System for predicting cholestasis of pregnancy

Abstract

The application belongs to the field of early diagnosis of diseases, and discloses a system related to prediction of intrahepatic cholestasis of pregnancy. The application constructs a simple ICP risk assessment model based on two variables of rs2296651 and age. Clinical large sample verification shows that the AUC value of the combined model for evaluating the risk of ICP is as high as 0.805. After excluding pre-pregnancy BMI, principal component and other variables, the evaluation performance is even better than that of the complex evaluation model combined with multiple SNP sites and multiple variables. The system relies on few variables, the data collection is simple, the clinical application threshold and detection cost are low. The technical scheme is clear and simple to operate, and can perform low-cost early risk screening on subjects in early pregnancy or even before pregnancy, thereby providing objective data support for early intervention, individualized health management and reasonable allocation of medical resources for pregnant women with high risk of ICP, and having high clinical application prospect and grassroots promotion value.

Description

Technical Field

[0001] This invention belongs to the field of early disease diagnosis, specifically relating to a system for predicting cholestasis of pregnancy. Background Technology

[0002] Intrahepatic cholestasis of pregnancy (ICP) is an obstetric complication that occurs in the mid-to-late stages of pregnancy. It often occurs in late pregnancy and is accompanied by pruritus and elevated serum total bile acid (TBA) levels, which can lead to serious complications such as fetal distress, stillbirth, and premature birth. Currently, the clinical diagnosis of ICP mainly uses fasting serum TBA ≥10 μmol / L or postprandial serum TBA ≥19 μmol / L as the diagnostic criteria. Pruritus is the most common clinical symptom, appearing before abnormal laboratory indicators. However, because this symptom usually occurs in late pregnancy, and ICP-induced pruritus typically does not present with a rash but rather as scratch marks, while severe pruritus can lead to nodular prurigo in pregnant women, it can easily be misdiagnosed as a rash. Therefore, ICP is prone to missed or delayed diagnosis in clinical practice, thus delaying the onset of the disease. [1] .

[0003] Currently, routine clinical screening and prediction of cholestasis of pregnancy mainly rely on clinical symptoms of pruritus and changes in serum biochemical indicators that appear in the second and third trimesters of pregnant women. [1] However, the initial clinical symptoms of ICP usually appear in the mid-to-late stages of pregnancy, and current routine clinical techniques face a serious predictive lag. Existing prediction and risk assessment models often incorporate numerous complex clinical characteristics or multi-omics variables, are highly dependent on the pregnant woman's serum biochemical indicators, are costly, and have low practicality. Therefore, there is an urgent clinical need for an early risk assessment tool for intrahepatic cholestasis of pregnancy that is less dependent on serum biochemical indicators, has fewer variables, is low-cost, and has stability.

[0004] Currently, there are three main categories of methods for risk assessment of cholestasis of pregnancy: (1) Strategies for constructing risk assessment models based on clinical indicators and machine learning: In recent years, several research teams have used machine learning methods to integrate clinical and laboratory test indicators to construct ICP risk assessment models. Ren et al. [2]A retrospective study was conducted, including 798 participants (498 with ICP and 495 controls). Initial screening through literature review included 43 variables, and LASSO regression analysis identified 11 clinical indicators (TBA, GGT, MP, LYMPH%, HCT, LYC, Neut%, PLT, PT, RBC, AST), which were then applied to 13 machine learning algorithms for model construction. The best model, CatBoost, achieved an AUC of 0.9614. (He et al.) [3] A retrospective analysis of 1092 subjects, using models built from single-period and multi-period data in early pregnancy, found that the RNN model achieved an accuracy of 90.50% in time-series data. Asali et al. [4] This study incorporated variables such as age, parity, gestational age, and previous ICP history during pregnancy, constructing models using XGBoost, KNN, SVC, and Logistic regression. The XGBoost model achieved an AUC of 0.90, outperforming traditional logistic regression models. However, building models for intrahepatic cholestasis of pregnancy using machine learning relies heavily on clinical and laboratory indicators, requiring extensive data collection for practical application and limiting its clinical applicability. Furthermore, the model structures are complex and difficult for non-specialists to interpret. Most models lack effective predictive ability in the early stages of the disease.

[0005] (2) Strategies for constructing predictive models based on demographic information or test indicators: Some studies use test indicators and demographic information to construct clinical indices or simple logistic regression to predict cholestasis of pregnancy. (Wu Wei et al.) [5] A nomogram model was constructed by incorporating six factors, including hepatitis B virus infection, gestational hypertension, low selenium intake, and family history of ICP. The AUC of the model group was 0.840, and the AUC of the validation group was 0.801. (Jiang Yuwei et al.) [6] An early ICP prediction model was constructed by incorporating four factors: education level, hepatitis B virus infection, progesterone use, and APRI score. The training set AUC was 0.83, and the validation set AUC was 0.72. However, the variable selection in such studies involves maternal demographic information and personal privacy, which may introduce data bias. (Gong Qingquan et al.) [7] An early ICP prediction model was constructed based on routine blood indicators and demographic information in early pregnancy. Five variables were included: history of miscarriage, neutrophil count, mean platelet volume, plateletcrit, and platelet distribution width. The AUC was 0.719, indicating poor accuracy when used alone. [8]Studies have found that the ALBI score (0.085 × (albumin g / L) + 0.66 × (total bilirubin μmol / L)) and MELD score (3.78 × ln(t total bilirubin μmol / L) + 11.2 × ln(INR) + 9.57 × ln(creatinine mg / dL) + 6.4) have good predictive power for ICP in early pregnancy (10-12 weeks), with AUC values ​​of 0.83 and 0.84, respectively. However, the model construction relies on a large number of clinical biochemical indicators. Peker et al. [9] Risk assessment models are constructed by calculating the APRI index (AST / platelet count), but this method has drawbacks such as small sample size and complete reliance on clinical biochemical indicators. Furthermore, clinical studies have preliminarily found a correlation between maternal age and the risk of ICP. Some studies indicate that maternal age under 25 years (RR...) adjust (2.01) or greater than or equal to 35 years old (RR) adjust A value of 1.34 indicates a risk factor for ICP. However, research on the relationship between age and ICP still has significant limitations. Empirical age segmentation is mostly based on statistical experience rather than deep pathological mechanisms, and cannot prove that a particular age group is an absolutely low-risk range. Furthermore, considering only age as a single demographic characteristic is highly subjective and one-sided.

[10] .

[0006] Constructing ICP risk assessment models using demographic data, obstetric information, and biochemical indicators is a widely used technique in clinical practice. However, model construction typically requires incorporating numerous clinical indicators and demographic variables. The collection of demographic information, which involves maternal privacy, can easily lead to information bias. Furthermore, the included clinical biochemical indicators can be affected by the patient's physiological and pathological conditions, resulting in numerical deviations. The testing instruments used by different medical institutions may also cause differences in test results, leading to measurement bias. Conversely, using a single clinical indicator to construct an ICP risk assessment model results in lower accuracy and limited guidance for clinical prediction. [2] .

[0007] (3) Strategies for constructing risk assessment models based on biomarkers: With the development of omics technologies, researchers are attempting to screen novel biomarkers for risk assessment of intrahepatic cholestasis of pregnancy using technologies such as proteomics and metabolomics. These mainly include serum proteins, exosomal proteins, and miRNAs. In proteomics, Chen et al.

[11] 4D-DIA technology was used to screen for secretory proteins common to the placenta and serum. Four key secretory proteins—NECTIN1, SPINT1, ITIH3, and SERPIND1—were incorporated into the model construction. The optimal triplet model (NECTIN1 + SPINT1 + SERPIND1) achieved an AUC of 0.9438. (Zou et al.)

[12] Using DIA technology to screen for differentially expressed proteins in serum, the AUC of a quadruple biomarker model (S100-A9, LDHA, APOA1, CHE) reached 0.962. (Feng et al.)

[13] Using iTRAQ technology to screen plasma exosomal differentially expressed proteins, the biomarker Clusterin was found to have high predictive power, with an AUC of 0.995. In the field of adipokines research, Yurtcu et al.

[14] The predictive ability of adipokines such as leptin, adiponectin, apelin, and ghrelin was systematically evaluated. Apelin showed a specificity of 96.9%, but the study also indicated that adipokines are significantly influenced by maternal BMI, requiring multifactorial correction for practical clinical application. Furthermore, serum Sortilin-1 was also confirmed to be significantly elevated in the serum of ICP patients, with an AUC of 0.814.

[15] .

[0008] Regarding miRNA biomarkers, Zou et al.

[16] A systematic study of serum miRNA profiles in ICP patients revealed that the AUC of the triplet model (miR-371a-5p, miR-6865-5p, miR-1182) was 0.845. (Dong et al.)

[17] Studies have found that the AUC of serum exosomal miRNA duplex models (hsa-miR-4271, hsa-miR-1275) is as high as 0.982. Furthermore, miRNA-7706...

[18] hsa-miR-767-3p

[19] miRNA-148a

[20] It also demonstrated a high ICP prediction capability.

[0009] In the field of metabolomics, Ruirui Dong et al.

[21] A combined metabolomics and proteomics study was conducted, revealing that the AUC of the triple model (ACOX1, L-palmitoylcarnitine, and glycocholic acid) reached 0.993 in the third trimester, compared to 0.932 in the first trimester. (Xiao Chen et al.)

[22] The analysis of urinary bile acid profiles revealed that the AUC of the dual model (GCA + T-ω-MCA) was as high as 0.960.

[0010] Existing predictive models for cholestasis of pregnancy based on novel biomarkers exhibit high predictive efficacy, with some models achieving an AUC exceeding 0.98. However, their limitations include the complexity and high cost of detection technologies such as DIA mass spectrometry and iTRAQ, hindering their widespread adoption in primary care hospitals. Furthermore, exosome extraction is a complex process requiring strict preservation of clinical samples. The use of different biomarkers in different studies also makes it difficult to establish unified clinical testing standards, thus lacking practical clinical guidance.

[0011] In summary, while current risk assessment models for intrahepatic cholestasis of pregnancy (ICP) each have their own characteristics, none fully meet the practical requirements for early ICP risk assessment in clinical practice. Therefore, exploring a new, cost-effective strategy for early risk assessment of ICP that is less reliant on biochemical testing and easier to implement at the grassroots level remains essential.

[0012] The inventors previously conducted two maternal-infant genomics studies to comprehensively analyze the genetic basis of ICP in the Chinese population, and for the first time confirmed through large-scale population studies that the rs2296651 variant on the SLC10A1 gene is a key genetic factor in the occurrence of ICP in the Chinese population. [23，24] However, current findings are limited to etiological exploration and epidemiological association analysis, and have not yet been transformed into an early risk assessment tool for ICP that can guide clinical practice.

[0013] Clinically, there is still a need for a low-cost approach that does not rely on complex biochemical indicators and can be used to predict ICP risk in early pregnancy or even before pregnancy. Summary of the Invention

[0014] The purpose of this invention is to overcome at least one deficiency of the prior art and to provide a system for predicting cholestasis of pregnancy.

[0015] The technical solution adopted in this invention is: A system for predicting cholestasis of pregnancy includes: The data acquisition module is used to collect the genotype of the rs2296651 locus and the age of the subject in the subject's sample. The mutant base of rs2296651 is T and its complementary base A, and the wild-type base is C and its complementary base G. The genotype is wild-type homozygous, heterozygous mutant, or mutant homozygous. The data analysis module determines the risk of cholestasis of pregnancy based on the genotype of the rs2296651 locus in the subject samples and the age of the subjects. The results output module is used to output the analysis results from the data analysis module.

[0016] In some instances, the mathematical expression of the risk assessment model of the data analysis module is: P = β0 + β1 × rs2296651 + β2 × Age; where P is the probability of ICP incidence; β0 is a constant term, β1 and β2 are regression coefficients, rs2296651 is the genotype assignment: 0 when the subject is wild-type homozygous, 1 when it is heterozygous mutation, and 2 when it is mutant homozygous; Age is the actual age of the pregnant woman in years.

[0017] In some instances, machine learning classification algorithms are used to determine the values ​​of β0, β1, and β2 based on known samples.

[0018] In some instances, the machine learning classification algorithm includes any one or more combinations of: logistic regression, support vector machine, decision tree, random forest, gradient boosting tree, artificial neural network, Naive Bayes, or K-nearest neighbor algorithm.

[0019] In some instances, the machine learning classification algorithm is a logistic regression algorithm.

[0020] In some instances, the characteristics are β0 = -7.386, β1 = 2.425 and β2 = 0.048; when the calculated P value is greater than the threshold of 0.024, it is assessed as high risk of ICP, and vice versa.

[0021] In some instances, the genotype at the rs2296651 locus is determined by PCR amplification or probes.

[0022] In some instances, the primer pairs used for PCR amplification are: (1) a forward primer, the nucleotide sequence of which is shown in SEQ ID NO. 1; and (2) a reverse primer, the nucleotide sequence of which is shown in SEQ ID NO. 2.

[0023] In some instances, the sample is peripheral blood.

[0024] In some instances, the subjects were from East Asian populations.

[0025] These features can be combined arbitrarily as long as they do not conflict with each other.

[0026] The beneficial effects of this invention are: The system for predicting cholestasis of pregnancy according to this invention has the following advantages: (1) Zero dependence on biochemical indicators, enabling early risk assessment of ICP Existing conventional clinical risk assessment techniques heavily rely on pruritus symptoms and abnormal serum total bile acid (TBA) levels in the mid-to-late stages of pregnancy, facing the challenge of clinical lag. Furthermore, biochemical indicators are easily affected by the patient's physiological and pathological state, as well as the testing equipment. This invention, however, focuses solely on genetics, eliminating the need for complex clinical biochemical tests on pregnant women to assess ICP risk. Individual nucleotide polymorphism (SNP) test results do not change with physiological state, overcoming the clinical lag inherent in existing technologies. Low-cost early risk screening can be conducted in early pregnancy or even before conception, providing objective data support for early intervention, personalized health management, and the rational allocation of medical resources for pregnant women at high risk of ICP.

[0027] (2) Very few variables, simple model, easy to generalize This invention constructs a simplified ICP risk assessment model based on two variables: rs2296651 and Age. Large-sample clinical validation shows that this combined model achieves an AUC of 0.805 in assessing ICP risk. Studies confirm that after removing pre-pregnancy BMI, principal components (PCs), and other multivariates, this simplified model's assessment efficacy is even superior to complex assessment models incorporating multiple SNP loci and multiple variables (AUC range 0.777-0.803). This model relies on very few variables, simplifies data collection, and lowers the threshold for clinical application and testing costs. This technical solution is clear, easy to operate, and has extremely high clinical application prospects and value for promotion at the grassroots level.

[0028] (3) This is the first time that the clinical application of SNP loci specific to East Asian populations has been filled. Currently, research on ICP in the field of genetics is limited to etiological exploration and epidemiology. This invention is the first to successfully transform an East Asian-specific ICP susceptibility locus (rs2296651) into a tool for assessing individual clinical risk. The established mathematical risk assessment model not only provides a cost-effective and easily implemented approach for ICP screening in the general population at the grassroots level, but also buys time for subsequent early clinical intervention. Attached Figure Description

[0029] Figure 1 The effectiveness of different risk assessment models in evaluating ICPs is discussed.

[0030] Figure 2 The effectiveness of different variable combination prediction models in evaluating ICP is assessed. Detailed Implementation

[0031] The technical solution of the present invention will be further illustrated below with examples. The direction of base mutation in the core SNP:

[0032] The core SNP site rs22922651 is a mutation of C (cytosine) base or its complementary base G (guanine) to T (thymine) or its complementary base A (adenine).

[0033] To facilitate evaluation and comparison, in addition to the core SNP site rs2296651, this proposal also includes another SNP site (rs147525203) that our team discovered is associated with bile acids and ICP as a candidate for evaluation.

[0034] In this embodiment, the method for obtaining the rs2296651 genotype in the sample of the pregnant woman to be tested is PCR amplification. Primers are used to specifically amplify the rs2296651 locus of the SLC10A1 gene and its flanking sequences. The nucleotide sequences of the PCR primer pairs are as follows: (1) Forward Primer: 5' - GGCAATGAGGAGAAGCCCTT- 3' (2) Reverse amplification primer: 5'-CAGCACTGGGACAAAGTTGC- 3' Amplification parameters: primer Tm value is approximately 60℃, and the expected amplification product length is 221 bp.

[0035] Verification results: Preliminary experiments have verified that the above primers exhibit good amplification specificity at an annealing temperature of 60℃, with no primer dimer interference, and the sequencing results can accurately distinguish the wild-type (CC or GG), heterozygous (CT or GA), and mutant (TT or AA) genotypes of the subjects.

[0036] Using the primers described above, PCR amplification was performed on the genomic DNA of the sample to be tested. After obtaining the target amplification product, the specific base sequence of the target site was read by Sanger sequencing or time-of-flight mass spectrometry (MALDI-TOF) to determine whether the subject's genotype at that site was wild-type homozygous (CC or GG), heterozygous mutant (CT or GA), or mutant homozygous (TT or AA).

[0037] Those skilled in the art will understand that the primer sequence described above is only one preferred embodiment of the present invention. Other primer sequences designed using conventional primer design software based on the rs2296651 site and its flanking sequences that can achieve specific amplification of this site (e.g., sequences with more than 90% homology to the above sequence, or sequences containing the core fragment of the above sequence) are all included within the scope of protection of the present invention. Determination of clinical covariates

[0038] Baseline characteristics of pregnant women, including age, pre-pregnancy BMI, and principal components (PCs), were jointly assessed. Through logistic regression analysis, the maternal age was selected as the covariate for the ICP risk assessment model based on the AUC value of the model. Establishment and verification of mathematical models

[0039] 1) Based on data from the Guangzhou Birth Cohort, this study included pregnant women who underwent routine pre-pregnancy and prenatal checkups at the Women and Children's Medical Center of Guangzhou Medical University between February 2012 and June 2025, and were fully followed up until delivery. These women joined the Guangzhou Birth Cohort before 20 weeks of gestation and planned to reside in Guangzhou for the next three years. According to the ICP case definition: pregnant women experiencing pruritus and fasting TBA concentrations >10 μmol / L during the same period, a total of 39 cases and 4068 controls were included. Data on age, height, and pre-pregnancy weight were obtained from all enrolled pregnant women, and biological samples were collected for gene sequencing to determine the genotype of each sample at two candidate loci.

[0040] 2) Based on the chromosomal location of SNP sites, the Tabix tool is used to extract the corresponding variants from the whole genome data of the samples and generate VCF files. The Python script is used to convert the genotype data into numerical data and assign values: 0 for no mutation, 1 for heterozygous mutation, and 2 for homozygous mutation, while retaining the sample ID.

[0041] 3) Based on the case definition of ICP, use a Python script to select samples that have both an ID and whether they have ICP. Then, unify whether they have ICP into numerical data, assigning a value of 0 to those who do not have ICP and a value of 1 to those who have ICP.

[0042] 4) Comparing the variable distributions of the ICP group and the control group, it can be found that the differences between the two candidate SNP sites in the ICP group and the control group are statistically significant (P values ​​< 0.001).

[0043] 5) Univariate logistic regression analysis was performed on rs2296651 and rs147525203, and it was found that both candidate SNP sites showed a risk effect in both the ICP group and the control group, with P values ​​< 0.001.

[0044] 6) Multivariate logistic regression analysis was performed on all candidate variables rs2296651, rs147525203, preBMI, Age, and PCs (1-5). It was found that rs147525203 lost its independent statistical significance. Linkage disequilibrium (LD) analysis was then performed on rs2296651 and rs147525203, and the correlation coefficient R between them was calculated. 2 The correlation coefficient was 0.36, indicating linkage disequilibrium. Therefore, only rs2296651 was retained as the sole SNP site for inclusion in the ICP core risk assessment model.

[0045] To verify the superiority of the model, this study constructed risk assessment models with seven different combinations of variables and conducted ROC curve prediction performance analysis (see appendix for details). Figure 1 (See attached) Figure 1 As shown, with the continuous addition of clinical covariates (such as pre-pregnancy BMI and population principal components PCs) and redundant SNP loci, the AUC values ​​of complex models (Model 3 to Model 7) fluctuated between 0.777 and 0.803. In contrast, the optimal model Model 2 (rs2296651 and Age) selected in this invention, after excluding the interference of clinical data, achieved the most efficient AUC value of 0.805 for predicting ICP. Therefore, after eliminating other redundant variables, a mathematical model for the final assessment of ICP risk is established: Logit(p=ICP)=-7.386 +2.425×rs2296651+0.048×Age In the formula, rs2296651 is the genotype assignment (wild type is assigned 0, heterozygous mutation is assigned 1, homozygous mutation is assigned 2), Age is the pregnant woman's age (years), and a P value greater than 0.024 is defined as high risk, and vice versa.

[0046] Validated in a large clinical cohort, this risk assessment model demonstrated an overall discrimination (AUC) of 0.805 for predicting intracranial cholestasis of pregnancy (ICP). At the optimal cutoff value of 0.024, the model's negative predictive value (NPV) reached 99.6%, meaning that when the model determines a low risk, the individual has a very high probability of not having ICP, demonstrating excellent value for early clinical screening.

[0047] To further verify the superiority of the Model 2 (rs2296651 + Age) joint assessment model proposed in this invention, this embodiment compares the receiver operating characteristic (ROC) curves of the single-indicator model and the joint prediction model, and uses the DeLong test to perform a statistical significance analysis of the area under each curve (AUC). See attached... Figure 2As shown, the AUC value of including only Age was 0.528, indicating no significant predictive ability for risk. The AUC value of including only rs2296651 was 0.777, demonstrating that this SNP locus has the ability to independently predict ICP risk. However, the AUC values ​​of both models were smaller than those of the combined rs2296651 and Age assessment model (AUC value = 0.805). The DeLong test results (Appendix Table 1) show that the combined model's power was significantly better than Age assessment and rs2296651 assessment alone (P < 0.05), proving that introducing the rs2296651 genetic indicator can greatly compensate for the lack of accuracy in traditional demographic screening. Although the inclusion of Age in this population did not produce a significant statistical difference in the AUC of the combined model (P = 0.277), its technical significance lies in avoiding the defect of mechanically setting the 25-34 age group as the reference group in existing studies.

[10] For high-risk pregnant women who do not carry the rs2296651 mutation but are at very advanced or very young age (<25 years or >35 years), the combined model can provide risk warnings through age. Moreover, maternal age, as an important and universally available population baseline information in routine screening, improves the model's inclusiveness of complex clinical situations without significantly increasing testing costs.

[0048] Table 1. Performance parameters and DeLong test results of prediction models with different variable combinations Age only vs. rs2296651 0.528 0.777 0.0004 Age only vs rs2296651 + age 0.528 0.805 1.49494E-06 rs2296651 vs rs2296651+ age only 0.777 0.805 0.2767 References: [1] Obstetrics Group, Obstetrics and Gynecology Branch, Chinese Medical Association; Perinatal Medicine Branch, Chinese Medical Association. Chinese Medical Journal Co., Ltd., 2024. Clinical Diagnosis and Management Guidelines for Intrahepatic Cholestasis of Pregnancy (2024 Edition) [J]. Chinese Journal of Obstetrics and Gynecology, 2024, 59(2): 97-107.

[0049] [2]Ren Y, Shan

[0050] [3]He J, Zhu X, Yang X, Wang H. Predictive efficacy of machine-learning algorithms on intrahepatic cholestasis of pregnancy based on clinical and laboratory indicators[J]. Journal of Maternal-Fetal and NeonatalMedicine, 2025, 38(1): 2413854.

[0051] [4]Asali A. Intrahepatic cholestasis of pregnancy: machine-learning algorithm to predict elevated bile acid based on clinical and laboratory data[J]. Archives of Gynecology and Obstetrics: 1-7.

[0052] [5] Wu Wei, Yu Jingwen, Jiang Hui. Research on the establishment of a nomogram model of risk of intrahepatic cholestasis of pregnancy in pregnant women [J]. Journal of Practical Clinical Medicine, 2022, 26(4): 18-23.

[0053] [6] Jiang Yuwei. Establishment and validation of an early risk prediction model for cholestasis of pregnancy [D]. Anhui Medical University, 2025. DOI:10.26921 / d.cnki.ganyu.2025.001084.

[0054] [7] Gong Qingquan, Liu Yanqing, Yuan Yuhong, Qi Hongbo, Shi Qi. Establishment and validation of an early prediction model for intrahepatic cholestasis of pregnancy [J]. Chinese Journal of Family Planning and Obstetrics and Gynecology, 2024, 16(6): 93-98, 114.

[0055] [8]Obut M, Kından A, Ibanoğlu M C, Çayönü Kahraman N, Arat Ö, KeleşA, Topkara S, Tokgöz Çakır B, Bucak M, İskender C T. Liver damage parametersand peripheral blood parameters for prediction and diagnosis of intrahepaticcholestasis in pregnancy[J]. Journal of Obstetrics and Gynaecology Research,2024, 50(2): 196-204.

[0056] [9]Peker A, Tanaçan A, İpek G, Ağaoğlu Z, Şahin D. Role of aspartateaminotransferase to platelet ratio in the prediction and prognosis ofintrahepatic cholestasis of pregnancy: a case–control study from a tertiarycenter[J]. International Journal of Gynecology & Obstetrics, 2024, 164(2):656-661.

[0057]

[10] Gao X-X, Ye M-Y, Liu Y, Li J-Y, Li L, Chen W, Lu X, Nie G, ChenY-H. Prevalence and risk factors of intrahepatic cholestasis of pregnancy ina chinese population[J]. Scientific Reports, 2020, 10(1): 16307.

[0058]

[11] Chen L, Li J, You Y, Qian Z, Liu J, Jiang Y, Gu Y, Xiao J, ZhangY. Secreted proteins in plasma and placenta as novel non-invasive biomarkersfor intrahepatic cholestasis of pregnancy: a case-control study[J]. Heliyon,2023, 9(11): e21616.

[0059]

[12] Zou S, Dong R, Wang J, Liang F, Zhu T, Zhao S, Zhang Y, Wang T,Zou P, Li N, Wang Y, Chen M, Zhou C, Zhang T, Luo L. Use of data-independentacquisition mass spectrometry for comparative proteomics analyses of serafrom pregnant women with intrahepatic cholestasis of pregnancy[J]. Journal ofProteomics, 2021, 236: 104124.

[0060]

[13] Feng F, Jiang M, Liu W, Xu D, Huang X-M, Shao Y. Study on plasmaexosome biomarkers of pregnant women with intrahepatic cholestasis ofpregnancy[J]. Scandinavian Journal of Gastroenterology, 2022, 57(6): 696-704.

[0061]

[14] Yurtcu N, Caliskan C S, Guvey H, Celik S, Hatirnaz S, Tinelli A.Predictive and diagnostic value of serum adipokines in pregnant women withintrahepatic cholestasis[J]. International Journal of Environmental Researchand Public Health, 2022, 19(4): 2254.

[0062]

[15] Ertaş S, Ünver G, Gonca Okumuş Z, Yılmaz S, Dağ İ, Turhan U.Maternal serum sortilin-1 level as a potential biomarker for intrahepaticcholestasis of pregnancy[J]. Gynecological Endocrinology, 2022, 38(11): 935-938.

[0063]

[16] Zou P, Luo L, Zhao C, Chen Z, Dong R, Li N, Wang Y, Wang J, WangT, Chen M, Zhang T, Chen D. The serum microRNA profile of intrahepaticcholestasis of pregnancy: identification of novel noninvasive biomarkers[J].Cellular Physiology and Biochemistry, 2018, 51(3): 1481-1488.

[0064]

[17] Dong R, Ye N, Wang J, Zhao S, Wang T, Wang G, Shi X, Cheng J, Zhang Y, Yao T, Chen M, Zhang T, Luo L. Serum exosomes MicroRNAs are novelnon-invasive biomarkers of intrahepatic cholestasis of pregnancy[J]. Frontiers in endocrinology, 2022, 13: 832577.

[0065]

[18] Zu Y, Guo S, Li G, Gao Q, Wang X, Zhang C, Liu D. Serum microRNAsas non-invasive diagnostic biomarkers for intrahepatic cholestasis ofpregnancy[J]American Journal of Translational Research, 2022,14(9):6763-6773.

[0066]

[19] Kong Y, Zhan Y, Chen D, Deng X, Liu X, Xu T, Wang X. UniquemicroRNA expression profiles in plasmic exosomes from intrahepaticcholestasis of pregnancy[J]. BMC Pregnancy and Childbirth, 2023, 23(1): 147-158.

[0067]

[20] Li Minxia, ​​Chen Hongmei, Guo Jingjing, He Wenjuan. Expression and clinical predictive value of MicroRNA-148a in intrahepatic cholestasis of pregnancy [J]. Chinese Journal of Modern Medicine, 2024, 34(4): 17-22.

[0068]

[21] Dong R, Ye N, Zhao S, Wang G, Zhang Y, Wang T, Zou P, Wang J, YaoT, Chen M, Zhou C, Zhang T, Luo L. Studies on novel diagnostic and predictivebiomarkers of intrahepatic cholestasis of pregnancy through metabolomics andproteomics[J]. Frontiers in Immunology, 2021, 12: 733225.

[0069]

[22] Chen X, Zhang X, Xu B, Cui Y, He Y, Yang T, Shao Y, Ding M. Theurinary bile acid profiling analysis of asymptomatic hypercholanemia ofpregnancy: a pseudo-targeted metabolomics study[J]. Clinica Chimica Acta,2019, 497: 67-75.

[0070]

[23] Huang S, Liu S, Huang M, He JR, Wang C, Wang T, Feng X, Kuang Y,Lu J, Gu Y, Xia X, Lin S, Born In Guangzhou Cohort Study (Bigcs) Group, ShenS, Liu X, Lu M, Yuan M, Xiao W, Zhou F, Wei D, Chen N, Wei Y, Deng L, Liu W, Shao D, Yang Y,Li Y, Zhou W, Fu Q, Xia H, Qiu X. The born in Guangzhou cohort study enables generational genetic discoveries[J]. Nature, 2024, 626(7999): 565-573.

[0071]

[24] Liu Y, Wei Y, Chen X, Huang S, Gu Y, Yang Z, Guo X, Zheng H, FengH, Huang M, Chen S, Xiao T, Hu L, Zhang Q, Zhang Y, Chen GB, Qiu Hepatology, 2025, 82(5): 826-835.

[0072] The above is a further detailed description of the present invention and should not be considered as a limitation on the specific implementation of the present invention. For those skilled in the art, simple deductions or substitutions without departing from the concept of the present invention are all within the protection scope of the present invention.

Claims

1. A system for predicting cholestasis of pregnancy, characterized in that, include: The data acquisition module is used to collect the genotype of the rs2296651 locus and the age of the subject in the subject's sample. The mutant base of rs2296651 is T and its complementary base A, and the wild-type base is C and its complementary base G. The genotype is wild-type homozygous, heterozygous mutant, or mutant homozygous. The data analysis module determines the risk of cholestasis of pregnancy based on the genotype of the rs2296651 locus in the subject samples and the age of the subjects. The results output module is used to output the analysis results from the data analysis module.

2. The system according to claim 1, characterized in that, The mathematical expression of the risk assessment model of the data analysis module is: P=β0+β1×rs2296651+β2×Age; where P is the probability of ICP incidence; β0 is a constant term, β1 and β2 are regression coefficients, rs2296651 is the genotype assignment: 0 when the subject is wild-type homozygous, 1 when it is heterozygous mutation, and 2 when it is mutant homozygous; Age is the actual age of the pregnant woman in years.

3. The system according to claim 2, characterized in that, Based on known samples, machine learning classification algorithms are used to determine the values ​​of β0, β1, and β2.

4. The system according to claim 3, characterized in that, The machine learning classification algorithm includes any one or more combinations of logistic regression, support vector machine, decision tree, random forest, gradient boosting tree, artificial neural network, Naive Bayes or K-nearest neighbor algorithm.

5. The system according to claim 4, characterized in that, The machine learning classification algorithm is a logistic regression algorithm.

6. The system according to claim 2, characterized in that... β0 = -7.386, β1 = 2.425 and β2 = 0.048; when the calculated P value is greater than the threshold of 0.024, it is assessed as high risk of ICP, otherwise it is low risk.

7. The system according to claim 1, characterized in that, The genotype of the rs2296651 locus was determined by PCR amplification or probe.

8. The system according to claim 7, characterized in that, The primer pairs used for PCR amplification are: (1) forward primer, whose nucleotide sequence is shown in SEQ ID NO.1; and (2) reverse primer, whose nucleotide sequence is shown in SEQ ID NO.

2.

9. The system according to claim 1, characterized in that, The sample was peripheral blood.

10. The system according to claim 1, characterized in that, The subjects were from East Asian populations.

Images