Serum metabolism marker combination for ectopic pregnancy diagnosis and rupture risk prediction and application thereof

Abstract

The invention discloses a serum metabolism marker combination for ectopic pregnancy diagnosis and rupture risk prediction and application of the serum metabolism marker combination, and relates to the technical field of biological science. The metabolic fingerprint diagnosis model (AUCgt, 0.91) and the metabolic marker panel (AUCgt, 0.92) have the advantages that the performance of the metabolic fingerprint diagnosis model (AUCgt, 0.91) and the performance of the metabolic marker panel (AUCgt, 0.92) are obviously superior to those of an existing'ultrasonic + beta-hCG 'diagnosis mode, the problems of diagnosis delay, misdiagnosis and missed diagnosis clinically at present are effectively solved, and the clinical rigid requirements of early and accurate diagnosis are met; the fracture risk prediction model AUC 0.885 constructed by the method disclosed by the invention fills the current clinical blank. Key decision support can be provided for doctors, early intervention on high-risk patients is achieved, excessive medical treatment on low-risk patients is avoided, and the mortality risk of pregnant and lying-in women is reduced. According to the invention, based on the nanoparticle-assisted laser desorption / ionization mass spectrometry technology, ultra-fast detection of only about 3 seconds for single sample detection is realized, and the method is expected to become a fast detection tool in emergency treatment and outpatient service scenes.

Description

Technical Field

[0001] This invention relates to the field of bioscience and technology, and in particular to a combination of serum metabolic markers for the diagnosis of ectopic pregnancy and the prediction of rupture risk, and their application. Background Technology

[0002] Ectopic pregnancy (EP) refers to an abnormal pregnancy in which the embryo implants outside the uterine cavity. Globally, it accounts for approximately 2% of all pregnancies and is a leading cause of maternal morbidity and mortality in early pregnancy. Early and accurate diagnosis is crucial for preventing life-threatening complications such as fallopian tube rupture, which occurs in approximately 20%-35% of ectopic pregnancies, leading to 6%-10% of pregnancy-related deaths. Currently, clinical diagnosis of ectopic pregnancy mainly relies on ultrasound examination (especially transvaginal ultrasound, TVUS) and blood parameter testing. However, ultrasound examination suffers from high operator dependence and low sensitivity (approximately 55%-85%), especially in patients in early pregnancy. Furthermore, continuous monitoring of blood β-human chorionic gonadotropin (β-hCG) levels overlaps between ectopic pregnancy, intrauterine pregnancy, and miscarriage, resulting in up to 40% of cases with delayed diagnosis and nearly 20% of cases experiencing rupture before diagnosis. Therefore, there is an urgent need to develop an efficient and accurate method for ectopic pregnancy diagnosis and risk prediction.

[0003] Blood biomarkers hold significant value in precision medicine, particularly for the diagnosis of ectopic pregnancy. Currently used blood biomarkers (such as β-hCG) are mostly protein-based, located upstream of regulatory pathways, and can only indirectly reflect disease status. In contrast, metabolite biomarkers, as end products of regulatory pathways, can directly and in real-time reflect pathophysiological changes, offering new possibilities for the diagnosis of ectopic pregnancy. Recent metabolomics studies have shown that ectopic pregnancy involves multiple metabolic pathway disorders, suggesting the potential of diagnostic strategies based on pathway-specific biomarkers. However, current research on metabolic biomarkers for ectopic pregnancy is still in its early stages, with limitations such as a limited cohort size (approximately 370 samples) and a lack of multifunctional tools integrating diagnostic and risk prediction functions.

[0004] Mass spectrometry is a primary method for metabolite analysis, enabling label-free, high-precision detection by measuring the mass-to-charge ratio (m / z) of metabolites. However, traditional mass spectrometry platforms typically require complex sample pretreatment steps (such as protein removal and chromatographic separation) to purify and enrich metabolites in biological samples, resulting in slow analysis speeds (approximately 25 minutes per sample) and limiting its application in large-scale clinical settings. Notably, nanoparticle-assisted laser desorption / ionization mass spectrometry (Nano-LDI-MS) utilizes engineered nanoparticles to selectively capture metabolites on a microarray chip, achieving rapid (approximately 3 seconds per sample) and high-throughput metabolite detection, providing a powerful alternative to traditional mass spectrometry techniques. Currently, serum metabolic fingerprinting based on Nano-LDI-MS technology has been successfully applied in the clinical management of various major diseases (such as gastric cancer, polycystic ovary syndrome, and endometrial cancer). Given the rich information contained in serum metabolic fingerprints, developing novel detection methods based on this technology to improve the clinical diagnosis and treatment of reproductive system diseases such as ectopic pregnancy holds significant potential.

[0005] In summary, existing diagnostic methods for ectopic pregnancy have significant limitations in terms of sensitivity, specificity, and diagnostic timeliness, while metabolic marker detection technology has not been fully applied to the clinical diagnosis and risk assessment of ectopic pregnancy. Therefore, there is an urgent need to develop a new technological system that can achieve rapid and accurate diagnosis, while also having the function of predicting the risk of rupture, in order to improve the clinical management of ectopic pregnancy and improve patient prognosis.

[0006] To address the aforementioned problems, this invention aims to develop a serum metabolite for the diagnosis of ectopic pregnancy and the prediction of rupture risk.

[0007] Thank you for the combination of symbols. Summary of the Invention

[0008] In view of the above-mentioned deficiencies of the prior art, the technical problem to be solved by the present invention is to develop a combination of serum metabolic markers for the diagnosis of ectopic pregnancy and the prediction of rupture risk.

[0009] To achieve the above objectives, the present invention provides a serum metabolite for the diagnosis of ectopic pregnancy and prediction of rupture risk.

[0010] Xie's marker combination and its application.

[0011] Furthermore, a combination of serum metabolic markers for the diagnosis and prediction of rupture risk in ectopic pregnancy includes: A. A combination of serum metabolic markers for the diagnosis of ectopic pregnancy, consisting of the following markers: Glycerol, Dihydrouracil, N-methyl-beta-alanine, Dihydrouracil, Urocanic acid, Lysine, Glucose, Glycylproline, and Fucose. B. A combination of serum metabolic markers for predicting the risk of ectopic pregnancy rupture, used to further predict the risk of fallopian tube rupture in patient samples that are determined to be ectopic pregnancy by the aforementioned combination of serum metabolic markers; specifically composed of the following markers: Glycine, Glyoxylic acid, Ureidoglycolate, Acetoacetic acid, Phenylalanine, Pyrophosphate, Leucine, Pyridoxine, Methionine, Semidehydroascorbic acid, Phosphorylcholine, and Docosatetraenylethanolamide.

[0012] Furthermore, the screening method for the combination of serum metabolic markers used to diagnose ectopic pregnancy includes the following steps: Step 1: Collect serum samples from the clinically confirmed ectopic pregnancy group and the intrauterine pregnancy control group. Use nanoparticle-assisted laser desorption / ionization mass spectrometry to perform metabolic detection on all the serum samples and obtain the serum metabolic fingerprint of each sample; Step 2: Based on key clinical characteristics such as gestational age, age, and BMI, the overall sample is divided into a model discovery set and an independent validation set using stratified random sampling. Step 3: Perform Student's t-test on the serum metabolic fingerprints of the discovery set for the EP group and the IUP group to screen out m / z features that show significant differences between the two groups, with the significance level set at P<0.05. Step 4: The m / z features obtained in Step 3 are further reduced in dimensionality and filtered using the Lasso regression model. The regularization strength is determined by cross-validation, and the feature coefficient threshold is set to ≥0.4. Finally, 20 core m / z features that contribute the most to EP classification are selected. Step 5: Use high-precision Fourier transform ion cyclotron resonance mass spectrometry to accurately determine the m / z features with significant differences selected in Step 4, ensuring that the mass error is <5ppm; compare the inferred molecular formula with the human metabolome database, and annotate these features with metabolites to obtain a combination of serum metabolic markers for diagnosing ectopic pregnancy.

[0013] Further, in step one, the method for obtaining the serum metabolic fingerprint is as follows: First, iron oxide nanoparticles are synthesized using a solvothermal method. Then, the iron oxide nanoparticles are dispersed in an ethanol / water solution to prepare a matrix solution with a concentration of 1 mg / mL. Next, 1.5 μL of serum sample diluted 10 times is spotted onto a 384-well polished steel target plate. After natural drying at room temperature, 1.5 μL of matrix solution is added, and the plate is dried naturally at room temperature again to obtain a serum sample for mass spectrometry analysis. Then, a Bruker Autoflex time-of-flight mass spectrometer is used for detection. The detection parameters are set as follows: laser wavelength 355 nm, pulse frequency 1000 Hz, and 2000 laser signals are collected at each point. Finally, the obtained raw mass spectrometry data is processed to obtain a set of stable m / z features, which constitute the serum metabolic fingerprint of each sample.

[0014] Furthermore, the preparation method of the iron oxide nanoparticles is as follows: 0.6g of ferric chloride hexahydrate, 0.15g of trisodium citrate dihydrate and 0.96g of anhydrous sodium acetate are dissolved in 25mL of ethylene glycol and subjected to hydrothermal reaction at 200°C for 10 hours; the reaction product is washed with deionized water and ethanol and then dried to obtain iron oxide nanoparticles.

[0015] Furthermore, the raw mass spectrometry data obtained in step one was processed using Python 3.8. Specifically, peaks with a signal-to-noise ratio ≥ 3 were extracted, peak alignment was performed, and peaks with a frequency lower than 2 / 3 in any group were filtered out, resulting in 348 stable m / z features that constitute the serum metabolic fingerprint of each sample.

[0016] Furthermore, the screening method for the combination of serum metabolic markers used to predict the risk of ectopic pregnancy rupture includes the following steps: a. From the total sample of ectopic pregnancy patients, select multiple samples with clear outcomes to construct a cohort. The cohort includes patients who experienced rupture and whose intraoperative bleeding volume was confirmed to be greater than 100 mL, as well as patients who were confirmed not to have ruptured after conservative treatment or surgery. The cohort is then randomly divided into a discovery set and a validation set in a 7:3 ratio. b. In the discovery set, the following two strategies were used to screen for features associated with rupture from the full spectrum of serum metabolic fingerprints: Use Lasso regression to select the first set of features whose regression coefficients are not zero. Using the random forest algorithm, the importance score of each m / z feature is calculated, and the set of second features with an importance score greater than zero is selected. c. Select the intersection of the first feature set and the second feature set to finally determine a serum metabolic marker combination consisting of 12 metabolic markers for predicting the risk of ectopic pregnancy rupture.

[0017] Furthermore, the application of the serum metabolic marker combination is in the preparation of detection products for the diagnosis of ectopic pregnancy and the prediction of rupture risk.

[0018] Furthermore, the detection product includes an ectopic pregnancy diagnostic model, and the method for constructing the diagnostic model includes the following steps: 1) Based on key clinical characteristics such as gestational age, age, and BMI, a random stratified sampling method was used to divide serum samples from the clinically diagnosed ectopic pregnancy group and the intrauterine pregnancy control group into a model discovery set and an independent validation set in a 7:3 ratio. 2) The relative abundance values ​​of the m / z features of the complete serum metabolic fingerprint or the combination of serum metabolic markers used to diagnose ectopic pregnancy were used in the discovery set. The Orange data mining platform was used to train the model using a variety of machine learning algorithms, and five-fold cross-validation was used to optimize the model parameters. 3) Test the trained model on an independent validation set to evaluate its diagnostic performance.

[0019] Furthermore, in step 2), the machine learning algorithms include Lasso regression, neural networks, gradient boosting, ridge regression, support vector machines, and random forests.

[0020] Furthermore, the detection product includes an ectopic pregnancy rupture risk prediction model. The method for constructing this model involves training a machine learning model on a discovery set using the relative abundance values ​​of the m / z features of the rupture risk metabolic biomarkers and four clinical indicators: serum β-hCG level, endometrial thickness, adnexal mass size, and BMI. The trained model is then tested on an independent validation set to evaluate its diagnostic performance.

[0021] In a preferred embodiment 1 of the present invention, the process of obtaining serum metabolic fingerprints (ESF) using nanoparticle-assisted laser desorption / ionization mass spectrometry (Nano-LDI-MS) technology and constructing a machine learning model to achieve accurate diagnosis of ectopic pregnancy (EP) is described in detail.

[0022] In another preferred embodiment 2 of the present invention, the construction and application process of the ectopic pregnancy-specific metabolic marker panel is described in detail.

[0023] In another preferred embodiment 3 of the present invention, the construction and application process of the metabolic biomarker panel for predicting the risk of ruptured ectopic pregnancy is described in detail.

[0024] Technical effects: 1. The metabolic fingerprint diagnostic model (AUC>0.91) and metabolic biomarker panel (AUC>0.92) provided by this invention significantly outperform the existing "ultrasound + β-hCG" diagnostic model in terms of performance, effectively solving the problems of diagnostic delay, misdiagnosis and missed diagnosis currently faced in clinical practice, and meeting the rigid clinical demand for early and accurate diagnosis.

[0025] 2. The rupture risk prediction model (AUC 0.885) constructed in this invention fills a gap in current clinical practice. It can provide doctors with crucial decision support, enabling early intervention for high-risk patients, avoiding overtreatment of low-risk patients, and reducing the risk of maternal mortality.

[0026] 3. This invention, based on nanoparticle-assisted laser desorption / ionization mass spectrometry, achieves ultra-fast detection of a single sample in just under 3 seconds, making it a promising rapid testing tool for emergency and outpatient settings. This significantly shortens the diagnostic decision-making cycle from several hours or even days to minutes, substantially improving the efficiency of hospital treatment processes and demonstrating broad market prospects.

[0027] The following will further explain the concept, specific structure, and technical effects of the present invention in conjunction with the accompanying drawings, so as to fully understand the purpose, features, and effects of the present invention. Attached Figure Description

[0028] Figure 1 This is a preferred embodiment of the present invention: an EP diagnostic ROC curve based on metabolic fingerprinting. Figure 2 This is a schematic diagram of the EP diagnostic performance (based on metabolic fingerprint) at different gestational weeks, ages, and BMIs in a preferred embodiment of the present invention. Figure 3 This is a preferred embodiment 2 of the present invention, showing the ROC results of EP diagnosis based on metabolic markers; Figure 4 This is a preferred embodiment 3 of the present invention, showing the ROC curve for predicting EP rupture risk based on 12 metabolic markers and 4 clinical indicators; Figure 5 This is a comparison chart of the rupture risk prediction performance of a metabolic model (12 metabolic biomarkers and 4 clinical indicators) and the rupture risk measurement performance of the 4 clinical indicators in a preferred embodiment 3 of the present invention. Detailed Implementation

[0029] The following description, with reference to the accompanying drawings, illustrates several preferred embodiments of the present invention to make its technical content clearer and easier to understand. The present invention can be embodied in many different forms, and the scope of protection of the present invention is not limited to the embodiments mentioned herein.

[0030] Example 1: A machine learning model based on serum metabolic fingerprinting for the diagnosis of ectopic pregnancy

[0031] This embodiment aims to illustrate how to use nanoparticle-assisted laser desorption / ionization mass spectrometry (Nano-LDI-MS) technology to obtain serum metabolic fingerprints (ESF) and construct a machine learning model to achieve accurate diagnosis of ectopic pregnancy (EP).

[0032] 1. Sample Cohort and Serum Sample Preparation: Serum samples from a total of 378 women were collected from the Pregnancy Biobank of the International Peace Maternity and Child Health Hospital. Among them, 183 cases were clinically diagnosed with EP, and 195 cases served as controls for intrauterine pregnancy (IUP). Diagnosis was confirmed by transvaginal ultrasound (TVUS) and serum β-human chorionic gonadotropin (β-hCG) levels, with postoperative pathology or subsequent pregnancy outcome serving as the gold standard.

[0033] All participants fasted for at least 8 hours before blood collection. Peripheral venous blood was collected and centrifuged at 3000 rpm for 10 minutes at 4°C within 2 hours to obtain serum, which was then aliquoted and stored at -80°C until analysis.

[0034] The 378 participants were randomly stratified into a discovery set (263 participants, including 128 EP and 135 IUP) and an independent validation set (115 participants, including 55 EP and 60 IUP). There were no significant differences between the two groups in baseline characteristics such as age and BMI (P>0.05).

[0035] 2. Collection of serum metabolic fingerprints: 1) Preparation of nanoparticle matrix: Ferric oxide nanoparticles were synthesized using a solvothermal method. Specifically, 0.6 g of ferric chloride hexahydrate, 0.15 g of trisodium citrate dihydrate, and 0.96 g of anhydrous sodium acetate were dissolved in 25 mL of ethylene glycol and subjected to a hydrothermal reaction at 200°C for 10 hours. The reaction product was washed with deionized water and ethanol and then dried to obtain nanoparticle powder. This powder was dispersed in an ethanol / water solution to prepare a homogeneous matrix suspension with a concentration of 1 mg / mL.

[0036] 2) Mass spectrometry detection: The serum sample was taken out from -80°C, thawed on ice, and then diluted 10 times with deionized water.

[0037] 3) Take 1.5 μL of diluted serum and spot it onto a 384-well polished steel target plate, then air dry at room temperature.

[0038] 4) Cover the dried serum spots with 1.5 μL of the above nanoparticle matrix solution and air dry again at room temperature to form a co-crystallized layer.

[0039] 5) Detection was performed using a Bruker Autoflex time-of-flight mass spectrometer. Parameter settings: laser wavelength 355nm, pulse frequency 1000Hz, 2000 laser signal acquisitions per point. The entire detection cycle (including laser irradiation and point movement) was approximately 3 seconds per sample.

[0040] 3. Data processing and model building: 1) Data preprocessing: The raw mass spectrometry data were processed using Python 3.8. Peaks with a signal-to-noise ratio (S / N) ≥ 3 were extracted, peak alignment was performed, and peaks with a frequency lower than 2 / 3 in any group were filtered out, resulting in 348 stable m / z features, which constitute the ESF of each sample.

[0041] 2) Based on the above ESF, machine learning model training and validation were performed. First, the queue was divided into a discovery set and a validation set in a 7:3 ratio. In the discovery set, the Orange data mining platform (v3.25.0) was used to train the model using six machine learning algorithms, including: Lasso regression, neural network, gradient boosting, ridge regression, support vector machine, and random forest.

[0042] 3) Five-fold cross-validation is used to optimize model parameters to prevent overfitting.

[0043] 4) Test the trained diagnostic model on an independent validation set to evaluate its diagnostic performance.

[0044] 4. Technical Effects

[0045] Diagnostic performance: On the independent validation set, the Lasso machine learning model based on full-spectrum ESF demonstrated excellent diagnostic capabilities, with an AUC of 0.913 and an accuracy of 88.3%. Figure 1 ).

[0046] General applicability: The model maintained consistently high diagnostic efficacy (AUC > 0.8 in all subgroups) across different gestational weeks (<5 weeks to ≥8 weeks), ages (≤35 years, >35 years), and BMIs (<18.5, 18.5-24.9, ≥25), demonstrating its broad clinical applicability. Figure 2 ).

[0047] High throughput: This method achieves ultra-high-speed analysis of approximately 3 seconds per sample from sample processing to data acquisition. The entire batch of 378 samples can be completed in approximately 3.2 hours, far exceeding traditional chromatography-mass spectrometry techniques.

[0048] Example 2: Construction and application of a panel of metabolic biomarkers specific to ectopic pregnancy

[0049] The core of this embodiment lies in screening and identifying a refined combination of metabolic biomarkers with clear biological significance from the full spectrum of metabolic fingerprints, for the development of a simplified and efficient EP diagnostic tool.

[0050] 1. Screening and identification of markers: (1) Initial screening of differential features: Based on the 348 m / z feature data of the discovery set (263 people) constructed in Example 1, the student t test of the EP group and the IUP group was performed, and 246 m / z features with significant differences between the two groups were screened out (P<0.05).

[0051] (2) Key feature selection: The Lasso regression model was used to further reduce the dimensionality and filter the above 246 features. The regularization strength was determined by cross-validation and the feature coefficient threshold was set to ≥0.4. Finally, 20 core m / z features that contributed the most to EP classification were selected.

[0052] (3) Metabolite identification: The above 20 m / z features were accurately measured using high-precision Fourier transform ion cyclotron resonance mass spectrometry (Bruker SolariX7.0T FT-ICR-MS) to ensure that the mass error was <5ppm.

[0053] (4) The inferred molecular formulas were compared with the Human Metabolome Database (HMDB), and metabolite annotations were performed on these features to obtain a clear panel of metabolic biomarkers for EP diagnosis (Table 1).

[0054] Table 1 Diagnostic markers for EP

[0055] 2. Performance validation of the diagnostic metabolic biomarker panel

[0056] Diagnostic performance was evaluated based solely on the aforementioned metabolic markers, and the model's diagnostic capability for patients with ectopic pregnancy was tested in an independent validation set.

[0057] 3. Technical Effects

[0058] In the independent validation set, the diagnostic model based on the metabolic biomarker panel in Table 1 achieved excellent performance with an AUC of 0.922, a sensitivity of 90.9%, and a specificity of 90%. Figure 3 ).

[0059] The panel demonstrated stable diagnostic capability across patients of different gestational weeks, ages, and BMIs (the mean predictive probability effectively distinguished between EP and IUP in all subgroups, P<0.05).

[0060] This refined panel, while maintaining high precision, greatly simplifies the complexity of subsequent development into a standardized clinical test kit.

[0061] Example 3: Construction and application of a panel of metabolic biomarkers for predicting the risk of ectopic pregnancy rupture.

[0062] The purpose of this embodiment is to extend the application of metabolic fingerprinting from diagnosis to prognosis, and to build a model that can predict the risk of fallopian tube rupture in EP patients.

[0063] 1. Risk prediction cohort establishment and feature selection

[0064] (1) Cohort setup: 120 patients with clear outcomes were selected from the total number of EP patients, including 45 patients who ruptured and whose intraoperative bleeding was confirmed to be greater than 100 mL, and 75 patients whose EP did not rupture after conservative treatment or surgery.

[0065] (2) The queue is randomly divided into a discovery set and a verification set in a ratio of 7:3.

[0066] (3) In the discovery set, a combination strategy is used to screen features related to breakage from the full spectrum ESF: the preprocessed mass spectrometry data is input into the LASSO regression model to screen out the first feature set with regression coefficients that are not zero; the mass spectrometry features are trained using a random forest classifier, the feature importance score is calculated, and the second feature set with an importance score that is greater than zero is screened out.

[0067] (4) Select the intersection of the first feature set and the second feature set to finally determine a “breakdown risk metabolic biomarker panel” consisting of 12 metabolic biomarkers (Table 2).

[0068] (5) Feature integration: In order to further improve the clinical applicability of the model, the above-mentioned metabolic markers are integrated with four conventional clinical indicators (serum β-hCG level, endometrial thickness, adnexal mass size, and BMI) to form a feature set, which are used as the model input.

[0069] 2. Construction and Evaluation of the Risk Prediction Model: Using the integrated feature set (12 metabolic biomarkers + 4 clinical indicators) and the discovery set samples as the training dataset, a random forest machine learning prediction model was constructed. The specific model parameters were set as follows: 100 decision trees, 5 features considered during splitting each decision tree, a maximum decision tree depth of 3 layers, and a minimum number of splits of 5 samples (i.e., no splitting occurs when the subset sample size is less than 5). This parameter combination constructs a simplified model structure, avoiding overfitting at the algorithmic level and ensuring the model's generalization ability. After model training, the predictive performance was evaluated on an independent validation set.

[0070] Table 2. Predictive markers for EP rupture risk

[0071] 3. Technical Effects

[0072] (1) The risk prediction model constructed in this invention achieved excellent performance with an AUC of 0.885 on the validation set, and an accuracy of 87.8% ( Figure 4 ).

[0073] (2) Compared with the model constructed using only 4 clinical indicators (AUC 0.702, accuracy 67.5%), the model performance of the present invention is significantly improved (DeLong test, P<0.05). Figure 5 ).

[0074] (3) To verify the robustness of the model, the dataset was randomly split 10 times. The AUC of the model remained stable above 0.8 in all validations.

[0075] (4) This model can be used to stratify the risk of diagnosed EP patients, identify patients with high rupture risk, and guide clinicians to take more proactive intervention measures (such as prioritizing surgery), while low-risk patients can try conservative treatment to achieve individualized and precise management.

[0076] The technical solution of this invention has significant practicality and industrialization prospects, mainly reflected in the following aspects:

[0077] I. Significant technical advantages and clinical value

[0078] 1. Precise Diagnosis, Addressing Clinical Pain Points: The metabolic fingerprint diagnostic model (AUC>0.91) and metabolic biomarker panel (AUC>0.92) provided by this invention significantly outperform the existing "ultrasound + β-hCG" diagnostic model. This effectively solves the current clinical problems of diagnostic delays, misdiagnosis, and missed diagnosis, directly serving the rigid clinical need for early and accurate diagnosis, and is expected to have extremely high market acceptance.

[0079] 2. Risk Prediction, Creating a New Market: The invention's unique rupture risk prediction function (AUC 0.885) fills a gap in current clinical practice. This function provides physicians with crucial decision support, enabling early intervention for high-risk patients and avoiding overtreatment for low-risk patients, thus directly reducing maternal mortality. This groundbreaking application not only has significant social value but also creates a completely new segment of the in vitro diagnostics (IVD) market.

[0080] 3. Fast detection speed, improving diagnostic efficiency: With a single sample detection speed of nearly 3 seconds, this invention has the potential to become a "rapid testing" tool in emergency and outpatient settings. This will greatly shorten the diagnostic decision-making cycle from the current hours or even days to minutes, significantly improving the efficiency of hospital diagnostic and treatment processes, and possessing strong technological appeal and competitiveness.

[0081] II. Excellent and quantifiable performance indicators

[0082] The practicality of this invention is based on solid and repeatable experimental data, and its performance indicators are clear and leading: 1. High accuracy: The diagnostic model achieved an AUC of over 0.913 in a multicenter validation set containing 378 samples, demonstrating its robust generalization ability.

[0083] 2. High robustness: The diagnostic metabolic marker panel maintained excellent performance in subgroups of different gestational weeks, ages and BMIs, indicating that it is applicable to a wide range of patients and has strong universality.

[0084] 3. Excellent predictive performance: The performance of the risk prediction model (AUC 0.885) is significantly better than that of the traditional clinical indicator model (AUC 0.702), and the difference is statistically significant (P<0.05), providing a scientific basis for the clinical application of this function.

[0085] These quantitative indicators provide strong data support for the future registration, approval, and clinical promotion of the product.

[0086] III. Clear Production Implementation Path and Industrialization Prospects

[0087] The technical solution of this invention is easily transformed into standardized in vitro diagnostic (IVD) products, and the industrialization path is clear: 1. Core materials can be produced in a standardized manner: The nanoparticle microarray chips on which the technology relies can be produced in a standardized and large-scale manner through mature microfabrication and nanomaterial synthesis processes, ensuring the stability of product quality and consistency between batches, while keeping costs under control.

[0088] 2. The instrument platform is easy to integrate and promote: The core of the Nano-LDI-MS system is the mass spectrometer. It can collaborate with existing commercial mass spectrometer manufacturers to develop dedicated benchtop or portable detectors. With the increasing prevalence of clinical mass spectrometry technology in precision medicine, hospitals' acceptance and adoption rates of this type of equipment are rising year by year, resulting in relatively low market education costs.

[0089] 3. It can form an integrated solution of "reagents + equipment + software": the final commercial product can be planned as a closed-loop detection system, including: Disposable test chip (reagent): pre-loaded with nanomaterials; users only need to add serum samples.

[0090] Dedicated mass spectrometer: Optimized parameters enable one-click rapid detection.

[0091] Intelligent diagnostic software: It has a built-in pre-trained machine learning model that automatically analyzes metabolic fingerprint data and directly outputs "diagnostic results" and "rupture risk score".

[0092] This all-in-one machine is easy to operate and can minimize the technical requirements for operators, making it very suitable for deployment in hospital laboratories and obstetrics and gynecology emergency departments.

[0093] 4. Diversified business models: A stable profit model that combines equipment sales with reagent and consumable sales.

[0094] Testing services: For primary hospitals that are not yet able to purchase equipment, they can cooperate with third-party testing centers (ICLs) to provide centralized testing services.

[0095] In summary, this invention is not only a laboratory-level technological breakthrough, but also a technical solution integrating innovation, practicality, and market scarcity. It precisely addresses unmet clinical needs, boasts superior technical performance, has a clear industrialization path, and possesses extremely high technology transfer value and broad market prospects. The successful commercialization of this product is expected to reshape the clinical diagnosis and treatment pathway for ectopic pregnancy, creating significant economic and social benefits.

[0096] The preferred embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make numerous modifications and variations based on the concept of the present invention without creative effort. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experimentation on the basis of existing technology should be within the scope of protection defined by the claims.

Claims

1. A serum metabolite marker combination for ectopic pregnancy diagnosis and rupture risk prediction, characterized in that, Comprise: A. A serum metabolite marker combination for diagnosing ectopic pregnancy, which is composed of the following markers: Glycerol, Dihydrouracil, N-methyl-beta-alanine, Dihydrouracil, Urocanic acid, Lysine, Glucose, Glycylproline and Fucose; B. A serum metabolite marker combination for predicting the risk of ectopic pregnancy rupture, which is used to further predict the risk of tubal rupture in a patient sample determined as ectopic pregnancy positive by the serum metabolite marker combination for diagnosing ectopic pregnancy; specifically composed of the following markers: Glycine, Glyoxylic acid, Ureidoglycolate, Acetoaceticacid, Phenylalanine, Pyrophosphate, Leucine, Pyridoxine, Methionine, Semidehydroascorbic acid, Phosphorylcholine, Docosatetraenylethanolamide.

2. The serum metabolic marker combination of claim 1, characterized in that, The screening method of the serum metabolite marker combination for diagnosing ectopic pregnancy comprises the following steps: Step one, collect serum samples of the ectopic pregnancy group and the intrauterine pregnancy control group diagnosed by clinic, and use nanoparticle assisted laser desorption / ionization mass spectrometry to detect the metabolites of all the above serum samples, and obtain the serum metabolite fingerprint of each sample; Step two, based on the key clinical characteristics of gestational age, age and BMI, the total samples are divided into model discovery set and independent verification set by random stratified sampling method; Step three, the serum metabolite fingerprints of the discovery set are subjected to Student's t test of EP group and IUP group, and the m / z characteristics with significant differences between the two groups are screened out, wherein the test level is set to P<0.05; Step four, Lasso regression model is used to further reduce and screen the m / z characteristics screened out in step three, the regularization strength is determined by cross-validation, and the feature coefficient threshold is set to ≥0.4, and finally 20 core m / z characteristics with the largest contribution to EP classification are selected; Step five, high-precision Fourier transform ion cyclotron resonance mass spectrometry is used to accurately measure the mass of the m / z characteristics selected in step four with significant differences, to ensure that the mass error is less than 5 ppm; compare the presumed molecular formula with the human metabolome database, annotate the metabolites of these characteristics, and obtain the serum metabolite marker combination for diagnosing ectopic pregnancy.

3. The serum metabolic marker combination of claim 2, wherein, In step one, the serum metabolic fingerprint acquisition method is as follows: first, Fe3O4 nanoparticles are synthesized by a solvothermal method, then the Fe3O4 nanoparticles are dispersed in an ethanol / water solution to prepare a matrix solution with a concentration of 1 mg / mL; 1.5 μL of a serum sample diluted 10 times is spotted on a 384-well polished steel target plate, and after natural drying at room temperature, 1.5 μL of the matrix solution is added and again naturally dried at room temperature to obtain a serum sample for mass spectrometry analysis; then a Bruker Autoflex time-of-flight mass spectrometer is used for detection, and the detection parameters are set as follows: laser wavelength 355 nm, pulse frequency 1000 Hz, and 2000 laser signals are collected for each point; finally, the obtained original mass spectrum data are processed to obtain a set of stable m / z characteristics, which constitute the serum metabolic fingerprint of each sample.

4. The serum metabolic marker combination of claim 3, wherein, The preparation method of the Fe3O4 nanoparticles is as follows: 0.6 g of iron trichloride hexahydrate, 0.15 g of citric acid trisodium dihydrate and 0.96 g of anhydrous sodium acetate are dissolved in 25 mL of ethylene glycol, and a hydrothermal reaction is carried out at 200°C for 10 hours; the reaction product is washed with deionized water and ethanol and then dried to obtain Fe3O4 nanoparticles.

5. The serum metabolic marker combination of claim 3, wherein, The original mass spectrum data obtained in step one are processed using Python 3.8, and the specific method is as follows: peaks with a signal-to-noise ratio of ≥3 are extracted, peak alignment is performed, and peaks with a frequency of less than 2 / 3 in any group are filtered out, finally 348 stable m / z characteristics are obtained, which constitute the serum metabolic fingerprint of each sample.

6. The serum metabolic marker combination of claim 2, wherein, The screening method of the serum metabolic marker combination for predicting the risk of ectopic pregnancy rupture comprises the following steps: a. From the total ectopic pregnancy patient samples, multiple samples with clear outcomes are selected to construct a cohort, which includes patients with rupture and intraoperative confirmed blood loss greater than 100 mL, and patients confirmed not to rupture by conservative treatment or surgery; the cohort is randomly divided into a discovery set and a validation set in a ratio of 7:3; b. In the discovery set, the following two strategies are used to screen features related to rupture from the full spectrum serum metabolic fingerprint: Using Lasso regression, the first feature set with non-zero regression coefficients is screened out; Using a random forest algorithm, the importance score of each m / z feature is calculated, and the second feature set with an importance score greater than zero is screened out; c. The intersection of the first feature set and the second feature set is selected, and finally a serum metabolic marker combination composed of 12 metabolic markers for predicting the risk of ectopic pregnancy rupture is determined.

7. The use of the serum metabolic marker combination of claim 2 in the preparation of a detection product for ectopic pregnancy diagnosis and rupture risk prediction.

8. The use according to claim 7, characterized in that, The detection product comprises an ectopic pregnancy diagnosis model, and the construction method of the diagnosis model comprises the following steps: 1) Based on the key clinical characteristics of gestational age, age and BMI, the serum samples of the clinically diagnosed ectopic pregnancy group and the intrauterine pregnancy control group are divided into a model discovery set and an independent validation set in a ratio of 7:3 by random stratified sampling. ​ 2) The relative abundance values of m / z features of the serum metabolic fingerprint or the serum metabolic marker combination for diagnosing ectopic pregnancy in the discovery set were analyzed using the Orange data mining platform, and various machine learning algorithms were used for model training, and five-fold cross-validation was used to optimize model parameters; 3) The trained model was tested on an independent validation set to evaluate its diagnostic performance.

9. Use according to claim 8, wherein the compound is ###0002### In step 2), the machine learning algorithms include Lasso regression, neural networks, gradient boosting, ridge regression, support vector machines, and random forests.

10. The use according to claim 7, wherein the compound is ###00010### or a pharmaceutically acceptable salt thereof. The detection product contains an ectopic pregnancy rupture risk prediction model, and the construction method of the ectopic pregnancy rupture risk prediction model is to train a machine learning model on the feature set of the relative abundance values of the m / z features of the rupture risk metabolic marker combination and four clinical indicators, namely serum β-hCG level, endometrial thickness, adnexal mass size, and BMI, and test the trained model on an independent validation set to evaluate its diagnostic performance.

Images