A method for predicting gestational diabetes in early pregnancy and use thereof

Abstract

This invention relates to a group of genes associated with gestational diabetes mellitus (GDM), a method for predicting GDM in early pregnancy using these genes, and related applications. Specifically, this invention obtains transcript start site features from at least one of the following genes from the cfDNA sample of a pregnant woman: LOC105376418, LOC105376756, DAGLA, NEU3, LOC105369949, TMEM220, CHRNE, LOC653631, IGFL4, LOC105373674, PCMTD1P5, PUS1, RHO, LOC100129931, PDZD2, RNU6-308P, MTCO3P1, TRGV1, TNS3, LOC124901922, MIR378D2, and CDK2AP2P2. These transcript start site features are then input into a prediction model, and the output results are used to determine the risk of GDM. The predictive model obtained by this invention has higher prediction accuracy than existing models, which can provide more accurate clinical intervention guidance and reduce medical costs.

Description

Technical Field

[0001] This invention relates to the field of biomedicine, specifically to a set of genes associated with gestational diabetes mellitus, methods for predicting gestational diabetes mellitus in early pregnancy using these genes, and their related applications. Background Technology

[0002] Gestational diabetes mellitus (GDM) refers to the first occurrence of impaired glucose tolerance in pregnant women during pregnancy, and it is one of the most common pregnancy-related diseases. GDM poses a significant threat to the health of both the mother and the newborn. Studies have shown that intervention before 20 weeks of gestation can reduce negative impacts on the newborn to some extent. Currently, there is a lack of non-invasive and convenient methods for predicting GDM in clinical practice. In recent years, some studies or patents have used cell-free DNA (cfDNA) in the blood for the prediction and research of GDM.

[0003] Among them, "A Model for Predicting Gestational Diabetes Mellitus Using Peripheral Blood Cell-Free DNA" (CN110387414B) discloses a model for predicting gestational diabetes mellitus using peripheral blood cell-free DNA. This model constructs a screening and prediction model for gestational diabetes mellitus based on peripheral blood cell-free DNA. Using an optimized combination of target genes—CC2D2B, NAT10, SIPA1, ZNF565, ZNF552, WDR35, MICALL1, CTNNB1, CLOCK, BCKDHB, and TGIF2LY—it can predict the onset of gestational diabetes mellitus before the appearance of clinical symptoms. The model's AUC in the validation set of the examples is 0.732. However, the predictive performance of this model is only around 0.7, indicating that the prediction effect is not ideal.

[0004] Therefore, how to better utilize cfDNA information to predict gestational diabetes in early pregnancy remains an urgent problem to be solved. Summary of the Invention

[0005] To address the above problems, this invention provides a set of genes associated with gestational diabetes, a method for predicting gestational diabetes in early pregnancy using these genes, and related applications.

[0006] The first aspect of the present invention provides the use of at least one of the genes LOC105376418, LOC105376756, DAGLA, NEU3, LOC105369949, TMEM220, CHRNE, LOC653631, IGFL4, LOC105373674, PCMTD1P5, PUS1, RHO, LOC100129931, PDZD2, RNU6-308P, MTCO3P1, TRGV1, TNS3, LOC124901922, MIR378D2, and CDK2AP2P2 in the prediction of gestational diabetes mellitus.

[0007] In some embodiments, the above applications include the use of reagents for detecting at least one of the genes LOC105376418, LOC105376756, DAGLA, NEU3, LOC105369949, TMEM220, CHRNE, LOC653631, IGFL4, LOC105373674, PCMTD1P5, PUS1, RHO, LOC100129931, PDZD2, RNU6-308P, MTCO3P1, TRGV1, TNS3, LOC124901922, MIR378D2, and CDK2AP2P2 in cfDNA samples in the preparation of gestational diabetes prediction products.

[0008] In some implementations, the above applications include the use of at least one of the genes LOC105376418, LOC105376756, DAGLA, NEU3, LOC105369949, TMEM220, CHRNE, LOC653631, IGFL4, LOC105373674, PCMTD1P5, PUS1, RHO, LOC100129931, PDZD2, RNU6-308P, MTCO3P1, TRGV1, TNS3, LOC124901922, MIR378D2, and CDK2AP2P2 in the construction of gestational diabetes mellitus prediction models.

[0009] In some implementations, the above application includes: obtaining transcript start site feature values ​​from at least one of LOC105376418, LOC105376756, DAGLA, NEU3, LOC105369949, TMEM220, CHRNE, LOC653631, IGFL4, LOC105373674, PCMTD1P5, PUS1, RHO, LOC100129931, PDZD2, RNU6-308P, MTCO3P1, TRGV1, TNS3, LOC124901922, MIR378D2, and CDK2AP2P2 in the cfDNA sample of the pregnant woman to be tested, inputting the transcript start site feature values ​​into a prediction model, and determining whether there is a risk of gestational diabetes based on the output results. Among them, the transcript start site feature is the difference coverage result between the average number of aligned sequences in the 6 windows closest to the transcription start site and the average number of aligned sequences in the 10 windows furthest from the transcription start site.

[0010] As one application of the above, a second aspect of the present invention provides a method for constructing a gestational diabetes prediction model, the method comprising:

[0011] Transcription start site features of at least one of the following genes from pregnant women's cfDNA samples—LOC105376418, LOC105376756, DAGLA, NEU3, LOC105369949, TMEM220, CHRNE, LOC653631, IGFL4, LOC105373674, PCMTD1P5, PUS1, RHO, LOC100129931, PDZD2, RNU6-308P, MTCO3P1, TRGV1, TNS3, LOC124901922, MIR378D2, and CDK2AP2P2—were obtained as relevant modeling factors. Training sample sets were constructed using these modeling factors and their corresponding pregnant women, including those with and without gestational diabetes.

[0012] Based on the above training sample set, several types of models are trained, and the trained models are evaluated. The best prediction model is determined based on the model evaluation index.

[0013] As one application of the above, a third aspect of the present invention provides a method for predicting gestational diabetes mellitus. The method includes: obtaining transcript start site feature values ​​of at least one of the genes LOC105376418, LOC105376756, DAGLA, NEU3, LOC105369949, TMEM220, CHRNE, LOC653631, IGFL4, LOC105373674, PCMTD1P5, PUS1, RHO, LOC100129931, PDZD2, RNU6-308P, MTCO3P1, TRGV1, TNS3, LOC124901922, MIR378D2, and CDK2AP2P2 from the cfDNA sample of the pregnant woman to be tested; inputting the transcript start site feature values ​​into a prediction model obtained based on the construction method of the second aspect of the present invention; and determining whether there is a risk of gestational diabetes mellitus based on the output results.

[0014] A fourth aspect of the present invention provides a predictive product for assessing the risk of gestational diabetes, the product comprising a memory for storing a program; and a processor for executing the program stored in the memory to implement the prediction method described in the third aspect of the present invention. A computer-readable storage medium is also provided, on which a program is stored that can be executed by a processor to implement the prediction method as described in the third aspect of the present invention.

[0015] A fifth aspect of the invention provides a computer-readable storage medium storing a prediction model obtained by the construction method provided in the second aspect above.

[0016] The sixth aspect of the present invention provides a method for detecting at least one of the following genes in a cfDNA sample: LOC105376418, LOC105376756, DAGLA, NEU3, LOC105369949, TMEM220, CHRNE, LOC653631, IGFL4, LOC105373674, PCMTD1P5, PUS1, RHO, LOC100129931, PDZD2, RNU6-308P, MTCO3P1, TRGV1, TNS3, LOC124901922, MIR378D2, and CDK2AP2P2. A gestational diabetes prediction kit comprising primers specifically amplifying at least one of the following genes: LOC105376418, LOC105376756, DAGLA, NEU3, LOC105369949, TMEM220, CHRNE, LOC653631, IGFL4, LOC105373674, PCMTD1P5, PUS1, RHO, LOC100129931, PDZD2, RNU6-308P, MTCO3P1, TRGV1, TNS3, LOC124901922, MIR378D2, and CDK2AP2P2.

[0017] The beneficial effects of this invention are as follows: This invention discloses 22 genes related to gestational diabetes mellitus. By using the transcript start site feature value (TSS score) of these genes, a random forest algorithm is used to construct a method that can effectively predict the risk of pregnant women developing gestational diabetes mellitus, providing a reliable means for the early prevention and treatment of gestational diabetes mellitus.

[0018] The key advantage of this invention is that the prediction method provided by this invention has a detection range of 10-17 weeks of gestation. +6 This invention provides a method and product for accurate prediction of gestational diabetes in early pregnancy, as early as 20 weeks of gestation. Studies have shown that intervention before 20 weeks of gestation can reduce negative impacts on newborns. Therefore, the prediction method and product provided by this invention can intervene in gestational diabetes in a timely manner, effectively improving maternal and infant outcomes. Furthermore, blood collection in early pregnancy is convenient, facilitating clinical application. The prediction model obtained by this invention has a validation set AUC of 0.85, specificity of 90%, and sensitivity of 64.18%, significantly higher than current detection performance, providing more accurate clinical intervention guidance and reducing medical costs. In addition, the TSS characteristic values ​​of the 22 genes involved in this invention can be detected simultaneously during NIPT testing, serving as an add-on product that reduces costs and facilitates commercial use. Attached Figure Description

[0019] Figure 1 This is the ROC diagram of the optimal model in this embodiment of the invention;

[0020] Figure 2 This is a schematic diagram showing the importance ranking of the TSS feature values ​​of the 22 genes in this embodiment of the invention. Detailed Implementation

[0021] This invention involves collecting maternal blood samples in early pregnancy and obtaining cfDNA information through routine sequencing. Based on these early pregnancy cfDNA samples, a set of genes associated with the risk of gestational diabetes was identified: LOC105376418, LOC105376756, DAGLA, NEU3, LOC105369949, TMEM220, CHRNE, LOC653631, IGFL4, LOC105373674, PCMTD1P5, PUS1, RHO, LOC100129931, PDZD2, RNU6-308P, MTCO3P1, TRGV1, TNS3, LOC124901922, MIR378D2, and CDK2AP2P2 (a total of 22 genes). Furthermore, by calculating the transcription start site characteristics of these 22 genes, a risk model for predicting gestational diabetes was constructed. The 22 genes provided by this invention can effectively and robustly predict the risk of gestational diabetes in early pregnancy, providing a powerful tool for subsequent clinical intervention and treatment.

[0022] Specifically, this invention extracts samples from pregnant women aged 10-17. +6 Blood samples were collected, and cfDNA plasma was extracted for library construction. After library construction, sequencing was performed. Specifically, in this embodiment of the invention, the sequencing platform used was the BGI Genomics T7 sequencing platform, PE100 sequencing, with a sequencing depth of approximately 20X. The purpose of sequencing was to obtain the transcription start site characteristic value (TSS score) of the aforementioned genes; therefore, the sequencing platform or method itself does not constitute a limitation of this invention. Those skilled in the art can determine the appropriate sequencing depth based on the specific sequencing method used. The FastQ data obtained from the sequencing were quality controlled, and the quality-controlled samples were used for the next step of reference genome alignment. After alignment, multiple alignment sequences and PCR repetitive sequences were removed. The transcription start site characteristic value (TSS score) was calculated from the aligned data. Specifically, the 1Kb region upstream and downstream of the transcription start site (TSS) region on the aligned genome was divided into 20 equal windows (bins), each 100bp in length. The difference coverage result is calculated using Formula 1 below, which shows the difference between the average number of aligned sequences in the 6 windows (midBin) closest to the transcription start site and the average number of aligned sequences in the 10 windows (sideBin) furthest from the transcription start site.

[0023]

[0024] After the above calculations, this invention selects 22 transcription start site features from the whole genome that are significantly different and stable in the training set, involving the genes LOC105376418, LOC105376756, DAGLA, NEU3, LOC105369949, TMEM220, CHRNE, LOC653631, IGFL4, LOC105373674, PCMTD1P5, PUS1, RHO, LOC100129931, PDZD2, RNU6-308P, MTCO3P1, TRGV1, TNS3, LOC124901922, MIR378D2, and CDK2AP2P2, as feature values ​​and predictive biomarkers for subsequent model construction.

[0025] Furthermore, the present invention uses the transcription start site feature values ​​obtained above as modeling factors, and constructs a training sample set with these modeling factors and their corresponding pregnant women, wherein the pregnant women include pregnant women with gestational diabetes and pregnant women without gestational diabetes.

[0026] Based on the above training sample set, several types of models are trained, and the trained models are evaluated. The best prediction model is determined based on the model evaluation index.

[0027] In some specific embodiments, a random forest model is used, taking the above 22 feature values ​​as model input values ​​and outputting whether the pregnant woman has gestational diabetes, to construct a gestational diabetes prediction model. Besides the random forest model mentioned above, other machine learning, deep learning, reinforcement learning, and other algorithms can also be used to construct prediction models suitable for this invention.

[0028] In some specific embodiments, the transcription start site features of all 22 genes can be used as input in model construction. In other specific embodiments, several of the aforementioned 22 genes, such as the 21 genes exemplified in Table 4 of this embodiment, can be used as input to construct the prediction model. In still other specific embodiments, any 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, or 20 of the aforementioned 22 genes can be used as input to construct the model, achieving similar prediction results. Therefore, different prediction models correspond to different inputs.

[0029] It should be noted that, as mentioned in the above description, the 22 genes and their various combinations can be fully utilized in model construction to obtain multiple different prediction models and ROC curves (Receiver Operating Characteristic curves). Based on the different ROC curves, the sensitivity, specificity, positive predictive value, and negative predictive value of each model in predicting gestational diabetes can be calculated, and the optimal model can be selected.

[0030] In one specific embodiment of the present invention, the threshold can be set as a threshold when the specificity of the model is 90%, or as a threshold when the specificity is 80% or 85%. It should be understood that other methods can also be used to determine the threshold.

[0031] It should be understood that once the prediction model is constructed using the above method, when predicting the risk of gestational diabetes, the combination of 22 genes used in the determined prediction model is used to calculate the relevant transcription start site feature values ​​of the gene sequencing results of the pregnant woman being tested. The model then automatically calculates the risk value and determines the risk by comparing it with a set threshold: values ​​above the threshold are considered high risk, and values ​​below the threshold are considered low risk. It should also be understood that the method for obtaining transcription start site feature values ​​in the sample from the gestational diabetes prediction model construction method described in this paper is also applicable to the gestational diabetes prediction method.

[0032] In some specific embodiments, the gestational diabetes prediction method disclosed in this invention includes:

[0033] 1) Obtain a sample from the pregnant woman to be tested. This sample includes, but is not limited to, plasma, whole blood, serum, urine, saliva, amniotic fluid, cerebrospinal fluid, and nipple aspiration fluid.

[0034] 2) Extract cfDNA from the sample and construct a sequencing library for sequencing. The sequencing depth is approximately 20X.

[0035] 3) The FASTQ data obtained from the sequencing were subjected to quality control. The quality control samples were then aligned with the reference genome. After alignment, multiple alignment sequences and PCR repetitive sequences were removed. The transcription start site eigenvalue (TSS score) was calculated from the aligned data.

[0036] The calculation of the transcription start site score (TSS score) involves dividing the 1 kb region upstream and downstream of the transcription start site (TSS) region on the aligned genome into 20 equal windows (bins), each 100 bp in length. The difference in sequence coverage between the average number of aligned sequences in the 6 windows (midBin) closest to the transcription start site and the average number of aligned sequences in the 10 windows (sideBin) furthest from the transcription start site is calculated using Formula 1. This difference is the transcription start site score.

[0037]

[0038] 4) Input the obtained transcription start site feature values ​​into the prediction model to obtain the gestational diabetes risk value, and compare the risk value with a predetermined threshold: values ​​higher than the threshold are considered high risk, and values ​​lower than the threshold are considered low risk.

[0039] Preferably, the gestational age of the pregnant women to be tested is 10-17 years. +6 week.

[0040] Meanwhile, this invention provides a gestational diabetes prediction kit for detecting at least one of the following genes in a sample: LOC105376418, LOC105376756, DAGLA, NEU3, LOC105369949, TMEM220, CHRNE, LOC653631, IGFL4, LOC105373674, PCMTD1P5, PUS1, RHO, LOC100129931, PDZD2, RNU6-308P, MTCO3P1, TRGV1, TNS3, LOC124901922, MIR378D2, and CDK2AP2P2. The sample contains cfDNA. The detection refers to the extraction or acquisition of the relevant gene from the sample. In one specific embodiment, the above-mentioned genes can be obtained from the sample through specific amplification. At this time, the kit includes specific amplifications of LOC105376418, LOC105376756, DAGLA, NEU3, LOC105369949, TMEM220, CHRNE, LOC653631, IGFL4, LOC105373674, PCMTD1P5, PUS1, RHO, LOC100129931, PDZD2, and R. Primers for any 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21 or all 22 genes from NU6-308P, MTCO3P1, TRGV1, TNS3, LOC124901922, MIR378D2 and CDK2AP2P2.

[0041] The present invention will now be described in further detail with reference to specific embodiments and accompanying drawings.

[0042] Example 1

[0043] 1. Sample collection

[0044] 1.1. Inclusion criteria:

[0045] Inclusion criteria for gestational diabetes mellitus: According to the standard definition of the International Association for the Study of Diabetes and Pregnancy (IADS), a universal test for gestational diabetes mellitus is performed between 24 and 28 weeks of gestation using a 75g 2-hour oral glucose tolerance test (OGTT). According to the IAD, at least one of the following criteria must be met: fasting blood glucose ≥ 5.1 mmol / L; 1-hour glucose ≥ 10.0 mmol / L; 2-hour glucose ≥ 8.5 mmol / L. Furthermore, pregnant women with type 1 or type 2 diabetes are excluded.

[0046] Healthy control inclusion criteria: The samples are full-term pregnancies without pregnancy complications, the fetus grows well at birth, and there are no obstetric, medical, or surgical complications during pregnancy. Exclusion criteria: ① Concurrent other pregnancy complications; ② Severe heart, liver, and kidney insufficiency; ③ Patients with autoimmune diseases or malignant tumor diseases. Exclude abnormal pregnant women caused by chromosomal, congenital abnormalities, premature birth, and multiple pregnancies.

[0047] 1.2. Participants:

[0048] According to the above inclusion criteria, collect the remaining blood samples of pregnant women who underwent NIPT screening in the first trimester of pregnancy, and screen out the samples detected at 10-17 +6 weeks of gestation.

[0049] Table 1. Brief introduction of sample sources

[0050]

[0051] 2. Sequencing of cfDNA:

[0052] Collect the blood samples of pregnant women, use EDTA-K2 blood collection tubes, and complete plasma separation within 6 hours after blood collection. Plasma separation conditions: Centrifuge at 1600g at 4°C for 10 minutes, then take the supernatant, and then centrifuge at 16000g at 4°C for 10 minutes. Finally, obtain the supernatant plasma, and store the processed plasma at -80°C. The input volume of plasma for cfDNA extraction is 200 microliters. Use nucleic acid extraction reagents (BGI [Hubei Medical Equipment Preparation No. 20150250]) to extract cfDNA. For library construction, use the MGIEasy cell-free DNA library preparation kit set (MGI). The library concentration greater than 8 ng / ul is considered qualified. Sequencing is performed using the MGISEQ T7 sequencing platform, with PE100 sequencing and a sequencing depth of about 20x.

[0053] 3. Obtaining cfDNA characteristics:

[0054] 3.1. Preprocessing of sequencing data:

[0055] Perform quality control on the obtained fastq data of sequencing, use the fastp software to obtain a quality control report, and perform the next step of reference genome alignment on the samples that pass the quality control. Use the bwa software to align the sequencing sequences to the hg38 version of the genome, and then remove multi-aligned sequences and PCR duplicate sequences.

[0056] 3.2. Calculation of cfDNA characteristic values:

[0057] The 1kb region upstream and downstream of the transcription start sites (TSS) on the genome was divided into 20 equal windows (bins), each 100bp in length. The TSS score was calculated using the following formula: specifically, the difference in sequence coverage between the average number of aligned sequences in the 6 windows (midBin) closest to the transcription start site and the average number of aligned sequences in the 10 windows (sideBin) furthest from the transcription start site was calculated.

[0058]

[0059] Preferably, 22 transcription start site features (TSS scores) that are significantly different and stable in the training set are selected from the whole genome as features and predictive biomarkers for subsequent model construction. Specific information is as follows:

[0060] Table 2. Information on 10 transcription start sites

[0061]

[0062]

[0063] 4. Construct a gestational diabetes prediction model

[0064] 4.1. Model Training

[0065] Using the transcription start site features of 22 selected genes, a random forest model was constructed to predict the risk of gestational diabetes mellitus. Specifically, using the sklearn package in Python, we first tuned the hyperparameters of the random forest model to select the most suitable combination of hyperparameter sizes. Secondly, with the hyperparameters fixed, we performed 100 10-fold crossover operations on the training set, selecting the model with the best AUC (Area Under Curve) result as the final model.

[0066] 4.2. Model Prediction

[0067] After training the model on the training set to obtain the optimal model, the optimal model was used to perform risk prediction on the validation set, resulting in an AUC of approximately 0.85. Specific metric results and the ROC (Receiver Operating Characteristic Curve) are shown below:

[0068] Table 3. Evaluation Metrics of the Optimal Model

[0069]

[0070] 4.3. Other Models

[0071] In addition to using the TSS features of all 22 genes as model features, this example evaluates the modeling effectiveness of other combinations of TSS features from the 22 genes. The importance ranking of the 22 gene features is as follows: Figure 2 As shown in Table 4, the same excellent prediction results can be achieved when using a partial model based on the TSS feature values ​​of 21 out of 22 genes (with one gene removed arbitrarily).

[0072] Table 4. Partial Prediction Results of the Combinatorial Model of 22 Genes for Gestational Diabetes

[0073]

[0074]

[0075]

[0076] The above examples illustrate the present invention only to aid in understanding it and are not intended to limit the scope of the invention. Those skilled in the art can make various simple deductions, modifications, or substitutions based on the principles of this invention.

Claims

1. The application of at least one of the following genes in the prediction of gestational diabetes mellitus: LOC105376418, LOC105376756, DAGLA, NEU3, LOC105369949, TMEM220, CHRNE, LOC653631, IGFL4, LOC105373674, PCMTD1P5, PUS1, RHO, LOC100129931, PDZD2, RNU6-308P, MTCO3P1, TRGV1, TNS3, LOC124901922, MIR378D2, and CDK2AP2P2.

2. Application of reagents for detecting at least one of the genes LOC105376418, LOC105376756, DAGLA, NEU3, LOC105369949, TMEM220, CHRNE, LOC653631, IGFL4, LOC105373674, PCMTD1P5, PUS1, RHO, LOC100129931, PDZD2, RNU6-308P, MTCO3P1, TRGV1, TNS3, LOC124901922, MIR378D2, and CDK2AP2P2 in cfDNA samples in the preparation of gestational diabetes prediction products.

3. Application of at least one of the genes LOC105376418, LOC105376756, DAGLA, NEU3, LOC105369949, TMEM220, CHRNE, LOC653631, IGFL4, LOC105373674, PCMTD1P5, PUS1, RHO, LOC100129931, PDZD2, RNU6-308P, MTCO3P1, TRGV1, TNS3, LOC124901922, MIR378D2, and CDK2AP2P2 in the construction of a gestational diabetes mellitus prediction model.

4. The application of at least one of the genes LOC105376418, LOC105376756, DAGLA, NEU3, LOC105369949, TMEM220, CHRNE, LOC653631, IGFL4, LOC105373674, PCMTD1P5, PUS1, RHO, LOC100129931, PDZD2, RNU6-308P, MTCO3P1, TRGV1, TNS3, LOC124901922, MIR378D2, and CDK2AP2P2 in the prediction of gestational diabetes mellitus, wherein the application includes: By obtaining transcript start site features of at least one of the genes LOC105376418, LOC105376756, DAGLA, NEU3, LOC105369949, TMEM220, CHRNE, LOC653631, IGFL4, LOC105373674, PCMTD1P5, PUS1, RHO, LOC100129931, PDZD2, RNU6-308P, MTCO3P1, TRGV1, TNS3, LOC124901922, MIR378D2, and CDK2AP2P2 from the cfDNA sample of the pregnant woman to be tested, the transcript start site features are input into the prediction model, and the risk of gestational diabetes is determined based on the output results.

5. The use according to claim 4, wherein the compound is ###0002### The transcript start site characteristic is the difference between the average number of aligned sequences in the 6 windows closest to the transcription start site and the average number of aligned sequences in the 10 windows furthest from the transcription start site.

6. A method for constructing a gestational diabetes mellitus prediction model, characterized in that, The construction method includes: Transcription start site features of at least one of the genes LOC105376418, LOC105376756, DAGLA, NEU3, LOC105369949, TMEM220, CHRNE, LOC653631, IGFL4, LOC105373674, PCMTD1P5, PUS1, RHO, LOC100129931, PDZD2, RNU6-308P, MTCO3P1, TRGV1, TNS3, LOC124901922, MIR378D2, and CDK2AP2P2 from pregnant women's cfDNA samples were obtained as relevant modeling factors. A training sample set was constructed using the modeling factors and their corresponding pregnant women, including pregnant women with gestational diabetes and pregnant women without gestational diabetes. Several types of models are trained based on the training sample set, and the trained models are evaluated. The best prediction model is determined based on the model evaluation index.

7. A method for predicting gestational diabetes, characterized by, The prediction method includes: Transcription start site features of at least one of the following in the cfDNA sample of the pregnant woman to be tested: LOC105376418, LOC105376756, DAGLA, NEU3, LOC105369949, TMEM220, CHRNE, LOC653631, IGFL4, LOC105373674, PCMTD1P5, PUS1, RHO, LOC100129931, PDZD2, RNU6-308P, MTCO3P1, TRGV1, TNS3, LOC124901922, MIR378D2, and CDK2AP2P2. These transcript start site features are then input into the prediction model obtained by the construction method described in claim 6, and the risk of gestational diabetes is determined based on the output results.

8. A prediction product for the assessment of the risk of gestational diabetes mellitus, characterized in that include: Memory, used to store programs; A processor for implementing the prediction method as described in claim 7 by executing a program stored in the memory.

9. A computer-readable storage medium, characterized in that, The medium stores a program that can be executed by a processor to implement the method as described in claim 7.

10. A computer-readable storage medium, characterized in that, The predictive model obtained by the construction method as described in claim 5 is stored on the medium.

11. A method for detecting at least one of the following genes in a cfDNA sample during pregnancy: LOC105376418, LOC105376756, DAGLA, NEU3, LOC105369949, TMEM220, CHRNE, LOC653631, IGFL4, LOC105373674, PCMTD1P5, PUS1, RHO, LOC100129931, PDZD2, RNU6-308P, MTCO3P1, TRGV1, TNS3, LOC124901922, MIR378D2, and CDK2AP2P2. A diabetes prediction kit comprising primers that specifically amplify at least one of the following: LOC105376418, LOC105376756, DAGLA, NEU3, LOC105369949, TMEM220, CHRNE, LOC653631, IGFL4, LOC105373674, PCMTD1P5, PUS1, RHO, LOC100129931, PDZD2, RNU6-308P, MTCO3P1, TRGV1, TNS3, LOC124901922, MIR378D2, and CDK2AP2P2.

Images