Cell subset proportion-based gestational age prediction model construction method and gestational age prediction system

Abstract

The invention relates to a cell subset proportion-based fetal age prediction model construction method and a fetal age prediction system. The method comprises the following steps: acquiring a single cell transcriptome data set; filtering low-quality data in the single-cell transcription data set, and carrying out data integration; performing maternal-fetal interface cell subset composition analysis to obtain a maternal-fetal interface cell composition map; calculating the cell subset proportion of each cell subset in different fetal age samples in the maternal-fetal interface cell composition map, and then calculating the regression coefficient of each cell subset proportion and the real fetal age; and constructing a gestational age prediction model based on the proportion of each cell subset and the corresponding regression coefficient, and predicting the real gestational age by using a calculation formula of the gestational age prediction model. According to the method, by obtaining a single cell transcriptome data set, a maternal-fetal interface single cell map is reconstructed, key cell subgroups influencing maternal-fetal interface stability are screened out, and the gestational age in actual biological significance is predicted according to the change of the proportion of the number of the cell subgroups.

Description

Technical Field

[0001] The present invention relates to the technical field of reproductive detection in obstetrics and gynecology, and in particular to a method for constructing a fetal age prediction model based on cell subpopulation ratios and a fetal age prediction system. Background Art

[0002] Pregnancy is a complex physiological process. The fetus will continue to grow and develop with the gestational age during pregnancy. Accurate assessment of the fetal gestational age is of great significance in ensuring the health of pregnant women and fetuses, optimizing pregnancy management, and accurately guiding delivery plans and postpartum care decisions. However, the current methods for determining gestational age are mostly based on the last menstrual date method and ultrasound examination evaluation. However, the existing methods for determining gestational age may lead to inaccurate judgments. For example, those with irregular menstruation cannot calculate the gestational age based on the last menstrual period. Even if the menstrual cycle is very regular, there is a certain difference between the gestational age calculated based on the last menstrual period and the actual gestational age, because there may be abnormal ovulation in the month of conception. The accuracy of ultrasound examination to determine the gestational age is affected by the quality of ultrasound images and the operation skills of ultrasound physicians. In the middle and late stages of pregnancy, fetal growth has greater variability and complexity. The reliability of ultrasound examination to determine the gestational age will gradually decrease with the increase of gestational age. And the existing methods for determining gestational age cannot determine the specific cause of abnormal gestational age. Summary of the invention

[0003] The technical problem to be solved by the present invention is to provide a method for constructing a fetal age prediction model based on cell subpopulation ratios to overcome the deficiencies in the above-mentioned prior art.

[0004] The technical solution of the present invention to solve the above technical problem is as follows: A method for constructing a fetal age prediction model based on cell subpopulation ratios comprises the following steps:

[0005] Step S01: obtaining a single-cell transcriptome dataset, which includes an internal dataset and an external dataset;

[0006] Step S02: filtering out low-quality data in the single-cell transcriptional dataset, and integrating the data of the filtered single-cell transcriptional dataset;

[0007] Step S03: Based on the data in the single-cell transcriptional dataset after data integration, perform a composition analysis of the subpopulations of cells at the maternal-fetal interface to obtain a composition map of cells at the maternal-fetal interface;

[0008] Step S04: Calculate the proportion of each cell subpopulation in the maternal-fetal interface cell composition map in samples of different gestational ages, and then calculate the regression coefficient between the proportion of each cell subpopulation and the actual gestational age;

[0009] Step S05: Based on the proportion of each cell subset and its corresponding regression coefficient, a gestational age prediction model is constructed, and the actual gestational age is predicted using the calculation formula of the gestational age prediction model.

[0010] The beneficial effects of the present invention are as follows: this method reconstructs the single-cell map of the maternal-fetal interface by acquiring a single-cell transcriptome data set, screens out key cell subpopulations that affect the stability of the maternal-fetal interface, and predicts the gestational age in the actual biological sense based on the changes in the proportion of the number of cell subpopulations. The current methods for determining gestational age are mostly based on the last menstrual date method and ultrasound examination evaluation, both of which may lead to inaccurate judgments. This method is based on the analysis of the actual cell composition of fetal placental tissue, so it is more accurate. If the predicted gestational age is abnormal, the specific cell composition that causes the abnormal gestational age can be traced back, thereby finding the specific cause of the abnormal gestational age.

[0011] Based on the above technical solution, the present invention can also be improved as follows.

[0012] Furthermore, the method for obtaining the internal data set in step S01 includes:

[0013] Step S11: collecting decidua and placental tissue samples from volunteers at 4 to 39 weeks of pregnancy; wherein, samples from early and mid-pregnancy are from women who undergo elective termination of pregnancy; and samples from late pregnancy are from women after delivery;

[0014] Step S12: The collected decidua and placenta tissues are lysed into single cells and then single-cell capture and single-cell sequencing are performed to obtain an internal data set.

[0015] Further, step S02 specifically includes the following steps:

[0016] Step S021: Perform low-quality data filtering analysis on each single-cell transcription data set to filter out low-quality data;

[0017] Step S022: Use the IntegrateData function in Seurat to integrate the single-cell transcriptional dataset after filtering out low-quality data;

[0018] Step S023: After completing the data integration, duplicate cells are filtered out using the R software package DoubletFinder.

[0019] Furthermore, the filtering conditions for filtering out low-quality data in step S021 include:

[0020] Low-quality data with the number of expressed genes less than 200 and the proportion of mitochondrial genes higher than 30% were filtered out.

[0021] Further, the analysis of the subpopulation composition of maternal-fetal interface cells in step S03 includes the following steps:

[0022] Step S31: using the R software package Seurat to perform LogNormalization on the data in the single-cell transcriptome dataset after data integration;

[0023] Step S32: using the FindVariableFeatures function to select genes with large expression variation in the standardized single-cell transcriptome data set, using the ScaleData function to normalize the data of the genes with large expression variation, performing principal component analysis PCA on the normalized genes with large expression variation, and screening out the genes with large expression variation contained in the principal components with significant significance;

[0024] Step S33: Perform unsupervised clustering on the selected genes with large expression variation, cluster them into different subgroups, and visualize the clustering results through umap / tSNE dimensionality reduction to obtain the maternal-fetal interface cell composition map. Further, the maternal-fetal interface cell composition map in step S33 includes the following 38 subgroups:

[0025] 22 immune cell subsets: CD4 + T, CD8 + T, Treg, Th1, Tfh, DC1, DC2, dM1, dM2, dM3, MO, dNK1, dNK2, dNK3, dNKp, CD16 - NK、CD16 + NK, NKT, HB, Plasma, ILC3 and Granulocytes;

[0026] 7 stromal cell subsets: dS1, dS2, dS3, fFB1, fFB2, dP1, and dP2;

[0027] 3 trophoblast subsets: EVT, VCT, and SCT;

[0028] 3 endothelial cell subsets: Endo(m), Endo(f), and Endo L;

[0029] 2 epithelial cell subsets: Epi1 and Epi2;

[0030] 1 subset of blood cells: Erythrocytes.

[0031] Further, the calculation steps of the regression coefficient in step S04 are as follows:

[0032] Based on the R software package stats, a multivariate linear regression model was constructed, and the regression coefficient between the proportion of each cell subset and the actual gestational age was calculated using the multivariate linear regression model.

[0033] Further, step S05 specifically includes: optimizing and screening variables by stepwise regression method and full subset regression method, taking the proportions of the top 8 cell subsets with the highest correlation and their corresponding regression coefficients to construct a fetal age prediction model,

[0034] The calculation formula of the gestational age prediction model is:

[0035]

[0036] Among them, gestational age is the gestational age, Prop (*) represents the proportion of cells of this subpopulation in the maternal-fetal interface tissue, β 0 To β 8 is the regression coefficient obtained from the training of the multivariate linear regression model. The top 8 cell subsets with the highest correlation include: EVT, fFB2, Endo(f), CD16 + NK, dP1, dNK1, dM2, and DC1.

[0037] The present invention also discloses a fetal age prediction system based on cell subpopulation ratio, comprising a data acquisition module, a data calculation module and a data output module;

[0038] Data acquisition module, which is used to collect EVT, fFB2, Endo(f), CD16 + The proportion of NK, dP1, dNK1, dM2, and DC1 cells in maternal-fetal interface tissues;

[0039] A data calculation module, which is used to substitute the cell ratio into the calculation formula of the fetal age prediction model constructed in the above-mentioned method for constructing a fetal age prediction model based on the cell subpopulation ratio, and calculate the gestational period of the pregnant woman;

[0040] The data output module is used to output the pregnancy period of the pregnant woman.

[0041] The beneficial effect of the present invention is that the current methods for determining gestational age are mostly based on the last menstrual period date method and ultrasound examination evaluation, both of which may lead to inaccurate judgments. The present method is based on the analysis of the true cell composition of fetal placental tissue and is therefore more accurate. For example, if the predicted gestational age is abnormal, the specific cell composition that causes the abnormal gestational age can be traced back, thereby finding the specific cause of the abnormal gestational age. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 is a flow chart of the method of the present invention;

[0043] Figure 2 is a model parameter diagram of the present invention;

[0044] Figure 3 It is the prediction scatter plot of the present invention. DETAILED DESCRIPTION

[0045] The principles and features of the present invention are described below in conjunction with the accompanying drawings. The examples given are only used to explain the present invention and are not used to limit the scope of the present invention.

[0046] like Figure 1 to Figure 3 As shown, Example 1, a method for constructing a fetal age prediction model based on cell subpopulation ratio, step S01: obtaining a single-cell transcriptome dataset, the single-cell transcriptome dataset including an internal dataset and an external dataset; data for constructing a single-cell atlas of the maternal-fetal interface come from the internal dataset and the external dataset, respectively, the dataset includes single-cell transcriptome data of decidua and placenta tissues, and the external dataset includes the existing datasets GSE130560, GSE198373, and GSE174481;

[0047] Step S02: filtering out low-quality data in the single-cell transcriptional dataset, and integrating the data of the filtered single-cell transcriptional dataset;

[0048] Step S03: Based on the data in the single-cell transcriptional dataset after data integration, perform a composition analysis of the subpopulations of cells at the maternal-fetal interface to obtain a composition map of cells at the maternal-fetal interface;

[0049] Step S04: Calculate the proportion of each cell subpopulation in the maternal-fetal interface cell composition map in samples of different gestational ages, and then calculate the regression coefficient between the proportion of each cell subpopulation and the actual gestational age;

[0050] Step S05: Based on the proportion of each cell subset and its corresponding regression coefficient, a gestational age prediction model is constructed, and the actual gestational age is predicted using the calculation formula of the gestational age prediction model.

[0051] This method reconstructs the single-cell map of the maternal-fetal interface by acquiring a single-cell transcriptome dataset, screens out key cell subpopulations that affect the stability of the maternal-fetal interface, and predicts the gestational age in the actual biological sense based on changes in the proportion of the number of cell subpopulations. Current methods for determining gestational age are mostly based on the last menstrual date method and ultrasound examination evaluation, both of which may lead to inaccurate judgments. This method is based on the analysis of the actual cell composition of fetal placental tissue, so it is more accurate. If the predicted gestational age is abnormal, the specific cell composition that causes the abnormal gestational age can be traced back to find the specific cause of the abnormal gestational age.

[0052] Embodiment 2: This embodiment is a further improvement on the basis of embodiment 1, and its details are as follows:

[0053] The method for obtaining the internal data set in step S01 includes:

[0054] Step S11: collecting decidua and placental tissue samples from volunteers at 4 to 39 weeks of pregnancy; wherein, samples from early and mid-pregnancy are from women who undergo elective termination of pregnancy; and samples from late pregnancy are from women after delivery;

[0055] Step S12: The collected decidua and placenta tissues are lysed into single cells and then single-cell capture and single-cell sequencing are performed to obtain an internal data set.

[0056] Embodiment 3, this embodiment is a further improvement on the basis of embodiment 1, and its details are as follows:

[0057] Step S02 specifically includes the following steps:

[0058] Step S021: Perform low-quality data filtering analysis on each single-cell transcription data set to filter out low-quality data;

[0059] Step S022: Use the IntegrateData function in Seurat to integrate the single-cell transcriptional dataset after filtering out low-quality data;

[0060] Step S023: After completing the data integration, duplicate cells are filtered out using the R software package DoubletFinder.

[0061] Embodiment 4: This embodiment is a further improvement on the basis of embodiment 3, and its details are as follows:

[0062] The filtering conditions for filtering out low-quality data in step S021 include:

[0063] Low-quality data with the number of expressed genes (nFeature) less than 200 and the proportion of mitochondrial genes (percent.mt) higher than 30% were filtered out.

[0064] Embodiment 5, this embodiment is a further improvement on the basis of embodiment 1, and its details are as follows:

[0065] The analysis of the subpopulation composition of maternal-fetal interface cells in step S03 includes the following steps:

[0066] Step S31: using the R software package Seurat to perform LogNormalization on the data in the single-cell transcriptome dataset after data integration;

[0067] Step S32: using the FindVariableFeatures function to select genes with large expression variation in the standardized single-cell transcriptome data set, using the ScaleData function to normalize the data of the genes with large expression variation, performing principal component analysis PCA on the normalized genes with large expression variation, and screening out the genes with large expression variation contained in the principal components with significant significance;

[0068] Step S33: Perform unsupervised clustering on the selected genes with large expression variation, cluster them into different subgroups, and visualize the clustering results through umap / tSNE dimensionality reduction to obtain a cell composition map of the maternal-fetal interface.

[0069] Embodiment 6: This embodiment is a further improvement on the basis of embodiment 5, and its details are as follows:

[0070] The maternal-fetal interface cell composition map in step S33 includes the following 38 subpopulations:

[0071] 22 immune cell subsets: CD4 + T (naive T cells), CD8 + T (naive T cells), Treg (regulatory T cells), Th1 (helper T cells), Tfh (helper T cells), DC1 (dendritic cells), DC2 (dendritic cells), dM1 (decidual macrophages), dM2 (decidual macrophages), dM3 (decidual macrophages), MO (mononuclear macrophages), dNK1 (natural killer cells), dNK2 (natural killer cells), dNK3 (natural killer cells), dNKp (natural killer cells), CD16 - NK (natural killer cells), CD16 + NK (natural killer cells), NKT (natural killer cells), HB (placental resident macrophages Hofbauer cells), Plasma (plasma cells), ILC3 (innate lymphocytes) and Granulocytes (granulocytes);

[0072] 7 stromal cell subsets: dS1 (decidual stromal cells), dS2 (decidual stromal cells), dS3 (decidual stromal cells), fFB1 (fetal fibroblasts), fFB2 (fetal fibroblasts), dP1 (pericytes), and dP2 (pericytes);

[0073] 3 trophoblast subsets: EVT (extravillous trophoblast), VCT (villous trophoblast), and SCT (syncytial trophoblast);

[0074] Three endothelial cell subsets: Endo(m) (maternal endothelial cells), Endo(f) (fetal endothelial cells) and Endo L (L-type endothelial cells);

[0075] 2 epithelial cell subsets: Epi1 (epithelial cells) and Epi2 (epithelial cells);

[0076] 1 blood cell subset: Erythrocytes (blood cells).

[0077] Embodiment 7, this embodiment is a further improvement on the basis of embodiment 1, and its details are as follows:

[0078] The calculation steps of the regression coefficient in step S04 are as follows:

[0079] Based on the R software package stats, a multivariate linear regression model was constructed, and the regression coefficient between the proportion of each cell subset and the actual gestational age was calculated using the multivariate linear regression model.

[0080] Embodiment 8, this embodiment is a further improvement on the basis of embodiment 7, and its details are as follows:

[0081] Step S05 specifically includes: optimizing and screening variables by using stepwise regression method and full subset regression method, taking the proportions of the top 8 cell subsets with the highest correlation and their corresponding regression coefficients to construct a fetal age prediction model;

[0082] The calculation formula of the gestational age prediction model is:

[0083]

[0084] Among them, gestational age is the gestational period (i.e. gestational age), Prop (*) represents the proportion of cells of this subpopulation in the maternal-fetal interface tissue, β 0 To β 8 is the regression coefficient obtained from the training of the multivariate linear regression model. The top 8 cell subsets in terms of correlation include: EVT (extravillous trophoblast), fFB2 (fetal fibroblast), Endo(f) (fetal vascular endothelial cells), CD16 + NK (natural killer cells), dP1 (pericyte), dNK1 (natural killer cells), dM2 (decidual macrophages), and DC1 (dendritic cells).

[0085] This technical solution uses the R software package caret to perform a 6-fold cross validation. The results show that changes in the cell composition of the human maternal-fetal interface are strongly correlated with fetal development and can better predict gestational age. See Figure 2 and Figure 3, the model R2 reached 0.7669. After 6-fold cross validation, the MAE was 6.1161, and the model performed well.

[0086] Figure 2 It is a visualization of the parameters of the multivariate linear regression model, showing the regression coefficients and their significance of 8 cell subsets as independent variables. The position of each independent variable point represents the estimated value of the regression coefficient of the variable, and the length of the horizontal line represents the 95% confidence interval of the regression coefficient. If the horizontal line passes through 0, it means that the effect of the independent variable on the dependent variable may not be significant. If the horizontal line does not pass through 0, it means that the variable has a significant effect on the dependent variable (greater than 0 is a positive effect, less than 0 is a negative effect). The lower right corner of the figure shows the AlC (Akaike Information Criterion) and BlC (Bayesian Information Criterion) of the model, which are used to evaluate the goodness of fit and complexity of the model, respectively.

[0087] Figure 3 This is a scatter plot of the predicted pregnancy period relative to the actual pregnancy weeks obtained by using the multivariate linear regression model analysis. The horizontal axis value corresponding to each point is the actual pregnancy period, and the vertical axis value is the predicted pregnancy period. The fitting trend line R between the predicted value and the actual value 2 The value is 0.7669, and the P value is 3.625*10 -10 . Figure 3 The relevant parameters are shown in Table 1 below:

[0088]

[0089] Table 1

[0090] Embodiment 9, a fetal age prediction system based on cell subpopulation ratio, comprising a data acquisition module, a data calculation module and a data output module;

[0091] Data acquisition module, which is used to collect EVT, fFB2, Endo(f), CD16 + The proportion of NK, dP1, dNK1, dM2, and DC1 cells in maternal-fetal interface tissues;

[0092] A data calculation module, which is used to substitute the cell ratio into the calculation formula of the fetal age prediction model constructed in the method for constructing a fetal age prediction model based on the cell subpopulation ratio in Example 8, and calculate the gestational period of the pregnant woman;

[0093] The data output module is used to output the pregnancy period of the pregnant woman.

[0094] The current methods for determining gestational age are mostly based on the last menstrual period date method and ultrasound examination evaluation, both of which may lead to inaccurate judgments. This method is based on the analysis of the true cell composition of fetal placental tissue and is therefore more accurate. If the predicted gestational age is abnormal, the specific cell composition that causes the abnormal gestational age can be traced back to find the specific cause of the abnormal gestational age.

[0095] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A method for constructing a fetal age prediction model based on cell subpopulation ratios, characterized in that: The steps include: Step S01: obtaining a single-cell transcriptome dataset, wherein the single-cell transcriptome dataset includes an internal dataset and an external dataset; Step S02: filtering out low-quality data in the single-cell transcriptome dataset, and performing data integration on the filtered data of the single-cell transcriptome dataset; Step S03: Based on the data in the single-cell transcriptome dataset after data integration, perform a composition analysis of maternal-fetal interface cell subpopulations to obtain a maternal-fetal interface cell composition map; Step S04: calculating the cell subpopulation ratio of each cell subpopulation in the maternal-fetal interface cell composition atlas in samples of different gestational ages, and then calculating the regression coefficient between each cell subpopulation ratio and the actual gestational age; Step S05: Based on the proportion of each cell subset and its corresponding regression coefficient, a gestational age prediction model is constructed, and the actual gestational age is predicted using the calculation formula of the gestational age prediction model.

2. The method for constructing a fetal age prediction model based on cell subpopulation ratio according to claim 1, characterized in that: The method for obtaining the internal data set in step S01 includes: Step S11: collecting decidua and placental tissue samples from volunteers at 4 to 39 weeks of pregnancy; wherein, samples from early and mid-pregnancy are from women who undergo elective termination of pregnancy; and samples from late pregnancy are from women after delivery; Step S12: The collected decidua and placenta tissues are lysed into single cells and then single cell capture and single cell sequencing are performed to obtain the internal data set.

3. The method for constructing a fetal age prediction model based on cell subpopulation ratio according to claim 1, characterized in that: The step S02 specifically includes the following steps: Step S021: performing low-quality data filtering analysis on each of the single-cell transcription data sets to filter out low-quality data; Step S022: using the IntegrateData function in Seurat to integrate the single-cell transcriptional dataset after filtering out low-quality data; Step S023: After completing the data integration, duplicate cells are filtered out using the R software package DoubletFinder.

4. The method for constructing a fetal age prediction model based on cell subpopulation ratio according to claim 3, characterized in that: The filtering conditions for filtering out low-quality data in step S021 include: Low-quality data with the number of expressed genes less than 200 and the proportion of mitochondrial genes higher than 30% were filtered out.

5. The method for constructing a fetal age prediction model based on cell subpopulation ratio according to claim 1, characterized in that: The analysis of the subpopulation composition of maternal-fetal interface cells in step S03 comprises the following steps: Step S31: using the R software package Seurat to perform LogNormalization on the data in the single-cell transcriptome dataset after data integration; Step S32: using the FindVariableFeatures function to select genes with large expression variation in the standardized single-cell transcriptome data set, using the ScaleData function to normalize the data of the genes with large expression variation, performing principal component analysis PCA on the normalized genes with large expression variation, and screening out the genes with large expression variation contained in the principal components with significant significance; Step S33: Perform unsupervised clustering on the selected genes with large expression variation, cluster them into different subgroups, and visualize the clustering results through umap / tSNE dimensionality reduction to obtain a cell composition map of the maternal-fetal interface.

6. The method for constructing a fetal age prediction model based on cell subpopulation ratio according to claim 5, characterized in that: The maternal-fetal interface cell composition map in step S33 includes the following 38 subgroups: 22 immune cell subsets: CD4 + T, CD8 + T, Treg, Th1, Tfh, DC1, DC2, dM1, dM2, dM3, MO, dNK1, dNK2, dNK3, dNKp, CD16 - NK、CD16 + NK, NKT, HB, Plasma, ILC3 and Granulocytes; 7 stromal cell subsets: dS1, dS2, dS3, fFB1, fFB2, dP1, and dP2; 3 trophoblast subsets: EVT, VCT, and SCT; 3 endothelial cell subsets: Endo(m), Endo(f), and Endo L; 2 epithelial cell subsets: Epi1 and Epi2; 1 subset of blood cells: Erythrocytes.

7. The method for constructing a fetal age prediction model based on cell subpopulation ratio according to claim 1, characterized in that: The calculation steps of the regression coefficient in step S04 are as follows: Based on the R software package stats, a multiple linear regression model was constructed, and the regression coefficient between the proportion of each cell subpopulation and the actual gestational age was calculated using the multiple linear regression model.

8. The method for constructing a fetal age prediction model based on cell subpopulation ratios according to claim 7, characterized in that: The step S05 specifically includes: optimizing and screening variables by using a stepwise regression method and a full subset regression method, taking the proportions of the top 8 cell subsets with the highest correlation and their corresponding regression coefficients to construct the fetal age prediction model, The calculation formula of the gestational age prediction model is: Among them, gestational age is the gestational period, Prop (*) represents the proportion of cells in the maternal-fetal interface tissue of this subpopulation, β0 to β8 are the regression coefficients obtained by training the multivariate linear regression model, and the top 8 cell subpopulations in terms of correlation include: EVT, fFB2, Endo(f), CD16 + NK, dP1, dNK1, dM2, and DC1.

9. A fetal age prediction system based on cell subpopulation ratio, characterized in that: It includes a data acquisition module, a data calculation module and a data output module; The data acquisition module is used to collect data on EVT, fFB2, Endo(f), CD16 + The proportion of NK, dP1, dNK1, dM2, and DC1 cells in maternal-fetal interface tissues; The data calculation module is used to substitute the cell ratio into the calculation formula of the fetal age prediction model constructed in the method for constructing a fetal age prediction model based on cell subpopulation ratios in claim 8 to calculate and obtain the gestational period of the pregnant woman; The data output module is used to output the pregnancy period of the pregnant woman.

Images