Diagnostic marker combination, and construction method and application of ovarian endometriosis prediction model

Abstract

The invention discloses a diagnostic marker combination, and a construction method and application of an ovarian endometriosis prediction model. A serum metabonomics technology and an artificial intelligence data analysis technology are adopted to construct the model for predicting the occurrence risk of the ovarian endometriosis patient, the diagnostic marker screening method is high in operability, the model construction method is simple, the obtained diagnostic model is good in effect, diagnosis can be conducted only through blood sampling, convenience and rapidness are achieved, no internal wound exists, and the method is suitable for large-scale popularization and application. The kit is high in sensitivity and good in specificity for predicting the occurrence risk of ovarian endometriosis, and has a very good clinical application value.

Description

Technical Field

[0001] The present invention belongs to the technical field of diagnostic marker screening, and specifically relates to a diagnostic marker combination, a construction method and application of an ovarian endometriosis prediction model. Background Art

[0002] Endometriosis is a common estrogen-dependent chronic inflammatory gynecological disease, which usually affects the ovaries, uterine ligaments, rectum, bladder, etc. (Giudice LC. Clinical practice. Endometriosis. N Engl J Med 2010, 362(25): 2389-2398.). Its prominent clinical manifestations are chronic pelvic pain, dysmenorrhea, dyspareunia and infertility, which seriously affect women's work and daily life (Vercellini P, Vigano P, Somigliana E, Fedele L. Endometriosis: pathogenesis and treatment. Nat Rev Endocrinol 2014, 10(5): 261-275.).

[0003] In particular, endometriosis (Ovarian Endometriosis, OvE) that is ectopic to the ovaries is one of the high-risk factors for ovarian cancer (Vercellini P, Vigano P, Somigliana E, Fedele L. Endometriosis: pathogenesis and treatment. Nat Rev Endocrinol 2014, 10 (5): 261-275.). Therefore, early diagnosis and early treatment of OvE are extremely important for women's healthy life. However, the prevalence of delayed diagnosis of OvE is not optimistic. Studies have shown that OvE is usually delayed for 8-10 years before being discovered (Hudelist G, Fritzer N, Thomas A, Niehues C, Oppelt P, Haas D, Tammaa A, Salzer H. Diagnostic delay for endometriosis in Austria and Germany: causes and possible consequences. Hum Reprod 2012, 27 (12): 3412-3416.). Delayed diagnosis not only directly affects disease treatment and prognosis, but may also increase the risk of OvE recurrence. Therefore, the diagnosis of OvE is of great significance for disease treatment and improving the quality of life of patients. At present, the clinical diagnosis of OvE can be carried out in combination with clinical manifestations, signs, imaging examinations and biomarkers (Giudice LC, Kao LC. Endometriosis. Lancet 2004, 364 (9447): 1789-1799.), but the clinical manifestations of OvE are atypical, and the results of gynecological examinations, ultrasonography and magnetic resonance imaging (MRI) are greatly affected by the experience and skills of doctors. In addition, there is currently no biomarker that can accurately diagnose OvE. Although serum CA125 has a certain correlation with OvE, its specificity and sensitivity are limited and it is not a specific indicator of OvE.The detection of this indicator has limited diagnostic significance for OvE (Hirsch M, Duffy J, Davis CJ, Nieves Plana M, Khan KS, International Collaboration to Harmonise O, Measures for E. Diagnostic accuracy of cancer antigen 125 for endometriosis: a systematic review and meta-analysis. BJOG 2016, 123(11): 1761-1768.). Its elevated level is more common in patients with severe OvE, adenomyosis, obvious pelvic inflammation, and combined with OvE cyst rupture. Clinical diagnosis is of great significance for the early intervention and treatment of OvE. Biomarkers are objective indicators that are not affected by physician experience and skills (DF, Flores I, Waelkens E, D'Hooghe T. Noninvasive diagnosis of endometriosis: Review of current peripheral blood and endometrial biomarkers. Best Pract Res Clin Obstet Gynaecol 2018, 50: 72-83.). Therefore, it is urgent to screen new indicators that can assist in improving the early noninvasive diagnosis of OvE.

[0004] Metabolomics, as a research field that has emerged in recent years, provides a powerful tool for the analysis of disease mechanisms and the mining of biomarkers by comprehensively analyzing the changes in small molecule metabolites in organisms. This technology can capture the subtle changes in the metabolic pathways of organisms in disease states, providing a new perspective for understanding OvE. In this context, the introduction of machine learning technology has made it possible to deeply mine metabolomics data. In particular, the Lasso method, as an advanced feature selection technology, can effectively screen out the most predictive features in high-dimensional data, while dealing with complex collinearity problems in the data and improving the stability and accuracy of the model. By combining metabolomics with the Lasso machine learning algorithm, the present invention aims to screen out key metabolic biomarkers from complex metabolite data. Summary of the invention

[0005] Purpose of the invention: The technical problem to be solved by the present invention is to provide diagnostic markers for ovarian endometriosis and their screening methods and applications.

[0006] In order to achieve the above-mentioned purpose, the present invention adopts the following technical scheme: The present invention provides a diagnostic marker combination, which includes N-acetyl-aspartic acid, lactic acid, (R)-3-hydroxybutyric acid, hippuric acid, indole-3-acetylvaline and ascorbic acid.

[0007] The present invention also provides the use of the diagnostic marker combination in the preparation of ovarian endometriosis prediction and / or diagnosis products.

[0008] The present invention also provides the application of the diagnostic marker combination in constructing a prediction model for ovarian endometriosis.

[0009] The present invention also provides a method for constructing an ovarian endometriosis prediction model, comprising the following steps:

[0010] 1) Compare the serum data of patients with ovarian endometriosis and normal healthy women through non-targeted metabolomics screening, and screen out metabolites with significant differences based on AUC>0.7;

[0011] 2) Further weighted gene co-expression network analysis was performed on the metabolites with significant differences, and the metabolites with significant differences were associated with the clinical characteristics of OvE according to their different expression patterns, and the parameters significantly correlated with the clinical characteristics of OvE were screened;

[0012] 3) Take the intersection of AUC>0.7 and Black model in WGCNA to further determine the metabolites with specific differences;

[0013] 4) Expand the sample size for targeted metabolomics validation, and further verify the metabolites with specific differences through multiple machine learning methods, further reduce the dimension through Lasso, and finally screen out the characteristic parameters for predicting ovarian endometriosis and build a prediction model;

[0014] The calculation formula of the prediction model is:

[0015]

[0016] Where n = the number of characteristic parameters for ovarian endometriosis prediction, exp i represents the expression value of the characteristic parameter for predicting ovarian endometriosis, β i Regression coefficients representing characteristic parameters for prediction of ovarian endometriosis.

[0017] Among them, the parameters significantly correlated with the clinical characteristics of OvE in step 2) include dysmenorrhea and CA125.

[0018] Among them, the metabolites with specific differences in step 3) include gallic acid, N-acetyl-aspartic acid, aspartic acid, alanine, lactic acid, isoflurane, arginine, N-acetyl-DL-tryptophan, (R)-3-hydroxybutyric acid, quinic acid, hippuric acid, 4-methylumbelliferyl sulfate potassium salt, ellagic acid, quercetin, magnolol, mysamine, β-creatinine, indole-3-acetylvaline, ascorbic acid, melatonin and sebacic acid.

[0019] Among them, the multiple machine learning methods in step 4) include one or more of convolutional neural network, naive Bayes algorithm, random forest algorithm, support vector machine algorithm and Lasso-logit regression method.

[0020] Among them, the characteristic parameters for predicting ovarian endometriosis in step 4) include N-acetyl-aspartic acid, lactic acid, (R)-3-hydroxybutyric acid, hippuric acid, indole-3-acetylvaline, ascorbic acid, 2 clinical indicators of dysmenorrhea and CA125.

[0021] Among them, the formula of the prediction model in step 4) is: Total score = 22.471×N-acetyl-aspartic acid + 0.011×lactic acid + 0.008×(R)-3-hydroxybutyric acid - 0.333×hippuric acid + 0.204×indole-3-acetylvaline - 0.685×ascorbic acid + 13.601×dysmenorrhea + 17.728×CA125 + 141.159, the threshold of Total score is 170.17, when the individual patient score is higher than 170.17, it is defined as a high-risk group for OvE, and lower than 170.17, it is defined as a low-risk group for OvE.

[0022] The present invention also provides a diagnostic marker kit for predicting OvE risk, which contains the diagnostic marker combination.

[0023] In the present invention, 34 OvE and 34 normal healthy women in the OvE cohort were screened for non-targeted metabolomics using high performance liquid chromatography-mass spectrometry in the early stage. After database comparison, a total of 56 metabolites with significant differences were screened out, and further WGCNA screening was performed on the basis of AUC>0.7, and finally 21 metabolites were selected; the sample was expanded (107 OvE and 130 HC) for targeted metabolomics, and further five machine learning methods showed that these 21 metabolites were significantly different in HC and OvE. Further Lasso dimensionality reduction was used to screen a total of 6 metabolites (N-acetyl-aspartic acid, lactic acid, (R)-3-hydroxybutyric acid, hippuric acid, indole-3-acetylvaline and ascorbic acid) and a prediction model consisting of two clinical indicators (dysmenorrhea and CA125) were screened out.

[0024] Beneficial effects: Compared with the prior art, the present invention has the following advantages: the present invention uses serum metabolomics technology and artificial intelligence data analysis technology to construct a prediction model for the risk of OvE patients, the diagnostic marker screening method of the present invention is highly operable, the model construction method is simple, the obtained diagnostic model has good effect, and the diagnosis can be performed only by blood sampling, which is convenient, fast and non-invasive, and has high sensitivity and good specificity for predicting the risk of OvE, and has good clinical application value. The present invention also provides a diagnostic marker kit containing the above-mentioned diagnostic marker suitable for predicting the risk of OvE, which can be used to predict the risk of OvE. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 Diagnostic performance diagram of differential metabolites of ovarian endometriosis (AUC>0.7); A, nomogram model for predicting the risk of OvE; BC, Ttrain model (70%) prediction; D, test model (30%) diagnosis prediction; EF, C-Index model predicting the risk of OvE.

[0026] Figure 2 The module-trait relationship heat map of WGCNA analyzes the correlation between metabolites and clinical pathological characteristics;

[0027] Figure 3 The graph shows the concentration differences of the 21 selected metabolites between the HC and OvE groups;

[0028] Figure 4 Results graphs of the diagnostic efficacy of the selected metabolites predicted by the machine learning algorithm. A. Decision curve analysis of the training set of the selected metabolites through five machine learning analyses; B. ROC curve of the training set performance of the selected metabolites obtained through machine learning analyses; C. PRC curve obtained from the training set performance of the selected metabolites through machine learning analyses; D. Decision curve analysis of the test set of the selected metabolites through CNN, NB, RF, SVM, and lasso-logit analyses; (E). ROC of the test set performance of the selected metabolites through machine learning analyses; (F). PRC of the test set performance of the selected metabolites through five machine learning analyses;

[0029] Figure 5 The following are the diagrams of the prediction model construction for predicting the occurrence of OvE: A. Forest plot based on multivariate LASSO Logistic regression analysis; B. Nomogram model for predicting the risk of OvE occurrence;

[0030] Figure 6is the diagnostic efficacy diagram of the prediction model for the occurrence of OvE. AB is the distribution of risk scores and the expression of metabolites involved in the formation of risk scores. The threshold was defined as 170.17, and all participants were divided into low-risk (<cutoff) and high-risk (>cutoff) score groups from the training and test sets: CD Confusion matrix of the diagnostic model in the training and test sets; EF, the C index of the training set was 0.93 and the C index of the test set was 0.92. DETAILED DESCRIPTION

[0031] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0032] It should be noted that the terms "including" and "having" and any variations thereof in the specification and claims of the present application and the above-mentioned drawings are intended to cover non-exclusive inclusions. For example, a process, method, system, product or apparatus comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or apparatuses.

[0033] Example 1 Screening method suitable for predicting the risk of OvE

[0034] In this embodiment, targeted metabolomics analysis was performed on each population sample using high performance liquid chromatography-mass spectrometry to obtain the original targeted metabolic fingerprint of each serum sample. According to the mass number and abundance of the fragment ions, the parent ion-daughter ion pair was selected by bioinformatics resource database, literature reports and mass spectrometry multiple reaction monitoring (MRM) technology prediction software, and multiple MRM channels were determined for each screened metabolite. The sample preparation process and conditions such as chromatography-mass spectrometry were adjusted, including selecting parent ion-daughter ion pairs, improving chromatographic peak shape, improving mass spectrometry signal-to-noise ratio, etc., to perform synchronous quantitative detection of 21 metabolites. Mass spectrometry multiple reaction monitoring technology was used for each analysis sample, and electrospray ionization cations and electrospray ionization anions were used in the positive ion mode of the electrospray ion source.

[0035] The specific steps are as follows:

[0036] 1) The serum data of 34 patients with ovarian endometriosis and 34 normal healthy women were compared by non-targeted metabolomics screening to screen for significantly different metabolites; 5153 and 6241 metabolic characteristic peaks were obtained in positive and negative ion modes, respectively. Using FC>2, P<0.05 and VIP>1.5 as screening conditions, the differential metabolic peaks were imported into PubChem and HMDB databases for comparison, and a total of 56 endogenous differential metabolites were identified, of which 15 were upregulated in the OvE group and 41 were downregulated. The specific methods are as follows:

[0037] 1.1) Serum metabolite extraction

[0038] 200 μL of serum stored at -80°C was taken out and fully thawed on ice at 4°C, and 800 μL of -80°C pre-cooled methanol (containing 12.5 μmol / L 13C5-glutamic acid: CLM-1822-H-PK, Cambridge Isotope Laboratories, Inc.) was added. The mixture was vortexed for 5 min and incubated in a -80°C refrigerator for 8 h to fully quench the protein and extract small molecule metabolites. Centrifuge at 16,000g for 15 min at 4℃, carefully remove the sample and transfer the supernatant to a new centrifuge tube, add 200μL of 80% methanol water precooled at -80℃ to the precipitate, ultrasonicate at 4℃ for 5min, vortex vigorously for 5min to fully resuspend the precipitate, centrifuge at 16,000g for 15min at 4℃, remove the supernatant, combine the two supernatants and concentrate to dryness by vacuum centrifugation at 4℃, re-dissolve the sample with 100μL of 80% methanol aqueous solution, centrifuge at 18,000g for 15min at 4℃, remove the supernatant, put it into a mass spectrometer bottle and store it at 4℃ until it is loaded on the machine.

[0039] 1.2) LC-MS / MS analysis

[0040] Each sample was analyzed by high performance liquid chromatography mass spectrometry for plasma targeted metabolism analysis, chromatographic separation using LC-MS / MS analysis, Phenomenex C18 (2.1x 100mm, 2.6μm) (Phenomene, USA) column for ultra-high performance liquid chromatography AB SCIEX ExionLC TMThe AD system was connected in series with the AB SCIEX QTRAP 6500 (AB SCIEX, USA) system mass spectrometer. The acquisition mode selected the positive and negative ion modes for simultaneous acquisition. The mobile phase A (aqueous phase) was 0.1% formic acid water, and the mobile phase B (organic phase) was 0.1% formic acid acetonitrile. The linear elution gradient program used in the analysis was as follows: B: 5% (0 min) → 70% (5.5 min) → 95% (6.5 min) → 95% (7 min) → 5% (7.1 min) → 5% (8 min). The column oven temperature was set at 40°C, the autosampler temperature was set at 4°C, the flow rate was set at 0.3 ml / min, and the injection volume was set at 3 μL. The IonSpray Voltage of the positive ion mode and the negative ion mode were set to 4500 V and -4500 V respectively, the ion source temperature was set to 450°C, the Curtain Gas was set to 35 psi, the Ion Source Gas1 was set to 50 psi, the Ion Source Gas2 was set to 50 psi, and the Collision Gas was set to Medium. The specific parameters are shown in Table 1.

[0041] Table 1

[0042]

[0043]

[0044] 2) The metabolites that showed differences through PRC curve diagnosis had a good effect in diagnosing OvE. Among them, there were 39 metabolites with an area under the curve AUC>0.7 ( Figure 1 );

[0045] 3) We further performed weighted gene co-expression network analysis (WGCNA) on the metabolites, divided the metabolites into 9 modules according to their different expression patterns, and then associated the 9 modules with the clinical characteristics of OvE. The results showed that the metabolites in the Black model were significantly correlated with the clinical characteristics of OvE: the correlation with dysmenorrhea was 0.55 (P = 1e-06), the correlation with CA125 was 0.73 (P = 2e-12), and the correlation with clinical qualitative OvE and staging was as high as 0.94 (P = 8e-33) and 0.91 (P = 1e-26) ( Figure 2 ).

[0046] 4) The intersection of AUC>0.7 and the Black model in WGCNA was taken, and a total of 21 metabolites were further tested: gallic acid, N-acetyl-aspartic acid, aspartic acid, alanine, lactic acid, isoflurane, arginine, N-acetyl-DL-tryptophan, (R)-3-hydroxybutyric acid, quinic acid, hippuric acid, 4-methylumbelliferyl sulfate potassium salt, ellagic acid, quercetin, magnolol, mysamine, β-creatinine, indole-3-acetylvaline, ascorbic acid, melatonin and sebacic acid;

[0047] 5) Expanded sample size (130 normal people: 107 OvE patients) for targeted metabolomics validation, the method is the same as above, and finally, 16 of the 21 metabolites were significantly different, namely: gallic acid, N-acetyl-aspartic acid, alanine, lactic acid, isoflurane, N-acetyl-DL-tryptophan, (R)-3-hydroxybutyric acid, quinic acid, hippuric acid, ellagic acid, quercetin, magnolol, β-creatinine, indole-3-acetylvaline, ascorbic acid, melatonin ( Figure 3 );

[0048] 6) The diagnostic effects of 21 metabolites were further verified by five machine learning algorithms: Convolutional Neural Networks (CNN), Naive Bayes (NB), Random Forest (RF), Support Vector Machine (SVM) and Lasso-logit. The AUC and AUPRC showed that they had good diagnostic effects in OvE and HC ( Figure 4 );

[0049] 7) The obtained patient two-dimensional mass spectrometry sample data matrix was randomly divided into a training set and a test set at a ratio of 70% and 30%, respectively. The Lasso regression variable screening was performed using the glmnet package in R. The variables were constrained using L1 norm regularization. The convergence model complexity was adjusted by optimizing the λ parameter. The number of iterations was 200, and λ.1se=0.1 was determined. A 10-fold cross validation was used to screen the characteristic variables based on the minimum model error. Finally, six non-zero regression coefficient characteristic variables including N-acetyl-aspartic acid, lactic acid, (R)-3-hydroxybutyric acid, hippuric acid, indole-3-acetylvaline and ascorbic acid were determined ( Figure 5 A), the coxph function in the survival package was used to perform multi-factor fitting regression analysis on the selected characteristic variables, and the OvE nomogram model was established ( Figure 5B). Finally, a total of 6 metabolites, N-acetyl-aspartate, lactate, (R)-3-hydroxybutyrate, hippuric acid, indole-3-acetylvaline and ascorbic acid, and 2 clinical indicators, dysmenorrhea and CA125, were screened to form the prediction model.

[0050] According to the above Lasso-cox model, a nomogram for predicting the risk of OvE was established by logistic regression. The Total score of individual patients was calculated by the obtained Total score linear equation (Total score = 22.471 × N-acetyl-aspartic acid + 0.011 × lactic acid + 0.008 × (R)-3-hydroxybutyric acid - 0.333 × hippuric acid + 0.204 × indole-3-acetylvaline - 0.685 × ascorbic acid + 13.601 × dysmenorrhea + 17.728 × carbohydrate antigen 125 + 141.159). According to the ROC curve of Total score, the optimal threshold point was determined to be the Youden value = 170.17 ( Figure 6 A and 6B), individual patients with scores above 170.17 are defined as high-risk patients for OvE, and those below 170.17 are defined as low-risk patients for OvE. According to the patient risk level analysis, patients in the OvE group have a higher Total score. Statistics show that the proportion of patients in the OvE group who are defined as high-risk is significantly higher than that in the normal healthy female group. Figure 2 The confusion matrix shows that using this model in the training set, 61 out of 68 people with high risk (AB) were diagnosed with endometriosis, 7 were normal, 88 were normal, and 13 were OvE in the low risk; in the test set, 30 out of 31 people with high risk (6C-6D) were OvE, and only one was normal; 34 were normal and 3 were OvE patients in the low risk. In the training set, the true positive rate of predicting EM was 61 / 68 (89.70%), and the true negative rate was 88 / 101 (87.12%). In the test set, the true positive rate of predicting EM was 30 / 31 (96.77%), and the true negative rate was 34 / 37 (91.89%). This shows that the diagnostic results are accurate.

[0051] Calibration curves were performed using 1000 bootstrap resamplings with replacement in the model development cohort, demonstrating the robustness and excellent predictive performance of our model ( Figure 6 E). In addition, the consistency index C-index of the training set and the test set is greater than 0.9 ( Figure 6 F), which shows that the model has high prediction accuracy.

Claims

1. A diagnostic marker combination, characterized in that: The marker combination includes N-acetyl-aspartate, lactate, (R)-3-hydroxybutyrate, hippuric acid, indole-3-acetylvaline, and ascorbic acid.

2. Use of the diagnostic marker combination according to claim 1 in the preparation of a prediction and / or diagnosis product for ovarian endometriosis.

3. Use of the diagnostic marker combination according to claim 1 in constructing a prediction model for ovarian endometriosis.

4. A method for constructing a prediction model for ovarian endometriosis, characterized in that: The steps include: 1) Compare the serum data of patients with ovarian endometriosis and normal healthy women through non-targeted metabolomics screening, and screen out metabolites with significant differences based on AUC>0.7; 2) Further weighted gene co-expression network analysis was performed on the metabolites with significant differences, and the metabolites with significant differences were associated with the clinical characteristics of OvE according to their different expression patterns, and the parameters significantly correlated with the clinical characteristics of OvE were screened; 3) Take the intersection of AUC>0.7 and Black model in WGCNA to further determine the metabolites with specific differences; 4) Expand the sample size for targeted metabolomics validation, and further verify the metabolites with specific differences through multiple machine learning methods, further reduce the dimension through Lasso, and finally screen out the characteristic parameters for predicting ovarian endometriosis and build a prediction model; The calculation formula of the prediction model is: Where n = the number of characteristic parameters for ovarian endometriosis prediction, exp i represents the expression value of the characteristic parameter for predicting ovarian endometriosis, β i Regression coefficients representing characteristic parameters for prediction of ovarian endometriosis.

5. The method for constructing a prediction model for ovarian endometriosis according to claim 4, characterized in that: Parameters in step 2) that were significantly associated with the clinical features of OvE included dysmenorrhea and CA125.

6. The method for constructing a prediction model for ovarian endometriosis according to claim 4, characterized in that: The metabolites with specific differences in step 3) include gallic acid, N-acetyl-aspartic acid, aspartic acid, alanine, lactic acid, isoflurane, arginine, N-acetyl-DL-tryptophan, (R)-3-hydroxybutyric acid, quinic acid, hippuric acid, 4-methylumbelliferyl sulfate potassium salt, ellagic acid, quercetin, magnolol, mysamine, β-creatinine, indole-3-acetylvaline, ascorbic acid, melatonin and sebacic acid.

7. The method for constructing a prediction model for ovarian endometriosis according to claim 4, characterized in that: The multiple machine learning methods in step 4) include one or more of convolutional neural network, naive Bayes algorithm, random forest algorithm, support vector machine algorithm and Lasso-logit regression method.

8. The method for constructing a prediction model for ovarian endometriosis according to claim 4, characterized in that: The characteristic parameters for predicting ovarian endometriosis in step 4) include N-acetyl-aspartic acid, lactic acid, (R)-3-hydroxybutyric acid, hippuric acid, indole-3-acetylvaline, ascorbic acid, 2 clinical indicators of dysmenorrhea and CA125.

9. The method for constructing a prediction model for ovarian endometriosis according to claim 4, characterized in that: The formula of the prediction model in step 4) is: Total score = 22.471×N-acetyl-aspartic acid + 0.011×lactic acid + 0.008×(R)-3-hydroxybutyric acid - 0.333×hippuric acid + 0.204×indole-3-acetylvaline - 0.685×ascorbic acid + 13.601×dysmenorrhea + 17.728×CA125 + 141.

159. The threshold of Total score is 170.

17. When the individual patient score is higher than 170.17, it is defined as a high-risk group for OvE, and lower than 170.17 is defined as a low-risk group for OvE.

10. A diagnostic marker kit for predicting OvE risk, characterized in that: It contains the diagnostic marker combination according to claim 1.

Images