Medical event probability determination method, apparatus, device, computer medium, and product

Abstract

The application discloses a medical event probability determination method, device, equipment, computer medium and product. The method comprises the following steps: acquiring at least one target metabolite of a target patient hospitalized due to a heart failure with preserved ejection fraction medical event and concentration data of each target metabolite in the at least one target metabolite; acquiring an effect value corresponding to each target metabolite from a corresponding relationship between the target metabolite and the effect value, wherein the effect value is used for representing the influence degree of the target metabolite on the heart failure with preserved ejection fraction medical event; and performing weighted summation on the concentration data of each target metabolite and the effect value corresponding to each target metabolite, so as to determine the probability of the target patient suffering from the heart failure with preserved ejection fraction medical event. The method can improve the accuracy of medical event probability determination, thereby improving the accuracy of the prediction result of the target patient suffering from the heart failure with preserved ejection fraction medical event.

Description

Technical Field

[0001] This application belongs to the field of heart failure disease research technology, and in particular relates to a method, device, equipment, computer storage medium and computer program product for determining the probability of medical events. Background Technology

[0002] Heart failure with preserved ejection fraction (HFpEF) is a subtype of heart failure characterized by preserved left ventricular ejection fraction (LVEF ≥ 50%), accounting for more than 50% of all heart failure cases. Compared with heart failure with reduced ejection fraction (HFrEF), the pathophysiological mechanisms of HFpEF are more complex, involving the interplay of multiple mechanisms such as left ventricular diastolic dysfunction, systemic inflammatory response, endothelial dysfunction, and myocardial fibrosis. Due to its highly heterogeneous etiology and diverse clinical manifestations, early diagnosis, risk stratification, and prognostic assessment of HFpEF remain significant challenges in the cardiovascular field.

[0003] In recent years, the development of metabolomics technology has provided a new perspective for the discovery of disease biomarkers. Metabolites, as downstream products of gene-environment interactions, can directly reflect the physiological and pathological state of the body. Existing studies have shown that metabolomics has broad application prospects in elucidating the molecular mechanisms of heart failure and discovering biomarkers. However, current technologies using metabolites to determine the probability of patients experiencing heart failure with partial angina pectoris (HFpEF) are not precise enough, affecting the accuracy of predicting HFpEF events.

[0004] Therefore, how to provide a method for determining the probability of medical events to improve the accuracy of determining the probability of medical events, thereby improving the accuracy of predicting the occurrence of HFpEF medical events in patients, is an urgent problem to be solved. Summary of the Invention

[0005] This application provides a method, apparatus, device, computer storage medium, and computer program product for determining the probability of medical events, which can improve the accuracy of determining the probability of medical events, thereby improving the accuracy of predicting the occurrence of medical events such as ejection fraction-preserving heart failure in target patients.

[0006] In a first aspect, embodiments of this application provide a method for determining the probability of a medical event, the method comprising: Acquire at least one target metabolite and concentration data of each target metabolite in a target patient hospitalized for a medical event of preserved ejection fraction heart failure, wherein the influence of the at least one target metabolite on the medical event of preserved ejection fraction heart failure is greater than a preset level; The effect value corresponding to each target metabolite is obtained from the correspondence between target metabolites and effect values. The effect value is used to characterize the degree of influence of the target metabolite on medical events of heart failure with preserved ejection fraction. The concentration data of each target metabolite and the corresponding effect value of each target metabolite are weighted and summed to determine the probability of the target patient experiencing a medical event of heart failure with preserved ejection fraction.

[0007] In some possible implementations, the at least one target metabolite is obtained in the following manner: Serum samples were obtained from the target patient, and metabolomics analysis was performed on the serum samples to obtain an initial pool of candidate metabolites. Data quality control processing is performed on each candidate metabolite in the initial candidate metabolite pool to obtain a second candidate metabolite pool; Differential expression screening was performed on each candidate metabolite in the second candidate metabolite pool to obtain the third candidate metabolite pool; Each candidate metabolite in the third candidate metabolite pool is subjected to clinical factor correction to obtain the fourth candidate metabolite pool; Each candidate metabolite in the fourth candidate metabolite pool is subjected to multiple feature screening to obtain the at least one target metabolite.

[0008] In some possible implementations, the step of performing data quality control processing on each candidate metabolite in the initial candidate metabolite pool to obtain a second candidate metabolite pool includes: Stability testing is performed on each candidate metabolite in the initial candidate metabolite pool to obtain the coefficient of variation; the coefficient of variation is used as an indicator to measure the degree of data dispersion. Candidate metabolites with a coefficient of variation greater than a preset threshold are removed to obtain a fifth candidate metabolite pool; Integrity testing is performed on each candidate metabolite in the fifth candidate metabolite pool to obtain the missing ratio; the missing ratio is used to represent the proportion of the total number of serum test samples in which each candidate metabolite was not detected. Candidate metabolites with a missing ratio greater than a preset threshold are removed to obtain a second candidate metabolite pool.

[0009] In some possible implementations, the differential expression screening process for each candidate metabolite in the second candidate metabolite pool to obtain a third candidate metabolite pool includes: The difference magnitude is detected for each candidate metabolite in the second candidate metabolite pool to obtain the difference fold; the difference fold is used to reflect the expression change of the metabolite between survivors and deceased subjects in the presence of medical events of heart failure with preserved ejection fraction; Statistical analysis was performed on each candidate metabolite in the second candidate metabolite pool to obtain significance indicators; A third candidate metabolite pool is obtained by screening candidate metabolites from the second candidate metabolite pool whose difference fold is greater than a preset threshold and whose significance index is less than a preset threshold.

[0010] In some possible implementations, the step of performing clinical factor correction on each candidate metabolite in the third candidate metabolite pool to obtain a fourth candidate metabolite pool includes: Each candidate metabolite in the third candidate pool and a preset clinical factor are input into a multivariate regression model to obtain the interference results of the clinical factors on medical events of heart failure with preserved ejection fraction; the interference results are used to characterize the degree of interference of the clinical factors on the association between metabolites and medical events of heart failure with preserved ejection fraction. Based on the interference results, candidate metabolites that are independently associated with medical events of heart failure with preserved ejection fraction are screened from the third candidate pool to obtain a fourth candidate metabolite pool.

[0011] In some possible implementations, the step of performing multiple feature screening on each candidate metabolite in the fourth candidate metabolite pool to obtain the at least one target metabolite includes: The optimal regularization parameter is selected by cross-validation. Using the optimal regularization parameter, metabolites whose feature coefficients are compressed to a preset value in the fourth candidate metabolite pool are removed to obtain the sixth candidate metabolite pool. Obtain the original features and shadow features of each candidate metabolite in the sixth candidate metabolite pool. The shadow features are generated by randomly shuffling the original features. The shadow features are similar in distribution to the original features but are unrelated to the original target. The original features and the shadow features are input into a random forest model to obtain the importance scores of the original features and the shadow features, respectively. Based on the importance scores of the original features and the shadow features, noisy features with importance scores lower than those of the shadow features in the original features are removed to obtain the seventh candidate metabolite pool. The at least one target metabolite is obtained from the seventh candidate metabolite pool.

[0012] In some possible implementations, obtaining the at least one target metabolite from the seventh candidate metabolite pool includes: Independent variable correlation verification was performed on each candidate metabolite in the seventh candidate metabolite pool to obtain the correlation coefficient between each candidate metabolite. Candidate metabolites with correlation coefficients higher than a preset threshold are removed to obtain at least one target metabolite.

[0013] In some possible implementations, the at least one target metabolite includes: N-methyl-trans-4-hydroxy-proline, rhamnoic acid, uridine triphosphate, 4-hydroxy-3-methoxyphenylacetic acid, hydroxyphenyl lactic acid, perilla alkaloid, N-oleoyl-L-serine, undecanoic acid (11:0), triglycerides (15:0-16:0-24:1), and triglycerides (15:0-18:1-24:1).

[0014] In some possible implementations, the weighted summation of the concentration data of each of the target metabolites and the corresponding effect values ​​of each target metabolite to determine the probability of the target patient experiencing a medical event of preserved ejection fraction heart failure includes: For each target metabolite, the probability of an initial ejection fraction-preserving heart failure medical event for each target metabolite is determined based on the product of the effect value corresponding to the target metabolite and the concentration data of the target metabolite. The probability of the target patient experiencing a preservative-ejection-fraction heart failure medical event is determined by summing the probabilities of each initial ejection fraction-preserving heart failure medical event.

[0015] Secondly, embodiments of this application provide a medical event probability determination device, the device comprising: The acquisition module is used to acquire at least one target metabolite of a target patient hospitalized due to a medical event of preserved ejection fraction heart failure, as well as the concentration data of each target metabolite in the at least one target metabolite, wherein the influence of the at least one target metabolite on the medical event of preserved ejection fraction heart failure is greater than a preset level; The acquisition module is also used to acquire the effect value corresponding to each of the target metabolites from the correspondence between target metabolites and effect values, wherein the effect value is used to characterize the degree of influence of the target metabolite on medical events of heart failure with preserved ejection fraction; The determination module is used to perform a weighted summation of the concentration data of each target metabolite and the effect value corresponding to each target metabolite to determine the probability of the target patient experiencing a medical event of heart failure with preserved ejection fraction.

[0016] Thirdly, embodiments of this application provide a medical event probability determination device, the device comprising: A processor and a memory storing computer program instructions; a method for determining the probability of medical events that, when the processor executes the computer program instructions, implements any of the above.

[0017] Fourthly, embodiments of this application provide a computer storage medium on which computer program instructions are stored, and when the computer program instructions are executed by a processor, the method for determining the probability of medical events described above is implemented.

[0018] Fifthly, embodiments of this application provide a computer program product in which instructions, when executed by a processor of an electronic device, enable the electronic device to perform any of the above-mentioned methods for determining the probability of medical events.

[0019] The medical event probability determination method, apparatus, device, computer storage medium, and computer program product of this application have the following advantages. Firstly, the target metabolites used in this method are those with a greater than preset impact on medical events of preserved ejection fraction heart failure. These metabolites are not randomly selected but have a direct and significant correlation with the occurrence of preserved ejection fraction heart failure. This eliminates interference from metabolites with no substantial or weak correlation to the medical event, ensuring that subsequent concentration detection and data calculation based on these metabolites revolve around the core influencing factors of the medical event. This avoids interference from invalid information on the probability determination results from the data source, guaranteeing a high degree of consistency between the probability determination process and the medical event of preserved ejection fraction heart failure. Secondly, the effect value, as a quantitative indicator characterizing the impact of the target metabolite on medical events of preserved ejection fraction heart failure, accurately reflects the magnitude of the role of different target metabolites in the occurrence of the medical event—metabolites with a greater impact on the medical event correspond to higher effect values, and their concentration data has a higher weight in the probability calculation; metabolites with a smaller impact correspond to lower effect values, and their weight percentage is correspondingly reduced. This approach breaks away from the drawbacks of traditional assessments that "average values" are assigned to each influencing factor. It allows the concentration data of each target metabolite to participate in the calculation based on its actual contribution, making the logic of probability calculation more consistent with the occurrence principle of medical events in heart failure with preserved ejection fraction. The calculation process is more scientific and reasonable. Furthermore, by weighted summing of concentration data and effect values, the probability of medical events is accurately quantified, thereby improving the accuracy of predicting the occurrence of medical events in heart failure with preserved ejection fraction in target patients. Attached Figure Description

[0020] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a flowchart illustrating a method for determining the probability of medical events provided in one embodiment of this application; Figure 2 This is a schematic diagram illustrating the process of obtaining at least one target metabolite according to an embodiment of this application; Figure 3 This is a volcano diagram provided in one embodiment of this application; Figure 4 This is a schematic diagram of OPLS-DA analysis results provided in one embodiment of this application; Figure 5 This is a metabolic feature map filtered by the Boruta algorithm provided in one embodiment of this application; Figure 6 This is a heatmap of differential metabolites associated with 4-year mortality events provided in one embodiment of this application; Figure 7 This is a receiver operating characteristic (ROC) curve of metabolites in the training set predicting the risk of all-cause mortality over 4 years, provided in one embodiment of this application. Figure 8 This is a receiver operating characteristic (ROC) curve provided in one embodiment of this application to validate the prediction of 4-year all-cause mortality risk by concentrated metabolites; Figure 9 This is a receiver operating characteristic (ROC) curve of metabolites in the training set used to predict the risk of cardiovascular death over 4 years, provided in one embodiment of this application. Figure 10 This is a receiver operating characteristic (ROC) curve provided in one embodiment of this application to validate the prediction of 4-year cardiovascular mortality risk using concentrated metabolites; Figure 11 This is a Kaplan-Meier curve of a 4-year all-cause mortality prediction model in the training set provided in one embodiment of this application; Figure 12 This is a Kaplan-Meier curve of a 4-year all-cause mortality prediction model provided in the validation set of one embodiment of this application; Figure 13 This is a schematic diagram of the structure of a medical event probability determination device provided in another embodiment of this application; Figure 14 This is a schematic diagram of the structure of an electronic device provided in another embodiment of this application. Detailed Implementation

[0022] The features and exemplary embodiments of various aspects of this application will be described in detail below. To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain this application and not to limit it. For those skilled in the art, this application can be implemented without some of these specific details. The following description of the embodiments is merely to provide a better understanding of this application by illustrating examples.

[0023] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes said element.

[0024] It should be noted that the acquisition, storage, use, and processing of data in this application embodiment all comply with the relevant provisions of national laws and regulations.

[0025] It should be noted that in the embodiments of this application, certain software, components, models and other existing solutions in the industry may be mentioned. These should be regarded as exemplary and are only intended to illustrate the feasibility of implementing the technical solution of this application. However, it does not mean that the applicant has used or necessarily used the solution.

[0026] To address the problems of the prior art, embodiments of the present invention provide a method, apparatus, device, and computer storage medium for determining the probability of medical events.

[0027] The method for determining the probability of medical events provided in the embodiments of the present invention will be introduced first below.

[0028] Figure 1 A flowchart illustrating a method for determining the probability of medical events according to an embodiment of the present invention is shown. Figure 1 As shown, the method may include the following steps: S110. Obtain at least one target metabolite and concentration data of each target metabolite in the at least one target metabolite from the target patient hospitalized due to a medical event of heart failure with preserved ejection fraction.

[0029] In S110, at least one target metabolite has a greater-than-preset impact on medical events related to preserved ejection fraction heart failure. This type of metabolite is directly and significantly associated with the occurrence and development of medical events related to preserved ejection fraction heart failure and is a core indicator for assessing the probability of such medical events. Concentration data refers to the numerical content of the target metabolite in the target patient's body. It is the basic data reflecting the level of this metabolite in the patient's body and is also the core basic parameter for subsequent probability calculations.

[0030] In practice, biological samples can be collected from target patients hospitalized due to medical events of heart failure with preserved ejection fraction. The samples are then subjected to metabolomics detection using a liquid chromatography-electrospray ionization-tandem mass spectrometry system. Based on preset impact criteria, target metabolites are screened from the test results. The concentration data corresponding to each target metabolite is then extracted from the test data to form a complete set of target metabolites and concentration data for subsequent analysis steps.

[0031] For example, at least one target metabolite may include: N-methyl-trans-4-hydroxy-proline, rhamnoic acid, uridine triphosphate, 4-hydroxy-3-methoxyphenylacetic acid, hydroxyphenyl lactic acid, perilla alkaloid, N-oleoyl-L-serine, undecanoic acid (11:0), triglycerides (15:0-16:0-24:1), and triglycerides (15:0-18:1-24:1).

[0032] It should be noted that the occurrence of heart failure with preserved ejection fraction is closely related to metabolic disorders in the body. Only some metabolites have a significant impact on this type of medical event. This step obtains metabolites and their concentrations that meet the criteria for the degree of impact on the medical event by targeting them. Metabolites with no substantial correlation are removed from the data source to ensure that the subsequent calculation data is highly correlated with the medical event, making the probability assessment more targeted and avoiding the bias caused by invalid data.

[0033] S120. Obtain the effect value corresponding to each target metabolite from the correspondence between target metabolites and effect values.

[0034] In S120, the effect size characterizes the impact of the target metabolite on medical events in patients with preserved ejection fraction (PEF). Its magnitude is positively correlated with the degree of impact of the metabolite on the medical event and serves as a key weighting factor in subsequent weighted summation calculations. The correspondence between target metabolites and effect sizes is based on a large-scale cohort study of PEF patients, and the matching relationship between each target metabolite and its corresponding effect size was obtained through statistical model analysis. This relationship was pre-established and validated and can be directly used for clinical assessment.

[0035] The correspondence between target metabolites and effect sizes is shown in Table 1: Table 1. Correspondence between target metabolites and effect sizes

[0036] In practice, a correspondence table between target metabolites and effect values ​​can be pre-established and stored. The selected target metabolites are used as the retrieval basis, and one by one they are matched in the correspondence table to extract the effect value corresponding to each target metabolite, forming a set of effect values ​​corresponding to the target metabolites and concentration data, ensuring that each target metabolite has a corresponding quantitative weight index.

[0037] It should be noted that the impact of different target metabolites on medical events related to heart failure with preserved ejection fraction varies. Directly using concentration data for calculations would not reflect the actual differences in the effects of each metabolite. This step quantifies the impact of each target metabolite by matching it to pre-determined effect values, providing a scientific basis for subsequent weighted calculations. This allows the concentration data of each metabolite to participate in probability calculations based on its actual contribution, improving the scientific rigor of the results.

[0038] S130. The concentration data of each target metabolite and the corresponding effect value of each target metabolite are weighted and summed to determine the probability of the target patient experiencing the medical event of heart failure with preserved ejection fraction.

[0039] In S130, weighted summation refers to the calculation method of multiplying the concentration data of each target metabolite with its corresponding effect value, and then summing all the product results. The effect value serves as the weight, reflecting the different contributions of each metabolite in the probability calculation. The probability of occurrence of medical events in heart failure with preserved ejection fraction refers to the quantitative value obtained by integrating the effects of various target metabolites, indicating the likelihood of this type of medical event in the target patient. The magnitude of the value directly reflects the patient's risk of developing this type of medical event.

[0040] In practice, for each target metabolite, its concentration data can be multiplied with its corresponding effect value to obtain a separate contribution value of each target metabolite to the probability of the occurrence of the medical event. After completing the multiplication of all target metabolites, all separate contribution values ​​are summed. The summation result is the probability of the target patient experiencing the medical event of heart failure with preserved ejection fraction. This value is directly output as the evaluation result.

[0041] In some embodiments, the probability of a target patient experiencing a heart failure event with preserved ejection fraction is determined by weighted summation of the concentration data of each target metabolite and the corresponding effect value of each target metabolite. This may include: for each target metabolite, determining the initial probability of a heart failure event with preserved ejection fraction for each target metabolite based on the product of the effect value of the target metabolite and the concentration data of the target metabolite; and summing the initial probabilities of heart failure events with preserved ejection fraction to determine the probability of a target patient experiencing a heart failure event with preserved ejection fraction.

[0042] The probability of a target patient experiencing a medical event of heart failure with preserved ejection fraction (HFpEF) can also be referred to as the HFpEF prognostic risk multimetabolite risk score (MRS):

[0043] Where i represents the metabolite, m represents the total number of metabolites, β represents the effect value of the metabolite on the prognostic risk of heart failure in the training set of 250 people through the Cox proportional hazards regression model, j represents the serum level of the metabolite, and the MRS value is the sum of the effect values ​​of all metabolites related to the prognostic risk of heart failure.

[0044] It should be noted that the concentration data of a single metabolite cannot fully reflect the overall risk of a patient developing heart failure with preserved ejection fraction. By using a weighted summation method, the concentration data of each target metabolite is combined with its quantified effect value, integrating the contribution of all core metabolites. This transforms the patient's metabolic risk into a specific probability value, achieving accurate quantification of the risk of medical events and making the assessment results more valuable. At the same time, the calculation logic is simple and easy to operate quickly in clinical practice.

[0045] In this embodiment of the application, the higher the probability of the target patient experiencing a medical event of heart failure with preserved ejection fraction, the greater the 4-year all-cause mortality risk of the target patient. That is, the higher the MRS score of the prognostic risk of HFpEF heart failure composed of metabolites, the greater the 4-year all-cause mortality risk of HFpEF heart failure patients.

[0046] The medical event probability determination method, apparatus, device, computer storage medium, and computer program product of this application have the following advantages. Firstly, the target metabolites used in this method are those with a greater than preset impact on medical events of preserved ejection fraction heart failure. These metabolites are not randomly selected but have a direct and significant correlation with the occurrence of preserved ejection fraction heart failure. This eliminates interference from metabolites with no substantial or weak correlation to the medical event, ensuring that subsequent concentration detection and data calculation based on these metabolites revolve around the core influencing factors of the medical event. This avoids interference from invalid information on the probability determination results from the data source, guaranteeing a high degree of consistency between the probability determination process and the medical event of preserved ejection fraction heart failure. Secondly, the effect value, as a quantitative indicator characterizing the impact of the target metabolite on medical events of preserved ejection fraction heart failure, accurately reflects the magnitude of the role of different target metabolites in the occurrence of the medical event—metabolites with a greater impact on the medical event correspond to higher effect values, and their concentration data has a higher weight in the probability calculation; metabolites with a smaller impact correspond to lower effect values, and their weight percentage is correspondingly reduced. This approach breaks away from the drawbacks of traditional assessments that "average values" are assigned to each influencing factor. It allows the concentration data of each target metabolite to participate in the calculation based on its actual contribution, making the logic of probability calculation more consistent with the occurrence principle of medical events in heart failure with preserved ejection fraction. The calculation process is more scientific and reasonable. Furthermore, by weighted summing of concentration data and effect values, the probability of medical events is accurately quantified, thereby improving the accuracy of predicting the occurrence of medical events in heart failure with preserved ejection fraction in target patients.

[0047] In some embodiments, such as Figure 2 As shown, at least one target metabolite can be obtained in the following manner: S210. Obtain serum samples from the target patient, perform metabolomics testing on the serum samples, and obtain an initial pool of candidate metabolites.

[0048] Serum samples are blood supernatants collected from and processed from the target patient. They do not contain blood cells or other components, but are enriched with various metabolites from the body, making them a commonly used sample for metabolomics testing.

[0049] The initial candidate metabolite pool is a collection of all metabolites obtained after metabolomics testing of serum samples, including all metabolites related to and unrelated to the medical event.

[0050] In practice, venous blood can be collected from target patients hospitalized due to medical events of heart failure with preserved ejection fraction. Serum samples can be obtained by centrifugation. A liquid chromatography-electrospray ionization-tandem mass spectrometry system can be used to perform broad-target full-spectrum metabolomics detection on the serum samples, collect information on all metabolites obtained, and integrate them to form an initial candidate metabolite pool.

[0051] By selecting serum samples as the target for metabolomics testing, which are enriched with various metabolites in the body, the results can directly reflect the physiological and pathological metabolic state of patients and are highly representative. By obtaining an initial candidate metabolite pool through broad-target full-spectrum metabolomics testing, the metabolite information in patients can be comprehensively captured, ensuring that no potential metabolites related to medical events of heart failure with preserved ejection fraction are missed, thus laying a complete and comprehensive metabolite data foundation for subsequent screening.

[0052] S220. Perform data quality control processing on each candidate metabolite in the initial candidate metabolite pool to obtain the second candidate metabolite pool.

[0053] Data quality control is the process of inspecting and screening metabolite detection data, with the core objective of ensuring data stability and integrity.

[0054] The second candidate metabolite pool is a collection of metabolites that meet the quality standards after the initial candidate metabolite pool has undergone data quality control processing.

[0055] In practice, for each candidate metabolite in the initial candidate metabolite pool, the coefficient of variation reflecting the degree of data dispersion can be calculated, and metabolites with a coefficient of variation exceeding a preset threshold can be removed; then the detection missing ratio of the remaining metabolites can be calculated, and metabolites with a missing ratio exceeding a preset threshold can be removed; the remaining metabolites after the above two steps of screening can be integrated to obtain the second candidate metabolite pool.

[0056] Data quality control was performed on the initial candidate metabolite pool to remove metabolites with poor detection stability and excessive missing values. This effectively avoided detection errors caused by factors such as experimental operation and instrument precision, ensuring that the data of metabolites entering subsequent screening all possessed good stability and completeness. High-quality basic data avoided the bias in screening results caused by low-quality data from the source, allowing all subsequent screening analyses to be based on reliable data.

[0057] S230. Differential expression screening is performed on each candidate metabolite in the second candidate metabolite pool to obtain the third candidate metabolite pool.

[0058] Differential expression screening is a process that detects the magnitude and statistical significance of differences in the expression levels of metabolites in different study populations, and then screens out metabolites with significant differences.

[0059] The third candidate metabolite pool is a collection of metabolites that show significant expression differences after differential expression screening of the second candidate metabolite pool.

[0060] In practice, for each candidate metabolite in the second candidate metabolite pool, the fold change in expression of each metabolite in different prognostic populations of heart failure with preserved ejection fraction can be calculated, and statistical tests can be performed to obtain significance indicators. Metabolites with fold changes exceeding a preset threshold and significance indicators meeting preset standards can be screened out and integrated to form the third candidate metabolite pool.

[0061] Based on the differences in metabolite expression among different prognostic populations of heart failure with preserved ejection fraction, differential expression screening was conducted to identify metabolites with significantly different expression levels, thus precisely narrowing down the range of metabolites potentially associated with the medical event. This step, through quantitative difference magnitude and statistical significance testing, eliminated irrelevant metabolites with no expression differences, significantly narrowing the subsequent screening dimensions and allowing for a more focused screening approach, effectively improving the initial association between metabolites and the medical event.

[0062] S240. Perform clinical factor correction on each candidate metabolite in the third candidate metabolite pool to obtain the fourth candidate metabolite pool.

[0063] Clinical factor correction is the process of incorporating clinical factors into the analysis model and eliminating their interference with the association between metabolites and medical events.

[0064] The fourth candidate metabolite pool is a collection of metabolites that are independently related to the medical event and are retained after the third candidate metabolite pool has been adjusted for clinical factors.

[0065] In practice, clinical factors associated with heart failure with preserved ejection fraction are identified, and each candidate metabolite in the third candidate metabolite pool is input into a multivariate regression model along with these clinical factors. Through model analysis, metabolites that are affected by clinical factors and are not independently associated with medical events are eliminated, and the remaining metabolites that are independently associated with medical events are integrated to obtain the fourth candidate metabolite pool.

[0066] Metabolite levels are easily influenced by clinical factors such as age, underlying diseases, and medication use. Clinical factor correction incorporates various clinical factors into the analysis model, eliminating metabolites whose expression differs due to clinical interference and which are not independently associated with medical events. This step ensures that the remaining metabolites are directly related to medical events in preserved ejection fraction heart failure and are unaffected by clinical factors, guaranteeing the specificity and independence of the association between metabolites and medical events, and preventing spurious associations from affecting subsequent assessment results.

[0067] S250. Perform multiple feature screening on each candidate metabolite in the fourth candidate metabolite pool to obtain at least one target metabolite.

[0068] Multiple feature screening is a process that uses multiple feature screening methods to comprehensively screen metabolite features. Its core is to remove noise and redundant features and retain the core features.

[0069] In practice, cross-validation can be used to select the optimal parameters, feature screening can be performed on the fourth candidate metabolite pool to remove metabolites whose feature coefficients are compressed to a preset value; then, by generating shadow features, noisy metabolites with importance scores lower than the shadow features can be removed; finally, independent variable correlation verification can be performed on the remaining metabolites to remove redundant metabolites with excessive correlation; the remaining metabolites after the above multi-step screening are determined as target metabolites.

[0070] Multi-feature screening employs a combination of methods to further eliminate weakly correlated, noisy, and redundant features, achieving final purification of core metabolites. This step ensures that the final target metabolites are all core indicators significantly impacting medical events related to heart failure with preserved ejection fraction, and that each metabolite is highly independent with no feature overlap. This avoids interference from invalid features in probability calculations and prevents computational biases caused by repeated weighting, maximizing the effectiveness and conciseness of the target metabolites.

[0071] The entire screening process is progressive and interconnected, with each step having a clear screening objective and scientific basis. Through multiple rounds of purification, precise screening of target metabolites is achieved. At the same time, the number of core metabolites is significantly reduced after multiple rounds of screening, decreasing the number of indicators for subsequent clinical testing, lowering testing costs and operational complexity, and making subsequent concentration detection and probability calculations simpler and more efficient. This aligns with the actual scenarios of routine clinical testing and data processing, improving the overall clinical operability and scalability of the method.

[0072] In some embodiments, performing data quality control processing on each candidate metabolite in the initial candidate metabolite pool to obtain a second candidate metabolite pool may include: First, stability testing is performed on each candidate metabolite in the initial candidate metabolite pool to obtain the coefficient of variation (CV). Stability testing is a process of analyzing the detection data of candidate metabolites to assess the overall dispersion of the data, and its core purpose is to determine the reliability of the detection data. The coefficient of variation (CV) is used to measure the degree of data dispersion; its value is positively correlated with the degree of data dispersion. The lower the CV value, the more stable the detection data of the metabolite, and the more reliable the detection results. A larger CV value indicates greater fluctuations in the detection results, which may be due to instrument instability, inconsistent sample processing, or instability of the metabolite itself.

[0073] In practice, all detection data of each candidate metabolite in the initial candidate metabolite pool can be retrieved, and the coefficient of variation of the detection data of each candidate metabolite can be calculated according to statistical methods. The stability test of all candidate metabolites can be completed, and the coefficient of variation value corresponding to each candidate metabolite can be recorded to form a complete stability test result.

[0074] Then, candidate metabolites with a coefficient of variation greater than a preset threshold are removed to obtain the fifth candidate metabolite pool.

[0075] The preset threshold is a value set in advance based on industry standards and research needs for metabolomics detection. It serves as a quantitative standard for judging whether metabolite detection data meets stability requirements. As an example, the preset threshold could be 0.3. The fifth candidate metabolite pool is the set of candidate metabolites with stable detection data remaining after removing metabolites with a coefficient of variation greater than the preset threshold from the initial candidate metabolite pool.

[0076] In practice, the coefficient of variation of each candidate metabolite can be compared with a preset threshold one by one. Candidate metabolites with a coefficient of variation less than or equal to the preset threshold are selected and integrated into a fifth candidate metabolite pool. Candidate metabolites with a coefficient of variation greater than the preset threshold are directly removed and will not be included in subsequent analysis.

[0077] Next, integrity checks were performed on each candidate metabolite in the fifth candidate metabolite pool to obtain the missing percentage.

[0078] Integrity testing is a process of statistically analyzing the detection data of candidate metabolites to assess the degree of data loss. Its core purpose is to determine the completeness of the detection data.

[0079] The missing percentage represents the proportion of serum samples in which each candidate metabolite was not detected out of the total number of samples. Its value is negatively correlated with data integrity; the lower the value, the more complete the metabolite detection data.

[0080] In practice, for each candidate metabolite in the fifth candidate metabolite pool, the number of samples that could not be detected in all serum samples can be counted. The ratio of this number to the total number of samples can be calculated to obtain the missing proportion for each candidate metabolite. The integrity detection of all candidate metabolites can be completed, and the missing proportion value for each candidate metabolite can be recorded.

[0081] Finally, candidate metabolites with a missing proportion greater than a preset threshold are removed to obtain a second candidate metabolite pool.

[0082] As an example, the preset threshold can be 50%. In practice, the missing proportion of each candidate metabolite can be compared with the preset threshold one by one. Candidate metabolites with a missing proportion less than or equal to the preset threshold are selected and integrated into a second candidate metabolite pool. Candidate metabolites with a missing proportion greater than the preset threshold are removed and will no longer participate in the subsequent target metabolite screening steps.

[0083] It should be noted that after obtaining the second candidate metabolite pool, the metabolite data in the second candidate metabolite pool also need to be imputed and normalized. For example, missing values ​​are imputed by half of the minimum value of the corresponding metabolite, and the metabolite values ​​are processed by Log2 to make the data conform to a normal distribution.

[0084] By employing a two-step progressive quality control design involving stability and integrity testing, candidate metabolites with substandard test data can be eliminated at each stage. This approach controls the quality of metabolite data from two core dimensions: stability and integrity, preventing low-quality data from causing deviations in subsequent screening results. It provides a high-quality and reliable metabolite data foundation for subsequent steps such as differential expression screening, ensuring the scientific rigor and accuracy of the entire target metabolite screening process.

[0085] In some embodiments, differential expression screening is performed on each candidate metabolite in the second candidate metabolite pool to obtain a third candidate metabolite pool, which may include: First, the difference magnitude is detected for each candidate metabolite in the second candidate metabolite pool to obtain the difference fold.

[0086] The differential amplitude detection is a process of analyzing the variation of the expression level of candidate metabolites in different prognostic populations of heart failure with preserved ejection fraction.

[0087] The fold difference is an indicator that reflects the magnitude of change in metabolite expression between survivors and deceased individuals in medical events of heart failure with preserved ejection fraction. The value directly reflects the degree of difference in metabolite expression between the two groups.

[0088] In practice, the expression data of each candidate metabolite in the second candidate metabolite pool in both surviving and deceased subjects can be retrieved. The fold change of each candidate metabolite between the two groups can be calculated using statistical methods. The difference magnitude of all candidate metabolites can be detected, and the fold change value corresponding to each candidate metabolite can be recorded.

[0089] Subsequently, statistical analysis was performed on each candidate metabolite in the second candidate metabolite pool to obtain significance indicators.

[0090] Statistical differential detection is a process that uses statistical testing methods to analyze whether the expression differences of candidate metabolites in different prognostic populations are statistically significant.

[0091] Significance indicators are indicators that reflect the statistical significance of differences in metabolite expression. Changes in their values ​​can reflect the probability that the expression differences are caused by random factors.

[0092] In practice, for each candidate metabolite in the second candidate metabolite pool, its expression data in surviving and dead subjects can be incorporated into a statistical test model to complete the differential statistical test, calculate the significance index corresponding to each candidate metabolite, record the significance index values ​​of all candidate metabolites, and form a complete test result.

[0093] Finally, candidate metabolites with a difference fold greater than a preset threshold and a significance index less than a preset threshold are selected from the second candidate metabolite pool to obtain the third candidate metabolite pool.

[0094] In practice, the fold change and significance index of each candidate metabolite can be compared with the corresponding preset thresholds. Candidate metabolites that simultaneously meet the requirements of a fold change greater than the preset threshold and a significance index less than the preset threshold can be selected. These metabolites are then integrated into a third candidate metabolite pool. Candidate metabolites that do not meet the dual threshold requirements are removed and will not be included in subsequent screening and analysis.

[0095] For example, Figure 3 and Figure 4 The volcano plot and OPLS-DA analysis results are presented. Finally, based on the threshold (|log2FC|>0.585 and FDR-P value<0.05), a third pool of candidate metabolites that significantly contributed to mortality outcomes was selected. These metabolites mainly include alcohols and amines, amino acids and their metabolites, carbohydrates and their metabolites, fatty acids, glycerides, nucleotides and their metabolites, organic acids and their derivatives, etc.

[0096] It should be noted that the preset fold change threshold and significance threshold are the qualification standards for the magnitude and statistical significance of metabolite expression differences, respectively. Only when a metabolite simultaneously meets the conditions of a fold change greater than the preset threshold and a significance threshold less than the preset threshold is its expression difference in different prognostic populations considered sufficiently significant and real. By using dual thresholds for screening, metabolites without significant expression differences can be accurately eliminated, while metabolites with potential association with medical events can be retained, forming a third pool of candidate metabolites, thus achieving precise narrowing of the screening scope.

[0097] By employing a dual approach of difference magnitude detection and difference statistical detection combined with threshold screening, metabolites potentially associated with heart failure with preserved ejection fraction were precisely identified based on both the magnitude of expression changes and statistical significance. This dual-detection method corroborates each other, avoiding the limitations of single-dimensional screening and effectively eliminating irrelevant metabolites without significant expression differences. This narrows the scope of subsequent screening, ensuring that the remaining metabolites exhibit characteristic expression in different prognostic populations. This provides a precise screening foundation for subsequent clinical factor correction steps, guaranteeing the association between target metabolites and medical events.

[0098] In some embodiments, each candidate metabolite in the third candidate metabolite pool is subjected to clinical factor correction to obtain a fourth candidate metabolite pool, which may include: First, each candidate metabolite in the third candidate pool and the preset clinical factors are input into a multivariate regression model to obtain the interference results of clinical factors on medical events of heart failure with preserved ejection fraction.

[0099] Pre-defined clinical factors are clinically relevant indicators that affect metabolism or disease progression, determined in advance based on the diagnostic and treatment characteristics of heart failure with preserved ejection fraction. They are the core reference factors for analyzing interference effects. Clinical factors may include age, sex, body mass index, coronary artery disease, diabetes, systolic blood pressure, dyslipidemia, statins, smoking, estimated glomerular filtration rate, stroke, atrial fibrillation, New York Heart Association functional classification, chronic obstructive pulmonary disease, and N-terminal pro-brain natriuretic peptide (PNP), etc.

[0100] Multivariate regression models are statistical models that can simultaneously analyze the relationship between multiple variables and the research objective, and can quantitatively determine the independent effect of each variable on the research objective.

[0101] Interference results are quantitative results used to characterize the degree of interference of clinical factors on the association between metabolites and medical events of heart failure with preserved ejection fraction, and can reflect whether the association of metabolites is affected by clinical factors.

[0102] In practice, pre-defined clinical factors related to heart failure with preserved ejection fraction can be identified, relevant data and corresponding clinical factor data for each candidate metabolite in the third candidate metabolite pool can be compiled, and the two types of data can be uniformly input into a multivariate regression model. Through model calculation and analysis, the interference results of clinical factors corresponding to each candidate metabolite can be obtained, complete interference result data can be recorded, and an analysis report can be generated.

[0103] Subsequently, based on the interference results, candidate metabolites independently associated with medical events of heart failure with preserved ejection fraction were screened from the third candidate pool to obtain the fourth candidate metabolite pool.

[0104] Independent association refers to the relationship between metabolites and medical events of heart failure with preserved ejection fraction, which is not affected by any pre-set clinical factors and is the direct effect of the metabolites themselves on the medical event.

[0105] The fourth candidate metabolite pool is a collection of candidate metabolites that are independently associated with medical events of heart failure with preserved ejection fraction, which are retained after the interference results are screened from the third candidate metabolite pool.

[0106] In practice, a criterion for judging the degree of interference from clinical factors can be set. The interference results of each candidate metabolite can be compared with the criterion one by one. Candidate metabolites whose interference results meet the criterion and have an independent correlation with the medical event of heart failure with preserved ejection fraction can be screened out. These metabolites can be integrated into a fourth candidate metabolite pool. Candidate metabolites that are significantly interfered with by clinical factors and have no independent correlation can be removed and will not be included in the subsequent target metabolite screening steps.

[0107] By incorporating metabolites and pre-defined clinical factors into a multivariate regression model analysis and combining interference results to screen for independently relevant metabolites, this design eliminates the interference of clinical factors such as age and underlying diseases on the association between metabolites and the medical event of heart failure with preserved ejection fraction. This accurately screens out metabolites that are directly and independently related to the medical event, avoids spurious association metabolites caused by clinical factors from entering subsequent screening, ensures the specificity of the association between retained metabolites and the medical event, and provides an accurate and reliable metabolite data foundation for subsequent multi-feature screening.

[0108] In some embodiments, performing multiple feature screening on each candidate metabolite in the fourth candidate metabolite pool to obtain at least one target metabolite may include: First, cross-validation is used to select the optimal regularization parameter. Using the optimal regularization parameter, metabolites whose feature coefficients in the fourth candidate metabolite pool are compressed to a preset value are removed to obtain the sixth candidate metabolite pool.

[0109] Cross-validation is a statistical parameter selection method that determines the optimal parameters of a model by validating data in groups, thus avoiding parameter bias caused by training on a single dataset.

[0110] The optimal regularization parameter is a model parameter obtained through cross-validation that can accurately distinguish the contribution of metabolite features and is used to compress the coefficients of weakly correlated features.

[0111] The sixth candidate metabolite pool is the set of candidate metabolites whose feature contribution meets the standard after removing metabolites whose feature coefficients have been compressed to a preset value from the fourth candidate metabolite pool.

[0112] In practice, the feature data of the fourth candidate metabolite pool can be incorporated into the model, and the cross-validation method can be used to conduct multiple rounds of data grouping training and validation to select the optimal regularization parameter of the model. This parameter is then applied to metabolite feature analysis to remove metabolites whose feature coefficients are compressed to a preset value (e.g., 0), and the remaining metabolites are integrated to obtain the sixth candidate metabolite pool.

[0113] Next, the original features and shadow features of each candidate metabolite in the sixth candidate metabolite pool are obtained. The shadow features are generated by randomly shuffling the original features. The shadow features are similar in distribution to the original features but are unrelated to the original target.

[0114] In practice, the original feature data corresponding to each candidate metabolite in the sixth candidate metabolite pool can be extracted to form an original feature set. According to the numerical distribution pattern of the original features, each original feature is randomly shuffled to generate corresponding shadow feature data, ensuring that each original feature has a matching shadow feature, thus forming a complete feature comparison set.

[0115] Then, the original features and shadow features are input into the random forest model to obtain the importance scores of the original features and shadow features, respectively.

[0116] Random forest is a statistical model of ensemble learning that can efficiently analyze the relationship between multiple features and research objectives and quantify the importance scores of each feature.

[0117] Importance score is a quantitative indicator that characterizes the degree of influence of a feature on medical events of heart failure with preserved ejection fraction. The higher the score, the stronger the feature's contribution to the research objective and its effectiveness.

[0118] In practice, the sorted original feature and shadow feature data can be uniformly input into the random forest model, set the model analysis target related to the medical event of heart failure with preserved ejection fraction, and run the model to perform feature contribution analysis; after the model calculation is completed, extract the importance score of each original feature and the corresponding shadow feature, and record the complete score data.

[0119] Subsequently, based on the importance scores of the original features and the shadow features, noisy features with lower importance scores than the shadow features in the original features were removed, resulting in the seventh candidate metabolite pool.

[0120] Noise features refer to raw features that appear to be relevant to the research objective but actually contribute nothing and are merely random data interference. These features can affect the accuracy of subsequent analysis.

[0121] The seventh candidate metabolite pool is the set of candidate metabolites with valid original features remaining after removing noisy features from the sixth candidate metabolite pool.

[0122] In practice, the importance score of each original feature can be compared with the corresponding shadow feature score one by one. Metabolites with original feature scores lower than shadow feature scores are eliminated, and only metabolites with original feature scores higher than shadow feature scores are retained. The retained metabolites are integrated to form the seventh candidate metabolite pool, in which all metabolites are core candidate metabolites with effective features.

[0123] Finally, at least one target metabolite is obtained from the seventh candidate metabolite pool.

[0124] The overall approach employs a progressive design that combines initial screening using cross-validation with fine-tuning using a random forest model and shadow features. This process refines core metabolites layer by layer from two dimensions: feature contribution and feature effectiveness. Weakly correlated features and noisy features are eliminated to ensure that the final target metabolites are all effective features that significantly impact the medical event of ejection fraction-preserving heart failure. This avoids invalid features interfering with the accuracy of subsequent probability calculations and provides a high-quality core indicator foundation for determining the probability of medical events.

[0125] In some embodiments, obtaining at least one target metabolite from the seventh candidate metabolite pool may include: First, independent variable correlation was performed on each candidate metabolite in the seventh candidate metabolite pool to obtain the correlation coefficients between each candidate metabolite.

[0126] Independent variable correlation verification is a process of analyzing the degree of linear correlation between different metabolites based on the feature data of multiple candidate metabolites. Its core purpose is to determine whether there is feature redundancy in metabolites.

[0127] The correlation coefficient is a quantitative indicator that characterizes the degree of association between two candidate metabolites. The value directly reflects the degree of feature overlap between metabolites. The higher the value, the stronger the association between metabolites and the higher the feature redundancy.

[0128] In practice, feature data of all candidate metabolites in the seventh candidate metabolite pool can be extracted, and each pair of metabolites can be grouped together. The correlation coefficient between each group of metabolites can be calculated using statistical methods to complete the independent variable correlation verification of all candidate metabolites. The correlation coefficient value of each group of metabolites can be recorded to form a complete verification result.

[0129] Finally, candidate metabolites with correlation coefficients higher than a preset threshold are removed to obtain at least one target metabolite.

[0130] The target metabolite is the core metabolite that remains in the seventh candidate metabolite pool after correlation verification and removal of highly correlated metabolites. It is highly independent of each other and highly correlated with the medical event of heart failure with preserved ejection fraction. It is the core indicator for subsequent calculation of the probability of medical events.

[0131] In practice, the correlation coefficient of each group of metabolites can be compared with a preset threshold one by one. For the group of metabolites with a correlation coefficient higher than the preset threshold, metabolites with a relatively weak impact on medical events of heart failure with preserved ejection fraction can be removed. All the remaining metabolites after screening can be integrated to obtain at least one target metabolite, which can be used in the subsequent steps of determining the probability of medical events.

[0132] It should be noted that, in this embodiment, hospitalized patients with heart failure from 20 provinces across China were included between 2016 and 2018. The inclusion criteria were heart failure patients aged 18 years and older diagnosed by collaborating hospitals, with heart failure being the primary cause of hospitalization, including newly diagnosed heart failure and exacerbations of chronic heart failure, as well as LVEF > 50% and NT-proBNP > 300 ng / L. Hospitalization case information of the included patients was extracted, and long-term follow-up was conducted until 2023, including 4 years of mortality data. After excluding patients with malignant tumors, renal failure (eGFR < 15 mL / min / 1.73 m²), and cirrhosis, 500 patients were randomly selected for metabolomics analysis. In this embodiment, the study population was randomly divided into a training set and a validation set (250 patients each) at a 1:1 ratio.

[0133] It should be noted that the prognostic risks of heart failure include at least one of the following situations: patients diagnosed with heart failure based on the "Chinese Guidelines for the Diagnosis and Treatment of Heart Failure".

[0134] Inclusion criteria: 1. Echocardiography showing a left ventricular ejection fraction (LVEF) >50%; 2. N-terminal pro-brain natriuretic peptide (NT-proBNP) >300 ng / L; 3. All-cause mortality risk within 4 years.

[0135] Exclusion criteria: 1. Malignant tumors; 2. Renal failure (eGFR <15mL / min / 1.73m²); 3. Liver cirrhosis.

[0136] In this embodiment, a broad-target full-spectrum metabolomics assay was performed on 500 individuals, detecting a total of 1896 metabolites. During quality control, metabolites with a coefficient of variation (CV) greater than 0.3 were removed (CV reflects the dispersion of metabolite levels in the quality control samples; the lower the value, the better the stability of the detection system). Metabolites with a missing proportion greater than 50% were removed, and half of the minimum value of each metabolite was used for imputation. The metabolite values ​​were then processed using Log2 to ensure they conformed to a normal distribution, resulting in 1884 metabolites that were subsequently analyzed.

[0137] To accurately screen differentially expressed metabolites in the training set, we combined multi-step statistical analysis with machine learning methods. Figure 3 and Figure 4 The volcano plot and OPLS-DA analysis results are presented. Ultimately, based on a threshold (|log2FC|>0.585 and FDR-P value<0.05), 102 candidate metabolites significantly contributing to mortality were selected. These metabolites mainly include alcohols and amines, amino acids and their metabolites, carbohydrates and their metabolites, fatty acids, glycerides, nucleotides and their metabolites, organic acids and their derivatives, etc. After adjusting for clinical factors (including age, sex, body mass index, coronary artery disease, diabetes, systolic blood pressure, dyslipidemia, statins, smoking, estimated glomerular filtration rate, stroke, atrial fibrillation, New York Heart Association functional class, chronic obstructive pulmonary disease, and N-terminal pro-brain natriuretic peptide), 41 metabolites were significantly associated with the 4-year all-cause mortality risk. The optimal parameter (λ) in the LASSO model was selected using 10-fold cross-validation (LASSO is an efficient embedded feature selection method that automatically compresses unimportant feature coefficients to 0 during model training, resulting in an accurate and concise model). After applying the Boruta algorithm (which uses an iterative process to filter out the feature variables that truly affect the results and remove random noise features with low information content, thus simplifying the model) and performing correlation tests on independent variables, 10 metabolites (perillyl alkaloid, N-oleoyl-L-serine, N-methyltrans-4-hydroxyproline, rhamnoic acid, free fatty acids (11:0), triglycerides (15:0-16:0-24:1), triglycerides (15:0-18:1-24:1), uridine triphosphate, 4-hydroxy-3-methoxyphenylacetic acid, and hydroxyphenyllactic acid) were finally identified as suitable for predicting 4-year survival status. Based on these 10 significant metabolites, a multi-metabolite risk score (MRS) suitable for Chinese people was constructed.

[0138] For example, Figures 3-6This is a metabolic profile of HFpEF patients with changes in all-cause mortality over 4 years, including... Figure 3 The volcano plot shows the five metabolites with the most significant downregulation and the five metabolites with the most significant upregulation in the 4-year all-cause mortality group. Red dots represent metabolites with an increasing trend in expression in the 4-year all-cause mortality group, and blue dots represent metabolites with a decreasing trend in expression. The names of the five metabolites with the most significant downregulation and the five metabolites with the most significant upregulation are also labeled. Figure 4 This is an orthogonal partial least squares discriminant analysis (OPLS-DA) plot, which distinguishes between patients in the 4-year mortality group and the control group; Figure 5 Metabolic features selected by the Boruta algorithm; Figure 6 Heatmap of differential metabolites associated with 4-year mortality events.

[0139] The relationship between the multimetabolite risk score (MRS) and 4-year all-cause mortality risk in heart failure patients with heart failure pEF was assessed in the validation set. Results showed that a higher MRS score was associated with a greater 4-year all-cause mortality risk. Figures 7-10As shown, the metabolite model composed of the aforementioned 10 significant metabolites performed well in predicting 4-year all-cause mortality on both the training and validation sets. The metabolite model (training set: AUC 0.929, 95% CI 0.896–0.961; validation set: AUC 0.878, 95% CI 0.835–0.921) outperformed existing risk models, including the MAGGIC score (training set: AUC 0.668, 95% CI 0.600–0.735; validation set: AUC 0.672, 95% CI 0.605–0.739) and the EFFECT score (training set: AUC 0.698, 95% CI 0.632–0.763; validation set: AUC 0.694, 95% CI 0.694). The statistical significance (AUC) was 0.629–0.759, and a clinical risk score comprised of traditional significant clinical variables (including age, sex, body mass index, coronary artery disease, diabetes, systolic blood pressure, dyslipidemia, statins, smoking, estimated glomerular filtration rate, stroke, atrial fibrillation, New York Heart Association classification, chronic obstructive pulmonary disease, and N-terminal pro-brain natriuretic peptide) was also assessed (training set: AUC 0.762, 95% CI 0.702–0.822; validation set: AUC 0.707, 95% CI 0.642–0.772). Furthermore, we constructed a comprehensive model with 10 metabolites and clinical variables, which demonstrated good predictive performance (training set: AUC 0.937, 95% CI 0.904–0.970; validation set: AUC 0.884, 95% CI 0.840–0.928). Next, the bootstrap method was used to validate these models across all samples, and the results remained robust, with the metabolite model showing superior performance (AUC 0.907, 95% CI 0.879–0.928). The metabolite model also robustly predicted 4-year cardiovascular mortality (training set: AUC 0.858, 95% CI 0.810–0.907; validation set: AUC 0.844, 95% CI 0.794–0.894). Furthermore, compared to the clinical risk score, the metabolite model significantly improved patient classification in both 4-year all-cause mortality and cardiovascular mortality (P<0.001).

[0140] The target metabolite categories and their associated risks are shown in Table 2, which illustrates the association between 10 significant metabolites in the prediction model and 4-year all-cause mortality.

[0141] Table 2 Target Metabolite Categories and Risk Ratios

[0142] Figures 7-10 The model demonstrates the metabolites and their performance in the prediction model, among which... Figure 7To train receiver operating characteristic (ROC) curves for predicting 4-year all-cause mortality risk using concentrated metabolites; Figure 8 To validate the receiver operating characteristic (ROC) curves for predicting 4-year all-cause mortality risk using concentrated metabolites; Figure 9 To train receiver operating characteristic (ROC) curves for predicting 4-year cardiovascular mortality risk using concentrated metabolites; Figure 10 To validate the receiver operating characteristic (ROC) curves of concentrated metabolites in predicting the risk of cardiovascular death over 4 years.

[0143] For ease of clinical use, the MRS can be further divided into high quantile, middle quantile, and low quantile groups according to tertiles. Table 3 shows the association between training set metabolite scores and 4-year mortality risk, and Table 4 shows the association between validation set metabolite scores and 4-year mortality risk.

[0144] Table 3. Association between training set metabolite scores and 4-year mortality risk

[0145] Table 4. Association between validation set metabolite scores and 4-year mortality risk

[0146] like Figure 11 and Figure 12 As shown, Figure 11 The Kaplan-Meier curves of the all-cause mortality prediction model over the four years of the training set are shown. Figure 12 Kaplan-Meier curves for the 4-year all-cause mortality prediction model in the validation set are presented. For example, based on Table 3, in the training set, compared to the low quantile group of HFpEF heart failure patients, the 4-year mortality risk in the middle quantile group was 5.93 times higher (HR=5.93, 95% CI: 2.24–15.65); P<0.001, and the 4-year mortality risk in the high quantile group was 32.59 times higher (HR=32.59, 95% CI: 12.66–83.89); P<0.001.

[0147] The validation set results showed that with increasing MRS scores, the higher the MRS score for HFpEF prognostic risk, the greater the 4-year all-cause mortality risk. For example, based on Table 4, compared to the low-quartile group of HFpEF patients, the 4-year mortality risk in the middle-quartile group was 2.78 times higher (HR=2.78, 95% CI: 1.32-5.86); P<0.001, and the 4-year mortality risk in the high-quartile group was 32.59 times higher (HR=22.95, 95% CI: 11.07-57.54); P<0.001.

[0148] The medical event probability determination method, apparatus, device, computer storage medium, and computer program product of this application have the following advantages. Firstly, the target metabolites used in this method are those with a greater than preset impact on medical events of preserved ejection fraction heart failure. These metabolites are not randomly selected but have a direct and significant correlation with the occurrence of preserved ejection fraction heart failure. This eliminates interference from metabolites with no substantial or weak correlation to the medical event, ensuring that subsequent concentration detection and data calculation based on these metabolites revolve around the core influencing factors of the medical event. This avoids interference from invalid information on the probability determination results from the data source, guaranteeing a high degree of consistency between the probability determination process and the medical event of preserved ejection fraction heart failure. Secondly, the effect value, as a quantitative indicator characterizing the impact of the target metabolite on medical events of preserved ejection fraction heart failure, accurately reflects the magnitude of the role of different target metabolites in the occurrence of the medical event—metabolites with a greater impact on the medical event correspond to higher effect values, and their concentration data has a higher weight in the probability calculation; metabolites with a smaller impact correspond to lower effect values, and their weight percentage is correspondingly reduced. This approach breaks away from the drawbacks of traditional assessments that "average values" are assigned to each influencing factor. It allows the concentration data of each target metabolite to participate in the calculation based on its actual contribution, making the logic of probability calculation more consistent with the occurrence principle of medical events in heart failure with preserved ejection fraction. The calculation process is more scientific and reasonable. Furthermore, by weighted summing of concentration data and effect values, the probability of medical events is accurately quantified, thereby improving the accuracy of predicting the occurrence of medical events in heart failure with preserved ejection fraction in target patients.

[0149] Based on the medical event probability determination method provided in the above embodiments, this application also provides specific implementation methods of the medical event probability determination device. Please refer to the following embodiments.

[0150] First see Figure 13 The medical event probability determination device 1300 provided in this application embodiment includes: The acquisition module 1310 is used to acquire at least one target metabolite of the target patient hospitalized due to the medical event of preserved ejection fraction heart failure, as well as the concentration data of each target metabolite in the at least one target metabolite, wherein the influence of at least one target metabolite on the medical event of preserved ejection fraction heart failure is greater than a preset level. The acquisition module 310 is also used to acquire the effect value corresponding to each target metabolite from the correspondence between target metabolites and effect values. The effect value is used to characterize the degree of influence of the target metabolite on medical events of heart failure with preserved ejection fraction. The determination module 1320 is used to perform a weighted summation of the concentration data of each target metabolite and the corresponding effect value of each target metabolite to determine the probability of the target patient experiencing the medical event of heart failure with preserved ejection fraction.

[0151] In some possible implementations, at least one target metabolite is obtained in the following manner: Obtain serum samples from the target patients, perform metabolomics analysis on the serum samples, and obtain an initial pool of candidate metabolites; Data quality control processing is performed on each candidate metabolite in the initial candidate metabolite pool to obtain the second candidate metabolite pool; Differential expression screening was performed on each candidate metabolite in the second candidate metabolite pool to obtain the third candidate metabolite pool; Each candidate metabolite in the third candidate metabolite pool was subjected to clinical factor correction to obtain the fourth candidate metabolite pool; Each candidate metabolite in the fourth candidate metabolite pool is subjected to multiple feature screening to obtain at least one target metabolite.

[0152] In some possible implementations, data quality control is performed on each candidate metabolite in the initial candidate metabolite pool to obtain a second candidate metabolite pool, including: Stability testing was performed on each candidate metabolite in the initial candidate metabolite pool to obtain the coefficient of variation; the coefficient of variation is used as an indicator to measure the degree of data dispersion. Candidate metabolites with a coefficient of variation greater than a preset threshold are removed to obtain the fifth candidate metabolite pool; Completeness testing was performed on each candidate metabolite in the fifth candidate metabolite pool to obtain the missing ratio; the missing ratio is used to represent the proportion of serum samples in which each candidate metabolite was not detected out of the total number of samples. Candidate metabolites with a missing proportion greater than a preset threshold are removed to obtain a second candidate metabolite pool.

[0153] In some possible implementations, differential expression screening is performed on each candidate metabolite in the second candidate metabolite pool to obtain a third candidate metabolite pool, including: The magnitude of difference was detected for each candidate metabolite in the second candidate metabolite pool to obtain the fold change; the fold change was used to reflect the magnitude of expression change of the metabolite between survivors and deceased subjects in the presence of medical events of heart failure with preserved ejection fraction. Statistical analysis was performed on each candidate metabolite in the second candidate metabolite pool to obtain significance indicators; The third candidate metabolite pool is obtained by screening candidate metabolites from the second candidate metabolite pool that have a difference fold greater than a preset threshold and a significance index less than a preset threshold.

[0154] In some possible implementations, each candidate metabolite in the third candidate metabolite pool is subjected to clinical factor correction to obtain a fourth candidate metabolite pool, including: Each candidate metabolite in the third candidate pool and a preset clinical factor are input into a multivariate regression model to obtain the interference results of clinical factors on medical events of heart failure with preserved ejection fraction; the interference results are used to characterize the degree of interference of clinical factors on the association between metabolites and medical events of heart failure with preserved ejection fraction. Based on the interference results, candidate metabolites that are independently associated with medical events of heart failure with preserved ejection fraction were screened from the third candidate pool to obtain the fourth candidate metabolite pool.

[0155] In some possible implementations, each candidate metabolite in the fourth candidate metabolite pool undergoes multiple feature screening to obtain at least one target metabolite, including: The optimal regularization parameter was selected by cross-validation. Using the optimal regularization parameter, metabolites whose feature coefficients were compressed to a preset value in the fourth candidate metabolite pool were removed to obtain the sixth candidate metabolite pool. Obtain the original features and shadow features of each candidate metabolite in the sixth candidate metabolite pool. The shadow features are generated by randomly shuffling the original features. The shadow features are similar in distribution to the original features but are unrelated to the original target. The original features and shadow features are input into the random forest model to obtain the importance scores of the original features and shadow features, respectively. Based on the importance scores of the original features and shadow features, noisy features with lower importance scores than shadow features in the original features are removed to obtain the seventh candidate metabolite pool. At least one target metabolite is obtained from the seventh candidate metabolite pool.

[0156] In some possible implementations, at least one target metabolite is obtained from the seventh candidate metabolite pool, including: Independent variable correlation verification was performed on each candidate metabolite in the seventh candidate metabolite pool to obtain the correlation coefficients between each candidate metabolite. Candidate metabolites with correlation coefficients higher than a preset threshold are removed to obtain at least one target metabolite.

[0157] In some possible implementations, at least one target metabolite includes: N-methyl-trans-4-hydroxy-proline, rhamnoic acid, uridine triphosphate, 4-hydroxy-3-methoxyphenylacetic acid, hydroxyphenyl lactic acid, perillaldehyde, N-oleoyl-L-serine, undecanoic acid (11:0), triglycerides (15:0-16:0-24:1), and triglycerides (15:0-18:1-24:1).

[0158] In some possible implementations, the determination module 1320 is used to determine the initial ejection fraction-preserving heart failure medical event probability for each target metabolite based on the product of the effect value corresponding to the target metabolite and the concentration data of the target metabolite. The probability of a pre-exhaustion fraction heart failure medical event is determined by summing the probabilities of each initial ejection fraction preserved heart failure medical event in the target patient.

[0159] The various modules of the medical event probability determination device provided in this application embodiment can achieve... Figure 1 It provides the functionality for each step of the method for determining the probability of medical events and achieves the corresponding technical effects. For the sake of brevity, it will not be elaborated here.

[0160] Figure 14 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0161] The electronic device 1400 may include a processor 1401 and a memory 1402 storing computer program instructions.

[0162] Specifically, the processor 1401 may include a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.

[0163] Memory 1402 may include mass storage for data or instructions. For example, and not limitingly, memory 1402 may include a hard disk drive (HDD), floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, memory 1402 may include removable or non-removable (or fixed) media. Where appropriate, memory 1402 may be internal or external to the integrated gateway disaster recovery device. In a particular embodiment, memory 1402 is non-volatile solid-state memory.

[0164] In certain embodiments, the memory may include read-only memory (ROM), random access memory (RAM), disk storage media devices, optical storage media devices, flash memory devices, and electrical, optical, or other physical / tangible memory storage devices. Thus, generally, memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., memory devices) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the method according to one aspect of this application.

[0165] The processor 1401 implements any of the medical event probability determination methods in the above embodiments by reading and executing computer program instructions stored in the memory 1402.

[0166] In some examples, the electronic device 1400 may also include a communication interface 1403 and a bus 1404. For example, Figure 14 As shown, the processor 1401, memory 1402, and communication interface 1403 are connected through bus 1404 and complete communication with each other.

[0167] The communication interface 1403 is mainly used to realize communication between various modules, devices, units and / or equipment in the embodiments of this application.

[0168] Bus 1404 includes hardware, software, or both, that couples components of an online data traffic metering device together. For example, and not as a limitation, bus 1404 may include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), HyperTransport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a Microchannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local (VLB) bus, or other suitable buses, or combinations of two or more of these. Where appropriate, bus 1404 may include one or more buses. Although specific buses are described and illustrated in embodiments of this application, any suitable bus or interconnect is contemplated herein.

[0169] For example, the electronic device 1400 can be a mobile phone, tablet computer, laptop computer, handheld computer, in-vehicle electronic device, ultra-mobile personal computer (UMPC), netbook, or personal digital assistant (PDA), etc.

[0170] The electronic device 1400 can execute the medical event probability determination method in the embodiments of this application, thereby achieving a combination of Figure 1 The method for determining the probability of medical events described.

[0171] In addition, in conjunction with the medical event probability determination method in the above embodiments, this application also provides a computer-readable storage medium for implementation. The computer-readable storage medium stores computer program instructions; when these computer program instructions are executed by a processor, they implement any of the medical event probability determination methods in the above embodiments. Examples of computer-readable storage media include non-transitory computer-readable storage media, such as portable disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, etc.

[0172] This application also provides a computer program product, which includes a computer program or instructions. When the computer program or instructions are executed by a processor, they implement any of the medical event probability determination methods in the above embodiments.

[0173] It should also be noted that the exemplary embodiments mentioned in this application describe methods or systems based on a series of steps or apparatus. However, this application is not limited to the order of the above steps; that is, the steps can be performed in the order mentioned in the embodiments, or in a different order, or several steps can be performed simultaneously.

[0174] The aspects of this disclosure have been described above with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It should be understood that each block in the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that these instructions, executable via the processor of the computer or other programmable data processing apparatus, enable the implementation of the functions / actions specified in one or more blocks of the flowchart illustrations and / or block diagrams. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field-programmable logic circuit. It is also understood that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can also be implemented by special-purpose hardware performing the specified functions or actions, or can be implemented by a combination of special-purpose hardware and computer instructions.

[0175] The above are merely specific embodiments of this application. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, modules, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. It should be understood that the protection scope of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the protection scope of this application.

Claims

1. A method for determining the probability of a medical event, characterized in that, include: Acquire at least one target metabolite and concentration data of each target metabolite in a target patient hospitalized for a medical event of preserved ejection fraction heart failure, wherein the influence of the at least one target metabolite on the medical event of preserved ejection fraction heart failure is greater than a preset level; The effect value corresponding to each target metabolite is obtained from the correspondence between target metabolites and effect values. The effect value is used to characterize the degree of influence of the target metabolite on medical events of heart failure with preserved ejection fraction. The concentration data of each target metabolite and the corresponding effect value of each target metabolite are weighted and summed to determine the probability of the target patient experiencing a medical event of heart failure with preserved ejection fraction.

2. The method according to claim 1, characterized in that, The at least one target metabolite is obtained in the following manner: Serum samples were obtained from the target patient, and metabolomics analysis was performed on the serum samples to obtain an initial pool of candidate metabolites. Data quality control processing is performed on each candidate metabolite in the initial candidate metabolite pool to obtain a second candidate metabolite pool; Differential expression screening was performed on each candidate metabolite in the second candidate metabolite pool to obtain the third candidate metabolite pool; Each candidate metabolite in the third candidate metabolite pool is subjected to clinical factor correction to obtain the fourth candidate metabolite pool; Each candidate metabolite in the fourth candidate metabolite pool is subjected to multiple feature screening to obtain the at least one target metabolite.

3. The method according to claim 2, characterized in that, The step of performing data quality control processing on each candidate metabolite in the initial candidate metabolite pool to obtain a second candidate metabolite pool includes: Stability testing is performed on each candidate metabolite in the initial candidate metabolite pool to obtain the coefficient of variation; the coefficient of variation is used as an indicator to measure the degree of data dispersion. Candidate metabolites with a coefficient of variation greater than a preset threshold are removed to obtain a fifth candidate metabolite pool; Integrity testing is performed on each candidate metabolite in the fifth candidate metabolite pool to obtain the missing ratio; the missing ratio is used to represent the proportion of the total number of serum test samples in which each candidate metabolite was not detected. Candidate metabolites with a missing ratio greater than a preset threshold are removed to obtain a second candidate metabolite pool.

4. The method according to claim 2 or 3, characterized in that, The differential expression screening process for each candidate metabolite in the second candidate metabolite pool to obtain the third candidate metabolite pool includes: The difference magnitude is detected for each candidate metabolite in the second candidate metabolite pool to obtain the difference fold; the difference fold is used to reflect the expression change of the metabolite between survivors and deceased subjects in the presence of medical events of heart failure with preserved ejection fraction; Statistical analysis was performed on each candidate metabolite in the second candidate metabolite pool to obtain significance indicators; A third candidate metabolite pool is obtained by screening candidate metabolites from the second candidate metabolite pool whose difference fold is greater than a preset threshold and whose significance index is less than a preset threshold.

5. The method according to claim 2, characterized in that, The process of performing clinical factor correction on each candidate metabolite in the third candidate metabolite pool to obtain a fourth candidate metabolite pool includes: Each candidate metabolite in the third candidate pool and a preset clinical factor are input into a multivariate regression model to obtain the interference results of the clinical factors on medical events of heart failure with preserved ejection fraction; the interference results are used to characterize the degree of interference of the clinical factors on the association between metabolites and medical events of heart failure with preserved ejection fraction. Based on the interference results, candidate metabolites that are independently associated with medical events of heart failure with preserved ejection fraction are screened from the third candidate pool to obtain a fourth candidate metabolite pool.

6. The method according to claim 2, characterized in that, The step of performing multiple feature screening on each candidate metabolite in the fourth candidate metabolite pool to obtain the at least one target metabolite includes: The optimal regularization parameter is selected by cross-validation. Using the optimal regularization parameter, metabolites whose feature coefficients are compressed to a preset value in the fourth candidate metabolite pool are removed to obtain the sixth candidate metabolite pool. Obtain the original features and shadow features of each candidate metabolite in the sixth candidate metabolite pool. The shadow features are generated by randomly shuffling the original features. The shadow features are similar in distribution to the original features but are unrelated to the original target. The original features and the shadow features are input into a random forest model to obtain the importance scores of the original features and the shadow features, respectively. Based on the importance scores of the original features and the shadow features, noisy features with importance scores lower than those of the shadow features in the original features are removed to obtain the seventh candidate metabolite pool. The at least one target metabolite is obtained from the seventh candidate metabolite pool.

7. The method according to claim 6, characterized in that, Obtaining the at least one target metabolite from the seventh candidate metabolite pool includes: Independent variable correlation verification was performed on each candidate metabolite in the seventh candidate metabolite pool to obtain the correlation coefficient between each candidate metabolite. Candidate metabolites with correlation coefficients higher than a preset threshold are removed to obtain at least one target metabolite.

8. The method according to any one of claims 1-7, characterized in that, The at least one target metabolite includes: N-methyl-trans-4-hydroxy-proline, rhamnoic acid, uridine triphosphate, 4-hydroxy-3-methoxyphenylacetic acid, hydroxyphenyl lactic acid, perilla alkaloid, N-oleoyl-L-serine, undecanoic acid (11:0), triglycerides (15:0-16:0-24:1), and triglycerides (15:0-18:1-24:1).

9. The method according to any one of claims 1-7, characterized in that, The step of weighted summing of the concentration data of each target metabolite and the corresponding effect value of each target metabolite to determine the probability of the target patient experiencing a medical event of heart failure with preserved ejection fraction includes: For each target metabolite, the probability of an initial ejection fraction-preserving heart failure medical event for each target metabolite is determined based on the product of the effect value corresponding to the target metabolite and the concentration data of the target metabolite. The probability of the target patient experiencing a preservative-ejection-fraction heart failure medical event is determined by summing the probabilities of each initial ejection fraction-preserving heart failure medical event.

10. A device for determining the probability of a medical event, characterized in that, The device includes: The acquisition module is used to acquire at least one target metabolite of a target patient hospitalized due to a medical event of preserved ejection fraction heart failure, as well as the concentration data of each target metabolite in the at least one target metabolite, wherein the influence of the at least one target metabolite on the medical event of preserved ejection fraction heart failure is greater than a preset level; The acquisition module is also used to acquire the effect value corresponding to each of the target metabolites from the correspondence between target metabolites and effect values, wherein the effect value is used to characterize the degree of influence of the target metabolite on medical events of heart failure with preserved ejection fraction; The determination module is used to perform a weighted summation of the concentration data of each target metabolite and the effect value corresponding to each target metabolite to determine the probability of the target patient experiencing a medical event of heart failure with preserved ejection fraction.

11. A device for determining the probability of a medical event, characterized in that, The device includes: a processor and a memory storing computer program instructions; the processor, when executing the computer program instructions, implements the medical event probability determination method as described in any one of claims 1-9.

12. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer program instructions that, when executed by a processor, implement the method for determining the probability of medical events as described in any one of claims 1-9.

13. A computer program product, characterized in that, When the instructions in the computer program product are executed by the processor of the electronic device, the electronic device is able to perform the medical event probability determination method as described in any one of claims 1-9.

Images