Diabetes risk prediction system and method based on big data analysis

Abstract

The invention relates to the technical field of medical health information data, and discloses a diabetes risk prediction system through big data analysis, a sample collection module collects a non-invasive biological sample of an individual, an original data acquisition module obtains data of individual molecular level comprehensive information from the biological sample through a detection technology, and the data is used for predicting the risk of diabetes through big data analysis. The data preprocessing module is used for preprocessing original molecular data to prepare for subsequent analysis, and a preprocessing method comprises a hypothesis space conversion method and false model processing; the expression pattern recognition module explores a sample expression pattern from multiple dimensions through a system biological network analysis method; and the multi-level risk indicator integration module is used for integrating personalized biological risk indicators extracted from samples and clinical data into a multi-level data representation form. According to the method, molecular information in an individual biological sample is analyzed by adopting a system biology method based on bioinformatics, so that individual personalized diabetes risk prediction is realized.

Description

Technical Field

[0001] The present invention relates to the field of medical health information data technology, and specifically to a diabetes risk prediction system and method through big data analysis. Background Art

[0002] Diabetes risk assessment is to predict the possibility of diabetes by evaluating an individual's lifestyle, weight, age, family history and other factors. This assessment helps to prevent and control diabetes in advance and reduce the risk of disease. Lifestyle is the main factor affecting diabetes risk. Bad lifestyle habits, such as long-term sitting, unbalanced diet, excessive smoking and drinking, etc., may increase the risk of diabetes; being overweight or obese is also an important risk factor for diabetes. People with excessive body mass index (BMI), especially those with abdominal obesity, are more likely to develop diabetes; in addition, age and family history are also important factors affecting diabetes risk. The older you are, the higher the risk of disease; people with a family history of diabetes also have a relatively high risk of disease; through the assessment of these risk factors, the possibility of an individual's disease can be predicted more accurately, so that preventive measures can be taken in advance, such as adjusting lifestyle habits and controlling weight. At the same time, blood sugar tests can also be performed regularly based on the assessment results to detect and treat diabetes in a timely manner. Diabetes has become one of the global public health issues, and early prediction and prevention are essential. However, current prediction methods are mainly based on demographic data and individual biomarkers, and have not yet fully utilized the potential of big data.

[0003] Existing diabetes risk prediction systems and methods have the following shortcomings: Existing methods mainly rely on physiological parameters and medical history to predict risk, but cannot go deep into biological mechanisms, make individual predictions with high accuracy, and cannot make comprehensive predictions from multi-level risk factors. Therefore, it is necessary to design corresponding technical solutions to solve this problem. Summary of the invention

[0004] 1. Technical issues to be resolved In response to the shortcomings of the existing technology, the present invention provides a diabetes risk prediction system and method through big data analysis, which solves the technical problems that the existing methods mainly rely on physiological parameters and medical history to predict risks, but cannot deeply understand the biological mechanisms, have difficulty in making individual predictions with high precision, and cannot make comprehensive predictions from multi-level risk factors.

[0005] (II) Technical solution To achieve the above objectives, the present invention is implemented through the following technical solutions: a diabetes risk prediction system through big data analysis, including a sample collection module, a raw data acquisition module, a data preprocessing module, an expression pattern recognition module, a multi-level risk index integration module, a prediction model training module, an individual prediction module and a result interpretation module. The sample collection module collects non-invasive biological samples of an individual, including sheath hair samples, saliva samples, oral flora samples, feces samples, urine samples, exhaled breath samples, expanded fibrous tissue samples, serum / plasma samples and sebum samples; The raw data acquisition module obtains data of comprehensive information at the individual molecular level from biological samples through detection technology, and the detection technology includes high-throughput genome technology, transcriptome technology, proteome technology, metabolome technology, epigenetic group technology and microbial group technology; The data preprocessing module preprocesses the raw molecular data to prepare for subsequent analysis, and the preprocessing method includes a hypothesis space conversion method and a pseudo-model processing; The expression pattern recognition module uses a systems biology network analysis method to explore the characteristic expression patterns of samples in multiple dimensions at the levels of genes, metabolic pathways, expression pattern recognition modules, protein interaction networks, epigenetic modifications, transcriptional regulatory factor networks, signal transduction networks, microbiome structures, and single cell subtypes; The multi-level risk index integration module integrates the individual biological risk index extracted from the sample with the clinical data into a multi-level data representation form; The prediction model training module trains the machine learning model based on a large amount of sample data to establish an algorithm framework for individual risk prediction; The individual prediction module uses the trained model to perform a personalized risk prediction analysis report on the newly input individual samples; The result interpretation module can interpret the prediction results and reveal the potential risk pathways of individuals at the molecular mechanism level.

[0006] Preferably, the raw data acquisition module can combine single-cell group technology to obtain multi-level data, and the way to obtain individual molecular level information is as follows: separation of single cells, single-cell whole genome sequencing, single-cell proteome analysis, single-cell carbon monoxide toluene measurement, single-cell epigenetic group technology and multi-data integrated analysis.

[0007] Preferably, the method step of assuming spatial transformation includes assuming that the original data contains a hidden nonlinear structure; Develop a set of biological hypotheses and assume that the data fits these hypotheses in the new feature space. Define new feature representation methods based on the hypothesis, based on topological position encoding of protein interaction networks; Transform the original data into a new feature space.

[0008] Preferably, the method steps of the pseudo-model processing include being applicable to sample data with a large number of missing values; Based on the existing samples, establish a hypothetical distribution model of the quality; Perform simulated sampling for each missing value point according to the model; The simulation was repeated several times for each missing point.

[0009] The distribution properties of possible values ​​are taken into account.

[0010] Preferably, the machine learning model uses neural network regression to perform risk value regression prediction, and the algorithm framework for individual risk prediction includes data collection, data fusion / feature engineering, model training and model evaluation / selection, and the feature engineering includes signal extraction and selection of feature combinations.

[0011] Preferably, the method comprises the following steps: S1, non-invasive biological samples of individual hair samples, saliva samples, oral flora samples, feces samples, urine samples, exhaled breath samples, expanded fibrous tissue samples, serum / plasma samples and sebum samples; S2, obtain comprehensive information at the individual molecular level through high-throughput genomics, transcriptomics, proteomics, metabolomics, epigenetics and microbiome technologies; S3, then use the system biology network inference method to analyze the expression pattern characteristics of samples at multiple levels: genes, metabolic pathways, expression pattern recognition modules, protein interaction networks, epigenetic modifications, transcriptional regulatory factor networks, signal transduction networks, microbiome structures, and single cell subtypes; S4, integrating these personalized biological risk indicators with peripheral medical history data through machine learning algorithms to establish a multi-level risk prediction model; S5, ultimately provides an individualized prediction report on potential risk pathways and risk values.

[0012] Preferably, the discovery of individual risk characteristics of diabetes also includes taking the following approaches: investigating lifestyle risk factors, collecting family history and medical history information, testing physiological indicators, analyzing drug use history, obtaining psychological condition data, using biomarker technology, obtaining environmental risk factor information, applying mobile technology and collecting demographic characteristics.

[0013] Preferably, the investigated lifestyle risk factors include diet structure, body fat percentage, and exercise habits; The collection of family history and medical history information includes environmental risks of family history of diabetes; The physiological indicators detected include physiological risk indexes of glucose tolerance, insulin level and blood lipid indicators; The analysis of drug use history includes risk characteristics of medical history of use of diabetes drugs; The acquisition of psychological condition data includes studying the effect of psychological stress on diabetes risk; The use of biomarker technology includes detecting plasma microRNA levels to discover biological risk signals; The environmental risk factor data obtained include the exogenous risk of long-term environmental pollutants; The application of mobile terminal technology to obtain real-time data on individual lifestyle; The demographic characteristics collected include gender, ethnicity, and social risk factors; The diabetes risk prediction method based on big data analysis adopts a more comprehensive and systematic approach, taking into account multi-level factors such as clinical, biological, and lifestyle factors. However, the specific prediction accuracy depends on the implementation of the following aspects: sample size / quality, feature selection, algorithm model, prediction model evaluation, and clinical validation.

[0014] Preferably, the multi-level risk prediction model is a deep neural network model or an integrated learning model; The multi-level risk prediction model is established through the following steps: Collect multi-class non-invasive biospecimens from individuals; Apply systems biology network analysis methods; Use machine learning algorithms; The personalized prediction report provides each individual with a predicted risk score / category, potential risk pathway analysis, and personalized risk management recommendations.

[0015] Preferably, the clinical application method of diabetes risk prediction includes the following important stages: Theoretical research and development and algorithm model design; Sample dataset construction; Algorithm training and tuning; Small-scale clinical validation; Large-scale multi-center clinical research; Productization and application promotion.

[0016] (III) Beneficial effects Compared with the prior art, the present invention has the following beneficial effects: it adopts a systems biology method based on bioinformatics to analyze the molecular information in individual biological samples to achieve personalized diabetes risk prediction for individuals; it reveals the individual biological risk mechanism, and the prediction results are scientifically explainable; it considers multi-level risk factors for comprehensive prediction to improve accuracy; it adopts non-invasive collection of biological samples to explore the risk mechanism, achieves breakthroughs, and is more economical and practical; it can be used for mass screening and early intervention of high-risk groups, and has broad application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1Schematic diagram of the system framework of the present invention. DETAILED DESCRIPTION

[0018] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0019] See also Figure 1 The embodiment of the present invention provides a technical solution: a diabetes risk prediction system through big data analysis, including a sample collection module, a raw data acquisition module, a data preprocessing module, an expression pattern recognition module, a multi-level risk index integration module, a prediction model training module, an individual prediction module and a result interpretation module, The sample collection module collects non-invasive biological samples of an individual, including sheath hair samples, saliva samples, oral flora samples, feces samples, urine samples, exhaled breath samples, expanded fibrous tissue samples, serum / plasma samples, and sebum samples; The raw data acquisition module obtains comprehensive information at the individual molecular level from biological samples through detection technologies, including high-throughput genomics technology, transcriptome technology, proteome technology, metabolome technology, epigenetic group technology and microbiome group technology; The data preprocessing module preprocesses the raw molecular data to prepare for subsequent analysis. The preprocessing methods include hypothesis space conversion method and pseudo-model processing; The expression pattern recognition module uses the system biology network analysis method to explore the characteristic expression patterns of samples in multiple dimensions at the levels of genes, metabolic pathways, expression pattern recognition modules, protein interaction networks, epigenetic modifications, transcriptional regulatory factor networks, signal transduction networks, microbiome structures, and single-cell subtypes. The multi-level risk index integration module integrates the individual biological risk index extracted from the sample with the clinical data into a multi-level data representation; The prediction model training module trains the machine learning model based on a large amount of sample data and establishes an algorithm framework for individual risk prediction; The individual prediction module uses the trained model to provide personalized risk prediction analysis reports for newly input individual samples; The result interpretation module can explain the prediction results and reveal the individual's potential risk pathways at the molecular mechanism level.

[0020] Oral colony samples: Oral cleaning fluid is used to collect the common bacteria in the individual's oral cavity to reflect the changes in the microecological environment; Fecal samples: Collecting feces can obtain information about intestinal microecology and metabolism; Urine samples: Urine analysis of carrier substances can provide insights into renal function and metabolic status; Exhaled breath samples: VOCs in the gas phase can reflect the respiratory system and metabolism; Expanded fibrous tissue samples: such as tissue around the nipple fissure, which contains a lot of tissue information; Serum / plasma samples: After collection, blood cells can be obtained by immunomagnetic bead method; Sebum samples: Analysis of bacteria and substances in sebum; The collection can be combined to obtain diverse information, which helps to fully reflect the biological characteristics of individuals. The selection of appropriate non-invasive samples is the key to realizing this system.

[0021] High-throughput genomic technology: used to sequence the entire genome of an individual quickly and at low cost, and to obtain genomic information such as gene expression levels and structural variations; Transcriptome technology: sequencing mRNA level, studying changes in gene expression levels, and revealing transcriptional regulatory mechanisms in biological processes; Proteomic technology: Review protein expression levels and modification status through mass spectrometry to understand the actual biological changes that occur; Metabolomics technology: detect the level of low molecular weight metabolites through metabolite analysis technology to understand the metabolic status and changes in metabolic diseases; Epigenetic group technology: study DNA methylation and other DNA epigenetic modifications, and explore the regulatory mechanism of environmental factors on gene expression; Microbiome technology: Analyze the microbial structure and function in individual microbial samples to understand the characteristics of individual microecological environments.

[0022] Protein interaction network level: Analyze the physical or regulatory interaction network between proteins to derive expression patterns; Epigenetic modification level: studying the patterns of epigenetic features such as DNA methylation in different samples; Transcriptional regulatory factor network level: explore the regulatory relationship network between transcription factors and the genes they control; Signal transduction network level: revealing the expression status of samples in higher-order regulatory networks such as cell localization and signal transduction; Microbiome structure level: Analyze the variation patterns of various microbial types and their functional loop networks within the sample; Single-cell subtype level: Identify specific expression patterns of different cell subtypes within a cell population.

[0023] The biological characteristics of samples in various links of the life network can be fully explored, providing important clues for subsequent work.

[0024] For further improvement, the raw data acquisition module can be combined with single-cell group technology to obtain multi-level data to provide support for system analysis. Single-cell group technology is a technical platform that obtains the characteristics of each cell type in a tissue or organ by performing comprehensive molecular-level analysis on individual cells. It can be used to obtain individual molecular-level information.

[0025] The raw data acquisition module can combine single-cell group technology to obtain multi-level data. The ways to obtain individual molecular level information are as follows: isolation of single cells, single-cell whole genome sequencing, single-cell proteome analysis, single-cell carbon monoxide toluene measurement, single-cell epigenetic group technology and multi-data integration analysis.

[0026] Isolate single cells, accurately separate individual cells in tissue samples by enzymatic hydrolysis or microfluidics technology; Single-cell whole genome sequencing: quantitative sequencing of the whole genome mRNA or DNA level of each cell to obtain subcellular level information; Single-cell proteome analysis uses flow cytometry or mass spectrometry to quantitatively detect the protein expression of each cell. Flow cytometry is a technology that detects the fluorescent signals of markers in single cells or other biological particles in suspension to achieve high-speed, cell-by-cell quantitative analysis and sorting. Single-cell carbon monoxide toluene measurements to quantify metabolic activity status at the subcellular level; Single-cell epigenetic technology to study the differences in chromatin modifications between different cell types; Multi-data integration analysis: integrating and analyzing data from multiple dimensions such as gene expression, protein expression, and relative cell position; These technologies can be used to extract the molecular characteristics of various cell types in tissues or organs, reflect the biological status at the subcellular level, and conduct deeper analysis of individuals. They can be well combined with other omics technologies.

[0027] In a further improvement, the method step of assuming the space transformation includes assuming that the original data contains a hidden nonlinear structure; Develop a set of biological hypotheses and assume that the data fits these hypotheses in the new feature space. Define new feature representation methods based on the hypothesis, based on topological position encoding of protein interaction networks; Transform the original data into a new feature space; Run regular preprocessing on the new space.

[0028] Directly consider biological hypotheses, and the feature expression is more realistic; It is possible to extract hidden nonlinear or high-order relationship features.

[0029] In a further improvement, the method steps of pseudo-model processing include being applicable to sample data with a large number of missing values; Based on the existing samples, establish a hypothetical distribution model of the quality; Simulate sampling for each missing value point according to the model to infer its possible value; Repeat the simulation several times for each missing point and take the average value to fill it.

[0030] The distribution characteristics of possible values ​​are taken into account, and the estimation is more accurate; For samples with more missing values, the model is more mature and the interpolation effect is better.

[0031] For further improvement, the machine learning model uses neural network regression to perform risk value regression prediction. The algorithm framework for individual risk prediction includes data collection, data fusion / feature engineering, model training and model evaluation / selection. Feature engineering includes signal extraction and selection of feature combinations.

[0032] Feature engineering includes signal extraction, feature combination selection, etc. to improve feature expression capabilities; Use grid search and cross-validation algorithm to adjust parameters and select the optimal model and parameters; The risk prediction score or risk category of each sample is extracted as the prediction result.

[0033] A diabetes risk prediction method based on big data analysis comprises the following steps: S1, non-invasive biological samples of individual hair samples, saliva samples, oral flora samples, feces samples, urine samples, exhaled breath samples, expanded fibrous tissue samples, serum / plasma samples and sebum samples; S2, obtain comprehensive information at the individual molecular level through high-throughput genomics, transcriptomics, proteomics, metabolomics, epigenetics and microbiome technologies; S3, then use the system biology network inference method to analyze the expression pattern characteristics of samples at multiple levels: genes, metabolic pathways, expression pattern recognition modules, protein interaction networks, epigenetic modifications, transcriptional regulatory factor networks, signal transduction networks, microbiome structures, and single cell subtypes; S4, integrating these personalized biological risk indicators with peripheral medical history data through machine learning algorithms to establish a multi-level risk prediction model; S5, ultimately provides an individualized prediction report on potential risk pathways and risk values.

[0034] For further improvement, the discovery of individual risk characteristics of diabetes may also include the following methods: investigating lifestyle risk factors, collecting family history and medical history information, testing physiological indicators, analyzing drug use history, obtaining psychological condition data, using biomarker technology, obtaining environmental risk factor information, applying mobile technology and collecting demographic characteristics.

[0035] Further improvements include investigating lifestyle risk factors including diet structure, body fat percentage, and exercise habits; Collect family and medical history information including family history of diabetes and environmental risks; Physiological indicators including glucose tolerance, insulin levels and physiological risk index of blood lipid index were tested; Analysis of medication history including risk characteristics of history of use of diabetes medications; Obtaining data on psychological conditions includes studying the effects of psychological stress on diabetes risk; Use biomarker technology including detection of plasma microRNA levels to detect biological risk signals; Obtain information on environmental risk factors including exogenous risks from long-term environmental pollutants; Apply mobile technology to obtain real-time data on individual lifestyles; Demographic characteristics including gender, ethnicity and social risk factors were collected.

[0036] Comprehensively combing through risk characteristics at multiple levels is conducive to improving the accuracy of personality prediction.

[0037] The diabetes risk prediction method based on big data analysis adopts a more comprehensive and systematic approach, taking into account multi-level factors such as clinical, biological, and lifestyle factors. However, the specific prediction accuracy depends on the implementation of the following aspects: sample size / quality, feature selection, algorithm model, prediction model evaluation, and clinical validation.

[0038] Sample size and quality: The more complete and larger the sample data, the better the model training effect, which helps improve accuracy; Feature selection: mining and selecting the features that most affect predictions, eliminating redundant or noisy features, which helps improve accuracy; Algorithm model: Selecting and tuning an algorithm model suitable for such multi-source and multi-factor problems, such as a deep learning model, has a greater impact on accuracy; Prediction model evaluation: Select evaluation indicators such as AUC and accuracy to perform accurate evaluation on the validation set and test set; Clinical validation: Evaluate its true predictive effect in clinical applications in large-scale clinical studies.

[0039] When each key step is implemented properly and verified with good quality samples, its prediction accuracy is better than the traditional method of single feature.

[0040] Further improved, the multi-level risk prediction model is a deep neural network model or an integrated learning model. Deep neural network models can integrate and learn nonlinear relationships between multi-source heterogeneous data; The ensemble learning model improves the prediction performance by integrating multiple base learners.

[0041] The multi-level risk prediction model was established through the following steps: Collect multi-classified non-invasive biological samples from individuals to obtain comprehensive information at the molecular level; Apply systems biology network analysis methods to mine sample feature expression patterns from multiple molecular and biological levels; Using machine learning algorithms, biological risk indicators are integrated with peripheral data from clinical medical history.

[0042] A personalized prediction report is given for each individual including a predicted risk score / category, potential risk pathway analysis, and personalized risk management recommendations.

[0043] Predicted risk score or category (e.g., high risk / intermediate risk / low risk); Potential risk pathway analysis, revealing the main influencing factors through feature selection; Personalized risk management advice, providing preventive measures tailored to individual circumstances; Driven by big data, truly personalized risk prediction and prevention guidance are achieved, and a comprehensive and actionable results report is provided.

[0044] Specifically, the clinical application method of diabetes risk prediction includes the following important stages: Theoretical research and development and algorithm model design, this stage describes the working principle and basic framework of the system; Sample data set construction requires the collection of a large number of individual biological samples, physiological parameters, and follow-up results to build a database for algorithm training and verification; Algorithm training and tuning: use data sets to train algorithm models and continuously optimize algorithm parameters to achieve ideal prediction performance; Small-scale clinical validation: preliminary performance evaluation of the algorithm in a limited number of actual cases to verify the consistency between the predicted results and the actual results; Large-scale multi-center clinical research, conducting more comprehensive and systematic clinical research with a large number of clinical samples and cooperation with multiple medical institutions; Productization and application promotion, through processes such as market review, promote the transformation of technological achievements into medical products that can be provided for clinical use.

[0045] Summary: A systems biology approach based on bioinformatics is used to analyze the molecular information in individual biological samples to achieve personalized diabetes risk prediction for individuals; the individual biological risk mechanism is revealed, and the prediction results are scientifically explainable; multi-level risk factors are considered for comprehensive prediction to improve accuracy; a breakthrough is achieved by non-invasively collecting biological samples to explore the risk mechanism, which is more economical and practical; it can be used for mass screening and early intervention of high-risk groups, and has broad application prospects.

[0046] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It is obvious to those skilled in the art that the present invention is not limited to the details of the above exemplary embodiments, and the present invention can be implemented in other specific forms without departing from the spirit or basic features of the present invention. Therefore, no matter from which point of view, the embodiments should be regarded as exemplary and non-restrictive. The scope of the present invention is defined by the attached claims rather than the above description, and it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims are included in the present invention. Any figure mark in the claims should not be regarded as limiting the claims involved.

[0047] In addition, it should be understood that although the present specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment may also be appropriately combined to form other implementation modes that can be understood by those skilled in the art.

Claims

1. A diabetes risk prediction system based on big data analysis, comprising a sample collection module, a raw data acquisition module, a data preprocessing module, an expression pattern recognition module, a multi-level risk index integration module, a prediction model training module, an individual prediction module and a result interpretation module, characterized in that: The sample collection module collects non-invasive biological samples of an individual, including sheath hair samples, saliva samples, oral flora samples, feces samples, urine samples, exhaled breath samples, expanded fibrous tissue samples, serum / plasma samples and sebum samples; The raw data acquisition module obtains data of comprehensive information at the individual molecular level from biological samples through detection technology, and the detection technology includes high-throughput genome technology, transcriptome technology, proteome technology, metabolome technology, epigenetic group technology and microbial group technology; The data preprocessing module preprocesses the raw molecular data to prepare for subsequent analysis, and the preprocessing method includes a hypothesis space conversion method and a pseudo-model processing; The expression pattern recognition module uses a systems biology network analysis method to explore the characteristic expression patterns of samples in multiple dimensions at the levels of genes, metabolic pathways, expression pattern recognition modules, protein interaction networks, epigenetic modifications, transcriptional regulatory factor networks, signal transduction networks, microbiome structures, and single cell subtypes; The multi-level risk index integration module integrates the individual biological risk index extracted from the sample with the clinical data into a multi-level data representation form; The prediction model training module trains the machine learning model based on a large amount of sample data to establish an algorithm framework for individual risk prediction; The individual prediction module uses the trained model to perform a personalized risk prediction analysis report on the newly input individual samples; The result interpretation module can interpret the prediction results and reveal the potential risk pathways of individuals at the molecular mechanism level.

2. A diabetes risk prediction system based on big data analysis according to claim 1, characterized in that: The raw data acquisition module can combine single-cell group technology to obtain multi-level data. The ways to obtain individual molecular level information are as follows: isolation of single cells, single-cell whole genome sequencing, single-cell proteome analysis, single-cell carbon monoxide toluene measurement, single-cell epigenetic group technology and multi-data integration analysis.

3. The diabetes risk prediction system based on big data analysis according to claim 1, characterized in that: The methodological steps of assuming space transformation include assuming that the original data contain hidden nonlinear structures; Develop a set of biological hypotheses and assume that the data fits these hypotheses in the new feature space. Define new feature representation methods based on the hypothesis, based on topological position encoding of protein interaction networks; Transform the original data into a new feature space.

4. The diabetes risk prediction system based on big data analysis according to claim 1, characterized in that: The method steps for pseudo-model processing include those applicable to sample data with more missing values; Based on the existing samples, establish a hypothetical distribution model of the quality; Perform simulated sampling for each missing value point according to the model; The simulation was repeated several times for each missing point.

5. The diabetes risk prediction system based on big data analysis according to claim 1, characterized in that: The machine learning model uses neural network regression to perform risk value regression prediction. The algorithm framework for individual risk prediction includes data collection, data fusion / feature engineering, model training and model evaluation / selection. The feature engineering includes signal extraction and selection of feature combinations.

6. A diabetes risk prediction method based on big data analysis, characterized in that: The following steps are involved: S1, non-invasive biological samples of individual hair samples, saliva samples, oral flora samples, feces samples, urine samples, exhaled breath samples, expanded fibrous tissue samples, serum / plasma samples and sebum samples; S2, obtain comprehensive information at the individual molecular level through high-throughput genomics, transcriptomics, proteomics, metabolomics, epigenetics and microbiome technologies; S3, then use the system biology network inference method to analyze the expression pattern characteristics of samples at multiple levels: genes, metabolic pathways, expression pattern recognition modules, protein interaction networks, epigenetic modifications, transcriptional regulatory factor networks, signal transduction networks, microbiome structures, and single cell subtypes; S4, integrating these personalized biological risk indicators with peripheral medical history data through machine learning algorithms to establish a multi-level risk prediction model; S5, ultimately provides an individualized prediction report on potential risk pathways and risk values.

7. The method for predicting diabetes risk by big data analysis according to claim 6, characterized in that: Individual risk characterization of diabetes mellitus includes the following approaches: investigating lifestyle risk factors, collecting family and medical history information, testing physiological indicators, analyzing medication use history, obtaining psychological condition data, using biomarker technology, obtaining environmental risk factor information, applying mobile technology, and collecting demographic characteristics.

8. The method for predicting diabetes risk by big data analysis according to claim 7, characterized in that: The survey investigated lifestyle risk factors including diet structure, body fat percentage, and exercise habits; Collect family and medical history information including family history of diabetes and environmental risks; Physiological indicators including glucose tolerance, insulin levels and physiological risk index of blood lipid index were tested; Analysis of medication history including risk characteristics of history of use of diabetes medications; Obtaining data on psychological conditions includes studying the effects of psychological stress on diabetes risk; Use biomarker technology including detection of plasma microRNA levels to detect biological risk signals; Obtain information on environmental risk factors including exogenous risks from long-term environmental pollutants; Apply mobile technology to obtain real-time data on individual lifestyles; Demographic characteristics including gender, ethnicity and social risk factors were collected.

9. The method for predicting diabetes risk by big data analysis according to claim 6, characterized in that: The multi-level risk prediction model is a deep neural network model or an integrated learning model; The multi-level risk prediction model was established through the following steps: Collect multi-class non-invasive biospecimens from individuals; Apply systems biology network analysis methods; Use machine learning algorithms; A personalized prediction report is given for each individual including a predicted risk score / category, potential risk pathway analysis, and personalized risk management recommendations.

10. The method for predicting diabetes risk through big data analysis according to claim 6, characterized in that: The clinical application method of diabetes risk prediction includes the following important stages: Theoretical research and development and algorithm model design; Sample dataset construction; Algorithm training and tuning; Small-scale clinical validation; Large-scale multi-center clinical research; Productization and application promotion.

Images