Metabolite target interaction prediction method and system for myocardial injury

By constructing a knowledge base of myocardial injury risk factors and metabolite targets, and combining AdaBoost and LSTM networks, the problem of integrating multi-source heterogeneous data was solved, enabling early, dynamic, and accurate assessment and personalized intervention of myocardial injury risk, and improving the risk identification and prevention of cardiovascular diseases.

CN121075411AActive Publication Date: 2025-12-05SICHUAN ACADEMY OF MEDICAL SCI SICHUAN PROVINCIAL PEOPLES HOSPITAL

Patent Information

Application Number
CN202511613066.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-11-06
Publication Date
2025-12-05
Estimated Expiration
2045-11-06

AI Technical Summary

Technical Problem

Existing technologies struggle to integrate multi-source heterogeneous data and capture the interactions of metabolite targets during myocardial injury, leading to inaccurate cardiovascular disease risk assessments and difficulty in identifying high-risk individuals and providing personalized intervention strategies in the early stages.

Method used

By acquiring multi-source heterogeneous data, a knowledge base of myocardial injury risk factors and metabolite targets is constructed. Machine learning and deep learning technologies are used, combined with AdaBoost and gradient boosting tree models, to conduct static risk assessment. Then, LSTM networks are used for time series analysis to generate personalized prevention and intervention strategies.

Benefits of technology

It enables early, dynamic, and precise assessment of myocardial injury risk, identifies metabolite target interactions in high-risk individuals, and provides personalized prevention and intervention strategies, thereby improving the accuracy and early identification capabilities of cardiovascular disease risk assessment.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121075411A_ABST
    Figure CN121075411A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of bioinformatics, particularly discloses a metabolite target interaction prediction method and system for myocardial injury, and aims to solve the problems that complex heterogeneous data processing is insufficient, and dynamic changes of a metabolic network are difficult to capture and fine time sequence prediction is difficult in the prior art. According to the method, multi-source heterogeneous data is integrated, a myocardial injury risk factor and metabolite target knowledge base is constructed, and machine learning and deep learning technologies are utilized to realize risk factor weighting and static and sequential dynamic risk assessment. By fusing the static risk level and the dynamic adjustment factor, the system can predict the metabolite target interaction related to the individual high-risk state and generate personalized prevention and intervention strategy suggestions according to the metabolite target interaction. The system comprises a plurality of functional modules including data acquisition and preprocessing, knowledge base construction, feature engineering, risk modeling, time sequence optimization, target prediction and the like, so that early, dynamic and accurate prediction of myocardial injury risks and recognition of personalized intervention targets are realized.
Need to check novelty before this filing date? Find Prior Art

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of bioinformatics, and specifically relates to a metabolite target interaction prediction method and system for myocardial injury. BACKGROUND

[0002] Cardiovascular disease is a major disease that seriously threatens human health worldwide, and its high morbidity and mortality have always been the focus of research and development in the medical field. Specifically, myocardial ischemia refers to insufficient blood perfusion to the heart, which leads to reduced oxygen supply to myocardial cells, energy metabolism disorder, and ultimately impaired cardiac function. On this basis, if myocardial ischemia recovers blood perfusion after a period of time, it will exacerbate the structural damage and functional disorder of myocardial cells, induce significant decline in heart function, and even induce malignant arrhythmia. This complex pathophysiological process is called myocardial ischemia-reperfusion injury. Given the serious consequences of myocardial ischemia-reperfusion injury, in-depth understanding of its mechanism, early warning, and development of effective intervention strategies are of great significance to reduce patient morbidity and mortality.

[0003] In existing research and clinical practice on myocardial ischemia-reperfusion injury, various technical approaches have been explored and applied to the diagnosis, treatment, and mechanism elucidation of the disease. For example, as a class of small molecule polypeptides with significant biological activity, cardiac peptides have been widely proven to have a clear preventive and therapeutic effect on myocardial ischemia-reperfusion injury. In-depth pharmacological studies have shown that cardiac peptides can effectively inhibit the abnormal release of myocardial enzymes after ischemia-reperfusion, thereby reducing the degree of myocardial cell damage. Its mechanism of action is that during ischemia-reperfusion injury, myocardial cells often undergo degeneration and necrosis, leading to impaired cell membrane function and large amounts of intracellular CK enzymes escaping into the serum, resulting in a significant increase in myocardial enzyme activity. Clinical application data show that after cardiac peptide intervention, the serum myocardial enzyme activity of patients is significantly reduced, which confirms its protective effect on the myocardium. In addition, arrhythmia, as a common complication of myocardial ischemia-reperfusion injury, one of its mechanisms is the dysfunction of myocardial cell membrane ion channels. In the pathological environment of ischemia-reperfusion injury, intracellular acidosis leads to an increase in hydrogen ion concentration, which promotes intracellular Ca2+ overload through H+ / Na+ exchange and Na+ / Ca2+ exchange mechanisms. These abnormal ion activities are prone to induce fatal arrhythmia. Cardiac peptides also show a positive effect in this regard, as they can effectively reduce the incidence of severe ventricular arrhythmia in rats with ischemia-reperfusion injury by inhibiting the inward calcium current of ventricular myocardial cells, thereby playing a myocardial protective role in multiple dimensions.

[0004] Therefore, how to construct a method and system capable of deeply integrating multi-source heterogeneous data, accurately predicting metabolite target point interactions related to myocardial injury through advanced machine learning and deep learning techniques, and achieving dynamic and accurate assessment and early identification of cardiovascular disease risk has become a key challenge and technical problem to be solved for those skilled in the art. SUMMARY

[0005] To achieve the above object, the present application provides the following technical solutions: According to the first aspect of the present application, a metabolite target point interaction prediction method for myocardial injury is requested to protect, comprising the following steps: S1, obtaining individual physiological and biochemical data, clinical diagnosis and treatment records, medical image data, genomics data and metabolomics raw data, and performing cleaning, format unification, standardization, missing value filling, outlier processing and noise reduction operation on the collected heterogeneous data; S2, constructing a myocardial injury risk factor and metabolite target point knowledge base, extracting risk factors, biomarkers, metabolic pathways, protein target points and known drug-target point interaction information related to myocardial injury through automatic text mining and knowledge graph construction technology, and forming a structured knowledge network; S3, fusing the multi-source data preprocessed by S1 and the knowledge base information constructed by S2 to generate a comprehensive feature set containing clinical risk features, genomics features, metabolomics features, time series features and graph structure embedding based features extracted from the knowledge base; and using a recursive feature elimination based feature selection method to select a key feature subset; S4, using an adaptive enhanced AdaBoost learning algorithm to calculate the weights of the key risk factors selected by S3 to construct a weighted risk factor model, and inputting the weights output by the weighted risk factor model and the key feature subset into a gradient boosting tree GBT model to construct a static risk level classification model, and the GBT model outputs the myocardial injury risk level of the patient at the current time point; S5, using a long short-term memory LSTM network to perform time series analysis on the individual historical physiological and biochemical data and metabolomics data preprocessed by S1 to construct a time series data dynamic optimization model, and the LSTM model outputs a dynamic adjustment factor; S6, fusing the static risk level output by the GBT model in S4 and the dynamic adjustment factor output by the LSTM model in S5, predicting the metabolite target point interaction related to the high-risk state of the individual through reasoning based on metabolomics features and the knowledge base and / or prediction based on a machine learning model, and generating personalized prevention and intervention strategy suggestions according to the dynamically adjusted myocardial injury risk level and the predicted metabolite target point interaction.

[0006] Further, in the S1, the physiological and biochemical data include numerical data of blood pressure, blood glucose, blood lipids, and heart rate regular indicators. The clinical diagnosis and treatment records include text data of medical history, diagnosis, medication, and treatment plan. The medical image data include echocardiogram, magnetic resonance imaging (MRI), and computed tomography (CT) image data. The genomic data include single nucleotide polymorphism (SNP) and gene expression profile sequence data, and the metabolomics raw data is obtained by liquid chromatography-mass spectrometry (LC-MS) or gas chromatography-mass spectrometry (GC-MS) technology.

[0007] Further, in the S1, the text data is processed by natural language processing technology for word segmentation, part-of-speech tagging, and named entity recognition, and is converted into structured information. The image data is processed by image processing algorithms for feature extraction. The sequence data is subjected to genotyping and variation identification. The metabolomics raw data is processed by peak extraction, peak alignment, retention time correction, isotope deconvolution, and peak area normalization, and the standard metabolite library is used for preliminary identification and quantification of metabolites.

[0008] Further, in the S2, the automated text mining uses a named entity recognition (NER) model based on deep learning to identify entities and a relation extraction (RE) model to identify semantic relationships between entities, and constructs triplets. The knowledge graph represents the extracted entities and their relationships in the form of nodes and edges, forming a structured knowledge network. The knowledge base integrates known physical interactions between metabolites and myocardial-related protein targets, signal pathway regulation relationships, enzymatic reaction relationships, and gene expression regulation relationships.

[0009] Further, in the S3, the comprehensive feature set includes: Clinical risk features, including physiological and biochemical indicators, demographic information, and structured risk factors extracted from clinical diagnosis and treatment records; Genomic features, including variation information at specific gene sites and gene expression levels related to myocardial injury; Metabolomics features, including relative concentrations of identified metabolites, metabolite ratios, and specific metabolic pathway activity scores obtained by pathway enrichment analysis; Temporal features, extracted by differential and moving average processing of historical physiological and biochemical data and metabolomics data, including trend and volatility features. The graph structure based embedding features extracted from the knowledge base are embedded into the knowledge graph by a graph neural network.

[0010] Further, in the S4, the AdaBoost algorithm trains weak classifiers by iteration, adjusts the weight of training samples in each iteration, and gives the weight to the weak classifier according to the classification performance of the weak classifier, and the final strong classifier is composed of the weighted voting combination of all weak classifiers. The GBT model sequentially constructs multiple decision trees, each tree attempts to correct the prediction error of the previous tree, and uses a gradient descent algorithm to optimize the loss function during training.

[0011] Further, in the S5, the LSTM network internally contains an input gate, a forget gate, an output gate, and a cell state, the input gate controls when and to what extent new information is written into the cell state, the forget gate controls how much information in the cell state at the previous time step is forgotten, the output gate controls how much information in the cell state is output to the hidden state at the current time step, and the cell state is responsible for saving and transmitting information over a long time sequence.

[0012] Further, in the S6, the fusion mechanism is to correct the static risk level output by the GBT model by taking the dynamic adjustment factor output by the LSTM model as an independent prediction variable or as a weight coefficient. The reasoning based on the metabolomics features and the knowledge base analyzes the significantly abnormal metabolites in the individual metabolomics data, identifies the affected metabolic pathways and protein targets in combination with the knowledge base, and the prediction based on the machine learning model trains a graph neural network or a random forest model to predict the existence of metabolite-target interactions by taking the individual comprehensive feature set as input.

[0013] According to the second aspect of the present application, the present application claims to protect a metabolite-target interaction prediction system for myocardial injury, comprising: A data acquisition and preprocessing module for performing the multi-source data acquisition and preprocessing function of step S1; A knowledge base construction module for performing the knowledge base construction function of step S2; A metabolomics data analysis module for performing spectrum processing, metabolite identification and quantification on raw metabolomics data; A feature engineering module for performing the feature engineering function of S3; A risk factor weighting and selection module for performing the feature selection function of S3 and the weighted risk factor model construction function of step S4; A risk level classification module for performing the static risk level classification model construction function of step S4; The time series data dynamic optimization module is configured to construct a time series data dynamic optimization model for performing step S5. The risk assessment and target prediction module is configured to perform a risk level dynamic adjustment and metabolite target interaction prediction function for performing step S6. The intervention strategy suggestion module is configured to perform a personalized intervention strategy suggestion generation function for performing step S6.

[0014] Further, the data acquisition and preprocessing module is configured with a plurality of data interfaces supporting DICOM, HL7, mzML, and FASTQ data formats; and the data acquisition and preprocessing module internally includes a data cleaning unit, a standardization unit, a missing value filling unit, an outlier processing unit, and a data conversion unit.

[0015] The present application relates to the technical field of bioinformatics, and specifically discloses a metabolite target interaction prediction method and system for myocardial injury, aiming to solve the problems of insufficient processing of complex heterogeneous data, difficulty in capturing dynamic changes of metabolic networks, and fine timing prediction in the prior art. The present application integrates multiple sources of heterogeneous data, constructs a myocardial injury risk factor and metabolite target knowledge base, and uses machine learning and deep learning techniques to realize risk factor weighting, static and timing dynamic risk assessment. By fusing the static risk level and the dynamic adjustment factor, the system can predict the metabolite target interaction related to the high-risk state of an individual, and generate personalized prevention and intervention strategy suggestions accordingly. The present application includes multiple functional modules such as data acquisition and preprocessing, knowledge base construction, feature engineering, risk modeling, timing optimization, and target prediction, thereby realizing early, dynamic, and accurate prediction of myocardial injury risk and identification of personalized intervention targets. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 A workflow diagram of a metabolite target interaction prediction method for myocardial injury as claimed in the embodiments of the present application; Figure 2 A structural module diagram of a metabolite target interaction prediction system for myocardial injury as claimed in the embodiments of the present application. DETAILED DESCRIPTION

[0017] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of protection of the present application.

[0018] The terms "first", "second", "third", etc. in the present application are only used for descriptive purpose and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the technical features indicated. Thus, the features defined with "first", "second", "third" can explicitly or implicitly include at least one of the features. In the description of the present application, the meaning of "a plurality of" is at least two, such as two, three, etc., unless otherwise explicitly and specifically limited. All directional indications (such as upper, lower, left, right, front, back, etc.) in the embodiments of the present application are only used to explain the relative position relationship, movement condition, etc. between components in a certain posture (as shown in the drawings), and if the certain posture changes, the directional indications also change accordingly. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but can optionally include steps or units not listed, or can optionally include other steps or units inherent to the process, method, product or device.

[0019] Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment can be included in at least one embodiment of the application. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment, nor are they necessarily mutually exclusive or alternative embodiments. It is expressly understood that the embodiments described herein are combinable with other embodiments.

[0020] Metabolomics, as a branch of systems biology, has become an important means of quantifying and characterizing small molecule metabolites in the body since its emergence in the 1990s. It is particularly noteworthy that non-targeted metabolomics has been widely used in the elucidation of disease pathophysiological mechanisms, the discovery of biomarkers, and the diagnosis of diseases due to its unique advantage of simultaneously quantifying large-scale metabolite data. This technology mainly analyzes the dynamic tracking of metabolites in cells, tissues or body fluids. Given the relatively stable form of metabolites in body fluids, metabolomics analysis has become a powerful tool for disease diagnosis research. Currently, the discovery of new metabolites related to myocardial ischemia-reperfusion injury through metabolomics methods undoubtedly has important theoretical and practical significance for understanding disease progression and developing new prevention and treatment strategies.

[0021] However, with the deepening of medical research and the increasing demand for precision and individualization in clinical diagnosis and treatment, some inherent characteristics of the above-mentioned prior art solutions in the principle level have gradually shown deep limitations in dealing with early warning and precise intervention of cardiovascular diseases, especially myocardial ischemia-reperfusion injury. Specifically, although the discovery and application of cardiotrophin provide an effective way for the treatment of myocardial injury, and its mechanism has been elucidated to some extent, it is more a post-treatment intervention, that is, a therapeutic protection after the injury has occurred or started. It does not provide a perfect and prospective mechanism to predict the risk of myocardial injury in individuals under complex physiological conditions or to accurately identify which type of individual benefits from which intervention. In other words, the existing research on cardiotrophin focuses on its therapeutic effect on known injury, rather than constructing a complex model that can predict the dynamic interaction between metabolites and targets before injury occurs.

[0022] Furthermore, although metabolomics, especially untargeted metabolomics, can efficiently identify and quantify a large number of metabolites and play an important role in disease diagnosis and new metabolite discovery, it still faces great challenges in constructing a fine, multi-dimensional, and time-series metabolite-target interaction prediction model. Traditional metabolomics analysis often focuses on the difference in metabolite levels at a certain time point, and has limited ability to capture the dynamic evolution of metabolic networks during disease development, making it difficult to fully reveal the complex and nonlinear interaction between metabolites and biological targets. The occurrence and development of cardiovascular diseases is a complex process involving multiple factors and stages, involving genetics, environment, lifestyle, and fine regulation of metabolic networks in the body. Simple correlation analysis of metabolite levels is difficult to accurately capture and quantify these complex risk factors, and it is even more difficult to establish a comprehensive prediction model that integrates text-mining clinical risk factors, metabolomics data, and time-series information. The existing technology lacks systematic methodology and efficient computing tools in dealing with high-dimensional, heterogeneous data and extracting key metabolite-target interactions with high predictive power. Traditional statistical methods or experience-based judgments often fail to fully exploit the deep hidden patterns in the data, especially when considering the weights, nonlinear relationships, and time dependencies of different risk factors, the prediction accuracy and robustness often fail to meet the needs of early identification of high-risk populations in clinical practice. This limitation makes it difficult to accurately identify high-risk individuals at an early stage, especially when the disease has not yet shown obvious clinical symptoms, thus missing the best opportunity for prevention and intervention.

[0023] According to the first embodiment of the present application, with reference to Figure 1 The present application claims a metabolite-target interaction prediction method for myocardial injury, comprising the following steps: S1, obtaining individual physiological and biochemical data, clinical diagnosis and treatment records, medical image data, genomics data, and metabolomics raw data, and performing cleaning, format unification, standardization, missing value filling, outlier processing, and noise reduction operations on the collected heterogeneous data; S2, constructing a myocardial injury risk factor and metabolite target knowledge base, extracting risk factors, biomarkers, metabolic pathways, protein targets, and known drug-target interaction information related to myocardial injury through automatic text mining and knowledge graph construction technology, and forming a structured knowledge network; S3, fusing the preprocessed multi-source data of S1 and the knowledge base information constructed in S2 to generate a comprehensive feature set containing clinical risk features, genomics features, metabolomics features, time series features, and graph structure embedding-based features extracted from the knowledge base; and using a recursive feature elimination-based feature selection method to select a key feature subset; S4, using an adaptive boosting AdaBoost learning algorithm to calculate the weights of the key risk factors selected in S3 to construct a weighted risk factor model, and inputting the weights output by the weighted risk factor model and the key feature subset into a gradient boosting tree GBT model to construct a static risk level classification model, and the GBT model outputs the myocardial injury risk level of the patient at the current time point; S5, using a long short-term memory LSTM network to perform time series analysis on the individual historical physiological and biochemical data and metabolomics data preprocessed in S1 to construct a time series data dynamic optimization model, and the LSTM model outputs a dynamic adjustment factor; S6, fusing the static risk level output by the GBT model in S4 and the dynamic adjustment factor output by the LSTM model in S5, predicting the metabolite target interaction related to the individual high-risk state through reasoning based on metabolomics features and the knowledge base and / or prediction based on a machine learning model, and generating personalized prevention and intervention strategy suggestions according to the dynamically adjusted myocardial injury risk level and the predicted metabolite target interaction.

[0024] In this embodiment, individual physiological and biochemical data, clinical diagnosis and treatment records, medical image data, genomics data, and metabolomics raw data are obtained. The physiological and biochemical data includes but is not limited to numerical data of blood pressure, blood glucose, blood lipids, and heart rate regular indicators. The clinical diagnosis and treatment records include but are not limited to text data of medical history, diagnosis, medication, and treatment plan. The medical image data includes but is not limited to image data of echocardiogram, magnetic resonance imaging (MRI), and computed tomography (CT). The genomics data includes but is not limited to sequence data of single nucleotide polymorphism (SNP) and gene expression profile. The metabolomics raw data includes but is not limited to raw mass spectrum and chromatogram obtained by liquid chromatography-mass spectrometry (LC-MS) or gas chromatography-mass spectrometry (GC-MS) technology, as well as corresponding peak intensity, retention time, and mass-to-charge ratio data. The collected heterogeneous data is cleaned, formatted, standardized, missing value filled, outlier processed, and denoised to ensure data quality and consistency for subsequent analysis. Text data is processed by natural language processing technology for word segmentation, part-of-speech tagging, and named entity recognition, and then converted into structured information. Image data is processed by image processing algorithms for feature extraction, such as texture features, morphological features, and gray level features. Sequence data is processed for genotyping and variation identification. Metabolomics raw data is processed by peak extraction, peak alignment, retention time correction, isotope deconvolution, and peak area normalization, and then identified and quantified using a standard metabolite library.

[0025] S2 extracts risk factors, biomarkers, metabolic pathways, protein targets, and known drug-target interaction information related to myocardial injury from public medical literature, clinical guidelines, and disease databases through automated text mining and knowledge graph construction technology. The risk factors include but are not limited to hypertension, diabetes, hyperlipidemia, smoking history, age, gender, body mass index (BMI), family history, and specific genetic susceptibility gene sites. The text mining technology uses a named entity recognition (NER) model based on deep learning to identify entities and a relation extraction (RE) model to identify semantic relationships between entities, constructing triplets (entity 1, relationship, entity 2). The knowledge graph represents the extracted entities and their relationships in the form of nodes and edges, forming a structured knowledge network. Meanwhile, the knowledge base integrates known physical interactions between metabolites and myocardial-related protein targets, signal pathway regulation relationships, enzymatic reaction relationships, and gene expression regulation relationships, forming a metabolite-target interaction network. The knowledge base provides structured prior knowledge for subsequent feature engineering, risk assessment model training, and target prediction.

[0026] S3 Feature engineering and selection: The multi-source data after the first step of preprocessing is fused with the knowledge base information constructed in the second step to generate a comprehensive feature set for the machine learning model. The comprehensive feature set includes: Firstly, clinical risk features, including physiological and biochemical indicators, demographic information, and structured risk factors extracted from clinical diagnosis and treatment records.

[0027] Secondly, genomic features, including mutation information of specific gene sites and gene expression levels related to myocardial injury.

[0028] Thirdly, metabolomics features, including the relative concentration of identified metabolites, metabolite ratios, and specific metabolic pathway activity scores obtained through pathway enrichment analysis.

[0029] Fourthly, time series features, which are extracted by differentiating and moving average processing of historical physiological and biochemical data and metabolomics data to capture trend and volatility features.

[0030] Fifthly, graph structure embedding-based features extracted from the knowledge base, which are learned by graph neural networks to map entities and relationships to low-dimensional vector spaces, capturing complex association information between entities.

[0031] For the comprehensive feature set, a recursive feature elimination (RFE) based feature selection method is used. The working mechanism of the RFE method is as follows: first, a basic machine learning model is trained on all features. Second, according to the model training results, the least important features are removed. Then, the model is retrained on the remaining features, and the step of removing the least important features is repeated. This iterative process continues until a pre-set feature number threshold is reached, or until the model performance indicator reaches the optimal. Through this method, a key feature subset with the highest correlation and greatest contribution to myocardial injury risk prediction is selected, thereby reducing the model complexity and improving the model generalization ability.

[0032] The variation information of the specific genetic locus in S1 and S3 refers to genetic variation at a locus that has been shown to have a significant association with myocardial injury, cardiovascular disease risk, or related pathophysiological processes such as lipid metabolism, inflammatory response, blood pressure regulation, coagulation function, etc. in scientific research or clinical practice through genomics analysis techniques such as genome-wide association analysis GWAS, targeted sequencing, etc. These sites usually exist in the form of single nucleotide polymorphism, insertion / deletion, etc. For example, known risk sites on chromosome region 9p21 (strongly associated with coronary heart disease and myocardial infarction risk), APOE gene (associated with lipid metabolism and atherosclerosis), CETP gene (associated with high-density lipoprotein cholesterol metabolism), PCSK9 gene (associated with low-density lipoprotein cholesterol levels), ACE gene (associated with angiotensin-converting enzyme activity and hypertension risk), etc. The variation information of the specific genetic locus can be embodied as the genotype of the locus (such as AA, AG, GG), allele frequency, or risk score based on a genetic model, etc. when constructing the feature set.

[0033] S4: Constructing a weighted risk factor model and a static risk level classification model: The present application uses an adaptive reinforcement learning algorithm to calculate the weights of the key risk factors selected in the third step to construct a weighted risk factor model. The AdaBoost algorithm trains a series of weak classifiers through iteration. In each iteration, the algorithm adjusts the weights of the training samples, giving higher weights to the samples that were misclassified by the weak classifiers in the previous iteration, so that they receive more attention in subsequent training. Each weak classifier is trained and then assigned a weight based on its classification performance. The final strong classifier is formed by the weighted voting of all weak classifiers. The AdaBoost model outputs the contribution weights of each key risk factor, which reflect the relative influence of different risk factors on the risk of myocardial injury.

[0034] Further, the present application combines the weights output by the weighted risk factor model with the key feature subset screened in S3 and inputs them into a gradient boosting tree model to construct a static risk level classification model. The GBT model is an ensemble learning model that constructs multiple decision trees sequentially, with each tree attempting to correct the prediction error of the previous tree. The GBT model optimizes the loss function using the gradient descent algorithm during training, where each new decision tree is designed to fit the negative gradient of the prediction residual of the previous tree. The GBT model finally outputs the risk level of myocardial injury for the patient at the current time point, which is divided into three categories: low risk, medium risk, and high risk. The classification result reflects the patient's risk of myocardial injury under static physiological conditions.

[0035] S5 constructing a time series data dynamic optimization model: the application uses a long short-term memory (LSTM) network to perform time series analysis on the individual historical physiological and biochemical data and metabolomics data preprocessed in the first step, to construct a time series data dynamic optimization model. The LSTM network is a special recurrent neural network (RNN) structure, which contains an input gate, a forget gate, an output gate and a cell state. The input gate controls when and to what extent new information is written into the cell state. The forget gate controls how much information in the cell state at the previous time step is forgotten. The output gate controls how much information in the cell state is output to the hidden state at the current time step. The cell state is responsible for saving and transmitting information over a long time series.

[0036] The LSTM model receives physiological and biochemical index sequences and metabolomics spectrum data sequences at consecutive time points of a patient as input. The LSTM network effectively captures and learns long-term dependencies and dynamic change patterns in time series data through its unique gating mechanism. After processing the sequence data, the LSTM model outputs a dynamic adjustment factor. The dynamic adjustment factor reflects the dynamic change trend of the patient's myocardial injury risk relative to the static assessment according to the latest time series data.

[0037] S6 dynamic adjustment of myocardial injury risk level and prediction of metabolite target interaction: The static risk level output by the GBT model in S4 is fused with the dynamic adjustment factor output by the LSTM model in the fifth step, to realize dynamic and accurate adjustment of the myocardial injury risk level. The fusion mechanism is that the dynamic adjustment factor output by the LSTM model is used as an independent prediction variable or as a weight coefficient to correct the static risk level output by the GBT model. For example, when the LSTM model predicts an upward trend in risk, even if the GBT model assesses the risk as medium, the risk level after dynamic adjustment is upgraded to high; vice versa. The dynamic adjusted risk level can more accurately reflect the actual risk state of myocardial injury of the patient at different time points.

[0038] On the basis of this dynamic risk assessment, the application further predicts metabolite target interactions related to the high-risk state of the individual. The prediction method includes the following aspects: First, reasoning based on metabolomics features and knowledge base: for individuals identified as high risk, analyze the significantly abnormal metabolites in their metabolomics data, and identify the metabolic pathways and related protein targets affected by these abnormal metabolites, in combination with the metabolite target knowledge base constructed in the second step. The analysis uses methods such as overexpression analysis or gene set enrichment analysis to identify metabolic pathways that are highly associated with myocardial injury pathways and are abnormally expressed in the patient's body.

[0039] Second aspect, prediction based on machine learning model: train a specialized machine learning model that takes the comprehensive feature set of high-risk individuals as input, and known metabolite-target interactions as training labels, to predict new or disturbed metabolite-target interactions present in the individual. The model can identify which metabolites exhibit abnormal binding, activation or inhibition of which protein targets under specific physiological and pathological conditions.

[0040] The prediction results will clearly indicate which metabolites and their corresponding protein targets have undergone significant dynamic changes or abnormal interactions under specific myocardial injury risk conditions.

[0041] Generate personalized intervention strategy recommendations: based on the dynamically adjusted myocardial injury risk level and the predicted metabolite-target interactions, generate personalized prevention and intervention strategy recommendations. The recommendations include but are not limited to lifestyle interventions, drug treatment regimen optimization, and subsequent monitoring programs. The recommendations are directly aimed at the individual's identified risk factors, abnormal metabolic pathways and disturbed targets, aiming to maximize the reduction of myocardial injury incidence or severity.

[0042] Further, in the S1, the physiological and biochemical data include numerical data of blood pressure, blood glucose, blood lipids, and heart rate routine indicators; The clinical diagnosis and treatment records include text data of medical history, diagnosis, medication, and treatment plan; The medical image data include echocardiogram, magnetic resonance imaging (MRI), and computed tomography (CT) image data; The genomics data include single nucleotide polymorphism (SNP) and gene expression profile sequence data; the metabolomics raw data is obtained by liquid chromatography-mass spectrometry (LC-MS) or gas chromatography-mass spectrometry (GC-MS) technology.

[0043] In this embodiment, this step is the basis of the entire prediction method, aiming to collect and normalize biomedical data from different sources and different formats. Specifically, the types of data collected are extremely rich, covering multiple dimensions that can fully reflect the individual's physiological state and disease progression. This includes individual physiological and biochemical data such as blood pressure, blood glucose, blood lipids, and heart rate regular indicators, which are usually stored in the form of numerical matrices. At the same time, clinical diagnosis and treatment records, such as detailed medical history descriptions, diagnosis results, medication, surgery records, and treatment plans, exist in the form of unstructured text. Medical imaging data, such as echocardiogram, cardiac magnetic resonance imaging (MRI) image sequence, and coronary computed tomography (CT) image, are image-based and carry important information about cardiovascular structure and function. Genomic data, including single nucleotide polymorphism (SNP) site information, gene copy number variation, and specific gene expression profile data, are usually presented in the form of sequence or expression matrix. Most importantly, metabolomics raw data, obtained through high-throughput liquid chromatography-mass spectrometry (LC-MS) or gas chromatography-mass spectrometry (GC-MS) techniques, include complex spectral data such as original mass spectra, chromatograms, corresponding peak intensities, retention times, and mass-to-charge ratios. Strict preprocessing of these collected heterogeneous data is crucial, and this process aims to eliminate noise, correct bias, fill in missing values, and unify data formats to ensure the quality and consistency of subsequent analysis. For text data, natural language processing (NLP) techniques are used to convert it into structured information through steps such as dictionary-based and statistical model-based word segmentation, part-of-speech tagging, named entity recognition (NER), and extraction of key entities such as disease names, drug names, and symptoms. Image-based data is processed using deep learning-driven image processing algorithms to segment and identify lesion regions and extract texture features, morphological features, and grayscale features. Sequence-based data requires bioinformatics processing such as sequence alignment, variant detection, genotyping, and expression quantification. The metabolomics raw data processing process is particularly complex, including baseline correction, noise filtering, accurate spectral peak detection, multi-sample spectral peak alignment, retention time correction, isotope peak deconvolution, and peak area or peak height normalization. Subsequently, the international standard metabolite library is used for preliminary metabolite identification and quantification to generate structured metabolite relative or absolute quantification data matrix. This preprocessing work uses a combination of automated processes and manual quality control to ensure data quality and reliability.

[0044] Further, in S1, the text data is segmented, part-of-speech tagged, and named entity recognized using natural language processing techniques, and is converted into structured information. The image data is feature-extracted using image processing algorithms. The sequence data is genotyped and variant-identified. The metabolomics raw data is processed by peak extraction, peak alignment, retention time correction, isotope deconvolution and peak area normalization, and a standard metabolite library is used for preliminary identification and quantification of metabolites.

[0045] Further, in S2, the automated text mining uses a named entity recognition (NER) model based on deep learning to identify entities and a relation extraction (RE) model to identify semantic relationships between entities, thereby constructing triplets. The knowledge graph represents the extracted entities and their relationships in the form of nodes and edges, forming a structured knowledge network. The knowledge base integrates known physical interactions between metabolites and myocardial-related protein targets, signal pathway regulation relationships, enzymatic reaction relationships, and gene expression regulation relationships.

[0046] In this embodiment, this step aims to structure the knowledge scattered in the vast medical literature and databases, providing rich prior knowledge for subsequent feature engineering and model training. The construction of this knowledge base relies on automated text mining and knowledge graph technology. By using advanced text mining engines, information related to risk factors, biomarkers, metabolic pathways, protein targets, and known drug-target interactions associated with myocardial injury is automatically identified and extracted from public medical literature databases, disease databases, and pathway databases such as PubMed, OMIM, KEGG, DrugBank, Reactome, etc. The risk factors include not only traditional clinical indicators such as hypertension, diabetes, hyperlipidemia, smoking history, age, gender, body mass index (BMI), family history, etc., but also specific genetic susceptibility gene site information. The core of the text mining technology is the named entity recognition (NER) model based on deep learning, for example, using a pre-trained biomedical language model BERT combined with a conditional random field (CRF) layer to accurately identify medical entities such as diseases, genes, proteins, metabolites, and drugs. Subsequently, a relationship extraction (RE) model, such as a Bi-LSTM model with an attention mechanism, is used to identify the semantic relationships between these entities, such as "gene A is associated with disease B" and "drug C acts on target D," to construct triples like (entity 1, relationship, entity 2). These triples are further used to construct a knowledge graph, representing entities as nodes and relationships as edges to form a highly structured knowledge network, which is stored and managed using a graph database such as Neo4j. At the same time, the knowledge base also integrates known physical interactions between metabolites and myocardial-related protein targets, signal pathway regulation relationships, enzyme reaction relationships, and gene expression regulation relationships to form a comprehensive metabolite-target interaction network. For example, information is extracted from protein interaction databases such as STRING and BioGRID, and metabolic pathway regulation relationships are obtained from pathway databases such as KEGG and Reactome. This knowledge base is not only a static knowledge storage body but also has a knowledge update and management unit that regularly updates and verifies information from the latest research literature and public databases to ensure its timeliness and accuracy.

[0047] Further, in S3, the comprehensive feature set includes: Clinical risk features, including physiological and biochemical indicators, demographic information, and structured risk factors extracted from clinical diagnosis and treatment records; Genomic features, including variation information at specific genetic loci and gene expression levels associated with myocardial injury; Metabolomic features, including relative concentrations of identified metabolites, metabolite ratios, and specific metabolic pathway activity scores obtained through pathway enrichment analysis; Temporal characteristics are extracted by performing differential and moving average processing on historical physiological and biochemical data and metabolomics data to identify trend and fluctuation features. Features based on graph structure embedding extracted from the knowledge base are used to embed the knowledge graph into a graph neural network.

[0048] In this embodiment, this step involves fusing the preprocessed data obtained in the previous two steps with structured knowledge, and extracting a comprehensive feature set that has the greatest information gain for predicting the risk of myocardial injury. This comprehensive feature set is carefully designed as a multi-dimensional, multi-level representation: Firstly, clinical risk characteristics are directly derived from the physiological and biochemical indicators, demographic information, and risk factors extracted from clinical records, such as smoking status and history of diabetes. These characteristics reflect the patient's basic health status and past medical history.

[0049] Secondly, genomic characteristics, including variation information of specific gene loci obtained through genotyping, and gene expression levels related to myocardial injury obtained through RNA sequencing or gene chips.

[0050] Third, metabolomics characteristics, including the relative or absolute concentrations of identified metabolites, the ratios of specific metabolites, and the activity fractions of specific metabolic pathways obtained through pathway enrichment analysis, directly reflect the metabolic state of cardiomyocytes.

[0051] Fourth, temporal characteristics: By performing differential, moving average, and exponential smoothing on the patient's historical physiological and biochemical data and metabolomics data sequences, dynamic change patterns such as trends, fluctuations, and periods can be extracted. For example, the long-term fluctuation of blood pressure or the postprandial peak change rate of blood glucose. These characteristics are crucial for capturing dynamic information on disease progression.

[0052] Fifthly, the features based on graph structure embedding extracted from the knowledge base are input into the graph neural network or graph attention network to learn the embedding of the knowledge graph constructed in the second step. This process maps the entity nodes and relationship edges in the knowledge graph into a low-dimensional vector space, capturing the complex association information, structural information, and semantic similarity between entities, and generating distributed feature vectors that can represent the position and contact of individuals in the knowledge graph. For the large and comprehensive feature set thus generated, a feature selection method based on recursive feature elimination is used to identify the key feature subset with the highest contribution and relevance to myocardial injury risk prediction, while reducing the dimension and complexity of the model and improving its generalization ability. The working mechanism of the RFE method is as follows: first, select a basic machine learning model, such as a support vector machine or a ridge regression model, and train it on the complete dataset containing all original features. After training, identify the feature with the lowest importance according to the model's importance evaluation results. Then, remove the feature with the lowest importance from the feature set. Then, retrain the basic model on the remaining feature subset and identify and remove the current feature with the lowest importance. This iterative process continues until the number of features reaches a pre-set threshold or the performance of the model on the validation set reaches the optimal value. Through RFE, redundant and low-relevance features can be systematically excluded, and a refined and efficient key feature subset can be obtained.

[0053] Further, in the S4, the AdaBoost algorithm trains weak classifiers by iteration, adjusts the weights of training samples in each iteration, and gives weights to weak classifiers according to their classification performance. The final strong classifier is composed of the weighted voting of all weak classifiers. The GBT model constructs multiple decision trees sequentially, each tree trying to correct the prediction error of the previous tree, and uses gradient descent algorithm to optimize the loss function during training.

[0054] In this embodiment, this step aims to quantitatively evaluate the identified key risk factors and preliminarily classify the static risk level of the patient based on this.

[0055] First, the adaptive boosting learning algorithm is used to calculate the weights of the key risk factors selected in the third step to build a weighted risk factor model. AdaBoost is an iterative ensemble learning method that trains a series of weak classifiers sequentially. In each iteration, AdaBoost adjusts the weights of the training samples according to the prediction results of the previous weak classifier: for the samples that are misclassified, their weights are increased so that they will receive more attention in subsequent iterations; for the samples that are correctly classified, their weights are reduced. After each weak classifier is trained, it is assigned a weight according to its classification performance: the better the classifier performs, the higher the weight it gets. The final strong classifier is composed of the weighted voting of all weak classifiers. The AdaBoost model outputs the contribution weights of each key risk factor, which objectively reflect the relative influence and prediction importance of different risk factors on the risk of myocardial injury. Then, the weights output by the weighted risk factor model are combined with the key feature subset selected in the third step and input into the gradient boosting tree model to build a static risk level classification model. GBT model is also a powerful ensemble learning model that constructs multiple decision trees sequentially, and the goal of each subsequent tree is to correct the prediction error of the previous tree.

[0056] Specifically, the GBT model uses the gradient descent algorithm to optimize the predefined loss function during training. Each new decision tree does not directly fit the output value, but fits the negative gradient of the prediction residual of the previous tree. This cumulative learning enables the GBT model to gradually approach the optimal solution and handle complex nonlinear relationships. After receiving the weighted key risk factors and other non-temporal features as input, the GBT model finally outputs the risk level of myocardial injury for the patient at the current time point. The risk level is divided into three discrete categories: low risk, medium risk, and high risk. The classification result reflects the patient's risk of myocardial injury in a static physiological state, providing an initial benchmark for subsequent dynamic adjustments.

[0057] Subsequently, a time-series data dynamic optimization model is constructed. This step aims to capture the dynamic evolution trend of the patient's myocardial injury risk, to make up for the deficiency of static models in dealing with time-dependent information. The present application uses a long short-term memory network to perform time-series analysis on the individual historical physiological and biochemical data and metabolomics data sequences pre-processed in the first step, to construct a time-series data dynamic optimization model. The LSTM network is a special recurrent neural network (RNN) structure, which ingeniously solves the gradient vanishing or gradient explosion problem of traditional RNN in processing long sequence data, and can effectively learn and remember the dependent relationship in long time intervals. The core of LSTM lies in its unique gating mechanism and cell state contained inside: the input gate is responsible for controlling how much new information at the current time step is written into the cell state; the forget gate is responsible for controlling how much information in the cell state at the previous time step should be forgotten or retained; the output gate is responsible for controlling how much information in the cell state at the current time step is output to the hidden state. The cell state is the memory unit of LSTM, which runs through the entire sequence and is responsible for saving and transmitting information in long time series. The LSTM model receives the physiological and biochemical index sequence and metabolomics spectrum data sequence at consecutive time points of the patient as input. Through these gating mechanisms, the LSTM network can effectively capture and learn the long-term dependencies and complex dynamic change patterns in time-series data, for example, identify whether a specific pattern of blood pressure fluctuations indicates deterioration of myocardial function, or whether certain metabolite concentration trends are related to myocardial injury progression. The LSTM model outputs a dynamic adjustment factor after deep learning processing of the sequence data. The dynamic adjustment factor is a continuous value that reflects the dynamic change trend of the patient's myocardial injury risk relative to the static assessment based on the latest time-series data, for example, a positive value indicates an increasing risk, a negative value indicates a decreasing risk, and a value close to zero indicates a stable risk, thereby providing a quantitative basis for dynamic adjustment of the risk level.

[0058] Further, in S5, the LSTM network internally contains an input gate, a forget gate, an output gate, and a cell state. The input gate controls when and to what extent new information is written into the cell state, the forget gate controls how much information in the cell state at the previous time step is forgotten, the output gate controls how much information in the cell state is output to the hidden state at the current time step, and the cell state is responsible for saving and transmitting information in long time series.

[0059] In this embodiment, this step is one of the core innovations of the method of the present application, which combines static risk assessment with dynamic trend and further predicts the molecular level interaction related to high-risk state. Specifically, the static risk level output by the GBT model in S4 is fused with the dynamic adjustment factor output by the LSTM model in the fifth step, thereby realizing the dynamic and accurate adjustment of the risk level of myocardial injury. The fusion mechanism can adopt various ways, for example, the dynamic adjustment factor output by the LSTM model is used as an independent prediction variable to perform weighted linear combination with the prediction probability of the static risk level, or is used as a correction coefficient applied to the classification boundary of the static risk level. More specifically, when the LSTM model predicts that the risk presents an upward trend, even if the GBT model preliminarily assesses it as medium risk, the risk level after dynamic adjustment will be upgraded to high risk to reflect the nature of disease progression; conversely, if the LSTM model predicts that the risk presents a downward trend, the medium risk will be downgraded to low risk. This dynamically adjusted risk level can more accurately reflect the actual risk state of myocardial injury of the patient at different time points, because it takes into account both the baseline risk of the patient and the dynamic changes of recent physiological and biochemical indicators. On the basis of this dynamic risk assessment, the present application further predicts the metabolite-target interaction related to the high-risk state of the individual. The prediction method comprises two complementary aspects: The first aspect is based on the reasoning of metabolomics features and knowledge base. For individuals identified as high risk, first analyze the significantly abnormal metabolites in their metabolomics data, which have concentrations higher or lower than the normal physiological range. Then, combined with the metabolite-target knowledge base constructed in the second step, identify the metabolic pathways and related protein targets affected by these abnormal metabolites. This process can reveal which metabolic imbalances are directly involved in the pathophysiological process of myocardial injury.

[0060] The second aspect is based on the prediction of machine learning models. A special machine learning model, such as a graph neural network or an ensemble learning model, is trained. This model takes the comprehensive feature set of high-risk individuals as input and known metabolite-target interactions as training labels. By learning these complex association patterns, the model can identify which metabolites exhibit abnormal binding, activation or inhibition to which protein targets under specific physiological and pathological conditions, and even predict new, previously unknown metabolite-target interactions. The prediction results will clearly indicate which metabolites and their corresponding protein targets have undergone significant dynamic changes or abnormal interactions under a specific myocardial injury risk state, providing molecular-level target information for subsequent precise intervention.

[0061] Further, in S6, the fusion mechanism is to modify the static risk level output by the GBT model by taking the dynamic adjustment factor output by the LSTM model as an independent prediction variable or as a weight coefficient. The reasoning based on the metabolomics features and the knowledge base identifies the affected metabolic pathways and protein targets by analyzing the significantly abnormal metabolites in the individual metabolomics data in combination with the knowledge base; and the prediction based on the machine learning model trains a graph neural network or a random forest model to predict the existing metabolite-target interaction by taking the individual comprehensive feature set as input.

[0062] In this embodiment, this step is the final goal of the method of the present application, aiming to convert the complex prediction results into operable clinical guidance. According to the dynamically adjusted myocardial injury risk level in S6 and the predicted metabolite-target interaction information, the present application can generate personalized prevention and intervention strategy suggestions. The suggestion content is very specific and targeted, including but not limited to lifestyle interventions, such as recommending specific dietary adjustment programs according to the individual metabolite profile, such as limiting the intake of high-sugar or high-fat foods and increasing dietary fiber; providing customized exercise recommendations, such as the duration and intensity of aerobic exercise per week; optimizing drug treatment programs, such as recommending the use of specific targeted drugs or adjusting the dosage of existing drugs to accurately correct metabolic imbalance or target pathological pathways; and subsequent monitoring programs, such as recommending regular detection of specific metabolite or target levels and imaging review to assess the effectiveness of the intervention and adjust the strategy in a timely manner. The suggestions are directly aimed at the risk factors, abnormal metabolic pathways and disturbed targets identified for the individual, aiming to maximize the reduction of the incidence or severity of myocardial injury and achieve truly precise medicine.

[0063] According to the second embodiment of the present application, referring to Figure 2 The present application claims a metabolite-target interaction prediction system for myocardial injury, comprising: A data acquisition and preprocessing module for performing the multi-source data acquisition and preprocessing function of step S1; A knowledge base construction module for performing the knowledge base construction function of step S2; A metabolomics data analysis module for performing spectrum processing, metabolite identification and quantification on the raw metabolomics data; A feature engineering module for performing the feature engineering function of S3; A risk factor weighting and selection module for performing the feature selection function of S3 and the weighted risk factor model construction function of step S4; A risk level classification module for performing the static risk level classification model construction function of step S4; a time series data dynamic optimization module for performing the time series data dynamic optimization model construction function of step S5; a risk assessment and target prediction module for performing the risk level dynamic adjustment and metabolite target interaction prediction function of step S6; and an intervention strategy suggestion module for performing the personalized intervention strategy suggestion generation function of step S6.

[0064] Further, the data collection and preprocessing module is configured with multiple data interfaces supporting DICOM, HL7, mzML, and FASTQ data formats. The data collection and preprocessing module internally includes data cleaning, standardization, missing value filling, outlier processing, and data conversion units.

[0065] In this embodiment, the data collection and preprocessing module is the first step. This module is the entrance of the system and is configured with multiple data interfaces to ensure seamless integration of data streams from various medical information systems and biological sample analysis equipment. These interfaces support multiple industry standard data formats, such as DICOM for medical images, HL7 for clinical information exchange, mzML for mass spectrometry data, and FASTQ for genetic sequencing data. The module internally integrates data cleaning, standardization, missing value filling, outlier processing, and data conversion units. The data cleaning unit uses rule-based or machine learning-based algorithms to identify and correct data errors, ensuring the accuracy of the original data. The standardization unit uses methods such as Z-score standardization or Min-Max normalization to unify the dimensions of the data and eliminate the impact of different dimensions. The missing value filling unit uses multiple strategies to complete the missing values. The data conversion unit is responsible for converting the original heterogeneous data into a unified, structured data format, such as using a pre-trained word embedding model to convert text data into word vectors or parsing it into entity relationship triples; image data is extracted into high-dimensional feature vectors through deep convolutional neural networks.

[0066] Secondly, the knowledge base construction module. This module is the key to the intelligence of the system, which is configured with a text mining engine, a knowledge graph construction engine, and a knowledge update and management unit. The text mining engine uses advanced deep learning models to realize named entity recognition (NER) and relation extraction (RE) in medical literature. Specifically, the NER model is fine-tuned based on a pre-trained biomedical language model to identify medical professional entities such as "myocardial infarction", "hypertension", "statins", "ACE inhibitors", "L-carnitine", and "mitochondria". The RE model uses a self-attention mechanism to capture the deep semantic relationships between entities, such as identifying the "inhibition" relationship between "beta-blockers" and "heart rate". The knowledge graph construction engine stores the extracted entities and relationships in the form of nodes and edges as a graph structure, and uses high-performance graph databases for management. These databases can efficiently store and query complex graph structure data. The knowledge base stores various entities related to myocardial injury and their relationships, forming a large and detailed knowledge network. The knowledge update and management unit regularly obtains incremental information from the latest research literature and public databases, automatically updates and expands the knowledge base, and ensures the timeliness and completeness of the knowledge.

[0067] Next is the metabolomics data analysis module. This module focuses on processing mass spectrometry raw data, which is configured with a raw spectrum processing unit, a metabolite identification unit, and a metabolite quantification unit. The raw spectrum processing unit performs a series of complex operations such as baseline correction, noise filtering, accurate peak detection, multi-sample peak alignment, isotope peak deconvolution, and data redundancy removal on LC-MS / GC-MS raw data. The metabolite identification unit uses standard spectral libraries and combines precise mass matching, fragment ion matching, and retention time matching to identify metabolite structures with high confidence from the extracted spectral peaks. The metabolite quantification unit performs relative quantification on identified metabolites, or absolute quantification through the addition of internal standards.

[0068] Subsequently, the feature engineering module. This module is a bridge between raw data and machine learning models, which is configured with a feature generation unit, a feature fusion unit, and a feature conversion unit. The feature generation unit extracts and generates various features from the outputs of the data acquisition and preprocessing module, the knowledge base construction module, and the metabolomics data analysis module. These features include but are not limited to numerical features, categorical features, time series features, and graph embedding features. The feature fusion unit uses multi-modal data fusion technology to integrate features of different modalities into a unified feature vector. Fusion strategies include early fusion, late fusion, or intermediate fusion. The feature conversion unit normalizes, standardizes, or performs dimensionality reduction such as principal component analysis (PCA) or t-SNE on the features to optimize model input, reduce redundancy and collinearity between features, and improve model training efficiency and generalization ability.

[0069] The risk factor weighting and selection module is the third module. It aims to identify and quantify the most important risk factors of myocardial injury. It is configured with a recursive feature elimination (RFE) engine and an adaptive boosting (AdaBoost) engine. The RFE engine receives the comprehensive feature set output by the feature engineering module and identifies the key feature subset that is optimal for predicting the risk of myocardial injury by iteratively training a base model and systematically removing the least weighted features. The optimal number of features is dynamically determined by cross-validation. The AdaBoost engine receives the key feature subset selected by the RFE engine and trains a series of weak classifiers. By iteratively adjusting the weights of the training samples and the weights of the weak classifiers, the AdaBoost model outputs the contribution weights of each key risk factor. These weights are represented as floating-point numbers and directly reflect the relative importance of the factor in predicting the risk of myocardial injury. The output of the module is a set of key risk factors with clear weight ordering, providing accurate input for subsequent risk level classification.

[0070] The risk level classification module is the fourth module. It is responsible for preliminary assessment of the static risk of myocardial injury in patients based on the weighted key risk factors. It is configured with a gradient boosting tree (GBT) model. The GBT model receives the weighted key risk factor set output by the risk factor weighting and selection module and other non-temporal features provided by the feature engineering module as input. The GBT model integrates multiple decision trees and uses gradient descent optimization to minimize prediction error to classify and predict the static risk level of myocardial injury. The loss function can use multi-class log loss. The classification result is a discrete class label of low risk, medium risk, or high risk. The GBT model outputs the static risk assessment result of the patient at the current time point as a benchmark for subsequent dynamic adjustment.

[0071] The temporal data dynamic optimization module is the fifth module. It focuses on processing longitudinal data of patients to capture the dynamic changes in risk. It is configured with a long short-term memory (LSTM) network model. The LSTM network model receives the individual historical physiological and biochemical data sequences and metabolomics data sequences provided by the data collection and preprocessing module. The LSTM network model performs deep learning on the time series data through its internal input gate, forget gate, output gate, and cell state mechanism. Its network structure usually includes multiple LSTM layers followed by a fully connected layer to capture long-term dependencies and dynamic change patterns. The model learns how to predict future risk trends based on historical data. The LSTM network model outputs a dynamic adjustment factor, which is a floating-point number between -1 and 1, reflecting the dynamic trend of the patient's risk of myocardial injury: a positive value indicates an increasing risk, a negative value indicates a decreasing risk, and a zero value indicates a stable risk.

[0072] The risk assessment and target prediction module is the core analysis unit of the entire system, which integrates static and dynamic information and goes deep into the molecular level. The module receives the static risk level output by the risk level classification module and the dynamic adjustment factor output by the time series data dynamic optimization module. The module applies the dynamic adjustment factor to the static risk level in a weighted fusion manner to generate a dynamically adjusted myocardial injury risk level. For example, the dynamic adjustment factor is mapped to the [0, 1] interval as an adjustment weight through a Sigmoid function, and then fused with the predicted probability of static risk. Further, the module combines the dynamically adjusted risk level, the metabolite spectrum provided by the metabolomics data analysis module, and the metabolite target knowledge base provided by the knowledge base construction module to predict specific metabolite target interactions related to high-risk states. The prediction is completed by performing pathway enrichment analysis, network topology analysis, and interaction prediction model based on graph neural network. For example, GNN can learn the potential connection patterns between metabolite nodes and target nodes to predict new or disturbed interactions.

[0073] The intervention strategy suggestion module follows. This module aims to translate complex prediction results into executable clinical recommendations. The module receives the dynamic risk level and predicted metabolite target interaction information output by the risk assessment and target prediction module. The module internally contains a rule-based expert system and a knowledge graph reasoning engine. The expert system encodes pre-set clinical guidelines and treatment paths, such as, "if the patient's risk level is high, and it is predicted that the cholesterol metabolism pathway is abnormal, then recommend using statins." The knowledge graph reasoning engine uses entities and relationships in the knowledge base, combined with the specific risk characteristics of individual patients, to perform logical reasoning and generate more detailed and personalized recommendations. These recommendations cover lifestyle interventions, drug treatment regimen optimization, and follow-up monitoring plans.

[0074] Finally, there are the data storage module and the computing resource management module. The data storage module is configured with a relational database for storing structured clinical data, feature data, and model parameters; a non-relational database for storing semi-structured or unstructured data such as raw mass spectrometry data, text analysis results; and a graph database for storing knowledge graphs. The database adopts a distributed storage architecture to ensure data high availability, scalability, and disaster recovery capabilities. The computing resource management module is responsible for scheduling and allocating the system's computing resources, including central processing unit (CPU) resources, graphics processing unit (GPU) resources, and memory resources. The module supports distributed computing frameworks to meet the parallel computing needs of large-scale data processing and complex model training, ensuring the stable operation of the system under high concurrency and large data volume.

[0075] The working principle of the present application is realized through the close cooperation of the above-mentioned modules: first, the data acquisition and preprocessing module integrates and normalizes the heterogeneous biomedical data from multiple sources. At the same time, the knowledge base construction module constructs a structured risk factor and metabolite target knowledge base through text mining and knowledge graph technology. The metabolomics data analysis module performs fine processing on the original mass spectrometry data to extract metabolomics features. The feature engineering module fuses multi-modal data and knowledge base information and generates a unified comprehensive feature set. Subsequently, the risk factor weighting and selection module uses RFE to screen key features and uses the AdaBoost algorithm to calculate the weights of these key risk factors, providing more accurate input for subsequent prediction. The risk level classification module classifies patients into preliminary static myocardial injury risk levels through the GBT model. At the same time, the time series data dynamic optimization module uses the LSTM network to perform deep learning on the patient's historical longitudinal data, capturing the dynamic evolution trend of myocardial injury risk, and outputting a dynamic adjustment factor. The risk assessment and target prediction module fuses static risk assessment and dynamic adjustment factors to generate a more accurate dynamic risk level, and on this basis, combined with individual metabolomics data and knowledge base, it predicts specific metabolite target interactions related to high-risk states. Finally, the intervention strategy suggestion module generates personalized prevention and treatment recommendations based on the predicted dynamic risk level and metabolite target interaction information. Through the collaborative work of the modules of the system, the present application realizes early, dynamic, and accurate prediction of myocardial injury risk and identification of individualized intervention targets, providing comprehensive and in-depth decision support for clinical practice.

[0076] In one specific embodiment, the present application applies the metabolite target interaction prediction method for myocardial injury to a cohort study of 2000 patients, of which 1000 are patients with confirmed myocardial injury, 1000 are healthy controls, and 500 are high-risk populations, such as those with hypertension, diabetes, or high blood lipids. The data collection period is 3 years, and physiological and biochemical indicators and metabolomics data are collected every quarter, and comprehensive genomics sequencing and medical imaging examinations are performed every year.

[0077] Specifically, in the data collection and preprocessing stage, we collected 20 routine physiological and biochemical indicators of patients, including systolic blood pressure, diastolic blood pressure, fasting blood glucose, total cholesterol, LDL-C, HDL-C, triglycerides, and heart rate. Clinical diagnosis and treatment records were identified by named entity recognition based on the BioBERT-CRF model, extracting 15 cardiovascular disease diagnoses including "angina", "heart failure", "coronary atherosclerosis", and 30 commonly used cardiovascular drugs including "aspirin", "metoprolol", and "losartan". Medical imaging data of cardiac MRI were automatically measured by U-Net deep learning model for key indicators such as left ventricular ejection fraction (LVEF) and left ventricular mass (LVM), and texture features such as entropy and contrast of myocardial fibrosis area were extracted. Genomics data identified 200 SNP sites related to myocardial injury through whole exome sequencing and quantified the expression levels of 50 core genes. Metabolomics raw data were collected by LC-MS / MS platform, and Progenesis QI software was used for peak alignment, retention time correction, and isotope deconvolution. HMDB and METLIN databases were used to identify 750 plasma metabolites and perform relative quantification.

[0078] In the knowledge base construction module, we extracted more than 100,000 entities and 500,000 relationships from PubMed, KEGG, and DrugBank databases using BERT-CRF and Bi-LSTM-Attention models, and constructed a knowledge graph containing myocardial injury, risk factors, genes, proteins, metabolites, and drugs. The graph is stored in the Neo4j graph database and is updated regularly. For example, it is clear that "high blood pressure" is a risk factor for "myocardial hypertrophy", "angiotensin-converting enzyme" ACE is the target of "captopril", and "L-carnitine" is involved in the "fatty acid beta-oxidation" metabolic pathway.

[0079] The feature engineering module integrates the above multi-source data. Clinical features include 50 numerical indicators and 10 categorical indicators. Genomics features include 200 SNP site encodings and 50 gene expression levels. Metabolomics features include 750 metabolite concentrations and 20 core metabolic pathway activity scores obtained by MetaboAnalyst pathway enrichment analysis. Time series features are generated by calculating the average, standard deviation, and linear trend slope of each physiological and biochemical indicator in the past year, resulting in 100 time series features. In addition, GCN is used to perform embedding learning on the constructed knowledge graph, generating a 256-dimensional graph embedding feature vector for each patient. Finally, a comprehensive feature set containing 1185 features is formed.

[0080] The risk factor weighting and selection module adopts support vector machine (SVM) as the basic model of RFE, and finally selects 250 key features with the highest predictive power through 5-fold cross-validation. The AdaBoost algorithm weights these 250 features, in which the weight of LDL-C is 0.12, the weight of hypertension history is 0.09, the weight of cardiac troponin T is 0.15, the weight of specific SNP site rs10046384 is 0.07, and the weight of plasma L-carnitine concentration is 0.08.

[0081] The risk level classification module adopts gradient boosting tree (GBT) model, receives the weighted 250 key features and part of the non-time series features. The model achieves a classification accuracy of 92.5% on the training set, and preliminarily predicts the static risk levels of myocardial injury of patients as low, medium, and high on the independent test set.

[0082] The time series data dynamic optimization module uses a three-layer LSTM network, each layer containing 128 units, and the input sequence length is 12 time steps, representing the physiological, biochemical, and metabolomics data of the past 12 quarters. After the LSTM model is trained, it can output a dynamic adjustment factor that quantifies the real-time changes of the patient's myocardial injury risk. For example, a patient with a static assessment of medium risk, if his LDL-C has been rising for the past two quarters and the concentration of plasma oxidative stress metabolites such as malondialdehyde has significantly increased, the LSTM model will output a positive dynamic adjustment factor, for example, +0.35.

[0083] The risk assessment and target prediction module fuses the static risk probability output by the GBT model, for example, the medium risk probability is 0.6, and the high risk probability is 0.3, with the dynamic adjustment factor output by the LSTM model. The fusion function is set as: dynamic risk probability = static risk probability + dynamic adjustment factor * 0.2. If the adjusted high risk probability exceeds 0.5, the final risk level is upgraded to high risk. For example, the adjusted high risk probability of the above-mentioned medium risk patient changes to 0.3 + 0.35 * 0.2 = 0.37, still medium risk. If another patient has a static high risk probability of 0.4 and a dynamic adjustment factor of +0.8, the adjusted high risk probability is 0.4 + 0.8 * 0.2 = 0.56, and the final risk level is upgraded to high risk.

[0084] On this basis, for individuals dynamically assessed as high risk, the system analyzes their metabolomics data. For example, it is found that the plasma free fatty acid (FFA) concentration of this high-risk patient is significantly elevated, and the sugar metabolism pathway such as glycolysis activity is reduced. Combined with the knowledge base, the system infers that high FFA levels cause lipotoxicity of cardiomyocytes and affect mitochondrial function through the PPARa pathway. At the same time, the GNN model predicts that FFA has abnormal binding or activity regulation with the CD36 receptor fatty acid transporter and ACS acyl-CoA synthetase fatty acyl CoA synthetase. The prediction result clearly indicates that the myocardial injury of this patient is closely related to the disorder of fatty acid metabolism and the abnormal interaction of its downstream CD36 / ACS target.

[0085] The intervention strategy suggestion module generates personalized suggestions for this high-risk patient based on the above prediction results: lifestyle intervention suggestions include strict restriction of saturated fatty acid and trans fatty acid intake, and increase of Omega-3 fatty acid-rich foods; it is recommended to do 150 minutes of moderate-intensity aerobic exercise per week. The drug treatment optimization suggestion considers using PPARa agonists to regulate fatty acid metabolism, and combining cholesterol absorption inhibitors to reduce plasma FFA levels. The follow-up monitoring scheme suggests monitoring plasma FFA and myocardial cell lipid deposition by MRI every 3 months, and regularly reviewing myocardial troponin levels.

[0086] Comparative Example: To further illustrate the non-obviousness and technical superiority of the present application, a comparative example is designed. In this comparative example, a traditional static risk assessment method is used, which has the following main characteristics: Single data source and simplified processing: only physiological and biochemical routine indicators and part of clinical text information are used, and no deep knowledge base construction and multi-modal feature fusion are performed. Metabolomics data are only subjected to simple peak area normalization, without metabolite identification and pathway enrichment analysis. Genomics data are not included in the analysis.

[0087] Simplified feature engineering: no knowledge graph embedding features are extracted, no time series features are deeply mined, and only statistical index-based features are used.

[0088] Single model: no weighted risk factor model is constructed, a standard support vector machine (SVM) is directly used for static risk classification, and no dynamic adjustment factor is introduced.

[0089] No metabolite target interaction prediction: no molecular-level metabolite target interaction prediction is performed, and only macroscopic risk levels can be given.

[0090] Generalized intervention suggestions: based on general clinical guidelines, non-personalized and non-targeted intervention suggestions are provided.

[0091] Specifically, the comparative example method uses the same high-risk population cohort as the embodiment for evaluation. In terms of data, only 20 physiological and biochemical indicators of patients and 5 main clinical risk factors are extracted. No detailed text mining and knowledge graph construction is performed. Metabolomics data is directly input into the model after total peak area normalization, without metabolite identification and pathway analysis. In the feature engineering stage, 30 features are directly input into the model. The classification model uses a linear support vector machine with optimized parameters through grid search to train, and outputs low, medium and high risk classification results. The model is evaluated on an independent test set, and its accuracy, F1 score and early myocardial injury event recognition rate are significantly lower than those of the embodiment. In addition, the comparative example method cannot provide metabolite target interaction information, nor can it adjust the risk level according to the dynamic physiological changes of patients. The intervention suggestions it gives are only general "control blood pressure, blood sugar, blood lipid", lacking of pertinence.

[0092] In the feature engineering stage, the variation information of the specific gene sites involved can include but is not limited to the following SNP sites related to myocardial injury risk: rs10757274 (located at 9p21 locus); rs7412 and rs429358 (jointly determine APOE ε2 / ε3 / ε4 genotype); rs708272 (located in the CETP gene); rs11591147 (located in the PCSK9 gene); rs1799752 (located in the ACE gene, I / D polymorphism); and sites closely related to other cardiovascular risk factors such as hypertension and diabetes, such as specific variations on AGT, NOS3, etc. genes.

[0093] The genotype information of these sites is extracted and encoded through bioinformatics processes and included as part of the genomic features in the subsequent comprehensive feature set. Those skilled in the art can update and expand the list of specific gene sites according to the latest research results and public databases.

[0094] Table 1 compares the key indicators of myocardial injury prediction performance and intervention strategy accuracy between the embodiment and the comparative example. The data comes from the same independent validation set, containing 300 patients, of which 50 patients had myocardial injury events during the follow-up period.

[0095] Table 1 Evaluation Comparison Table

[0096] From the above data table, it can be clearly seen that the embodiment of the present application is significantly superior to the traditional static SVM method of the comparative example in the prediction accuracy of myocardial injury events, F1 score and early identification rate. Especially in terms of early identification rate, the embodiment of the present application can identify 85% of myocardial injury events three months in advance, while the comparative example is only 40%, which fully embodies the advantages of the present application in the aspects of time series data analysis, multi-modal feature fusion and knowledge graph utilization. In addition, the present application can provide fine metabolite target interaction prediction results and highly personalized intervention schemes, and simulate a higher average risk reduction rate after intervention, which directly translates into better clinical application value. Although the present application has relatively high computational resource consumption in the model training stage, the great improvement in prediction performance and accuracy of clinical decision support fully proves its technical advancement and practical value.

[0097] In several embodiments provided in the present application, it should be understood that the disclosed system, device and method can be implemented in other ways. For example, the device embodiments described above are only illustrative, and for example, the division of units is only a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0098] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The above integrated unit can be realized in the form of hardware or in the form of software functional unit. The above is only an embodiment of the present application, and does not limit the patent scope of the present application, and any equivalent structure or equivalent process transformation using the content of the present application specification and drawings, or direct or indirect application in other related technical fields, are also included in the patent protection scope of the present application.

[0099] The specific embodiments of the application are described in detail above, but they are only examples, and the present application is not limited to the above described specific embodiments. For those skilled in the art, any equivalent modification or replacement of the present application is also within the scope of the present application, therefore, any equivalent transformation and modification, improvement, etc. made without departing from the spirit and principle range of the present application should be covered in the scope of the present application.

Claims

1. A metabolite target interaction prediction method for myocardial injury, characterized by, The method comprises the following steps: S1, obtaining individual physiological and biochemical data, clinical diagnosis and treatment records, medical image data, genomics data and metabolomics raw data, and performing cleaning, format unification, standardization, missing value filling, outlier processing and noise reduction operation on the collected heterogeneous data; S2, constructing a myocardial injury risk factor and metabolite target knowledge base, extracting risk factors, biomarkers, metabolic pathways, protein targets and known drug-target interaction information related to myocardial injury through automatic text mining and knowledge graph construction technology, and forming a structured knowledge network; S3, fusing the multi-source data preprocessed in S1 and the knowledge base information constructed in S2 to generate a comprehensive feature set containing clinical risk features, genomics features, metabolomics features, time series features and graph structure embedded features extracted from the knowledge base; and using a recursive feature elimination based feature selection method to select a key feature subset; S4, using an adaptive boosting AdaBoost learning algorithm to calculate the weights of the key risk factors selected in S3 to construct a weighted risk factor model, and inputting the weights output by the weighted risk factor model and the key feature subset into a gradient boosting tree GBT model to construct a static risk level classification model, and the GBT model outputs the myocardial injury risk level of the patient at the current time point; S5, using a long short-term memory LSTM network to perform time series analysis on the individual historical physiological and biochemical data and metabolomics data preprocessed in S1 to construct a time series data dynamic optimization model, and the LSTM model outputs a dynamic adjustment factor; S6, fusing the static risk level output by the GBT model in S4 and the dynamic adjustment factor output by the LSTM model in S5, predicting the metabolite target interaction related to the individual high-risk state through reasoning based on metabolomics features and the knowledge base and / or prediction based on a machine learning model, and generating personalized prevention and intervention strategy suggestions according to the dynamically adjusted myocardial injury risk level and the predicted metabolite target interaction.

2. The method of claim 1, wherein, In S1, the physiological and biochemical data includes numerical data of blood pressure, blood glucose, blood lipids and heart rate; The clinical diagnosis and treatment records include text data of medical history, diagnosis, medication and treatment plan; The medical image data includes echocardiogram, magnetic resonance imaging MRI and computed tomography CT image data; The genomics data includes single nucleotide polymorphism SNP and gene expression profile sequence data; and the metabolomics raw data is obtained by liquid chromatography-mass spectrometry LC-MS or gas chromatography-mass spectrometry GC-MS technology.

3. The method of claim 2, wherein, In S1, the text data is processed by natural language processing technology for word segmentation, part-of-speech tagging, named entity recognition, and converted into structured information; The image data is processed by image processing algorithm for feature extraction; The sequence data is genotyped and variation recognized; The metabolomics raw data is processed by peak extraction, peak alignment, retention time correction, isotope deconvolution and peak area normalization, and a standard metabolite library is used for preliminary identification and quantification of metabolites.

4. The method of claim 1, wherein, In S2, the automatic text mining uses a named entity recognition (NER) model based on deep learning to identify entities and a relation extraction (RE) model to identify semantic relationships between entities, thereby constructing triples. The knowledge graph represents the extracted entities and their relationships in the form of nodes and edges, forming a structured knowledge network. The knowledge base integrates known physical interactions between metabolites and myocardial-related protein targets, signal pathway regulation relationships, enzymatic reaction relationships, and gene expression regulation relationships.

5. The method of claim 1, wherein, In S3, the comprehensive feature set includes: Clinical risk features, including physiological and biochemical indicators, demographic information, and structured risk factors extracted from clinical diagnosis and treatment records; Genomics features, including variation information at specific genetic loci and gene expression levels related to myocardial injury; Metabolomics features, including the relative concentrations of identified metabolites, metabolite ratios, and specific metabolic pathway activity scores obtained through pathway enrichment analysis; Temporal features, which are extracted by differentiating and moving average processing of historical physiological and biochemical data and metabolomics data to capture trend and volatility characteristics; Graph structure embedding features extracted from the knowledge base, which are learned through graph neural networks.

6. The method of claim 1, wherein, In S4, the AdaBoost algorithm iteratively trains weak classifiers, adjusts the weights of training samples at each iteration, and assigns weights to weak classifiers based on their classification performance. The final strong classifier is composed of the weighted votes of all weak classifiers. The GBT model constructs multiple decision trees sequentially, with each tree attempting to correct the prediction error of the previous tree and optimizing the loss function using gradient descent during training.

7. The method of claim 1, wherein, In S5, the LSTM network contains input gates, forget gates, output gates, and cell states. The input gate controls when and how much new information is written to the cell state. The forget gate controls how much information from the previous time step is forgotten. The output gate controls how much information in the cell state is output to the hidden state at the current time step. The cell state is responsible for preserving and transmitting information over long time series.

8. The method of claim 1, wherein, In S6, the fusion mechanism modifies the static risk level output by the GBT model by using the dynamic adjustment factor output by the LSTM model as an independent prediction variable or as a weight coefficient. The reasoning based on metabolomics features and the knowledge base identifies the metabolic pathways and protein targets affected by significantly abnormal metabolites in individual metabolomics data, combined with the knowledge base. The prediction based on machine learning models trains graph neural networks or random forest models to predict the existence of metabolite-target interactions using the individual comprehensive feature set as input.

9. A metabolite target interaction prediction system for myocardial injury, characterized by, It includes: Data acquisition and preprocessing module: used to perform the multi-source data acquisition and preprocessing function of step S1 in claim 1; Knowledge base construction module: used for performing the knowledge base construction function of step S2 in claim 1; Metabolomics data analysis module: used for performing the spectrum processing, metabolite identification and quantification of metabolomics raw data; Feature engineering module: used for performing the feature engineering function of S3 in claim 1; Risk factor weighting and selection module: used for performing the feature selection function of S3 and the weighted risk factor model construction function of step S4 in claim 1; Risk level classification module: used for performing the static risk level classification model construction function of step S4 in claim 1; Time series data dynamic optimization module: used for performing the time series data dynamic optimization model construction function of step S5 in claim 1; Risk assessment and target prediction module: used for performing the risk level dynamic adjustment and metabolite target interaction prediction function of step S6 in claim 1; and Intervention strategy suggestion module: used for performing the individualized intervention strategy suggestion generation function of step S6 in claim 1.

10. The system of claim 9, wherein, The data acquisition and preprocessing module is configured with a plurality of data interfaces supporting DICOM, HL7, mzML and FASTQ data formats; and the data acquisition and preprocessing module internally comprises a data cleaning unit, a standardization unit, a missing value filling unit, an outlier processing unit and a data conversion unit.

Citation Information

Patent Citations

  • Adverse drug reaction trace management method and system

    CN118280602A

  • Machine learning for collaborative medical data metrics

    US20180342323A1

Cited By

  • Kidney rejection monitoring system based on wireless implantable biosensor

    CN121971079A

  • Machine learning-based clinical mass spectrum risk prediction method and equipment

    CN122084815A