Methods, products, and applications for classifying exercise training intensity based on blood metabolites.

Abstract

This disclosure provides a method, product, and application for classifying exercise training intensity based on blood metabolites. It combines metabolomics detection results with conventional hematological indicators for athlete training type classification, breaking through the limitations of traditional methods that rely solely on single physiological or functional indicators. Through statistical analysis and machine learning stability selection, core indicators are screened, a robust discriminant panel is established, and an interpretable one-to-many logistic regression discriminant model is constructed. A graded threshold method is introduced, forming a scoring system that can be directly used for practical classification. Furthermore, an "uncertainty / retest" discriminant mechanism is added to the model output to reduce the risk of misjudgment in practical applications. This disclosure not only targets the training classification of competitive athletes but also has strong application extensibility, which can be extended to the selection of youth athletes, public health and physical fitness assessment, and chronic disease rehabilitation exercise prescription design, expanding the application boundaries of metabolomics results in the field of sports and health.

Description

Technical Field

[0001] This disclosure relates to the field of blood metabolism technology, and in particular to methods, products and applications for classifying exercise training intensity based on blood metabolites. Background Technology

[0002] In the field of sports medicine and training, different types of exercise (such as high-intensity anaerobic, mixed, and endurance exercises) have significantly different effects on the body's metabolic adaptation. Current athlete classification and selection primarily rely on traditional physiological and athletic performance indicators, such as maximum oxygen uptake (VO2 max). These methods include lactate threshold, cardiopulmonary function testing, and specialized performance evaluation. While these methods have some practical value, they still have the following shortcomings:

[0003] 1. Reliance on high-intensity exercise load: Many traditional tests require athletes to perform under conditions close to their limits, which carries significant risks and is not suitable for adolescents, those recovering from injuries, or the general population;

[0004] 2. Results are greatly affected by external factors: athletic performance and functional indicators are easily affected by training status, psychological factors, competition environment, etc., and lack stability and repeatability, making them difficult to use as a long-term stable classification basis.

[0005] 3. Lack of objective indicators at the molecular level: Existing methods are mostly based on macroscopic physiological data, which cannot reveal the molecular mechanisms behind exercise adaptation and make it difficult to achieve precise, mechanism-driven athlete classification;

[0006] 4. Limitations of existing metabolomics research: Although some studies have shown that metabolites can be used to reflect exercise status, they often remain at the stage of discovering differential metabolites, lacking a systematic approach to integrate hematological indicators with multivariate modeling, and failing to develop stable and generalizable discrimination tools.

[0007] In other words, existing methods mostly rely on single or few physiological signals such as heart rate, velocity, and lactate, which have drawbacks such as limited methods, dependence on workload, insufficient objectivity and stability of results, and lack of directly applicable tools, which are not conducive to practical application. Summary of the Invention

[0008] The purpose of this disclosure is to provide a method for classifying exercise training intensity based on blood metabolites, related products and applications, in order to solve the above-mentioned technical problems.

[0009] To achieve the above objectives, in a first aspect, this disclosure provides a marker for classifying human exercise intensity, said marker being human blood metabolites, including purine metabolites, Metabolic related substances and medium- or long-chain acylcarnitine lipids;

[0010] The purine metabolites include AMP, hypoxanthine, and inosine; Metabolite-related substances include nicotinamide and 1-methylnicotinamide.

[0011] In this disclosure, the marker may also be referred to as a composition of markers, and may be used interchangeably in this disclosure.

[0012] In a preferred embodiment, the number of the biomarkers is 10 to 15, and the biomarkers include adenosine, AMP, hypoxanthine, inosine, nicotinamide, 1-methylnicotinamide, leucine, valine, isoleucine, at least one medium-chain or long-chain acylcarnitine lipid, at least one phospholipid, and at least one ceramide.

[0013] In a preferred embodiment, the markers include PC 34:3, PC 39:6, L-Glutaminic acid, Palmitoleic acid | (Z)-14-Methylpentadec-6-enoic acid, PC 36:4, D-Tagatose | D-chiro-Inositol, PE 36:4, Nervonic acid, LysoPC 17:1, PC 35:1, PC 34:2e, FA 22:6, SMd40:2 (d16:1 / 24:1), PC 35:3 | MePC 32:0, FA 20:5, PC 32:1e, L-Tyrosine, alpha-Ketoisovaleric acid, Hypoxanthine, PC 38:5, SMd38:5 (d18:2 / 20:3), Norleucine, L-Prolylglycine | Glycyl-L-proline, 3-Amino-3-(4-hydroxyphenyl)propanoic acid, PC38:7, PC 38:6 | MePC 35:3, TAG 49:0, PC 34:4, 2,6-Diaminopimelic acid, Linolenic acid | Gamolenic acid (FA 18:3), alpha-D-Talose | D-chiro-Inositol, 2-Oxoglutaric acid | alpha-Ketoglutaric acid, Lauric acid, LysoPC 20:0, TAG 42:2, Glycoursodeoxycholic acid, Cer d41:2 (d17:1 / 24:1) | Cer d41:2 (d18:2 / 23:0), PC36:5, Pyroglutamic acid and TAG 56:1.

[0014] Secondly, this disclosure provides a method for classifying exercise training intensity based on blood metabolites, comprising the following steps:

[0015] Collect blood samples from the subjects;

[0016] The blood sample is tested, and the metabolite signal intensity of several markers is extracted from the test results, wherein the markers are those described in the first aspect embodiment above;

[0017] The metabolite signal intensity of the marker is input into a pre-constructed classification and discrimination model to obtain the posterior probability corresponding to each type of exercise training intensity.

[0018] The comparison results are obtained by comparing the posterior probabilities corresponding to various types of sports training intensity with the preset thresholds corresponding to various types of sports training intensity.

[0019] Based on the comparison results, the exercise training intensity type of the subject is output.

[0020] In a preferred embodiment, the exercise training intensity types include High Intensity Anaerobic (HIAN), Medium-High Intensity Mixed (MIX), and Moderate Endurance (MODEN), with preset thresholds θ corresponding to each. HIAN =0.46、θ MIX =0.46、θ MODEN =0.44.

[0021] In a preferred embodiment, the detection is performed using chromatography-mass spectrometry or nuclear magnetic resonance.

[0022] In a preferred embodiment, pre-constructing a classification and discrimination model includes:

[0023] Based on the metabolite signal intensity x of the marker j (j=1...N, 40≥N≥10) Establish a one-to-many logistic regression discriminant model for various sports training intensity types:

[0024]

[0025] Where, x j It is the metabolite signal intensity of the j-th biomarker. k Represents the type of exercise training intensity, β 0,k Is a type k The intercept of the corresponding model, β k,j type k The coefficient of the metabolite signal intensity of the j-th biomarker in the corresponding model, logit k Is a type k Discriminant scores of linear predicted values;

[0026] The discrimination scores for each type of exercise training intensity are input into the softmax function to obtain the posterior probability of each type of exercise training intensity.

[0027] In a preferred embodiment, the step of comparing the posterior probabilities corresponding to various types of exercise training intensity with preset thresholds corresponding to various types of exercise training intensity to obtain comparison results includes:

[0028] If there is only one type of exercise training intensity with a posterior probability greater than or equal to its corresponding preset threshold, then that type of exercise training intensity is selected as the comparison result.

[0029] If there is a posterior probability of more than one type of exercise training intensity that is greater than or equal to its corresponding preset threshold, then the exercise training intensity type with the highest posterior probability is selected as the comparison result.

[0030] If the posterior probability of all types of exercise training intensity is less than its corresponding preset threshold, the comparison result is "uncertain / retest".

[0031] In another preferred embodiment, the step of comparing the posterior probabilities corresponding to various types of exercise training intensity with preset thresholds corresponding to various types of exercise training intensity to obtain comparison results further includes:

[0032] If the posterior probability of the maximum exercise training intensity type is less than its corresponding preset threshold minus 0.05, or less than the confidence interval threshold after calibration, the comparison result is "retest". If the retest still does not meet the threshold requirement, a second sampling is performed.

[0033] Thirdly, this disclosure provides a sports training intensity classification system based on blood metabolites, including:

[0034] The data collection module is used to collect blood samples from the subjects.

[0035] A biomarker metabolomics detection module is used to detect and extract the metabolite signal intensity of the biomarkers described in the first aspect embodiment above;

[0036] The discrimination module is used to input the metabolite signal intensity of the marker into a pre-constructed classification and discrimination model to obtain the posterior probability corresponding to various types of exercise training intensity.

[0037] The comparison module is used to compare the posterior probabilities corresponding to various types of sports training intensity with the preset thresholds corresponding to various types of sports training intensity, and obtain the comparison results.

[0038] The output module is used to output the type of exercise training intensity to which the subject belongs based on the comparison results.

[0039] Fourthly, this disclosure also provides an electronic device, including: a memory and one or more processors; the memory is used to store one or more computer programs; when the one or more computer programs are executed by the one or more processors, they implement the exercise training intensity classification method based on blood metabolites as described in any embodiment of the second aspect of this disclosure.

[0040] Fifthly, this disclosure also provides a computer storage medium storing a computer program; when the computer program is executed by a processor, it implements the exercise training intensity classification method based on blood metabolites as described in any embodiment of the second aspect of this disclosure.

[0041] In a sixth aspect, this disclosure also provides a computer program product, including a computer program or instructions, which, when executed by a processor, implement the exercise training intensity classification method based on blood metabolites as described in any embodiment of the second aspect of this disclosure.

[0042] Seventhly, this disclosure also provides applications of the aforementioned markers, the aforementioned sports training intensity classification system, the aforementioned electronic devices, or the aforementioned computer storage media, including for distinguishing athletes from different training modes, for assessing the adaptive changes in the body's metabolism caused by sports training, for predicting individual sports adaptation types to provide sports intervention recommendations, for assisting in the development of training cycles and rehabilitation programs for competitive athletes, for the selection of adolescent athletes, or for providing a reference for rehabilitation exercise prescriptions for patients with chronic diseases.

[0043] Beneficial effects

[0044] Compared to existing technologies, the exercise training intensity classification method, related products, and applications based on blood metabolites disclosed in this publication enable rapid and objective identification of athletes' training types (high-intensity anaerobic, medium-high intensity mixed, and moderate endurance) without relying on extreme exercise loads. By employing multidimensional information based on metabolomics and hematology, it reveals the molecular mechanisms of exercise adaptation, improving the scientific rigor and explanatory power of the classification. Through machine learning modeling and feature selection, it can establish highly accurate and repeatable discrimination formulas and thresholds, forming an operable discrimination system with a wide range of applications. It is not only suitable for the selection and training classification of competitive athletes but can also be extended to rehabilitation medicine and public sports health management. Attached Figure Description

[0045] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of this disclosure and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0046] Figure 1 A flowchart of the exercise training intensity classification method based on blood metabolites provided in this disclosure;

[0047] Figure 2 The volcano plot of differentially expressed metabolites provided in this disclosure shows the distribution of significant differences among the three groups;

[0048] Figure 3 Heatmaps of significantly different metabolites provided in this disclosure illustrate the differences in metabolic characteristic patterns among the three athlete groups;

[0049] Figure 4 The KEGG pathway enrichment analysis results provided in this disclosure reveal that differential metabolites are mainly concentrated in amino acid metabolism and energy metabolism pathways.

[0050] Figure 5 The ROC curve and confusion matrix of the discriminant model provided in this disclosure show the classification performance and prediction accuracy of the model;

[0051] Figure 6 The importance ranking plot of the core panel features provided in this disclosure highlights the metabolites and hematological indicators that contribute the most to the model.

[0052] Figure 7 The learning curves for the relationship between panel size and classification performance provided in this disclosure show that approximately 40 core metrics represent the optimal balance between performance and practicality.

[0053] Figure 8 This is a block diagram illustrating the principle of classifying exercise training intensity based on blood metabolites provided in this disclosure;

[0054] Figure 9 A schematic block diagram of the electronic device provided in this disclosure. Detailed Implementation

[0055] The technical solutions of the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this disclosure, and not all embodiments. The components of the embodiments of this disclosure described and shown in the accompanying drawings can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this disclosure provided in the accompanying drawings is not intended to limit the scope of the claimed disclosure, but merely represents selected embodiments of this disclosure. All other embodiments obtained by those skilled in the art based on the embodiments of this disclosure without inventive effort are within the scope of protection of this disclosure.

[0056] The term "marker" or "biomarker" as used in this disclosure refers to a biomolecule, biomolecular fragment, or clinical variable whose changes and / or detection can be associated with a specific physical condition or state. Throughout this disclosure, the terms "marker" and "biomarker" are used interchangeably. These biomarkers include any suitable analyte, but are not limited to, biomolecules, including nucleotides, nucleic acids, nucleosides, amino acids, sugars, fatty acids, steroids, metabolites, peptides, polypeptides, proteins, carbohydrates, lipids, hormones, antibodies, regions of interest used as substitutes for biological macromolecules, and combinations thereof (e.g., glycoproteins, ribonucleoproteins, lipoproteins). The term also covers miRNAs and portions or fragments of miRNAs.

[0057] In this disclosure, the biomarkers are generally also referred to as groups or compositions of biomarkers, and these terms may be used interchangeably throughout the disclosure. The term "group" refers to a composition comprising one or more biomarkers, such as an array or collection. The term may also refer to an expression pattern profile or index of one or more biomarkers described herein. The number of biomarkers useful for a group of biomarkers is based on the sensitivity and specificity of a particular combination of biomarker values.

[0058] Embodiment 1 of this disclosure provides a marker for classifying human exercise intensity, wherein the marker is a human blood metabolite, and the human blood metabolite includes purine metabolites, Metabolic related substances and medium- or long-chain acylcarnitine lipids;

[0059] The purine metabolites include AMP, hypoxanthine, and inosine; Metabolite-related substances include nicotinamide and 1-methylnicotinamide.

[0060] In this disclosure, the term "purine" refers to a heterocyclic aromatic compound formed by the fusion of a pyrimidine ring and an imidazole ring, with the following chemical structural formula: The term "purine metabolite" generally refers to the final and intermediate metabolites of purines in the human body. The compounds referred to in this term include both naturally occurring or produced forms within organisms and their salts, hydrates, and crystalline forms. Furthermore, compounds with the same structure obtained by mimicking the aforementioned metabolic pathways through in vitro chemical or enzymatic methods also fall within the scope of this definition. Examples include, but are not limited to, uric acid, hypoxanthine, xanthine, inosine monophosphate (IMP), adenosine monophosphate (AMP), and guanosine monophosphate (GMP).

[0061] The term "amino acid" in this disclosure includes naturally occurring carboxylated α-amino acids or combinations thereof, such as alanine (three-letter code: ala, single-letter code: A), arginine (arg, R), asparagine (asn, N), aspartic acid (asp, D), cysteine ​​(cys, C), glutamine (gln, Q), glutamic acid (glu, E), glycine (gly, G), histidine (his, H), isoleucine (ile, I), leucine (leu, L), lysine (lys, K), methionine (met, M), phenylalanine (phe, F), proline (pro, P), serine (ser, S), threonine (thr, T), tryptophan (trp, W), tyrosine (tyr, Y), or valine (val, V).

[0062] The term "lipid molecules" in this disclosure refers to a broad class of natural or synthetic organic compounds with diverse chemical structures that are poorly soluble in water but soluble in nonpolar solvents. This class can include, for example, simple lipids (primarily fatty acids and their esters with alcohols, such as triglycerides), complex lipids (molecules containing other important groups besides fatty acids and alcohols, such as phospholipids and glycolipids), derived lipids and precursors (molecules produced by the hydrolysis or metabolism of the above lipids, or molecules with similar properties, such as steroids and their derivatives, fat-soluble vitamins, terpenoids, and derivatives of fatty acids). Furthermore, the term also encompasses all possible chemical modifications (such as hydrogenation, sulfonation, PEGylation, and fluorination), salt forms, isomers (including stereoisomers and optical isomers), complex forms (such as lipoproteins formed with proteins), and artificially designed structural analogs. In specific contexts, supramolecular structures formed by the self-assembly of lipid molecules (such as liposomes, micelles, and chylomicrons) are also within the scope of this definition.

[0063] In some embodiments, the number of the markers is 10 to 15, and the markers include adenosine, AMP, hypoxanthine, inosine, nicotinamide, 1-methylnicotinamide, leucine, valine, isoleucine, at least one medium-chain or long-chain acylcarnitine lipid, at least one phospholipid, and at least one ceramide.

[0064] In some embodiments, the markers include PC 34:3, PC 39:6, L-Glutaminic acid, Palmitoleic acid | (Z)-14-Methylpentadec-6-enoic acid, PC 36:4, D-Tagatose | D-chiro-Inositol, PE 36:4, Nervonic acid, LysoPC 17:1, PC 35:1, PC 34:2e, FA 22:6, SMd40:2 (d16:1 / 24:1), PC 35:3 | MePC 32:0, FA 20:5, PC 32:1e, L-Tyrosine, alpha-Ketoisovaleric acid, Hypoxanthine, PC 38:5, SMd38:5 (d18:2 / 20:3), Norleucine, L-Prolylglycine | Glycyl-L-proline, 3-Amino-3-(4-hydroxyphenyl)propanoic acid, PC38:7, PC 38:6 | MePC 35:3, TAG 49:0, PC 34:4, 2,6-Diaminopimelic acid, Linolenic acid | Gamolenic acid (FA 18:3), alpha-D-Talose | D-chiro-Inositol, 2-Oxoglutaric acid | alpha-Ketoglutaric acid, Lauric acid, LysoPC 20:0, TAG 42:2, Glycoursodeoxycholic acid, Cer d41:2 (d17:1 / 24:1) | Cer d41:2 (d18:2 / 23:0), PC36:5, Pyroglutamic acid, and TAG 56:1. Specifically, the number of the markers is 40.

[0065] The metabolomics indicators of the 40 biomarkers in this embodiment showed stable performance and good calibration in multi-model and independent hold-out validation, effectively distinguishing three training types: high-intensity anaerobic (HIAN), medium-to-high-intensity mixed (MIX), and moderate-to-endurance (MODEN). This demonstrates its value for practical application and training monitoring. Understandably, any subset of 10-15 biomarkers disclosed herein can also be used to construct a predictive model. Verification has shown that this subset model can still effectively distinguish different training types, and both this subset and any subset formed using several indicators from the 40 biomarkers belong to the same concept as this disclosure.

[0066] The first embodiment of this disclosure provides a marker for classifying human exercise intensity, which can not only be used to distinguish athletes with different training modes, but also to assess the adaptive changes in the body's metabolism caused by exercise training, to predict individual exercise adaptation types to provide exercise intervention suggestions, to assist in the development of training cycles and rehabilitation programs for competitive athletes, to be used for the selection of young athletes, and to provide a reference for rehabilitation exercise prescriptions for patients with chronic diseases.

[0067] Please see Figure 1 This is a flowchart of a method for classifying exercise training intensity based on blood metabolites, provided in Embodiment 2 of this disclosure. It should be noted that the method of this disclosure is not limited to the order of the following steps, and in other embodiments, the method of this disclosure may include only a portion of the following steps, or some steps may be deleted.

[0068] The exercise training intensity classification method based on blood metabolites provided in this embodiment can be applied to exercise training intensity classification systems based on blood metabolites. This method collects blood samples from athletes, combines metabolomics characteristics with conventional hematological indicators, and uses statistical and machine learning modeling to achieve rapid and objective classification of athletes' training types (high-intensity anaerobic HIAN, medium-high intensity mixed MIX, and moderate endurance MODEN), and can be further used for athlete selection, training intervention, and public sports health management.

[0069] The exercise training intensity classification method based on blood metabolites provided in this embodiment includes the following steps:

[0070] Step S10: Collect blood samples from the subject. Specifically, collect venous blood samples from the athlete, which can be done automatically using a collection device.

[0071] Step S20: Detect the blood sample and extract the metabolite signal intensity of several markers from the detection results; the markers are the markers described in any one embodiment of the present disclosure.

[0072] Specifically, after collecting venous blood samples from athletes, serum / plasma is separated to obtain metabolite detection data and routine hematological indicators. Metabolomics detection is then performed using liquid chromatography-mass spectrometry (LC-MS) or gas chromatography-mass spectrometry (GC-MS) platforms to extract metabolic characteristic peaks. Data processing is then performed to obtain metabolic indicator data of biomarkers.

[0073] In some embodiments of this disclosure, nuclear magnetic resonance (NMR) can also be used to detect the isolated metabolites. NMR has unique advantages such as being non-destructive, highly reproducible, providing structural information, and offering absolute quantitative capabilities. Commonly used NMR analysis methods include ¹H-NMR (proton NMR).

[0074] In some embodiments of this disclosure, the data processing described above may include preprocessing steps (e.g., peak identification, alignment, normalization) of metabolite signal intensity data of the biomarker, as well as multivariate statistical analysis.

[0075] In some embodiments of this disclosure, the metabolic index data of the above-mentioned biomarkers include: Compound Name, Coef perSD, OR perSD (95% CI), direction, Superclass, Class, Subclass, PubChem CID, CAS Number, and Chemical Formula.

[0076] Step S30: Input the metabolite signal intensity of the biomarker into a pre-constructed classification model to obtain the posterior probability Pk corresponding to each type of exercise training intensity. Specifically, based on the metabolite signal intensity of the biomarker, a one-to-many logistic regression model is established to output the posterior probability of each training type.

[0077] Among them, the pre-built classification and discrimination model includes:

[0078] Based on the metabolite signal intensity x of the marker j (j=1...N, 40≥N≥10) Establish a one-to-many logistic regression discriminant model for various sports training intensity types:

[0079]

[0080] Where, x j It is the metabolite signal intensity of the j-th biomarker. k Represents the type of exercise training intensity, β 0,k Is a type k The intercept of the corresponding model, β k,j type k The coefficient of the metabolite signal intensity of the j-th biomarker in the corresponding model, logit k Is a type k Discriminant scores of linear predicted values;

[0081] The discrimination scores for each type of exercise training intensity are input into the softmax function to obtain the posterior probability of each type of exercise training intensity.

[0082] Understandably, the classification and discrimination model is not limited to logistic regression, but can also be ElasticNet regression, support vector machine (SVM), random forest (RF), gradient boosting tree (XGBoost / LightGBM) or other classifiers with equivalent functions.

[0083] Step S40: Compare the posterior probabilities corresponding to various types of exercise training intensity with the preset thresholds corresponding to various types of exercise training intensity to obtain the comparison results.

[0084] Specifically, step S40 includes the following sub-steps:

[0085] If there is only one type of exercise training intensity with a posterior probability greater than or equal to its corresponding preset threshold, then that type of exercise training intensity is selected as the comparison result.

[0086] If there is a posterior probability of more than one type of exercise training intensity that is greater than or equal to its corresponding preset threshold, then the exercise training intensity type with the highest posterior probability is selected as the comparison result.

[0087] If the posterior probability of all types of exercise training intensity is less than its corresponding preset threshold, the comparison result is "uncertain / retest".

[0088] Among them, the posterior probability of only one type of exercise training intensity being greater than or equal to its corresponding preset threshold includes the following three cases:

[0089] If P HIAN ≥θ HIAN And P MIX <θ MIX P MODEN <θ MODEN The comparison result is that it meets the requirements of high-intensity anaerobic HIAN.

[0090] If P MIX ≥θ MIX And P HIAN <θ HIAN P MODEN <θ MODEN The comparison result is that it satisfies the medium-to-high intensity mixed type MIX;

[0091] If P MODEN ≥θ MODEN And P HIAN <θ HIAN P MIX <θ MIX The comparison result indicates that it meets the requirements for a moderately endurance-oriented MODEM.

[0092] Among them, there are cases where the posterior probability of one or more types of exercise training intensity is greater than or equal to its corresponding preset threshold, that is, there are two or three types where the posterior probability is greater than or equal to its corresponding preset threshold, including the following four cases:

[0093] If P HIAN ≥θ HIAN P MIX ≥θ MIX And P MODEN <θ MODEN Then it is necessary to compare P. HIAN P MIX The final comparison result is P. HIAN P MIX The type of exercise training intensity corresponding to the largest value in the table;

[0094] If P MIX ≥θ MIX P MODEN ≥θ MODEN And P HIAN <θ HIAN Then it is necessary to compare P. MIX P MODEN The final comparison result is P. MIX P MODEN The type of exercise training intensity corresponding to the largest value in the table;

[0095] If P MODEN ≥θ MODEN P HIAN ≥θ HIAN And P MIX <θ MIX Then it is necessary to compare P. MODEN P HIAN The final comparison result is P. MODEN P HIAN The type of exercise training intensity corresponding to the largest value in the table;

[0096] If P HIAN ≥θ HIAN P MIX ≥θ MIX P MODEN ≥θ MODEN Then it is necessary to compare P. HIAN P MIX P MODEN The final comparison result is P. HIAN P MIX P MODEN The type of exercise training intensity corresponds to the largest value in the table.

[0097] Among them, the posterior probability of all sports training intensity types is less than their corresponding preset threshold, i.e., P HIAN<θ HIAN P MIX <θ MIX P MODEN <θ MODEN If so, the comparison result will be "uncertain / retest".

[0098] Step S50: Output the type of exercise training intensity to which the subject belongs based on the comparison results. Specifically, set classification thresholds for different training types (e.g., anaerobic 0.46, mixed 0.46, endurance 0.44), and determine the type based on the posterior probability output.

[0099] Understandably, exercise training intensity types include High Intensity Anaerobic (HIAN), Medium-High Intensity Mixed (MIX), and Moderate Endurance (MODEN). The preset thresholds for HIAN, MIX, and MODEN are respectively θ. HIAN =0.46、θ MIX =0.46、θ MODEN =0.44, the classification threshold is based on the ROC curve analysis of the training set, using the Youden Index (...). Determining the optimal point: θ HIAN =0.46, θ MIX =0.46, θ MODEN =0.44, the threshold was tested for average consistency using 10-fold cross-validation.

[0100] This embodiment provides a novel method for athlete training type classification by integrating metabolomics and hematological indicators. The core of the method is to standardize and screen blood metabolite characteristics (such as purine metabolites, amino acids, lipid molecules, etc.) to form a set of core indicators that can significantly distinguish different training types. Using plasma metabolomics as the information carrier, statistical and machine learning modeling is used to construct 40 core biomarkers plus threshold discrimination rules to achieve rapid and objective classification of athlete training types (high-intensity anaerobic, medium-high intensity mixed, and moderate endurance). This method can be further used for athlete selection, training intervention, and public sports health management.

[0101] This disclosure overcomes the shortcomings of existing athlete training type classification methods, such as reliance on high-intensity load tests, significant susceptibility to external factors, lack of objective molecular-level indicators, and lack of generalizable and stable discrimination tools. It proposes a comprehensive classification method combining metabolomics and hematological indicators. The training type classification described in this disclosure (high-intensity anaerobic / medium-high intensity mixed / moderate endurance) is a general classification method based on exercise physiology, considering the dominant roles of the three major energy systems (phosphagen / anaerobic short-term energy supply, glycolysis / mixed energy supply, and aerobic oxidation) in different events. It also incorporates long-term specialized training content and physical fitness tests (such as short-term explosive power / 30-second power output, lactate threshold, and maximum oxygen uptake). As a standard for confirming the classification label, this classification is consistent with the classification of energy supply types in sports in international exercise physiology textbooks and academic reviews.

[0102] The exercise training intensity classification method based on blood metabolites provided in this embodiment can not only be used to prepare tools for classifying athlete training types, but also to prepare products for athlete selection and specific adaptability assessment, auxiliary assessment systems for exercise training monitoring and rehabilitation prescription formulation, and detection products for public sports health management and individualized exercise prescription recommendations.

[0103] The technical solution of this disclosure will be implemented and verified below:

[0104] (1) Subject recruitment and grouping

[0105] Sample size and criteria: 200 outstanding athletes (aged 12–18; ≥2 years of specialized training; no chronic diseases / medications affecting metabolism) were included. The 200 samples were randomly divided in a 7:3 ratio: 140 cases for training set, used for modeling and cross-validation, and 60 cases for test set, used for independent validation.

[0106] Grouping principle: The subtyping labels are first determined based on the subject's specific training program, and then by physical fitness tests. Consistency was verified using lactate threshold and 30-second explosive power, and the results were categorized into three types based on the characteristics of specialized training:

[0107] High-intensity anaerobic group: 100m and 200m sprints, weightlifting, boxing, taekwondo; Medium-high intensity mixed group: 100-400m swimming, short-distance cycling, rock climbing; Moderate endurance group: 800m, 1500m and 3000m middle-distance running, race walking, medium-distance kayaking.

[0108] Accompanying records: General information and routine hematological parameters (creatine kinase, red blood cell count, hemoglobin, etc.) for subsequent covariate / sensitivity analysis.

[0109] (2) Sample collection and preprocessing

[0110] Data collection window: fasting for 10-12 hours, and ≥24 hours since the last training session.

[0111] Processing conditions: 4–6 mL of antecubital venous blood (EDTA), centrifuged at 1,200 g × 10 min × 4℃, plasma / serum separated; single freeze-thaw cycle. save.

[0112] Quality Control (QC): Prepare equal-volume QC pooled samples, inserting one QC random injection every 5–10 samples; internal standard is used for retention time / response correction; QC median. ≤ (Low abundance ≤20%).

[0113] (3) Metabolomics platform and peak extraction

[0114] Platform: LC–MS (ESI±) is preferred; GC–MS is an equivalent alternative implementation.

[0115] Peak processing: XCMS / MZMine / etc. are used to detect, align, and remove isotopes / adducts; features with a missing rate of ≤20% are retained; logarithmic transformation → median imputation → z-score normalization.

[0116] Drift / Batch Assessment: Perform QC-LOESS for single-batch sequences; if multiple batches exist, use ComBat for sensitivity assessment.

[0117] Feature size: Approximately 300–600 metabolic features were detected (428 were retained in this study).

[0118] (4) Statistical screening and pathway / network analysis

[0119] Difference test: ANOVA or Kruskal–Wallis between groups, BH-FDR correction (q<0.05).

[0120] Results: 164 significantly differentially expressed metabolites were identified; the maximum value was calculated. Effect size (Cohen's d / Cliff's δ).

[0121] Pathway enrichment: mapping KEGG / HMDB / Reactome, hypergeometric test + FDR, revealing amino acids, energy / purines, Significant effects were observed in pathways such as CoA and lipids.

[0122] Related networks: Spearman correlation (|ρ|≥0.30, q<0.05), Louvain modular representation of BCAA, nucleotide turnover, lipid modules, etc.

[0123] (5) Stability selection and determination of core indicators

[0124] Procedure: Perform B=200 no-replacement 70% subsamplings; fit an L1-regularized multinomial logistic regression (One-vs-Rest) for each iteration.

[0125] Selection criteria: Any category coefficient ≠ 0 is denoted as "selected"; the cumulative selection frequency π is used as a stability index; the threshold π_thr = 0.6 (0.8 for high confidence).

[0126] Core panel: 40 items are selected in descending order of π to form the core discrimination panel.

[0127] Table 1 below lists 40 biomarkers (including Feature ID, metabolite name, orientation, coef_perSD, OR / SD (95% CI), pathway classification, chemical formula, PubChem / CAS, etc.):

[0128] Table 1

[0129]

[0130] (6) Discriminant model training and validation (example results)

[0131] Candidate models: L1-Logistic (OvR), SVM-RBF, Random Forest.

[0132] Validation strategy: nested cross-validation (outer layer 5-fold; inner layer parameter tuning); all data processing is performed within the training fold to prevent information leakage.

[0133] Metrics: macro-AUC, macro-F1, macro PR-AUC, accuracy, Brier score, calibration slope / intercept.

[0134] Performance example: 40-marker L1-Logistic achieves accuracy ≥90%, macro-AUC≈0.96, macro-F1≈0.73, Brier≈0.10, calibration slope≈1.0, and intercept≈0 in outer / holdout validation.

[0135] In the test set (n=60), based on previous training intensity criteria, the samples were further divided into three categories: high-intensity anaerobic (34 cases), moderate-to-high-intensity mixed (15 cases), and moderate-to-endurance (11 cases). Forty core metabolic biomarkers locked after modeling were used for classification prediction. The results showed that the model's overall accuracy in the test set was 88.3% (53 / 60), with the prediction results for the three categories as follows: high-intensity anaerobic (30 / 34), moderate-to-high-intensity mixed (13 / 15), and moderate-to-endurance (10 / 11). ROC analysis showed a multi-class macro-average AUC of 0.95, indicating that the model has high discriminative performance and generalization ability among independent subjects. These results demonstrate that the final classification results are highly consistent with the initial sample grouping, possessing operability and reproducibility. In other words, the biomarker combination and classification method proposed in this invention can stably identify subjects with different training intensity types and has good universality.

[0136] The above steps are aimed at training intensity classification for outstanding young athletes, constructing a reproducible metabolic fingerprint + discrimination panel, and providing a complete list of core indicators and implementation discrimination criteria, forming a panel of 40 core metabolic indicators. It has shown stable performance and good calibration in multi-model and independent hold-out validation, and can effectively distinguish between the three training types of HIAN / MIX / MODEN, and has the value of implementation discrimination and training monitoring.

[0137] (7) Discriminant formula and probability output

[0138] For standardized feature x j (j=1...40) Establish three types of OvR models:

[0139]

[0140] The three posterior probabilities P are obtained through softmax. k Output: {P HIAN , P MIX , P MODEN}, final category, Top-N feature contribution (by (Sorting). The parameters of the logistic regression model are shown in Table 1 above.

[0141] (8) Discrimination based on classification threshold

[0142] Threshold setting: θ HIAN =0.46、θ MIX =0.46、θ MODEN =0.44;

[0143] Judgment rules:

[0144] If P exists HIAN ≥θ HIAN It was classified as such.

[0145] If multiple categories satisfy the condition simultaneously, choose the one with the highest probability.

[0146] If none of the three categories reach the threshold, output "Uncertain / Retest".

[0147] Among them, the gray area suggestion (practice enhancement): if maximum a posteriori <max(θ k If the threshold is crossed by a confidence interval of -0.05 or after calibration, it is marked "retest / resample".

[0148] (9) Verification results and consistency

[0149] In outer-layer cross-validation and hold-out validation, the threshold mechanism significantly reduces the risk of misclassification of boundary samples, maintaining the same ≥90% classification accuracy and good calibration as in Example 1.

[0150] This rule can be directly used in laboratory reports / software module outputs to meet the deployment needs of clinical or physical fitness monitoring systems.

[0151] This embodiment clarifies the discrimination formula, threshold, and uncertainty handling, realizing a closed loop from "metabolic fingerprint → quantitative discrimination", and is operable and reproducible.

[0152] like Figure 2 As shown, Figure 2 This is a volcano plot of differentially expressed metabolites obtained during a metabolomic comparison analysis between groups with different training intensities. The figure uses... With the horizontal axis as the base, With the vertical axis as the y-axis, the significantly upregulated and downregulated metabolites are distributed on both sides of the graph, showing a clear separation trend. This indicates that there are stable and significant metabolic differences among people with different training intensities, which can be used for subsequent intensity classification modeling.

[0153] like Figure 3 As shown, Figure 3 This is a heatmap of metabolic features obtained after screening for significantly different metabolites. The color intensity in the graph represents the relative abundance level of metabolites. Different training intensity categories show a clear and distinguishable distribution of metabolic patterns, indicating that various metabolites have a consistent population difference structure among the three groups, providing a reliable biological basis for the classification model.

[0154] like Figure 4 As shown, Figure 4 This is a metabolic pathway distribution map obtained during KEGG enrichment analysis of differential metabolites. The significance and enrichment degree of each pathway are shown by the size and color changes of the points, indicating that differential metabolites are mainly enriched in energy-related pathways such as purine metabolism, amino acid metabolism, and fatty acid oxidation, suggesting that these pathways play a key regulatory role in differences in training intensity.

[0155] like Figure 5 As shown, Figure 5 The figure shows the ROC curve and confusion matrix obtained when constructing a training intensity classification model based on metabolites. The ROC curve shows that the model has high discriminative performance among the three training intensities, and the predicted distribution in the confusion matrix shows good accuracy. The slightly greater than 1 sum in the first row is due to rounding errors during display. These results indicate that the model has stable and reliable classification ability and can be used for practical training intensity discrimination.

[0156] like Figure 6 As shown, Figure 6 This is a ranking chart of the importance of core features obtained when building the classification model. Metabolites are sorted from highest to lowest according to their contribution to the model output. High-contribution features are mainly concentrated in purine metabolites. The relevant metabolites and some lipid indicators indicate that these core metabolites play a major role in training intensity discrimination and are key biomarkers of this method.

[0157] like Figure 7 As shown, Figure 7 This is a graph showing the relationship between metabolite panel size and classification performance during model optimization. The graph shows that as the number of features increases, the model performance gradually improves, reaching its optimal value (F1 score exceeding 0.91) at approximately 40 features. This indicates that approximately 40 markers achieve the best balance between performance and detection cost, making it suitable for training intensity classification in real-world applications.

[0158] It is understood that the technical solutions described in this disclosure are not limited to specific detection platforms, feature selection methods, and modeling approaches. Regarding metabolomics detection platforms, in addition to liquid chromatography-mass spectrometry (LC-MS), gas chromatography-mass spectrometry (GC-MS), nuclear magnetic resonance (NMR), or multi-omics joint detection methods can also be used. Regarding data processing and feature selection, in addition to ANOVA+FDR correction and L1-logistic regression, conventional algorithms such as PLS-DA, Elastic Net, LASSO regression, and random forest recursive feature elimination can also be used. Regarding modeling, in addition to logistic regression, support vector machines, and random forests, gradient boosting trees (XGBoost, LightGBM), deep learning neural networks, or Bayesian classifiers can also be used. Regarding hematological indicators, in addition to creatine kinase, red blood cell count, and hemoglobin, indicators such as lactate dehydrogenase, serum lactate, and hematocrit can also be introduced as supplementary or alternative indicators. Therefore, any method that classifies athletes’ training types based on metabolomics and hematological parameters should be considered an equivalent alternative to this disclosure and is within the scope of protection of this disclosure.

[0159] Please see Figure 8This is a schematic diagram of the exercise training intensity classification system based on blood metabolites provided in Embodiment 3 of this disclosure. The exercise training intensity classification system 100 based on blood metabolites provided in this disclosure includes:

[0160] The data collection module 10 is used to collect blood samples from the subject.

[0161] The blood metabolism detection product 20 described in any one embodiment of the present disclosure is used to detect the blood sample and extract the metabolite signal intensity of several markers from the detection results;

[0162] The discrimination module 30 is used to input the metabolite signal intensity of the marker into a pre-constructed classification discrimination model to obtain discrimination parameters corresponding to various types of exercise training intensity.

[0163] Comparison module 40 is used to compare the discrimination parameters corresponding to various types of sports training intensity and the preset thresholds corresponding to various types of sports training intensity to obtain comparison results;

[0164] Output module 50 is used to output the type of exercise training intensity to which the subject belongs based on the comparison results.

[0165] Understandably, the aforementioned functional modules can be stored in memory as software programs and executed by a processor. In alternative embodiments, the aforementioned functional modules can also be hardware with specific functions, such as chips programmed with specific software.

[0166] It should be noted that, in implementation, the exercise training intensity classification method based on blood metabolites can be implemented using the aforementioned exercise training intensity classification system 100 based on blood metabolites. The exercise training intensity classification system 100 based on blood metabolites can be implemented using one or more specific embodiments of the exercise training intensity classification method based on blood metabolites described in the above embodiments. That is, all embodiments of the exercise training intensity classification method based on blood metabolites provided in this disclosure are applicable to the exercise training intensity classification system 100 based on blood metabolites provided in this disclosure, and all can achieve the same or similar beneficial effects; further details are omitted here.

[0167] Please see Figure 9 This disclosure also provides an electronic device, including: a memory 210 and one or more processors 220.

[0168] Specifically, the memory 210 is used to store one or more computer programs; when the one or more computer programs are executed by one or more processors 220, the exercise training intensity classification method based on blood metabolites described in Embodiment 2 is implemented.

[0169] The memory 210 may be, but is not limited to, Random Access Memory (RAM), Read Only Memory (ROM), Programmable Read-Only Memory (PROM), Erasable Programmable Read-Only Memory (EPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), etc. The memory 210 stores programs, and the processor 220 runs these programs after receiving execution instructions to implement the exercise training intensity classification method based on blood metabolites described in Embodiment 1. It is understood that access to the memory 210 by the processor 220 and other possible components can be performed under the control of a storage controller.

[0170] The processor 220 may be an integrated circuit chip with signal processing capabilities. The processor 220 may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc., or it may be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components, capable of implementing or executing the methods and steps disclosed in Embodiment 2 of this disclosure.

[0171] This disclosure also provides a computer storage medium storing a computer program, which, when executed by a processor, implements the exercise training intensity classification method based on blood metabolites described in Embodiment 2 above.

[0172] This disclosure also provides a computer program product, including a computer program or instructions, which, when executed by a processor, implement the exercise training intensity classification method based on blood metabolites described in Embodiment 2 above.

[0173] This disclosure also provides an application of the blood metabolism testing product described in Embodiment 1 above, including distinguishing athletes with different training modes, assessing the adaptive changes in the body's metabolism caused by exercise training, predicting individual exercise adaptation types to provide exercise intervention suggestions, assisting in the development of training cycles and rehabilitation programs for competitive athletes, selecting young athletes, and providing a reference for rehabilitation exercise prescriptions for patients with chronic diseases.

[0174] This disclosure presents a method and system for classifying athlete training types based on metabolomics and hematological indicators, which has broad application prospects, including but not limited to the following scenarios:

[0175] 1. Sports training monitoring and guidance: By regularly testing athletes' hematological and metabolomics indicators, we can determine their training type and suitability, thereby providing a basis for personalized training load adjustment, specific ability development, and scientific talent selection.

[0176] 2. Competitive Status Assessment and Early Warning: Using the classification model constructed in this disclosure, athletes' competitive status can be assessed in real time or in stages, and fatigue, overtraining or sub-health status can be detected, and intervention suggestions can be made in a timely manner.

[0177] 3. Sports and Health for Adolescents and the General Public: This method can be extended to adolescent sports training and the general fitness population to help individuals identify suitable types of exercise, optimize exercise prescriptions, and reduce exercise risks.

[0178] 4. Anti-doping and objective assessment: By using biological indicators for typing, it is possible to help determine whether changes in athletic performance are due to training adaptation or exogenous factors, providing a reference for anti-doping testing.

[0179] 5. Sports nutrition and intervention research: Combining the classification results can be used for sports nutrition formulation optimization and intervention research, exploring the metabolic needs and nutritional supplementation programs corresponding to different training types.

[0180] In other words, this disclosure is applicable not only to the field of professional competitive sports, but also to public health, rehabilitation medicine and sports science research, and has significant promotional value and industrial application potential.

[0181] The innovations of this disclosure are mainly reflected in the following aspects: First, this disclosure is the first to propose combining metabolomics detection results with routine hematological indicators for athlete training type classification, breaking through the traditional method that relies solely on a single physiological or functional indicator. Second, through statistical analysis and stability selection using machine learning, this disclosure screens approximately 40 core indicators, establishing a robust discriminant panel, addressing the problem that existing metabolomics research often focuses on "discovering differential metabolites" without applicable tools. Third, this disclosure constructs an interpretable one-to-many logistic regression discriminant model and introduces a graded threshold method, forming a scoring system that can be directly used for practical classification. Simultaneously, this disclosure incorporates an "uncertainty / retest" discrimination mechanism into the model output, reducing the risk of misjudgment in practical applications. Finally, this disclosure not only targets the training classification of competitive athletes but also possesses strong application extensibility, extending to the selection of youth athletes, public health and physical fitness assessment, and chronic disease rehabilitation exercise prescription design, expanding the application boundaries of metabolomics results in the field of sports and health.

[0182] Compared with existing technologies, this disclosure has significant advantages. First, this method avoids the exercise risks associated with traditional extreme load testing, offering high safety and making it particularly suitable for adolescents and those in rehabilitation. Second, based on multidimensional index fusion and molecular-level discrimination, the classification results are more objective and stable, avoiding interference from psychological state and environmental factors on performance indicators. Third, the model's classification accuracy can reach over 90%, significantly better than traditional classification methods based on a single indicator. Fourth, the number of core indicators proposed in this disclosure is moderate, and the detection methods are mature, making it applicable to laboratory testing, sports and health centers, and research institutions, demonstrating strong feasibility. Fifth, the differential indicators are mainly concentrated in amino acid metabolism, energy metabolism, and hormone levels, revealing the molecular mechanisms of exercise adaptation and providing scientific explanations for sports science and metabolic research. Sixth, this disclosure has significant transformation potential; it can not only serve as a research result but also be further developed into practical testing products, software algorithm modules, or sports and health assessment services, possessing industrialization and commercial application value.

[0183] In other words, this disclosure avoids the risks of relying on extreme exercise load tests in traditional methods; it provides objective and stable molecular indicators, enhancing the scientific nature of training type classification; it establishes a portable and scalable discrimination model with a classification accuracy of over 90%; and it has broad application prospects in competitive sports, sports medicine, rehabilitation, and public health.

[0184] The above description is merely an embodiment of this disclosure and does not limit the patent scope of this disclosure. Any equivalent structural or procedural transformations made using the content of this disclosure and its drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this disclosure.

Claims

1. A marker for classifying human exercise intensity, characterized in that, The markers are human blood metabolites, including purine metabolites. Metabolic related substances and medium- or long-chain acylcarnitine lipids; The purine metabolites include AMP, hypoxanthine, and inosine; Metabolite-related substances include nicotinamide and 1-methylnicotinamide; The number of the biomarkers is 10 to 15, and the biomarkers include AMP, hypoxanthine, inosine, nicotinamide, 1-methylnicotinamide, leucine, valine, isoleucine, at least one medium-chain or long-chain acylcarnitine lipid, at least one phospholipid and at least one ceramide. The markers include PC 34:3, PC 39:6, L-Glutaminic acid, Palmitoleic acid | (Z)-14-Methylpentadec-6-enoic acid, PC 36:4, D-Tagatose | D-chiro-Inositol, PE 36:4, Nervonic acid, LysoPC 17:1, PC 35:1, PC 34:2e, FA 22:6, SM d40:2 (d16:1 / 24:1), PC35:3 | MePC 32:0, FA 20:5, PC 32:1e, L-Tyrosine, alpha-Ketoisovaleric acid, Hypoxanthine, PC 38:5, SM d38:5 (d18:2 / 20:3), Norleucine, and L-Prolylglycine. |Glycyl-L-proline, 3-Amino-3-(4-hydroxyphenyl)propanoic acid, PC 38:7, PC 38:6 |MePC 35:3, TAG 49:0, PC 34:4, 2,6-Diaminopimelic acid, Linolenic acid | Gamolenicacid (FA 18:3), alpha-D-Talose | D-chiro-Inositol, 2-Oxoglutaric acid | alpha-Ketoglutaric acid, Lauric acid, LysoPC 20:0, TAG 42:2, Glycoursodeoxycholic acid, Cer d41:2 (d17:1 / 24:1) | Cer d41:2 (d18:2 / 23:0), PC 36:5, Pyroglutamic acid and TAG 56:

1.

2. A method for classifying exercise training intensity based on blood metabolites, characterized in that, Includes the following steps: Collect blood samples from the subjects; The blood sample is tested, and the metabolite signal intensity of several markers is extracted from the test results, wherein the markers are those described in claim 1; The metabolite signal intensity of the marker is input into a pre-constructed classification and discrimination model to obtain the posterior probability corresponding to each type of exercise training intensity. The comparison results are obtained by comparing the posterior probabilities corresponding to various types of sports training intensity with the preset thresholds corresponding to various types of sports training intensity. Based on the comparison results, the exercise training intensity type of the subject is output.

3. The exercise training intensity classification method based on blood metabolites as described in claim 2, characterized in that, The exercise training intensity types include High Intensity Anaerobic (HIAN), Medium-High Intensity Mixed (MIX), and Moderate Endurance (MODEN). The preset thresholds corresponding to HIAN, MIX, and MODEN are respectively θ. HIAN =0.46、θ MIX =0.46、θ MODEN =0.

44.

4. The exercise training intensity classification method based on blood metabolites as described in claim 2, characterized in that, The detection is performed using chromatography-mass spectrometry or nuclear magnetic resonance.

5. The exercise training intensity classification method based on blood metabolites as described in claim 2, characterized in that, Pre-built classification and discrimination models include: Based on the metabolite signal intensity x of the marker j (j=1...N, 40≥N≥10) Establish a one-to-many logistic regression discriminant model for various sports training intensity types: ， Where, x j It is the metabolite signal intensity of the j-th biomarker. k Represents the type of exercise training intensity, β 0,k Is a type k The intercept of the corresponding model, β k,j type k The coefficient of the metabolite signal intensity of the j-th biomarker in the corresponding model, logit k It is the discriminant score of the linear predictor of type k; The discrimination scores for each type of exercise training intensity are input into the softmax function to obtain the posterior probability of each type of exercise training intensity.

6. The method for classifying exercise training intensity based on blood metabolites as described in any one of claims 2-5, characterized in that, The steps for comparing the posterior probabilities corresponding to various types of exercise training intensity and the preset thresholds corresponding to various types of exercise training intensity to obtain the comparison results include: If there is only one type of exercise training intensity with a posterior probability greater than or equal to its corresponding preset threshold, then that type of exercise training intensity is selected as the comparison result. If there is a posterior probability of more than one type of exercise training intensity that is greater than or equal to its corresponding preset threshold, then the exercise training intensity type with the highest posterior probability is selected as the comparison result. If the posterior probability of all types of exercise training intensity is less than its corresponding preset threshold, the comparison result is "uncertain / retest".

7. The exercise training intensity classification method based on blood metabolites as described in claim 6, characterized in that, The step of comparing the posterior probabilities corresponding to various types of exercise training intensity with the preset thresholds corresponding to various types of exercise training intensity to obtain the comparison results also includes: If the posterior probability of the maximum exercise training intensity type is less than its corresponding preset threshold minus 0.05, or less than the confidence interval threshold after calibration, the comparison result is "retest". If the retest still does not meet the threshold requirement, a second sampling is performed.

8. A sports training intensity classification system based on blood metabolites, characterized in that, include: The data collection module is used to collect blood samples from the subjects. A biomarker metabolomics detection module is used to detect and extract the metabolite signal intensity of the biomarker as described in claim 1; The discrimination module is used to input the metabolite signal intensity of the marker into a pre-constructed classification and discrimination model to obtain the posterior probability corresponding to various types of exercise training intensity. The comparison module is used to compare the posterior probabilities corresponding to various types of sports training intensity with the preset thresholds corresponding to various types of sports training intensity, and obtain the comparison results. The output module is used to output the type of exercise training intensity to which the subject belongs based on the comparison results.

9. An electronic device, comprising: Memory and one or more processors; The memory is used to store one or more computer programs; characterized in that, when the one or more computer programs are executed by the one or more processors, they implement the exercise training intensity classification method based on blood metabolites as described in any one of claims 2-7.

10. A computer storage medium storing a computer program; characterized in that, When the computer program is executed by the processor, it implements the exercise training intensity classification method based on blood metabolites as described in any one of claims 2-7.

Images