Use of biomarkers in the manufacture of a diagnostic reagent for gestational diabetes mellitus

Abstract

The present application relates to the field of medical diagnosis, and in particular to screening biomarkers for differential diagnosis of gestational diabetes mellitus by using metabolomics, and establishing a model for diagnosing whether a pregnant individual is diabetic based on the biomarkers. The present application provides biomarkers and a diagnostic model for differential diagnosis of gestational diabetes mellitus, which can be applied to early diagnosis or prediction of gestational diabetes mellitus, and is of great significance for prevention or treatment of gestational diabetes mellitus.

Description

[0001] Priority

[0002] The present application claims the priority of two Chinese prior applications, which are

Application No. 202210083776.9, filing date: 2022.01.24

Application No. 202210083970.7, filing date: 2022.01.24

[0003] The present application relates to the field of medical diagnosis, in particular, to the use of metabolomics to screen biomarkers for diabetes and use in the diagnosis of diabetes, especially the diagnosis of gestational diabetes mellitus and a system for diagnosing whether a pregnant individual has diabetes. BACKGROUND

[0004] Metabolomics is a discipline that qualitatively and quantitatively analyzes small molecular metabolites in the body with a relative molecular weight less than 1000. Metabolomics analysis can reflect the physiological and pathological conditions of the body and can also distinguish differences between different individuals. With the development of mass spectrometry technology, liquid chromatography-mass spectrometry (LC-MS) has become the most important research tool in metabolomics research. Currently, metabolomics has been widely used in the field of clinical diagnosis, mainly to find metabolic markers related to disease diagnosis and treatment.

[0005] Gestational diabetes mellitus (GDM) is the most common metabolic abnormality during pregnancy, which significantly increases the risk of premature birth, fetal growth restriction, fetal malformation, and maternal postpartum development of type 2 diabetes, etc. From the perspective of birth defect prevention and control, maternal GDM as a direct "unfavorable environmental factor" hinders the normal development of the fetus and can lead to embryonic-derived adult diseases. Therefore, early diagnosis of GDM is very important and is an important basis for timely clinical intervention, prevention of birth defects, and adverse maternal-fetal outcomes. The traditional GDM diagnosis is based on the oral 75g glucose tolerance test at 24-28 weeks of gestation, which is a late opportunity, and the patient's compliance is poor. High blood sugar and related metabolic disorders have already had adverse effects on the mother and child, which is not conducive to improving the maternal and infant outcomes. Since the pathogenesis of GDM is still unclear, there is still a lack of markers and early detection methods that accurately reflect the phenotype of GDM.

[0006] Therefore, it is necessary to provide a system for predicting whether a pregnant individual has a risk of diabetes at an earlier stage and to use new markers for prediction. SUMMARY

[0007] The application collects serum samples of normal pregnant women and pregnant women with gestational diabetes mellitus, uses liquid chromatography-high resolution mass spectrometry (LC-HRMS) technology to perform metabolomics analysis on the above samples, and screens new differential metabolites between normal pregnant women and pregnant women with gestational diabetes mellitus through statistical analysis, and further establishes a model for differential diagnosis of gestational diabetes mellitus. The identification here means recognition and distinction, that is, to distinguish the system of healthy normal and gestational diabetes mellitus. The system includes the model company to automatically calculate the output result. At the same time, the application also uses the markers to establish a system for predicting the blood glucose value of a pregnant person 1 hour and 2 hours after fasting. The system also includes a mathematical model, so that the value can be used to judge whether the pregnant person is diabetic.

[0008] The purpose of the application is to diagnose gestational diabetes mellitus, including a method for finding new biomarkers of gestational diabetes mellitus, finding biomarkers and models for the differential diagnosis of gestational diabetes mellitus, and a method for early diagnosis of gestational diabetes mellitus.

[0009] In the first aspect of the application, a method for screening biomarkers of gestational diabetes mellitus based on serum metabolomics is provided, and the specific steps are as follows:

[0010] (1) Collect serum samples of pregnant women with gestational diabetes mellitus and normal pregnant women;

[0011] (2) Extract serum metabolites;

[0012] (3) Detect and data preprocess the extracted serum metabolites by liquid chromatography-mass spectrometry;

[0013] (4) Use partial least squares discriminant analysis to group the samples, and combine significance analysis to screen differential metabolites or differential biomarkers of different groups;

[0014] (5) According to the screened differential metabolites, mine biomarkers that can be used to diagnose gestational diabetes mellitus and the application of the markers, for example: how to use the markers to diagnose or predict gestational diabetes mellitus patients, or to diagnose gestational diabetes mellitus patients from healthy people.

[0015] In some modes, the step (1) is specifically implemented as follows: the serum samples are from normal pregnant women and pregnant women with gestational diabetes mellitus. The normal pregnant women and pregnant women with gestational diabetes mellitus mentioned here are those who have been diagnosed and confirmed, such as normal pregnant women and pregnant women with gestational diabetes mellitus confirmed by blood glucose detection and glucose tolerance test.

[0016] In some embodiments, the step (2) is specifically implemented as follows: adding the methanol precipitant containing multiple isotopic internal standards to the serum sample at a ratio of 1:4, mixing for 3 minutes, and then centrifuging at 4000xg for 10 minutes at 20°C. Four 100 μL aliquots of the supernatant are taken from each sample and transferred to four sample plates, dried by nitrogen blowing, and then multiple isotopic internal standard-containing resuspension solutions are added for subsequent UPLC-MS / MS detection.

[0017] The step (3) is specifically implemented as follows: extracting m / z ions and retention time from the original mass spectrum data, identifying metabolites by database retrieval after correction of the retention time, checking the peak area obtained by metabolite chromatographic peak integration, and performing data normalization and missing value filling to obtain a data matrix for subsequent bioinformatics analysis.

[0018] The step (4) is specifically implemented as follows: performing data filtering on the data matrix, and using orthogonal partial least squares discriminant analysis on the remaining data to cluster the samples. The normal pregnant women and the pregnant women with gestational diabetes can be clearly clustered.

[0019] In some embodiments, the step (5) is specifically implemented as follows: screening compounds with an FDR value less than 0.05 and a VIP greater than 1 as differential metabolites based on biological significance, mining biomarkers for pregnant women with gestational diabetes, and performing metabolic pathway analysis.

[0020] Further, after screening the biomarkers for gestational diabetes, one or more biomarkers are selected to establish a model for the differential diagnosis of gestational diabetes.

[0021] Further, a non-linear model is established for predicting the blood glucose value of a pregnant individual 1 hour or 2 hours after fasting, and the model is optimized. The predicted value of the model is compared with the measured value to verify the accuracy of the model diagnosis, and a more optimal diagnostic model is finally obtained.

[0022] Therefore, in a second aspect, the present application provides the use of biomarkers in a detection reagent for diagnosing whether a pregnant individual has diabetes, and the markers are selected from one or more of the following:

[0023] R-3-hydroxybutyrylcarnitine, 1,5-anhydroglucitol, 1-arachidonoylglycerophosphocholine, 1-arachidonoylglycerophosphoinositol, 1-linoleoylglycerophosphocholine, 1-palmitoylglycerophosphatidic acid, 1-palmitoylglycerophosphocholine, 2- aminoadipic acid, 2-hydroxybutyric acid, 3-(4-hydroxyphenyl)lactic acid, 3-hydroxybutyric acid, 3-methyl-2-oxobutyric acid, 3-methyl-2-oxovaleric acid, 4-methyl-2- oxovaleric acid, 8-methoxykynurenine, carnitine, cis-3,4-methyleneglutaryl carnitine, cystathionine, cysteine disulfide, deoxycholic acid, gamma-glutamyl-epsilon-lysine, glucose, glycerophosphoinositol, glycine, glycocholic acid sulfate, glycohyocholic acid sulfate, histidylalanine, indoleacetic acid, isoleucine, isohomocholic acid sulfate, isovaleric acid, sulfided di-alanine, leucine, N6-acetyllysine, N-acetyltaurine, N-acety tryptophan, N-acety valine, oleic acid, orotidine, oxalic acid, palmitoylcarnitine, pantothenic acid, phenylalanine, pyroglutamine, serine, threonic acid, tyrosine, valine.

[0024] Preferably, the gestational diabetes, especially early-mid gestational diabetes, the early-mid being within 28 weeks of gestation, such as within 1-20 weeks of conception, such as within 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 12, or 14 weeks of conception.

[0025] In some aspects, the biomarker is selected from one or more of:

[0026] R-3-hydroxybutyrylcarnitine, 1-arachidonoylglycerophosphocholine, 1- arachidonoylglycerophosphoinositol, 1-linoleoylglycerophosphocholine, 1- palmitoylglycerophosphatidic acid, 1-palmitoylglycerophosphocholine, 3-(4- hydroxyphenyl)lactic acid, 3-methyl-2-oxobutyric acid, 8-methoxykynurenine, cis-3,4-methyleneglutaryl carnitine, cysteine disulfide, gamma-glutamyl-epsilon- lysine, glycerophosphoinositol, histidylalanine, indoleacetic acid, isohomocholic acid sulfate, N6-acetyllysine, N-acetyltaurine, N-acetyl tryptophan, N-acety valine, orotidine, oxalic acid, pantothenic acid.

[0027] In some aspects, the biomarker is selected from one or more of:

[0028] 1-arachidonoylglycerophosphoinositol, 1-palmitoylglycerophosphatidic acid, 8- methoxykynurenine, cis-3,4-methyleneglutaryl carnitine, cysteine disulfide, gamma-glutamyl-epsilon-lysine, glycerophosphoinositol, histidylalanine, isohomocholic acid sulfate, N6-acetyllysine, N-acetyltaurine, N-acetyl tryptophan, N-acety valine, orotidine, oxalic acid.

[0029] In some embodiments, the biomarkers are selected from one or more of the following:

[0030] 8-methoxykynurenine, cysteine disulfide glycine, histidylalanine, isoursodeoxycholate sulfate.

[0031] In a third aspect of the present application, a classification model for diagnosing whether a pregnant individual has gestational diabetes mellitus is established. The classification model can distinguish between gestational diabetes mellitus and non-gestational diabetes mellitus, and can classify a plurality of pregnant animals or pregnant women into a gestational diabetes mellitus group and a non-gestational diabetes mellitus group, or identify pregnant animals or pregnant women with gestational diabetes mellitus from a plurality of pregnant animals or pregnant women. For example, when there are 300 pregnant women, of which there may be about 20 patients with gestational diabetes mellitus, and it is necessary to identify the patients with gestational diabetes mellitus or to classify the 300 pregnant women into a gestational diabetes mellitus group and a non-gestational diabetes mellitus group, the classification model can be used to achieve this.

[0032] In some embodiments, the classification model is one of Model A, Model B, Model C, Model D, Model E and Model F, and the equations of the classification models are as follows:

[0033] Model A: Score = 3.425*glucose + 0.854*palmitoylcarnitine - 4.598*oleic acid - 1.307*glycine + 0.309*phenylalanine - 2.253*serine - 0.335*tyrosine - 0.172*isoleucine - 1.273*leucine - 0.422*valine + 0.622*2-oxoglutaric acid - 0.882*1,5-anhydroglucitol + 0.898*3-methyl-2-oxobutanoic acid + 2.292*3-hydroxybutyric acid + 2.919*2-hydroxybutyric acid + 1.319*pantothenic acid - 0.103*3-methyl-2-oxopentanoic acid + 0.856*4-methyl-2-oxopentanoic acid + 1.256*1-palmitoylglycerophosphatidylcholine;

[0034] Model B: Score = 0.199*Deoxycholic acid + 2.903*N-acetylvaline + 4.004*Carnitine - 1.611*Cystathionine + 2.441*Indoleacetic acid - 1.765*Oxalic acid - 1.271*Threonic acid - 0.588*3-(4-hydroxyphenyl)lactic acid + 0.284*Glycocholic sulfate - 0.678*Glycohyocholic sulfate - 4.844*1-arachidonoyl glycerophosphocholine + 4.307*Gamma-glutamyl-epsilon-lysine - 0.769*N-acetyltryptophan - 0.495*1-arachidonoyl glycerophosphoinositol + 2.144*Linoleoyl glycerophosphocholine + 0.635*1-palmitoyl glycerophosphatidic acid + 0.171*Ornithine + 4.154*N6-acetyllysine - 0.039*Thionitrile + 0.893*Histidylalanine + 0.995*R-3-hydroxybutyrylcarnitine + 0.756*Isovalerate - 1.463*Pyroglutamic acid - 0.503*Glycerophosphoinositol - 0.513*N-acetylmethionine + 0.205*Disulfide cysteinylglycine - 1.874*8-methoxykynurenine + 0.361*Isohyodeoxycholic sulfate - 1.421*Cis-3,4-methenylheptanoylcarnitine + 3.425*Glucose + 0.854*Palmitoylcarnitine - 4.598*Oleic acid - 1.307*Glycine + 0.309*Phenylalanine - 2.253*Serine - 0.335*Tyrosine - 0.172*Isoleucine - 1.273*Leucine - 0.422*Valine + 0.622*2-oxoglutarate - 0.822*1,5-anhydroglucitol + 0.898*3-methyl-2-oxobutanoate + 2.292*3-hydroxybutyrate + 2.919*2-hydroxybutyrate + 1.319*pantothenic acid - 0.103*3-methyl-2-oxopentanoate + 0.856*4-methyl-2-oxopentanoate + 1.256*Palmitoylglycerophosphocholine;

[0035] Model C: Score = 0.163*1-palmitoylglycerophosphocholine + 1.775*Palmitoylcarnitine + 0.455*Oleic acid - 0.723*Glycine + 0.203*Phenylalanine + 0.085*Serine - 1.599*Tyrosine - 0.271*Isoleucine - 1.177*Leucine + 0.506*Valine + 1.622*2-oxoglutarate;

[0036] Model D: Score = 1.847*Palmitoylcarnitine + 0.447*Oleic acid - 0.757*Glycine + 0.235*Phenylalanine + 0.057*Serine - 1.606*Tyrosine - 0.285*Isoleucine - 1.103*Leucine + 0.491*Valine + 1.622*2-oxoglutarate;

[0037] Model E: Score = 0.688*oleic acid - 0.78*glycine + 0.484*phenylalanine + 0.146*serine - 0.781*tyrosine + 0.383*isoleucine - 1.431*leucine + 0.303*valine + 1.27*oxalate;

[0038] Model F: Score = -0.0199*deoxycholic acid - 0.290*N-acetyl valine - 0.400*carnitine + 0.161*cystathionine - 0.244*indoleacetic acid + 0.177*oxalic acid + 0.127*threonic acid + 0.00588*3-(4-hydroxyphenyl)lactic acid - 0.0284*glycocholic acid sulfate + 0.0678*glycohyocholic acid sulfate + 0.484*1-arachidonoyl glycerophosphocholine - 0.431*gamma-glutamyl-epsilon-lysine + 0.177*N-acetyl tryptophan + 0.0495*1-arachidonoyl glycerophosphoinositol - 0.214*linoleoyl glycerophosphocholine - 0.0635*1-palmitoyl glycerophosphatidic acid - 0.0171*orotidine + 0.415*N6-acetyl lysine + 0.0039*dipropyl sulfide - 0.0893*homoarginine + 0.0995*R-3-hydroxybutyrylcarnitine - 0.0756*isovalerate + 0.146*pyroglutamic acid + 0.0503*glycerophosphoinositol + 0.0513*N-acetylcysteine - 0.0205*cysteine-glutamylglycine + 0.187*8-methoxykynurenine - 0.0361*isohomocholic acid sulfate + 0.142*cis-3,4-methenylheptanoylcarnitine; wherein the biomarker name represents the relative abundance of the corresponding biomarker in serum.

[0039] In some ways, the critical value of the model A is 0.515, and the relative abundance of the corresponding compound in the serum is input into the model A: when the score (Score) > 0.515, the possibility of diagnosing gestational diabetes is high, and when the score (Score) ≤ 0.515, the possibility of diagnosing gestational diabetes is low or normal pregnant women.

[0040] The critical value of the model B is 5.368, and the relative abundance of the corresponding compound of the marker in the serum is input into the model F: when the score (Score) > 5.368, the possibility of diagnosing gestational diabetes is high, and when the score (Score) ≤ 5.368, the possibility of diagnosing gestational diabetes is low.

[0041] The critical value of the model C is 0.662, and the relative abundance of the corresponding compound of the marker in the serum is input into the model C: when the score (Score) > 0.662, the possibility of diagnosing gestational diabetes is high, and when the score (Score) ≤ 0.662, the possibility of diagnosing gestational diabetes is low.

[0042] The critical value of the model D is 0.661, and the relative abundance of the corresponding compound of the marker in the serum is input into the model D: when the score (Score) > 0.661, the possibility of diagnosing gestational diabetes is high, and when the score (Score) ≤ 0.661, the possibility of diagnosing gestational diabetes is low.

[0043] The critical value of the model E is 0.671, and the relative abundance of the corresponding compound of the marker in the serum is input into the model E: when the score (Score) > 0.671, the possibility of diagnosing gestational diabetes is high, and when the score (Score) ≤ 0.671, the possibility of diagnosing gestational diabetes is low.

[0044] The critical value of the model F is 0.463, and the relative abundance of the corresponding compound of the marker in the serum is input into the model B: when the score (Score) > 0.463, the possibility of diagnosing gestational diabetes is high or can be directly judged as gestational diabetes, and when the score (Score) ≤ 0.463, the possibility of diagnosing gestational diabetes is low or can be directly judged as normal pregnancy.

[0045] The above models are used to verify the accuracy of the differentiation of diabetes in the actual pregnant women population and the correlation of the actual data is 0.950, which indicates that the above models can be used to distinguish between normal and diabetic pregnant women.

[0046] In some modes, the detection step of the relative abundance of the corresponding marker compound in the serum includes:

[0047] Step 1: Collecting serum samples to be tested;

[0048] Step 2: Extracting serum metabolites;

[0049] Step 3: Detecting the serum metabolites of step 2 by liquid chromatography-mass spectrometry and processing data;

[0050] In some modes, the specific operation of extracting serum metabolites in step 2 includes: adding methanol precipitant to the serum sample in a ratio of 1:4, mixing uniformly after oscillation for 3 minutes, and centrifuging at 20℃ 4000×g for 10 minutes. Take 4 portions of 100 μL supernatant from each sample into 4 sample plates, dry by nitrogen blowing, and add resolvent for subsequent detection.

[0051] In some aspects, the reconstitution solution comprises a plurality of isotopic internal standards.

[0052] In some aspects, the liquid chromatography-mass spectrometry detection conditions comprise:

[0053] Detection is performed using UPLC-Q Exactive with a scan range of 70-1000 m / z.

[0054] For detection in positive ion electrospray ionization mode, a C18 column is used for separation, mobile phase A is water containing 0.05% PFPA and 0.1% FA, and mobile phase B is methanol or a mixture of methanol, acetonitrile and water.

[0055] For detection in negative ion electrospray ionization mode, a C18 column is used for separation, mobile phase A is water containing 6.5 mM ammonium bicarbonate, and mobile phase B is methanol (B); or a HILIC column is used for separation, mobile phase A is water containing 10 mM ammonium formate, and mobile phase B is acetonitrile.

[0056] It should be noted that the relative abundance referred to herein is the relative value of the amount of each biomarker, for example, the total amount of all biomarkers of model A in a sample is X, and the amount of glucose is A, and the relative abundance of glucose is the ratio of A to X. The amount can be expressed in terms of concentration, content or weight; or in terms of the strength of ultraviolet absorption, the strength of fluorescence emission, the area of chromatographic peak or peak height. In some aspects, the relative abundance of each biomarker can also be referenced to the amount of an internal standard or a certain compound, and the relative abundance of a certain biomarker is the ratio of the amount of the biomarker to the amount of the internal standard or the compound, in which case the critical value of the model needs to be changed accordingly or the model coefficients of each biomarker in the model need to be changed proportionally. In some aspects, the relative abundance herein can also be the absolute value of the amount of each biomarker, for example, the concentration or content in the sample to be tested, in which case the critical value of the model needs to be changed accordingly or the model coefficients of each biomarker in the model need to be changed proportionally.

[0057] In some embodiments, ROC curves can be established for each biomarker, and those biomarkers with larger areas under the curve can be selected to establish a diagnostic model, or a more reliable diagnostic result. In general, it can be understood that the more biomarkers selected, the more reliable the model established, for example, the higher the accuracy and specificity, and the higher the sensitivity. However, a single or several important biomarkers can also be selected for diagnosis, or for preliminary screening. Such detection methods can be various, for example, using the liquid chromatography-mass spectrometry of the present application for combined detection, high-throughput methods can be used to detect one or more biomarkers of the present application at one time, and of course, detection of a small number of biomarkers is not excluded. Of course, immunological methods can also be used to detect a small number of important biomarkers, for example, combined detection of 1, 2, 3, 4, or 5 biomarkers, which also has diagnostic value, for example, using the amount of glucose in blood as a single marker to measure whether it is possible to have diabetes, which is also a gold standard, but does not mean that glucose is the only marker, and of course other markers can be selected for measurement, for example, the newly discovered markers of the present application to diagnose whether an individual has the possibility of developing gestational diabetes.

[0058] In a fourth aspect of the present application, a prediction model for predicting blood glucose values is provided. The prediction model can be a linear model or a non-linear model, for example, one of random forest regression, polynomial regression, support vector regression, and gradient boosting regression tree, preferably a support vector regression model.

[0059] The prediction model can be established by selecting one or more biomarkers of gestational diabetes, for example, a combination of one or more of the above-mentioned biomarkers, and of course, other reported biomarkers of gestational diabetes or combinations thereof. The blood glucose level of a pregnant individual can be predicted by the prediction model, for example, by inputting the detection value or test concentration value of the biomarker into the prediction model, the blood glucose value can be predicted. In some embodiments, a plurality of markers can be used in combination to predict the marker substance of a pregnant individual, and the concentration value of the marker substance in a fasting blood sample can be input into the model of the present application to predict the blood glucose value 1 hour or 2 hours after fasting.

[0060] In some embodiments, the prediction model is a support vector regression model, and the equation of the support vector regression model is:

[0061]

[0062] where Y is the predicted blood glucose value, i represents the ith biomarker, m represents the number of biomarkers, W i represents the weight of the ith biomarker, and K ibiomarker, b is a constant.

[0063] In some ways, the K i is calculated by the following formula:

[0064] K i = (γ·μ i ·ν i + coef) degree

[0065] Wherein, γ, coef and degree are parameters to be adjusted, μ i ·ν i is a linear model of independent variables, μ i is the linear coefficient of the i-th biomarker, ν i is the detection value or test value of the i-th biomarker.

[0066] In some ways, when 19 biomarkers are selected to establish the support vector regression model, m is 19.

[0067] Preferably, the 19 biomarkers are composed of glucose, 1,5-anhydroglucitol, 3-methyl-2-oxobutanoic acid, 3-hydroxybutyric acid, 2-hydroxybutyric acid, pantothenic acid, 3-methyl-2-oxovaleric acid, 4-methyl-2-oxovaleric acid, palmitoylglycerophosphatidylcholine, palmitoylcarnitine, oleic acid, glycine, phenylalanine, serine, tyrosine, isoleucine, leucine, valine, 2-oxoglutarate.

[0068] W1, W2…W 19 are the weights of glucose, 1,5-anhydroglucitol, 3-methyl-2-oxobutanoic acid, 3-hydroxybutyric acid, 2-hydroxybutyric acid, pantothenic acid, 3-methyl-2-oxovaleric acid, 4-methyl-2-oxovaleric acid, palmitoylglycerophosphatidylcholine, palmitoylcarnitine, oleic acid, glycine, phenylalanine, serine, tyrosine, isoleucine, leucine, valine, 2-oxoglutarate, respectively.

[0069] μ1, μ2…μ 19 are the linear coefficients of glucose, 1,5-anhydroglucitol, 3-methyl-2-oxobutanoic acid, 3-hydroxybutyric acid, 2-hydroxybutyric acid, pantothenic acid, 3-methyl-2-oxovaleric acid, 4-methyl-2-oxovaleric acid, palmitoylglycerophosphatidylcholine, palmitoylcarnitine, oleic acid, glycine, phenylalanine, serine, tyrosine, isoleucine, leucine, valine, 2-oxoglutarate, respectively. ν1, ν2…ν 19the measured values of glucose, 1,5-anhydro-D-glucitol, 3-methyl-2-oxobutanoic acid, 3-hydroxybutyric acid, 2-hydroxybutyric acid, pantothenic acid, 3-methyl-2-oxovaleric acid, 4-methyl-2-oxovaleric acid, palmitoylglycerophosphocholine, palmitoylcarnitine, oleic acid, glycine, phenylalanine, serine, tyrosine, isoleucine, leucine, valine, 2-oxoglutarate in the serum of the individual, respectively.

[0070] In some embodiments, the measured values are the measured serum concentrations of the biomarkers at fasting 0 hour when predicting the blood glucose value at a time point after fasting 0 hour.

[0071] In some embodiments, the support vector regression model is used to predict the blood glucose value at 1 hour after oral glucose at fasting 0 hour, b = 0.0628, g = 0.037, coef = 1, degree = 3, W1, W2…W 19 The values of b, g, coef, degree, W1, W2…W 19 The values of b, g, coef, degree, W1, W2…W

[0072] In some embodiments, the support vector regression model is used to predict the blood glucose value at 2 hours after oral glucose at fasting 0 hour, b = 0.0797, g = 0.037, coef = 2, degree = 3; W1, W2…W 19-98.73331703, 3.187643085, -0.21586202, -12.36378322, -2.548963953, -0.290267916, 0.192553693, -0.858808125, 15.2824188, -0.009794368, -17.62280907, -6.38707688, -5.655502071, -2.357173357, -0.809820523, 1.810651075, 0.243270797, -3.100345313, -0.150868078, μ1, μ2... μ 19 The values are as follows: -0.000356085, -0.033072209, 0.35070682, -0.003628292, 0.227143481, 10.14047839, 0.684864863, -0.675811046, 0.011682708, 3.438905244, 0.011565459, 0.050289956, -0.022181694, -0.038488107, -0.175825644, 0.165291135, 0.077198939, -0.070012174, 1.889998371.

[0073] It should be noted that in some modes, the detection value is obtained, and the detection value is substituted into the support vector regression model, and the system can predict the blood glucose concentration of 1 hour and 2 hours of fasting through the model, and can also be used to predict the result value (OGTT) of the oral glucose tolerance test. By comparing the prediction result of the model with the standard value of blood glucose, it can be judged whether the pregnant individual has gestational diabetes. In some modes, the detection value here can be the serum concentration of the biomarker described above, but also the serum content, the relative abundance in the serum, or the concentration or content or relative abundance of the biomarker in other body fluids, such as urine. The method for obtaining the detection value can be liquid chromatography-mass spectrometry technology, or other analysis methods such as gas chromatography, ultraviolet, infrared, nuclear magnetic resonance or immune detection that can detect the biomarker.

[0074] Using such formula or model to predict, for example, fasting sample, measure the concentration or content of selected markers, then through the above selected markers of each value, for example, each marker weight value (calculated in advance), concentration value (measured sample), coefficient value (calculated in advance by modeling), through the prediction formula of the present application to calculate to obtain the predicted blood glucose value, so as not to use oral glucose to detect tolerance in the traditional way, avoid the defects of the traditional method, in addition, it can also be predicted early, for example, 1 week, 2 weeks, 3 weeks, 4 weeks, 5 weeks, 8 weeks, 12 weeks, or 18 weeks, or 20 weeks after conception. Finally, such a way only needs to take one sample, while the traditional OGTT method needs to take at least two samples to test.

[0075] In some ways, a system containing the above model is provided, which includes the above model or model formula, and then by inputting the measured value of the marker substance, the system automatically calculates the predicted blood glucose value, or the computer automatically calculates whether the tester is gestational diabetes. Therefore, the system includes an input module for inputting the measured value of the marker substance, such as concentration value, or a storage module for storing the concentration value of the marker substance, and then the calculation system automatically extracts the concentration value and brings it into the model formula for operation. The coefficient value and weight value of these marker substances can be stored in the memory in advance, and when calculating, the model formula is automatically extracted and calculated to obtain the result, such as obtaining the blood glucose value, or obtaining the normal or diabetic pregnant individual.

[0076] In some ways, the present application provides a method for establishing a support vector regression model, which comprises the following steps:

[0077] Step (1) obtaining sample data, such as detecting the concentration of multiple biomarkers of patients at 0 hours of fasting and the blood glucose value of patients at 1 hour and 2 hours after oral glucose;

[0078] Step (2) randomly dividing the sample data obtained in step (1) into training set and validation set;

[0079] Step (3) using a polynomial kernel function K(X) for high-dimensional mapping, the mathematical expression of the polynomial kernel function is:

[0080] K(X) = (γ·X + coef) degree

[0081] X = μ i ·ν i

[0082] Wherein, γ, coef and degree are to be adjusted parameters, μ i ·ν iis a linear model of independent variables, μ i is a linear coefficient of the i-th biomarker, v i is a data value of the i-th biomarker obtained in step (1).

[0083] Step (4) training the model by the training set data: the parameter adjustment adopts a combination of grid search and gradient descent, which delimits the most possible range of the optimal parameters, and traverses all parameter combinations within the delimited range to obtain a support vector regression model containing a certain number of support vectors, and then verifies the accuracy of the model by the validation set. Among them, the support vectors constitute a "separation band" in the high-dimensional space, and when predicting new samples, the distance between the markers and the edge of the "separation band" is calculated by the formula to obtain the final predicted value, i.e. the predicted blood glucose value.

[0084] In some ways, the training set accounts for 80% of the total number of sample data, and the validation set accounts for 20% of the total number of samples. In addition, the sample data can also be divided into training set and validation set in the ratio of 1:1 or other ratios, and generally the sample data of the training set is not less than that of the validation set. The sample should have a certain quantity and representativeness, so as to have statistical significance, for example, the total number of samples is not less than 20.

[0085] In a fifth aspect of the present application, the present application provides an early diagnosis system for gestational diabetes. The diagnosis system comprises an operation module.

[0086] In some ways, the operation module comprises one or more of the classification models for jointly differentiating whether a pregnant individual has diabetes according to the plurality of biomarkers in the third aspect. The relative abundance of the biomarkers of the measured pregnant individual is input into the system, and the system can distinguish whether the pregnant woman has gestational diabetes or non-gestational diabetes; or the relative abundance of the biomarkers of a plurality of pregnant women (including pregnant women with gestational diabetes and normal pregnant women) is input into the system, and the system can divide the plurality of pregnant women into a gestational diabetes group and a non-gestational diabetes group.

[0087] In some ways, the operation module comprises the prediction model for predicting blood glucose value in the third aspect. The detection value of the biomarkers of the pregnant individual measured at 0 hours of fasting is input into the system, and the system can predict the predicted blood glucose value at a certain time (such as 1 hour and / or 2 hours) after oral glucose at 0 hours of fasting. Preferably, the system can also compare the predicted blood glucose value with the standard value to determine whether the pregnant individual has diabetes or the possibility of having diabetes.

[0088] In some ways, the diagnosis system further comprises an input module for inputting one or more detection results of the aforementioned biomarkers. Such detection results can be quantitative detection results or qualitative results.

[0089] In some embodiments, the diagnostic system comprises a detection module for detecting the sample, by which the amount of each biomarker, such as concentration, relative abundance, etc., is measured. The system comprises a measurement module for detecting the specific biomarker in the sample, which can be a liquid chromatography-mass spectrometer or a fluorescence instrument for detecting antigen-antibody reaction, etc. The specific diagnostic or detection method herein can use conventional methods, such as liquid chromatography, gas chromatography, capillary electrophoresis, supercritical fluid chromatography, ion chromatography, etc. chromatographic methods, mass spectrometry and the combination of mass spectrometry and the above chromatographic methods, nuclear magnetic, ultraviolet, infrared, etc. spectroscopy, immunological methods, etc., wherein the immunological methods include enzyme-linked immunoassay, dry chemistry method, dry test strip method, or electrochemical method.

[0090] In some embodiments, the diagnostic system comprises a judgment module for judging the relationship between the calculation result of the operation module and the critical value, and obtaining the diagnostic result.

[0091] In some embodiments, the diagnostic system further comprises an output module for outputting the diagnostic result.

[0092] In some embodiments, the diagnostic system further comprises a negative control or reference data module.

[0093] In a seventh aspect of the present application, a kit for detecting gestational diabetes is provided, which comprises a reagent for detecting one or more of the aforementioned biomarkers, which can be a blood processing reagent, such as a reagent for filtering or extracting the aforementioned biomarker, or a reagent for directly detecting the presence or amount of the biomarker, such as an antibody, antigen, or marker.

[0094] The present application has the advantages that the present application screens small molecule differential metabolites between pregnant women with gestational diabetes and normal pregnant women using serum metabolomics methods, as biomarkers, for differential diagnosis of gestational diabetes and related applications. In addition, the present application also provides a model for accurately diagnosing gestational diabetes. BRIEF DESCRIPTION OF DRAWINGS

[0095] Figure 1 is an analysis flowchart.

[0096] Figure 2 is a total ion current chart in positive ion mode.

[0097] Figure 3 is a total ion current chart in negative ion mode.

[0098] Figure 4AFigure 4A is a result plot of 19 biomarkers, and Figure 4B is a result plot of 48 biomarkers (OPLS-DA statistical result plot of CM of normal pregnant women and GDM of pregnant women with gestational diabetes mellitus).

[0099] Figure 5A Figure 5A is CAS numbers of 48 biomarkers, and Figure 5B is structural formulas of biomarkers without CAS numbers.

[0100] Figure 6 Figure 6 is a ROC curve of Model A.

[0101] Figure 7 Figure 7 is a ROC curve of Model B.

[0102] Figure 8 Figure 8 is a ROC curve of Model C.

[0103] Figure 9 Figure 9 is a ROC curve of Model D.

[0104] Figure 10 Figure 10 is a ROC curve of Model E.

[0105] Figure 11 Figure 11 is a ROC curve of Model F.

[0106] Figure 12 Figure 12 is a scatter plot and a lowess fitting curve of 1-hour blood glucose value and concentration value of 19 markers.

[0107] Figure 13 Figure 13 is a scatter plot and a lowess fitting curve of 2-hour blood glucose value and concentration value of 19 markers.

[0108] Figure 14 Figure 14 is a schematic diagram of prediction of 1-hour blood glucose value of a validation set by a support vector regression model.

[0109] Figure 15 Figure 15 is a schematic diagram of prediction of 2-hour blood glucose value of a validation set by a support vector regression model.

[0110] DETAILED DESCRIPTION

[0111] (1) Diagnosis or detection

[0112] The diagnosis or detection herein refers to detecting or assaying the biomarker in the sample, or the amount of the biomarker of interest, such as absolute amount or relative amount, and then indicating whether the individual providing the sample is likely to have or suffer from a certain disease, or the likelihood of having a certain disease, by whether the target biomarker is present or not, or the amount of the biomarker. The diagnosis and detection herein can be interchangeable. The result of the detection or diagnosis is not a direct result of the disease, but an intermediate result, and if a direct result is to be obtained, other auxiliary means such as pathology or dissection are needed to confirm the disease. For example, the present application provides a plurality of new biomarkers associated with gestational diabetes, and the change in the amount of the biomarkers is directly associated with whether the individual has gestational diabetes.

[0113] (2) Association of the biomarker with gestational diabetes

[0114] The biomarker and the biomarker have the same meaning in the present application. The association herein refers to the direct association between the presence or the change in the amount of a certain biomarker in the sample and a certain disease, such as relative increase or decrease in the amount, indicating that the likelihood of having the disease is higher than that of a healthy person.

[0115] If a plurality of different biomarkers are present in the sample or the relative change in the amount of the biomarkers is present, it indicates that the likelihood of having the disease is also higher than that of a healthy person. That is, among the biomarkers, some biomarkers are strongly associated with the disease, some biomarkers are weakly associated with the disease, or some biomarkers are not associated with a certain disease. One or more of the biomarkers with strong association can be used as a biomarker for diagnosing the disease, and the biomarkers with weak association can be combined with the strong biomarkers to diagnose a certain disease, thereby increasing the accuracy of the detection result.

[0116] For the biomarkers discovered in the serum of the present application, these markers can be used to distinguish between pregnant women with gestational diabetes and healthy pregnant women. The markers herein can be used individually as a single marker for direct detection or diagnosis, and the selection of such markers indicates that the relative change in the content of the marker is strongly associated with gestational diabetes. Of course, it is understood that the simultaneous detection of one or more markers with strong association with gestational diabetes can be selected. It is understood that in some ways, the selection of biomarkers with strong association for detection or diagnosis can achieve a certain standard accuracy, such as 60%, 65%, 70%, 80%, 85%, 90% or 95% accuracy, which indicates that these markers can obtain the intermediate value of diagnosing a certain disease, but it does not mean that it can be directly confirmed as having a certain disease. For example, in the present application, the differential metabolites in Table 2 can be selected as markers for diagnosing whether it is gestational diabetes or as markers for screening gestational diabetes from healthy people.

[0117] Of course, the differential metabolites with larger ROC values can also be selected as diagnostic markers. The so-called strong and weak are generally calculated and confirmed by some algorithms, such as marker contribution rate or weight analysis associated with gestational diabetes. Such calculation methods can be significance analysis (p value or FDR value) and fold change (fold change), multivariate statistical analysis mainly includes principal component analysis (PCA), partial least squares discriminant analysis (PLS-DA) and orthogonal partial least squares discriminant analysis (OPLS-DA), and of course other methods such as ROC analysis. Of course, other model prediction methods are also possible, and when selecting biomarkers, the differential metabolites disclosed in the present application can be selected, or other known markers can be selected or combined.

[0118] (3) Diabetes and gestational diabetes

[0119] Diabetes is a metabolic disease characterized by chronically high blood sugar in patients. Different types of diabetes, including type I diabetes, type II diabetes, gestational diabetes and other types of diabetes, are caused by the inability of the pancreas to produce enough insulin or cell insensitivity to insulin. Type I diabetes is a condition in which patients cannot produce enough insulin or cannot produce insulin at all, also known as insulin-dependent diabetes. Type II diabetes is a condition in which the patient's pancreas has no pathological problems, but the cells do not respond normally, are not sensitive to insulin, or do not respond to insulin, also known as non-insulin-dependent diabetes. Gestational diabetes refers to pregnant women who have no history of diabetes, but their blood sugar is higher than normal during pregnancy.

[0120] From the above classification, gestational diabetes is significantly different from type I diabetes and type II diabetes. It occurs in a specific period and specific population. The metabolism of this pregnant population is different from that of normal people, and their serum metabolites are also different. Therefore, the biomarkers (including serum biomarkers) used for the diagnosis of type I diabetes or type II diabetes are not necessarily suitable for gestational diabetes. For example, Joanna Hajduk et al. reported that alanine showed no significant difference between the control group and the gestational diabetes group (p>0.1) (A Combined Metabolomic and Proteomic Analysis of Gestational Diabetes Mellitus, International Journal of Molecular Sciences, 2015, 16, 30034-30045); but Sanmei Chen et al. reported that alanine was closely related to type II diabetes and was a potential biomarker (Serum amino acid profiles and risk of type 2 diabetes among Japanese adults in the Hitachi Health Study, Scientific Reports, 2019, 9:7010). Similarly, the biomarkers suitable for the diagnosis of gestational diabetes are not necessarily suitable for the diagnosis of type I diabetes or type II diabetes.

[0121] In addition, obesity is a high-risk factor for diabetes (including gestational diabetes), but not an absolute factor. The biomarker of obesity is not necessarily a biomarker of diabetes or gestational diabetes. For example, patent US16 / 375834 discloses that glutamate is a differential metabolite of obesity, while Kalliopi I. Pappa et al. reported that there was no significant difference in glutamate between the normal pregnant women and the gestational diabetes pregnant women (Intermediate metabolism in association with the amino acid profile during the third trimester of normal pregnancy and diet-controlled gestational diabetes, American Journal of Obsetrics & Gynecology, 2007, 1).

[0122] (4) Fasting

[0123] Empty stomach means not eating anything. In 2020, the National Health Commission of the People's Republic of China issued the latest “Guidelines for Collection of Venous Blood Samples WS / T 661-2020”, which states that “empty stomach requires at least 8 hours of fasting, preferably 12-14 hours, but not more than 16 hours. It is recommended to arrange blood collection between 7:00 and 9:00 in the morning. During the empty stomach period, you can drink a small amount of water.” Therefore, the empty stomach in this application means at least 8 hours without eating anything. DETAILED DESCRIPTION

[0124] In order to describe the present application more specifically, the technical solutions of the present application will be described in detail below in combination with the drawings and specific embodiments. These descriptions are only to show how the present application is implemented, and cannot limit the specific scope of the present application. The scope of the present application is defined in the claims.

[0125] Example 1: Collection of serum samples

[0126] Serum samples of normal pregnant women and gestational diabetes were collected, and these individuals were normal pregnant women and gestational diabetes samples confirmed by gold standard testing, each 30 cases, and all in the second trimester of pregnancy (20-28 weeks).

[0127] Example 2: Extraction of serum metabolites

[0128] According to the ratio of 1:4, methanol precipitant containing multiple isotopic internal standards was added to the serum sample, and after oscillation for 3 minutes, 10 minutes of centrifugation at 20℃ 4000xg was carried out. From each sample, 4 aliquots of 100 μL supernatant were taken into 4 sample plates, nitrogen was blown dry, and multiple isotopic internal standard-containing resuspension solution was added for subsequent UPLC-MS / MS detection.

[0129] Example 3: Detection of extracted serum metabolites and data preprocessing

[0130] (1) Liquid chromatography / mass spectrometry conditions

[0131] Four UPLC-MS / MS methods were all carried out using ACQUITY 2D UPLC (ultra-high performance liquid chromatography; Waters, Milford, MA, USA) combined with Q Exactive (QE) high-resolution mass spectrometry (Thermo Fisher Scientific, San Jose, USA). The mass spectrometry parameters are: scan resolution 35000, scan range 70-1000 m / z.

[0132] The specific parameters of the four UPLC-MS / MS methods are as follows:

[0133] Method 1: QE was detected using positive ion electrospray ionization (ESI) mode. Liquid chromatography was performed using a C18 column (UPLCBEH C18, 2.1 x 100 mm, 1.7 μm; Waters). The mobile phase consisted of water (A) and methanol (B) containing 0.05% PFPA (pentafluoropropionic anhydride) and 0.1% FA (formic acid).

[0134] Method 2: QE was detected using negative ion electrospray ionization (ESI) mode. Liquid chromatography was performed using a C18 column (UPLCBEH C18, 2.1 x 100 mm, 1.7 μm; Waters). The mobile phase consisted of water (A) and methanol (B) containing 6.5 mM ammonium bicarbonate.

[0135] Method 3: QE was detected using positive ion electrospray ionization (ESI) mode. Liquid chromatography was performed using a C18 column (UPLCBEH C18, 2.1x100mm, 1.7μm; Waters). The mobile phases were water (A) and methanol / acetonitrile / water (B) containing 0.05% PFPA and 0.1% FA.

[0136] Method 4: QE was detected using negative ion electrospray ionization (ESI) mode. Liquid chromatography was performed using a HILIC column (UPLC BEH Amide, 2.1 x 150 mm, 1.7 μm; Waters). The mobile phase consisted of water (A) containing 10 mM ammonium formate and acetonitrile (B).

[0137] (2) Data preprocessing

[0138] After obtaining the raw peak area of ​​each metabolite, standardization was performed for subsequent statistical and bioinformatics analysis. First, the raw peak area of ​​each metabolite was log-2 transformed to reduce the overall skewness of the distribution and make the data closer to a normal distribution. Then, the median was used for normalization. Finally, the minimum value of all samples was used to fill missing values.

[0139] Example 4: Orthogonal partial least squares discriminant analysis was used to cluster the samples, and significance analysis was combined to screen differentially expressed metabolites.

[0140] Metabolomics generally employs a combination of univariate and multivariate statistical analyses to screen for differentially expressed metabolites. Univariate analyses mainly include significance analysis (p-value or FDR value) and fold change of characteristic ions in different groups. Multivariate statistical analyses mainly include principal component analysis (PCA), partial least squares discriminant analysis (PLS-DA), and orthogonal partial least squares discriminant analysis (OPLS-DA), as shown in Figure 4.

[0141] All statistical analysis was completed using R, and the specific R-related information is shown in Table 1 below.

[0142] Table 1 R used in the present application and its related information

[0143] Name Version R 3.4.1 Rstudio 1.4.1717 mixOmics 6.10.9 ropls 1.18.1

[0144] Variable Importance for the Projection (VIP) was calculated to measure the influence strength and explanation ability of the expression pattern of each metabolite on the classification discrimination of each group of samples, and Wilcoxon rank sum test was further performed to obtain the corrected p value (FDR).

[0145] According to the screening criteria of differential metabolites: (1) VIP>1; (2) FDR<0.05, that is, when VIP>1 or FDR<0.05, it is determined that there is a significant difference between the two groups, and the metabolite is a differential metabolite between the two groups.

[0146] The present application found that the main significant differential metabolites are:

[0147] Table 2 Differential metabolites of pregnant women with gestational diabetes and normal pregnant women

[0148]

[0149]

[0150]

[0151] The 48 serum differential metabolites (their CAS numbers or structural formulas are shown in Figure 5) of gestational diabetes and normal pregnant women in the above table can be used as candidate biomarkers for the differential diagnosis of gestational diabetes, and the combination of one or more of them can be used for the auxiliary diagnosis of gestational diabetes. The smaller the FDR value and / or the larger the VIP value in the table, the more significant the difference between the two groups, and at the same time, it also indicates that the differential compound may have higher diagnostic value. As can be seen from the above table, the 48 differential metabolites are mainly related to metabolic pathways such as glucose metabolism, fatty acid metabolism, phospholipid metabolism, and amino acid metabolism. The difference of a certain compound or metabolite on the relevant metabolic pathway may affect other metabolites on the metabolic pathway, so further biomarkers for gestational diabetes can be found on these metabolic pathways.

[0152] Example 5: Classification model for differentiating pregnant women with gestational diabetes from normal pregnant women and its establishment

[0153] 1. Classification model for differentiating pregnant women with gestational diabetes from normal pregnant women using a single differential metabolite and its establishment.

[0154] ROC curves of each metabolite in Table 3 of Example 4 were established, and the experimental results were judged by the size of the area under the curve (AUC). AUC of 0.5 indicates that a single metabolite has no diagnostic value; AUC greater than 0.5 indicates that a single metabolite has diagnostic value; the larger the AUC, the higher the diagnostic value of a single metabolite.

[0155] Table 3 ROC analysis of the ROC values and related information of each differential metabolite of pregnant women with gestational diabetes and normal pregnant women

[0156]

[0157]

[0158]

[0159] 2. Classification model for differential diagnosis of pregnant women with gestational diabetes and normal pregnant women by multiple differential metabolites and establishment thereof

[0160] Based on the relative abundance of the differential metabolites in Table 3 in pregnant women with gestational diabetes and normal pregnant women, a model for differential diagnosis of pregnant women with gestational diabetes and normal pregnant women was established by orthogonal partial least squares regression (Rstudio software, NIPALS algorithm) (19 metabolite variables were used as the independent variable matrix, and the logical values of "pregnant women with gestational diabetes" and "normal pregnant women" were used as the dependent variable). The variable projection importance (VIP) and FDR value were calculated to measure the influence strength and explanatory power of the expression pattern of each metabolite on the classification and discrimination of each group of samples. Wilcoxon rank-sum test was further performed to obtain the corrected p value (FDR) for screening the optimal model parameters (Rstudio software) for differential diagnosis of gestational diabetes, and the prediction models A to F were obtained.

[0161] ① Model A was established by selecting 19 differential metabolites, and the model variables and related parameters are shown in Table 4:

[0162] Table 4 Variable and parameter list of model A

[0163]

[0164]

[0165] Model A equation: Score = 3.425*glucose + 0.854*palmitoylcarnitine - 4.598*oleate - 1.307*glycine + 0.309*phenylalanine - 2.253*Serine - 0.335*tyrosine - 0.172*isoleucine - 1.273*leucine - 0.422*valine + 0.622*2-aminoadipic acid - 0.882*1,5-anhydroglucitol (1,5-AG) + 0.898*3-methyl-2-oxobutyrate + 2.292*3-hydroxybutyrate (BHBA) + 2.919*2-hydroxybutyrate (AHB) + 1.319*pantothenic acid - 0.103*3-methyl-2-oxovalerate + 0.856*4-methyl-2-oxopentanoate + 1.256*1-palmitoyl-GPC (16:0)

[0166] In the model equation, the names of glucose, 1,5-anhydroglucitol, etc. represent the relative abundance of the corresponding biomarkers. As Figure 6 ROC analysis was performed, and the AUC was 0.910, and the sensitivity (Sensitivity) and specificity (Specficity) were 0.825 and 0.930, respectively, indicating that Model A can well distinguish between pregnant women with gestational diabetes and normal pregnant women. The critical value of Model A is 0.515, and the relative abundance of each biomarker is substituted into the above equation. When Score > 0.515, it is highly likely to be diagnosed as gestational diabetes; when Score ≤ 0.515, it is less likely to be diagnosed as gestational diabetes.

[0167] ②Select all 48 differential metabolites to establish Model B, and the model variables and related parameters are shown in Table 5:

[0168] Table 5 Variable and parameter list of Model B

[0169]

[0170]

[0171]

[0172]

[0173] The model B equation is: Score = 0.199*deoxycholate + 2.903*N-acetylvaline + 4.004*carnitine - 1.611*cystathionine + 2.441*indolelactate - 1.765*oxalate (ethanedioate) - 1.271*threonate - 0.588*3-(4-hydroxyphenyl)lactate (HPLA) + 0.284*glycocholenate sulfate - 0.678*glycolithocholate sulfate - 4.844*1-arachidonoyl-GPC (20:4) + 4.307*gamma-glutamyl-epsilon-lysine - 1.769*N-acetyltryptophan - 0.495*1-arachidonoyl-GPI (20:4) + 2.144*1-linoleoyl-GPC (18:2) + 0.635*1-palmitoyl-GPA (16:0) + 0.171*orotidine - 4.154*N6-acetyllysine - 0.039*lanthionine + 0.893*histidylalanine + 0.995*(R)-3-hydroxybutyrylcarnitine + 0.756*isovalerate (C5) - 1.463*pyroglutamine - 0.503*glycerophosphoinositol - 0.513*N-acetyltaurine + 0.205*cysteinylglycine disulfide - 1.874*8-methoxykynurenate + 0.361*isoursodeoxycholate sulfate (2) - 1.421*cis-3,4-methyleneheptanoylcarnitine + 3.425*glucose + 0.854*palmitoylcarnitine - 4.598*oleate - 1.307*glycine + 0.309*phenylalanine - 2.253*serine - 0.335*tyrosine - 0.172*isoleucine - 1.273*leucine - 0.422*valine + 0.622*2-aminoadipic acid - 0.822*1,5-anhydroglucitol (1,5-AG) + 0.898* 3-methyl-2-oxobutyrate + 2.292* 3-hydroxybutyrate (BHBA) + 2.919* 2- hydroxybutyrate (AHB) + 1.319* pantothenic acid - 0.103* 3-methyl-2-oxovalerate + 0.856* 4-methyl-2-oxopentanoate + 1.256* 1-palmitoyl-GPC (16:0).

[0174] In the model equation, the biomarker names such as oleate, glycine represent the relative abundance of the corresponding biomarker. For example, Figure 7 The ROC analysis was performed, and the AUC of model B was 0.954, and the sensitivity and specificity were 0.833 and 1.000, respectively, indicating that model B can well distinguish pregnant women with gestational diabetes mellitus from normal pregnant women. The critical value of model B is 5.368, and the relative abundance of each biomarker is substituted into the above equation. When Score > 5.368, the possibility of diagnosing gestational diabetes mellitus is high; when Score ≤ 5.368, the possibility of diagnosing gestational diabetes mellitus is low.

[0175] ③Select 11 differential metabolites to establish model C, and the model variables and related parameters are shown in Table 6:

[0176] Table 6 Variable and parameter list of model C

[0177]

[0178] The model C equation is:

[0179] Score = 1-palmitoyl-GPC (16:0)*0.163 + palmitoylcarnitine*1.775 + oleate*0.455 - glycine*0.723 + phenylalanine*0.203 + Serine*0.085 - tyrosine*1.599 - isoleucine*0.271 - leucine*1.177 + valine*0.506 + 2-aminoadipic acid*1.622

[0180] In the model equation, the biomarker names such as oleate, glycine represent the relative abundance of the corresponding biomarker. For example, Figure 8, ROC analysis was performed, and the AUC was 0.884, the sensitivity (Sensitivity) and specificity (Specficity) were 0.747 and 0.841, respectively, indicating that model C can well distinguish pregnant women with gestational diabetes mellitus from normal pregnant women. The critical value of model C is 0.662, and the relative abundance of each biomarker is substituted into the above equation. When Score>0.662, it is highly likely to be diagnosed as gestational diabetes mellitus; when Score≤0.662, it is less likely to be diagnosed as gestational diabetes mellitus.

[0181] IV. Ten differential metabolites were selected to establish model D, and the model variables and related parameters are shown in Table 7:

[0182] Table 7. Variable and parameter list of model D

[0183]

[0184] The equation of model D is:

[0185] Score = 1.847*palmitoylcarnitine + 0.447*oleate - 0.757*glycine + 0.235*phenylalanine + 0.057*Serine - 1.606*tyrosine - 0.285*isoleucine - 1.103*leucine + 0.491*valine + 1.622*2-aminoadipic acid

[0186] In the model equation, the biomarker names such as oleate, glycine, etc. represent the relative abundance of the corresponding biomarker. For example, Figure 9 ROC analysis was performed, and the AUC was 0.879, the sensitivity (Sensitivity) and specificity (Specficity) were 0.774 and 0.841, respectively, indicating that model D can well distinguish pregnant women with gestational diabetes mellitus from normal pregnant women. The critical value of model D is 0.661, and the relative abundance of each biomarker is substituted into the above equation. When Score>0.661, it is highly likely to be diagnosed as gestational diabetes mellitus; when Score≤0.661, it is less likely to be diagnosed as gestational diabetes mellitus.

[0187] V. Nine differential metabolites were selected to establish model E, and the model variables and related parameters are shown in Table 8:

[0188] Table 8. Variable and parameter list of model E

[0189]

[0190]

[0191] The model E equation is:

[0192] Score = 0.688*oleate - 0.78*glycine + 0.484*phenylalanine + 0.146*Serine - 0.781*tyrosine + 0.383*isoleucine - 1.431*leucine + 0.303*valine + 1.27*2-aminoadipic acid

[0193] In the model equation, the biomarker names such as oleate, glycine, etc. represent the relative abundance of the corresponding biomarker. For example, Figure 10 ROC analysis was performed, and the AUC was 0.782, and the sensitivity and specificity were 0.642 and 0.854, respectively, indicating that model E can well distinguish between pregnant women with gestational diabetes and normal pregnant women. The critical value of model E is 0.671. When the relative abundance of each biomarker is substituted into the above equation, when Score > 0.671, it is highly likely to be diagnosed as gestational diabetes; when Score ≤ 0.671, it is less likely to be diagnosed as gestational diabetes.

[0194] (6) Model F was established by selecting all 29 differential metabolites, and the model variables and related parameters are shown in Table 9

[0195] Table 9. Variable and parameter list of model F

[0196]

[0197]

[0198]

[0199] The model F equation is: Score = -0.0199*deoxycholate - 0.290*N-acetylvaline - 0.400*carnitine + 0.161*cystathionine - 0.244*indolelactate + 0.177*oxalate (ethanedioate) + 0.127*threonate + 0.00588*3-(4-hydroxyphenyl)lactate (HPLA) - 0.0284*glycocholenate sulfate + 0.0678*glycolithocholate sulfate + 0.484*1-arachidonoyl-GPC (20:4) - 0.431*gamma-glutamyl-epsilon-lysine + 0.177*N-acetyltryptophan + 0.0495*1-arachidonoyl-GPI (20:4) - 0.214*1-linoleoyl-GPC (18:2) - 0.0635*1-palmitoyl-GPA (16:0) - 0.0171*orotidine + 0.415*N6-acetyllysine + 0.0039*lanthionine - 0.0893*histidylalanine - 0.0995*(R)-3-hydroxybutyrylcarnitine - 0.0756*isovalerate (C5) + 0.146*pyroglutamine + 0.0503*glycerophosphoinositol + 0.0513*N-acetyltaurine - 0.0205*cysteinylglycine disulfide + 0.187*8-methoxykynurenate - 0.0361*isoursodeoxycholate sulfate (2) + 0.142*cis-3,4-methyleneheptanoylcarnitine

[0200] As Figure 11 ROC analysis was performed, and the AUC of model F was 0.947, and the sensitivity and specificity were 0.967 and 0.867, respectively, indicating that model B can well distinguish pregnant women with gestational diabetes mellitus from normal pregnant women. The critical value of model F is 0.463. When the relative abundance of each biomarker is substituted into the above equation, when Score > 0.463, it is highly likely to be diagnosed as gestational diabetes mellitus; when Score≤0.463, it is less likely to be diagnosed as gestational diabetes mellitus.

[0201] It should be noted that in the above models, when the model coefficient of the biomarker (or differential metabolite) is positive, it indicates that the biomarker is positively correlated with the occurrence of gestational diabetes, i.e. the greater the relative abundance of the biomarker, the higher the probability of being diagnosed with gestational diabetes; when the model coefficient is negative, it indicates that the biomarker is negatively correlated with the occurrence of gestational diabetes, i.e. the greater the relative abundance of the biomarker, the lower the probability of being diagnosed with gestational diabetes. The greater the absolute value of the model coefficient of the biomarker, the higher the diagnostic value of the biomarker in the model for gestational diabetes, for example: in model A, the model coefficient of oleic acid is -4.598 and the model coefficient of pantothenic acid is 1.319, indicating that the higher the relative abundance of oleic acid, the lower the probability of being diagnosed with gestational diabetes; the higher the relative abundance of pantothenic acid, the higher the probability of being diagnosed with gestational diabetes; the diagnostic value of oleic acid is higher than that of pantothenic acid, i.e. when the relative abundance of oleic acid and pantothenic acid changes by the same amount, the influence of oleic acid on the diagnostic result of model A is greater than that of pantothenic acid.

[0202] Generally speaking, when establishing a disease diagnosis model, the more biomarkers (or differential metabolites) selected, the higher the diagnostic accuracy of the model. However, in the process of clinical actual use, technical difficulty of clinical detection, clinical charges, difficulty of clinical report interpretation, and complexity of the modeling process also need to be considered. Therefore, the higher the accuracy, the higher the value of the model is not necessarily. In addition, in the process of selecting biomarkers, the metabolic pathways related to the biomarkers also need to be considered, including the concentration of related metabolic pathways and the interpretability of the biological significance of the pathways. According to the principle, the simpler the model, the better, i.e. under the premise of meeting the accuracy, the fewer biomarkers used in the model and the easier to be detected, the higher the value of the model.

[0203] Comparing the above model A, model B, model C, model D, model E and model F, (1) when establishing a model, the biomarker selected is not necessarily the compound with higher diagnostic value, for example: the AUC values of N-acetylaminethanesulfonic acid and phenylalanine in Table 3 are 0.773 and 0.512, respectively, and the diagnostic value of single N-acetylaminethanesulfonic acid for gestational diabetes is higher than that of single phenylalanine, but model A selects phenylalanine and does not select N-acetylaminethanesulfonic acid.

[0204] (2) The biomarker with higher single diagnostic value may not have better diagnostic value in the model, for example: the AUC values of serine and 2-aminoglycollic acid in Table 3 are 0.631 and 0.691, respectively, and the diagnostic value of single 2-aminoglycollic acid for gestational diabetes is higher than that of single 2-aminoglycollic acid; in model A, the model coefficient of serine is -2.253 and the model coefficient of 2-aminoglycollic acid is 0.622, and the contribution of serine in model A is much higher than that of 2-aminoglycollic acid.

[0205] (3) Model B includes all 48 biomarkers, with an AUC value of 0.954, a sensitivity of 0.833, and a specificity of 1.000, and has the best diagnostic performance; Model A includes 19 biomarkers, with an AUC value of 0.910, a sensitivity of 0.825, and a specificity of 0.930. The number of biomarkers in Model A is much lower than that in Model B, and the diagnostic performance is only slightly reduced. Considering the difficulty of comprehensive detection technology, detection cost and other factors, the actual value of Model A may be higher than that of Model B.

[0206] (4) The AUC values of Models C, D and E are all lower than those of Models A and B, but they still have high diagnostic values. The number of biomarkers in Models C, D and E is 11, 10 and 9, respectively, and the actual application process is more convenient and the detection cost is lower.

[0207] (5) Model D reduces 1-palmitoyl glycerophosphatidylcholine compared with Model C, and the AUC value decreases slightly (from 0.884 to 0.879), and the sensitivity and specificity do not change significantly, indicating that there is no significant difference in the diagnostic value of Model C compared with Model D. Model E reduces palmitoyl carnitine compared with Model D, and the AUC value decreases significantly (from 0.879 to 0.782), indicating that the diagnostic value of Model E is lower than that of Model D. Therefore, considering the number of biomarkers, Model D has a higher value.

[0208] In order to verify the classification accuracy of the above models, a blind selection experiment was performed, and 150 pregnant women were randomly selected, and the time of pregnancy was between 20-28 weeks. The abundance values of each marker were tested by the embodiments of the application, and the classification models of Models A-F were used to identify and classify by inputting the relative abundance of different markers. The data possibly belonging to gestational diabetes were obtained, in which the number of A model was 34, the number of B model was 35, the number of C model was 36, the number of D model was 35, the number of E model was 38, and the number of F model was 35. Among these classified positive gestational diabetes, after confirming by the gold standard of 150 pregnant women tested by diabetes, the actual number of gestational diabetes was 35. It is indicated that the probability of missed detection or false positive is very low by using the above models for classification and prediction, and the accuracy is above 97%, so it is considered that the classification model established by the above markers has certain accuracy and can be used for preliminary classification in clinic.

[0209] Example 6: Prediction model for predicting blood glucose value of pregnant women and establishment thereof

[0210] 1. Data acquisition

[0211] A total of 499 pregnant women were obtained, which included serum concentration values of 19 (listed in Table 12) biomarkers measured by liquid chromatography tandem mass spectrometry (unit: pg / mL) at fasting T0 hour (without oral sugar), and biochemical blood glucose values (unit mmol / L, biochemical blood glucose values can be tested using general kits) at 1 hour and 2 hours after fasting T0 hour oral glucose (75g). The detection process of the 19 biomarkers includes: extracting metabolites in serum samples, and detecting by liquid chromatography tandem mass spectrometry to obtain the concentration values of the 19 markers in Table 12 (specific data is omitted), and the specific sample extraction and processing method and parameters can use the method introduced in Examples 1-3.

[0212] 2. Model establishment

[0213] The blood glucose values at 1 hour and 2 hours after fasting oral glucose are the gold standard for the diagnosis of gestational diabetes (Yang Huixia et al., New Milestone in the Diagnosis Standard of Gestational Diabetes, Chinese Journal of Perinatal Medicine, May 2010, Vol. 3, No. 12). We selected gestational diabetes biomarkers to establish a prediction model, and brought the measured values of each biomarker at fasting T0 hour into the prediction model to predict the blood glucose values at 1 hour and 2 hours after fasting T0 hour, and compared the predicted values with the measured values to determine whether the prediction model is accurate.

[0214] First consider multiple linear regression, and include 19 variables (concentrations of markers) in the linear regression model, for example

[0215]

[0216] Where m is the number of biomarkers, μ i is the linear coefficient of the i-th biomarker, v i is the detection value of the i-th biomarker (is the concentration of 19 markers tested by taking blood samples at fasting), and b is a constant.

[0217] After testing and parameter optimization, two linear regression models were obtained for predicting the blood glucose values at 1 hour and 2 hours after fasting T0 hour, respectively, but there was a large gap between the predicted blood glucose values at 1 hour and 2 hours after fasting by the two models and the measured values, with R-squared values of 0.32 (1 hour) and 0.36 (2 hours), and RMSE values of 1.32 (1 hour) and 1.15 (2 hours), which did not meet the requirements of clinical modeling. The general evaluation basis is that the larger the R-squared value is, the better, and the smaller the RMSE is, the better, and these two parameters are the parameters for evaluating the prediction index of the model.

[0218] After observing the trend between blood glucose value and each independent variable (serum concentration of biomarker), it is determined that there may be a nonlinear relationship between blood glucose value and independent variable (see Figure 12 , Figure 13 ), especially the polynomial of variables such as M-1, M-2, M-4, M-11, etc. and the change trend of dependent variable relatively obviously show a correlation (lowess fitting curve), so a nonlinear model is considered to be used, and methods such as random forest regression, polynomial regression, support vector regression and gradient boosting regression tree are tried in turn, among which the support vector regression performs relatively optimally (see Table 10). It can be understood that the R-squared value is relatively maximum, and the RMSE value is relatively minimum, indicating that the nonlinear model, especially the support vector regression model, is the best for predicting gestational diabetes mellitus, and the specific selection calculation process is omitted.

[0219] Table 10 Comparison of nonlinear model fitting effects

[0220] Model R-squared RMSE Multiple linear regression 0.32 1.32 Random forest regression 0.21 1.51 Polynomial regression 0.22 1.22 Gradient boosting regression tree 0.39 1.24 Support vector regression 0.434 0.83

[0221] When using the support vector regression model to predict blood glucose value, the support vector matrix data of each independent variable, such as each biomarker, needs to be calculated to calculate the weight and the coefficient of each marker, and these matrix data are used to predict the final blood glucose value through specific calculation of the prediction model.

[0222] In this embodiment, 19 specific biomarkers are selected to establish support vector regression models for predicting fasting 1-hour and 2-hour blood glucose values, and the optimization process is as follows:

[0223] (1) All samples are randomly divided into training set and validation set in advance, wherein the training set accounts for 80% of the total number of samples, and the validation set accounts for 20%, i.e. there are 398 samples in the training set and 101 samples in the validation set.

[0224] (2) The tool used for modeling is R language (version 3.6.2), and Rstudio operation interface is used. The support vector regression function comes from e1071 software package (version 1.7.7), and high-dimensional mapping is performed using polynomial kernel function (polynomial), and its mathematical expression is:

[0225] K(X) = (γ·X + coef) degree

[0226] X = μ i ·ν i

[0227] Where γ, coef and degree are parameters to be adjusted; μ i ·ν i is a linear model of independent variables; μi is the linear coefficient of the i-th biomarker; v i is the serum concentration value of the i-th biomarker at 0 hour of fasting. In this example, 19 biomarkers of gestational diabetes were selected, i is 1 to 19 and corresponds to biomarkers with serial numbers M-1 to M-19, respectively, as shown in Table 12.

[0228] (3) The independent variable matrix (training set) was input into the SVR function (K(X)) in the form of continuous numerical values, and the dependent variables were 1-hour glucose value (1hPG) and 2-hour glucose value (2hPG), respectively, and the kernel was set to the polynomial kernel function "polynomial" with initial parameters of γ = 1, coef = 0, and degree = 1. After modeling, the R-square values were 0.43 (1 hour) and 0.46 (2 hours), and the RMSE values were 0.83 (1 hour) and 0.78 (2 hours) under the initial parameters. Such results are not optimal, and the R-square value is smaller than the RMSE, which is not the optimal parameter.

[0229] Parameter adjustment used a combination of grid search and gradient descent to define the most likely range of optimal parameters, and all parameter combinations within the defined range were traversed. The two final models contained 469 (1 hour) and 454 (2 hour) support vectors (support vector matrix (SVR SV )) and variable parameters, and these support vectors constituted the "interval zone" in the above high-dimensional space. When predicting new samples, the distance from the edge of the "interval zone" was calculated, which was the predicted value.

[0230] The equation of the final support vector regression model is:

[0231]

[0232] where Y is the predicted glucose value (mmol / L), i represents the i-th biomarker, m represents the number of biomarkers (m = 19), W i represents the weight of the i-th biomarker (Table 13), K i represents the coefficient of the i-th biomarker, and b is a constant. K i The coefficient is calculated by the following formula (K(x)):

[0233] K i = (γ·μ i ·ν i +coef) degree

[0234] where γ, coef, and degree are parameters to be adjusted, μ i ·ν iLinear model for independent variables, μ i Linear coefficient of the i-th biomarker, v i Detection value of the i-th biomarker (μg / mL). The parameter-related information obtained after optimization is shown in Tables 11 to 14. The RMSE of the support vector regression model under the optimal parameters was 0.67 (1 hour) and 0.53 (2 hours), respectively.

[0235] Table 11 Tuning results of support vector regression modeling

[0236] Parameter 1hPG 2hPG degree 3 3 gamma 0.037 0.037 coef 1 2 b 0.0628 0.0797

[0237] Table 12 Modeling test results of support vector regression

[0238]

[0239] Table 13 Linear coefficients of 19 independent variables

[0240]

[0241] Table 14 Weight values of 19 independent variables

[0242]

[0243] Note: The calculation formula of the weight value is: Wi = SVR coef % * % SVR sv ; SVR coef is the coefficient of 1 (1 hour) and 2 (2 hours), SVR sv is the support vector matrix of the SVR model; % * % represents matrix multiplication, and the specific values and calculation process are omitted, thereby obtaining the weight of each index.

[0244] The RMSE of the support vector regression model for predicting fasting 1-hour and 2-hour blood glucose values obtained after final optimization was 0.67 (1 hour) and 0.53 (2 hours), respectively. Using the predicted blood glucose values obtained with the above model and the actually measured blood glucose values, the Pearson correlation coefficient between the actual value and the predicted value was 0.95 (1 hour) and 0.93 (2 hours), and the RMSE was 0.59 (1 hour) and 0.59 (2 hours). The specific data results and graphical representations are shown in Figure 14 and Figure 15, the black dots are the measured blood glucose values, the gray dots are the predicted values of the model, the gray dotted line is the standard for clinical diagnosis, and the length of the line between the black dots and the red dots represents the difference between them. The prediction results of the support vector regression model for the validation set samples show that the difference between the predicted blood glucose values and the measured blood glucose values of most samples is within an acceptable range. The fitting degree of the training set is the highest compared to other methods. The difference between the predicted values and the measured values of 86% of the samples in the validation set is less than 1. The number of samples with larger differences that affect the positive and negative of gestational diabetes is also in line with expectations.

[0245] This fully demonstrates that the nonlinear regression model of the present application, combined with the weight of the marker, the coefficient and the actual measured concentration of the fasting marker, can predict the blood glucose values 1 hour and 2 hours after fasting. According to the predicted values, it can be determined whether the pregnancy is a diabetic patient. This can predict early and reduce the number of measurements. Clinical trials have been conducted using the above model. The blood glucose values predicted by the prediction model of the present application have high correlation with the actual test values, and can be practically applied in clinical practice.

Claims

1. Use of a combination of biomarkers in the manufacture of a reagent for predicting blood glucose values in a pregnant individual, characterized in that, The biomarker combination comprises glucose, 1,5-anhydro-D-glucitol, 3-methyl-2-oxobutanoic acid, 3-hydroxybutyric acid, 2-hydroxybutyric acid, pantothenic acid, 3-methyl-2-oxovaleric acid, 4-methyl-2-oxovaleric acid, palmitoylglycerophosphocholine, palmitoylcarnitine, oleic acid, glycine, phenylalanine, serine, tyrosine, isoleucine, leucine, valine, 2-oxoglutaric acid; the blood glucose value refers to the blood glucose value at 1 hour and / or 2 hours after fasting; the blood glucose value of the pregnant individual is predicted by using a support vector regression model combined with the weight, coefficient and actual measured serum concentration value of the fasting biomarker; the equation of the support vector regression model is: Y = b + åi=1mWiK i i i i i wherein Y is the predicted blood glucose value, i represents the i th biomarker, m represents the number of biomarkers, Wi represents the weight of the i th biomarker, Ki represents the coefficient of the i th biomarker, and b is a constant; the Ki is calculated by the following formula: Ki = åj=1n i ∙ν i wherein γ, coef and degree are to-be-adjusted parameters, μ i i i is the linear model of the independent variable, μ i is the linear coefficient of the i th biomarker, and ν i is the serum concentration value of the i th biomarker.

2. Use according to claim 1, characterized in that, W1, W2,... W 19 are the weights of glucose, 1,5-anhydroglucitol, 3-methyl-2-oxobutanoic acid, 3-hydroxybutyric acid, 2-hydroxybutyric acid, pantothenic acid, 3-methyl-2-oxopentanoic acid, 4-methyl-2-oxopentanoic acid, palmitoylglycerophosphatidylcholine, palmitoylcarnitine, oleic acid, glycine, phenylalanine, serine, tyrosine, isoleucine, leucine, valine, 2-aminoglutaric acid, respectively; μ1, μ2,... μ 19 are the linear coefficients of glucose, 1,5-anhydroglucitol, 3-methyl-2-oxobutanoic acid, 3-hydroxybutyric acid, 2-hydroxybutyric acid, pantothenic acid, 3-methyl-2-oxopentanoic acid, 4-methyl-2-oxopentanoic acid, palmitoylglycerophosphatidylcholine, palmitoylcarnitine, oleic acid, glycine, phenylalanine, serine, tyrosine, isoleucine, leucine, valine, 2-aminoglutaric acid, respectively.

3. Use according to claim 2, characterized in that, ν1, ν2...ν 19 respectively the serum concentration values of glucose, 1,5-anhydro-D-glucitol, 3-methyl-2- oxobutanoic acid, 3-hydroxybutyric acid, 2-hydroxybutyric acid, pantothenic acid, 3-methyl- 2-oxopentanoic acid, 4-methyl-2-oxopentanoic acid, palmitoylglycerophosphatidylcholine, palmitoylcarnitine, oleic acid, glycine, phenylalanine, serine, tyrosine, isoleucine, leucine, valine, 2-oxoglutaric acid in the serum of said individual.

4. Use according to claim 3, characterized in that, When predicting the blood glucose value of the individual 1 hour after the meal, b = 0.0628, γ = 0.037, coef = 1, degree = 3, W1, W2...W 19 The values of b, γ, coef, degree, W1, W2...W are respectively: -140.1367461, -18.20203701, -0.266373135, -3.780820943, 0.703137151, 0.012695848, 0.390205074, -0.34291643, -8.627272594, 0.012476258, 4.889600901, -0.140125414, -2.270950842, -3.66914922, 1.697783174, -1.961842966, -6.56784338, -4.497375666, -0.037450268, μ1, μ2...μ 19 The values of b, γ, coef, degree, W1, W2...W are respectively: -140.1367461, -18.20203701, -0.266373135, -3.780820943, 0.703137151, 0.012695848, 0.390205074, -0.34291643, -8.627272594, 0.012476258, 4.889600901, -0.140125414, -2.270950842, -3.66914922, 1.697783174, -1.961842966, -6.56784338, -4.497375666, -0.037450268, μ1, μ2...μ The values of b, γ, coef, degree, W1, W2...W are respectively: -140.1367461, -18.20203701, -0.266373135, -3.780820943, 0.703137151, 0.012695848, 0.390205074, -0.34291643, -8.627272594, 0.012476258, 4.889600901, -0.140125414, -2.270950842, -3.66914922, 1.697783174, -1.961842966, -6.56784338, -4.497375666, -0.037450268, μ1, μ2...μ 5. The use according to claim 3, characterized in that, When predicting the blood glucose value of the individual 2 hours after the meal, b = 0.0797, γ = 0.037, coef = 2, degree = 3; W1, W2... W 19 -0.000356085, -0.033072209, 0.35070682, -0.003628292, 0.227143481, 10.14047839, 0.684864863, -0.675811046, 0.011682708, 3.438905244, 0.011565459, 0.050289956, -0.022181694, -0.038488107, -0.175825644, 0.165291135, 0.077198939, -0.070012174, 1.889998371, respectively. 19 -0.000356085, -0.033072209, 0.35070682, -0.003628292, 0.227143481, 10.14047839, 0.684864863, -0.675811046, 0.011682708, 3.438905244, 0.011565459, 0.050289956, -0.022181694, -0.038488107, -0.175825644, 0.165291135, 0.077198939, -0.070012174, 1.889998371, respectively.

6. Use of a combination of biomarkers for the manufacture of a reagent for determining whether a pregnant individual is a diabetic, characterized in that, The support vector regression model is used to predict the blood glucose value 1 hour and 2 hours after fasting by combining the weight, coefficient and actual measured serum concentration value of the fasting biomarker, and the predicted value is used to determine whether the pregnant woman is a diabetic patient; the biomarker combination comprises glucose, 1,5-anhydroglucitol, 3-methyl-2-oxobutanoic acid, 3-hydroxybutyric acid, 2-hydroxybutyric acid, pantothenic acid, 3-methyl-2-oxovaleric acid, 4-methyl-2-oxovaleric acid, palmitoylglycerophosphatidylcholine, palmitoylcarnitine, oleic acid, glycine, phenylalanine, serine, tyrosine, isoleucine, leucine, valine, 2-aminoglycolic acid; the equation of the support vector regression model is: Wherein, Y is the predicted blood glucose value, i represents the i th biomarker, m represents the number of biomarkers, W i represents the weight of the i th biomarker, K i represents the coefficient of the i th biomarker, and b is a constant; the K i is calculated by the following formula: Wherein, γ, coef and degree are to be adjusted parameters, μ i ∙ν i is a linear model of independent variables, μ i is the linear coefficient of the i th biomarker, and ν i is the serum concentration value of the i th biomarker.

Images