Female individual biological age speculation method and system based on blood sample

Abstract

The invention provides a female individual biological age speculation method and system based on a blood sample, and relates to the technical field of biology, and the method comprises the following steps: obtaining methylation chip data; an Elastic Net regression model is established; inputting the data of each methylated chip into an Elastic Net regression model, outputting a first predicted age, and extracting a first CpG site set related to the first predicted age; respectively carrying out intersection screening on the first CpG site set and the plurality of site sets to obtain a second CpG site set; respectively inputting the second CpG site set into a plurality of models to respectively obtain a plurality of predicted ages; performing accuracy evaluation on the plurality of models through each prediction age, and determining an optimal prediction model; and outputting the predicted age of the optimal prediction model as a speculation result of the biological age of the individual female. The method can effectively improve the efficiency and accuracy of female individual biological age speculation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of biotechnology, in particular to a female individual biological age estimation method and system based on blood samples. BACKGROUND

[0002] The female individual biological age estimation method based on blood samples refers to analyzing the methylation beta value of specific age-related CpG sites, mining the regularity characteristics between the methylation beta value and the actual age, and establishing a biological age prediction model using various machine learning algorithms, aiming to scientifically quantify the health status, aging speed and physiological age of female individuals.

[0003] With the acceleration of global population aging, how to accurately assess the true physiological age of individuals and then assist health management and disease warning has become an important direction of aging and precision medicine research. A female individual biological age estimation method based on blood samples, specific, few CpG sites, controllable cost, and high prediction accuracy can effectively improve the scientificity and individualization level of female health management, and provide important technical support and application value for life health big data research and precision intervention strategies under the background of an aging society.

[0004] However, the existing age estimation method relies on a large number of CpG site numbers, resulting in high detection cost, high experimental complexity, which is not conducive to clinical and large-scale population application. Most models do not consider gender differences, ignore the unique physiological and epigenetic characteristics of female individuals in the aging process, affect the pertinence and accuracy of prediction, and the existing models generally only use a single Elastic Net regression algorithm for model construction, limiting the optimal performance exploration and improvement of the model. SUMMARY

[0005] In view of the above deficiencies of the prior art, the purpose of the embodiments of the present application is to provide a female individual biological age estimation method based on blood samples, which can solve the technical problems of the prior art that the existing age estimation method relies on a large number of CpG site numbers, resulting in high detection cost, high experimental complexity, which is not conducive to clinical and large-scale population application. Most models do not consider gender differences, ignore the unique physiological and epigenetic characteristics of female individuals in the aging process, affect the pertinence and accuracy of prediction, and the existing models generally only use a single Elastic Net regression algorithm for model construction, limiting the optimal performance exploration and improvement of the model.

[0006] The first aspect of the embodiments of the present application provides a female individual biological age estimation method based on blood samples, comprising:

[0007] S1: obtain methylation chip data, wherein the methylation chip data comprises a plurality of age-related female CpG sites;

[0008] S2: establish an Elastic Net regression model;

[0009] S3: input the methylation chip data into the Elastic Net regression model, output a first predicted age, and extract a first CpG site set comprising 137 female CpG sites related to the first predicted age;

[0010] S4: perform intersection filtering on the first CpG site set with a Horvath site set, a Hannum site set, and a PhenoAge Clock CpG site set respectively, to obtain a second CpG site set comprising 18 female CpG sites, wherein the 18 female CpG sites are respectively: cg23606718, cg07553761, cg16867657, cg16419235, cg04875128, cg14692377, cg20822990, cg06639320, cg03607117, cg08097417, cg14361627, cg21120249, cg07850604, cg10917602, cg04084157, cg01511567, cg22736354, and cg04474832;

[0011] S5: input the second CpG site set into the Elastic Net regression model, an SVM machine learning algorithm, a GBR regression model, and an ANN regression model respectively, to output a second predicted age, a third predicted age, a fourth predicted age, and a fifth predicted age;

[0012] S6: perform accuracy evaluation on the Elastic Net regression model, the SVM machine learning algorithm, the GBR regression model, and the ANN regression model by using the second predicted age, the third predicted age, the fourth predicted age, and the fifth predicted age, to determine an optimal prediction model;

[0013] S7: output the predicted age of the optimal prediction model as a result of inferring the biological age of the female individual.

[0014] In a second aspect of the embodiment of the present application, a female individual biological age inference system based on a blood sample is provided, comprising a processor and a memory.

[0015] The memory stores programs or instructions that can run on the processor, and the programs or instructions are executed by the processor to implement the steps of the female individual biological age estimation method based on a blood sample according to the first aspect.

[0016] According to a third aspect of the embodiments of the present application, a readable storage medium is provided, and the readable storage medium stores programs or instructions, and the programs or instructions are executed by a processor to implement the steps of the female individual biological age estimation method based on a blood sample according to the first aspect.

[0017] The technical scheme provided by the embodiments of the present application has at least the following beneficial effects:

[0018] In the embodiments of the present application, the Elastic Net is used to automatically screen the relevant CpG sites, and the intersection screening is performed with the Horvath, Hannum and PhenoAge Clock site sets, so that the number of CpG sites required by the model is significantly reduced, the detection cost and complexity are reduced, and the Elastic Net, support vector machine (SVM), gradient boosting regression (GBR) and artificial neural network (ANN) and other machine learning algorithms are innovatively used for the construction of the biological age prediction model, the feature learning ability and prediction advantage of different algorithms are fully mined, each model is fully evaluated, and the best prediction model selected finally has the optimal prediction accuracy and stability, so that the efficiency and accuracy of the female individual biological age estimation can be effectively improved. BRIEF DESCRIPTION OF DRAWINGS

[0019] The accompanying drawings are only for the purpose of illustrating specific embodiments and are not considered as limiting the present application. Throughout the drawings, the same reference signs indicate the same components. Obviously, the accompanying drawings described below are only some embodiments described in the embodiments of the present application, and other drawings can be obtained according to these drawings without creative labor for those skilled in the art.

[0020] Figure 1 is a flowchart of a female individual biological age estimation method based on a blood sample provided by the embodiments of the present application;

[0021] Figure 2 is a scatter plot of actual age and predicted age in the female 137-CpG site Elastic net model training set provided by the embodiments of the present application;

[0022] Figure 3 is a Wayne diagram of the female 137-CpG site intersection with Horvath, Hannum and PhenoAge Clock provided by the embodiments of the present application;

[0023] Figure 4 is a scatter plot of actual age and predicted age in the female 18-CpG site elastic net model training set provided by the embodiment of the present application;

[0024] Figure 5 is a scatter plot of actual age and predicted age in the female 18-CpG site SVM model training set provided by the embodiment of the present application;

[0025] Figure 6 is a scatter plot of actual age and predicted age in the female 18-CpG site GBR model training set provided by the embodiment of the present application;

[0026] Figure 7 is a scatter plot of actual age and predicted age in the female 18-CpG site ANN model training set provided by the embodiment of the present application;

[0027] Figure 8 is a structural diagram of a female individual biological age estimation system based on a blood sample provided by the embodiment of the present application. DETAILED DESCRIPTION

[0028] In order for those skilled in the art to better understand the technical solutions in the embodiments of the present application, the technical solutions of the present application will be described clearly and completely in conjunction with the drawings. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. It should be understood that these descriptions are only exemplary and are not intended to limit the scope of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative labor should be within the scope of protection of the present application.

[0029] The female individual biological age estimation method based on a blood sample provided by the embodiment of the present application will be described in detail below in conjunction with the drawings, through specific embodiments and their application scenarios.

[0030] Referring to the drawings attached to the specification, Figure 1 , a flowchart of a female individual biological age estimation method based on a blood sample provided by the embodiment of the present application is shown.

[0031] The female individual biological age estimation method based on a blood sample provided by the embodiment of the present application can include the following steps:

[0032] S1: Obtain methylation chip data, wherein the methylation chip data includes a plurality of female CpG sites related to age.

[0033] The methylation chip data refers to data obtained by using a methylation chip detection technology to perform high-throughput detection on the methylation level (beta value) of a CpG site in a human DNA sample. The CpG site refers to a region in a DNA sequence in which cytosine (C) and guanine (G) are connected by a phosphodiester bond. DNA methylation mainly occurs at the CpG site. The female CpG site refers to a CpG site that is closely related to the age change of a female individual and has a significant methylation change rule in a female population.

[0034] It should be noted that by obtaining the age-related female CpG site methylation chip data on a large scale, the dynamic change characteristics of DNA methylation in the aging process of a female individual can be comprehensively reflected, the comprehensiveness, scientificity and representativeness of the data source for subsequent model construction are ensured, and the reliability and applicability of the prediction results of the model are improved.

[0035] In one possible implementation, S1 specifically includes:

[0036] Obtaining the methylation chip data related to the age from a public database.

[0037] The public database includes an EWAS database and a GEO database.

[0038] The EWAS database is a professional database of global epigenome-wide association study (EWAS), mainly including DNA methylation data and the results of association studies between the data and environmental factors, age, diseases and other phenotypes under different populations, different diseases and different physiological states, and providing rich data resources for the study of methylation and health or aging. The GEO database is a public high-throughput genomic data storage platform established and maintained by the National Center for Biotechnology Information (NCBI) of the United States, widely including various types of genomic data such as gene expression, methylation and chip sequencing, covering sample data of different tissues and different populations, and being an important data source for biological research and model construction.

[0039] After S1, the following is further included:

[0040] The methylation chip data is pre-processed, including sample screening, methylation site filtering, interpolation and correction.

[0041] The methylation site filtering refers to screening out CpG sites related to the research target and having high reliability, and removing irrelevant, poor-quality or excessive missing value CpG sites.

[0042] It should be noted that by sample screening, methylation site filtering, interpolation and correction on the methylation chip data, the data quality is improved, the sample screening removes irrelevant or poor quality samples, the methylation site filtering removes irrelevant CpG sites, the interpolation solves the data missing problem, and the correction eliminates the probe type difference, optimizes the consistency and accuracy of the data, thereby providing high-quality input data for the subsequent biological age prediction model, and enhancing the reliability and accuracy of the model.

[0043] The interpolation specifically includes:

[0044] The methylation chip data is filled using the KNN algorithm.

[0045] The correction specifically includes:

[0046] The Type II probe distribution of the methylation data is corrected to the Type I probe distribution difference by BMIQ.

[0047] Specifically, the KNN interpolation algorithm fills in the missing values by using similar samples, reducing the information loss caused by missing data. The BMIQ correction adjusts the distribution difference between different types of probes, eliminates the data bias caused by probe type difference, and ensures the standardization and consistency of the data.

[0048] For example, the sample screening specifically includes:

[0049] Sample screening:

[0050] (1) Preliminary screening: Extract samples with tissue type as whole blood (n = 1801), and filter out samples with more than 5% of CpG sites with beta value of NA for each sample.

[0051] (2) Further screening conditions: Remove Parkinson's patients, HIV+ patients, and those with a family history of Parkinson's disease. Remove pain_facilitation, pain_inhibition and pmi evaluators. Remove samples with CD8+ cell proportion almost 0 or less than one thousandth. Remove patients who have received treatment. Remove fetal and twin samples. Remove samples less than 10 years old. After screening, the number of samples is 1451.

[0052] The methylation site filtering specifically includes:

[0053] The initial number of sites is 485,512, and the number of filtered sites is 333, with the following filtering conditions:

[0054] (1) Filter CpG sites with more than 10 NA.

[0055] (2) Filter bad probes, cross-react probes, and variation sites.

[0056] (3) Filter sites with a spearman coefficient < 0.6 for age and CpG sites.

[0057] S2: Establish an Elastic Net regression model.

[0058] The Elastic Net regression model is a regression model that combines Lasso regression (L1 regularization) and Ridge regression (L2 regularization), which can both perform feature selection (by L1 regularization to compress the coefficients of unimportant features to zero) and handle the problem of multicollinearity between features (by L2 regularization).

[0059] It should be noted that the establishment of the Elastic Net regression model can effectively process high-dimensional CpG site data, avoid overfitting problems caused by excessively high data dimensions, automatically filter out key features related to biological age, and maintain the stability and prediction accuracy of the model in the presence of multiple collinearity between features.

[0060] For example, 1451 samples were obtained through sample screening, including 678 female samples. According to the methylation site filtering conditions of Example 1, 328 CpG sites were selected for female individual biological age predictor modeling, and the specific operation steps were as follows:

[0061] A linear data set was generated using the make_regression function, and the data set was divided into a training set and a test set.

[0062] An ElasticNet object was created, and the fit method was called to fit the training data set. The test set was used for prediction.

[0063] The mean squared error (MSE) of the prediction results was calculated.

[0064] A scatter plot of the actual data and the model fitting line was drawn to visualize the performance of the model.

[0065] Referring to the accompanying drawings Figure 2 , a scatter plot of actual age and predicted age in the training set of the female 137-CpG site Elastic net model provided by the embodiment of the present application is shown.

[0066] As Figure 2As shown, 678 Samples indicates that there are 678 samples in the dataset. 137 CpG Markers indicates that 137 CpG sites are used for biological age prediction. r = 0.97 indicates that the Pearson correlation coefficient is 0.97, indicating a high correlation between predicted age and actual age. MAD = 2.71 indicates that the median absolute deviation (MAD) is 2.71, reflecting the error range between predicted age and actual age. The horizontal axis represents the actual age of the samples, and the vertical axis represents the biological age predicted by the Elastic Net model. The data points in the scatter plot show a distribution close to the diagonal line, indicating that the predicted age is highly consistent with the actual age, and the model shows good fitting degree. The Pearson correlation coefficient (r) is 0.97, indicating a strong linear relationship between the two. The MAD value is 2.71, indicating that the model has small prediction error, high accuracy and reliability. The red diagonal line represents perfect prediction (predicted age equals actual age), and the closer the data points are to this line, the more accurate the model's prediction.

[0067] After S2, further comprising:

[0068] Training the Elastic Net regression model by grid search.

[0069] S3: inputting the methylation chip data into the Elastic Net regression model, outputting the first predicted age, and extracting the first CpG site set related to the first predicted age, including 137 female CpG sites.

[0070] In a possible implementation, the 137 female CpG sites specifically include: cg06784991, cg17879376, cg23115387, cg24466241, cg05864140, cg16832267, cg21117668, cg23606718, cg00664406, cg07553761, cg12934382, cg02662828, cg23995914, cg12325536, cg26886462, cg13782301, cg14176752, cg16867657, cg00094518, cg00292135, cg14871932, cg12873476, cg15427886, cg16419235, cg21159778, cg10804656, cg12610471, cg13327545, cg23016129, cg06943835, cg08160331, cg10151367, cg15114474, cg18987410, cg19947104, cg09692396, cg05404236, cg16519487, cg07489048, cg04875128, cg01337391, cg04503319, cg14692377, cg20286236, cg25309759, cg26161329, cg20192747, cg03593550, cg05213896, cg23368787, cg14345012, cg01557798, cg03132729, cg05721773, cg07178825, cg10397932, cg11194994, cg15393490, cg16686396, cg18933331, cg19261426, cg20822990, cg22059812, cg22820188, cg27232482, cg03545227, cg06639320, cg14674720, cg27213509, cg01844642, cg03607117, cg04474832, cg12623930, cg22927302, cg10242160, cg15957394, cg17343374, cg25148589, cg00059225, cg08316825, cg19505546, cg21874213, cg21572722, cg22736354, cg26729380, cg04084157, cg08097417, cg13278478,cg14361627, cg18064714, cg21406967, cg24853724, cg09388605, cg00755588, cg01594949, cg03262554, cg12052661, cg13649056, cg13861180, cg21120249, cg00533891, cg06291867, cg14209784, cg15903032, cg00757070, cg01511567, cg02318971, cg03638795, cg04959790, cg12189835, cg13319938, cg17592231, cg04604946, cg22358580, cg04792813, cg07850604, cg27050153, cg01763090, cg16717122, cg25584930, cg00259097, cg00448560, cg00451635, cg02610723, cg10917602, cg11220950, cg01502244, cg04209460, cg06247837, cg09164913, cg13683374, cg16477091, cg27401724, cg01352175, cg14074174, cg05691152, and cg19855470.

[0071] 137 female CpG sites corresponding to the regression coefficients specifically include: 2.291495, 0.882302, -1.492299, 0.432044, -0.425845, 1.030312, 0.51782, 14.456299, 0.932463, 8.789924, 0.080978, -1.247315, 6.086003, -5.273948, -1.18283, 1.95882, 1.773583, 26.371531, 4.782331, 7.940453, 1.299914, -5.136421, -1.155972, 1.20862, 8.820084, 10.00791, -1.119056, 4.62902, -0.103743, -1.136442, 1.88895, 9.419272, -0.138146, 0.031852, -1.284144, -3.267388, 4.639806, -3.149571, -1.455748, 17.774459, -2.626975, -2.359211, 2.006525, -2.851064, -0.06897, 4.205647, -2.822424, -1.177477, 0.067755, -0.843595, -2.518978, -0.684228, -0.015192, -0.989606, 9.126192, -1.12541, -3.814128, -4.25528, -6.568031, -6.214926, -1.81056, -4.392583, -0.362821, -3.245837, 7.113492, 0.05882, 8.263629, 4.749182, -3.249873, 0.602561, 7.595085, -2.666076, -3.612885, -1.456128, 1.45474, 5.407755, -1.423017, 1.063415, 4.983844, 0.399332, 2.083004, -0.356364, 4.235597, 7.771332, -3.686111, 7.40769, 8.728696, -5.65028, 20.292396, 1.553976, -8.229331, 0.687853, -3.177324, -1.986662, -2.731578, 0.429622, -0.031781, 0.476553, -3.548825, -1.272669, -3.428887, 6.091485, -7.576764, -0.751705, -0.579197, -6.322823, -2.517449, -1.831233, -6.763124, 0.023965, 1.476804, 3.161699, -6.039722, 7.668402, -0.453372, 7.214178, -2.139882, 4.774778, 4.333632, 1.930233, 0.352275, -2.182114, -2.184846, -5.228092, -6.414498, 7.062557, -1.485686, -0.543866, -7.115084, -1.454716, -1.17938, 0.63142, -4.717871, -4.548705, -5.900412, 3.045458, 2.097387.

[0072] In actual operation, 10x-fold training is performed on 678 samples and 328 methylation site data by using the Elastic net regression model, the optimal model parameters alpha=0.01 and L1_ration=0.9 obtained by grid search, 137 CpG sites are automatically screened out for calculating DNAmAge, the mean square error (MSE) of the optimal model is 29.66, the intercept is 61.38, the Pearson correlation coefficient r between the predicted age and the actual age is 0.97, the median of the absolute value of the difference between the predicted age and the actual age is MAD=2.71, and the scatter diagram of the actual age and the predicted age in the training set of the 137-CpG site Elastic net model is shown in FIG. 1. Figure 2

[0073] In a possible implementation, the calculation formula of the first predicted age is specifically:

[0074]

[0075] wherein DNAmAge represents the first predicted age, b0 represents an intercept term, bi represents a regression coefficient of the i th CpG site, β CpGi represents a methylation β value of the i th site, i=1, 2, …, n, and n represents the total number of CpG sites.

[0076] FIG. 2 shows a scatter diagram of the actual age and the predicted age of the female 137-CpG site according to the present application. Figure 3 FIG. 3 shows a Wayne diagram of the female 137-CpG site intersected with Horvath, Hannum, and PhenoAge Clock according to the present application.

[0077] ​S4: Intersect the first CpG site set with the Horvath site set, the Hannum site set, and the PhenoAgeClock CpG site set respectively to obtain a second CpG site set including 18 female CpG sites, wherein the 18 female CpG sites are respectively: cg23606718, cg07553761, cg16867657, cg16419235, cg04875128, cg14692377, cg20822990, cg06639320, cg03607117, cg08097417, cg14361627, cg21120249, cg07850604, cg10917602, cg04084157, cg01511567, cg22736354, and cg04474832.

[0078] In a possible implementation, the prediction age calculation formula based on the 18 female CpG sites is specifically:

[0079] DNAmAge2 = 24.368339 + 23.87795β cg23606718 + 13.018577β cg07553761

[0080] + 43.418649β cg16867657 + 5.323545β cg16419235 + 25.920519β cg04875128

[0081] + 5.456631β cg14692377 - 24.671208β cg20822990 + 19.352971β cg06639320

[0082] + 12.056049β cg03607117 + 16.060441β cg08097417 + 36.607789β cg14361627

[0083] - 14.83355β cg21120249 + 14.210864β cg07850604 - 17.753059β cg10917602

[0084] + 16.100622β cg04084157 - 15.738963β cg01511567 + 13.98999β cg22736354

[0085] - 14.712983β cg04474832

[0086] wherein DNAmAge2 represents the predicted age based on 18 female CpG sites, β cg23606718 , β cg07553761 , β cg16867657 , β cg16419235 , β cg04875128 , β cg14692377 , β cg20822990 , β cg06639320 , β cg03607117 , β cg08097417 , β cg14361627 , β cg21120249 , β cg07850604 , β cg10917602 , β cg04084157 , β cg01511567 , β cg22736354 , β cg04474832 each represents the methylation rate of the corresponding CpG site.

[0087] Referring to the accompanying drawings and Figure 4 , a scatter plot of actual age and predicted age in the training set of the female 18-CpG site Elastic net model provided by the embodiments of the present application is shown.

[0088] As shown in Figure 4 , the Venn diagram shows the intersection of different biological age prediction models (Horvath, Hannum, PhenoAge) and the female 137-CpG site set used by the present application. The number of sites unique to each model is shown, as well as the intersection between them.

[0089] Referring to the accompanying drawings and Figure 5 , a scatter plot of actual age and predicted age in the training set of the female 18-CpG site SVM model provided by the embodiments of the present application is shown.

[0090] As shown in Figure 5 , the horizontal axis represents the actual age of the sample, and the vertical axis represents the biological age predicted by the SVM model. The red scatter points in the figure are closely distributed near the diagonal line, indicating that there is a strong linear relationship between the prediction results of the model and the actual age. The Pearson correlation coefficient (r) is 0.98, indicating that there is a very high correlation between the predicted and actual ages. The MAD value is 2.22, indicating that the error of the model is relatively small, and the prediction accuracy is high. The MSE is 21.67 and the RMSE is 4.66, further indicating the reliability of the model in prediction. The determination coefficient R 2 is 0.96, indicating that the model can explain 96% of the data variation, with a high degree of fitting.

[0091] Referring to the accompanying drawings and Figure 6, which shows a scatter plot of actual age and predicted age in the female 18-CpG site GBR model training set provided by an embodiment of the present invention.

[0092] like Figure 6 As shown, the horizontal axis represents the actual age of the sample, and the vertical axis represents the biological age predicted by the GBR model. The green scattered points in the figure are closely distributed close to the diagonal line, indicating that there is a very strong linear correlation between the predicted age and the actual age. The Pearson correlation coefficient (r) is 0.99, showing extremely high prediction accuracy. The MAD value is 1.74, indicating that the prediction error of the model is very small and has good stability. The MSE is 7.95 and the RMSE is 2.82. These low error values ​​further demonstrate the high accuracy of the model. The coefficient of determination R 2 The value is 0.99, which means that the model can explain 99% of the data variation and has a very high degree of fit, indicating its reliability and effectiveness in prediction.

[0093] Reference Manual Figure 7 , shows a scatter plot of actual age and predicted age in the female 18-CpG site ANN model training set provided by an embodiment of the present invention.

[0094] like Figure 7 As shown, the horizontal axis represents the actual age of the sample, and the vertical axis represents the biological age predicted by the ANN model. The purple scatter points in the figure are close to the diagonal line, indicating that there is a strong linear correlation between the model's prediction results and the actual age. The Pearson correlation coefficient (r) is 0.98, showing a high correlation. The MAD value is 2.69, indicating that the prediction error is small and the accuracy is high. The MSE is 20.36 and the RMSE is 4.51, which indicates that the prediction error of the model is moderate. The coefficient of determination R 2 It is 0.96, indicating that the ANN model can explain 96% of the data changes, has a high degree of fit, and exhibits good predictive ability.

[0095] S5: Input the second CpG site set into the Elastic Net regression model, the SVM machine learning algorithm, the GBR regression model, and the ANN regression model respectively, and output the second predicted age, the third predicted age, the fourth predicted age, and the fifth predicted age.

[0096] Among them, the SVM machine learning algorithm (Support Vector Machine) is a supervised learning algorithm commonly used for classification and regression tasks, aiming to find a hyperplane that maximizes the class interval, suitable for high-dimensional data and complex nonlinear problems. The GBR regression model (Gradient Boosting Regression) is an ensemble learning method that trains multiple weak learners (usually decision trees) sequentially to gradually reduce prediction errors, suitable for handling nonlinear relationships. The ANN regression model (Artificial Neural Network) is a machine learning model that simulates the neural network of the human brain, learning complex patterns in data through a multi-layer neural network structure, suitable for large-scale and complex regression tasks.

[0097] It should be noted that the selected second CpG site set is input into the Elastic Net regression model, SVM machine learning algorithm, GBR regression model and ANN regression model for prediction, which can utilize the advantages of different algorithms to obtain multiple prediction results. By comparing the output results of different models, the accuracy of the prediction can be evaluated from multiple angles, and the optimal model can be selected to improve the stability and accuracy of the prediction, avoid the limitations of a single algorithm, and make the final biological age prediction more robust and reliable.

[0098] Based on the 137 sites automatically screened, 18 sites were obtained by intersecting with Horvath, Hannum, and PhenoAge Clock. The 137-CpG site and the intersection of Horvath, Hannum, and PhenoAge Clock are shown in the Wayne diagram as shown in the specification Figure 3 Elastic net 10x-fold model training was performed on the data of 18 methylation sites of 678 samples. The best model parameters obtained by grid search were alpha = 0.01 and L1_ration = 0.9. Automatic screening of 18 CpG sites can be used to calculate DNAmAge, and the mean square error (MSE) of the best model was 31.81, the intercept was 24.37, the Pearson correlation coefficient r between the predicted age and the actual age was 0.97, the median of the absolute value of the difference between the predicted age and the actual age was MAD = 3.31, and the scatter plot of the actual age and the predicted age in the 18-CpG site Elastic net model training set is shown in Figure 4 .

[0099] The 18 CpG sites selected by the SVM machine learning algorithm were used for regression modeling for female individual biological age predictor modeling, and the specific operation steps were as follows:

[0100] Load the dataset, split the data into training and test sets with test_size=0.3: the test set is 30% of the total dataset. random_state=42: set the seed of the random number generator to ensure that the data split is the same each time the code is run, which helps with reproducibility of experimental results.

[0101] Standardize the data, StandardScaler(): create a standardization object. fit_transform(): calculate the mean and standard deviation on the training set and apply standardization. transform(): use the mean and standard deviation calculated on the training set to standardize the test set.

[0102] Create an SVM classifier, SVC(kernel='linear'): create a support vector machine classifier using a linear kernel.

[0103] Train the model and make predictions on the test set, Fit(): train the SVM classifier on the training set. Predict(): make predictions using the trained model.

[0104] Evaluate the model and output the prediction results, including accuracy, recall, F1 score, and other indicators. Classification_report(): print a detailed classification report of the prediction results.

[0105] Visualize the results and plot the fitting graph.

[0106] The mean square error (MSE) of the 18-CpG site best model based on the SVM algorithm is 21.67, the Pearson correlation coefficient r between the predicted age and the actual age is 0.98, the median of the absolute value of the difference between the predicted age and the actual age is MAD=2.22, and the scatter plot of the actual age and the predicted age in the training set of the 18-CpG site SVM model is shown in Figure 5 .

[0107] The 18-CpG sites selected are used to build a regression model for female individual biological age predictor using GBR machine learning algorithm, and the specific operation steps are as follows:

[0108] Load the data and perform data preprocessing, split the dataset into 90% for training and the rest for testing.

[0109] Set the regression model parameters, the regression model parameters are as follows:

[0110] n_estimators: the number of boosting times to be performed.

[0111] max_depth: limit the number of nodes in the tree, the optimal value depends on the interaction between input variables.

[0112] min_samples_split: the minimum number of samples required to split an internal node.

[0113] learning_rate: how much the contribution of each tree will decrease.

[0114] loss: loss function optimization, using the least squares function.

[0115] Initialize a gradient boosting regressor and fit it to the training data.

[0116] Calculate the test set bias.

[0117] Visualize the results based on the boosting iterations, drawing a fitted plot.

[0118] The mean squared error (MSE) of the 18-CpG site optimal model based on the GBR algorithm was 7.95, the Pearson correlation coefficient r between predicted age and actual age was 0.99, the median of the absolute value of the difference between predicted age and actual age was MAD = 1.74, and the scatter plot of actual age and predicted age in the training set of the 18-CpG site GBR model is shown in Figure 6 .

[0119] The 18 selected CpG sites were used to build a regression model for female individual biological age predictor using ANN machine learning algorithm, and the specific operation steps were as follows:

[0120] (1) Data preprocessing: use the get_dummies() method of the Pandas library to decompose the variables into binary fields that can be used, and use the train_test_split method of Scikit-learn to divide the dataset into training set and test set.

[0121] (2) Build neural network model: use the Sequential model of Keras to build the neural network model, which contains two fully connected hidden layers and uses the ReLU layer with activation function.

[0122] (3) Compile and train the model: before training the model, you need to compile the model, specify the loss function (loss function) and optimizer (optimizer) and evaluation metrics (metrics), use the fit() method to train the model, and save it to the history object for subsequent analysis.

[0123] (4) Evaluate the model using the test dataset and save the results to the y_pred variable. Observe the loss of the test set and the validation set in the history object, and calculate the average error.

[0124] (5) Visualize the results and plot the fitting graph.

[0125] The mean square error (MSE) of the 18-CpG site optimal model constructed based on the ANN algorithm is 20.36, the Pearson correlation coefficient r between the predicted age and the actual age is 0.98, the median MAD of the absolute value of the difference between the predicted age and the actual age is 2.69, and the scatter plot of the actual age and the predicted age in the training set of the 18-CpG site ANN model is as shown in Figure 7

[0126] S6: Perform accuracy evaluation on the Elastic Net regression model, SVM machine learning algorithm, GBR regression model and ANN regression model by the second predicted age, third predicted age, fourth predicted age and fifth predicted age, and determine the optimal prediction model.

[0127] It should be noted that by comparing the results of the second predicted age, the third predicted age, the fourth predicted age and the fifth predicted age, the prediction performance of the Elastic Net regression model, the SVM machine learning algorithm, the GBR regression model and the ANN regression model can be comprehensively evaluated. By calculating various evaluation indexes, the prediction accuracy of each model is quantitatively analyzed to ensure that the best model is selected. The optimal prediction model can be effectively selected, the accuracy and stability of the biological age estimation can be maximized, and errors caused by a single model can be avoided.

[0128] The accuracy evaluation is specifically:

[0129] The performance of the Elastic Net regression model, the SVM machine learning algorithm, the GBR regression model and the ANN regression model is compared to evaluate the accuracy.

[0130] The performance indicators specifically include: Pearson correlation coefficient, median absolute error, determination coefficient, mean square error and root mean square error.

[0131] In the present application, the final optimal prediction model also needs to be verified.

[0132] ​Specifically, the Pearson correlation coefficient r between the predicted age and the actual age is 0.95, and the median absolute deviation (MAD) of the difference between the predicted age and the actual age is 3.61 in the Chinese female methylation chip dataset based on the 18-CpG site model of the Elastic net algorithm. The Pearson correlation coefficient r between the predicted age and the actual age is 0.96, and the median absolute deviation (MAD) of the difference between the predicted age and the actual age is 4.38 in the Chinese female methylation chip dataset based on the 18-CpG site model of the SVM algorithm. The Pearson correlation coefficient r between the predicted age and the actual age is 0.96, and the median absolute deviation (MAD) of the difference between the predicted age and the actual age is 3.72 in the Chinese female methylation chip dataset based on the 18-CpG site model of the GBR algorithm. The Pearson correlation coefficient r between the predicted age and the actual age is 0.95, and the median absolute deviation (MAD) of the difference between the predicted age and the actual age is 8.09 in the Chinese female methylation chip dataset based on the 18-CpG site model of the ANN algorithm.

[0133] After verification, it is found that the performance indicators of the female 18-CpG site model under the four algorithms all decrease compared with the performance in the training set, that is, the prediction accuracy decreases, and the prediction accuracy based on the ANN algorithm decreases significantly (R2=0.11), which cannot effectively predict the biological age, suggesting that there may be overfitting in the training set, and it is not recommended to be used for biological age prediction. Although the prediction accuracy based on the Elastic net, SVM and GBR algorithms decreases compared with the training set, it can still effectively predict the biological age, and the MAD values of the prediction models based on the Elastic net and GBR algorithms are both less than 4 and the R2 values are both higher than 0.8, so they can be used as the preferred choice for biological age prediction.

[0134] S7: output the predicted age of the best prediction model as the estimated result of the biological age of the female individual.

[0135] The technical scheme provided by the embodiments of the present application has at least the following beneficial effects:

[0136] In the embodiment of the present application, by adopting Elastic Net to automatically screen relevant CpG sites, and performing intersection screening with the Horvath, Hannum and PhenoAge Clock site sets, the number of CpG sites required by the model is significantly reduced, and the detection cost and complexity are reduced. The Elastic Net, support vector machine (SVM), gradient boosting regression (GBR) and artificial neural network (ANN) and other machine learning algorithms are innovatively used in the construction of the biological age prediction model, the feature learning ability and prediction advantage of different algorithms are fully mined, each model is comprehensively evaluated, and the finally selected best prediction model has optimal prediction accuracy and stability, which can effectively improve the efficiency and accuracy of the biological age prediction of female individuals.

[0137] In summary, the present application provides a method for predicting the biological age of female individuals based on methylation chip data, which realizes high-precision age prediction by constructing and optimizing multiple machine learning models. First, methylation data containing multiple age-related female CpG sites are obtained from public databases such as EWAS and GEO. After sample screening, missing value imputation (KNN algorithm) and BMIQ correction, the data are input into the Elastic Net regression model for training. Then, the CpG sites significantly related to the predicted age are screened from the preliminary modeling, and the intersection with the site sets in the Horvath, Hannum and PhenoAge Clock models is performed to form an optimized CpG site set. The site set is input into the Elastic Net, SVM, GBR and ANN models for retraining and prediction, and multiple sets of predicted ages are output. Based on performance indicators such as Pearson correlation coefficient, MAD, R 2 , MSE and RMSE, the model accuracy is evaluated, and the best prediction model is finally selected as the prediction tool for the biological age of female individuals.

[0138] Reference is made to the accompanying drawings Figure 8 , which shows a structural schematic diagram of a female individual biological age prediction system based on blood samples provided by the embodiment of the present application.

[0139] The embodiment of the present application provides a female individual biological age prediction system 20 based on blood samples, which comprises a processor 201 and a memory 202.

[0140] The memory 202 stores programs or instructions executable on the processor 201, and the programs or instructions are executed by the processor 201 to realize the steps of the female individual biological age prediction method based on blood samples described above, and achieve the same technical effects. To avoid repetition, the present application will not be described again.

[0141] It is to be understood that the processor 201 in the embodiments of the present application can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic components, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor.

[0142] It is also to be understood that the memory 202 in the embodiments of the present application can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memory. Among them, the non-volatile memory can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically EPROM (EEPROM) or a flash memory. The volatile memory can be a random access memory (RAM) used as an external cache. By way of example, but not limitation, many forms of random access memory (RAM) are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced SDRAM (ESDRAM), synchlink DRAM (SLDRAM) and direct rambus RAM (DR RAM).

[0143] The above-described embodiments can be implemented in part or in whole through software, hardware (e.g., circuitry), firmware, or any combination thereof. When implemented in software, the above-described embodiments can be implemented in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When loaded and executed by a computer, the computer instructions or computer programs cause the computer to perform the processes or functions described above according to the embodiments of the present application. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable apparatus. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium, such as from a website, a computer, a server, or a data center to another website, computer, server, or data center through a wired (e.g., infrared, wireless, microwave, or the like) manner. The computer-readable storage medium can be any available medium or a collection of medium accessible by a computer or a data storage device such as a server, data center, or the like, containing one or more of the available medium. The available medium can be a magnetic medium (e.g., a floppy diskette, a hard disk, a magnetic tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0144] It should be understood that the size of the serial number of each process described above does not mean the order of execution, and the execution order of each process should be determined by its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0145] Those skilled in the art can realize that the units and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be realized in electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. A person skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0146] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described devices, apparatuses, and units can refer to the corresponding processes in the foregoing method embodiments, which will not be described here.

[0147] In several embodiments provided by the present application, it should be understood that the disclosed devices, apparatuses and methods can be implemented in other manners. For example, the embodiments of the apparatus described above are merely schematic. For example, the division of the units is only a logical function division. There can be another division manner for the actual implementation. For example, a plurality of units or components can be combined or integrated into another device, or some features can be ignored or not executed. In addition, the displayed or discussed mutual couplings or direct couplings or communication connections between different units, can be indirect couplings or communication connections through some interfaces, devices or units, and can be in electrical, mechanical or other forms.

[0148] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, i.e., can be located in one place, or can be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiments.

[0149] In addition, each functional unit in the various embodiments of the present application can be integrated into a processing unit, or each unit can exist physically, or two or more units can be integrated into one unit.

[0150] If the functions are realized in the form of software function units and sold or used as independent products, they can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application or the part of the technical solutions that essentially contribute to the prior art can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0151] The embodiments of the present application provide a readable storage medium, which includes: a program or instructions stored on the readable storage medium, the program or instructions are executed by a processor to implement the steps of the female individual biological age estimation method based on a blood sample described above, and can achieve the same technical effects. To avoid repetition, the present application will not be described again.

[0152] It should be pointed out finally that the above embodiments are only used to illustrate the technical solutions of the embodiments of the present application, but not to limit them. Although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those skilled in the art that the technical solutions recorded in the foregoing embodiments can be modified, or some technical features can be replaced equivalently; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application. Any changes or replacements that can be easily thought of by those skilled in the art within the technical scope disclosed by the present application should be covered within the protection scope of the present application.

Claims

1. A method for estimating the biological age of a female individual based on a blood sample, characterized in that: include: S1: Acquire methylation chip data, wherein the methylation chip data includes a plurality of age-related female CpG sites; S2: Build an Elastic Net regression model; S3: inputting the methylation chip data into the Elastic Net regression model, outputting a first predicted age, and extracting a first CpG site set including 137 female CpG sites associated with the first predicted age; S4: The first CpG site set is respectively combined with the Horvath site set, the Hannum site set and the PhenoAgeClock Intersection screening of the CpG site sets was performed to obtain a second CpG site set including 18 female CpG sites, wherein the 18 female CpG sites were: cg23606718, cg07553761, cg16867657, cg16419235, cg04875128, cg14692377, cg20822990, cg06639320, cg03607117, cg08097417, cg14361627, cg21120249, cg07850604, cg10917602, cg04084157, cg01511567, cg22736354, and cg04474832; S5: inputting the second CpG site set into the Elastic Net regression model, the SVM machine learning algorithm, the GBR regression model, and the ANN regression model respectively, and outputting a second predicted age, a third predicted age, a fourth predicted age, and a fifth predicted age; S6: Evaluate the accuracy of the Elastic Net regression model, the SVM machine learning algorithm, the GBR regression model, and the ANN regression model based on the second predicted age, the third predicted age, the fourth predicted age, and the fifth predicted age to determine the best prediction model; S7: Outputting the predicted age of the optimal prediction model as the estimated result of the biological age of the female individual.

2. The method for estimating the biological age of a female individual based on a blood sample according to claim 1, characterized in that: After S1, the following steps are also included: The methylation chip data is preprocessed including sample screening, methylation site filtering, interpolation and correction.

3. The method for estimating the biological age of a female individual based on a blood sample according to claim 2, characterized in that: The interpolation is specifically as follows: The methylation chip data was filled using the KNN algorithm; The correction is specifically: The Type II probe distribution of the methylation data was corrected for Type I probe distribution differences using BMIQ.

4. The method for estimating the biological age of a female individual based on a blood sample according to claim 1, wherein: The 137 female CpG sites specifically include: cg06784991, cg17879376, cg23115387, cg24466241, cg05864140, cg16832267, cg21117668, cg23606718, cg00664406, cg07553761, cg12934382, cg02662828, cg23995914, cg12325536, cg26886462, cg13782301, cg14176752, cg16867657, cg00094518, cg00292135, cg14871932, cg12873476, cg15427886, cg16419235, cg21159778, cg10804656, cg12610 471, cg13327545, cg23016129, cg06943835, cg08160331, cg10151367, cg1 5114474, cg18987410, cg19947104, cg09692396, cg05404236, cg16519487 , cg07489048, cg04875128, cg01337391, cg04503319, cg14692377, cg20286 236, cg25309759, cg26161329, cg20192747, cg03593550, cg05213896, cg2 3368787, cg14345012, cg01557798, cg03132729, cg05721773, cg07178825 , cg10397932, cg11194994, cg15393490, cg16686396, cg18933331, cg1926 1426, cg20822990, cg22059812, cg22820188, cg27232482, cg03545227, cg0 6639320, cg14674720, cg27213509, cg01844642, cg03607117, cg04474832 , cg12623930, cg22927302, cg10242160, cg15957394, cg17343374, cg25148 589, cg00059225, cg08316825, cg19505546, cg21874213, cg21572722, cg2 2736354, cg26729380, cg04084157, cg08097417, cg13278478, cg14361627,cg18064714, cg21406967, cg24853724, cg09388605, cg00755588, cg01594949, cg03262554, cg12052661, cg13649056, cg13861180, cg21120249, cg00533891, cg06291867, cg14209784, cg15903032, cg00757070, cg01511567, cg02318971, cg03638795, cg04959790, cg12189835, cg13319938, cg17592231, cg04604946, c g22358580, cg04792813, cg07850604, cg27050153, cg01763090, cg16717122, c g25584930, cg00259097, cg00448560, cg00451635, cg02610723, cg10917602, c g11220950, cg01502244, cg04209460, cg06247837, cg09164913, cg13683374, cg16477091, cg27401724, cg01352175, cg14074174, cg05691152, and cg19855470; The regression coefficients corresponding to the 137 female CpG sites specifically include: 2.291495, 0.882302, -1.492299, 0.432044, -0.425845, 1.030312, 0.51782, 14.456299, 0.932463, 8.789924, 0.080978, -1.247315, 6.086003, -5.273948, -1.18283, 1.95882, 1.773583, 26.371531, 4.782331, 7.940453, 1.299914, -5.136421, -1.155972, 1.20862, and 8.8 20084, 10.00791, -1.119056, 4.62902, -0.103743, -1.136442, 1.88895, 9.419272, -0.138146, 0.031852, -1.284144, -3.267388, 4.639806, -3.149 571, -1.455748, 17.774459, -2.626975, -2.359211, 2.006525, -2.851064, -0.06897, 4.205647, -2.822424, -1.177477, 0.067755, -0.843595, -2.51 8978, -0.684228, -0.015192, -0.989606, 9.126192, -1.12541, -3.814128, -4.25528, -6.568031, -6.214926, -1.81056, -4.392583, -0.362821, -3. 245837, 7.113492, 0.05882, 8.263629, 4.749182, -3.249873, 0.602561, 7.595085, -2.666076, -3.612885, -1.456128, 1.45474, 5.407755, -1.42301 7, 1.063415, 4.983844, 0.399332, 2.083004, -0.356364, 4.235597, 7.771332, -3.686111, 7.40769, 8.728696, -5.65028, 20.292396, 1.553976, -8.2 29331, 0.687853, -3.177324, -1.986662, -2.731578, 0.429622, -0.031781, 0.476553, -3.548825, -1.272669, -3.428887, 6.091485, -7.576764, -0.751705、-0.579197、-6.322823、-2.517449、-1.831233、-6.763124、0.023965、1.476804、3.161699、-6.039722、7.668402、-0.453372、7.214178、-2.139882、4.774778、4.333632、1.930233、0.352275、-2.182114、-2.184846、-5.228092、-6.414498、7.062557、-1.485686、-0.543866、-7.115084、-1.454716、-1.17938、0.63142、-4.717871、-4.548705、-5.900412、3.045458、2.097387。.

5. The method for estimating the biological age of a female individual based on a blood sample according to claim 1, wherein: The calculation formula for the first predicted age is specifically: Where, DNAmAge represents the first predicted age, b0 represents the intercept term, bi represents the regression coefficient of the i-th CpG site, β CpGi The methylation β value of the i-th site, i = 1, 2, ..., n, where n represents the total number of CpG sites.

6. The method for estimating the biological age of a female individual based on a blood sample according to claim 1, characterized in that: The specific calculation formula for predicted age based on the 18 female CpG sites is: DNAmAge2=24.368339+23.87795b cg23606718 +13.018577b cg07553761 +43.418649b cg16867657 +5.323545b cg16419235 +25.920519b cg04875128 +5.456631b cg14692377 -24.671208b cg20822990 +19.352971b cg06639320 +12.056049b cg03607117 +16.060441b cg08097417 +36.607789b cg14361627 -14.83355b cg21120249 +14.210864b cg07850604 -17.753059b cg10917602 +16.100622b cg04084157 -15.738963b cg01511567 +13.98999b cg22736354 -14.712983b cg04474832 Among them, DNAmAge2 represents the predicted age based on 18 female CpG sites, β cg23606718 , β cg07553761 , β cg16867657 , β cg16419235 , β cg04875128 , β cg14692377 , β cg20822990 , β cg06639320 , β cg03607117 , β cg08097417 , β cg14361627 , β cg21120249 , β cg07850604 , β cg10917602 , β cg04084157 , β cg01511567 , β cg22736354 , β cg04474832 All represent the methylation rates of the corresponding CpG sites.

7. The method for estimating the biological age of a female individual based on a blood sample according to claim 1, characterized in that: The accuracy assessment is specifically: The accuracy was evaluated by comparing the performance indicators of the Elastic Net regression model, the SVM machine learning algorithm, the GBR regression model, and the ANN regression model.

8. The method for estimating the biological age of a female individual based on a blood sample according to claim 7, characterized in that: The performance indicators specifically include: Pearson correlation coefficient, median absolute error, determination coefficient, mean square error and root mean square error.

9. A female individual biological age estimation system based on blood samples, characterized in that: include: processor and memory; The memory stores programs or instructions that can be run on the processor, and when the programs or instructions are executed by the processor, the steps of the method for estimating the biological age of a female individual based on a blood sample as described in any one of claims 1 to 8 are implemented.

10. A readable storage medium, characterized in that: The readable storage medium stores a program or instruction, which, when executed by a processor, implements the steps of the method for estimating the biological age of a female individual based on a blood sample as described in any one of claims 1 to 8.

Images