System for constructing identification model for cardiometabolic risk factor profile, and storage medium and kit

Abstract

The present invention relates to the field of biomedicine. Provided are a system for constructing an identification model for a cardiometabolic risk factor profile, and a storage medium and a kit. The system for constructing an identification model for a cardiometabolic risk factor profile comprises a collection module, a feature selection module, a training module and an identification module. A prediction model for a medium-risk type or a high-risk type of the cardiometabolic risk factor profile obtained in the present invention demonstrates excellent risk assessment capabilities with high sensitivity, specificity and accuracy. The present invention only requires a small amount of plasma as a test sample, features relatively low costs, is easily accepted by subjects, is suitable for large-scale application of the method, and can also be used in the screening of drugs targeting the cardiometabolic risk factor profile.

Description

System for constructing identification model of cardiovascular metabolic risk factor spectrum, storage medium and kit TECHNICAL FIELD

[0001] The present application relates to the biomedical field, in particular to a system for constructing an identification model of a cardiovascular metabolic risk factor spectrum, a storage medium and a kit. BACKGROUND

[0002] The cardiovascular metabolic risk factor spectrum includes obesity, hypertension, hyperglycemia and hyperlipidemia.

[0003] The cardiovascular metabolic risk factor spectrum is not only a key risk factor for cardiovascular disease, but also has been reported to be associated with an increased incidence of various cancers and neurodegenerative diseases. Multiple risk factors in the cardiovascular metabolic risk factor spectrum often coexist. According to the classification standard recommended by the Chinese Medical Association Diabetes Branch in 2004 (CDS2004), the cardiovascular metabolic risk factor spectrum can be divided into low risk (0 risk factors), medium risk (1-2 risk factors) and high risk (3-4 risk factors) according to the number of cardiovascular metabolic risk factors (obesity, hypertension, hyperglycemia and hyperlipidemia).

[0004] However, this traditional risk assessment method relies on the appearance of obvious clinical symptoms and involves complex procedures such as physical examination, questionnaire survey, hematology examination, etc., which is time-consuming and requires the patient's full cooperation. At present, there is still a lack of a fast and convenient cardiovascular metabolic risk factor spectrum risk assessment tool. SUMMARY

[0005] The purpose of the present application is to provide a system for constructing an identification model of a cardiovascular metabolic risk factor spectrum, a storage medium and a kit.

[0006] To solve the above problems, the present application provides a system for constructing an identification model of a cardiovascular metabolic risk factor spectrum, comprising:

[0007] A collection module is used to collect plasma samples of subjects of low, medium and high risk types of cardiovascular metabolic risk factor spectrum, and obtain plasma small molecule metabolic biomarkers and their relative concentration values of low, medium and high risk types from each plasma sample;

[0008] A feature selection module is used to perform machine learning based on the plasma small molecule metabolic biomarkers and their relative concentration values of low, medium and high risk types, to screen plasma small molecule metabolic biomarkers that have significant differences between subjects of low, medium and high risk types of cardiovascular metabolic risk factor spectrum, as plasma small molecule metabolic biomarkers after feature selection;

[0009] a training module configured to train a prediction model for the medium or high risk type of the cardiovascular metabolic risk factor spectrum based on the feature-selected plasma small molecule metabolic biomarkers and their relative concentration values of each subject of the low, medium, and high risk types;

[0010] a recognition module configured to obtain the feature-selected plasma small molecule metabolic biomarkers and their relative concentration values of the to-be-detected subject, and input the feature-selected plasma small molecule metabolic biomarkers and their relative concentration values of the to-be-detected subject into the prediction model to obtain a prediction result of whether the to-be-detected subject belongs to the medium or high risk type of the cardiovascular metabolic risk factor spectrum;

[0011] The feature-selected plasma small molecule metabolic biomarkers at least include the following metabolite biomarkers in peripheral plasma: acetoacetate, dimethylglycine, sarcosan, daucic acid, serine, homocysteine sulfate lactone, glyceric acid, nicotinic acid, homocysteine, glucose, aminoethanesulfonic acid, uric acid, succinic acid, threonine, cysteine, piperidinic acid, aminoethyl sulfinate, and erythrulose.

[0012] Further, in the identification model construction system for the cardiovascular metabolic risk factor spectrum, the collection module is configured to obtain a first subject group and a second subject group which are mutually non-overlapping, wherein the first subject group and the second subject group each include subjects of the low, medium, and high risk types of the cardiovascular metabolic risk factor spectrum which are mutually non-overlapping; the collection module is configured to collect first plasma samples of the first subject group, and obtain plasma small molecule metabolic biomarkers and their relative concentration values from each first plasma sample as a first set; and the collection module is configured to collect second plasma samples of the second subject group, and obtain plasma small molecule metabolic biomarkers and their relative concentration values from each second plasma sample as a test set.

[0013] Further, in the identification model construction system for the cardiovascular metabolic risk factor spectrum, the feature selection module is configured to use Kruskal-Wallis rank sum test to perform feature selection on the plasma metabolite biomarkers based on the first set, and set a significance threshold to obtain the feature-selected plasma small molecule metabolic biomarkers.

[0014] Further, in the identification model construction system for the cardiovascular metabolic risk factor spectrum, the training module is configured to train a candidate prediction model for the medium or high risk type of the cardiovascular metabolic risk factor spectrum based on the feature-selected plasma small molecule metabolic biomarkers and their relative concentration values of each subject of the low, medium, and high risk types in the first set; and obtain a final prediction model for the medium or high risk type of the cardiovascular metabolic risk factor spectrum based on the test set and the first set.

[0015] Further, in the above-mentioned system for constructing a recognition model of cardiovascular metabolic risk factor spectrum, the training module is configured to classify the subjects with low risk and medium risk in the first set as a third group of subjects to obtain a third group of subjects; and classify the subjects with high risk in the first set as a fourth group of subjects to obtain a fourth group of subjects; the training module is configured to train a first candidate prediction model of cardiovascular metabolic risk factor spectrum based on the selected plasma small molecule metabolic biomarkers and their relative concentration values of the third group of subjects and the fourth group of subjects; and obtain a final prediction model of high risk of cardiovascular metabolic risk factor spectrum, i.e., a first prediction model, based on the performance of the prediction results of the first candidate prediction model on the test set, the third group of subjects and the fourth group of subjects; the training module is configured to classify the subjects with low risk in the first set as a fifth group of subjects to obtain a fifth group of subjects; and classify the subjects with medium risk and high risk in the first set as a sixth group of subjects to obtain a sixth group of subjects; the training module is configured to train a second candidate prediction model of cardiovascular metabolic risk factor spectrum based on the selected plasma small molecule metabolic biomarkers and their relative concentration values of the fifth group of subjects and the sixth group of subjects; and obtain a final prediction model of medium-high risk of cardiovascular metabolic risk factor spectrum, i.e., a second prediction model, based on the performance of the prediction results of the second candidate prediction model of cardiovascular metabolic risk factor spectrum on the test set, the third group of subjects and the fourth group of subjects.

[0016] Further, in the above-mentioned system for constructing a model for identifying a cardiovascular metabolic risk factor profile, the training module is configured to divide the third group of subjects and the fourth group of subjects into five mutually exclusive data sets, and each time, one of the five mutually exclusive data sets that has not been selected is used as a first internal validation set, and the remaining four data sets are used as a first training set; wherein the first internal validation set and the first training set each include third subjects and fourth subjects that are mutually exclusive; the training module is configured to use the first internal validation set and the first training set each time and use an extreme gradient boosting algorithm to train the high-risk prediction model for the cardiovascular metabolic risk factor profile for five cycles to obtain five first candidate prediction models that meet the performance requirements; wherein the extreme gradient boosting algorithm XGBoostd hyperparameters are set as a learning rate eta = 0.3, a minimum weight of all observation values of a subset min_child_weight = 1, a maximum depth of a tree max_depth = 6, a minimum target function reduction required for further branching at a leaf node of the tree gamma = 0, a sample rate of samples subsample = 0.8 when constructing each tree, a feature sampling rate colsample_bytree = 1 when constructing each tree, a feature sampling rate colsample_bylevel = 1 when constructing each layer, a feature sampling rate colsample_bynode = 1 when constructing each leaf node, an L1 regularization weight alpha = 0, an L2 regularization weight lambda = 1, and a maximum number of iterations nrounds = 200; the training module is configured to input the test set into the five first candidate prediction models that meet the performance requirements to obtain prediction results of the first candidate prediction models on the test set, wherein the prediction results of the first candidate prediction models on the test set are determined by voting of the five first candidate prediction models that meet the performance requirements, and a minority submits to a majority principle to obtain the prediction results of the first candidate prediction models on the test set; if the performance value of the prediction results of the first candidate prediction models on the test set is higher than a preset performance threshold, and the performance of the prediction results on the test set and the performance of the prediction results on the first internal validation set differ by less than a preset difference threshold, then the five first candidate prediction models that meet the performance requirements are used as the final first prediction model.

[0017] Further, in the model construction system of the cardiovascular metabolic risk factor spectrum above, the training module is configured to divide the fifth type of subject group and the sixth type of subject group into five non-overlapping data, and each time, one of the five non-overlapping data that is not selected is taken as a second internal validation set, and each time, the remaining four data are taken as a second training set; the second internal validation set and the second training set each include fifth type of subjects and sixth type of subjects that are non-overlapping; the training module is configured to, based on the second internal validation set and the second training set each time, and using an extreme gradient boosting algorithm, cyclically train the prediction model of the cardiovascular metabolic risk factor spectrum for 5 rounds to obtain five second candidate prediction models that meet the requirements; wherein the extreme gradient boosting algorithm XGBoostd hyperparameter settings are learning rate eta = 0.3, minimum weight of all observation values of the subset min_child_weight = 1, maximum depth of the tree max_depth = 6, minimum target function reduction required for further branching at the leaf node of the tree gamma = 0, sample rate of the sample subsample = 0.8 when building each tree, feature sampling rate colsample_bytree = 1 when building each tree, feature sampling rate colsample_bylevel = 1 when building each layer, feature sampling rate colsample_bynode = 1 when building each leaf node, L1 regularization weight alpha = 0, L2 regularization weight lambda = 1, and maximum number of iterations nrounds = 200; the training module is configured to input the test set into the five second candidate prediction models that meet the requirements for prediction to obtain the prediction results of the second candidate prediction models on the test set, wherein the prediction results of the first candidate prediction models on the test set are determined by voting of the five first candidate prediction models that meet the requirements, and the minority principle is adopted to obtain the prediction results of the second candidate prediction models on the test set; if the performance value of the prediction results of the second candidate prediction models on the test set is higher than the preset performance threshold, and the performance of the prediction results on the test set and the performance of the prediction results on the second internal validation set differ by less than the preset difference threshold, then the five second candidate prediction models that meet the requirements are taken as the final second prediction model.

[0018] According to another aspect of the present application, there is also provided a computer readable storage medium having computer executable instructions stored thereon, wherein the computer executable instructions, when executed by a processor, cause the processor to perform the following steps:

[0019] The collection module collects plasma samples of each subject of low-risk, medium-risk and high-risk types of the cardiovascular metabolic risk factor spectrum, and obtains plasma small molecule metabolic biomarkers and relative concentration values of low-risk, medium-risk and high-risk types from each plasma sample; the feature selection module performs machine learning based on the plasma small molecule metabolic biomarkers and relative concentration values of low-risk, medium-risk and high-risk types to screen plasma small molecule metabolic biomarkers that have significant differences between subjects of low-risk, medium-risk and high-risk types of the cardiovascular metabolic risk factor spectrum as the selected plasma small molecule metabolic biomarkers; the training module trains a prediction model of medium-risk or high-risk types of the cardiovascular metabolic risk factor spectrum based on the selected plasma small molecule metabolic biomarkers and relative concentration values of each subject of low-risk, medium-risk and high-risk types; the identification module obtains the selected plasma small molecule metabolic biomarkers and relative concentration values of the to-be-detected subject, and inputs the selected plasma small molecule metabolic biomarkers and relative concentration values of the to-be-detected subject into the prediction model to obtain a prediction result of whether the to-be-detected subject belongs to the medium-risk or high-risk type of the cardiovascular metabolic risk factor spectrum; the selected plasma small molecule metabolic biomarkers at least include the following metabolite biomarkers in peripheral plasma: acetoacetate, dimethyl glycine, creatinine, daucic acid, serine, homocysteine sulfate lactone, glyceric acid, nicotinic acid, homocysteine, glucose, aminothiophene acid, uric acid, succinic acid, threonine, cysteine, piperidine acid, aminoethyl sulfinate and erythrone.

[0020] According to another aspect of the present application, the present application also provides a detection kit for identifying the cardiovascular metabolic risk factor spectrum, and the metabolite biomarkers in the detection kit at least include the following metabolite biomarkers in peripheral plasma: acetoacetate, dimethyl glycine, creatinine, daucic acid, serine, homocysteine sulfate lactone, glyceric acid, nicotinic acid, homocysteine, glucose, aminothiophene acid, uric acid, succinic acid, threonine, cysteine, piperidine acid, aminoethyl sulfinate and erythrone.

[0021] Compared with the prior art, the present application obtains a prediction model of medium-risk or high-risk types of the cardiovascular metabolic risk factor spectrum through machine learning of plasma metabolite biomarkers, and has excellent sensitivity, specificity and risk assessment ability. The present application has high efficient evaluation ability for the cardiovascular metabolic risk factor spectrum, and only needs a small amount of plasma as a detection sample. Compared with the traditional cumbersome evaluation process, the present application is simple and convenient, low in cost, easy to be accepted by subjects, and suitable for wide promotion of the method. In addition, the high sensitivity and specificity enable the present application to be also used in screening of drugs for the cardiovascular metabolic risk factor spectrum. BRIEF DESCRIPTION OF DRAWINGS

[0022] Figure 1 shows the performance of a first prediction model in one embodiment of the present application in identifying high risk of cardiovascular metabolic risk factor profile using a combination of 18 metabolic biomarkers on a first internal validation set and a test set;

[0023] Figure 2 shows the performance of a second prediction model in one embodiment of the present application in identifying intermediate or high risk of cardiovascular metabolic risk factor profile using a combination of 18 metabolic biomarkers on a second internal validation set and a test set;

[0024] Figure 3 shows the schematic diagram of the module structure of the system for constructing the identification model of the cardiovascular metabolic risk factor profile in one embodiment of the present application;

[0025] Figure 4 shows the schematic diagram of the flow of the method for identifying the cardiovascular metabolic risk factor profile in one embodiment of the present application. DETAILED DESCRIPTION

[0026] The present application will be further described by a non-limiting example with reference to the accompanying drawings.

[0027] In one typical configuration of the present application, the terminal, the device of the service network and the trusted party each comprises one or more processors (CPU), input / output interface, network interface and memory.

[0028] The memory can include non-persistent memory in computer readable media, random access memory (RAM), and / or non-volatile memory, such as read only memory (ROM) or flash memory. The memory is an example of computer readable media.

[0029] Computer readable media includes permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology for storing information. The information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read only memory (ROM), electrically erasable programmable read only memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette, magnetic tape disk storage or other magnetic storage device, or any other non-transmission medium that can be used to store information accessible by a computing device. According to the definition herein, computer readable media does not include non-transitory computer readable media, such as modulated data signals and carriers.

[0030] In order to make the purposes, technical solutions and advantages of the embodiments of the present application clearer, the various embodiments of the present application will be described in detail below with reference to the embodiments. The experimental methods not specified in the embodiments are usually performed according to the conventional conditions, such as the conditions described in the textbooks and experimental guidelines, or the conditions recommended by the manufacturers. The recommended conditions of the matching software are run.

[0031] As shown in FIG. 3, the present application provides a recognition model construction system of cardiovascular metabolic risk factor spectrum, which comprises the following acquisition module 10, feature selection module 20, training module 30 and recognition module 40 described below. In an embodiment, the acquisition module is used to collect the plasma samples of each subject of low-risk, medium-risk and high-risk types of cardiovascular metabolic risk factor spectrum, and obtain the plasma small molecule metabolic biomarkers of low-risk, medium-risk and high-risk types and their relative concentration values from each plasma sample.

[0032] In an embodiment, as shown in FIG. 4, the present application provides a recognition method of cardiovascular metabolic risk factor spectrum, comprising: step S1, the acquisition module collects the plasma samples of each subject of low-risk, medium-risk and high-risk types of cardiovascular metabolic risk factor spectrum, and obtains the plasma small molecule metabolic biomarkers of low-risk, medium-risk and high-risk types and their relative concentration values from each plasma sample;

[0033] Step S2, the feature selection module performs machine learning based on the plasma small molecule metabolic biomarkers of low-risk, medium-risk and high-risk types and their relative concentration values, to screen the plasma small molecule metabolic biomarkers that have significant differences between the subjects of low-risk, medium-risk and high-risk types of cardiovascular metabolic risk factor spectrum, as the plasma small molecule metabolic biomarkers after feature selection;

[0034] Step S3, the training module trains the prediction model of medium-risk or high-risk type of cardiovascular metabolic risk factor spectrum based on the plasma small molecule metabolic biomarkers after feature selection and their relative concentration values of each subject of low-risk, medium-risk and high-risk types;

[0035] Step S4, the recognition module obtains the plasma small molecule metabolic biomarkers after feature selection and their relative concentration values of the to-be-detected subject, inputs the plasma small molecule metabolic biomarkers after feature selection and their relative concentration values of the to-be-detected subject into the prediction model, to obtain the prediction result of whether the to-be-detected subject belongs to the medium-risk or high-risk type of cardiovascular metabolic risk factor spectrum.

[0036] Metabolomics is a new branch of biology that emerged in the early 21st century after genomics, transcriptomics and proteomics. It aims to comprehensively analyze metabolic products, small molecule substrates, intermediates and chemical processes of cellular metabolites, and plays an important role in the framework of systems biology. Plasma is an ideal biological fluid for finding biomarkers in metabolomics. At present, plasma metabolomics has shown broad application prospects in many research fields including disease diagnosis, drug development, drug efficacy and toxicity evaluation. Patients with risk and high risk in cardiovascular metabolic risk factor spectrum show different metabolic models from low risk population due to metabolic disorders. Therefore, plasma metabolomics can be a useful resource for risk assessment of cardiovascular metabolic risk factor spectrum, which will strongly promote disease prevention and control.

[0037] The relative concentration value refers to the relative concentration value of a certain type of plasma small molecule metabolic biomarker in plasma, which can be in %.

[0038] Preferably, the collection module can be used to obtain a first subject group and a second subject group that do not overlap with each other, wherein the first subject group and the second subject group each include respective subjects of low risk, medium risk and high risk types of vascular metabolic risk factor spectrum that do not overlap with each other.

[0039] Here, the first subject group includes respective subjects of low risk, medium risk and high risk types of vascular metabolic risk factor spectrum; the second subject group includes respective subjects of low risk, medium risk and high risk types of vascular metabolic risk factor spectrum; the first subject group and the second subject group do not overlap with each other.

[0040] Preferably, the collection module can be used to collect first plasma samples of the first subject group, and obtain plasma small molecule metabolic biomarkers and their relative concentration values from each first plasma sample as a first set.

[0041] Preferably, the collection module can be used to collect second plasma samples of the second subject group, and obtain plasma small molecule metabolic biomarkers and their relative concentration values from each second plasma sample as a test set.

[0042] Specifically, the population of subjects can be recruited from January 2020 to December 2020 in three centers approved by the ethics committee of a certain hospital.

[0043] The first set of subjects for training and validation of the prediction model can be 1244 subjects with low risk of cardiovascular metabolic risk factor profile from center 1 (first community), 1336 subjects with medium risk of cardiovascular metabolic risk factor profile from center 2 (second community), and 1144 subjects with high risk of cardiovascular metabolic risk factor profile; the training set can be further split into a training set and an internal validation set which do not overlap with each other. Among them, the 1244 subjects with low risk of cardiovascular metabolic risk factor profile exclude subjects with medium and high risk to be identified by the present application, as a control without cardiovascular metabolic risk factor profile.

[0044] The test set for the prediction model can be 584 subjects with low risk of cardiovascular metabolic risk factor profile, 623 subjects with medium risk of cardiovascular metabolic risk factor profile, and 531 subjects with high risk of cardiovascular metabolic risk factor profile from center 3 (third community).

[0045] Plasma can be collected from subjects classified as low risk, medium risk, and high risk of cardiovascular metabolic risk factor profile according to the criteria of the Chinese Medical Association Diabetes Branch (CDS2004) in 2004, and metabolite biomarkers are detected in a nanoparticle-enhanced laser desorption / ionization mass spectrometer.

[0046] The inclusion criteria for subjects can be formulated according to the criteria of the Chinese Medical Association Diabetes Branch (CDS2004) in 2004, and subjects meeting three or more of the following four items are classified as high risk of cardiovascular metabolic risk factor profile; subjects meeting one or two of the following items are classified as medium risk of cardiovascular metabolic risk factor profile; and subjects without any of the following items are classified as low risk of cardiovascular metabolic risk factor profile.

[0047] ① Hyperglycemia: fasting blood glucose ≥ 6.1 mmol / L (110 mg / dl) and / or two-hour postprandial blood glucose ≥ 7.8 mmol / L (140 mg / dl), and / or confirmed as diabetic and treated;

[0048] ② Hypertension: systolic / diastolic blood pressure ≥ 140 / 90 mmHg, and / or confirmed as hypertensive and treated;

[0049] ③ Dyslipidemia: fasting serum triglycerides ≥ 1.7 mmol / L (150 mg / dl), and / or fasting blood high-density lipoprotein < 0.9 mmol / L (35 mg / dl) (male) or < 1.0 mmol / L (39 mg / dl) (female).

[0050] 1.0 mmol / L (39 mg / dl) (female).

[0051] (4) Overweight and / or obesity: BMI≥25.0 kg / m2;

[0052] Preferably, the exclusion criteria within the enrollment criteria are subjects with acute and infectious clinical symptoms within three weeks prior to sampling, including but not limited to fever, headache, cough, sore throat, loss of smell, runny nose, abdominal pain and diarrhea.

[0053] The plasma samples of the subject group are collected, and the plasma small molecule metabolic biomarkers and their relative concentration values are obtained from each plasma sample by the steps S121 to S128 described below.

[0054] In step S121, the collection module is used to collect the plasma sample of the peripheral venous blood of the subject in the morning on an empty stomach, which is treated with EDTA anticoagulation and stored at -80°C.

[0055] Step S122, preparation of instruments and reagents: prepare the nanoparticle-enhanced laser desorption / ionization time-of-flight mass spectrometer (MALDI-TOF-MS) and related reagents.

[0056] Here, the nanoparticle-enhanced laser desorption / ionization time-of-flight mass spectrometer can be, for example, the product of Bruker Company in Germany.

[0057] In step S123, the collection module is used to dilute the plasma sample: the plasma sample is diluted by a standard proportion with deionized water to obtain a diluted plasma sample, ensuring that the metabolite concentration is suitable for mass spectrometry analysis. The plasma sample is diluted.

[0058] Specifically, 100 nL of the plasma sample can be diluted 10 times with deionized water to obtain the diluted plasma sample.

[0059] In step S124, the collection module is used to prepare inorganic nanoparticles into a nanoparticle matrix solution for enhancing mass spectrometry signal.

[0060] Specifically, the matrix (inorganic nanoparticles) is prepared into a 1 mg / mL matrix solution with deionized water.

[0061] In step S125, the collection module is used to prepare samples on the mass spectrometry target plate: the diluted plasma sample is spotted onto the mass spectrometry target plate and dried at room temperature.

[0062] Specifically, sample preparation can be performed on the mass spectrometry target plate, with 500 nL of each diluted plasma sample being spotted and dried at room temperature.

[0063] In step S126, the collection module is used to prepare the matrix on the mass spectrometry target plate: the nanoparticle matrix solution is spotted onto the mass spectrometry target plate to ensure that the nanoparticle matrix solution uniformly covers the plasma sample spot to obtain a sample.

[0064] Specifically, matrix preparation can be performed on a mass spectrometry target plate, 500 nL of each matrix solution is spotted on each plasma sample, and dried at room temperature.

[0065] In step S127, the acquisition module is configured to perform data acquisition on the mixture in the nanoparticle-enhanced laser desorption / ionization time-of-flight mass spectrometer: using MALDI-TOF-MS to acquire mass spectrum data of metabolites in the plasma, and obtaining a metabolite peak map in the sample by delayed extraction and time-of-flight analysis.

[0066] Specifically, data acquisition is performed in the nanoparticle-enhanced laser desorption / ionization time-of-flight mass spectrometer, the data is acquired in positive ion mode extraction, the delayed extraction is adopted, the repetition rate is 1000 hz, the acceleration voltage is 20 kV, the delay time is 250 ns, and the number of laser emission times for each analysis is 2000 times.

[0067] In step S128, the acquisition module is configured to run an embedded data preprocessing pipeline based on the metabolite peak map in the sample under the recommended conditions of the software of the nanoparticle-enhanced laser desorption / ionization time-of-flight mass spectrometer, including: data resampling, spectrum smoothing, baseline correction, and spectrum peak matching, so as to obtain relative concentration values of 303 metabolite biomarkers in the full spectrum.

[0068] Specifically, the Savitzky-Golay (S-G) filter can be used for spectrum smoothing; then an adaptive iterative algorithm is used to identify baseline points, a polynomial is fitted to the identified baseline points in the metabolite peak map, the fitted baseline is subtracted from the original spectrum, and the corrected spectrum is checked to ensure that the baseline is flat and has no negative value; then the peak detection parameters are set: signal-to-noise ratio threshold = 3, mass accuracy tolerance 50 ppm, continuous wavelet transform (CWT) is used to identify possible metabolite marker peak positions, and further peak type fitting is performed to obtain corresponding peak positions, peak heights and peak areas, and metabolite marker levels are relatively quantified according to peak height size; finally, Min-Max normalization is performed to obtain the relative concentration values of 303 markers.

[0069] Here, the relative concentration values obtained can be peak intensity values after min-max normalization, with units of %, which are used to measure the relative concentration values of metabolites in the plasma, and normalization ensures that all metabolites have the same dimension.

[0070] The 303 plasma small molecule metabolite biomarkers can be identified by comparing the HMDB (Human Metabolome Database) database.

[0071] In one embodiment, the feature selection module is used to perform machine learning based on the low risk, medium risk, high risk type of plasma small molecule metabolic biomarkers and their relative concentration values, to screen out the plasma small molecule metabolic biomarkers that have significant differences between the subjects of low risk, medium risk, high risk type of cardiovascular metabolic risk factor spectrum as the plasma small molecule metabolic biomarkers after feature selection.

[0072] Preferably, the feature selection module is used to perform feature selection on the plasma metabolite biomarkers on the first set using the Kruskal-Wallis rank sum test, and set a significance threshold to obtain the plasma small molecule metabolic biomarkers after feature selection.

[0073] More preferably, the significance threshold setting can be Benjamini-Hochberg corrected P value <0.05.

[0074] Specifically, the Kruskal-Wallis rank sum test can be used on the R (4.3.1) software to perform feature selection on the 303 plasma small molecule metabolic biomarkers in the training set, and the Benjamini-Hochberg method is used for multiple comparisons to control the occurrence rate of the first type of error (false positive). Finally, 18 plasma small molecule metabolic biomarkers were screened out that had significant differences between subjects of low risk, medium risk, high risk type of cardiovascular metabolic risk factor spectrum (P <0.05).

[0075] R software is a language and environment for statistical computing and graphics. Cross-platform support for Windows Mac Linux, etc. It is a GNU project, similar to the S language and environment developed by John Chambers and his colleagues at Bell Labs (formerly AT&T, now Lucent Technologies). R can be considered a different implementation of S. There are some important differences, but much of the code written for S remains unchanged when run under R. R provides a wide range of statistical (linear and nonlinear modeling, classical statistical tests, time series analysis, classification, clustering, etc.) and graphical techniques, and is highly extensible. The S language is usually the preferred tool for statistical method research, and R language provides an open source approach to participating in this activity.

[0076] Preferably, the plasma small molecule metabolic biomarkers after feature selection include at least the following metabolite biomarkers in peripheral plasma:

[0077] Acetoacetic acid, Dimethylglycine, Creatinine, Malonic acid, L-Serine, Homocysteine thiolactone, Glyceric acid, Nicotinic acid, Homocysteine, D-Glucose, Taurine, Uric acid, Succinic acid, L-Threonine, L-Cysteine, Pipecolic acid, Hypotaurine and Erythronic acid.

[0078] wherein, the normalized relative concentration value range of Acetoacetic acid is 17.48±13.09 for low risk type, 16.33±12.99 for medium risk type and 15.12±11.76 for high risk type, the unit of relative concentration value is %;

[0079] Dimethylglycine, the normalized relative concentration value range of Dimethylglycine is 25.58±15.20 for low risk type, 24.20±15.22 for medium risk type and 23.24±14.47 for high risk type;

[0080] Creatinine, the normalized relative concentration value range of Creatinine is 8.46±5.14 for low risk type, 8.15±5.64 for medium risk type and 8.18±5 for high risk type;

[0081] Malonic acid, the normalized relative concentration value range of Malonic acid is 15.16±10.78 for low risk type, 14.76±11.07 for medium risk type and 14.24±10.32 for high risk type;

[0082] L-Serine, the normalized relative concentration value range of L-Serine is 16.42±12.66 for low risk type, 15.35±12.55 for medium risk type and 14.32±11.20 for high risk type;

[0083] Homocysteine thiosulfate, low risk type range of normalized relative concentration values of 20.92 ± 14.00, medium risk type range of normalized relative concentration values of 19.45 ± 13.42, high risk type range of normalized relative concentration values of 18.81 ± 12.74;

[0084] Glyceric acid, low risk type range of normalized relative concentration values of 16.15 ± 12.77, medium risk type range of normalized relative concentration values of 15.38 ± 12.54, high risk type range of normalized relative concentration values of 14.82 ± 12.01;

[0085] Nicotinic acid, low risk type range of normalized relative concentration values of 19.57 ± 12.96, medium risk type range of normalized relative concentration values of 18.60 ± 13.05, high risk type range of normalized relative concentration values of 18.13 ± 12.68;

[0086] Homocysteine, low risk type range of normalized relative concentration values of 7.80 ± 5.76, medium risk type range of normalized relative concentration values of 7.51 ± 6.08, high risk type range of normalized relative concentration values of 7.21 ± 6.08;

[0087] Glucose, low risk type range of normalized relative concentration values of 15.26 ± 9.09, medium risk type range of normalized relative concentration values of 17.22 ± 11.46, high risk type range of normalized relative concentration values of 21.47 ± 14.01;

[0088] Aminoethanesulfonic acid, low risk type range of normalized relative concentration values of 19.69 ± 13.79, medium risk type range of normalized relative concentration values of 18.75 ± 14.04, high risk type range of normalized relative concentration values of 17.94 ± 13.27;

[0089] Uric acid, low risk type range of normalized relative concentration values of 14.14 ± 10.69, medium risk type range of normalized relative concentration values of 13.59 ± 10.73, high risk type range of normalized relative concentration values of 13.92 ± 11.16;

[0090] Succinic acid, low risk type range of normalized relative concentration values of 16.09 ± 11.56, medium risk type range of normalized relative concentration values of 15.17 ± 11.56, high risk type range of normalized relative concentration values of 14.62 ± 11.14;

[0091] Threonine, the normalized relative concentration value range of low-risk type is 18.82±10.81, the normalized relative concentration value range of medium-risk type is 17.75±10.20, and the normalized relative concentration value range of high-risk type is 17.82±10.23;

[0092] Cysteine, the normalized relative concentration value range of low-risk type is 16.42±12.66, the normalized relative concentration value range of medium-risk type is 15.35±12.55, and the normalized relative concentration value range of high-risk type is 14.32±11.20;

[0093] Piperidinic acid, the normalized relative concentration value range of low-risk type is 17.34±12.43, the normalized relative concentration value range of medium-risk type is 16.85±12.78, and the normalized relative concentration value range of high-risk type is 16.84±12.87;

[0094] Aminoethylsulfinic acid, the normalized relative concentration value range of low-risk type is 17.04±13.30, the normalized relative concentration value range of medium-risk type is 15.77±13.28, and the normalized relative concentration value range of high-risk type is 14.53±11.85;

[0095] Erythrone, the normalized relative concentration value range of low-risk type is 18.42±12.16, the normalized relative concentration value range of medium-risk type is 18.10±12.60, and the normalized relative concentration value range of high-risk type is 17.73±12.42.

[0096] Specific information is shown in Tables 1 and 2.

[0097] Table 1: Information related to 18 metabolic biomarkers

[0098] Table 2: Concentration mean range of 18 metabolic biomarkers

[0099] Here, the range of normalized relative concentration values (relative peak intensity) of low-risk, medium-risk, and high-risk types is listed in Table 2, in %, expressed as mean plus or minus standard deviation. Based on the range of normalized relative concentration values (relative peak intensity) of the low-risk type reference group as a normal reference, the cardiovascular metabolic risk factor spectrum identification model construction system of the present application can identify medium-high risk types or high-risk types based on model training.

[0100] In the mass spectrum, the horizontal coordinate represents the mass-to-charge ratio (m / z) value of the ion, and the value of the mass-to-charge ratio increases from left to right. For ions with a single charge, the value represented by the horizontal coordinate is the mass of the ion. The vertical coordinate represents the intensity of the ion current, which is usually expressed in relative intensity, i.e., the strongest ion current intensity is set to 100%, and the intensities of other ion currents are expressed as percentages. Sometimes, the total ion current intensity of all recorded ions is taken as 100%, and various ions are expressed as percentages.

[0101] In the present application, the average ion current intensity ranges corresponding to low-risk, medium-risk, and high-risk types of each marker can be obtained, and then the average ion current intensities of low-risk, medium-risk, and high-risk types are normalized by min-max normalization to obtain the relative concentration value ranges corresponding to low-risk, medium-risk, and high-risk types, which are used as input values for model training in machine learning.

[0102] In this application, 18 plasma metabolic biomarkers are used to train a prediction model for medium-risk or high-risk types of cardiovascular metabolic risk factor spectrum, which can exhibit excellent risk assessment ability. Machine learning of the feature-selected plasma small molecule metabolic biomarkers can achieve efficient assessment of the risk of cardiovascular metabolic risk factor spectrum.

[0103] In one embodiment, the training module is configured to train a prediction model for medium-risk or high-risk types of cardiovascular metabolic risk factor spectrum based on the feature-selected plasma small molecule metabolic biomarkers and their relative concentration values of each subject in the low-risk, medium-risk, and high-risk types.

[0104] Preferably, the training module is configured to train a candidate prediction model for medium-risk or high-risk types of cardiovascular metabolic risk factor spectrum based on the feature-selected plasma small molecule metabolic biomarkers and their relative concentration values of each subject in the low-risk, medium-risk, and high-risk types in the first set; and select one of the candidate prediction models as the final prediction model for medium-risk or high-risk types of cardiovascular metabolic risk factor spectrum based on the test set.

[0105] More preferably, the training module is configured to classify the low-risk and medium-risk subjects in the first set into a third group of subjects to obtain a third group of subjects, and classify the high-risk subjects in the first set into a fourth group of subjects to obtain a fourth group of subjects.

[0106] More preferably, the training module is configured to train a first candidate prediction model of the cardiovascular metabolic risk factor profile based on the feature-selected plasma small molecule metabolite biomarkers and their relative concentration values of the third group of subjects and the fourth group of subjects; and obtain a final prediction model of high risk of the cardiovascular metabolic risk factor profile, i.e., the first prediction model, based on the performance of the prediction results of the first candidate prediction model on the first internal validation set of the test set.

[0107] Here, the first candidate prediction is a high-risk candidate prediction model.

[0108] In the first prediction model, the low-risk and medium-risk subjects in the first set are classified into one label, i.e., the third group of subjects; the high-risk subjects in the first set are classified into another label, i.e., the fourth group of subjects, and then training based on the two labels can predict patients with high risk of the cardiovascular metabolic risk factor profile.

[0109] More preferably, the training module is configured to classify the low-risk subjects in the first set into a fifth group of subjects to obtain the fifth group of subjects; and classify the medium-risk and high-risk subjects in the first set into a sixth group of subjects to obtain the sixth group of subjects.

[0110] More preferably, the training module is configured to train a second candidate prediction model of the cardiovascular metabolic risk factor profile based on the feature-selected plasma small molecule metabolite biomarkers and their relative concentration values of the fifth group of subjects and the sixth group of subjects; and obtain a final prediction model of medium-high risk of the cardiovascular metabolic risk factor profile, i.e., the second prediction model, based on the performance of the prediction results of the second candidate prediction model on the second internal validation set of the test set.

[0111] Here, unlike the first prediction model, in the second prediction model, the low-risk subjects in the first set are classified into one label, i.e., the fifth group of subjects; the medium-risk and high-risk subjects in the first set are classified into another label, i.e., the sixth group of subjects, and then training based on the two labels can predict patients with medium-high risk of the cardiovascular metabolic risk factor profile;

[0112] The performance of the first prediction model of machine learning in evaluating whether the subjects belong to the high risk of the cardiovascular metabolic risk factor profile is evaluated on the first internal validation set and the test set of the third group of subjects and the fourth group of subjects, respectively. The performances of AUC, sensitivity, specificity, and accuracy are selected as the model evaluation indexes. Among them, the model diagnosis performance curve is coincided in the first internal validation set and the test set, as shown in FIG. 1. The specific test data is shown in Table 3.

[0113] Table 3. Performance of machine learning model one on the internal validation set and test set of metabolic biomarkers

[0114] The model prediction results are represented using a 2x2 confusion matrix, where the rows represent the actual labels and the columns represent the predicted labels. The confusion matrix divides the samples into four categories:

[0115] TP (True Positive): The number of samples that are actually positive and are predicted to be positive.

[0116] TN (True Negative): The number of samples that are actually negative and are predicted to be negative.

[0117] FP (False Positive): The number of samples that are actually negative but are predicted to be positive.

[0118] FN (False Negative): The number of samples that are actually positive but are predicted to be negative.

[0119] Accuracy: The proportion of samples that are predicted correctly out of the total number of samples, calculated as (TP + FP) / (FN + TN + TP + TN).

[0120] Sensitivity: The proportion of samples that are actually positive and are predicted to be positive, calculated as TP / (TP + FN).

[0121] Specificity: The proportion of samples that are actually negative and are predicted to be negative, calculated as TN / (TN + FP).

[0122] AUC: The area under the ROC curve, which is a plot of the true positive rate versus the false positive rate for different thresholds. The AUC is not affected by the proportion of positive and negative samples and reflects the overall performance of the model across different thresholds,

[0123] ranging from 0 to 1, with a larger value indicating better overall performance of the model. The AUC can be calculated using the following formula:

[0124] where TPR = TP / (TP + FN) and FPR = FP / (TN + FP).

[0125] The performance of machine learning model two in assessing whether a subject belongs to the high-risk or very high-risk type of the cardiovascular metabolic risk factor spectrum was evaluated on the second internal validation set and test set of the fifth and sixth subject groups, respectively. The AUC, sensitivity, specificity, and accuracy were selected as the model evaluation indicators. The model diagnosis performance curves of the second internal validation set and test set coincide, as shown in FIG. 2. The specific test data is shown in Table 4.

[0126] Table 4. Performance of the second prediction model of machine learning obtained from the internal validation set and the test set of metabolic biomarkers

[0127] In one embodiment, the training module is configured to divide the third group of subjects and the fourth group of subjects into five non-overlapping data sets, each time taking one of the five non-overlapping data sets that has not been selected as the first internal validation set, and each time taking the remaining four data sets as the first training set; wherein the first internal validation set and the first training set each include non-overlapping third subjects and fourth subjects.

[0128] Specifically, the first internal validation set includes third subjects and fourth subjects; the first training set includes third subjects and fourth subjects; and the first internal validation set and the first training set are non-overlapping.

[0129] In one embodiment, the training module is configured to train the prediction model of high risk of cardiovascular metabolic risk factors based on the first internal validation set and the first training set each time, and using the extreme gradient boosting (XGBoost, eXtreme Gradient Boosting) algorithm, for 5 rounds, to obtain five first candidate prediction models that meet the performance requirements.

[0130] Here, meeting the performance requirements can mean that the performance meets the requirements, i.e., the performance value is relatively high, such as greater than a preset performance threshold. The first candidate prediction model is a candidate prediction model of high risk.

[0131] In one embodiment, the training module is configured to input the test set into the five first candidate prediction models that meet the performance requirements for prediction, to obtain the prediction results of the first candidate prediction models on the test set, wherein the prediction results of the first candidate prediction models on the test set are determined by voting of the five first candidate prediction models that meet the performance requirements, and the minority submits to the majority principle to obtain the prediction results of the first candidate prediction models on the test set; if the performance value of the prediction results of the first candidate prediction models on the test set is higher than the preset performance threshold, and the performance of the prediction results on the test set and the performance of the prediction results on the first internal validation set differ by less than a preset difference threshold, then the five first candidate prediction models that meet the performance requirements are taken as the final first prediction model.

[0132] In one specific embodiment, the training module is configured to divide the fifth group of subjects and the sixth group of subjects into five non-overlapping data sets, and each time, one of the five non-overlapping data sets that has not been selected is used as the second internal validation set, and the remaining four data sets are used as the second training set; the second internal validation set and the second training set each include non-overlapping fifth group of subjects and sixth group of subjects.

[0133] Specifically, the second internal validation set includes the fifth group of subjects and the sixth group of subjects; the second training set includes the fifth group of subjects and the sixth group of subjects; the second internal validation set and the second training set are non-overlapping.

[0134] In one specific embodiment, the training module is configured to use the extreme gradient boosting algorithm to train the prediction model for the medium-high risk of the cardiovascular and metabolic risk factor spectrum for 5 rounds based on the second internal validation set and the second training set each time, to obtain five second candidate prediction models that meet the performance requirements.

[0135] Here, the second candidate prediction is the candidate prediction model for the medium-high risk.

[0136] In one specific embodiment, the training module is configured to input the test set into the five second candidate prediction models that meet the performance requirements to obtain the prediction results of the second candidate prediction models on the test set, wherein the prediction results of the second candidate prediction models on the test set are determined by voting of the five second candidate prediction models that meet the performance requirements, and the minority principle is used to obtain the prediction results of the second candidate prediction models on the test set; if the performance value of the prediction results of the second candidate prediction models on the test set is higher than the preset performance threshold value, and the performance of the prediction results on the test set and the performance of the prediction results on the second internal validation set differ by less than the preset difference threshold value, the five second candidate prediction models that meet the performance requirements are used as the final second prediction model.

[0137] Here, if the performance meets the requirements, the value is relatively high, and the performance of the prediction results of the first prediction model or the second prediction model on the test set is relatively close to the performance on the corresponding internal validation set, without the risk of overfitting, indicating that the model can be generalized, and the trained model meets the requirements.

[0138] The first set includes 3724 subjects, including 1244 subjects with a low risk of cardiovascular and metabolic risk factor spectrum, 1336 subjects with a medium risk of cardiovascular and metabolic risk factor spectrum, and 1144 subjects with a high risk of cardiovascular and metabolic risk factor spectrum.

[0139] The test set is 1738 subjects, including: 584 subjects with low risk of cardiovascular and metabolic risk factor spectrum, 623 subjects with medium risk of cardiovascular and metabolic risk factor spectrum, and 531 subjects with high risk of cardiovascular and metabolic risk factor spectrum.

[0140] The third and fourth groups of subjects or the fifth and sixth groups of subjects are divided into 5 parts, and each time one different part is taken as an internal validation set, and the remaining 4 parts are taken as a training set, and 5 rounds of training are performed to obtain 5 first or second candidate prediction models meeting the requirements;

[0141] The model training on the first set can be performed using the extreme gradient boosting (XGBoost) algorithm on the R (4.3.1) software, and the performance of the prediction results of the model on the internal validation set is obtained through five-fold cross-validation;

[0142] Preferably, the hyperparameters of the extreme gradient boosting algorithm XGBoost are set as learning rate eta = 0.3, minimum weight of all observation values of the subset min_child_weight = 1, maximum depth of the tree max_depth = 6, minimum target function reduction required for further branching at the leaf node of the tree gamma = 0, sample rate of the sample subsample = 0.8 when building each tree, feature sampling rate of the sample colsample_bytree = 1 when building each tree, feature sampling rate of the sample colsample_bylevel = 1 when building each layer, feature sampling rate of the sample colsample_bynode = 1 when building each leaf node, L1 regularization weight alpha = 0, L2 regularization weight lambda = 1, and maximum number of iterations nrounds = 200; thereby obtaining 5 candidate prediction models meeting the requirements.

[0143] More preferably, the first set is divided into 5 parts: 4 parts as a training set and 1 part as an internal validation set; each time a different 1 part of the internal validation set is taken, and the remaining 4 parts are taken as an internal training set; 5 rounds of training are performed to obtain 5 candidate prediction models meeting the requirements.

[0144] The hyperparameters of XGBoost are set, and the specific information is shown in Table 5. The hyperparameters of the first prediction model and the second prediction model are the same, and the parameters are different.

[0145] Table 5: Hyperparameters of XGBoost internal hyperparameters

[0146] 5 rounds of cycle training, the maximum number of iterations in each round is nrounds = 200, and each round of training is verified by the corresponding 1 internal validation set;

[0147] After 5 rounds of cycle training, if the performances of the 5 models obtained are all good, the 5 models are taken as 5 candidate prediction models whose performances meet the requirements; if some of the performances of the five models are good and some are poor, 5 rounds of cycle training need to be performed again until five models with good performances are obtained;

[0148] Here, the input of the five models is the relative concentration value of the 18 metabolic biomarkers in the test set, and the output is the low, medium or high risk of the cardiovascular metabolic risk factor spectrum.

[0149] The test set can be used to test the accuracy of the trained model.

[0150] The trained first prediction model or second prediction model is used to predict the test set, and the performance of the prediction result of the model on the test set is obtained; if the performance meets the requirements, that is, the performance value is relatively high, and the performance of the prediction result of the first prediction model or the second prediction model on the test set is close to the performance on the corresponding internal validation set, without the risk of overfitting, it indicates that the model can be generalized, and the trained model meets the requirements.

[0151] The performance can include sensitivity, specificity and accuracy.

[0152] The first prediction model is used to monitor people with high degree of metabolic abnormalities, and its sensitivity in the first internal validation set is 89.03%, the specificity is 76.61%, the accuracy is 85.23%, and the AUC value is 0.934; its sensitivity in the test set is 85.58%, the specificity is 80.60%, the accuracy is 84.06%, and the AUC value is 0.925.

[0153] The second prediction model is used to monitor people with metabolic abnormalities, that is, to monitor people with medium or high degree of metabolic abnormalities. When one of the medium risk or high risk cannot be determined, it can be assisted by the first prediction model to determine whether it is indeed high risk. In order to do this, its sensitivity in the second internal validation set is 85.58%, the specificity is 85.20%, the accuracy is 85.33%, and the AUC value is 0.917; its sensitivity in the test set is 89.72%, the specificity is 83.36%, the accuracy is 85.55%, and the AUC value is 0.909. The model for risk assessment of cardiovascular metabolic risk factor spectrum provided by the application has the characteristics of less sample consumption and high reproducibility.

[0154] The prediction model of the medium risk or high risk type of the cardiovascular metabolic risk factor spectrum is used for training evaluation of whether the to-be-detected person belongs to the medium risk or high risk of the cardiovascular metabolic risk factor spectrum in a subsequent step.

[0155] In an embodiment, the identification module is used to obtain the feature-selected plasma small molecule metabolic biomarkers and the relative concentration values of the to-be-detected person, and input the feature-selected plasma small molecule metabolic biomarkers and the relative concentration values of the to-be-detected person into the prediction model to obtain a prediction result of whether the to-be-detected person belongs to the medium risk or high risk type of the cardiovascular metabolic risk factor spectrum.

[0156] Specifically, the first prediction model can predict a to-be-detected person of a high risk of the cardiovascular metabolic risk factor spectrum; the second prediction model can predict a to-be-detected person of a medium-high risk of the cardiovascular metabolic risk factor spectrum; if a to-be-detected person is predicted by the first prediction model and the second prediction model as a high-risk patient and a medium-high-risk patient respectively, the prediction result of the patient is a high-risk patient.

[0157] The feature-selected plasma small molecule metabolic biomarkers and the relative concentration values of the to-be-detected person can be obtained in a manner similar to that described above in combination with the collection module.

[0158] According to another aspect of the present application, a computer readable storage medium is also provided, which stores computer executable instructions, wherein the computer executable instructions are executed by a processor to make the processor perform the following steps:

[0159] Step S1, the collection module collects plasma samples of subjects of low risk, medium risk, and high risk types of the cardiovascular metabolic risk factor spectrum, and obtains the plasma small molecule metabolic biomarkers and the relative concentration values of the low risk, medium risk, and high risk types from each plasma sample;

[0160] Step S2, the feature selection module performs machine learning based on the plasma small molecule metabolic biomarkers and the relative concentration values of the low risk, medium risk, and high risk types to screen the plasma small molecule metabolic biomarkers that have significant differences between the subjects of the low risk, medium risk, and high risk types of the cardiovascular metabolic risk factor spectrum as the feature-selected plasma small molecule metabolic biomarkers;

[0161] Step S3, the training module trains a prediction model of the medium risk or high risk type of the cardiovascular metabolic risk factor spectrum based on the feature-selected plasma small molecule metabolic biomarkers and the relative concentration values of the subjects of the low risk, medium risk, and high risk types;

[0162] Step S4, the identification module acquires the feature-selected plasma small molecule metabolic biomarkers and the relative concentration values of the to-be-detected person, inputs the feature-selected plasma small molecule metabolic biomarkers and the relative concentration values of the to-be-detected person into the prediction model, to obtain a prediction result of whether the to-be-detected person belongs to a medium-risk or high-risk type of cardiovascular metabolic risk factor spectrum;

[0163] The feature-selected plasma small molecule metabolic biomarkers at least include the following metabolite biomarkers in peripheral plasma:

[0164] Acetoacetic acid, dimethylglycine, sarcosan, daucic acid, serine, homocysteine sulfate lactone, glyceric acid, nicotinic acid, homocysteine, glucose, aminoethanesulfonic acid, uric acid, succinic acid, threonine, cysteine, piperidinic acid, aminoethyl sulfinate and erythrone.

[0165] wherein, the normalized relative concentration value range of acetoacetic acid is 17.48±13.09 for low risk type, 16.33±12.99 for medium risk type, and 15.12±11.76 for high risk type, the relative concentration value unit is %; the normalized relative concentration value range of dimethylglycine is 25.58±15.20 for low risk type, 24.20±15.22 for medium risk type, and 23.24±14.47 for high risk type; the normalized relative concentration value range of creatine is 8.46±5.14 for low risk type, 8.15±5.64 for medium risk type, and 8.18±5 for high risk type; the normalized relative concentration value range of carotenic acid is 15.16±10.78 for low risk type, 14.76±11.07 for medium risk type, and 14.24±10.32 for high risk type; the normalized relative concentration value range of serine is 16.42±12.66 for low risk type, 15.35±12.55 for medium risk type, and 14.32±11.20 for high risk type; the normalized relative concentration value range of homocysteine sulphoxide is 20.92±14.00 for low risk type, 19.45±13.42 for medium risk type, and 18.81±12.74 for high risk type; the normalized relative concentration value range of glyceric acid is 16.15±12.77 for low risk type, 15.38±12.54 for medium risk type, and 14.82±12.01 for high risk type; the normalized relative concentration value range of nicotinic acid is 19.57±12.96 for low risk type, 18.60±13.05 for medium risk type, and 18.13±12.68 for high risk type; the normalized relative concentration value range of homocysteine is 7.80±5.76 for low risk type, 7.51±6.08 for medium risk type, and 7.21±6.08 for high risk type; and the normalized relative concentration value range of glucose is 15.26±9.09 for low risk type, 17.22±11.46 for medium risk type, and 21.47±14.01; the normalized relative concentration value range of aminoethanesulfonic acid is 19.69±13.79, the normalized relative concentration value range of the medium risk type is 18.75±14.04, and the normalized relative concentration value range of the high risk type is 17.94±13.27; the normalized relative concentration value range of uric acid is 14.14±10.69, the normalized relative concentration value range of the medium risk type is 13.59±10.73, and the normalized relative concentration value range of the high risk type is 13.92±11.16; the normalized relative concentration value range of succinic acid is 16.09±11.56, the normalized relative concentration value range of the medium risk type is 15.17±11.56, and the normalized relative concentration value range of the high risk type is 14.62±11.14; the normalized relative concentration value range of threonine is 18.82±10.81, the normalized relative concentration value range of the medium risk type is 17.75±10.20, and the normalized relative concentration value range of the high risk type is 17.82±10.23; the normalized relative concentration value range of cysteine is 16.42±12.66, the normalized relative concentration value range of the medium risk type is 15.35±12.55, and the normalized relative concentration value range of the high risk type is 14.32±11.20; the normalized relative concentration value range of piperidinic acid is 17.34±12.43, the normalized relative concentration value range of the medium risk type is 16.85±12.78, and the normalized relative concentration value range of the high risk type is 16.84±12.87; the normalized relative concentration value range of aminoethylsulfinic acid is 17.04±13.30, the normalized relative concentration value range of the medium risk type is 15.77±13.28, and the normalized relative concentration value range of the high risk type is 14.53±11.85; the normalized relative concentration value range of erythrionic acid is 18.42±12.16, the normalized relative concentration value range of the medium risk type is 18.10±12.60, and the normalized relative concentration value range of the high risk type is 17.73±12.42.

[0166] According to another aspect of the present application, a detection kit for identification of cardiovascular metabolic risk factor spectrum is also provided, and the metabolite biomarkers in the detection kit comprise at least the following metabolite biomarkers in peripheral plasma:

[0167] Acetoacetic acid, Dimethylglycine, Creatinine, Malonic acid, L-Serine, Homocysteine thiolactone, Glyceric acid, Nicotinic acid, Homocysteine, D-Glucose, Taurine, Uric acid, Succinic acid, L-Threonine, L-Cysteine, Pipecolic acid, Hypotaurine, and Erythronic acid.

[0168] The metabolic biomarkers in the detection kit at least include the following metabolic biomarkers in peripheral blood plasma: acetoacetate, the normalized relative concentration value range of the low-risk type is 17.48±13.09, the normalized relative concentration value range of the medium-risk type is 16.33±12.99, the normalized relative concentration value range of the high-risk type is 15.12±11.76, the relative concentration value unit is %; dimethylglycine, the normalized relative concentration value range of the low-risk type is 25.58±15.20, the normalized relative concentration value range of the medium-risk type is 24.20±15.22, the normalized relative concentration value range of the high-risk type is 23.24±14.47; creatinine, the normalized relative concentration value range of the low-risk type is 8.46±5.14, the normalized relative concentration value range of the medium-risk type is 8.15±5.64, the normalized relative concentration value range of the high-risk type is 8.18±5; carotinic acid, the normalized relative concentration value range of the low-risk type is 15.16±10.78, the normalized relative concentration value range of the medium-risk type is 14.76±11.07, the normalized relative concentration value range of the high-risk type is 14.24±10.32; serine, the normalized relative concentration value range of the low-risk type is 16.42±12.66, the normalized relative concentration value range of the medium-risk type is 15.35±12.55, the normalized relative concentration value range of the high-risk type is 14.32±11.20; homocysteine sulfate lactone, the normalized relative concentration value range of the low-risk type is 20.92±14.00, the normalized relative concentration value range of the medium-risk type is 19.45±13.42, the normalized relative concentration value range of the high-risk type is 18.81±12.74; glyceric acid, the normalized relative concentration value range of the low-risk type is 16.15±12.77, the normalized relative concentration value range of the medium-risk type is 15.38±12.54, the normalized relative concentration value range of the high-risk type is 14.82±12.01; nicotinic acid, the normalized relative concentration value range of the low-risk type is 19.57±12.96, the normalized relative concentration value range of the medium-risk type is 18.60±13.05, the normalized relative concentration value range of the high-risk type is 18.13±12.68; homocysteine, the normalized relative concentration value range of the low-risk type is 7.80±5.76, the normalized relative concentration value range of the medium-risk type is 7.51±6.08, the normalized relative concentration value range of the high-risk type is 7.21±6.08; glucose, the normalized relative concentration value range of the low-risk type is 15.26±9.09, the normalized relative concentration value range of the medium-risk type is 17.22±11.46, the normalized relative concentration value range of the low risk type is 21.47±14.01; the normalized relative concentration value range of the low risk type is 19.69±13.79, the normalized relative concentration value range of the medium risk type is 18.75±14.04, and the normalized relative concentration value range of the high risk type is 17.94±13.27; the normalized relative concentration value range of the low risk type is 14.14±10.69, the normalized relative concentration value range of the medium risk type is 13.59±10.73, and the normalized relative concentration value range of the high risk type is 13.92±11.16; the normalized relative concentration value range of the low risk type is 16.09±11.56, the normalized relative concentration value range of the medium risk type is 15.17±11.56, and the normalized relative concentration value range of the high risk type is 14.62±11.14; the normalized relative concentration value range of the low risk type is 18.82±10.81, the normalized relative concentration value range of the medium risk type is 17.75±10.20, and the normalized relative concentration value range of the high risk type is 17.82±10.23; the normalized relative concentration value range of the low risk type is 16.42±12.66, the normalized relative concentration value range of the medium risk type is 15.35±12.55, and the normalized relative concentration value range of the high risk type is 14.32±11.20; the normalized relative concentration value range of the low risk type is 17.34±12.43, the normalized relative concentration value range of the medium risk type is 16.85±12.78, and the normalized relative concentration value range of the high risk type is 16.84±12.87; the normalized relative concentration value range of the low risk type is 17.04±13.30, the normalized relative concentration value range of the medium risk type is 15.77±13.28, and the normalized relative concentration value range of the high risk type is 14.53±11.85; the normalized relative concentration value range of the low risk type is 18.42±12.16, the normalized relative concentration value range of the medium risk type is 18.10±12.60, and the normalized relative concentration value range of the high risk type is 17.73±12.42.

[0169] In summary, the present application obtains a prediction model of the medium risk or high risk type of the cardiovascular metabolic risk factor spectrum by machine learning of the plasma metabolite biomarker, and exhibits excellent risk evaluation capability of sensitivity, specificity and accuracy. The present application realizes efficient evaluation capability of the cardiovascular metabolic risk factor spectrum risk. The present application only needs a small amount of plasma as a detection sample, is simple and convenient compared with the traditional cumbersome evaluation process, has low cost, is easy to be accepted by subjects, and is suitable for wide promotion of the method. And the high sensitivity and specificity enable the present application to also be used in screening of the cardiovascular metabolic risk factor spectrum drugs.

[0170] Obviously, various modifications and changes can be made to the present application by those skilled in the art without departing from the spirit and scope of the present application. Thus, if these modifications and changes of the present application fall within the scope of the claims and their equivalents, they are intended to be included therein.

[0171] It should be noted that the present application can be implemented in software and / or a combination of software and hardware, for example, can be implemented by using an application specific integrated circuit (ASIC), a general purpose computer or any other similar hardware device. In one embodiment, the software program of the present application can be executed by a processor to implement the steps or functions described above. Similarly, the software program of the present application (including related data structures) can be stored in a computer readable recording medium, for example, a RAM memory, a magnetic or optical drive or a soft disk and the like. In addition, some steps or functions of the present application can be implemented by using hardware, for example, as a circuit cooperating with the processor to execute the respective steps or functions.

[0172] In addition, a part of the present application can be applied as a computer program product, for example, computer program instructions, when executed by a computer, through the operation of the computer, can invoke or provide the method and / or technical solutions according to the present application. The program instructions invoking the method of the present application can be stored in a fixed or removable recording medium, and / or transmitted through a data stream in a broadcast or other signal bearing medium, and / or stored in the working memory of the computer device running according to the program instructions. Here, one embodiment according to the present application includes a device including a memory for storing computer program instructions and a processor for executing program instructions, wherein when the computer program instructions are executed by the processor, the device triggers the operation of the method and / or technical solutions based on the foregoing according to the plurality of embodiments of the present application.

[0173] It will be obvious to a person skilled in the art that the application is not limited to the details of the above-described exemplary embodiments, but that the application can be implemented in other concrete forms without deviating from the spirit or the basic characteristics of the application. The embodiments are therefore to be considered in all respects as illustrative and not restrictive, the scope of the application being defined by the appended claims rather than by the above Description, which is therefore intended merely as a specification. All changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein. Any reference signs in the claims should not be construed as limiting the claim concerned. Furthermore, it is to be noted that the term "comprising" does not exclude other elements or steps, that the term "a" or "an" does not exclude a plurality, and that a single processor or other unit can fulfil the functions of several units recited in the claims. The terms first, second and the like do not denote any ordering, but rather are used as names for naming different units.

Claims

1. A system for constructing a recognition model of a cardiovascular metabolic risk factor profile, characterized by, The system comprises: a collection module configured to collect plasma samples of subjects of low-risk, medium-risk and high-risk types of a cardiovascular metabolic risk factor spectrum, and obtain plasma small molecule metabolic biomarkers and relative concentration values of the plasma small molecule metabolic biomarkers of the low-risk, medium-risk and high-risk types from the plasma samples; a feature selection module configured to perform machine learning based on the plasma small molecule metabolic biomarkers and the relative concentration values of the plasma small molecule metabolic biomarkers of the low-risk, medium-risk and high-risk types, and screen plasma small molecule metabolic biomarkers that have significant differences between the subjects of the low-risk, medium-risk and high-risk types of the cardiovascular metabolic risk factor spectrum as feature-selected plasma small molecule metabolic biomarkers; a training module configured to train a prediction model of the medium-risk or high-risk type of the cardiovascular metabolic risk factor spectrum based on the feature-selected plasma small molecule metabolic biomarkers and the relative concentration values of the feature-selected plasma small molecule metabolic biomarkers of the subjects of the low-risk, medium-risk and high-risk types; an identification module configured to obtain feature-selected plasma small molecule metabolic biomarkers and relative concentration values of the feature-selected plasma small molecule metabolic biomarkers of a to-be-detected subject, and input the feature-selected plasma small molecule metabolic biomarkers and the relative concentration values of the feature-selected plasma small molecule metabolic biomarkers of the to-be-detected subject into the prediction model to obtain a prediction result of whether the to-be-detected subject belongs to the medium-risk or high-risk type of the cardiovascular metabolic risk factor spectrum; the feature-selected plasma small molecule metabolic biomarkers at least include the following metabolite biomarkers in peripheral plasma: acetoacetate, dimethylglycine, sarcosan, daucic acid, serine, homocysteine sulfate lactone, glyceric acid, nicotinic acid, homocysteine, glucose, aminoethanesulfonic acid, uric acid, succinic acid, threonine, cysteine, piperidinic acid, aminoethyl sulfinate, and erythrulose.

2. The identification model construction system of the cardiovascular metabolic risk factor spectrum according to claim 1, wherein the collection module is configured to obtain a first subject group and a second subject group that are mutually exclusive, and the first subject group and the second subject group each comprise subjects of low-risk, medium-risk and high-risk types of the cardiovascular metabolic risk factor spectrum that are mutually exclusive; the collection module is configured to collect first plasma samples of the first subject group, and obtain plasma small molecule metabolic biomarkers and relative concentration values of the plasma small molecule metabolic biomarkers from the first plasma samples as a first set; the collection module is configured to collect second plasma samples of the second subject group, and obtain plasma small molecule metabolic biomarkers and relative concentration values of the plasma small molecule metabolic biomarkers from the second plasma samples as a test set.

3. The identification model construction system of the cardiovascular metabolic risk factor profile according to claim 2, wherein, the feature selection module is configured to perform feature selection on the plasma metabolite biomarkers on the first set using Kruskal-Wallis rank sum test, and set a significance threshold to obtain feature-selected plasma small molecule metabolic biomarkers.

4. The identification model construction system of cardiovascular metabolic risk factor profiles according to claim 3, characterized in that, The training module is configured to train a candidate prediction model of the medium-risk or high-risk type of the cardiovascular metabolic risk factor spectrum based on the feature-selected plasma small molecule metabolic biomarkers and their relative concentration values of the subjects of the low-risk, medium-risk and high-risk types in the first set; and obtain a final prediction model of the medium-risk or high-risk type of the cardiovascular metabolic risk factor spectrum based on the test set and the first set.

5. The identification model construction system of the cardiovascular metabolic risk factor spectrum according to claim 4, wherein, The training module is configured to classify the subjects of the low-risk and medium-risk types in the first set into a third type of subjects to obtain a third type of subject group; The training module is configured to classify the subjects of the high-risk type in the first set into a fourth type of subject group to obtain a fourth type of subject group; The training module is configured to train a first candidate prediction model of the cardiovascular metabolic risk factor spectrum based on the feature-selected plasma small molecule metabolic biomarkers and their relative concentration values of the third type of subject group and the fourth type of subject group; and obtain a final prediction model of the high-risk type of the cardiovascular metabolic risk factor spectrum, i.e., a first prediction model, based on the performance of the prediction results on the test set, the third type of subject group and the fourth type of subject group of the first candidate prediction model. The training module is configured to classify the subjects of the low-risk type in the first set into a fifth type of subject to obtain a fifth type of subject group; and classify the subjects of the medium-risk and high-risk types in the first set into a sixth type of subject to obtain a sixth type of subject group; The training module is configured to train a second candidate prediction model of the cardiovascular metabolic risk factor spectrum based on the feature-selected plasma small molecule metabolic biomarkers and their relative concentration values of the fifth type of subject group and the sixth type of subject group; and obtain a final prediction model of the medium-high-risk type of the cardiovascular metabolic risk factor spectrum, i.e., a second prediction model, based on the performance of the prediction results on the test set, the third type of subject group and the fourth type of subject group of the second candidate prediction model of the cardiovascular metabolic risk factor spectrum.

6. The identification model construction system of cardiovascular metabolic risk factor profiles according to claim 5, characterized in that, The training module is configured to divide the third type of subject group and the fourth type of subject group into five non-overlapping data each time, select one piece of non-overlapping data that has not been selected as a first internal validation set each time, and select the remaining four pieces of data as a first training set each time; wherein the first internal validation set and the first training set each include non-overlapping third type of subjects and fourth type of subjects. The training module is configured to train the high-risk prediction model of the cardiovascular metabolic risk factor spectrum for 5 rounds based on the first internal validation set and the first training set each time and using an extreme gradient boosting algorithm, so as to obtain five first candidate prediction models that meet the performance requirements; wherein the XGBoostd hyperparameters of the extreme gradient boosting algorithm are set as a learning rate eta=0.3, a minimum weight sum of all observation values of a subset min_child_weight=1, a maximum depth of a tree max_depth=6, a minimum target function reduction required for further branching at a leaf node of the tree gamma=0, a sample rate of samples subsample=0.8 when constructing each tree, a feature sampling rate of the tree colsample_bytree=1 when constructing each tree, a feature sampling rate of the tree colsample_bylevel=1 when constructing each layer, a feature sampling rate of each leaf node colsample_bynode=1, an L1 regularization weight alpha=0, an L2 regularization weight lambda=1, and a maximum number of iterations nrounds=200; The training module is configured to input the test set into the five first candidate prediction models to obtain prediction results of the first candidate prediction models on the test set, wherein the prediction results of the first candidate prediction models on the test set are determined by voting of the five first candidate prediction models, and a minority submits to the majority principle is adopted to obtain the prediction results of the first candidate prediction models on the test set; if the performance value of the prediction results of the first candidate prediction models on the test set is higher than a preset performance threshold value, and the performance of the prediction results on the test set and the performance of the prediction results on the first internal validation set differ by less than a preset difference threshold value, the five first candidate prediction models that meet the performance requirements are taken as the final first prediction model.

7. The identification model construction system of cardiovascular metabolic risk factor profiles according to claim 5, wherein, The training module is configured to divide the fifth and sixth subject groups into five mutually exclusive data, and each time, one of the five mutually exclusive data that is not selected is taken as a second internal validation set, and the remaining four data are taken as a second training set; the second internal validation set and the second training set each include mutually exclusive fifth and sixth subjects. The training module is configured to train the prediction model for the medium and high risk of the cardiovascular and metabolic risk factor spectrum for 5 rounds based on the second internal validation set and the second training set each time and using the extreme gradient boosting algorithm, so as to obtain 5 second candidate prediction models that meet the performance requirements; wherein the XGBoostd hyperparameters of the extreme gradient boosting algorithm are set as a learning rate eta = 0.3, a minimum weight sum of all observation values of a subset min_child_weight = 1, a maximum depth of a tree max_depth = 6, a minimum target function reduction required for further branching at a leaf node of the tree gamma = 0, a sample rate of samples when constructing each tree subsample = 0.8, a feature sampling rate when constructing each tree colsample_bytree = 1, a feature sampling rate when constructing each level colsample_bylevel = 1, a feature sampling rate when constructing each leaf node colsample_bynode = 1, an L1 regularization weight alpha = 0, an L2 regularization weight lambda = 1, and a maximum number of iterations nrounds = 200; The training module is configured to input the test set into the 5 second candidate prediction models that meet the performance requirements respectively for prediction, so as to obtain the prediction results of the second candidate prediction models on the test set, wherein the prediction results of the first candidate prediction models on the test set are determined by voting of the 5 first candidate prediction models that meet the performance requirements, and the minority is subject to the majority principle to obtain the prediction results of the second candidate prediction models on the test set; if the performance value of the prediction results of the second candidate prediction models on the test set is higher than a preset performance threshold value, and the performance of the prediction results on the test set and the performance of the prediction results on the second internal validation set differ by less than a preset difference threshold value, the 5 second candidate prediction models that meet the performance requirements are taken as the final second prediction model.

8. A computer-readable storage medium having stored thereon computer- executable instructions, wherein, The computer executable instructions, when executed by the processor, cause the processor to perform the following steps: The acquisition module collects plasma samples of subjects of low, medium and high risk types of the cardiovascular and metabolic risk factor spectrum, and obtains low, medium and high risk type plasma small molecule metabolic biomarkers and their relative concentration values from each plasma sample; The feature selection module performs machine learning based on the low, medium and high risk type plasma small molecule metabolic biomarkers and their relative concentration values, to screen low, medium and high risk type plasma small molecule metabolic biomarkers that have significant differences between subjects of the cardiovascular and metabolic risk factor spectrum, as the selected plasma small molecule metabolic biomarkers; The training module trains the prediction model for the medium and high risk of the cardiovascular and metabolic risk factor spectrum based on the selected plasma small molecule metabolic biomarkers and their relative concentration values of the subjects of the low, medium and high risk types; and The training module trains the prediction model for the medium and high risk of the cardiovascular and metabolic risk factor spectrum based on the selected plasma small molecule metabolic biomarkers and their relative concentration values of the subjects of the low, medium and high risk types. The identification module obtains the characteristic-selected plasma small molecule metabolic biomarkers and relative concentration values of the to-be-detected person, inputs the characteristic-selected plasma small molecule metabolic biomarkers and relative concentration values of the to-be-detected person into the prediction model, so as to obtain a prediction result of whether the to-be-detected person belongs to a medium-risk or high-risk type of cardiovascular metabolic risk factor spectrum. The characteristic-selected plasma small molecule metabolic biomarkers at least include the following metabolite biomarkers in peripheral plasma: Acetoacetic acid, dimethylglycine, sarcosine, daucic acid, serine, homocysteine sulfate lactone, glyceric acid, nicotinic acid, homocysteine, glucose, aminoethanesulfonic acid, uric acid, succinic acid, threonine, cysteine, piperidinic acid, aminoethyl sulfinate and erythrone.

9. A test kit for the identification of a cardiovascular metabolic risk profile, characterized in that, The metabolite biomarkers in the detection kit at least include the following metabolite biomarkers in peripheral plasma: Acetoacetic acid, dimethylglycine, sarcosine, daucic acid, serine, homocysteine sulfate lactone, glyceric acid, nicotinic acid, homocysteine, glucose, aminoethanesulfonic acid, uric acid, succinic acid, threonine, cysteine, piperidinic acid, aminoethyl sulfinate and erythrone; wherein the metabolic biomarkers and relative concentration values of the to-be-detected person are input into a prediction model, so as to obtain a prediction result of whether the to-be-detected person belongs to a medium-risk or high-risk type of cardiovascular metabolic risk factor spectrum.

Images