Pre-eclampsia risk assessment and prediction model based on metabonomics and application thereof

Abstract

The invention provides a metabonomics-based preeclampsia risk assessment prediction model construction method, a prediction model, application of metabolites, a metabolite detection kit, a prediction product, a prediction system and related application thereof. Specifically, clinical indexes and metabolite data of a pregnant woman serve as input, whether the pregnant woman suffers from preeclampsia or not serves as output, model construction is conducted through a machine learning method, and a preeclampsia risk assessment and prediction model is obtained. The metabolite is selected from at least one of 4-hydroxyphenyllactic acid, homovanillic acid, ornithine, sarcosine, adenosine monophosphate and 3-methylbutyryl carnitine. According to the invention, not only are six metabolite markers strongly associated with preeclampsia found, but also effective prediction of early preeclampsia and late preeclampsia is realized by establishing a prediction model, and the method has extremely high clinical value.

Description

Technical Field

[0001] The present invention relates to the field of preeclampsia analysis, and particularly to a construction method of a preeclampsia risk assessment and prediction model based on metabolomics, a prediction model, an application of metabolites, a metabolite detection kit, a prediction product, a prediction system and related applications thereof. Background Art

[0002] Preeclampsia (PE) refers to new-onset hypertension and proteinuria that occur in patients with normal blood pressure after 20 weeks of pregnancy or postpartum, as well as dysfunction of other organs. The continuous progression of the condition will cause a series of complications, endangering the lives of the mother and fetus. It is the second leading cause of maternal death after postpartum hemorrhage, accounting for more than 10% of maternal death cases. The overall prevalence of preeclampsia globally is about 2% - 4%, and about 2.3% in China. However, the proportion of severe PE patients in China is much higher than that in other countries. Preeclampsia can be divided into two categories according to the onset time: early-onset preeclampsia is less than 34 weeks and late-onset preeclampsia is greater than or equal to 34 weeks. Currently, the pathological mechanism of early-onset preeclampsia is not fully understood and may be related to genetic factors and immune factors. Current research believes that in the first trimester of pregnancy, due to factors such as shallow placental implantation, maternal immune dysfunction, defective trophoblast differentiation, and oxidative stress, abnormal remodeling of maternal spiral arteries will be triggered, ultimately resulting in insufficient uteroplacental perfusion after 20 weeks of pregnancy. Placental ischemia and hypoxia will release a variety of cytokines into the mother's body, thereby changing the function of endothelial cells in the mother's body, and then triggering the characteristic systemic symptoms of preeclampsia, leading to fetal growth restriction and iatrogenic premature birth and other outcomes. Late-onset preeclampsia is mostly caused by placental perfusion disorders caused by maternal vascular diseases, such as chronic hypertension, vascular diseases, pre-gestational diabetes, etc. Due to its heterogeneity and complexity, the current main treatment method is still termination of pregnancy and delivery of the placenta, and its risk prediction and early diagnosis are also very difficult. The serious consequences associated with PE are not only reflected in the pregnancy symptoms of pregnant women, but may also bring a series of complications to the newborn after childbirth. Severe complications also include internal bleeding, epilepsy, stroke, death, etc. Therefore, developing a more early and accurate PE risk prediction method, screening high-risk populations, and then timely performing drug intervention and disease course monitoring is the most valuable solution, which is of great significance for the prevention and effective management of preeclampsia in pregnant women.

[0003] Currently, many related teams have published work on predicting the risk of preeclampsia. In the widely validated model of the Fetal Medicine Foundation in the UK, the current best prediction model combines various basic factors of the mother with mean arterial pressure, uterine artery pulsatility index, and serum placental growth factor (PLGF), with a 90% prediction rate for very early PE and also a good detection rate for preterm and term preeclampsia. However, the uterine artery pulsatility index cannot be guaranteed to be generally obtained in domestic hospitals. There are also various detection methods for preeclampsia based on metabolites (metabolite markers) in the prior art, but they generally have deficiencies in aspects such as unrepresentative index selection and lack of effective validation. Therefore, it is of great significance to find more representative markers to develop a more universal and easy-to-implement prediction method. Summary of the Invention

[0004] Based on the metabolomic test data of pregnant women, this invention uses liquid chromatography-tandem mass spectrometry technology to detect multiple metabolites related to preeclampsia, analyzes the expression profiles of specific metabolites related to preeclampsia in the subjects, and obtains a set of 6 metabolite markers that can be used to predict the risk of preeclampsia. By combining these metabolites with the clinical indicators of pregnant women such as mean arterial pressure (MAP) and maternal factors, the risk of preeclampsia (including early-onset preeclampsia and late-onset preeclampsia) in the subjects is evaluated.

[0005] In the first aspect, this invention provides a method for constructing a preeclampsia risk assessment prediction model based on metabolomics, which includes using the clinical indicators and metabolite data of pregnant women as inputs and whether the pregnant woman has preeclampsia as the output, and constructing the model through machine learning methods to obtain a preeclampsia risk assessment prediction model.

[0006] Among them, the metabolites are selected from at least one of 4-hydroxyphenyllactic acid, homovanillic acid, ornithine, sarcosine, adenosine monophosphate, and 3-methylbutyrylcarnitine. The clinical indicators of pregnant women include age, height at 12-17 +6 weeks of gestation, weight at 12-17 +6 weeks of gestation, systolic / diastolic blood pressure at the time of detection at 12-17 +6 weeks of gestation, parity, number of deliveries, history of adverse pregnancy, history of adverse past events, and clinical data of assisted reproduction; the history of adverse past events includes at least one of a history of gestational diabetes mellitus, preeclampsia, familial preeclampsia, chronic hypertension, systemic lupus erythematosus, and antiphospholipid syndrome.

[0007] Preferably, for the risk assessment prediction of early-onset preeclampsia, the above metabolites are 4-hydroxyphenyllactic acid, homovanillic acid, ornithine, and sarcosine. For the risk assessment prediction of late-onset preeclampsia, the above metabolites are 3-methylbutyrylcarnitine, adenosine monophosphate, and ornithine.

[0008] Preferably, the clinical indicators of pregnant women include age, height of pregnant women at 12-17 +6 weeks of gestation, weight of pregnant women at 12-17 +6 weeks of gestation, systolic blood pressure / diastolic blood pressure measured at 12-17 +6 weeks of gestation, parity, number of deliveries, history of adverse pregnancy, history of adverse past medical conditions (including previous gestational diabetes, preeclampsia, familial preeclampsia, chronic hypertension, systemic lupus erythematosus, and antiphospholipid syndrome), and clinical data such as assisted reproduction.

[0009] Preferably, the above machine learning method is a random forest model.

[0010] After obtaining the above prediction model, a threshold can be set according to the model evaluation index of the prediction model, such as specificity. This threshold is used as the standard for classifying and judging the risk value output by the prediction model when using the prediction model for prediction. Specifically, the threshold is set based on the specificity of the prediction model being 80%-98%.

[0011] In a second aspect, the present invention provides the use of metabolites related to preeclampsia in the preparation of detection products related to preeclampsia, and the above metabolites are selected from at least one of 4-hydroxyphenyllactic acid, homovanillic acid, ornithine, sarcosine, adenosine monophosphate, and 3-methylbutyrylcarnitine.

[0012] In a third aspect, the present invention provides a detection kit for metabolites related to preeclampsia, and the metabolites are selected from at least one of 4-hydroxyphenyllactic acid, homovanillic acid, ornithine, sarcosine, adenosine monophosphate, and 3-methylbutyrylcarnitine. This kit is based on the LC-MS / MS method to achieve quantitative detection of the above metabolite indicators.

[0013] In a fourth aspect, based on the prediction model obtained in the first aspect above, the present invention provides a method for predicting preeclampsia risk based on metabolomics (as Figure 1 shown), and this method includes:

[0014] 1) Obtain the data of the clinical indicators and metabolites of the pregnant woman to be tested;

[0015] 2) Input the data of the clinical indicators and metabolites of the above pregnant woman into the prediction model for processing to obtain the risk value of preeclampsia of the above pregnant woman. When the risk value is higher than the set threshold, it is determined that the pregnant woman has a high risk of preeclampsia.

[0016] In a fifth aspect, based on the prediction method in the fourth aspect above, the present invention provides a prediction system for preeclampsia risk assessment based on metabolomics, and this system includes:

[0017] 1) A device for obtaining data of clinical indicators and metabolites of pregnant women to be tested;

[0018] 2) A device for performing predictive model processing on the data of clinical indicators and metabolites of pregnant women obtained in 1) above;

[0019] 3) A device for outputting a prediction result.

[0020] In a sixth aspect, the present invention provides a predictive product for preeclampsia risk assessment based on metabolomics, the product comprising: a memory and a processor. The memory is used for storing a program; the processor is used for implementing the preeclampsia risk assessment prediction method as mentioned in the fifth aspect above by executing the program stored in the above memory.

[0021] Meanwhile, the present invention also provides a computer-readable storage medium, on which a program is stored, and the program can be executed by a processor to implement the preeclampsia risk assessment prediction method as mentioned in the fourth aspect above.

[0022] In addition, the present invention also provides a computer-readable storage medium, on which a predictive model obtained by the construction method of the first aspect is stored.

[0023] The beneficial effects of the present invention are as follows:

[0024] 1. The present invention realizes the discovery of preeclampsia metabolite markers. It has been verified in multiple batches of test data that there are differences in relevant marker indicators between early-onset preeclampsia samples and healthy control samples, indicating a strong correlation between them and the risk of pregnant women suffering from preeclampsia. Based on the relevant metabolite markers, the research and preparation of a special kit based on metabolite markers have been completed, making the data information of relevant markers easier to obtain and analyze, and bringing a new interpretation in the metabolic dimension for the risk prediction of preeclampsia.

[0025] 2. The present invention establishes a predictive model for early pregnancy preeclampsia risk assessment based on metabolite marker data and maternal clinical information, and uses a relatively large sample size (see the examples) to evaluate the prediction effects of the above early-onset preeclampsia and late-onset preeclampsia models. The results show that the sensitivity and specificity of the early-onset preeclampsia prediction model are 63.58% and 90% respectively, and the sensitivity and specificity of the late-onset preeclampsia prediction model are 53% and 90% respectively. The model has good accuracy and has clinical value. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is a schematic diagram of the preeclampsia risk assessment prediction method based on metabolomics of the present invention;

[0027] Figure 2Differential expression analysis of four early-onset preeclampsia metabolite markers in the datasets of three hospitals (a, b, c) in the training set of an embodiment of the present invention, namely, the MoM value of Homovanillic acid (Figure A), the MoM value of Sarcosine (Figure B), the MoM value of Hydroxyphenyllactic acid (Figure C), the MoM value of Ornithine (Figure D), and the product of the MoM values of the four metabolites (Figure E), where the white part is the normal group and the black part is the diseased group;

[0028] Figure 3 Differential expression analysis of three late-onset preeclampsia metabolite markers in the datasets of three hospitals (a, b, c) in the training set of an embodiment of the present invention, namely, the MoM value of Ornithine (Figure A), the MoM value of Isovelarylcarnitine (Figure B), the MoM value of Adenosine monophosphate (Figure C), and the product of the MoM values of the three metabolites (Figure D), where the white part is the normal group and the black part is the diseased group;

[0029] Figure 4 Clinical high-risk indicators of preeclampsia in the pilot test (hospitals a and b) of an embodiment of the present invention. Figures A - C: respectively show the differences in age, BMI, and mean arterial pressure (MAP) among the healthy control (control), early-onset preeclampsia (EPE), and late-onset preeclampsia (LPE) groups;

[0030] Figure 5 Clinical high-risk indicators of preeclampsia in the mid-term test (hospitals d, e, and f) of an embodiment of the present invention. Figures A - C: respectively show the differences in age, BMI, and mean arterial pressure (MAP) among the healthy control (control), early-onset preeclampsia (EPE), and late-onset preeclampsia (LPE) groups;

[0031] Figure 6 Differential expression of the MoM value of Homovanillic acid (Figure A), the MoM value of Sarcosine (Figure B), the MoM value of Hydroxyphenyllactic acid (Figure C), the MoM value of Ornithine (Figure D), and the product of the MoM values of the four metabolites (Figure E) of the early-onset preeclampsia metabolite markers in the early-onset preeclampsia (EPE) and healthy control (control) samples in the training set, validation set (pilot test), and test set (mid-term test) of an embodiment of the present invention;

[0032] Figure 7In an embodiment of the present invention, the MoM values of the metabolites Ornithine, Isovelarylcarnitine, and Adenosine monophosphate, and the product of the MoM values of the three metabolites (Figure D) in late-onset preeclampsia, and their differential expression in late-onset preeclampsia (LPE) and healthy control samples in the training set, validation set (pilot test), and test set (pilot scale).

[0033] Figure 8 In an embodiment of the present invention, the prediction effects of the preeclampsia metabolic prediction models (early preeclampsia prediction model (Figure A) and late preeclampsia prediction model (Figure B)) in all datasets, where Train is the training set, vali is the validation set (pilot test), and test is the test set (pilot scale). Detailed implementation manners

[0034] In recent years, many research results have been published on the use of multi-omics methods including metabolomics to study preeclampsia prediction. For example, "Application of Arachidonic Acid Metabolites in the Preparation of a Preeclampsia Detection Kit" involves the application of arachidonic acid metabolites in the preparation of a preeclampsia detection kit. By detecting the arachidonic acid metabolites (5-HETE, 15-HETE, LTB4) in 49 serum samples of preeclampsia patients, it further verifies the good correlation with the blood samples of preeclampsia patients. It is considered that arachidonic acid metabolites can be used as a biomarker for preeclampsia detection, and by quantitatively measuring the content, it can be used for preeclampsia detection. In addition, "Detection of Preeclampsia Risk" involves a test method for early detection of the risk of preeclampsia in pregnant women, including the determination of the following metabolic biomarkers in samples taken from women at 11 to 17 weeks of pregnancy: 5-hydroxytryptophan, monosaccharides, decanoyl carnitine, methylglutaric acid and / or adipic acid, oleic acid, docosahexaenoic acid, docosatriynoic acid, γ-butyrolactone, dihydro-3(2H)-furanone, 2-oxovaleric acid, oxomethylbutyric acid, acetoacetic acid, hexadecenoyl-eicosatetraenyl-sn-glycerol, sphingosine 1-phosphate, dihydrosphingosine 1-phosphate, vitamin D3 derivatives. However, the current research on the characterization analysis of preeclampsia metabolites itself still has deficiencies and lacks verification data with a large sample size, and subsequent applications, especially model construction, lack universality.

[0035] Based on the limitations of the prior art, the object of the present invention is to develop metabolite markers for preeclampsia through metabolomics research methods, in order to analyze a series of disease-related metabolic indicators that can be obtained more easily and quickly, and combine relevant clinical information to obtain a model that can accurately predict the risk of preeclampsia during pregnancy, so as to realize the prediction of preeclampsia patients and help clinical early intervention judgment. Specifically, the present invention uses LS-MS / MS technology to detect 6 specific preeclampsia-related metabolite markers, combines the mean arterial pressure (MAP) and maternal factors to evaluate the risk of early-onset preeclampsia and late-onset preeclampsia in the subjects; the target population is all pregnant women. Risk assessment can help the high-risk population to carry out drug (aspirin) intervention in advance, reduce the incidence probability, and has important clinical value and significance.

[0036] Based on the R & D data in the exploration stage, the present invention analyzes the content of each metabolite in the peripheral blood plasma samples of pregnant women at 12-17 +6 weeks of gestation, discovers and validates 6 metabolite markers that can be used to predict the risk of preeclampsia, including 4-hydroxyphenyllactic acid, homovanillic acid, ornithine, sarcosine, adenosine monophosphate, and 3-methylbutyrylcarnitine. The physiological significance of each index is shown in Table 1 below.

[0037] Table 1 Metabolites and Their Clinical Significance

[0038]

[0039] To better achieve the precise detection of the above metabolites, the present invention establishes a method for quantitative detection of metabolite indicators based on the LC-MS / MS platform, and conducts a series of optimizations on the detection method. The 6 predictive metabolite markers for the risk of preeclampsia are all discovered based on the previous targeted metabolomics detection data, and then the targeted detection mass spectrometry methods for these 6 metabolic indicators are further confirmed and debugged. The characteristic detection ion pair information of each index in the method is shown in Table 2 below. And, based on the retention time distribution of each index in the liquid chromatography column, the mobile phase ratio gradient is further optimized, and finally a 5-minute liquid chromatography method is determined to simultaneously quantitatively detect these 6 metabolic indicators in the sample. All sample detections are carried out on the UPLC-QTRAP5500 platform (chromatographic column: BEH C18 chromatographic column (2.1X50mm, 1.7μm, Waters)), and the detailed mass spectrometry and liquid chromatography parameters are shown in Table 3 below.

[0040] Table 2 Ion Pair Information

[0041] Compound Name Q1 / Q3 RT(min) DP CE Hydroxyphenyllactic acid 316.11 / 137.21 2.3 -90 -30 Homovanillic acid 316.01 / 151.81 2.27 -120 -24 Sarcosine 223.01 / 137.11 0.98 -90 -30 Ornithine 266.01 / 137.11 1.00 -100 -34 Adenosine monophosphate 483.01 / 135.91 1.82 100 40 Isovelarylcarnitine 381.23 / 220.13 2.29 80 25

[0042] Table 3 Gradient Elution Conditions of Liquid Chromatography Method

[0043] (Mobile phase A: 0.1% formic acid aqueous solution; Mobile phase B: acetonitrile: isopropanol = 7:3 (v:v))

[0044] Time(min) Flow(mL / min) ％A ％B Curve Initial 0.400 95.0 5.0 Initial 0.50 0.400 95.0 5.0 6 1.20 0.400 70.0 30.0 6 2.50 0.400 50.0 50.0 6 3.00 0.400 22.0 78.0 6 3.80 0.400 5.0 95.0 6 4.00 0.600 0.0 100.0 1 4.50 0.600 0.0 100.0 6 4.60 0.400 95.0 5.0 1 5.00 0.400 95.0 5.0 1

[0045] Based on the conventional operation process of the carbonyl-derived metabolomics kit and the development of the determination and quantification methods for the above indicators, the present invention separately formulates 6 metabolite indicators and develops a kit for quantitative detection of metabolite for predicting the risk of preeclampsia, which includes two major categories: a sample releasing agent and calibrators and quality control products as shown in Table 4. Among them, the sample releasing agent specifically includes a releasing agent, reagents A, B, C, and D, and a diluent, and the calibrators and quality control products specifically include calibrators, calibrator reconstitution solutions, diluents, and each quality control product. Referring to the methods and requirements in "Development and Validation of Clinical Detection Methods for Liquid Chromatography-Tandem Mass Spectrometry", the present invention evaluates the specificity, matrix effect, carryover, detection limit and lower limit of quantification, linear range, precision, trueness, stability, etc. of the detection method of this kit, which can ensure the accurate and effective quantification of the indicators to be detected by the detection method.

[0046] Table 4 Configuration Information of the Kit

[0047]

[0048] Furthermore, based on the obtained metabolite concentration data of the above pregnant women, the present invention corrects the median of the metabolite concentration (MoM value) of the above metabolites, and combines the output data with maternal factors for statistical analysis, and establishes a high-risk prediction model for early-onset and late-onset preeclampsia in the first trimester of pregnancy.

[0049] Regarding the basic classification of preeclampsia, it includes early-onset preeclampsia (EPE, early-onset preeclampsia) and late-onset preeclampsia (LPE, late-onset preeclampsia). The present invention discovers through experiments that there are also different correlations between the above metabolites and different types of preeclampsia. Therefore, in model construction, the corresponding models can be constructed more precisely according to different types of preeclampsia.

[0050] In a specific embodiment of the present invention, the differences in metabolite data obtained through independent analysis among early-onset preeclampsia, late-onset preeclampsia, and healthy control groups in different hospitals are considered. Based on the same significant differential expression trends (P value < 0.05) presented by at least two or more hospitals, and the data of the third hospital showing at least the same differential expression trend as the screening criteria. Among them, the metabolite data is obtained by correcting the detected metabolite concentration obtained above with the median concentration of the metabolite. Specifically, the MoM correction of the concentration of a certain metabolite marker for each pregnant woman is obtained by dividing the original value of the concentration of this metabolite marker by the median concentration of this metabolite in the samples run on the same batch of machines, and then the median multiple MoM value can be calculated.

[0051] In a specific embodiment of the present invention, after screening and verification, the metabolite markers preferably related to early-onset preeclampsia include: homovanillic acid, sarcosine, ornithine, and hydroxyphenyllactic acid. The metabolite markers preferably related to late-onset preeclampsia include: isovelarylcarnitine, adenosine monophosphate, and ornithine. Specifically, the product of the data obtained by correcting the median of the concentration values of each metabolite is used as the input of the prediction model.

[0052] In addition to the above metabolites, there is also a certain correlation between the clinical indicators of pregnant women and preeclampsia, such as age, height at 12 - 17 +6 weeks of gestation, weight at 12 - 17 +6 weeks of gestation, systolic / diastolic blood pressure at 12 - 17 +6 weeks of gestation, parity, number of deliveries, history of adverse pregnancy, history of adverse past events (including previous gestational diabetes, preeclampsia history, family preeclampsia history, chronic hypertension history, systemic lupus erythematosus, and antiphospholipid syndrome), assisted reproduction, and other clinical data. In a specific embodiment of the present invention, through the analysis of the collected clinical indicators of pregnant women, based on the differences in the clinical information of pregnant women with preeclampsia and healthy control pregnant women, after screening and verification, the following specific data forms of the above clinical indicators of pregnant women are used as the input of the prediction model, including: age, BMI, mean arterial pressure, primipara, number of abortions or inductions more than 2 times, presence or absence of adverse past events, and presence or absence of in vitro assisted reproduction IVF. Among them, age, BMI, and mean arterial pressure all need to be further corrected by the median (i.e., their respective MoM values).

[0053] It should be particularly noted that the above BMI is calculated from the height at 12 - 17 +6 weeks of gestation and the weight at 12 - 17 +6The weekly weight is further calculated, where BMI = weight ÷ height 2 . The MoM values of age and BMI are calculated with reference to formula (Formula 1):

[0054]

[0055] where Clin is the original value of the pregnant woman's age or BMI, and P(Median) Clin is the median of the population age or BMI (calculated using the control pregnant woman data in the training set in this application), and is assigned 29 and 20.69 respectively, and MoM Clin is the age or BMI MoM value calibrated by the median.

[0056] The systolic / diastolic blood pressure at 12-17 +6 weeks of pregnancy is converted to the mean arterial pressure and calculated by the following formula 2.

[0057]

[0058] In formula 2, MAP is the mean arterial pressure, DBP is the diastolic blood pressure at 12-17 +6 weeks of pregnancy, and PP is the pulse pressure (obtained from the difference between the systolic and diastolic blood pressures at 12-17 +6 weeks of pregnancy). The MAP calibrated by the population MAP median is calculated by the following formula (Formula 3):

[0059]

[0060] where MAP is the original value of the mean arterial pressure measured by the pregnant woman, and P(Median) MAP is the population MAP median (calculated using the control pregnant woman data in the training set in this application). In a specific embodiment of the present invention, the population MAP median is assigned 82.98, and MoM MAP is the MAP MoM value calibrated by the median.

[0061] The data forms of parity, gravidity, and history of adverse pregnancy are as follows: the value for a primipara is 1 or 0, representing a primipara or not respectively; the value for more than 2 abortions or inductions is 1 or 0, representing more than 2 abortions or inductions or not respectively.

[0062] The data form of the history of adverse past events is as follows: the value for having or not having a history of adverse past events is 1 or 0, where 1 represents that the pregnant woman has had one or more of gestational diabetes, preeclampsia, familial preeclampsia, chronic hypertension, systemic lupus erythematosus, and antiphospholipid syndrome diseases in the past, and 0 represents no history of adverse past events.

[0063] The data form of assisted reproduction is as follows: the value of in vitro assisted reproduction (IVF) is 1 or 0, representing that the pregnant woman has undergone in vitro fertilization or not, respectively.

[0064] Those skilled in the art should understand that the above content makes an exemplary description of the screening of clinical indicators and metabolites of pregnant women. Using the data of the clinical indicators and metabolites of pregnant women obtained above as input, and whether the pregnant woman has preeclampsia as output, a model is constructed through machine learning methods such as a random forest model to obtain a preeclampsia risk assessment prediction model. Specifically, in a specific embodiment of the present invention, for the prediction model of early-onset preeclampsia, the product of the median-corrected data (MoM value) of the concentrations of 4 metabolites (homovanillic acid, sarcosine, ornithine, and hydroxyphenyllactic acid) and 7 clinical indicators (age MoM value, BMI MoM value, mean arterial pressure MoM value, primipara, number of abortions or induced labors more than 2 times, presence or absence of adverse medical history, and presence or absence of in vitro assisted reproduction (IVF)) are used as input, and whether there is early-onset preeclampsia is used as output, and a prediction model is constructed through a random forest model. For the prediction model of late-onset preeclampsia, the product of the median-corrected data (MoM value) of the concentrations of 3 metabolites (isovelarylcarnitine, adenosine monophosphate, and ornithine) and 7 clinical indicators (age MoM value, BMI MoM value, mean arterial pressure MoM value, primipara, number of abortions or induced labors more than 2 times, presence or absence of adverse medical history, and presence or absence of in vitro assisted reproduction (IVF)) are used as input, and whether there is late-onset preeclampsia is used as output, and a prediction model is constructed through a random forest model. For whether there is early-onset or late-onset preeclampsia, if yes (affected), the output during model construction is set to 1; if no (normal and not affected), the output during model construction is set to 0. After constructing the model using the training set, the validation set is then used to validate the prediction model. At the same time, after obtaining the prediction model, the threshold for judging the preeclampsia risk can be set according to the model evaluation index. This threshold is used as the judgment criterion when using the prediction model for prediction, that is, for the prediction sample, its early-onset preeclampsia risk value calculated by the constructed prediction model is compared with the set early-onset model threshold, and the late-onset preeclampsia risk value is compared with the set late-onset model threshold. Higher than the threshold is judged as high risk, and lower than the threshold is judged as low risk. In a specific embodiment of the present application, the setting of the threshold can be the thresholds under different specificities of the model - for example, specificities of 80%, 85%, 90%, 95%, 96%, 98%.

[0065] It should be noted that by classifying the data, the above two types of prediction models can also be integrated together.

[0066] It should be understood that after the model is constructed by the above method, when performing the prediction of preeclampsia risk assessment, the above two prediction models can be used independently, that is, independently predict early-onset preeclampsia and late-onset preeclampsia for the same pregnant woman respectively. At this time, the data input of the relevant pregnant women can be independent, that is, the data of different metabolites are input for different types of prediction models respectively. The two models can also be combined into one, and after uniformly inputting the metabolite data of the pregnant woman, the model respectively predicts early-onset preeclampsia and late-onset preeclampsia after data recognition and classification processing.

[0067] In addition, the clinical indicators and metabolite data of the above-mentioned pregnant women to be tested are input into the constructed prediction model to predict whether preeclampsia occurs in this sample. The output result of the model is the risk value of early-onset and late-onset preeclampsia. According to the defined threshold, the risk of early-onset and late-onset preeclampsia in the pregnant woman can be determined. In an embodiment of the present invention, the threshold for early-onset preeclampsia is 0.33, and the threshold for late-onset preeclampsia is 0.42. When inputting the relevant data of the pregnant woman to be tested, if the risk value output by the established random forest model is higher than the set threshold, it can be determined that the pregnant woman has a high risk of early-onset or late-onset preeclampsia.

[0068] The present invention will be further described in detail below through specific embodiments in conjunction with the accompanying drawings.

[0069] Example:

[0070] The equipment adopted in the present invention is a conventional equipment in the technical field, and all reagents involved in the invention are conventional products that can be commercially purchased through regular channels.

[0071] The present invention discloses a technical method for predicting the risk of preeclampsia by combining metabolic marker indicators and pregnant women's clinical indicators in the first trimester of pregnancy. The specific steps are as follows:

[0072] 1. Sample collection

[0073] 1.1 Inclusion criteria

[0074] Inclusion criteria for preeclampsia: Singleton pregnant women, after 20 weeks of pregnancy, the pregnant woman has systolic blood pressure ≥ 140 mmHg and / or diastolic blood pressure ≥ 90 mmHg, accompanied by any one of the following: urinary protein quantification ≥ 0.3 g / 24 h, or urinary protein / creatinine ratio ≥ 0.3, or random urinary protein ≥ (+) (examination method when protein quantification is not possible); no proteinuria but accompanied by any one of the following organ or system involvements: important organs such as heart, lung, liver, kidney, or abnormal changes in the hematological system, digestive system, nervous system, and involvement of the placenta-fetus, etc.

[0075] Inclusion criteria for early-onset preeclampsia: Abnormal blood pressure in preeclampsia or abnormal 24-hour proteinuria confirmed at ≤ 34 weeks of gestation.

[0076] Inclusion criteria for late-onset preeclampsia: Abnormal blood pressure in preeclamptic women or abnormal 24-hour proteinuria confirmed > 34 weeks of gestation.

[0077] Inclusion criteria for healthy controls: The samples were from term singleton pregnancies without pregnancy complications, with the fetus growing well at birth and no obstetric, medical, or surgical complications during pregnancy. Exclusion criteria: ① Concurrent other pregnancy complications; ② Severe heart, liver, and kidney insufficiency; ③ Patients with autoimmune diseases or malignant tumor diseases. Exclude abnormal pregnant women caused by chromosomal, congenital abnormalities, preterm birth, and multiple pregnancies.

[0078] 1.2 Participants:

[0079] Residual plasma samples of singleton pregnant women who underwent NIPT screening at 12 - 17 +6 weeks were recruited through three hospitals a, b, and c, with a total of about 800 cases. At the same time, the corresponding sample clinical information was retrieved, such as maternal age, height, weight at the time of NIPT, systolic / diastolic blood pressure at the time of NIPT, gravidity, parity, history of adverse pregnancy, history of adverse past events (including previous gestational diabetes, preeclampsia history, family history of preeclampsia, chronic hypertension history, systemic lupus erythematosus, and antiphospholipid syndrome), and clinical data such as assisted reproduction.

[0080] According to the above inclusion criteria, about 90 cases of early-onset preeclampsia, 153 cases of late-onset preeclampsia, and 553 healthy controls were recruited from the three hospitals as the training set for metabolite biomarker discovery and model training (Table 5). The pilot samples were from hospitals a and b, including 21 cases of early-onset preeclampsia, 106 cases of late-onset preeclampsia, and 176 healthy controls, for metabolite biomarker verification and model validation (Table 5). The mid-scale samples were from hospitals d, e, and f, including 151 cases of early-onset preeclampsia, 217 cases of late-onset preeclampsia, and 2950 healthy controls, as the test set for independent testing of the metabolic model (Table 5).

[0081] Table 5. Introduction to sample sources

[0082]

[0083] EPE: Early-onset preeclampsia, LPE: Late-onset preeclampsia, Control: Healthy control

[0084] 2. Discovery and verification of metabolite biomarkers

[0085] 2.1 Discovery of metabolite biomarkers in the training set

[0086] Independently analyze and compare the targeted metabolite data of the training set from three hospitals through the Wilcoxon rank sum test (the detection method of metabolites was carried out according to the aforementioned kits and methods, which will not be repeated here). The differences among early-onset preeclampsia, late-onset preeclampsia, and the control group were analyzed. According to the same significant differential expression trends presented by at least two or more hospitals (P value < 0.05), and the data of the third hospital showing at least the same differential expression trend as the screening criteria for markers, the present invention preliminarily determined potential preeclampsia metabolite markers (such as Figure 2 , 3).

[0087] 2.2 Verification of relevant indicators in the validation set and test set:

[0088] To further confirm the predictive value of the metabolite markers, this example further enrolled a total of about 300 samples from hospitals a and b. According to the sample information and inclusion criteria, this batch of pilot samples included 21 cases of early-onset preeclampsia, 106 cases of late-onset preeclampsia, and 176 healthy controls. The differences in the clinical information of pregnant women with preeclampsia and healthy controls in the pilot samples were compared, and the differences in clinical information such as advanced age, higher BMI, higher mean arterial pressure, primipara, more than 2 abortions or inductions, adverse past history, and in vitro assisted reproduction IVF in preeclampsia and healthy control samples were verified. These factors all increased the risk of preeclampsia in pregnant women, showing a stable predictive value for preeclampsia (Table 6, Figure 4 ).

[0089] Table 6. Statistical analysis of clinical information of pilot samples

[0090]

[0091]

[0092] *, **, and *** represent P values lower than 0.05, 0.01, and 0.001 respectively. For continuous variables: Wilcoxon rank sum test; for categorical variables: fisher’s exact test. Adverse past history includes previous gestational diabetes, preeclampsia history, family preeclampsia history, chronic hypertension history, systemic lupus erythematosus, and antiphospholipid syndrome

[0093] For further evaluation, this example enrolled a total of about 3300 samples from another three hospitals (hospitals d, e, and f). By collecting the corresponding sample clinical information. According to the inclusion criteria for early-onset preeclampsia, late-onset preeclampsia, and healthy controls, the pilot samples consisted of 151 cases of early-onset preeclampsia, 217 cases of late-onset preeclampsia, and 2950 healthy controls to form the test set. The differences in the clinical information of pregnant women with preeclampsia and healthy controls were verified as above. The results showed that most clinical information significantly increased the risk of disease occurrence, showing a stable predictive value (Table 7, Figure 5 ).

[0094] Table 7. Statistical Clinical Information of Pilot Samples

[0095]

[0096] *, **, and *** represent P values less than 0.05, 0.01, and 0.001, respectively. For continuous variables: Wilcoxon rank sum test; for categorical variables: Fisher's exact test. Adverse previous history includes previous gestational diabetes, preeclampsia history, family preeclampsia history, chronic hypertension history, systemic lupus erythematosus, and antiphospholipid syndrome.

[0097] By analyzing the expression of the aforementioned metabolic markers in the validation set and test set samples, the examples further determined that there were also significant differential expressions of 6 metabolite markers in the new samples, namely, the metabolite markers for early-onset preeclampsia: homovanillic acid, sarcosine, ornithine, and hydroxyphenyllactic acid; the metabolite markers for late-onset preeclampsia: isovelarylcarnitine, adenosine monophosphate, and ornithine. The product of the MoM values of these metabolic markers and the MoM values of the metabolites were all significantly upregulated in pregnant women with preeclampsia, showing a stable value for predicting preeclampsia ( Figure 6 , 7).

[0098] 3. Establishment and Validation of the Preeclampsia Risk Prediction Model

[0099] 3.1 Model Construction

[0100] To evaluate the effectiveness of the above-screened metabolic and clinical prediction markers in predicting preeclampsia risk, this study used the samples from three hospitals used in the discovery stage as the training set, and the product of the median-corrected data (MoM values) of the concentrations of 4 metabolite markers (median-corrected data of homovanillic acid concentration * median-corrected data of sarcosine concentration * median-corrected data of ornithine concentration * median-corrected data of hydroxyphenyllactic acid concentration) and the data of 7 clinical indicators (age MoM value, BMI MoM value, mean arterial pressure MoM value, primipara, number of abortions or induced labors more than 2 times, presence or absence of adverse previous history, and presence or absence of in vitro assisted reproduction IVF) as the input features of the prediction model. A prediction model was established using the random forest model, and the threshold was determined using the validation set (pilot test), and independent validation was performed using the test set (pilot scale).

[0101] Through model training and preliminary verification using the training set data, the results show that for the product of the combined clinical information and the median-corrected concentration values of 4 metabolite markers for early-onset preeclampsia, the AUC of the risk prediction model for early-onset preeclampsia in the training set can reach 0.98 (as shown in Figure 8 early-onset preeclampsia train in Figure 8 ); for the product of the combined clinical information and the median-corrected data (MoM values) of the concentration values of 3 metabolite markers for late-onset preeclampsia (the product of the median-corrected concentration value of 3-methylbutyrylcarnitine * the median-corrected concentration value of adenosine monophosphate * the median-corrected concentration value of ornithine), the AUC of the risk prediction model for late-onset preeclampsia in the training set reaches 0.99 (as shown in Figure 8 late-onset preeclampsia train in

[0102] 3.2 Validation set sample validation

[0103] The prediction effect of the above preeclampsia risk assessment model was evaluated using the pilot samples and the threshold was determined. The results show that: in the pilot samples, the above two models for early-onset and late-onset preeclampsia also achieved high prediction accuracy. Among them, the AUC of the prediction model for early-onset preeclampsia in the validation set is 0.91 (as shown in Figure 8 early-onset preeclampsia vali in Figure 8 ). When the threshold is set to 0.33, the sensitivity and specificity of the model are 66.67% and 96.59% respectively (Table 8); the AUC of the prediction model for late-onset preeclampsia in the validation set is 0.82 (as shown in Figure 8 late-onset preeclampsia vali in

[0104] Table 8. Prediction results of early-onset preeclampsia in pilot samples

[0105]

[0106] a. Positive and negative predictive values calculated based on 21 cases of early-onset preeclampsia and 176 control samples;

[0107] b. Positive and negative predictive values of the existing prediction models for early-onset preeclampsia estimated based on an incidence rate of 1.2%.

[0108] Table 9. Prediction results of late-onset preeclampsia in pilot samples

[0109]

[0110]

[0111] a. Positive and negative predictive values calculated based on 106 cases of late-onset preeclampsia and 176 control samples;

[0112] b. Estimate the positive predictive value and negative predictive value of the existing prediction model for late-onset preeclampsia at an incidence rate of 2.8%.

[0113] 3.3 Validation of the test set samples

[0114] Finally, the prediction effects of the above-mentioned early-onset preeclampsia and late-onset preeclampsia models were evaluated using the pilot samples. The results showed that the above models also achieved high accuracy in the pilot samples. Among them, the AUC of the early-onset preeclampsia prediction model was 0.89 (as shown in the early-onset preeclampsia test in Figure 8 ), and when the threshold was set to 0.33, the sensitivity and specificity of the model were 63.58% and 90% respectively (Table 10); the AUC of the late-onset preeclampsia prediction model was 0.82 (as shown in the late-onset preeclampsia test in Figure 8 ), and when the threshold was set to 0.42, the sensitivity and specificity of the model were 53% and 90% respectively (Table 11).

[0115] Table 10. Prediction results of early-onset preeclampsia in the pilot samples

[0116]

[0117] a. Positive predictive value and negative predictive value calculated based on 151 cases of early-onset preeclampsia and 2950 control samples;

[0118] b. Estimate the positive predictive value and negative predictive value of the existing prediction model for early-onset preeclampsia at an incidence rate of 1.2%.

[0119] Table 11. Prediction results of late-onset preeclampsia in the pilot samples

[0120]

[0121] a. Positive predictive value and negative predictive value calculated based on 217 cases of late-onset preeclampsia and 2950 control samples;

[0122] b. Estimate the positive predictive value and negative predictive value of the existing prediction model for late-onset preeclampsia at an incidence rate of 2.8%.

[0123] The above uses specific examples to elaborate on the present invention, which is only used to help understand the present invention and is not intended to limit the present invention. For those skilled in the technical field to which the present invention pertains, based on the idea of the present invention, several simple deductions, deformations, or substitutions can also be made.

Claims

1. A method for constructing a prediction model for preeclampsia risk assessment based on metabolomics, characterized in that using the clinical indicators and metabolite data of pregnant women as input and whether the pregnant woman has preeclampsia as output, a prediction model for preeclampsia risk assessment is constructed by machine learning method; the metabolites are selected from at least one of 4-hydroxyphenyllactic acid, homovanillic acid, ornithine, sarcosine, adenosine monophosphate, and 3-methylbutyrylcarnitine.

2. The construction method according to claim 1, characterized in that The clinical indicators include the age of the pregnant woman, the height of the pregnant woman at 12-17 +6 weeks of gestation, the weight of the pregnant woman at 12-17 +6 weeks of gestation, the systolic blood pressure / diastolic blood pressure measured at 12-17 +6 weeks of gestation, several items among the clinical data of the number of pregnancies, the number of deliveries, history of adverse pregnancy, history of adverse past medical conditions, and assisted reproduction; the history of adverse past medical conditions includes at least one of a history of previous gestational diabetes, preeclampsia, familial preeclampsia, chronic hypertension, systemic lupus erythematosus, and antiphospholipid syndrome; preferably, for the risk assessment and prediction of early-onset preeclampsia, the metabolites are 4-hydroxyphenyllactic acid, homovanillic acid, ornithine, and sarcosine; for the risk assessment and prediction of late-onset preeclampsia, the metabolites are 3-methylbutyrylcarnitine, adenosine monophosphate, and ornithine.

3. The construction method according to claim 1 or 2, characterized in that the machine learning method is a random forest model.

4. The construction method according to claims 1-3, characterized in that after obtaining the prediction model, set the threshold according to the model evaluation index of the prediction model; preferably, the model evaluation index is the specificity of the prediction model; preferably, the threshold is set when the specificity of the prediction model is 80%-98%; preferably, the threshold is set based on the specificity of the prediction model being 90%.

5. Use of metabolites related to preeclampsia in the preparation of detection products related to preeclampsia, wherein the metabolites are selected from at least one of 4-hydroxyphenyllactic acid, homovanillic acid, ornithine, sarcosine, adenosine monophosphate, and 3-methylbutyrylcarnitine.

6. A detection kit for metabolites related to preeclampsia, characterized in that the metabolites are selected from at least one of 4-hydroxyphenyllactic acid, homovanillic acid, ornithine, sarcosine, adenosine monophosphate, and 3-methylbutyrylcarnitine.

7. A method for predicting preeclampsia risk assessment based on metabolomics, characterized in that comprising: obtaining the clinical indicators and metabolite data of the pregnant woman to be tested; inputting the obtained clinical indicators and metabolite data of the pregnant woman into the prediction model for processing to obtain the risk value of preeclampsia of the pregnant woman, and when the risk value is higher than the set threshold, it is determined that the pregnant woman has a high risk of preeclampsia; the metabolites are selected from at least one of 4-hydroxyphenyllactic acid, homovanillic acid, ornithine, sarcosine, adenosine monophosphate, and 3-methylbutyrylcarnitine; The clinical indicators include the age of the pregnant woman to be tested, the height of the pregnant woman at 12-17 +6 weeks of gestation, the weight of the pregnant woman at 12-17 +6 weeks of gestation, the systolic blood pressure / diastolic blood pressure at the time of examination at 12-17 +6 weeks of gestation, the number of pregnancies, the number of deliveries, the history of adverse pregnancy, the history of adverse past events, and several items of clinical data on assisted reproduction; the history of adverse past events includes at least one of a history of previous gestational diabetes, a history of preeclampsia, a family history of preeclampsia, a history of chronic hypertension, systemic lupus erythematosus, and antiphospholipid syndrome; the prediction model is obtained by the prediction model construction method according to any one of claims 1-4.

8. The preeclampsia risk assessment prediction method according to claim 6, characterized in that the metabolites are 4-hydroxyphenyllactic acid, homovanillic acid, ornithine, sarcosine, adenosine monophosphate, and 3-methylbutyrylcarnitine; The clinical indicators are the age of the pregnant woman, the height of the pregnant woman at 12-17 +6 weeks of gestation, the weight of the pregnant woman at 12-17 +6 weeks of gestation, the systolic blood pressure / diastolic blood pressure measured at 12-17 +6 weeks of gestation, the number of pregnancies, the number of deliveries, the clinical data of adverse pregnancy history, adverse past history, and assisted reproduction; the adverse past history includes a history of previous gestational diabetes, preeclampsia, familial preeclampsia, chronic hypertension, systemic lupus erythematosus, and antiphospholipid syndrome.

9. A prediction system for preeclampsia risk assessment based on metabolomics, characterized in that comprising: a device for obtaining the clinical indicators and metabolite data of the pregnant woman to be tested; a device for processing the clinical indicators and metabolite data of the pregnant woman by the prediction model; a device for outputting the prediction result; The prediction model is obtained by the prediction model construction method described in any one of claims 1-4.

10. A prediction product for preeclampsia risk assessment based on metabolomics, characterized in that it includes: a memory for storing programs; a processor for implementing the method described in claim 7 or 8 by executing the programs stored in the memory.

11. A computer-readable storage medium, characterized in that a program is stored on the medium, and the program can be executed by a processor to implement the method described in claim 7 or 8.

12. A computer-readable storage medium, characterized in that a prediction model obtained by the construction method described in any one of claims 1-4 is stored on the medium.

Images