Gene marker combination and use thereof in early-onset preeclampsia risk prediction model

Abstract

The present application provides a gene marker combination and a use thereof in an early-onset preeclampsia risk prediction model. The gene marker combination comprises at least one selected from a combination of GTF2H1, CD48, PBX1, PLA2G2A, ACVR1B, HMGB1, ERBB2, XRCC5, TGFBR2, NFKB1, EGF, TLR6, EZH2, IKZF1, EGFR, and CD36, a combination of HIPK1, MTHFR, IL6R, RAB27A, UBE2I, CREBBP, SOX8, HNF4A, ICOS, LIF, TMPRSS6, DNMT3A, WWTR1, RASSF1, FOXP1, TLR2, and MAPK14, a combination of AGRN, ATP2B1, ELAC2, HDAC5, GABARAP, FXR2, NFE2L2, DGCR2, MAPK1, CASR, MYLIP, TUBB, DOCK4, PTGS1, HSPA5, ABL1, and SYK, or a combination of chr1:110,000,000-115,000,000, chr4:150,000,000-155,000,000, chr6:160,000,000-165,000,000, chr6:60,000,000-65,000,000, and chr7:125,000,000-130,000,000.

Description

Gene marker combination and its application in the risk prediction model for early-onset preeclampsia Technical Field

[0001] The present application relates to the field of bioinformatics. Specifically, the present application relates to a gene marker combination and its application in a risk prediction model for early-onset preeclampsia. Background Art

[0002] Preeclampsia is a pregnancy-related disorder that typically develops after 20 weeks of gestation and presents as new-onset hypertension accompanied by proteinuria or damage to other organs (such as the kidneys, liver, or brain). This disorder and its accompanying hypertension are among the leading causes of maternal mortality. Currently, early termination of pregnancy and delivery of the placenta are the main treatments. However, premature delivery can lead to immature fetal organs and various adverse consequences.

[0003] The incidence of preeclampsia is approximately 2% to 8% worldwide, and approximately 2.3% in my country. This disease is divided into early-onset and late-onset preeclampsia based on the time of onset. Currently, clinical screening often uses a pregnant woman's medical history (including a history of eclampsia, chronic hypertension, diabetes, etc.), age, body mass index (BMI), mode of pregnancy, and blood pressure as risk factors. However, using these clinical risk indicators alone is generally inaccurate and prone to false positives.

[0004] Recent studies have shown that the genome-wide promoter profile of cell-free DNA (cfDNA) in maternal peripheral blood can reflect intracellular gene expression, providing a new approach for predicting pregnancy-related diseases. However, existing prediction methods have low accuracy.

[0005] Therefore, new prediction methods need to be developed to solve the above problems.

[0006] Summary of the Invention

[0007] This application is made based on the inventor's discovery of the following problems and facts:

[0008] Although existing patents have made progress in the field of preeclampsia prediction (CN110305954A, CN114822682A), there are still problems such as insufficient subtype classification, insufficient comprehensiveness of features, and limited number of samples.

[0009] Therefore, it is necessary to develop more accurate and effective new detection methods that comprehensively consider the omics characteristics of multiple cfDNA fragments to improve the accuracy and universality of prediction models. Therefore, this application proposes a method for predicting the risk of early-onset preeclampsia using genetic markers combined with machine learning models.

[0010] Specifically, this application provides the following technical solutions:

[0011] In the first aspect of the present application, the present application proposes a gene marker combination. According to an embodiment of the present application, the gene marker combination includes GTF2H1, CD48, PBX1, PLA2G2A, ACVR1B, HMGB1, ERBB2, XRCC5, TGFBR2, NFKB1, EGF, TLR6, EZH2, IKZF1, EGFR, CD36, HIPK1, MTHFR, IL6R, RAB27A, UBE2I, CREBBP, SOX8, HNF4A, ICOS, LIF, TMPRSS6, DNMT3A, WWTR1, RASSF1, FOXP1, TLR2, MAPK14, AGRN , ATP2B1, ELAC2, HDAC5, GABARAP, FXR2, NFE2L2, DGCR2, MAPK1, CASR, MYLIP, TUBB, DOCK4, PTGS1, HSPA5, ABL1, SYK, chr1:1100000000-115000000, chr4:150000000-155000000, chr6:160000000-165000000, chr6:60000000-65000000 and chr7:1250000000-130000000. The inventors analyzed plasma cfDNA sequencing data from early pregnancy and screened for a target gene and target region combination associated with the occurrence of early-onset eclampsia, which can be used as a marker combination for effectively screening for early-onset preeclampsia.

[0012] In the second aspect of the present application, a kit is provided. According to an embodiment of the present application, the kit includes reagents for detecting at least one of the abundance of cfDNA fragments in the promoter region of transcripts, the distribution spectrum of cfDNA fragment abundance, the distribution spectrum of cfDNA fragment length, and the distribution spectrum of cfDNA fragment length in a 5M window of the genome, for the combination of gene markers described in the first aspect of the present application. In some examples of the present application, the reagents in the kit can be used to detect early-onset preeclampsia in some ex vivo samples.

[0013] According to an embodiment of the present application, the above-mentioned kit may further include at least one of the following technical features:

[0014] According to an embodiment of the present application, the kit further includes a reference dataset of the gene marker combination described in the first aspect of the present application. By comparing the sample test results with the reference data results, the risk prediction of early-onset preeclampsia of the sample can be determined.

[0015] The so-called reference dataset refers to standard data that has been verified by sequencing. It can be a DNA and / or RNA dataset that has been pre-determined and assembled by oneself, or a DNA and / or RNA dataset that has been publicly determined by others. It can be any reference template in the biological category to which the sample source individual / target individual belongs, for example, all or at least part of the publicly available genome assembly sequence of the same biological category.

[0016] In some examples of the present application, the above-mentioned kit also includes a sample processing solution and a kit instruction manual, etc.

[0017] In any of the above aspects or embodiments, the risk prediction of early-onset preeclampsia can be achieved by preparing primers or probes for the above-mentioned target genes and target regions, as well as other related reagents that use the above-mentioned cfDNA fragment omics characteristics as the amplification target fragments.

[0018] In a third aspect of this application, the present application proposes the use of the kit described in the second aspect for detecting early-onset preeclampsia, assessing the prognosis of early-onset preeclampsia, or providing medication guidance. In some examples of this application, the kit can be used to detect the risk of early-onset preeclampsia and to specify an individualized treatment plan based on the test results.

[0019] In the fourth aspect of the present application, the present application proposes a method for training a machine learning model, which is used to predict the risk of early-onset preeclampsia. According to an embodiment of the present application, the method includes: obtaining feature information of a training data set, the feature information including text features and cfDNA fragment features of the gene marker combination described in the first aspect of the present application; inputting the feature information into the machine learning model, and using the known risk of early-onset preeclampsia of the training sample as a marker to perform supervised training on the machine learning model. This method provides a comprehensive analysis of multi-source data by integrating the text features of the training samples and the cfDNA fragment features of the gene marker combination, thereby improving the accuracy and precision of predicting the risk of early-onset preeclampsia.

[0020] According to an embodiment of the present application, the above-mentioned machine learning model training method may further include at least one of the following technical features:

[0021] According to an embodiment of the present application, the text feature includes at least one selected from the group consisting of the pregnant woman's age, body mass index (BMI), mean arterial pressure, past medical history, and pregnancy pattern.

[0022] According to an embodiment of the present application, the cfDNA fragmentation feature includes at least one selected from the group consisting of the abundance of cfDNA fragments in the promoter region of the transcript, the distribution spectrum of cfDNA fragment abundance, the distribution spectrum of cfDNA fragment length, and the distribution spectrum of cfDNA fragment length in a 5M window of the genome. The inventors conducted an in-depth analysis of a large amount of sequencing data from healthy samples and early-onset preeclampsia samples and identified features that showed stable differences between healthy samples and early-onset preeclampsia samples.

[0023] In any aspect or embodiment of the present application, the transcript promoter region is selected from the region 1 kb upstream to 1 kb downstream of the transcription start site of each transcript.

[0024] According to an embodiment of the present application, the cfDNA fragment characteristics of the gene marker combination include the abundance of cfDNA fragments in the promoter regions of GTF2H1, CD48, PBX1, PLA2G2A, ACVR1B, HMGB1, ERBB2, XRCC5, TGFBR2, NFKB1, EGF, TLR6, EZH2, IKZF1, EGFR and CD36 transcripts, and the distribution spectrum of the abundance of cfDNA fragments in the promoter regions of HIPK1, MTHFR, IL6R, RAB27A, UBE2I, CREBBP, SOX8, HNF4A, ICOS, LIF, TMPRSS6, DNMT3A, WWTR1, RASSF1, FOXP1, TLR2 and MAPK14 transcripts. , AGRN, ATP2B1, ELAC2, HDAC5, GABARAP, FXR2, NFE2L2, DGCR2, MAPK1, CASR, MYLIP, TUBB, DOCK4, PTGS1, HSPA5, ABL1 and SYK transcript promoter regions, and at least one of the distribution profiles of cfDNA fragment lengths on chr1:1100000000-115000000, chr4:1500000000-1550000000, chr6:1600000000-1650000000, chr6:60000000-65000000 and chr7:1250000000-130000000 In order to reduce the calculation of the model and increase the representativeness of the features, the inventors combined the target genes and target regions with features with stable differences between healthy samples and early-onset preeclampsia samples, and finally obtained the above-mentioned representative features.

[0025] According to an embodiment of the present application, the abundance of cfDNA fragments in the promoter region of the transcript is calculated by the following formula: pTSS icoverage=(reads number mapped to pTSS i ×10 9 ) / (total reads × 2000), formula (1);

[0026] Among them, pTSS i Coverage indicates the abundance of cfDNA fragments in the promoter region of transcript i; read number mapped to pTSS i Represents the number of sequencing reads that match the promoter region of transcript i; total reads represents the number of valid reads in the sequencing data.

[0027] According to an embodiment of the present application, the distribution spectrum of cfDNA fragment abundance in the promoter region of the transcript is calculated by the following formula:

[0028] Among them, pTSS i "score" represents the distribution of cfDNA fragment abundance within the promoter region of transcript i; "depth of midBin" represents the coverage of the Z windows closest to the transcription start site, where Z represents the number of windows; "depth of sideBin" represents the coverage of the W windows farthest from the transcription start site, where W represents the number of windows; "x / y" represents the length of the windows bisecting the transcript promoter region; and "total reads" represents the number of valid reads in the sequencing data. The distribution of cfDNA fragment abundance within the transcript promoter region is a key feature first discovered by the inventors for machine learning models that can be used to predict the risk of early-onset preeclampsia.

[0029] According to an embodiment of the present application, the value of Z is selected from any integer between 3 and 9. For example, the value of Z can be 3, 4, 5, 6, 7, 8, or 9. In some preferred examples of the present application, the value of Z is 6.

[0030] According to an embodiment of the present application, the value of W is selected from any integer between 8 and 15. For example, the value of W can be 8, 9, 10, 11, 12, 13, 14, or 15. In some preferred examples of the present application, the value of W is 10.

[0031] According to an embodiment of the present application, the value of x / y is selected from any integer between 80 and 120. For example, the value of x / y can be 80, 90, 100, 110, or 120. In some preferred examples of the present application, the value of x / y is 100.

[0032] According to an embodiment of the present application, the distribution spectrum of the length of cfDNA fragments on the transcript promoter region is obtained by analyzing the dispersion of the length of the cfDNA fragments covering each transcript promoter region within a first predetermined range. Specifically, within a specific predetermined range (100 to 300 bp), by observing and comparing the length of the cfDNA fragments on each transcript promoter region, the dispersion of the lengths of these fragments within the range (i.e., the distribution of the lengths) is analyzed.

[0033] According to an embodiment of the present application, the first predetermined range is selected from 100 to 300 bp. In some examples of the present application, the first predetermined range is optionally 100 bp, 150 bp, 200 bp, 250 bp, or 300 bp.

[0034] According to embodiments of the present application, the distribution profile of cfDNA fragment lengths in a 5M window of the genome is obtained by: a. dividing the genome into 5M bp windows; b. calculating the ratio of the number of cfDNA fragments in each window within a first predetermined range and a second predetermined range. In some examples of the present application, the cfDNA fragments are GC-corrected.

[0035] According to an embodiment of the present application, the second predetermined range is selected from 100 to 150 bp.

[0036] According to an embodiment of the present application, the third predetermined range is selected from 151 to 220 bp.

[0037] According to an embodiment of the present application, the machine learning model is selected from at least one of a random forest, a decision tree, and a neural network. In some preferred embodiments of the present application, the machine learning model is selected from a random forest.

[0038] In a fifth aspect of the present application, a prediction model for early-onset preeclampsia is proposed. According to an embodiment of the present application, the prediction model includes: obtaining sequencing data of a sample to be tested; obtaining feature information of the sequencing data; and inputting the feature information into a pre-trained machine learning model to obtain a prediction result; wherein the pre-trained machine learning model is trained using the method described in the fourth aspect of the present application. In some examples of the present application, the prediction model is used to predict the risk of early-onset preeclampsia.

[0039] According to an embodiment of the present application, the above prediction model may further include at least one of the following technical features:

[0040] According to an embodiment of the present application, the sample to be tested is selected from at least one of plasma, whole blood, amniotic fluid, serum or urine of a pregnant woman.

[0041] According to an embodiment of the present application, the prediction result is determined as follows: if the machine learning model (or prediction model) outputs a result of 0, the sample is judged to be at low risk for early-onset preeclampsia; if the machine learning model (or prediction model) outputs a result of 1, the sample is judged to be at high risk for early-onset preeclampsia. In some examples of the present application, the prediction model can be used to efficiently and accurately determine the risk of early-onset preeclampsia, allowing for early intervention.

[0042] In the sixth aspect of the present application, the present application proposes a machine learning model training device, wherein the machine learning model is used to predict the risk of occurrence of early-onset preeclampsia. According to an embodiment of the present application, the device includes: a feature acquisition unit for acquiring feature information of a training data set, wherein the feature information includes text features and cfDNA fragment features of the gene marker combination described in the first aspect of the present application; and a training unit for inputting the feature information into the machine learning model, using the known risk of occurrence of early-onset preeclampsia of the training sample as a label, so as to supervise the training of the machine learning model. In some examples of the present application, the device can automatically extract text features and cfDNA fragment features of the gene marker combination in the training data set. Subsequently, the training unit inputs these features into the machine learning model, uses the risk of occurrence of early-onset preeclampsia of the known sample as a label, and performs supervised training. This makes the training of the model more efficient while ensuring the accuracy of risk prediction.

[0043] The structural diagram of the so-called model training device is shown in FIG1 , where the feature acquisition unit S100 is connected to the training unit S200 .

[0044] In the seventh aspect of the present application, the present application proposes a prediction model for early-onset preeclampsia. According to an embodiment of the present application, the model includes: obtaining sequencing data of a sample to be tested; obtaining feature information of the sequencing data; inputting the feature information into a pre-trained machine learning model to obtain a prediction result; wherein the pre-trained machine learning model is trained using the method described in the fourth aspect of the present application. This model combines multiple cfDNA fragment omics features with clinical indicators as model input values, and its prediction accuracy is superior to existing early-onset preeclampsia prediction models.

[0045] According to an embodiment of the present application, the above-mentioned early-onset preeclampsia prediction model may further include at least one of the following technical features:

[0046] According to an embodiment of the present application, the sample to be tested is selected from at least one of plasma, whole blood, amniotic fluid, serum or urine of a pregnant woman.

[0047] According to an embodiment of the present application, the prediction result is determined as follows: if the output result of the machine learning model is 0, the sample is judged to be at low risk for early-onset preeclampsia; if the output result of the machine learning model is 1, the sample is judged to be at high risk for early-onset preeclampsia.

[0048] In an eighth aspect of the present application, an electronic device is provided. According to an embodiment of the present application, the electronic device includes: a memory and a processor, the memory being configured to store a computer program; and the processor being configured to execute the computer program to implement the method described in the fourth aspect of the present application. In some examples of the present application, the processor reads executable program code stored in the memory to execute a program corresponding to the executable program code.

[0049] In a ninth aspect of the present application, the present application provides a computer-readable storage medium. According to an embodiment of the present application, the computer-readable storage medium stores computer program instructions, which, when executed on a processor, cause the processor to execute the method according to the fourth aspect of the present application.

[0050] It should be noted that the various embodiments of the methods and models described above can be implemented in digital electronic circuit systems, integrated circuit systems, FPGAs (Field Programmable Gate Arrays), ASICs (Application-Specific Integrated Circuits), ASSPs (Application Specific Standard Products), SOCs (System on Chips), CPLDs (Complex Programmable Logic Devices), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special-purpose or general-purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0051] The program code for implementing the method disclosed in the present application can be written in any combination of one or more programming languages. Such program code can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device so that when the program code is executed by the processor or controller, the functions / operations specified in the flow chart and / or block diagram are implemented. The program code can be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0052] In the context disclosed herein, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or apparatus. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or apparatus, or any suitable combination of the foregoing. More specific examples of machine-readable storage media may include an electrical connection based on one or more wires, a portable computer disk, a hard disk, RAM, ROM, EPROM (Electrically Programmable Read-Only-Memory) or flash memory, optical fiber, CD-ROM (Compact Disc Read-Only Memory), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0053] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (Cathode-Ray Tube) or LCD (Liquid Crystal Display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0054] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: LAN (Local Area Network), WAN (Wide Area Network), the Internet, and blockchain networks.

[0055] A computer system may include a client and a server. The client and server are generally remote from each other and typically interact via a communication network. This client-server relationship is established by computer programs running on the respective computers and establishing a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, a host product within a cloud computing service ecosystem that addresses the management difficulties and limited scalability of traditional physical hosts and VPS services ("Virtual Private Servers" or "VPS"). The server may also be a server in a distributed system or a server integrated with blockchain.

[0056] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this disclosure can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved. This is not limited herein.

[0057] It should be noted that the features and technical effects described in this article for different aspects can be used as reference for each other and will not be repeated here.

[0058] Additional aspects and advantages of the present invention will be set forth in part in the description which follows and, in part, will be obvious from the description which follows, or may be learned by practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which:

[0060] FIG1 is a schematic diagram of a machine learning model training device according to an embodiment of the present application;

[0061] FIG2 is a schematic diagram of validation error results of analyzing different omics features using a training set and a validation set through ten-fold cross validation according to one embodiment of the present application. DETAILED DESCRIPTION

[0062] The following describes embodiments of the present invention in detail, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and are not to be construed as limiting the present invention.

[0063] Definition and Description

[0064] As used herein, unless otherwise indicated, the singular forms "a," "an," and the like include plural referents (more than one); "a set" or "a plurality" refers to two or more.

[0065] In this document, unless otherwise specified, the terms "first", "second", "third", "fourth", etc. are used for descriptive purposes only and are not to be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated; features specified as "first", "second", etc. may explicitly or implicitly include one or more of the said features.

[0066] As used herein, the term "gene marker panel" refers to a set of specific genes or gene fragments whose expression or mutational status in an organism is associated with a specific disease, biological process, or drug response. In this application, one use of the gene marker panel is to construct a prediction model for early-onset preeclampsia.

[0067] In this article, "predictive model" and "machine learning model" refer to computational models or algorithms that automatically perform tasks such as prediction, classification, recognition, or decision-making by learning and analyzing input data. The model's learning process is based on statistical principles and data pattern recognition, using a training set to adjust model parameters and optimize the model to improve its predictive or inference capabilities. Machine learning models can employ various algorithms and techniques, such as neural networks, support vector machines, decision trees, random forests, and deep learning. These models can be trained and optimized through supervised learning, unsupervised learning, or reinforcement learning. In practical applications, machine learning models are used in a variety of fields, such as natural language processing, image recognition, pattern recognition, data mining, recommender systems, and predictive analytics. They hold significant potential for large-scale data processing, automated decision-making, and intelligent systems. However, it should be noted that in the specific application of machine learning models, in-depth research on the features and models used for prediction is required to achieve satisfactory prediction results. Otherwise, various problems may arise, such as overfitting, underfitting, and poor generalization. After years of research and combined with research experience, the inventors of this application have discovered a predetermined feature combination that can be combined with a machine learning model to assess the risk of early-onset preeclampsia.

[0068] In this article, supervised learning is used to train the machine learning model. So-called "supervised learning" is one of the most common and widely used learning methods in machine learning. The model is trained using training data with correct answers (labels), allowing the model to learn the relationship between input and output from the data, thereby making predictions or classifications on unseen data. In supervised learning, each training sample consists of input data and a corresponding label. The label in the training sample indicates what output should be produced for the given input data. The goal of the model is to learn a function from the training data that can map the input data to the corresponding output label. This process can be seen as learning the mapping relationship between input and output, also known as the model's "hypothesis function" or "mapping function."

[0069] In this article, "supervised learning" is used to solve the classification problem, and the goal of the machine learning model is to map the input data to one of the predefined discrete categories. The training data contains a series of input samples and the category labels to which they belong. After training, the model is able to classify new input samples and assign them to the correct category. For example, the cfDNA fragment characteristics and clinical information of the sample to be tested are input into the early-onset preeclampsia prediction model obtained in this application. If the model output result is 0, the sample is judged to be at low risk for early-onset preeclampsia; if the model output result is 1, the sample is judged to be at high risk for early-onset preeclampsia.

[0070] The training of machine learning models requires a training set and a test set. The so-called "training set" is a data set used to train the machine learning model. It contains a series of input samples (feature vectors) and corresponding known outputs (labels or target values). During the training phase, the machine learning model uses the samples in the training set to learn and adjust the parameters of the model to minimize the error between the predicted results and the true labels. By constantly trying different parameters and algorithms, the model gradually learns the patterns and characteristics between the samples, thereby obtaining more accurate prediction capabilities. In this application, the inventors perform model selection and tuning and prevent the occurrence of overfitting by adding an internal validation set (Training Validation Set, part of the training set).

[0071] The so-called "test set" or external validation set is a dataset used to evaluate the performance of a machine learning model. It also contains input samples and their corresponding true output labels, but the samples in the test set are not used during the model training phase, ensuring that the test set data is unknown to the model. After training is complete, the test set is used to evaluate the model's performance. This involves making predictions for the samples in the test set and comparing them with the true labels. By comparing the predicted results with the true labels, the model's performance on unknown data is evaluated, and its generalization ability and prediction accuracy are determined.

[0072] Given the poor universality and insufficient sample size of existing preeclampsia prediction models, the inventors conducted in-depth research on the various characteristics of cfDNA fragment omics in the peripheral blood of pregnant women based on whole-genome sequencing data of cfDNA in early pregnancy. They innovatively discovered several cfDNA fragment omics features associated with early-onset preeclampsia, which showed significant differences between patients with early-onset preeclampsia and healthy controls. To this end, the inventors of this application combined the cfDNA fragment omics features of specific genes and specific regions with machine learning models, and ultimately obtained a prediction model with better universality and higher detection accuracy.

[0073] For ease of understanding, the process of constructing the early-onset eclampsia prediction model of this application is described in detail below.

[0074] 1. cfDNA fragment sequencing data analysis

[0075] 1.1 Sampling and Sequencing: First-trimester plasma samples were collected from pregnant women with early-onset preeclampsia (diagnosed <34 weeks) and healthy controls, and 20-30X cfDNA whole-genome sequencing was performed.

[0076] In this article, there is no particular limitation on the sequencing platform, and it can be any common first-generation, second-generation, or third-generation sequencing platform. In the embodiments of the present application, the sequencing platform that can be used to perform the above sequencing method can be selected from, but not limited to, Illumina's HiSeq / Miseq / Nextseq / Novaseq sequencing platforms, Thermo Fisher / Life Technologies' Ion Torrent platforms, BGISEQ and MGISEQ / DNBSEQ platforms, and single-molecule sequencing platforms.

[0077] In one example of this application, the MGIEasy cell-free DNA library preparation kit was used to construct a double-stranded DNA library from maternal plasma cell-free DNA. The double-stranded DNA libraries from 10 pregnant women were then pooled and the MGIEasy circularization kit was used to prepare a single-stranded circular DNA library for the matrix nanochip of the MGI sequencing platform. 20-30X double-ended 100bp sequencing data was obtained using the DNBSEQ-T7 sequencer.

[0078] 1.2 Data processing: The whole-genome sequencing data of early pregnancy plasma cfDNA obtained in 1.1 were aligned with the reference genome (e.g., GRCh38), and multiple matching results and PCR duplications were removed to obtain valid sequencing sequences.

[0079] The so-called reference genome is a determined sequence, which can be a DNA and / or RNA sequence pre-determined and assembled by oneself, or a DNA and / or RNA sequence publicly determined by others, or any reference template in the biological category to which the sample source individual / target individual belongs, for example, all or at least a portion of the publicly disclosed genome assembly sequence of the same biological category. If the sample source individual or target individual is human, its genome reference sequence (also referred to as a reference genome or reference chromosome group) can be selected from the human reference genome provided by the UCSC, NCBI or ENSEMBL databases, such as HG19, HG38, GRCh36, GRCh37, GRCh38, etc. Those skilled in the art can understand the correspondence between the above-mentioned reference genome versions through the description of the database and select the version to use. Furthermore, a resource library containing more reference sequences can also be pre-configured. For example, before performing the comparison, a sequence that is closer to or has more characteristics in a certain aspect can be selected or determined and assembled based on factors such as the gender, race, and region of the target individual to serve as a reference sequence, which helps to obtain more accurate sequence analysis results later.

[0080] The so-called reference sequence can be constructed when testing the target sample, or it can be pre-constructed and stored and called when testing a prepared sample. In some examples of this application, the sample to be tested is from the human body, and the reference sequence is a human reference genome or a human autosome group.

[0081] In one example of the present application, the number of effective sequencing sequences should be no less than 10X, optionally 20X, 30X, 40X or 50X.

[0082] 2. Obtaining cfDNA fragment omics characteristics

[0083] 2.1 cfDNA fragment omics feature 1 (abundance of cfDNA fragments in the promoter region of transcripts): Using the total number of valid sequencing sequences obtained in 1.2 as the normalization parameter, the abundance of cfDNA fragments in each transcript promoter region was calculated using formula (1), and sequencing depth correction was performed to obtain the cfDNA fragment abundance spectrum of each promoter region. These data were median-corrected in the same batch of data; pTSS i coverage=(reads number mapped to pTSS i ×10 9 ) / (total reads × 2000), formula (1);

[0084] Among them, pTSSi coverage represents the abundance of cfDNA fragments in the promoter region of transcript i; reads number mapped to pTSSi represents the number of sequencing reads matching the cfDNA fragments in the promoter region of transcript i; total reads represents the number of valid reads in the sequencing data.

[0085] It should be noted that in the screening of cfDNA fragment omics features in this application, the transcript promoter regions were selected from the region 1 kb upstream to 1 kb downstream of the transcription start site of each transcript.

[0086] 2.2 cfDNA fragment omics feature 2 (distribution profile of cfDNA fragment abundance in transcript promoter regions): Each transcript promoter region was divided into windows. The average coverage of the Z windows closest to the transcription start site and the W windows farthest from the transcription start site were calculated based on formula (2). The ratio of the two was normalized to obtain the cfDNA abundance distribution trend of each promoter region. Similarly, median correction was performed within the same batch of data.

[0087] Among them, pTSS iscore represents the distribution of cfDNA fragment abundance in the promoter region of transcript i; depth of midBin represents the coverage of the Z windows closest to the transcription start site; depth of sideBin represents the coverage of the W windows farthest from the transcription start site; x / y represents the length of the equally divided windows in the transcript promoter region; and total reads represents the number of valid reads in the sequencing data. The values ​​of Z and W can be customized based on experimental requirements.

[0088] For example, the 2 kb promoter region of each transcript in 2.1 is divided into 20 equally divided windows, each window is 100 bp, and the ratio of the average coverage of the 6 windows closest to the transcription start site to the average coverage of the 10 windows farthest from the transcription start site is calculated and normalized according to formula (2), as shown in formula (3). The cfDNA abundance distribution trend of each transcript promoter region in the same sample is obtained, and median correction is performed on the same batch of data.

[0089] Among them, depth of midBin and depth of sideBin represent the coverage of the 6 windows closest to the transcription start site and the 10 windows farthest from the transcription start site, respectively; total reads represents the total number of valid sequencing sequences obtained by 1.2 screening.

[0090] 2.3 cfDNA fragmentomic feature 3 (distribution of cfDNA fragment lengths in transcript promoter regions): Using the R language gini function, we analyzed the dispersion of cfDNA fragment lengths within the 100-300 bp range covering each promoter region. These data were median-corrected within the same data set.

[0091] 2.4 cfDNA fragmentomic feature 4 (cfDNA fragment length distribution in 5M genomic windows): The genome was divided into 5M bp windows. Based on the fragment lengths of the valid sequencing reads screened in 1.2, the ratio of the number of 100-150 bp and 151-220 bp cfDNA fragments aligned to each window after GC correction was calculated. Median correction was also performed within the same data batch.

[0092] 3. Screening of optimal omics features

[0093] 3.1 Data Grouping and Differential Analysis: The diseased group, healthy group, and control group were divided into a training set, an internal validation set, and an external validation set in a ratio of 6:2:2. In this application, the differential analysis method can be adaptively selected. For example, the wilcox.test was used for differential analysis to obtain promoter regions and genomic regions with different cfDNA abundance, abundance distribution, and fragment length distribution between the diseased group and the healthy control group.

[0094] In one example of the present application, the above difference analysis is repeated 1000 times, and 75% of the samples of the training set and the internal validation set are randomly sampled for difference analysis each time. Other sampling methods can also be used to implement the present invention.

[0095] 3.2 Feature Screening: Based on the results of 1000 differential analyses, the upper limit of the 95% confidence interval of the P value for each cfDNA fragment omics feature was calculated. CFDNA fragment omics features that showed the same trend in the 1000 differential analyses in the diseased group and had a 95% confidence interval of less than 0.05 were selected.

[0096] In some examples of the present application, the screened transcript promoter regions with differences are concentrated in the promoter regions of genes currently reported to be associated with eclampsia in the three major databases (genecard, dbPEC and cPE).

[0097] 3.3 Feature Combination Optimization: We used the Boruta feature selection algorithm (a feature selection algorithm in the random forest model) to rank the importance of the top-ranked cfDNA fragment omics features. We then used the regsubsets method (a function in the R language that traverses all possible feature combinations) to select the optimal feature combination. Specifically, for each cfDNA fragment omics feature, the top 20 most important features from the Boruta feature selection method were selected as input for the regsubsets method to select the optimal feature combination.

[0098] 4. Optimization of clinical indicator screening

[0099] Clinical parameter processing: Clinical factors such as maternal BMI, age, mean arterial pressure, medical history, and mode of pregnancy were collected. Clinical parameters for each pregnant woman were median-corrected using the median of a sample of healthy pregnant women from the same hospital.

[0100] 5. Model construction

[0101] Model Construction: Combining cfDNA fragment omics features with clinical indicators, a random forest model was used to predict early-onset preeclampsia. During model construction, the training set was used to optimize model parameters, and the model was constructed using Python.

[0102] It is worth noting that through the above method, a combination of optimal cfDNA fragment omics features was obtained, which involved cfDNA features of several target genes and genomic target regions (including: cfDNA fragment abundance in the promoter regions of GTF2H1, CD48, PBX1, PLA2G2A, ACVR1B, HMGB1, ERBB2, XRCC5, TGFBR2, NFKB1, EGF, TLR6, EZH2, IKZF1, EGFR and CD36 transcripts, distribution spectrum of cfDNA fragment abundance in the promoter regions of HIPK1, MTHFR, IL6R, RAB27A, UBE2I, CREBBP, SOX8, HNF4A, ICOS, LIF, TMPRSS6, DNMT3A, WWTR1, RASSF1, FOXP1, TLR2 and MAPK14 transcripts, and A The results also examined the distribution profiles of cfDNA fragment lengths in the promoter regions of GRN, ATP2B1, ELAC2, HDAC5, GABARAP, FXR2, NFE2L2, DGCR2, MAPK1, CASR, MYLIP, TUBB, DOCK4, PTGS1, HSPA5, ABL1, and SYK transcripts, as well as the distribution profiles of cfDNA fragment lengths in the chr1:110000000-115000000, chr4:1500000000-1550000000, chr6:160000000-1650000000, chr6:60000000-650000000, and chr7:1250000000-1300000000), as well as multiple clinical indicators (including maternal BMI, age, mean arterial pressure, medical history, and pregnancy mode).

[0103] 6. Result determination

[0104] In the prediction model, these characteristic data are input into the constructed model. A prediction result of 0 indicates that the sample is at low risk for early-onset preeclampsia, and a prediction result of 1 indicates that the sample is at high risk for early-onset preeclampsia. Beneficial effects

[0105] Based on whole-genome sequencing data of cfDNA in pregnant women's plasma during early pregnancy, this study systematically analyzed multiple cfDNA fragment omics features, including the abundance of cfDNA fragments in transcript promoter regions, the distribution spectrum of cfDNA fragment abundance in transcript promoter regions, the distribution spectrum of cfDNA fragment length in transcript promoter regions, and the distribution spectrum of cfDNA fragment length in the 5M window of the genome. The results of this application showed that samples from patients with early-onset preeclampsia showed significantly different characteristics from healthy controls in all four cfDNA fragment omics features, indicating that these cfDNA fragment omics features have the potential to serve as markers for early-onset preeclampsia screening in early pregnancy. In addition, through machine learning methods, a model for predicting early-onset preeclampsia was constructed by comprehensively considering multiple cfDNA fragment omics features and clinical indicators. Compared with prediction based on clinical characteristics alone, the prediction model of this application has higher accuracy and outperforms the existing early-onset preeclampsia prediction model. This study not only provides new possibilities for screening for early-onset preeclampsia, but also has broad application prospects in the development of related products.

[0106] The embodiments of the present invention will be described in more detail below, and examples of the embodiments are shown in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, but are not to be construed as limiting the present invention. Where specific techniques or conditions are not specified in the embodiments, they are performed according to the techniques or conditions described in the literature in this area or according to the product specifications. Reagents or instruments used that do not specify the manufacturer are conventional products that can be obtained commercially.

[0107] Example 1: Construction and evaluation of a prediction model for early-onset preeclampsia

[0108] 1. Sample Collection

[0109] 1.1 Inclusion criteria

[0110] Inclusion criteria for early-onset preeclampsia: 20 to 34 weeks of gestation, systolic blood pressure ≥140 mmHg and / or diastolic blood pressure ≥90 mmHg, accompanied by any of the following: urine protein ≥0.3 g / 24 h, or urine protein / creatinine ratio ≥0.3, or random urine protein ≥(+) (examination method when protein quantification is unconditional); or no proteinuria, but accompanied by any of the following organ or system involvement: important organs such as the heart, lungs, liver, and kidneys, or abnormal changes in the blood system, digestive system, or nervous system, or placental-fetal involvement, etc.

[0111] The inclusion criteria for the healthy control group were: normal fetal growth at birth and no obstetric, medical or surgical complications during pregnancy; exclusion criteria included: severe heart, liver, or kidney dysfunction; autoimmune diseases or malignant tumors; and abnormal pregnant women due to chromosomal and congenital abnormalities.

[0112] 1.2 Participants

[0113] This study employed a disease-control approach, using a unique matching principle for pregnant women at four hospitals. Clinical data from pregnant women aged 24 to 45 years from 2021 to 2022 were extracted from the system. Residual plasma samples from gestational weeks 12 to 25 were also collected. A total of 585 eligible samples were screened according to the inclusion criteria in step 1.1, as shown in Table 1. This study used retrospective samples, and informed consent for generalization was obtained.

[0114] Table 1: Statistics of various hospitals

[0115] 2. Plasma Cell-free DNA Library Construction and Sequencing

[0116] Double-stranded DNA libraries were generated from 200 μL of plasma samples collected from pregnant women at gestational ages 12-25 weeks using the MGIEasy Cell-Free DNA Library Preparation Kit (Brand: MGI; Catalog No. 1000007037). Subsequently, double-stranded DNA libraries from 10 pregnant women were pooled and single-stranded circular DNA libraries were prepared using the MGIEasy Circularization Kit for use on the MGI sequencing platform's matrix nanochip. Sequencing data were obtained using the DNBSEQ-T7 sequencer with 20-30X coverage and paired-end 100bp sequencing.

[0117] 3. Analysis of Omic Characteristics of Plasma Cell-free DNA Fragments

[0118] After obtaining the whole genome data of the first trimester plasma cfDNA of the above-mentioned pregnant women, we compared it with the reference genome GRCh38, removed multiple matching results and PCR duplications, and performed cfDNA fragment omics analysis, including:

[0119] cfDNA fragment omics feature 1 (abundance of cfDNA fragments in the promoter region of transcripts, hereinafter referred to as pTSS coverage): The total number of valid sequencing sequences obtained after alignment and filtering was used as the normalization parameter. The abundance of each 2kb transcript promoter region of the same sample was calculated and corrected for sequencing depth according to formula (1), and median correction was performed in the same batch of data. i coverage=(reads number mapped to pTSS i ×109 ) / (total reads × 2000), formula (1);

[0120] Among them, pTSSi coverage represents the abundance of cfDNA fragments in the promoter region of transcript i; reads number mapped to pTSSi represents the number of sequencing reads matching the cfDNA fragments in the promoter region of transcript i; total reads represents the number of valid reads obtained in the sequencing data (obtained by aligning and filtering the offline sequencing data).

[0121] cfDNA fragment omics feature 2 (distribution spectrum of cfDNA fragment abundance in transcript promoter regions, subsequently referred to as pTSS score): The 2 kb promoter region of each transcript was divided into 20 equally divided windows, each window was 100 bp, and the ratio of the average coverage of the 6 windows closest to the transcription start site to the average coverage of the 10 windows farthest from the transcription start site was calculated and normalized according to formula (2), and median correction was performed in the same batch of data.

[0122] Among them, depth of midBin and depth of sideBin represent the coverage of the 6 windows closest to the transcription start site and the 10 windows farthest from the transcription start site, respectively; total reads represents the number of valid reads obtained in the sequencing data (obtained by aligning and filtering the offline sequencing data).

[0123] cfDNA fragment omics feature 3 (distribution spectrum of cfDNA fragment lengths in transcript promoter regions, subsequently referred to as gini): The fragment lengths of the valid sequencing sequences obtained after alignment and filtering were used to analyze the discreteness of the cfDNA fragment lengths covering each 2kb transcript promoter region in the range of 100bp-300bp using the R language gini function, and median correction was performed on the same batch of data.

[0124] cfDNA fragment omics feature 4 (distribution spectrum of cfDNA fragment length in 5M genomic windows, subsequently referred to as DELFI): The genome is evenly divided into 5M bp windows, and the fragment lengths of the valid sequencing sequences obtained after alignment and filtering are used to obtain the GC-corrected ratio of the number of cfDNA fragments with a length of 100-150 bp to the number of cfDNA fragments with a length of 151-220 bp in each window, and median correction is performed in the same batch of data.

[0125] 4. Screening of cfDNA chromatin profiles for predicting early-onset preeclampsia

[0126] 4.1 The four data sets (from four hospitals) in Table 1 were combined to include samples from the diseased group and healthy controls. The data were divided into training, internal validation, and external validation sets in a 6:2:2 ratio (Table 2). Samples from the training and internal validation sets were randomly sampled 1000 times with replacement. Wilcox.test was used to analyze the four cfDNA fragment omics features that were significantly different between the early-onset preeclampsia group and the healthy control group using 75% of the samples from each random sample.

[0127] Table 2: Division and statistics of each data set

[0128] 4.2 By combining the P value results of 1000 differential analyses and taking the upper limit of the 95% confidence interval of the P value <0.05 as the standard, we screened out the features in which the 1000 differential analysis results in the diseased group showed the same change trend compared with the healthy control group and the upper limit of the 95% confidence interval of the P value was less than 0.05.

[0129] 4.3 Filter the differential cfDNA fragment omics features obtained in step 4.2 based on the genes associated with eclampsia collected in at least two of the three major databases (genecard, dbPEC, and cPE).

[0130] 4.4 Use the Boruta feature selection method to rank the importance of each cfDNA fragment omics feature screened in step 4.3. For each cfDNA fragment omics feature, select the top 20 important features and use the regsubsets method. Use ten-fold cross-validation to select the optimal feature combination that minimizes the cross-validation error. The optimal number of combinations of the four cfDNA fragment omics features was 16, 17, 17, and 5, respectively (Figure 2), including: cfDNA fragment abundance in the promoter regions of GTF2H1, CD48, PBX1, PLA2G2A, ACVR1B, HMGB1, ERBB2, XRCC5, TGFBR2, NFKB1, EGF, TLR6, EZH2, IKZF1, EGFR, and CD36 transcripts; cfDNA fragment abundance in the promoter regions of HIPK1, MTHFR, IL6R, RAB27A, UBE2I, CREBBP, SOX8, HNF4A, ICOS, LIF, TMPRSS6, DNMT3A, WWTR1, RASSF1, FOXP1, TLR2, and MAPK14 transcripts. The distribution profiles of fragment abundance, the distribution profiles of cfDNA fragment lengths in the promoter regions of AGRN, ATP2B1, ELAC2, HDAC5, GABARAP, FXR2, NFE2L2, DGCR2, MAPK1, CASR, MYLIP, TUBB, DOCK4, PTGS1, HSPA5, ABL1 and SYK transcripts, and the distribution profiles of cfDNA fragment lengths in the chr1:1100000000-1150000000, chr4:1500000000-1550000000, chr6:1600000000-1650000000, chr7:60000000-650000000 and chr7:1250000000-130000000

[0131] 5. Screening of clinical indicators for predicting early-onset preeclampsia

[0132] 5.1 The differences in clinical information of pregnant women in the diseased group and the healthy control group in the four groups of data in Table 1 were calculated separately (Table 3). Wilcox.test was used to perform difference analysis on continuous variables, including: maternal age, height, weight, BMI, mean arterial pressure, and gestational age at delivery. Fisher's test was used to perform difference analysis on discrete variables, including pregnancy number, parity, past medical history, and pregnancy mode. Past medical history is the sum of the number of previous histories of chronic hypertension, previous preeclampsia, previous systemic lupus erythematosus, and previous antiphospholipid syndrome. If the pregnancy mode is assisted reproductive method, it is 1, and if it is natural pregnancy method, it is 0. Based on the above difference analysis results, maternal age, BMI, mean arterial pressure, past medical history, and pregnancy mode were used as clinical indicators for subsequent model construction.

[0133] Table 3: Clinical information statistics of pregnant women in four data sets Note: *, *, *** represent statistical differences of P < 0.05, P < 0.01 and P < 0.001, respectively.

[0134] 5.2 Correction of clinical indicators of pregnant women: The median age, BMI and mean arterial pressure of the maternal age, BMI and mean arterial pressure of each sample were corrected using the median of the maternal age, BMI and mean arterial pressure of the control sample in each group of data in Table 3.

[0135] 6. Construction of a model for predicting early-onset preeclampsia

[0136] Combining the clinical indicators screened in 5.1 with one or more cfDNA fragment omics features screened in 4.4, and using the training set in Table 2, a machine learning approach was used to construct a model and adjust model parameters. Ultimately, the optimal model was identified based on the area under the curve (AUC) calculated using the receiver operating characteristic (ROC) curve. The optimal parameter results are shown in Table 4.

[0137] Table 4: Optimal parameters of each random model

[0138] 7. Internal and External Validation of the Model for Predicting Early-Onset Preeclampsia

[0139] Using the two sets of data from the validation set in Table 2 , the abundance of cfDNA fragments in the promoter regions of GTF2H1, CD48, PBX1, PLA2G2A, ACVR1B, HMGB1, ERBB2, XRCC5, TGFBR2, NFKB1, EGF, TLR6, EZH2, IKZF1, EGFR, and CD36 transcripts in pregnant women’s plasma obtained as described in steps 3 and 5, the distribution profiles of the abundance of cfDNA fragments in the promoter regions of HIPK1, MTHFR, IL6R, RAB27A, UBE2I, CREBBP, SOX8, HNF4A, ICOS, LIF, TMPRSS6, DNMT3A, WWTR1, RASSF1, FOXP1, TLR2, and MAPK14 transcripts, and ATP2B1, ELAC2, HDAC5, GABA, and ATPase inhibitors were analyzed. The distribution spectrum of cfDNA fragment length in the promoter regions of ABARAP, FXR2, NFE2L2, DGCR2, MAPK1, CASR, MYLIP, TUBB, DOCK4, PTGS1, HSPA5, ABL1 and SYK transcripts, as well as the distribution spectrum of cfDNA fragment length in the chr1:110000000-115000000, chr4:150000000-155000000, chr6:160000000-165000000, chr6:60000000-65000000 and chr7:125000000-130000000, maternal BMI, age, mean arterial pressure, medical history and pregnancy mode were used as input values ​​(input matrix) for model validation of the optimal random forest model (Table 4). The AUC, accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of the three data sets are shown in Table 5. As shown in Table 5, the early-onset preeclampsia prediction model of the present application can effectively predict preeclampsia, and the prediction effect of the prediction model combining one or more cfDNA fragment omics features with clinical risk factors is better than the prediction effect of using clinical information alone;

[0140] Table 5: Evaluation indicators of the model for predicting early-onset preeclampsia at first diagnosis Note: 95% CI stands for 95% confidence interval.

[0141] In this example, the inventors used the Bootstrap method to repeatedly extract 1000 samples from each dataset, each time extracting the same disease and healthy control samples from the corresponding dataset. The median and 95% confidence interval of the 1000 results of the random forest model under the optimal parameters were calculated.

[0142] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of the technical features being referred to. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one such feature. In the description of the present invention, "plurality" means at least two, such as two, three, etc., unless otherwise specifically defined.

[0143] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in a suitable manner in any one or more embodiments or examples. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples, unless they are mutually inconsistent. Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations of the present invention. Those skilled in the art may change, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. A gene marker combination, characterized in that: Including GTF2H1, CD48, PBX1, PLA2G2A, ACVR1B, HMGB1, ERBB2, XRCC5, TGFBR2, NFKB1, EGF, TLR6, EZH2, IKZF1, EGFR, CD36, HIPK1, MTHFR, IL6R, RAB27A, UBE2I, CREBBP, SOX8, HNF4A, ICOS, LIF, TMPRSS6, DNMT3A, WWTR1, RASSF1, FOXP1, TLR2, MAPK14, AGRN, ATP2B1, EL AC2, HDAC5, GABARAP, FXR2, NFE2L2, DGCR2, MAPK1, CASR, MYLIP, TUBB, DOCK4, PTGS1, HSPA5, ABL1, SYK, chr1:1100000000-115000000, chr4:150000000-155000000, chr6:160000000-165000000, chr6:60000000-65000000 and chr7:125000000-130000000.

2. A kit, characterized in that: Comprising a reagent for detecting at least one of the abundance of cfDNA fragments in the promoter region of the transcript of the gene marker combination of claim 1, the distribution spectrum of the abundance of cfDNA fragments, the distribution spectrum of the length of cfDNA fragments, and the distribution spectrum of the length of cfDNA fragments in the 5M window of the genome.

3. The kit according to claim 2, characterized in that Further comprising a reference data set of the gene marker combination according to claim 1.

4. Use of the kit according to claim 2 or 3 in detecting early-onset preeclampsia, evaluating the prognosis of early-onset preeclampsia or providing medication guidance.

5. A method for training a machine learning model, wherein the machine learning model is used to predict the risk of early-onset preeclampsia, characterized in that: The method comprises: Acquiring feature information of a training data set, the feature information comprising text features and cfDNA fragment features of the gene marker combination of claim 1; The characteristic information is input into a machine learning model, and the known risk of occurrence of early-onset preeclampsia in the training samples is used as a marker to perform supervised training on the machine learning model.

6. The method according to claim 5, characterized in that The text features include at least one selected from the group consisting of maternal age, body mass index, mean arterial pressure, past medical history, and pregnancy pattern.

7. The method according to claim 5, characterized in that The cfDNA fragment characteristics include at least one selected from the abundance of cfDNA fragments on the promoter region of the transcript, the distribution spectrum of cfDNA fragment abundance, the distribution spectrum of cfDNA fragment length, and the distribution spectrum of cfDNA fragment length in the genomic 5M window.

8. The method according to claim 5 or 7, characterized in that: The cfDNA fragment characteristics of the gene marker combination include the abundance of cfDNA fragments in the promoter regions of GTF2H1, CD48, PBX1, PLA2G2A, ACVR1B, HMGB1, ERBB2, XRCC5, TGFBR2, NFKB1, EGF, TLR6, EZH2, IKZF1, EGFR and CD36 transcripts, the distribution spectrum of the abundance of cfDNA fragments in the promoter regions of HIPK1, MTHFR, IL6R, RAB27A, UBE2I, CREBBP, SOX8, HNF4A, ICOS, LIF, TMPRSS6, DNMT3A, WWTR1, RASSF1, FOXP1, TLR2 and MAPK14 transcripts, and the distribution spectrum of the abundance of cfDNA fragments in the promoter regions of AGRN , ATP2B1, ELAC2, HDAC5, GABARAP, FXR2, NFE2L2, DGCR2, MAPK1, CASR, MYLIP, TUBB, DOCK4, PTGS1, HSPA5, ABL1 and SYK transcript promoter regions, and at least one of the distribution profiles of cfDNA fragment lengths in chr1:110000000-115000000, chr4:150000000-155000000, chr6:160000000-165000000, chr6:60000000-65000000 and chr7:125000000-130000000 9. The method according to claim 7, characterized in that: The abundance of cfDNA fragments in the promoter region of the transcript is calculated by the following formula: pTSS i coverage=(reads number mapped to pTSS i ×10 9 ) / (total reads×2000), formula (1); in, pTSS i Coverage represents the abundance of cfDNA fragments in the promoter region of transcript i; reads number mapped to pTSS i Represents the number of sequencing reads matching the cfDNA fragments on the promoter region of transcript i; total reads The number of valid reads in the sequencing data.

10. The method according to claim 7, characterized in that The distribution spectrum of cfDNA fragment abundance in the promoter region of the transcript is calculated by the following formula: in, pTSS i score represents the distribution spectrum of cfDNA fragment abundance in the promoter i region of the transcript; Depth of midBin represents the coverage of the Z windows closest to the transcription start site; Depth of sideBin represents the coverage of the W windows farthest from the transcription start site; x / y represents the length of the equally divided window in the promoter region of the transcript; total reads The number of valid reads in the sequencing data.

11. The method according to claim 10, characterized in that The value of Z is selected from any integer between 3 and 9, preferably 6; the value of W is selected from any integer between 8 and 15, preferably 10; the value of x / y is selected from any integer between 80 and 120, preferably 100.

12. The method according to claim 7, characterized in that The distribution spectrum of the cfDNA fragment length on the transcript promoter region is obtained by analyzing the discreteness of the cfDNA fragment length covering each transcript promoter region within a first predetermined range.

13. The method according to claim 7, characterized in that The distribution spectrum of cfDNA fragment length in the genomic 5M window is obtained as follows: a. Divide the genome into 5M bp windows; b. Calculate the ratio of the number of cfDNA fragments in each window between the second predetermined range and the third predetermined range.

14. The method according to claim 12 or 13, characterized in that The first predetermined range is selected from 100 to 300 bp; the second predetermined range is selected from 100 to 150 bp; and the third predetermined range is selected from 151 to 220 bp.

15. The method according to claim 13, characterized in that The cfDNA fragments were GC corrected.

16. The method according to claim 5, characterized in that The machine learning model is selected from at least one of a random forest, a decision tree, and a neural network; a random forest model is preferred.

17. A prediction model for early-onset preeclampsia, characterized in that: include: Obtaining sequencing data of the sample to be tested; Acquiring characteristic information of the sequencing data; Inputting the feature information into a pre-trained machine learning model to obtain a prediction result; Wherein, the pre-trained machine learning model is obtained by training through the method described in any one of claims 5 to 16.

18. The prediction model according to claim 17, characterized in that The sample to be tested is selected from at least one of plasma, whole blood, amniotic fluid, serum or urine of a pregnant woman.

19. The prediction model according to claim 17, characterized in that The prediction result is determined as follows: If the output result of the machine learning model is 0, the sample is judged to be at low risk for early-onset preeclampsia; If the output result of the machine learning model is 1, the sample is judged to be at high risk for early-onset preeclampsia.

20. An electronic device, characterized in that: include: Memory and processor; The memory is used to store computer programs; The processor is used to execute the computer program to implement the method according to any one of claims 5 to 16.

21. A computer-readable storage medium, characterized in that: The storage medium stores computer program instructions, and when the computer program instructions are executed on a processor, the processor is caused to execute the method according to any one of claims 5 to 16.

Images