Classifier for early recognition and severity prediction of placenta implantable diseases and training system thereof

Abstract

The invention discloses a classifier for early recognition and severity prediction of placenta implantable diseases and a training system thereof. The system performs normalization and discretization based on pregnant woman plasma free whole genome DNA promoter region coverage rate analysis, identifies the optimal target combination of 15 genes, including MAPK10, MIR3169, MIR12133 and the like, and has application potential in preparation of diagnostic kits. By applying a machine learning algorithm, the constructed classifier shows high sensitivity and specificity in early prediction of PAS and severity of PAS, the AUC (area under operation characteristic curve) of a receiver exceeds 0.85, early risk assessment of high-risk pregnant women is effectively realized, and heavy PAS pregnant women are identified to reduce crisis of obstetrics. The classifier provides a non-invasive prediction tool for clinic, and has remarkable clinical application value and medical significance.

Description

Technical Field

[0001] The invention belongs to the field of model prediction, in particular to a classifier for early identification and severity prediction of placenta accreta disease and a training system thereof. Background Art

[0002] Placenta accreta spectrum (PAS) is divided into placenta accreta (PA), placenta accreta (PI) and placenta percreta (PP). PI and PP are one of the main causes of obstetric emergency and maternal death. Prenatal diagnosis of severe PAS mainly relies on imaging examinations. Existing methods have many limitations, including: ① B-ultrasound has advantages in evaluating the blood flow of placental implantation, but it is limited in evaluating the depth and breadth of implantation; ② Magnetic resonance imaging (MRI) has no advantages in evaluating the blood flow of placental implantation, is expensive, is not available in all specialized hospitals, and has high requirements for imaging doctors; ③ The diagnostic time window of imaging detection is in the middle and late stages of pregnancy, and a clear diagnosis is often a dilemma; ④ In 36% of patients, the preoperative imaging diagnosis and intraoperative findings are inconsistent. Therefore, early identification of severe PAS and advancing the PAS accurate prediction and diagnosis time window to the early and mid-pregnancy period are of great clinical significance for high-risk referral of PI and PP pregnant women, multidisciplinary joint consultation, and reducing the risks to pregnant women and perinatal infants.

[0003] Non-invasive prenatal DNA testing (NIPT) based on whole genome low coverage sequencing technology is a common prenatal screening content in clinical practice. In recent years, in addition to being used for fetal chromosomal abnormality screening, NIPT has shown great value in the early prediction of pregnancy complications, such as fetal growth restriction, macrosomia, and preeclampsia, but has not yet been applied to the early prediction of PAS. Extracting nucleosome coverage footprints from the promoter region of cell-free DNA (cfDNA) in early pregnancy plasma can reflect the gene expression pattern of origin tissues such as the placenta and decidua, and shows extremely high predictive potential for placental diseases, especially PAS and its severity. Summary of the invention

[0004] In view of actual needs and deficiencies in the prior art, the present invention provides a classifier and a training system for early identification and severity prediction of placenta accreta disease, in order to solve the problem that there is currently no method to accurately predict the occurrence and severity of PAS in the early and middle stages of pregnancy.

[0005] Technical solution:

[0006] To achieve the above object, the present invention adopts the following technical solutions:

[0007] In a first aspect, the present invention provides a classifier for early identification of the risk and severity of PAS in high-risk pregnant women based on plasma cfDNA promoter coverage, wherein the target gene combination includes MAPK10, MIR3169, MIR12133, ZDHHC6, LOC644634, LOC728739, KRT33A, C1GALT1C1, MIR1263, RBFOX2, CEP192, LINC02330, LPCAT3, OR7E2P, and OR4N2.

[0008] The present invention uses NIPT data to discover the whole genome promoter coverage characteristic spectrum of cfDNA in the early and middle stage plasma of pregnant women with PAS, obtains the optimal gene combination and the optimal cutoff value of each gene through machine learning strategy (see the second aspect), trains the classifier, and predicts that the area under the curve (AUC) of the receiver operating characteristic curve (ROC) in an independent validation data set reaches above 0.85, showing good potential as a screening method for pregnant women at high risk of PAS.

[0009] Methods for assessing the risk and severity of placenta accreta in high-risk women include:

[0010] Data collection and preprocessing: Low-coverage whole-genome sequencing data from high-risk pregnant women undergoing NIPT were collected and necessary preprocessing was performed to ensure data quality.

[0011] Promoter region identification and coverage extraction: NIPT data were aligned with the human reference genome hg19 using bwa-mem, SAMtools, BEDtools software and sequence alignment algorithms. The promoter regions pTSS within 1000 bp above and below the transcription start sites (TSS) of 15 genes including MAPK10, MIR3169, MIR12133, ZDHHC6, LOC644634, LOC728739, KRT33A, C1GALT1C1, MIR1263, RBFOX2, CEP192, LINC02330, LPCAT3, OR7E2P, and OR4N2 were determined, and the raw read coverage of the pTSS regions of the 15 genes was obtained.

[0012] Characteristic factor normalization processing: The original read coverage of the pTSS region of 15 genes was normalized by a TPM-like method to obtain the normalized promoter coverage (Normalized pTSS Coverage, NPC) of the 15 genes.

[0013]

[0014] Among them, NPC i represents the normalized promoter coverage of the pTSS region of the ith gene, q i represents the raw read coverage of the pTSS region, l i represents the transcript length (all 2000), ∑ j (q j / l j ) represents the sum of the read coverage of the pTSS regions of all genes in a sample after normalization of transcript length.

[0015] Characteristic factor discretization processing: The NPC values ​​of the 15 genes are compared with the optimal cutoff values ​​of each gene (obtained from the second aspect). When the characteristic factor NPC is greater than the corresponding optimal cutoff value, it is set to 1; otherwise, it is set to 0.

[0016] Risk assessment: The discretized NPC values ​​of the 15 genes of the pregnant woman were input into the classifier to calculate the risk of severe or mild PAS.

[0017] This method is suitable for singleton pregnant women with at least one of the following high-risk factors for PAS: ① previous history of uterine surgery, such as cesarean section, myomectomy, uterine septum resection, etc.; ② previous history of uterine cavity operation, such as hysteroscopic surgery, uterine curettage, etc.; ③ in vitro fertilization-embryo transfer (IVF-ET).

[0018] This process provides a non-invasive means of assessment based on the extraction of pTSS coverage from NIPT data. This suggests that the specific combination and expression pattern of the above 15 genes have great potential in the preparation of diagnostic kits. In the development of diagnostic kits, the expression of these genes can be detected through precise molecular diagnostic technology, which can be designed as a highly sensitive and specific detection method to accurately identify and predict the risk of PAS high-risk pregnant women at an early stage, which is expected to become a routine clinical testing project and serve a wider patient population.

[0019] Secondly, this patent provides a three-classifier training system for stratified prediction of PAS based on early and mid-pregnancy plasma cfDNA promoter sequencing;

[0020] A three-classifier training system for stratified prediction of PAS based on early and mid-pregnancy plasma cfDNA promoter sequencing, including:

[0021] Dataset division module: extract the medical data of PAS high-risk pregnant women who underwent NIPT on different sequencing platforms in advance, match PAS pregnant women and non-PAS pregnant women according to age, gestational age at NIPT, fetal sex and high-risk factor distribution, randomly divide the samples of the main platform into training data set and internal validation data set, and the samples of the remaining platforms are used as independent external validation data sets. Referring to the latest FIGO diagnostic criteria, the obstetric expert group divided PAS pregnant women into PP, PI, and PA based on the surgical records. Pregnant women with PP and PI were defined as recombinant, and pregnant women with PA were defined as mild group.

[0022] Influencing factor extraction module: The pre-collected NIPT data of PAS high-risk pregnant women are used to annotate the cfDNA promoter nucleosome coverage spectrum and extract features.

[0023] NIPT is a low-depth high-throughput sequencing of free DNA in the peripheral blood of pregnant women. Using an accurate sequence alignment algorithm, bwa-mem, SAMtools, and BEDtools software were used to align NIPT data with the human reference genome hg19, delete PCR duplicates, determine the region 1000bp above and below the transcription start site (TSS) as the promoter region pTSS, and calculate the raw read coverage of the pTSS region.

[0024] The normalized promoter coverage (NPC) of each gene was obtained by TPM-like calculation to reduce the impact of sequencing depth on data extraction and analysis.

[0025]

[0026] Among them, NPC i represents the normalized promoter coverage of the pTSS region of the ith gene, q i represents the raw read coverage of the pTSS region, l i represents the transcript length (all 2000), ∑ j (q j / l j ) represents the sum of the read coverage of the pTSS regions of all genes in a sample after normalization of transcript length.

[0027] All genes corresponding to NPC were input into the system as influencing factors.

[0028] Feature screening module: Use the following strategies to screen the impact factors:

[0029] The propensity score was used to perform 1:1 matching between high-risk pregnant women with PAS recombinant and those without PAS, and between high-risk pregnant women with PAS mild group and those without PAS in the training data set, and the three matched groups were subjected to pairwise difference analysis. The influencing factors were analyzed by DESeq2, limma-voom and rank sum test between the two groups, and the influencing factors with p values ​​of <0.05 calculated by the three difference analysis methods were selected as the characteristic factors between the two groups. The three groups of characteristic factors screened out by the three comparisons were subjected to intersection analysis, and the characteristic factors screened out by the three comparisons or at least two comparisons were selected as the best characteristic factors and included in the next module.

[0030] Characteristic factor discrete module: In order to enhance the universality and clinical practicality of the classifier for different sequencing platforms, a discretization strategy is adopted. The group whose gene NPC mean is significantly higher or lower than the other two groups among the three groups is taken as the gene-specific group. The optimal cutoff value is defined as the NPC value when the Youden index between the specific group and the other two groups is the largest (that is, the sum of sensitivity and specificity is the largest). When the characteristic factor NPC is greater than the corresponding optimal cutoff value, it is set to 1; otherwise, it is set to 0.

[0031] Model acquisition module: The screened feature factors are input into machine learning feature selection processes such as recursive feature elimination (RFE), and SVM-Gaussian kernel function (Radial Basis Function, RBF) machine learning is used to gradually build a PAS disease prediction classifier.

[0032] The PAS recombinant pregnant women were set to 2, the PAS mild group pregnant women were set to 1, and the non-PAS high-risk pregnant women were set to 0. The input system trained two binary classifiers for 0-(1+2) and 1-2. To improve the sensitivity of the classifier, the parameter value of class_weight was set to {0:0.1,1:0.3}. The best feature factor combination was extracted respectively, and the classifier and its evaluation results were output.

[0033] The best feature factor combinations of the two binary classifiers 0-(1+2) and 1-2 were combined and input into the machine learning feature selection process such as recursive feature elimination (RFE) again, and multiple machine learning methods such as SVM-Gaussian kernel function (Radial Basis Function, RBF) were used to construct a 0-1-2 three-classification PAS disease prediction classifier. The best feature factor combination was extracted, and the classifier and its evaluation results were output, including the average performance of the classifier and the recognition of PAS and various types of PAS.

[0034] Beneficial effects of the present invention

[0035] PAS is a general term for a group of diseases in which the placenta abnormally adheres to or invades the myometrium. In patients with PAS, the placenta cannot be detached normally after the fetus is delivered, causing massive bleeding on the placental detachment surface. It is one of the main causes of emergency hysterectomy, multiple organ failure, disseminated intravascular coagulation, shock and other obstetric emergencies and even perinatal death. In recent years, with the increase in cesarean sections and the progress of intrauterine operations and assisted reproductive technology, the incidence of PAS has increased year by year. According to statistics, one case occurs in every 300 to 400 pregnancies. Although PAS cases diagnosed prenatally based on existing clinical methods often invade the myometrium to a greater extent, the emergency cesarean section rate, blood loss, and blood transfusion volume are all lower than those of PAS cases that have not been diagnosed prenatally. Therefore, prenatal identification and perinatal management of PAS are crucial.

[0036] This study developed a classifier based on 15 gene NPCs, which can effectively stratify and predict the risk of PAS in high-risk pregnant women based on non-invasive blood tests in the early and middle stages of pregnancy, and has important clinical application value. The introduction of this innovative method is expected to be directly applied to clinical practice, providing a new and safe prediction method for the early prevention, clinical treatment and health management of PAS, and has important clinical significance for identifying pregnancy risks, promoting maternal and child health, and preventing and controlling the source of obstetric critical illness. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 This is the Venn diagram for differential gene screening in Example 1

[0038] Figure 2 Schematic diagram of the performance of the 23-gene SVM-RBF kernel 0-(1+2) classifier obtained by inputting the NPC values ​​of 211 genes in Example 1

[0039] Figure 3 Schematic diagram of the performance of the 15-gene SVM-RBF kernel 1-2 classifier obtained by inputting the NPC values ​​of 211 genes in Example 2

[0040] Figure 4 Venn diagram of 23 genes for the 0-(1+2) classifier and 15 genes for the 1-2 classifier

[0041] Figure 5 The confusion matrix of the SVM-RBF kernel three-classifier of 15 genes is obtained by inputting the NPC values ​​of 39 genes in Example 3 DETAILED DESCRIPTION

[0042] The present invention will be further described below in conjunction with embodiments, but the protection scope of the present invention is not limited thereto:

[0043] Example 1

[0044] A PAS high-risk cohort was established in the Maternal and Child Health Hospital of City A (BGI sequencing platform), the Maternal and Child Health Hospital of City B (Illumina sequencing platform), and the Maternal and Child Health Hospital of City C (Life platform). NIPT data and clinical information of pregnant women who underwent non-invasive DNA screening and had PAS high-risk factors (at least one of the following: previous cesarean section, history of uterine surgery such as uterine fibroids, history of hysteroscopy or intrauterine operation, assisted reproductive conception) were prospectively collected and followed up until 28 days after delivery.

[0045] Exclusion criteria: ① Stillbirth, neonatal death and other non-viable pregnancy outcomes; ② Fetal chromosomal abnormalities or structural developmental abnormalities; ③ Multiple pregnancy. Pregnant women with PAS and those without PAS were matched according to age, gestational age at NIPT, fetal sex and distribution of high-risk factors.

[0046] Among the high-risk pregnant women who underwent NIPT using the BGI sequencing platform, 70 PAS pregnant women and 210 non-PAS high-risk pregnant women were included in the training data set, and 35 PAS pregnant women and 95 non-PAS high-risk pregnant women were included in the internal validation data set for difference analysis and model training. In addition, 51 PAS pregnant women and 163 non-PAS high-risk pregnant women who underwent NIPT PLUS testing (higher sequencing depth) on the BGI sequencing platform of City A Maternal and Child Health Hospital, 54 PAS pregnant women and 162 non-PAS high-risk pregnant women enrolled in City B Maternal and Child Health Hospital, and 55 PAS pregnant women and 165 non-PAS high-risk pregnant women enrolled in City C Maternal and Child Health Hospital were included in three independent external validation data sets (NIPT PLUS data set, Illumina validation set, Life validation set).

[0047] The NIPT data of the enrolled pregnant women were annotated with cfDNA promoter nucleosome coverage spectra, and the human reference genome hg19 sequence was aligned, PCR duplicates were deleted, and the raw read coverage of the pTSS region was calculated to obtain the NPC value of each gene and input it into the classifier training system.

[0048]

[0049] Among them, NPC i represents the normalized promoter coverage of the pTSS region of the ith gene, q i represents the raw read coverage of the pTSS region, l i represents the transcript length (all 2000), ∑ j (q j / l j ) represents the sum of the read coverage of the pTSS regions of all genes in a sample after normalization of transcript length.

[0050] The propensity score was used to perform 1:1 matching between high-risk pregnant women with PAS recombinant and those without PAS, and between high-risk pregnant women with PAS mild group and those without PAS in the training data set, and pairwise difference analysis was performed on the three matched groups. The influencing factors were analyzed by DESeq2, limma-voom and rank sum test between the two groups, and the influencing factors with p values ​​of <0.05 calculated by the three difference analysis methods were selected as the characteristic factors between the two groups. The three groups of characteristic factors screened out by the three comparisons were subjected to intersection analysis, and the characteristic factors screened out by at least two comparisons were selected as the best characteristic factors, resulting in 211 factors, which were included in the next module. Figure 1 Shown is a Venn diagram for differential gene screening.

[0051] In order to enhance the universality and clinical practicality of the classifier for different sequencing platforms, a discretization strategy was adopted. The group whose gene NPC mean was significantly higher or lower than the other two groups among the three groups was taken as the gene-specific group. The optimal cutoff value was defined as the NPC value when the Youden index between the specific group and the other two groups was the largest (i.e., the sum of sensitivity and specificity was the largest). When the characteristic factor NPC was greater than the corresponding optimal cutoff value, it was set to 1; otherwise, it was set to 0.

[0052] The 211 selected feature factors were input into the RFE process, and the SVM-RBF kernel was selected for classifier training. PAS pregnant women were set to 1, and non-PAS high-risk pregnant women were set to 0 and input into the system for classifier training. To improve the sensitivity of the classifier, the parameter value of class_weight was set to {0:0.1,1:0.3}. The disease risk assessment results of the target to be predicted were obtained, and k-fold cross validation (k=10) was applied to increase the robustness of the assessment. The best feature factor combination was extracted, and the optimal classifier and its evaluation results were output.

[0053] The SVM-RBF kernel training method was used, and the output result was a 23-gene SVM-RBF kernel 0-(1+2) classifier. The AUC, accuracy, sensitivity, and specificity of the classifier were calculated in the training data set, the internal validation data set, and the external validation data set. See Table 1. Figure 2 The results suggest that detecting the expression levels of these 23 genes can accurately identify pregnant women at high risk of PAS.

[0054] Table 1

[0055]

[0056]

[0057] Example 2

[0058] The methods of cohort construction, sample collection, data set division, data annotation, and feature factor screening were the same as those in Example 1, and the NPC values ​​of 211 genes were input into the classifier training system.

[0059] 211 characteristic factors were input into the RFE process, and the SVM-RBF kernel was selected for classifier training. PAS recombinant pregnant women were set to 2 and PAS light pregnant women were set to 1 for input into the system for classifier training. To improve the sensitivity of the classifier, the parameter value of class_weight was set to {1:0.1,2:0.3}. The disease risk assessment results of the target to be predicted were obtained, and k-fold cross validation (k=10) was applied to increase the robustness of the assessment. The best characteristic factor combination under different modeling methods was extracted, and the classifier and its evaluation results were output.

[0060] The SVM-RBF kernel training method was used, and the output result was a 15-gene SVM-RBF kernel 1-2 classifier. The AUC, accuracy, sensitivity, and specificity of the classifier were calculated in the training data set, the internal validation data set, and the external validation data set. See Table 2. Figure 3 The overlap between the 15 target genes of the 1-2 classifier and the 23 target genes of the 0-(1+2) classifier is shown in Figure 4 The results suggest that detecting the expression levels of these 15 genes can distinguish high-risk pregnant women who develop severe PAS from those who develop mild PAS.

[0061] Table 2

[0062]

[0063] Example 3

[0064] The methods of cohort construction, sample collection, data set division, data annotation, and feature factor screening are the same as those in Example 1. The NPC values ​​of the 23 genes in Example 1 and the 15 genes in Example 2 (the discretization thresholds are shown in Table 3) are input into the RFE process, and the SVM-RBF kernel is selected for classifier training. Severe PAS pregnant women are set to 2, mild PAS pregnant women are set to 1, and non-PAS high-risk pregnant women are set to 0 and input into the system for classifier training. To improve the sensitivity of the classifier, the parameter value of class_weight is {0:1, 1:3, 2:4}. The disease risk assessment results of the target to be predicted are obtained, and k-fold cross validation (k=10) is applied to increase the robustness of the assessment. The best feature factor combination under different modeling methods is extracted, and the classifier and its evaluation results are output.

[0065] Table 3

[0066]

[0067] The SVM-RBF kernel training method was used to extract the best feature factor combination and obtain the 15-gene SVM-RBF kernel three-classifier. The classifier evaluation results were output, including the average performance of the classifier and the recognition of PAS, severe PAS and mild PAS, as shown in Table 4. Figure 5 The results suggest that integrating the expression levels of 15 genes, including MAPK10, MIR3169, MIR12133, ZDHHC6, LOC644634, LOC728739, KRT33A, C1GALT1C1, MIR1263, RBFOX2, CEP192, LINC02330, LPCAT3, OR7E2P, and OR4N2, can accurately identify PAS at an early stage and predict its risk, and design highly sensitive and specific detection methods, which is expected to be used in clinical diagnosis kits and promote the development of personalized medicine and precision treatment.

[0068] Table 4

[0069]

[0070] The specific embodiments described herein are merely examples of the spirit of the present invention. Those skilled in the art may make various modifications or additions to the specific embodiments described or replace them in similar ways, but they will not deviate from the spirit of the present invention or exceed the scope defined by the appended claims.

Claims

1. A classifier training system for early identification and severity prediction of placenta accreta disease, characterized in that It includes: Dataset division module: extract the pre-collected medical data of pregnant women at high risk of placenta accreta disease (PAS), and divide it into discovery data set, training data set, internal validation data set, and external validation data set; Influencing factor extraction module: The pre-collected non-invasive prenatal DNA test NIPT data of PAS high-risk pregnant women are used to annotate the free DNA genome-wide promoter nucleosome coverage spectrum and extract features, calculate the original read coverage of the pTSS region, perform TPM-like normalization on the original read coverage of the pTSS region to obtain the standardized promoter coverage NPC, and the NPC corresponding to all genes is used as the influencing factor; Feature screening module: Screen the influencing factors of high-risk pregnant women with severe, mild and no PAS in the discovery data set to determine the characteristic factors; Characteristic factor discretization module: determine the optimal cutoff value for each characteristic factor and discretize the characteristic factor based on the optimal cutoff value; Model acquisition module: Input the screened characteristic factors into the feature recursive elimination (RFE) process, use the machine learning algorithm to screen the characteristic gene sets representing the occurrence and typing of PAS diseases respectively, merge the two gene sets and input them into the RFE process again, and gradually build the PAS disease three-classifier; obtain the comprehensive assessment results of the disease risk of the target to be predicted, and apply k-fold cross validation to increase the robustness of the assessment; extract the best combination of characteristic factors, and output the classifier and its evaluation results.

2. The system according to claim 1, characterized in that In the data set division module, pregnant women with severe PAS, mild PAS and those without PAS were matched according to age, gestational age at NIPT, fetal sex and distribution of high-risk factors. The samples of the main center were randomly divided into a training data set and an internal validation data set. The samples of each branch center served as an independent external validation data set. The discovery data set was obtained by performing a 1:1 matching between the gestational age at NIPT and the fetal sex in the training data set.

3. The system according to claim 1, characterized in that In the impact factor extraction module, the NIPT data were aligned with the human reference genome hg19 using a sequence alignment algorithm, PCR duplicates were removed, the region 1000 bp above and below the transcription start site TSS was determined as the promoter region pTSS, and the raw read coverage of the pTSS region was calculated.

4. The system according to claim 3, characterized in that In the impact factor extraction module, the raw read coverage of the pTSS region is normalized in a TPM-like manner, and NPC is obtained by the following formula: Among them, NPC i represents the normalized promoter coverage of the pTSS region of the ith gene, q i represents the raw read coverage of the pTSS region, l i represents the transcript length, ∑ j (q j / l j ) represents the sum of the read coverage of the pTSS regions of all genes in a sample after normalization of transcript length.

5. The system according to claim 1, characterized in that the feature screening In the module, the propensity score was used to perform 1:1 matching between high-risk pregnant women with PAS reorganization and those without PAS, and between high-risk pregnant women with PAS mild group and those without PAS in the training data set, and pairwise difference analysis was performed on the three matched groups; The influencing factors were analyzed by DESeq2, limma-voom and rank sum test between the two groups, and the influencing factors with p values ​​of < 0.05 calculated by the three difference analysis methods were selected as the characteristic factors between the two groups; The three groups of characteristic factors screened out by the three comparisons were subjected to intersection analysis, and the characteristic factors screened out by the three comparisons or at least by two comparisons were selected as the best characteristic factors.

6. The system according to claim 1, characterized in that the characteristic factor In the discrete module, in order to enhance the universality and clinical practicality of the classifier for different sequencing platforms, the group whose gene NPC mean is significantly higher or lower than the other two groups is taken as the gene-specific group; the optimal cutoff value is defined as the NPC value when the Youden index between the specific group and the other two groups is the largest; when the characteristic factor NPC is greater than the corresponding optimal cutoff value, it is set to 1; otherwise, it is set to 0.

7. The system according to claim 1, characterized in that In the model acquisition module: The best characteristic factor combination is the target genes: MAPK10, MIR3169, MIR12133, ZDHHC6, LOC644634, LOC728739, KRT33A, C1GALT1C1, MIR1263, RBFOX2, CEP192, LINC02330, LPCAT3, OR7E2P, OR4N2; The classifier uses the support vector machine SVM-RBF kernel.

8. A classifier for early identification of the risk of placenta accreta disease in high-risk pregnant women, characterized in that It is obtained by the training system described in any one of claims 1-7.

9. Reagents for detecting the expression of MAPK10, MIR3169, MIR12133, ZDHHC6, LOC644634, LOC728739, KRT33A, C1GALT1C1, MIR1263, RBFOX2, CEP192, LINC02330, LPCAT3, OR7E2P, and OR4N2 genes are used in the preparation of kits for early identification and severity diagnosis of placenta accreta disease.

Images