Construction method and evaluation system of polycystic ovarian syndrome prediction model

Abstract

The invention relates to a construction method of a polycystic ovarian syndrome prediction model and an evaluation system, the method and the system based on the method are included, in the model construction part of the method, polycystic ovarian syndrome patients are divided into overweight obese polycystic ovarian syndrome patients and non-overweight obese polycystic ovarian syndrome patients, the influence of overweight obesity on the metabolome of the patient is avoided to a certain extent, so that the accurate diagnosis of the polycystic ovarian syndrome is realized, and the constructed prediction model fully considers the clinical characteristics of the disease. Patients are divided into overweight obese polycystic ovarian syndrome and non-overweight obese polycystic ovarian syndrome, which are unified with a new evidence-based diagnosis guide and thought proposed by an international polycystic ovarian syndrome expert group, and the convenience of clinical practical operation of a nuclear magnetic technology is fully utilized; an important reference tool can be provided for auxiliary diagnosis of the polycystic ovarian syndrome, and the application range is wide.

Description

Technical Field

[0001] The present invention relates to the field of medical technology, and in particular to a method for constructing a prediction model for polycystic ovary syndrome and an evaluation system. Background Art

[0002] Polycystic Ovary Syndrome (PCOS) affects 5-20% of women of childbearing age worldwide. Its clinical manifestations are highly heterogeneous, and the definition of diagnosis is still controversial. Due to the limitations of knowledge about PCOS and the large number of unknown scientific issues about the disease, it is greatly difficult to accurately diagnose PCOS.

[0003] In the existing technology, the clinical diagnosis of PCOS is mostly based on the 2003 Rotterdam criteria, but this standard is specified by European and American countries and is based on the European and American race. Since the androgen levels and metabolism of Chinese women are significantly different from those of European and American women, this diagnostic standard has a certain possibility of missed diagnosis and misdiagnosis.

[0004] At the same time, although overweight and obesity are not one of the criteria for diagnosing PCOS, overweight and obesity themselves will affect the patient's metabolites and thus affect the diagnosis of the disease. The metabolism of overweight and obese PCOS patients and non-overweight and obese PCOS patients shows certain differences, but the existing technology does not classify PCOS into overweight and obese and non-overweight and obese categories, resulting in a lack of diagnostic accuracy. Summary of the invention

[0005] The technical problem to be solved by the present invention is how to construct a polycystic ovary syndrome prediction model and an evaluation system with high-precision diagnostic capabilities. In order to overcome the defects of the above-mentioned prior art, the present invention provides a method for constructing a polycystic ovary syndrome prediction model and an evaluation system, which includes a method for constructing a polycystic ovary syndrome prediction model and a system for constructing and evaluating a polycystic ovary syndrome prediction model.

[0006] The present invention provides a method for constructing a prediction model for polycystic ovary syndrome, comprising the following steps:

[0007] S1: Several overweight and obese patients with PCOS and several normal women were used as the first comparison group, and several non-overweight and obese patients with PCOS and several normal women were used as the second comparison group. Blood metabolites of each comparison group were extracted by blood metabolomics detection to obtain the blood metabolite information corresponding to each comparison group.

[0008] S2: The corresponding blood metabolite information of each comparison group was subjected to principal component analysis, partial least squares discriminant analysis and independent sample t-test to screen the differential metabolites corresponding to each comparison group through variable importance weight value and p-value;

[0009] S3: A disease prediction model is constructed by binary logistic regression using the differential metabolites corresponding to each comparison group, and the model discrimination and model calibration are tested to evaluate the disease prediction model to obtain an evaluation result.

[0010] The method for constructing a PCOS prediction model disclosed in the present invention can avoid the impact of overweight and obesity on the patient's metabolic group to a certain extent by dividing PCOS patients into overweight and obese PCOS patients and non-overweight and obese PCOS patients, thereby achieving accurate diagnosis of PCOS.

[0011] At the same time, the method for constructing a prediction model for polycystic ovary syndrome disclosed in the present invention extracts blood metabolites from each comparison group through a blood metabolomics detection method to obtain the blood metabolite information corresponding to each comparison group. The blood lipid indicators in traditional biochemical detection only include TG, total cholesterol (total cholesterol, TC), LDL, HDL. Although the technology has improved and can be expanded to apolipoprotein B (Apolipoprotein B, ApoB), etc., the detection indicators are still less and cannot meet the needs of clinical model construction. At the same time, in recent years, mass spectrometry technology has also received certain attention for the detection of small molecules. However, compared with nuclear magnetic resonance detection technology, mass spectrometry technology has a complex pre-treatment process, destructive detection, and serious data drift, requiring batch detection. However, the blood metabolomics detection method of the present invention does not require pre-treatment, and the sample can be recovered and stored for a long time, the data drift is less than 2%, and more detectable indicators can be detected, which is particularly suitable for the detection of individual samples.

[0012] Finally, the method for constructing a prediction model for PCOS disclosed in the present invention fully considers the clinical characteristics of the disease and divides patients into overweight and obese PCOS and non-overweight and obese PCOS. This model is consistent with the new evidence-based diagnostic guidelines proposed by the International PCOS Expert Group and fully utilizes the convenience of nuclear magnetic resonance technology in clinical practice. It can provide an important reference tool for the auxiliary diagnosis of PCOS and is suitable for a wide range of people.

[0013] In a possible implementation, step S1 includes the following steps:

[0014] S11: According to the overweight standard of body mass index, multiple PCOS patients diagnosed based on the Rotterdam diagnostic criteria were divided into a group of overweight and obese PCOS patients and a group of non-overweight and obese PCOS patients;

[0015] S12: Randomly select a number of normal women as the control group;

[0016] Randomly select a number of overweight, obese, and polycystic ovary syndrome patients from the set of overweight, obese, and polycystic ovary syndrome patients, and use them and the control group as the first comparison group;

[0017] Randomly select a number of non-overweight, obese, polycystic ovary syndrome patients from the non-overweight, obese, polycystic ovary syndrome patient set, and use them and the control group as the second comparison group;

[0018] S13: Blood samples from the first comparison group and the second comparison group were collected respectively, and targeted nuclear magnetic resonance metabolomics based on nuclear magnetic resonance spectroscopy was used to analyze the blood samples. 1 H-NMR spectrum detects the blood metabolite information corresponding to each comparison group from the blood samples of each comparison group, wherein the blood metabolite information includes lipid metabolite information and small molecule metabolite information;

[0019] This solution fully considers the clinical characteristics of PCOS at the beginning of building the disease prediction model, divides PCOS patients into overweight and obese PCOS and non-overweight and obese PCOS, and then extracts blood samples, thereby ensuring the prediction accuracy and wide application of the disease prediction model.

[0020] In a possible implementation, the overweight standard of the body mass index is expressed as follows:

[0021] Body mass index greater than or equal to 24 kg / m 2 of PCOS patients are overweight and obese PCOS patients;

[0022] Body mass index less than 24 kg / m 2 of the PCOS patients were non-overweight or obese PCOS patients;

[0023] This scheme provides a basis for dividing PCOS patients into overweight and obese PCOS patients and non-overweight and obese PCOS patients. Moreover, this basis for division fully considers the physical characteristics of patients and can further ensure the improvement of the accuracy of the obtained model.

[0024] In a possible implementation manner, step S2 includes the following steps:

[0025] S21: performing principal component analysis on the blood metabolite information corresponding to the first comparison group and the blood metabolite information corresponding to the second comparison group obtained in step S13, respectively, to obtain respective corresponding principal component analysis results;

[0026] S22: performing partial least squares discriminant analysis and independent sample t-test on the principal component analysis results corresponding to the first comparison group and the principal component analysis results corresponding to the second comparison group obtained in step S21, respectively, to obtain the variable importance weight value distribution and p-value distribution corresponding to each;

[0027] S23: according to the variable importance weight value distribution and p-value distribution corresponding to the first comparison group obtained in step S22, the metabolites that meet the difference criteria are used as the differential metabolites in the lipid metabolites and the differential metabolites in the small molecule metabolites corresponding to the first comparison group; according to the variable importance weight value distribution and p-value distribution corresponding to the second comparison group obtained in step S22, the metabolites that meet the difference criteria are used as the differential metabolites in the lipid metabolites and the differential metabolites in the small molecule metabolites corresponding to the second comparison group, to obtain differential metabolite information;

[0028] This scheme can distinguish the overall differences in metabolic profiles between groups, screen differential metabolites between groups, and obtain differential metabolite information by performing partial least squares discriminant analysis.

[0029] In a possible implementation, the difference standard is:

[0030] Metabolites that simultaneously satisfy the variable importance weight value greater than 1 and the p value less than 0.05 are considered differential metabolites, otherwise they are considered non-differential metabolites;

[0031] Since the variable importance weight value is the variable weight value of the partial least squares discriminant analysis model variable, it can be used to measure the influence and explanatory power of the difference in the accumulation of each metabolite on the classification and discrimination of each group of samples. The common screening criteria for differential metabolites are that the variable importance weight value is greater than 1. Differential metabolites must also meet the independent sample t test, and a P value less than 0.05 is considered statistically significant. When both criteria are met at the same time, the corresponding metabolites can be selected as differential metabolites. Therefore, this scheme ensures the acquisition of differential metabolite information.

[0032] In a possible implementation manner, step S3 includes the following steps:

[0033] S31: Based on the differential metabolite information obtained in step S23, the disease prediction model is constructed by binary logistic regression, wherein the regression analysis of the binary logistic regression adopts a stepwise screening strategy, and the basis for variables to enter the equation is the score test statistic, and the basis for variables to be eliminated from the equation is the likelihood chi-square ratio under the maximum likelihood estimation principle;

[0034] S32: evaluating the calibration of the disease prediction model by the area under the fitting curve and the AUC value corresponding to the disease prediction model, and evaluating the calibration ability of the disease prediction model by the Hosmer-Lemeshow goodness of fit test to obtain the evaluation result;

[0035] This solution constructs a disease prediction model using binary logistic regression, and tests the model discrimination and model calibration to evaluate the disease prediction model, thereby obtaining evaluation results, providing a basis for model acquisition and evaluation, and ensuring improved model prediction accuracy and wide application.

[0036] Another technical solution of the present invention is to provide a system for constructing and evaluating a prediction model for polycystic ovary syndrome, based on the method for constructing a prediction model for polycystic ovary syndrome in the present invention, comprising:

[0037] The metabolite extraction module uses a number of overweight and obese patients with polycystic ovary syndrome and a number of normal women as a first comparison group, and a number of non-overweight and obese patients with polycystic ovary syndrome and a number of normal women as a second comparison group, and extracts blood metabolites from each comparison group using a blood metabolomics detection method to obtain blood metabolite information corresponding to each comparison group;

[0038] The differential metabolite extraction module performs principal component analysis and partial least squares discriminant analysis on the blood metabolite information corresponding to each comparison group, so as to screen the differential metabolites corresponding to each comparison group through the variable importance weight value and p value;

[0039] A model generation and evaluation module, which uses the differential metabolites corresponding to each comparison group to construct a disease prediction model through binary logistic regression, and tests the model discrimination and model calibration to evaluate the disease prediction model and obtain an evaluation result;

[0040] The differential metabolite extraction module is electrically connected to the metabolite extraction module, and the model generation and evaluation module is electrically connected to the differential metabolite extraction module.

[0041] The system for constructing and evaluating the prediction model of polycystic ovary syndrome in the present invention, by setting a metabolite extraction module, a differential metabolite extraction module and a model generation and evaluation module, the differential metabolite extraction module is electrically connected to the metabolite extraction module, and the model generation and evaluation module is electrically connected to the differential metabolite extraction module, so as to fully consider the clinical characteristics of the polycystic ovary syndrome disease before constructing the prediction model, divide the polycystic ovary syndrome patients into overweight and obese polycystic ovary syndrome and non-overweight and obese polycystic ovary syndrome, and is suitable for polycystic ovary syndrome patients diagnosed according to the new evidence-based guidelines proposed by the International Polycystic Ovary Syndrome Expert Group, and can provide an important reference tool for the auxiliary diagnosis of polycystic ovary syndrome.

[0042] In a possible implementation, the metabolite extraction module includes:

[0043] A classification storage device, taking a number of normal women randomly selected as a control group, storing blood samples of the control group, and storing blood samples of a number of overweight and obese patients with polycystic ovary syndrome and a number of non-overweight and obese patients with polycystic ovary syndrome, wherein the blood samples of the control group and the blood samples of the overweight and obese patients with polycystic ovary syndrome are used as blood samples of a first comparison group, and the blood samples of the control group and the blood samples of the non-overweight and obese patients with polycystic ovary syndrome are used as blood samples of a second comparison group;

[0044] An information collection device is used to detect corresponding blood metabolite information from the blood samples of the first comparison group, and to detect corresponding blood metabolite information from the blood samples of the second comparison group, wherein the blood metabolite information includes lipid metabolite information and small molecule metabolite information;

[0045] The information collection device is electrically connected to the differential metabolite extraction module.

[0046] In a possible embodiment, the differential metabolite extraction module includes:

[0047] A principal component analysis unit performs principal component analysis on the blood metabolite information corresponding to the first comparison group and the blood metabolite information corresponding to the second comparison group obtained by the information acquisition device, respectively, to obtain respective corresponding principal component analysis results;

[0048] A discriminant analysis unit performs partial least squares discriminant analysis on the principal component analysis results corresponding to the first comparison group and the principal component analysis results corresponding to the second comparison group obtained by the principal component analysis unit, respectively, to obtain the variable importance weight value distribution and p value distribution corresponding to each of them;

[0049] a difference information acquisition unit, which, according to the variable importance weight value distribution and p-value distribution corresponding to the first comparison group obtained by the discriminant analysis unit, takes the metabolites that meet the difference criteria as the differential metabolites in the lipid metabolites and the differential metabolites in the small molecule metabolites corresponding to the first comparison group; and, according to the variable importance weight value distribution and p-value distribution corresponding to the second comparison group obtained by the discriminant analysis unit, takes the metabolites that meet the difference criteria as the differential metabolites in the lipid metabolites and the differential metabolites in the small molecule metabolites corresponding to the second comparison group, to obtain differential metabolite information;

[0050] The principal component analysis unit is electrically connected to the information acquisition device, the discriminant analysis unit is electrically connected to the principal component analysis unit, and the difference information acquisition unit is electrically connected to the discriminant analysis unit and the model generation and evaluation module respectively.

[0051] In a possible implementation, the model generation and evaluation module includes:

[0052] A generating unit, based on the differential metabolite information obtained by the differential information obtaining unit, constructs the disease prediction model through binary logistic regression, wherein the regression analysis of the binary logistic regression adopts a stepwise screening strategy, and the basis for variables to enter the equation is the score test statistic, and the basis for variables to be eliminated from the equation is the likelihood chi-square ratio under the maximum likelihood estimation principle;

[0053] An evaluation unit, which evaluates the calibration of the disease prediction model by the area under the fitting curve and the AUC value corresponding to the disease prediction model, and evaluates the calibration ability of the disease prediction model by the Hosmer-Lemeshow goodness of fit test to obtain the evaluation result;

[0054] The generating unit is electrically connected to the difference information acquiring unit, and the evaluating unit is electrically connected to the generating unit. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 This is a flow chart of the method for constructing a prediction model for polycystic ovary syndrome disclosed in Example 1 of the present invention;

[0056] Figure 2 This is a flow chart of step S1 disclosed in Embodiment 1 of the present invention;

[0057] Figure 3 This is a flow chart of step S2 disclosed in Embodiment 1 of the present invention;

[0058] Figure 4 This is a flow chart of step S3 disclosed in Embodiment 1 of the present invention;

[0059] Figure 5 This is a diagram showing the results of the principal component analysis of lipids in the first comparison group disclosed in Example 1 of the present invention;

[0060] Figure 6 This is a graph showing the partial least squares discriminant analysis results of the first comparison group of lipids disclosed in Example 1 of the present invention;

[0061] Figure 7 This is a diagram showing the principal component analysis results of the first comparison group of small molecule metabolites disclosed in Example 1 of the present invention;

[0062] Figure 8This is a diagram showing the partial least squares discriminant analysis results of the first comparison group of small molecule metabolites disclosed in Example 1 of the present invention;

[0063] Fig. 9 This is a diagram showing the results of the principal component analysis of lipids in the second comparison group disclosed in Example 1 of the present invention;

[0064] Fig.10 This is a diagram showing the partial least squares discriminant analysis results of the second comparison group of lipids disclosed in Example 1 of the present invention;

[0065] Fig.11 This is a diagram showing the principal component analysis results of the second comparison group of small molecule metabolites disclosed in Example 1 of the present invention;

[0066] Fig.12 This is a diagram showing the partial least squares discriminant analysis results of the second comparison group of small molecule metabolites disclosed in Example 1 of the present invention;

[0067] Fig.13 A random forest diagram of the overweight, obese, polycystic ovary syndrome patient prediction model in Example 1 of the present invention;

[0068] Fig.14 The ROC curve diagram of the overweight and obese polycystic ovary syndrome patient prediction model in Example 1 of the present invention;

[0069] Fig.15 A random forest diagram of the prediction model for non-overweight and obese polycystic ovary syndrome patients in Example 1 of the present invention;

[0070] Fig.16 The ROC curve diagram of the prediction model for non-overweight and obese polycystic ovary syndrome patients in Example 1 of the present invention;

[0071] Fig.17 This is a schematic diagram of the structure of the polycystic ovary syndrome prediction model and evaluation system disclosed in Example 2 of the present invention. DETAILED DESCRIPTION

[0072] First, those skilled in the art should understand that these implementations are only used to explain the technical principles of the embodiments of the present application, and are not intended to limit the protection scope of the embodiments of the present application. Those skilled in the art can make adjustments to them as needed to adapt to specific application scenarios.

[0073] In the description of the embodiments of the present application, it should be noted that, unless otherwise clearly specified and limited, the terms "electrical connection" and "establishing an electrical connection relationship" should be understood in a broad sense, that is, it should be understood that two or more of them have an electrical relationship, which can be achieved through a wire, a wireless connection, or a combination of the two; it can be directly connected or indirectly connected through an intermediate medium. For ordinary technicians in this field, the specific meanings of the above terms in the embodiments of the present application can be understood according to specific circumstances.

[0074] Two embodiments will be used below, and the present application will be further described in detail in conjunction with the corresponding drawings and specific embodiments in each embodiment.

[0075] Embodiment 1

[0076] See also Figures 1 to 16 As shown, the present application embodiment discloses a method for constructing a prediction model for polycystic ovary syndrome, see Figure 1 As shown, the method comprises the following steps:

[0077] S1: Several overweight and obese patients with PCOS and several normal women were used as the first comparison group, and several non-overweight and obese patients with PCOS and several normal women were used as the second comparison group. Blood metabolites were extracted from each comparison group using blood metabolomics detection to obtain the corresponding blood metabolite information of each comparison group.

[0078] See also Figure 2 As shown, in this embodiment, step S1 includes the following steps:

[0079] S11: According to the overweight standard of body mass index, multiple PCOS patients diagnosed based on the Rotterdam diagnostic criteria were divided into a group of overweight and obese PCOS patients and a group of non-overweight and obese PCOS patients.

[0080] In this step, the overweight criteria for body mass index are expressed as follows:

[0081] Body mass index greater than or equal to 24 kg / m 2 of PCOS patients are overweight and obese PCOS patients;

[0082] Body mass index less than 24 kg / m 2 The patients with PCOS were non-overweight or obese patients with PCOS.

[0083] Correspondingly, the Rotterdam diagnostic criteria are:

[0084] On the premise of excluding other diseases that cause ovulation disorders and clinical or biochemical hyperandrogenism, patients who meet at least two of the following conditions are considered to have PCOS:

[0085] (1) Clinical manifestations of hyperandrogenism or hyperandrogenism;

[0086] (2) Infrequent ovulation or anovulation;

[0087] (3) Polycystic ovarian changes.

[0088] S12: Randomly select a number of normal women as the control group; randomly select a number of overweight and obese PCOS patients from the set of overweight and obese PCOS patients, and use them together with the control group as the first comparison group; randomly select a number of non-overweight and obese PCOS patients from the set of non-overweight and obese PCOS patients, and use them together with the control group as the second comparison group.

[0089] S13: Blood samples were collected from the first and second comparison groups respectively, and targeted nuclear magnetic resonance metabolomics based on nuclear magnetic resonance spectroscopy was used to analyze the blood samples. 1 The H-NMR spectrum detects the blood metabolite information corresponding to each comparison group from the blood samples of each comparison group, and the blood metabolite information includes lipid metabolite information and small molecule metabolite information.

[0090] Small molecule metabolites are molecules with a molecular weight range of 50-1500Da. 1 H-NMR spectra can detect a variety of substances (serial numbers 1 to 114), including lipids and small molecule metabolites (alcohols and derivatives, amino acids and derivatives, choline and other substances, serial numbers 115 to 153). The specific names, abbreviations and classifications of the detected substances are shown in Table 1.

[0091] Table 1: Utilization 1 List of blood metabolites detected by H-NMR spectrum

[0092]

[0093]

[0094]

[0095]

[0096]

[0097]

[0098]

[0099]

[0100] For the convenience of implementation, this example collects 22 blood samples from normal women (control group) as group D; collects 197 blood samples from overweight and obese patients with polycystic ovary syndrome as group B; collects 30 blood samples from non-overweight and obese patients with polycystic ovary syndrome as group C. Furthermore, the first comparison group is group D and group B, and the second comparison group is group D and group C.

[0101] S2: The blood metabolite information corresponding to each comparison group was subjected to principal component analysis, partial least squares discriminant analysis and independent sample t-test respectively, so as to screen the differential metabolites corresponding to each comparison group through variable importance weight value and p-value.

[0102] See also Figure 3 As shown, in this embodiment, step S2 includes the following steps:

[0103] S21: performing principal component analysis on the blood metabolite information corresponding to the first comparison group and the blood metabolite information corresponding to the second comparison group obtained in step S13, respectively, to obtain respective corresponding principal component analysis results.

[0104] S22: Perform partial least squares discriminant analysis and independent sample t-test on the principal component analysis results corresponding to the first comparison group obtained in step S21 and the principal component analysis results corresponding to the second comparison group, respectively, to obtain the corresponding variable importance weight value distribution and p-value distribution.

[0105] S23: According to the variable importance weight value distribution and p-value distribution corresponding to the first comparison group obtained in step S22, the metabolites that meet the difference criteria are taken as the differential metabolites among the lipid metabolites and the differential metabolites among the small molecule metabolites corresponding to the first comparison group; according to the variable importance weight value distribution and p-value distribution corresponding to the second comparison group obtained in step S22, the metabolites that meet the difference criteria are taken as the differential metabolites among the lipid metabolites and the differential metabolites among the small molecule metabolites corresponding to the second comparison group, to obtain differential metabolite information.

[0106] In this embodiment, the difference criterion is: metabolites that simultaneously satisfy the variable importance weight value greater than 1 and the p value less than 0.05 are differential metabolites, otherwise they are non-differential metabolites.

[0107] Since blood metabolite information has the characteristics of "high dimension, high noise, and high variation", multivariate statistical analysis methods are generally used to "simplify and reduce the dimension" of high-dimensional and complex data on the basis of retaining the original information to the greatest extent, and to establish a reliable mathematical model to summarize and conclude the metabolic spectrum characteristics of the research object. This embodiment uses partial least squares discriminant analysis (PLS-DA), which is a multivariate statistical analysis method for supervised pattern recognition. Multidimensional data can be grouped according to the difference factors to be found before compression (pre-setting Y values ​​for target classification and discrimination), so that the variables most relevant to the grouping factors can be found, while reducing the influence of some other factors.

[0108] Partial least squares discriminant analysis is often used to distinguish the overall differences in metabolic profiles between groups and screen for differential metabolites between groups. The specific process is: use SIMCA software to perform logarithmic transformation and UV formatting on the data, first perform partial least squares discriminant analysis on the first principal component in the principal component analysis results, and use cross-validation to test the quality of the discriminant analysis; then use the R obtained after cross-validation 2 X (how well the model explains the categorical variables) and R 2 Y (response modeling robustness) is used to judge the effectiveness of discriminant analysis; finally, the permutation test is performed to randomly change the order of the categorical variable Y 200 times to obtain different random Q 2 The Y (predictiveness of the model) value is used to further test the effectiveness of the discriminant analysis.

[0109] The variable importance weight value (hereinafter referred to as VIP) is the variable weight value of the partial least squares discriminant analysis model variable, which can be used to measure the influence and explanatory power of the difference in the accumulation of each metabolite on the classification and discrimination of each group of samples. VIP ≥ 1 is a common screening standard for differential metabolites. Differential metabolites must also meet the independent sample t test, and P value < 0.05 is considered statistically significant. When both criteria are met at the same time, that is, VIP> 1, P < 0.05, the corresponding metabolite can be selected as a differential metabolite.

[0110] Considering that blood metabolite information includes lipid metabolite information and small molecule metabolite information, in this embodiment, the analysis in step S2 and the modeling process in step S3 both adopt the method of analyzing lipid metabolite (serial number 1 to 114) information and small molecule metabolite (false high 115 to 153) information separately. The principal component analysis results and partial least squares discriminant analysis results in step S2 are shown below.

[0111] Figure 5 This is the result of principal component analysis of lipids in the first comparison group; Figure 6This is the result of the partial least squares discriminant analysis of lipids in the first comparison group. It can be seen from the figure that the lipid metabolism profiles of different groups (group B and group D) are well distinguished, especially the partial least squares discriminant analysis:

[0112] R 2 X=0.641,

[0113] R 2 Y=0.855,

[0114] Q 2 Y = 0.802,

[0115] This indicates that the analysis results have good stability and good predictive ability. When VIP>1 and P<0.05 are satisfied at the same time, the corresponding lipid metabolites can be selected as differential metabolites. Table 2 shows the differential metabolite information in the lipid metabolites selected by partial least squares discriminant analysis.

[0116] Table 2: Differential metabolite information in lipid metabolites of the first comparison group of this example

[0117]

[0118] Figure 7 This is the result of principal component analysis of small molecule metabolites in the first comparison group; Figure 8 This is the result of the partial least squares discriminant analysis of small molecule metabolites in the first comparison group. It can be seen from the figure that the lipid metabolism profiles of different groups (group B and group D) are well distinguished. The partial least squares discriminant analysis is:

[0119] R 2 X=0.847,

[0120] R 2 Y=0.91,

[0121] Q 2 Y = 0.826,

[0122] Table 3 shows the differential metabolite information among the small molecule metabolites selected by partial least squares discriminant analysis.

[0123] Table 3: Differential metabolite information among the small molecule metabolites selected in the first comparison group

[0124]

[0125] Fig. 9 This is the result of principal component analysis of lipids in the second comparison group; Fig.10 The results of the partial least squares discriminant analysis of lipids in the second comparison group are shown in the figure. It can be seen from the figure that the lipid metabolism profiles of different groups (group C and group D) are well distinguished. The partial least squares discriminant analysis is:

[0126] R 2 X=0.783,

[0127] R 2 Y=0.774,

[0128] Q 2 Y = 0.672,

[0129] Table 4 shows the differential metabolite information among the lipid metabolites selected by partial least squares discriminant analysis.

[0130] Table 4: Differential metabolite information in the selected lipid metabolites of the second comparison group

[0131]

[0132]

[0133] Fig.11 This is the result of principal component analysis of small molecule metabolites in the second comparison group; Fig.12 The results of the partial least squares discriminant analysis of small molecule metabolites in the second comparison group are shown in the figure. It can be seen from the figure that the lipid metabolism profiles of different groups (group C and group D) are well distinguished. The partial least squares discriminant analysis is:

[0134] R 2 X=0.729,

[0135] R 2 Y=0.898,

[0136] Q 2 Y=0.807,

[0137] Table 5 shows the differential metabolite information among the small molecule metabolites selected by partial least squares discriminant analysis.

[0138] Table 5: Differential metabolite information among the small molecule metabolites selected in the second comparison group

[0139]

[0140] S3: Use the differential metabolites corresponding to each comparison group to construct a disease prediction model through binary logistic regression, and test the model discrimination and model calibration to evaluate the disease prediction model and obtain the evaluation results.

[0141] See also Figure 4 As shown, in this embodiment, step S3 includes the following steps:

[0142] S31: Based on the differential metabolite information obtained in step S23, a disease prediction model is constructed through binary logistic regression, wherein the regression analysis of the binary logistic regression adopts a stepwise screening strategy, and the basis for variables to enter the equation is the score test statistic, and the basis for variables to be eliminated from the equation is the likelihood chi-square ratio under the maximum likelihood estimation principle.

[0143] S32: The calibration of the disease prediction model is evaluated by the area under the fitting curve and the AUC value corresponding to the disease prediction model, and the calibration ability of the disease prediction model is evaluated by the Hosmer-Lemeshow goodness of fit test to obtain an evaluation result.

[0144] After binary logistic regression, the disease prediction model for overweight and obese patients with PCOS was as follows: Fig.13 As shown in the figure, the OR value (odds ratio) and P value after binary logistic regression are also presented. Fig.13 In the forest diagram shown in Figure 1, a good disease prediction model is not just a simple mathematical combination of dependent variables and independent variables. The actual clinical significance behind it is the key point to grasp. This requires that the prediction model must not only have good discrimination, but also good calibration. The characteristic curve (ROC curve) of the prediction model is shown in Figure 1. Fig.14 As shown in the figure, the area under the curve and AUC value are 0.977>0.75, and the 95% confidence level CI is 0.959-0.995, indicating that the disease prediction model has good discrimination ability. Model calibration (Cal ibration) was evaluated by the Hosmer-Lemeshow goodness of fit test. The results showed that the Hosmer-Lemeshowχ 2 =4.538, P=0.806>0.05, indicating that there is no statistically significant difference between the predicted value of the disease prediction model and the actual observed value, and the prediction model has good calibration ability.

[0145] After binary logistic regression, the disease prediction model for non-overweight and obese PCOS patients was as follows: Fig.15 As shown, the ROC curve of the prediction model is as follows Fig.16 As shown, the area under the curve and AUC value were 0.921>0.75, and the 95% confidence level CI was 0.862-0.979, indicating that the prediction model had good discrimination ability.

[0146] Embodiment 2

[0147] On the basis of Example 1, this example further discloses a system for constructing and evaluating a prediction model for polycystic ovary syndrome. The structural diagram of the system is shown in FIG. Fig.17As shown, it includes a metabolite extraction module, a differential metabolite extraction module and a model generation and evaluation module. The differential metabolite extraction module is electrically connected to the metabolite extraction module, and the model generation and evaluation module is electrically connected to the differential metabolite extraction module.

[0148] In this system, the functions of each module are as follows:

[0149] The metabolite extraction module is configured to use a number of overweight and obese patients with polycystic ovary syndrome and a number of normal women as a first comparison group, and a number of non-overweight and obese patients with polycystic ovary syndrome and a number of normal women as a second comparison group, and extract blood metabolites from each comparison group using a blood metabolomics detection method to obtain blood metabolite information corresponding to each comparison group;

[0150] The differential metabolite extraction module is set to perform principal component analysis and partial least squares discriminant analysis on the blood metabolite information corresponding to each comparison group, so as to screen the differential metabolites corresponding to each comparison group through the variable importance weight value and p value;

[0151] The model generation and evaluation module is configured to construct a disease prediction model by binary logistic regression using the differential metabolites corresponding to each comparison group, and to test the model discrimination and model calibration to evaluate the disease prediction model and obtain evaluation results;

[0152] In this embodiment, the metabolite extraction module includes a classification storage device and an information collection device, and the information collection device is electrically connected to the differential metabolite extraction module. The classification storage device is configured to use a number of randomly selected normal women as a control group, store blood samples of the control group, and store blood samples of a number of overweight and obese polycystic ovary syndrome patients and a number of non-overweight and obese polycystic ovary syndrome patients, wherein the blood samples of the control group and the blood samples of the overweight and obese polycystic ovary syndrome patients are used as blood samples of the first comparison group, and the blood samples of the control group and the blood samples of the non-overweight and obese polycystic ovary syndrome patients are used as blood samples of the second comparison group; the information collection device is configured to detect corresponding blood metabolite information from the blood samples of the first comparison group, and detect corresponding blood metabolite information from the blood samples of the second comparison group, and the blood metabolite information includes lipid metabolite information and small molecule metabolite information.

[0153] In this embodiment, the differential metabolite extraction module includes a principal component analysis unit, a discriminant analysis unit and a differential information acquisition unit. The principal component analysis unit is electrically connected to the information acquisition device, the discriminant analysis unit is electrically connected to the principal component analysis unit, and the differential information acquisition unit is electrically connected to the discriminant analysis unit and the model generation and evaluation module, respectively.

[0154] In the differential metabolite extraction module, the functions of each unit are as follows:

[0155] The principal component analysis unit is configured to perform principal component analysis on the blood metabolite information corresponding to the first comparison group and the blood metabolite information corresponding to the second comparison group obtained by the information acquisition device, respectively, to obtain respective corresponding principal component analysis results;

[0156] The discriminant analysis unit is configured to perform partial least squares discriminant analysis on the principal component analysis results corresponding to the first comparison group and the principal component analysis results corresponding to the second comparison group obtained by the principal component analysis unit, respectively, to obtain the variable importance weight value distribution and p value distribution corresponding to each of them;

[0157] The difference information acquisition unit is configured to, according to the variable importance weight value distribution and p-value distribution corresponding to the first comparison group obtained by the discriminant analysis unit, take the metabolites that meet the difference criteria as the differential metabolites in the lipid metabolites and the differential metabolites in the small molecule metabolites corresponding to the first comparison group; according to the variable importance weight value distribution and p-value distribution corresponding to the second comparison group obtained by the discriminant analysis unit, take the metabolites that meet the difference criteria as the differential metabolites in the lipid metabolites and the differential metabolites in the small molecule metabolites corresponding to the second comparison group, to obtain differential metabolite information.

[0158] In this embodiment, the model generation and evaluation module includes a generation unit and an evaluation unit, the generation unit is electrically connected to the difference information acquisition unit, and the evaluation unit is electrically connected to the generation unit. The generation unit is configured to construct a disease prediction model through binary logistic regression based on the difference metabolite information obtained by the difference information acquisition unit, wherein the regression analysis of the binary logistic regression adopts a stepwise screening strategy, and the basis for the variable to enter the equation is the score test statistic, and the basis for eliminating the variable from the equation is the likelihood chi-square ratio under the maximum likelihood estimation principle; the evaluation unit is configured to evaluate the calibration of the disease prediction model through the area under the fitting curve corresponding to the disease prediction model and the AUC value, and evaluate the calibration ability of the disease prediction model through the Hosmer-Lemeshow goodness of fit test method to obtain an evaluation result.

[0159] The PCOS prediction model construction and evaluation system disclosed in this embodiment, by setting a metabolite extraction module, a differential metabolite extraction module and a model generation and evaluation module, the differential metabolite extraction module is electrically connected to the metabolite extraction module, and the model generation and evaluation module is electrically connected to the differential metabolite extraction module, so as to fully consider the clinical characteristics of the PCOS disease before constructing the prediction model, divide the PCOS patients into overweight and obese PCOS and non-overweight and obese PCOS, and is suitable for PCOS patients diagnosed according to the new evidence-based guidelines proposed by the International PCOS Expert Group, and can provide an important reference tool for the auxiliary diagnosis of PCOS.

[0160] In the description of the embodiments of the present application, it should be noted that in the description of the present application, terms such as "inside" and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the drawings. This is only for the convenience of description, and does not indicate or imply that the device or component must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation on the present application.

[0161] In the description of the present application, the description with reference to the terms "one embodiment", "some embodiments", "in the present embodiment", "specific example", or "some examples" etc. means that the specific features, mechanisms, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, mechanisms, 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 the different embodiments or examples, without contradiction.

[0162] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. A method for constructing a prediction model for polycystic ovary syndrome, characterized in that: The steps include: S1: Several overweight and obese patients with PCOS and several normal women were used as the first comparison group, and several non-overweight and obese patients with PCOS and several normal women were used as the second comparison group. Blood metabolites of each comparison group were extracted by blood metabolomics detection to obtain the blood metabolite information corresponding to each comparison group. S2: The corresponding blood metabolite information of each comparison group was subjected to principal component analysis, partial least squares discriminant analysis and independent sample t-test to screen the differential metabolites corresponding to each comparison group through variable importance weight value and p-value; S3: A disease prediction model is constructed by binary logistic regression using the differential metabolites corresponding to each comparison group, and the model discrimination and model calibration are tested to evaluate the disease prediction model to obtain an evaluation result.

2. The method for constructing a prediction model for polycystic ovary syndrome according to claim 1, characterized in that: The step S1 comprises the following steps: S11: According to the overweight standard of body mass index, multiple PCOS patients diagnosed based on the Rotterdam diagnostic criteria were divided into a group of overweight and obese PCOS patients and a group of non-overweight and obese PCOS patients; S12: Select a number of normal women as the control group; Randomly select a number of overweight, obese, and polycystic ovary syndrome patients from the set of overweight, obese, and polycystic ovary syndrome patients, and use them and the control group as the first comparison group; Randomly select a number of non-overweight, obese, polycystic ovary syndrome patients from the non-overweight, obese, polycystic ovary syndrome patient set, and use them and the control group as the second comparison group; S13: Blood samples from the first comparison group and the second comparison group were collected respectively, and targeted nuclear magnetic resonance metabolomics based on nuclear magnetic resonance spectroscopy was used to analyze the blood samples. 1 The H-NMR spectrum detects the blood metabolite information corresponding to each comparison group from the blood samples of each comparison group, and the blood metabolite information includes lipid metabolite information and small molecule metabolite information.

3. The method for constructing a prediction model for polycystic ovary syndrome according to claim 2, characterized in that: The overweight standard of the body mass index is expressed as follows: Body mass index greater than or equal to 24 kg / m 2 of PCOS patients are overweight and obese PCOS patients; Body mass index less than 24 kg / m 2 The patients with PCOS were non-overweight or obese patients with PCOS.

4. The method for constructing a prediction model for polycystic ovary syndrome according to claim 2 or 3, characterized in that: The step S2 comprises the following steps: S21: performing principal component analysis on the blood metabolite information corresponding to the first comparison group and the blood metabolite information corresponding to the second comparison group obtained in step S13, respectively, to obtain respective corresponding principal component analysis results; S22: performing partial least squares discriminant analysis and independent sample t-test on the principal component analysis results corresponding to the first comparison group and the principal component analysis results corresponding to the second comparison group obtained in step S21, respectively, to obtain respective corresponding variable importance weight values ​​and p-values; S24: According to the variable importance weight value and p-value corresponding to the first comparison group obtained in step S22, the metabolites that meet the difference criteria are taken as the differential metabolites among the lipid metabolites and the differential metabolites among the small molecule metabolites corresponding to the first comparison group; according to the variable importance weight value distribution and p-value distribution corresponding to the second comparison group obtained in step S22, the metabolites that meet the difference criteria are taken as the differential metabolites among the lipid metabolites and the differential metabolites among the small molecule metabolites corresponding to the second comparison group, to obtain differential metabolite information.

5. The method for constructing a prediction model for polycystic ovary syndrome according to claim 4, characterized in that: The difference criteria are: Metabolites that simultaneously satisfy the variable importance weight value greater than 1 and the p value less than 0.05 are considered differential metabolites, otherwise they are considered non-differential metabolites.

6. The method for constructing a prediction model for polycystic ovary syndrome according to claim 5, characterized in that: The step S3 comprises the following steps: S31: Based on the differential metabolite information obtained in step S24, the disease prediction model is constructed by binary logistic regression, wherein the regression analysis of the binary logistic regression adopts a stepwise screening strategy, and the basis for variables to enter the equation is the score test statistic, and the basis for variables to be eliminated from the equation is the likelihood chi-square ratio under the maximum likelihood estimation principle; S32: Evaluate the calibration of the disease prediction model by the area under the fitting curve and the AUC value corresponding to the disease prediction model, and evaluate the calibration ability of the disease prediction model by the Hosmer-Lemeshow goodness of fit test to obtain the evaluation result.

7. A system for constructing and evaluating a prediction model for polycystic ovary syndrome, characterized in that: The method for constructing a polycystic ovary syndrome prediction model based on any one of claims 1 to 6 comprises: The metabolite extraction module uses a number of overweight and obese patients with polycystic ovary syndrome and a number of normal women as a first comparison group, and a number of non-overweight and obese patients with polycystic ovary syndrome and a number of normal women as a second comparison group, and extracts blood metabolites from each comparison group using a blood metabolomics detection method to obtain blood metabolite information corresponding to each comparison group; The differential metabolite extraction module performs principal component analysis, partial least squares discriminant analysis, and independent sample t-test on the blood metabolite information corresponding to each comparison group, so as to screen the differential metabolites corresponding to each comparison group through the variable importance weight value and p-value; A model generation and evaluation module, which uses the differential metabolites corresponding to each comparison group to construct a disease prediction model through binary logistic regression, and tests the model discrimination and model calibration to evaluate the disease prediction model and obtain an evaluation result; The differential metabolite extraction module is electrically connected to the metabolite extraction module, and the model generation and evaluation module is electrically connected to the differential metabolite extraction module.

8. The system for constructing and evaluating a prediction model for polycystic ovary syndrome according to claim 7, characterized in that: The metabolite extraction module comprises: A classification storage device, taking a number of normal women randomly selected as a control group, storing blood samples of the control group, and storing blood samples of a number of overweight and obese patients with polycystic ovary syndrome and a number of non-overweight and obese patients with polycystic ovary syndrome, wherein the blood samples of the control group and the blood samples of the overweight and obese patients with polycystic ovary syndrome are used as blood samples of a first comparison group, and the blood samples of the control group and the blood samples of the non-overweight and obese patients with polycystic ovary syndrome are used as blood samples of a second comparison group; An information collection device is used to detect corresponding blood metabolite information from the blood samples of the first comparison group, and to detect corresponding blood metabolite information from the blood samples of the second comparison group, wherein the blood metabolite information includes lipid metabolite information and small molecule metabolite information; The information collection device is electrically connected to the differential metabolite extraction module.

9. The system for constructing and evaluating a prediction model for polycystic ovary syndrome according to claim 8, characterized in that: The differential metabolite extraction module comprises: A principal component analysis unit performs principal component analysis on the blood metabolite information corresponding to the first comparison group and the blood metabolite information corresponding to the second comparison group obtained by the information acquisition device, respectively, to obtain respective corresponding principal component analysis results; The discriminant analysis unit performs partial least squares discriminant analysis and independent sample t-test on the principal component analysis results corresponding to the first comparison group and the principal component analysis results corresponding to the second comparison group obtained by the principal component analysis unit, respectively, to obtain the corresponding variable importance weight value distribution and p-value distribution; a difference information acquisition unit, which, according to the variable importance weight value distribution and p-value distribution corresponding to the first comparison group obtained by the discriminant analysis unit, takes the metabolites that meet the difference criteria as the differential metabolites in the lipid metabolites and the differential metabolites in the small molecule metabolites corresponding to the first comparison group; and, according to the variable importance weight value distribution and p-value distribution corresponding to the second comparison group obtained by the discriminant analysis unit, takes the metabolites that meet the difference criteria as the differential metabolites in the lipid metabolites and the differential metabolites in the small molecule metabolites corresponding to the second comparison group, to obtain differential metabolite information; The principal component analysis unit is electrically connected to the information acquisition device, the discriminant analysis unit is electrically connected to the principal component analysis unit, and the difference information acquisition unit is electrically connected to the discriminant analysis unit and the model generation and evaluation module respectively.

10. The system for constructing and evaluating a prediction model for polycystic ovary syndrome according to claim 9, characterized in that: The model generation and evaluation module includes: A generating unit, based on the differential metabolite information obtained by the differential information obtaining unit, constructs the disease prediction model through binary logistic regression, wherein the regression analysis of the binary logistic regression adopts a stepwise screening strategy, and the basis for variables to enter the equation is the score test statistic, and the basis for variables to be eliminated from the equation is the likelihood chi-square ratio under the maximum likelihood estimation principle; An evaluation unit, which evaluates the calibration of the disease prediction model by the area under the fitting curve and the AUC value corresponding to the disease prediction model, and evaluates the calibration ability of the disease prediction model by the Hosmer-Lemeshow goodness of fit test to obtain the evaluation result; The generating unit is electrically connected to the difference information acquiring unit, and the evaluating unit is electrically connected to the generating unit.

Images