high-risk SNP sites screened out are further marked for their specific mechanisms of action in diabetes.

2. The diabetes typing method based on big data according to claim 1, characterized in that: The method of generating a personalized omics feature vector for a patient is performed by weighted merging of key features of metabolism, genes, and proteomics.

3. A diabetes typing system based on big data, based on the diabetes typing method based on big data described in any one of claims 1 to 2, characterized in that: Also includes: Data collection module, used to collect multi-dimensional data of patients and perform standardized processing; The feature fusion module is used to fuse the standardized multi-source data and the extracted features to generate a multi-dimensional feature vector; Omics analysis module, which is used to conduct in-depth analysis of patients’ individualized biomarkers based on omics technology and extract specific omics features; A model building module, used to combine the multidimensional feature vector and the omics data to build a diabetes typing model; The personalized treatment recommendation module is used to input the data of new patients using the diabetes typing model after the diabetes typing model training is completed and passed the verification, obtain personalized typing results, and make personalized treatment recommendations based on the patient's specific characteristics.

4. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the diabetes typing method based on big data described in any one of claims 1 to 2 are implemented.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the diabetes typing method based on big data described in any one of claims 1 to 2 are implemented.

Images