Protein markers for predicting the risk of premature birth and their applications

Abstract

The present invention belongs to the field of biomedicine, and discloses protein markers for predicting the risk of premature birth and their applications, wherein the protein is selected from one or more of PLA2G15, PPP1R7 and AGRN. The present invention also discloses a prediction method comprising: analyzing the expression level of a protein in a maternal blood sample, and predicting whether a subject will have premature birth based on the expression level. The present invention is used as an auxiliary means to predict the risk of premature birth in pregnant women, and to perform risk assessment and testing on patients. The present invention is the first to use the detection of the expression level of a protein with a high marker effect in a pregnant woman's body fluid sample, which has high sensitivity, high specificity, and can be monitored multiple times, and has better clinical application prospects than traditional methods.

Description

Technical Field

[0001] The present invention belongs to the field of biomedicine and relates to protein markers for predicting the risk of premature birth and applications thereof. Background Art

[0002] Preterm birth (PTB) is an abnormal disease of gestational duration and one of the most common adverse pregnancy outcomes. WHO defines it as delivery before 37 weeks of gestation. In my country, premature birth is defined as delivery that occurs between 28 weeks and less than 37 weeks of gestation, and is divided into spontaneous premature birth and medical premature birth. Among them, spontaneous preterm birth (SPTB) accounts for as high as 75% of premature births, which is extremely dangerous clinically. Effective measures need to be taken in time to prevent the occurrence of premature birth. Preterm birth (PTB) is an abnormal disease of gestational duration and one of the most common adverse pregnancy outcomes. WHO defines it as delivery before 37 weeks of gestation. In my country, premature birth is defined as delivery that occurs between 28 weeks and less than 37 weeks of gestation, and is divided into spontaneous premature birth and medical premature birth. Among them, spontaneous preterm birth (SPTB) accounts for up to 75% of premature births, which is extremely dangerous clinically. Effective measures need to be taken in time to prevent premature births. Therefore, the prevention of premature births is crucial. The gestational age of premature babies is significantly shorter than that of normal full-term fetuses, and their birth weight is also significantly lighter. They are more likely to have various growth and development problems, and even cause more serious complications and related sequelae, which will cause great economic pressure and medical burden on families and society.

[0003] Premature birth is a complex syndrome. In 2016, the World Health Organization designated the prediction and prevention of premature birth as a "priority research topic", and diagnostic and treatment guidelines have emerged in various countries, but we still lack understanding of premature birth. It is widely believed that premature birth is caused by the interaction of multiple factors, including genetic, immune, physiological, biochemical, metabolic and environmental factors. Influencing factors produced by various biological pathways, such as interleukin 6 (IL-6), stimulate the maternal-fetal interface to destroy its structure, or stimulate the uterus to contract prematurely, leading to premature birth. Although countless domestic and foreign scholars have made great efforts to prevent premature birth, there is still no treatment strategy that can effectively reduce the incidence of premature birth and the resulting neonatal morbidity and mortality, and current prevention and control measures cannot reduce the medical burden ratio caused by premature birth. Therefore, premature birth is still an extremely complex global obstetric problem.

[0004] Therefore, the prevention of premature birth is of vital importance. Early prediction and diagnosis, and identification of pregnant women at high risk of premature birth are the key to reducing premature birth and related complications. At the same time, excluding pregnant women at low risk of premature birth in the short term is the key to reducing overtreatment. Due to the limited ability of fetal fibronectin (fFN) and cervical length (CL) to predict premature birth, further research is needed to explore other possible screening methods.

[0005] In recent years, with the development of the concept of omics research and precision medicine, modern medical research and exploration have been continuously led into a new era, playing an increasingly important role in biomedicine. Among them, the technology of screening biomarkers using non-labeled protein quantitative technology is at the forefront of precision medicine. It is simple to operate and requires less sample volume. It shows unique advantages and great prospects in revealing the molecular mechanism of the disease and developing specific biomarkers. The research results are stable and reliable, which provides a guarantee for analyzing the molecular regulatory mechanism of premature birth. Proteins can be used as target molecules to study the biological state of tissues and cells of hosts or special groups to reflect the potential and immediate biological changes of premature patients. At present, in clinical practice, there are some protein markers that can effectively predict premature birth, including fetal fibronectin (fFN) and placental alpha microglobulin-1 (PAMG-1), which can effectively help doctors preliminarily identify and evaluate the possibility of premature birth. For example, the patent application number CN202110804975.X discloses a protein molecular marker for predicting premature birth and a prediction method thereof. The protein marker includes at least one of IL-1ra protein, TR protein, CCL3 protein, and FS protein. The prediction method includes: analyzing the expression spectrum of any one or a set of at least two proteins in the maternal sample, and predicting whether the subject will be premature based on the expression spectrum.

[0006] Over the past decades, research on predictive factors of premature birth has achieved certain results, but its mechanism of occurrence has not been clearly explained. At the same time, the clinical application of current biomarkers is limited due to the special application scenarios and non-standardized uterine and vaginal secretion sampling methods (such as fFN, PAMG-1, etc.). Blood protein markers are easy to obtain and have high detection accuracy, but currently only a small number of literature reports predictive markers related to premature birth in pregnant women's blood; at the same time, the use of proteomics technology to screen biomarkers of premature birth is still in its infancy, and there are no reliable predictive markers for clinical use; if premature birth can be predicted in advance and prevented before it happens, the risk of the general population can be identified and the key observation objects can be determined. Therefore, further exploring the etiology and molecular basis of premature birth, early and accurate prediction and intervention of premature birth are of great significance in reducing the incidence of premature birth and improving the survival rate of premature infants.

[0007] PPP1R7, protein phosphatase 1 regulatory subunit 7 (also known as SDS22), is a member of the protein serine / threonine phosphatase PP1 family. PP1 plays an important role in a wide range of cellular functions through the dephosphorylation of key proteins, and has been confirmed in previous studies on yeast cells. Its role in cells includes glucose regulation, kinetochore function, glycogen synthesis, meiosis and cell wall integrity. PPP1R7 (SDS22) is a highly conserved regulatory subunit in PP1 and is also one of the oldest PPI regulators. Studies have shown that SDS22 is a key regulator of G2 / M cell cycle progression. Cell division requires the precise coordination of chromosome separation and cell division by the actomyosin regulatory factor Ect2. Some scholars have demonstrated the existence of a second signaling pathway parallel to the actomyosin regulatory factor Ect2 by studying the proliferation and division of Drosophila and human cells, that is, through PP1 phosphatase and its regulatory subunit SDS22, promoting late growth and orderly cell division. Jiang et al. found in a Drosophila system model that SDS22 is a new Drosophila tumor suppressor gene that eliminates cell migration, invasion and tumor growth by inhibiting myosin II and JNK activity downstream of PPI, and dephosphorylating Thr308 and phosphorylating Ser473 through PP1. Beroukhim et al. showed that PPP1R7 is often deleted in various cancer subtypes, including breast cancer, ovarian cancer and melanoma; PPP1R7 is also often deleted in oral cancer and cervical cancer, and consistent with its gene expression, PPP1R7-RNA expression is also significantly downregulated in various cancer types. In general, PPP1R7 (SDS22) is usually locally deleted or downregulated in multiple human cancers. After searching, ARGN, Ppp1r7, and PLA2G15 have not been used to predict the risk of premature birth.

[0008] AGRN (also known as Agrin) is a protein-coding gene widely expressed in 21 tissues including kidney, lung and others. Its related pathways include chondroitin sulfate / dermatan sulfate metabolism and glycosaminoglycan metabolism. Wu et al.'s study proved that the extracellular proteoglycan (heparan sulfate proteoglycan) encoded by Agrin is synthesized in motor neurons, transported along the axon, and finally released into the synaptic basal layer to guide postsynaptic differentiation, including acetylcholine receptor (AChR) aggregation; Gros et al. and Zhang et al. also believed that Agrin can promote the association of muscle-specific kinase (MuSK) activated by low-density lipoprotein-related receptor (Lrp4) as an allosteric and paracrine regulator, thereby benefiting the formation and maintenance of neuromuscular junctions. Therefore, Agrin is essential for the formation and stability of neuromuscular junctions (NMJ). Agrin also regulates skeletal muscle regeneration and helps maintain skeletal muscle mass and function. Changes in the degradation of Agrin may lead to certain forms or stages of sarcopenia, and its fragmented C-terminal fragment can be used as a potential marker for sarcopenia caused by neuromuscular junction degeneration. In mouse models, excessive degradation of Agrin will lead to NMJ instability and loss of skeletal muscle mass; Wang Wenqing et al. believe that one of the causes of postsynaptic membrane-related CMS (congenital myasthenic syndrome) is the end plate synapse development and maintenance disorders caused by Agrin deficiency.

[0009] PLA2G15, phospholipase A2 group XV (also known as GXVPLA2, LLPL, LYPLA3, ACS, LPLA2), belongs to the lipolytic enzyme family. The gene is overexpressed in liver and amniotic cells, and its mRNA expression is mainly in peripheral tissues, among which it is highly expressed in heart, placenta, skeletal muscle, kidney and pancreas; Lysosomal phospholipase A2 (LPLA2 / PLA2G15) has dual calcium-dependent phospholipase and O-acyltransferase activities, and has a potential role in the remodeling of glycerophospholipids of lipophilic alcohol acyl groups present in acidic cell compartments, catalyzing the hydrolysis of the ester bond of fatty acyl groups attached to the sn-1 or sn-2 position of phospholipids (phospholipase A1 or A2 activity) and transferring it to the hydroxyl group at the first carbon of lipophilic alcohol (O-acyltransferase activity). Studies have shown that it plays an important role in regulating inflammation, signal transduction, and promoting tumorigenesis.

[0010] After searching, no results were found for the application of PLA2G15, PPP1R7 and AGRN proteins separately or in combination in the prediction of premature birth. Summary of the invention

[0011] In view of the above technical problems, the present invention intends to provide protein markers for predicting the risk of premature birth and their applications.

[0012] The technical solution of the present invention is as follows:

[0013] A protein marker for predicting the risk of premature birth, wherein the protein is selected from any one or more of PLA2G15, PPP1R7, and AGRN.

[0014] The present invention also protects the use of the protein marker for predicting the risk of premature birth in preparing a kit for predicting or evaluating the risk of premature birth.

[0015] Furthermore, the kit comprises a detection reagent for detecting the expression level of a biomarker associated with the risk of premature birth in a subject's sample.

[0016] Furthermore, the detection reagent is a reagent for quantitatively detecting the expression amount of any one or several of the protein markers PLA2G15, PPP1R7, and AGRN.

[0017] Furthermore, the subject is a pregnant woman with a gestational age of 28-40 weeks.

[0018] Furthermore, the subject sample is blood from a pregnant subject.

[0019] This study adopted a case-control study design. Blood samples from 24 premature pregnant women and 24 full-term pregnant women who visited the obstetrics department of the First Affiliated Hospital of Kunming Medical University from September 2019 to December 2020 and gave birth were collected for proteomic testing. The functions of differentially expressed proteins were systematically analyzed through GO (Gene Ontology) analysis, KEGG (Kyoto Encyclopedia of Genes and Genomes) analysis, PPI interaction network analysis (Protein-Protein Interaction Networks) and GSEA analysis (Gene Set Enrichment Analysis), to clarify the gene expression, metabolic pathways and protein-protein interactions involved in the differentially expressed proteins, and to further understand the molecular biological processes and biological functions involved in the differentially expressed proteins. The differentially expressed proteins PLA2G15, PPP1R7 and AGRN were selected as candidate biomarkers. Finally, the LASSO algorithm and random forest method were used to screen the most important differentially expressed proteins as candidate biomarkers, and a premature birth prediction model was established. The receiver operating characteristic (ROC) curve, clinical decision curve, calibration, discrimination, R 2 The performance of the model was evaluated by using the Kaplan Meier curve for survival analysis of the relationship between protein expression level and premature birth. The data of the subject were statistically analyzed using the software R language (Version 4.2.0) and GraphPad Prism (Version 9.0.0).

[0020] The results of ROC curve analysis showed that the AUC values ​​of the differential proteins PLA2G15, PPP1R7, and AGRN as single indicators for predicting premature delivery were 0.76 (95% CI: 0.63-0.90), 0.78 (95% CI: 0.64-0.92), and 0.81 (95% CI: 0.68-0.94), respectively; the AUC value of the combined indicator prediction model composed of (PLA2G15+PPP1R7+AGRN) in distinguishing the premature group from the normal full-term delivery group was 0.91 (95% CI: 0.83-1.00), with a sensitivity of 0.92 and a specificity of 0.83. The discrimination was 0.913, R 2 =0.624, the calibration curve showed that the model had good performance, and the DCA curve analysis results suggested that the model had high clinical applicability.

[0021] The internal validation results of the combined marker model showed that the average AUC value of the Bootstrap method was 0.92 (95% CI: 0.91-0.92), and the average C-Index was 0.92 (95% CI: 0.91-0.92); the average AUC value of the cross-validation method was 0.90 (95% CI: 0.87-0.92), and the average C-Index was 0.89 (95% CI: 0.86-0.91), indicating that the combined marker model has good stability.

[0022] The results of Kaplan Meier survival analysis showed that low expression levels of proteins PPP1R7 (HR=6.02, 95%CI: 1.69-14.40, P<0.001), PLA2G15 (HR=2.70, 95%CI: 0.70-5.50, P=0.005), and AGRN (HR=2.57, 95%CI: 0.87-5.40, P<0.001) were potentially associated with the occurrence of premature birth in pregnant women. This suggests that low expression of PPP1R7, PLA2G15, and AGRN is associated with an increased risk of premature birth. When the expression levels of the three proteins in pregnant women are reduced, the occurrence of premature birth should be vigilant.

[0023] Compared with the prior art, the present invention has the following beneficial effects:

[0024] The present invention discovers for the first time that the proteins PLA2G15, PPP1R7 and AGRN in serum and their combination can be used to effectively predict the risk of premature birth in pregnant women. By detecting the expression levels of proteins with high marker effects in blood samples of pregnant women, the detection has high sensitivity and specificity and can be monitored multiple times. Compared with traditional methods, it has better clinical application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 It is the result of PCA analysis of protein quantification between repeated samples;

[0026] Figure 2 It is the result of Pearson correlation coefficient analysis of protein quantification between two samples;

[0027] Figure 3 is the result of quality control analysis of mass spectrometry testing;

[0028] Figure 4 It is the statistical result of differential protein between groups;

[0029] Figure 5 is the volcano plot of differential proteins between groups;

[0030] Figure 6 is a heat map of differential protein expression;

[0031] Figure 7 is the subcellular structural localization and distribution of differentially expressed proteins;

[0032] Figure 8 is the result of GO enrichment analysis of differentially expressed proteins;

[0033] Fig. 9 is the result of KEGG enrichment analysis of differentially expressed proteins;

[0034] Fig.10 is a protein interaction network diagram;

[0035] Fig.11 It is a correlation analysis matrix diagram between differential protein expression abundance. Note: The numbers in the matrix diagram are correlation coefficient values, and the colors represent the degree of correlation. The darker the color, the stronger the correlation.

[0036] Fig.12 It is the result of LASSO regression analysis; A: LASSO regression coefficient change curve, B: LASSO regression cross-validation curve

[0037] Fig.13 It is the importance of random forest analysis of proteins; the bar graph on the left shows the average importance ranking of 100 random forest repeated results; the heat map on the right shows the average importance performance of each differential protein in each random forest; the importance value is distinguished by color, the redder the color, the greater the importance value.

[0038] Fig.14 It is a scatter plot and box plot of the relative expression levels of candidate protein markers in the two groups of samples;

[0039] Fig.15 It is the AUC value of the protein pairwise combination;

[0040] Fig.16It is the protein marker prediction model nomogram;

[0041] Fig.17 This is the ROC analysis of candidate protein markers, including A: PLA2G15, B: PPP1R7, C: AGRN, D: PLA2G15 vs combination, E: PPP1R7 vs combination, F: AGRN vs combination,

[0042] Fig.18 Clinical decision analysis curve of marker prediction model,

[0043] Fig.19 Survival analysis based on the cut-off value of relative expression of protein markers,

[0044] Fig. 20 Survival analysis of different expression level combinations of protein markers,

[0045] Fig.21 Linear regression analysis of the length of the prenatal and perinatal period and the relative expression levels of protein markers. Note: The length of the prenatal and perinatal period is in weeks. A: PLA2G15; B: PPP1R7; C: AGRN. DETAILED DESCRIPTION

[0046] The present invention will be further described in detail below in conjunction with the accompanying drawings and examples. The following examples are only used to illustrate the present invention and are not intended to limit the scope of the present invention. The experimental methods in the examples where specific conditions are not indicated are usually carried out according to conventional conditions such as those described in Sambrook et al., Molecular Cloning: A Laboratory Manual (New York: Cold Spring Harbor Laboratory Press, 1989), or according to the conditions recommended by the manufacturer.

[0047] Example 1 Screening for protein markers associated with premature birth

[0048] 1. Sample

[0049] A total of 48 samples with complete clinical information were screened from the samples preserved in the hospital, including 24 normal pregnant women and 24 pregnant women with premature delivery. They were randomly divided into a training group (24 cases) and a validation group (24 cases). The clinical information of the included patients included age, BMI, gestational age at blood sampling, education level, place of residence, occupation, fetal sex, whether it was a planned pregnancy, etc. The results are shown in Tables 1 and 2.

[0050] Table 1 General demographic characteristics of the study subjects

[0051]

[0052]

[0053]

[0054] Table 2 Statistics of relative expression of differentially expressed proteins in the premature group and the normal group

[0055]

[0056] 2. Proteomic analysis

[0057] 1) Sample repeatability test: principal component analysis (PCA) and Pearson correlation analysis were used to evaluate the stability of protein quantification results. Figure 1 and Figure 2 shown.

[0058] 2) Quality control of mass spectrometry experiments. Refer to the general rules of HCD fragmentation and trypsin digestion. If too few fragment ions are generated, an effective identification sequence cannot be formed, so peptides consisting of less than 5 amino acids cannot be identified. Peptides consisting of more than 20 amino acids are also not suitable for HCD fragmentation due to their large mass and relatively high number of free electrons. Figure 3 shown.

[0059] 3) Protein qualitative and quantitative results: The expression profiles of 1059 proteins in 48 samples were screened, and the expression multiples FC ≥ 1.5 or FC ≤ 1 / 1.5 (0.67) and P < 0.05 (R language statistical software "Limma" package) were used as the screening criteria for differential expression. Differential proteins with different expression levels were screened out, and the R language software "ggplot2" package was used to draw a volcano map of the differential proteins between the groups with the expression multiples FC and P value as the reference standard. The relative expression levels of the differential proteins between the premature group and the normal full-term delivery group were statistically analyzed. The R language software "pheatmap" package was used to classify and group the differential proteins between the two groups using the hierarchical clustering algorithm (Hierarchical Cluster), and the clustering results were displayed in the form of a tree heat map. The results are shown in Figure 2. Figure 4-6 shown.

[0060] 3. Analysis of biological functions of proteins

[0061] 1) Subcellular localization: Subcellular localization refers to the specific location of proteins or gene expression products in cells, such as the nucleus, cytoplasm, cell membrane or specific organelles, which are important places for proteins to play different functions. Figure 7 shown.

[0062] 2) GO enrichment analysis: DAVID (https: / / david.nc ifcrf.gov / ) was used to perform GO functional enrichment analysis on the differentially expressed proteins from three aspects: biological process, cellular component, and molecular function. The results were visualized using the R language software “ggpubr” package. Figure 8 shown.

[0063] 3) KEGG enrichment analysis: Through the KEGG database (https: / / www.kegg.jp / kegg / ), the differentially expressed proteins were systematically analyzed and enriched, which can be used to analyze the changing patterns of proteins in biological processes, disease mechanisms, and mechanisms of action. The proteins were hierarchically clustered according to their expression patterns and functional categories, and the relationship between proteins and pathways was visualized using the R language software "GOCluster" package. The results are shown in Figure 2. Fig. 9 shown.

[0064] 4) Protein interaction network analysis: All database numbers or protein sequences of the screened differentially expressed proteins were input into the GeneMANIA protein database website (http: / / genemania.org / ) and analyzed according to the default parameter settings to obtain the interaction relationship between proteins. Fig.10 as shown.

[0065] Example 2 Screening of protein markers

[0066] Different methods were used to screen protein markers.

[0067] 2.1 Protein-protein correlation analysis

[0068] The Pearson correlation analysis was used to evaluate the relationship between the abundance of differentially expressed proteins. The results showed that some proteins had a high correlation due to their similar functions, while the correlation between most other genes was medium or below, with a median correlation coefficient of 0.16 ( Fig.11 ). The numbers in the matrix are the correlation coefficient values, and the colors represent the degree of correlation. The darker the color, the stronger the correlation.

[0069] 2.2 LASSO regression and cross validation method

[0070] In view of the complex relationship between proteins and the high practical application value of using these biomarkers to predict premature birth, the LASSO regression algorithm of the R language software "glmnet" package is combined with a penalty function to reduce the dimension of the model to determine the optimal protein marker and optimize the model. Fig.12 As can be seen from A, when log(λ) increases, the coefficient of each independent variable (protein) will also change accordingly, where the vertical axis indicates the value of the coefficient. When Log(λ) reaches a certain threshold, the regression coefficient (i.e., the value of the vertical axis) will gradually approach zero and eventually reach 0. Fig.12 In B, the horizontal axis is the logarithm of the penalty coefficient Log(λ), and the vertical axis is the target parameter (Deviance) when selecting the model for cross-validation, that is, -2 times Log-likelihood. The smaller the value of the Y-axis, the better the fitting effect of the regression equation. For each λ value, an effective parameter estimation range can be obtained. When the value of λ reaches the lowest point, the dotted line on the left indicates that the model's fitting ability has reached the optimal level. Within the range of λ-se, the optimal λ value of the model is obtained by calculating the lowest standard error. The top number is the number of variables (proteins) remaining in the equation when λ is different values. In this study, λ-se was used as a reference, and three proteins with non-zero regression coefficients were retained, namely PLA2G15: -0.357, PPP1R7: -0.298, and AGRN: -0.372.

[0071] 2.3 Random Forest Method

[0072] The 28 differentially expressed proteins were used as input variables, and the random forest algorithm of the R language software "randomForest" package was used to determine the most important features in the model. The mean decrease in accuracy (MDA) refers to the degree to which the prediction accuracy of the random forest decreases when the value of a variable becomes a random number. The larger the value, the more important the variable is. This study averaged the mean decrease in accuracy for each protein in 100 random forest repetitions to minimize randomness. Finally, the importance ranking results of the variables (proteins) analyzed by the random forest model ( Fig.13 ), the top five important proteins with greater impact and stability on premature birth were selected, namely PPP1R7, AGRN, QSQX1, PLA2G15, and GSTP1, which were considered to have good research value.

[0073] Based on the above analysis, the top five proteins in the LASSO results and the random forest importance ranking were overlapped and intersected, and finally the proteins PPP1R7, AGRN, and PLA2G15 were selected as candidate biomarkers for predicting preterm birth. The specific information of the candidate markers is shown in Table 3, and their specific expression is shown in Fig.14 .

[0074] Table 3 Mass spectrometry information of candidate protein markers

[0075]

[0076] Note: Protein accession: The ID number corresponding to the protein in the database, which is the characteristic number of different proteins;

[0077] Protein description: detailed description of protein name; Gene name: gene name;

[0078] FC: Fold Change; Regulated type: protein expression type.

[0079] Example 3 Construction of prediction model

[0080] Logistic Estimation Process

[0081] 1) variable assignment, i.e. 0 indicates a normal pregnant woman, and 1 indicates a pregnant woman at risk of premature birth;

[0082] 2) Parameter estimation: The estimation of partial regression coefficients often uses maximum likelihood estimation, and its sample likelihood function is:

[0083]

[0084] When the sample size n is large, β j The sampling distribution of is approximately normal. If x j If there are only two levels, normal and premature, then the odds ratio (OR) j The 100(1-α)% confidence interval for

[0085]

[0086] 3) Hypothesis test of the model, the test hypothesis of the test model (overall test) is Ho: β1 = β2 = ... = βm = 0, and the test method is the likelihood ratio G test. The likelihood ratio G test can be used to test the joint effect of all independent variables (including constant terms) on the dependent variable. The test statistic is: The result is as follows Fig.13 shown.

[0087]

[0088] 5) ROC curves were used to determine the accuracy of single feature value and multiple feature value combined to determine premature birth. The results are shown in Table 4. As the number of feature genes increased, the area under the curve (AUC) and the accuracy of judgment gradually increased. The AUC value of PPP1R7+AGRN was 0.82, the AUC value of PLA2G15+AGRN was 0.82, and the AUC value of PPP1R7+PLA2G15 was 0.90. The AUC and accuracy formed by the combination of the three proteins were the highest, at 0.91. Fig.14 and 15 shown.

[0089] Based on the above results, the screened protein markers were added to the model for fitting to confirm their relationship with premature birth. The results of the logistic regression model showed that high-level expression of proteins AGRN (OR = 0.28, 95% CI: 0.07-0.83, P = 0.031), PPP1R7 (OR = 0.07, 95% CI: 0.07-0.31, P = 0.003), and PLA2G15 (OR = 0.05, 95% CI: 0.04-0.35, P = 0.009) were protective factors for premature birth, while low-expression proteins were associated with the occurrence of premature birth. Then, the regression model was displayed in the form of a nomogram.

[0090] like Fig.16 In the sample shown in , the expression value of protein AGRN in the peripheral blood of this pregnant woman is 0.9, the expression value of PPP1R7 is 1.1, and the expression value of PLA2G15 is 0.3. The total score of the corresponding scores is 254. Odds (Odds Ratio) represents the corresponding OR value. The risk of premature birth for this pregnant woman is 2.98 times that of other pregnant women.

[0091] Example 4 Evaluation of prediction model

[0092] 4.1 ROC curve analysis

[0093] ROC analysis was performed on all differentially expressed proteins to observe their efficacy and performance in predicting premature birth. The results showed that the AUC values ​​of most differentially expressed proteins in predicting premature birth were between 0.48 and 0.74, and the AUC values ​​of the candidate protein markers were all >0.75. The specific results are shown in Table 4.

[0094] Table 4 ROC prediction efficiency of all differentially expressed proteins

[0095]

[0096]

[0097] Note: combination: PLA2G15+PPP1R7+AGRN

[0098] In order to further evaluate the effectiveness of candidate protein markers in predicting premature birth, ROC curves were drawn for testing. The ROC curve (Receiver operating characteristic curve) is mainly used to evaluate the classification / prediction ability of a continuous indicator for a binary outcome (such as premature birth / normal full-term delivery, positive / negative), and to find the best indicator critical value (Cut-off point) to achieve the best classification effect of the indicator. First, we analyzed the performance of a single candidate protein marker in predicting premature birth. The results are as follows: Fig.17 A and Fig.17 As shown in B, the areas under the ROC curve (AUC) of candidate proteins PLA2G15 and PPP1R7 for predicting preterm delivery alone were both greater than 0.75, 0.76 (95% CI: 0.63-0.90) and 0.78 (95% CI: 0.64-0.92), respectively; while the AUC value of protein AGRN for distinguishing between preterm pregnant women and normal full-term pregnant women was 0.81 (95% CI: 0.68-0.94, Fig.17 C).

[0099] At the same time, the combined prediction model of multiple candidate markers was analyzed by logistic regression algorithm, hoping to obtain a better performance in the prediction of premature birth. The results showed that the combined marker prediction model composed of (PLA2G15+PPP1R7+AGRN) was the best choice to distinguish pregnant women with premature birth from pregnant women with normal full-term delivery. The area under the ROC curve (AUC) for predicting premature birth reached 0.91 (95% CI: 0.83-1.00), the cut-off value of the model was 0.47, the sensitivity was 0.92, and the specificity was 0.83.

[0100] The results of Delong test showed that the prediction model composed of (PLA2G15+PPP1R7+AGRN) was superior to the prediction model composed of protein PLA2G15 and PPP1R7 as single markers, and the difference was statistically significant (P D =0.022, P E =0.021). When compared with the AGRN single prediction model, the Delong test result was P F=0.115, the difference was not statistically significant; however, its net reclassification improvement index NRI (Net Reclassification improvement) value was 1.21 (95% CI: 0.76-1.58, P<0.0001), and the integrated discrimination improvement index IDI (Integrated Discrimination Improvement) value was 0.33 (95% CI: 0.20-0.47, P<0.0001), which suggested that the predictive performance of our combined marker model was better than that of the model using only AGRN as a predictive marker (see Table 5).

[0101] Table 5 Comparison between single protein model and joint model

[0102]

[0103] 4.2DCA curve analysis

[0104] By using clinical decision curve analysis (DCA), the clinical application scope of the model and the patient's benefit level and effect can be more accurately identified. Fig.18 ) results show that when the probability of premature birth in pregnant women is between 0.03 and 0.85, the benefit level of the combined marker model is much higher than other single protein models in most cases, and is higher than Treat-All and Treat-None, indicating that it is the model with the best benefits. According to the above content, the premature birth rate in China is between 0.06 and 0.1. Therefore, the use of the combined model for prediction and treatment can obtain better net benefits, which shows that the combined marker model of this subject has good clinical application value.

[0105] Example 5 Validation of the prediction model

[0106] 5.1 Bootstrap Validation

[0107] Through the Bootstrap method (Bootstrap Validation), we can use the model development cohort to perform random sampling with replacement, create a resampled set with the same amount of data, and use it to train the model, and then use it as a validation set to evaluate the performance of the model. After N iterations, we can finally obtain the performance of the model during internal validation. This study used the Bootstrap self-service sampling method (500 times) for internal validation, and the results showed that the average AUC value of the combined marker prediction model was 0.92 (95% CI: 0.91-0.92), and the average C-Index was 0.92 (95% CI: 0.91-0.92).

[0108] 5.2 Cross-Validation

[0109] Using the cross-validation method, we can divide the original data set into K incompatible subsets of similar size and consistent data distribution. We can train these subsets K times to obtain their average values, so as to better understand and improve the accuracy of the prediction model. In this study, K was set to 10 and cross-validation was performed. The results showed that the average AUC value of the combined marker model was 0.90 (95% CI: 0.87-0.92) and the average C-Index was 0.89 (95% CI: 0.86-0.91).

[0110] Example 6 Protein expression level and survival analysis of premature birth

[0111] The gestational age and outcome at delivery were set as OS. According to the optimal cutoff value, we divided pregnant women into two categories: high expression and low expression, plotted KM survival curves, and performed Log-rank tests. The results showed that among the three protein expression level groups, compared with the low expression group, the high expression group had a longer interval from the median gestational age of admission or prenatal examination sampling to delivery, and the differences between the groups were significant ( Fig.19 The results suggest that pregnant women with high levels of three proteins in their blood have a lower risk of premature birth: PPP1R7 (HR=6.02, 95%CI: 1.69-14.40, P<0.001), PLA2G15 (HR=2.70, 95%CI: 0.70-5.50, P=0.005), and AGRN (HR=2.57, 95%CI: 0.87-5.40, P<0.001).

[0112] When we combined the different expression levels of proteins PLA2G15, PPP1R7, and AGRN in the blood samples of pregnant women for survival analysis, we found that the difference in survival rate between the groups was statistically significant (χ 2 =21.5, P=0.0007), among which, the difference between the groups was the largest when the proteins PLA2G15, PPP1R7, and AGRN were all low-expressed and all high-expressed (P=0.001, Benjamini-Hochberg correction), indicating that pregnant women with high expression of the three proteins in their blood had the longest interval from the median gestational age of admission or pregnancy test sampling to delivery, that is, the gestational age at delivery was older, while pregnant women with low expression of all three proteins had the shortest gestational age at delivery, as shown in Figure 2. Fig. 20 .

[0113] Linear regression analysis revealed that there was a significant positive linear relationship between the expression levels of the three proteins and the length of the perinatal period (all P<0.05), that is, the higher their expression levels, the longer the length of the perinatal period of the pregnant woman, which means that the possibility of the fetus developing to full term at delivery is greater; conversely, it may indicate the occurrence of premature birth, see Fig.21 .

[0114] In summary, this study was based on proteomics technology and screened out differential blood proteins of premature birth through biological information analysis. Through analysis, it was found that the expression levels of proteins PPP1R7, AGRN, and PLA2G15 were significantly downregulated in pregnant women with premature birth, and were shown to be associated with an increased risk of premature birth. Changes in their expression levels in the peripheral blood of pregnant women may lead to dysregulation of signaling pathways, changes in the body's physiological environment, and ultimately trigger premature birth. Therefore, for pregnant women with low expression of PLA2G15, PPP1R7, and AGRN in peripheral blood at admission, especially those with low expression of all three, the occurrence of premature birth should be vigilant. PPP1R7, AGRN, and PLA2G15 can be used as important indicators for predicting premature birth.

Claims

1. Use of a protein marker for predicting the risk of premature birth in the preparation of a kit for predicting or evaluating the risk of premature birth, characterized in that: The protein markers are PLA2G15, PPP1R7 and AGRN, and the kit contains detection reagents for detecting the expression of biomarkers associated with the risk of premature birth in subject samples, and the subject samples are blood of pregnant subjects.

2. The use according to claim 1, characterized in that: The detection reagent is a reagent for quantitatively detecting the protein markers PLA2G15, PPP1R7 and AGRN protein expression amounts.

Images