A method and system for constructing a cell aging prediction model based on cell metabolism

Abstract

The present invention discloses a method and system for constructing a cell aging prediction model based on cell metabolism, and relates to the field of prediction model construction. The present invention predicts cell aging from four aspects by selecting content data of four metabolites, and finally determines whether the cell will age in the future by combining the four prediction results, thereby improving the comprehensiveness of the prediction and further improving the accuracy of the prediction; by constructing a first final multi-layer perceptron classification model, a classification model is provided for predicting the corresponding aging level of a first aging characterization difference degree data set matrix; a first final BP prediction model, a second final BP prediction model, a third final BP prediction model, and a fourth final BP prediction model are used to predict multiple metabolite data of a cell sample to be predicted; and a data basis is provided for subsequent determination of the aging status of the cell sample to be predicted in the future.

Description

Technical Field

[0001] The present invention belongs to the field of prediction model construction, and in particular, relates to a method and system for constructing a cell aging prediction model based on cell metabolism. Background Art

[0002] Chinese patent application CN110955961A discloses an immune damage analysis method, which obtains the proportion of immune cell subpopulations; constructs an original data panel of an immune damage model based on the proportion and performs preprocessing to generate panel data of specific immune cell subpopulations for modeling; establishes an immune damage model to score the degree of immune damage of samples, which solves the problem of complex operation, strict experimental conditions, and high requirements for reagent ratios in judging immune cell aging using a cell aging detection kit. It can also quantify the proportion of immune cell subpopulations and the degree of immune damage in the peripheral blood samples of the examinee to determine whether there is premature immune aging.

[0003] There are many factors that affect the occurrence of cell aging. Since the metabolite production content in aging cells is quite different from that in normal cells, it indicates that there is a strong correlation between the metabolite content and the occurrence of cell aging. However, the aging metabolites discovered so far may differ between different cell types, so it is impossible to predict the aging of multiple cells based on a single aging metabolite. Summary of the invention

[0004] In response to the problems in the related art, the present invention proposes a method and system for constructing a cell aging prediction model based on cell metabolism to overcome the above-mentioned technical problems existing in the existing related technology.

[0005] In order to solve the above technical problems, the present invention is achieved through the following technical solutions:

[0006] The present invention is a method for constructing a cell aging prediction model based on cell metabolism, comprising the following steps:

[0007] S1. Collect multiple metabolite data from multiple types of cell samples;

[0008] Constructing a characterization data processing model, wherein the characterization data processing model calculates the degree of difference of aging characterization data of various types of cell samples and constructs a first final multi-layer perceptron classification model to obtain a first aging characterization difference degree data set matrix;

[0009] S2, constructing a first classification model, wherein the first classification model uses the first final multi-layer perceptron classification model to classify the first aging representation difference degree data set matrix to obtain a second cell aging level data matrix; using multiple metabolite data and the second cell aging level data matrix to construct a second final multi-layer perceptron classification model, a third final multi-layer perceptron classification model, a fourth final multi-layer perceptron classification model and a fifth final multi-layer perceptron classification model;

[0010] S3, constructing a first final BP prediction model, a second final BP prediction model, a third final BP prediction model, and a fourth final BP prediction model using multiple metabolite data;

[0011] S4, constructing a prediction model, wherein the prediction model uses the first final BP prediction model, the second final BP prediction model, the third final BP prediction model, and the fourth final BP prediction model to predict multiple metabolite data of the cell sample to be predicted, and obtain a metabolite future data matrix; constructing a second classification model, wherein the second classification model uses the multiple final multi-layer perceptron classification models to classify the metabolite future data matrix, and obtain a third cell aging level data matrix;

[0012] S5, constructing an aging determination model, wherein the aging determination model determines whether the cell sample to be predicted will age in the future according to the third cell aging level data matrix;

[0013] The multiple metabolite data include SDHA antibody content data, oxidized lipid content data, pyruvate dehydrogenase kinase 4 content data and phosphoglycerate dehydrogenase content data; the content of these metabolites will change significantly when the cell ages; therefore, the present invention selects the content data of four metabolites to predict the aging of the cell from four aspects, and finally combines the four prediction results to determine whether the cell will age in the future; based on the above scheme, the comprehensiveness of the prediction is improved, and then the accuracy of the prediction is improved; at the same time, since the morphology and color of the cell will change when the cell ages, therefore, this scheme collects the characterization data of the cell at each growth time point, and compares the difference between the characterization data and the characterization data of the cell in the non-aging period, to obtain a first aging characterization difference degree data set matrix that can reflect whether the cell has aged; at the same time, the first aging characterization difference degree data of multiple cell samples and the corresponding cell aging level data are collected to facilitate the subsequent construction of the first final data set that can predict the cell aging level data based on the cell aging characterization difference degree data. The multi-layer perceptron classification model provides data support; by constructing the first final multi-layer perceptron classification model, a classification model is provided for predicting the corresponding aging level of the first aging characterization difference degree data set matrix; data support is provided for the subsequent construction of the second final multi-layer perceptron classification model, the third final multi-layer perceptron classification model, the fourth final multi-layer perceptron classification model and the fourth final multi-layer perceptron classification model that can predict the corresponding aging level according to the metabolite content data; among them, the metabolite data collected at each time point also provide data support for the subsequent construction of the first final BP prediction model, the second final BP prediction model, the third final BP prediction model and the fourth final BP prediction model that can predict the metabolite content data; by using the constructed first final BP prediction model, the second final BP prediction model, the third final BP prediction model and the fourth final BP prediction model to predict multiple metabolite data of the cell sample to be predicted, the metabolite future data matrix is ​​obtained; it provides a data basis for the subsequent judgment of the aging status of the cell sample to be predicted at a future moment.

[0014] Preferably, the S1 comprises the following steps:

[0015] S11, collecting multiple types of cell samples and cell type label data to obtain a first cell sample set and a first cell type data set;

[0016] S12, configuring a culture medium corresponding to each cell sample in the first cell sample set, and setting a culture time point set;

[0017] Set up aging metabolite type set , , , , SDHA antibody, oxidized lipids, pyruvate dehydrogenase kinase 4, and phosphoglycerate dehydrogenase, respectively;

[0018] Each cell sample in the first cell sample set is placed in a corresponding culture medium for culture; during the culture process, at each culture time point in the culture time point set, the content data of each aging metabolite in the aging metabolite type set of each cell sample is measured respectively to obtain a first metabolite data matrix, a second metabolite data matrix, a third metabolite data matrix, and a fourth metabolite data matrix;

[0019] S13, photographing and collecting image data of each cell sample using a microscope at each culture time point in the culture time point set to obtain a cell image data matrix;

[0020] Setting a cell aging characterization type set; extracting cell aging characterization data of each cell image data in a cell image data matrix using a CNN neural network model according to the cell aging characterization type set to obtain an aging characterization data set matrix;

[0021] S14, calculating the difference between the aging characterization data of each type of cell sample at each culture time point in the aging characterization data set matrix and the corresponding aging characterization data at the first culture time point in the culture time point set, to obtain a first aging characterization difference degree data set matrix;

[0022] S15, constructing the first final multi-layer perceptron classification model;

[0023] Preferably, the cell aging characterization types include cell volume, cell internal water content, cell internal pigment distribution, etc.;

[0024] By collecting image data of various cell samples at different stages in the culture medium, data support is provided for extracting characterization data of the cell samples.

[0025] Preferably, the S15 comprises the following steps:

[0026] S151, setting a cell senescence level set; collecting multiple types of cell samples and corresponding cell senescence level data according to the first cell type data set to obtain a second cell sample set;

[0027] According to the cell aging characterization type set, cell aging characterization difference degree data and corresponding cell aging grade data of each cell in the second cell sample set at each period are obtained to obtain a second aging characterization difference degree data set matrix and a first cell aging grade data matrix;

[0028] S152, constructing a first initial multi-layer perceptron classification model; using the second aging representation difference degree data set matrix and the first cell aging level data matrix to train and test the first initial multi-layer perceptron classification model; after the training and testing are completed, a first final multi-layer perceptron classification model is obtained;

[0029] By using the second aging characterization difference degree data set matrix and the first cell aging level data matrix to train and test the first initial multi-layer perceptron classification model, the constructed first final multi-layer perceptron classification model has a good ability to predict the corresponding aging level data based on the cell aging characterization difference degree data.

[0030] Preferably, the S152 comprises the following steps:

[0031] S1521, setting a first training data ratio; dividing the second aging representation difference degree data set matrix and the first cell aging level data matrix according to the first training data ratio to obtain an aging representation difference degree training data set matrix, an aging representation difference degree test data set matrix, a cell aging level training data matrix, and a cell aging level test data matrix;

[0032] S1522, setting a first training error threshold; inputting the aging representation difference degree training data set matrix and the cell aging level training data matrix into the first initial multi-layer perceptron classification model for training; during the training process, when the training error is less than the first training error threshold, stopping the training; otherwise, continuing the training; after the training is completed, obtaining a trained multi-layer perceptron classification model;

[0033] S1523, setting a first accuracy threshold; inputting the aging characterization difference degree test data set matrix and the cell aging level test data matrix into the trained multi-layer perceptron classification model for testing; after the test is completed, obtaining a first accuracy rate;

[0034] When the first accuracy rate is greater than or equal to the first accuracy rate threshold, the trained multilayer perceptron classification model is used as the first final multilayer perceptron classification model; otherwise, return to S1522 to continue training until the first accuracy rate is greater than or equal to the first accuracy rate threshold, and obtain the first final multilayer perceptron classification model;

[0035] Data partitioning is performed on the second aging characterization difference degree data set matrix and the first cell aging level data matrix, providing training data and test data for training and testing the first initial multi-layer perceptron classification model.

[0036] Preferably, S2 comprises the following steps:

[0037] S21, using the first final multi-layer perceptron classification model to classify the first aging representation difference degree data set matrix Perform classification operation on each aging characterization difference degree data in the data to obtain a second cell aging grade data matrix;

[0038] S22, constructing a second initial multilayer perceptron classification model, a third initial multilayer perceptron classification model, a fourth initial multilayer perceptron classification model, and a fifth initial multilayer perceptron classification model;

[0039] The second initial multi-layer perceptron classification model is trained and tested using the first metabolite data matrix and the second cell aging level data matrix; after the training and testing are completed, a second final multi-layer perceptron classification model is obtained;

[0040] The third initial multi-layer perceptron classification model is trained and tested using the second metabolite data matrix and the second cell aging level data matrix; after the training and testing are completed, a third final multi-layer perceptron classification model is obtained;

[0041] The fourth initial multi-layer perceptron classification model is trained and tested using the third metabolite data matrix and the second cell aging level data matrix; after the training and testing are completed, a fourth final multi-layer perceptron classification model is obtained;

[0042] The fifth initial multi-layer perceptron classification model is trained and tested using the fourth metabolite data matrix and the second cell aging level data matrix; after the training and testing are completed, a fifth final multi-layer perceptron classification model is obtained;

[0043] Since the correspondence between the content data of each metabolite and the corresponding cell aging level data is different, the second initial multi-layer perceptron classification model, the third initial multi-layer perceptron classification model, the fourth initial multi-layer perceptron classification model and the fifth initial multi-layer perceptron classification model are constructed to analyze the content data of the four metabolites respectively. Each model only processes the data of one of the metabolites, thereby improving the accuracy of classification.

[0044] Preferably, S3 comprises the following steps:

[0045] S31, constructing a first initial BP prediction model, a second initial BP prediction model, a third initial BP prediction model and a fourth initial BP prediction model;

[0046] S32, setting a second training error threshold, a third training error threshold, a fourth training error threshold, and a fifth training error threshold;

[0047] Partitioning the first metabolite data matrix, the second metabolite data matrix, the third metabolite data matrix, and the fourth metabolite data matrix respectively to obtain a first metabolite training data matrix, a first metabolite test data matrix, a second metabolite training data matrix, a second metabolite test data matrix, a third metabolite training data matrix, a third metabolite test data matrix, a fourth metabolite training data matrix, and a fourth metabolite test data matrix;

[0048] S33, inputting the first metabolite training data matrix into the first initial BP prediction model for training; during the training process, when the training error is less than the second training error threshold, stopping the training to obtain the first trained BP prediction model; otherwise, continuing the training;

[0049] Inputting the second metabolite training data matrix into the second initial BP prediction model for training; during the training process, when the training error is less than a third training error threshold, stopping the training to obtain a second trained BP prediction model; otherwise, continuing the training;

[0050] Inputting the third metabolite training data matrix into the third initial BP prediction model for training; during the training process, when the training error is less than a fourth training error threshold, stopping the training to obtain the third trained BP prediction model; otherwise, continuing the training;

[0051] Inputting the fourth metabolite training data matrix into the fourth initial BP prediction model for training; during the training process, when the training error is less than the fifth training error threshold, stopping the training to obtain the fourth trained BP prediction model; otherwise, continuing the training;

[0052] Setting initial parameters of the first trained BP prediction model, the second trained BP prediction model, the third trained BP prediction model, and the fourth trained BP prediction model to obtain an initial parameter matrix;

[0053] S34, setting an initial parameter set to be optimized; setting a second accuracy threshold, a third accuracy threshold, a fourth accuracy threshold and a fifth accuracy threshold;

[0054] Input the first metabolite test data matrix into the first trained BP prediction model for testing, input the second metabolite test data matrix into the second trained BP prediction model for testing, input the third metabolite test data matrix into the third trained BP prediction model for testing, and input the fourth metabolite test data matrix into the fourth trained BP prediction model for testing; after all tests are completed, the final parameter set to be optimized is obtained;

[0055] S35, iteratively optimizing each parameter to be optimized in the final parameter set to be optimized to obtain an optimized parameter set; substituting the optimized parameter set into the corresponding trained BP prediction model to obtain a first final BP prediction model, a second final BP prediction model, a third final BP prediction model and a fourth final BP prediction model;

[0056] Since the content data of different types of metabolites have different changing trends as the cells grow, this scheme improves the prediction accuracy by constructing the first final BP prediction model, the second final BP prediction model, the third final BP prediction model and the fourth final BP prediction model to separately predict the content data of different types of metabolites.

[0057] Preferably, in S35, iteratively optimizing each parameter to be optimized in the final parameter set to be optimized to obtain the optimized parameter set comprises the following steps:

[0058] S351, constructing a Harris Hawk population; setting the maximum number of iterations of the Harris Hawk population to , the current number of iterations is And the search space dimension is ;

[0059] S352, setting the initial position of each Harris hawk in the Harris hawk population according to the final parameter set to be optimized to obtain an initial position matrix;

[0060] S353, setting the fitness function of the Harris hawk population;

[0061] S354, start iterating; in each round of iteration, according to the fitness function Calculating the fitness value of each Harris hawk in the Harris hawk population to obtain a fitness value set; selecting the maximum fitness value in the fitness value set as the prey, and executing a global search strategy or a local development strategy according to the position of the prey;

[0062] S355, when When , stop the iteration operation; output the final global optimal position; otherwise, continue to iterate;

[0063] Substituting the final global optimal position into the corresponding trained BP prediction model to obtain the BP prediction model after substitution; testing the BP prediction model after substitution in the manner of S34;

[0064] when , , , If all of them are satisfied, the final global optimal position is used as the optimized parameter set; otherwise, return to S354 and continue to iterate until , , , until all are satisfied, the optimized parameter set is obtained;

[0065] The Harris Eagle optimization algorithm requires few parameters to be adjusted, and can effectively explore the entire search space through random search strategies in the initial stage, which helps to find the global optimal solution. It can also effectively explore the entire search space in the exploration stage to avoid falling into the local optimum, and can quickly converge to the optimal solution in the development stage. Based on the above advantages, the Harris Eagle optimization algorithm is used in this scheme to continuously iteratively optimize the number of nodes in the hidden layer, the number of hidden layers, and the learning rate of the BP prediction model, so as to quickly search for a set of optimal BP prediction model parameters, thereby improving the accuracy of the BP prediction model in predicting metabolite content data.

[0066] Preferably, S4 comprises the following steps:

[0067] S41, collecting a cell sample to be predicted; and setting a historical culture time point set for the cell sample to be predicted;

[0068] S42, placing the cell sample to be predicted in a culture dish for culture; during the culture process, at each historical culture time point in the historical culture time point set, measuring the content data of each aging metabolite in the aging metabolite type set of the cell sample to be predicted, to obtain a metabolite historical data matrix;

[0069] S43, setting a future training time point set;

[0070] Inputting each row of data in the metabolite historical data matrix into the first final BP prediction model, the second final BP prediction model, the third final BP prediction model, and the fourth final BP prediction model to predict the SDHA antibody content data, the oxidized lipid content data, the pyruvate dehydrogenase kinase 4 content data, and the phosphoglycerate dehydrogenase content data at each future culture time point to obtain a metabolite future data matrix;

[0071] S44, using the second final multi-layer perceptron classification model, the third final multi-layer perceptron classification model, the fourth final multi-layer perceptron classification model and the fifth final multi-layer perceptron classification model to classify each row of data in the metabolite future data matrix respectively, to obtain a third cell aging level data matrix;

[0072] Through prediction, the content data of multiple metabolites of the cell sample to be predicted at future moments are obtained, and then the content data of multiple metabolites at future moments are combined with the second final multi-layer perceptron classification model, the third final multi-layer perceptron classification model, the fourth final multi-layer perceptron classification model and the fifth final multi-layer perceptron classification model that have been trained and tested before to obtain the aging level data of the cell sample to be predicted at future moments, which provides a data basis for subsequent judgment of the aging of the cell sample to be predicted at future moments.

[0073] Preferably, the S5 comprises the following steps:

[0074] S51, setting the aging level data threshold;

[0075] S52, when there is cell aging level data in the third cell aging level data matrix that is greater than or equal to the aging level data threshold, it is determined that the cell sample to be predicted will age at the future culture time point corresponding to the cell aging level data, otherwise, no aging will occur;

[0076] The method provides a method for comparing the third cell aging level data matrix with the set aging level data threshold; from the perspective of the reflection of the four metabolites, as long as one metabolite reflects that the cell sample to be predicted will age at a future moment, it is determined that the cell sample to be predicted will age at this moment; thus improving the comprehensiveness and accuracy of the determination.

[0077] A cell aging prediction model construction system based on cell metabolism includes a first metabolite data acquisition module, a characterization data processing module, an aging grade classification model construction module, a first classification module, a metabolite data classification module, a metabolite data prediction model construction module, a metabolite data prediction model construction module, a second metabolite data acquisition module, a prediction module and an aging determination module.

[0078] The present invention has the following beneficial effects:

[0079] 1. The present invention selects content data of four metabolites and multiple cell types to predict the aging of various types of cells from four aspects, and finally combines the four prediction results to determine whether the cells will age in the future, thereby improving the comprehensiveness and universality of the prediction, and thus improving the accuracy of the prediction; by constructing a first final multi-layer perceptron classification model, a classification model is provided for predicting the corresponding aging level of the first aging characterization difference degree data set matrix; the first final BP prediction model, the second final BP prediction model, the third final BP prediction model, and the fourth final BP prediction model are used to predict multiple metabolite data of the cell sample to be predicted; and a data basis is provided for subsequent determination of the aging status of the cell sample to be predicted in the future.

[0080] 2. In the present invention, the first final BP prediction model, the second final BP prediction model, the third final BP prediction model and the fourth final BP prediction model are constructed to separately predict the content data of different types of metabolites, thereby improving the accuracy of the prediction.

[0081] 3. In the present invention, the Harris Eagle optimization algorithm is used to continuously iteratively optimize the number of nodes in the hidden layer, the number of layers in the hidden layer, and the learning rate of the BP prediction model, so that a set of optimal parameters of the BP prediction model can be quickly searched, thereby improving the accuracy of the BP prediction model in predicting metabolite content data.

[0082] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0083] In order to more clearly illustrate the technical solutions of the embodiments of the invention, the drawings required for describing the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the invention. For ordinary technicians in this field, they can also obtain drawings based on these drawings without paying creative work.

[0084] Figure 1 A schematic diagram of the process of constructing a cell aging prediction model by a cell aging prediction model construction system based on cell metabolism of the present invention. DETAILED DESCRIPTION

[0085] The following will be combined with the drawings in the embodiments of the invention to clearly and completely describe the technical solutions in the embodiments of the invention. Obviously, the described embodiments are only part of the embodiments of the invention, not all of the embodiments. Based on the embodiments in the invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the invention.

[0086] Embodiment 1

[0087] This embodiment is a method for constructing a cell aging prediction model based on cell metabolism, comprising the following steps:

[0088] S1. Collect multiple metabolite data from multiple types of cell samples;

[0089] Constructing a characterization data processing model, wherein the characterization data processing model calculates the degree of difference of aging characterization data of various types of cell samples and constructs a first final multi-layer perceptron classification model to obtain a first aging characterization difference degree data set matrix;

[0090] The S1 comprises the following steps:

[0091] S11. Collect multiple types of cell samples and cell type label data to obtain a first cell sample set And the first cell type dataset ; and Respectively represent the collected i types of cell samples and corresponding cell type label data, Indicates the total number of cell sample types collected;

[0092] S12. Configuring the first cell sample set The culture medium corresponding to each cell sample in the culture time point set , Indicates the setting i The cultivation time point, b Indicates the total number of set culture time points;

[0093] Set up aging metabolite type set , , , , SDHA antibody, oxidized lipids, pyruvate dehydrogenase kinase 4, and phosphoglycerate dehydrogenase, respectively;

[0094] The first cell sample set Each cell sample in the culture medium is placed in the corresponding culture medium for culture; during the culture process, at the culture time point, At each culture time point in the aging metabolite type set, the content data of each aging metabolite in each cell sample is measured separately to obtain the first metabolite data matrix , the second metabolite data matrix , the third metabolite data matrix And the fourth metabolite data matrix ; They are as follows,

[0095] ; ; ; ;

[0096] in, , , , Indicated in i At the culturing time point j SDHA antibody content data, oxidized lipid content data, pyruvate dehydrogenase kinase 4 content data, and phosphoglycerate dehydrogenase content data of the cell samples of different types;

[0097] S13, at each culture time point in the culture time point set, the image data of each cell sample is photographed and collected using a microscope to obtain a cell image data matrix ;as follows,

[0098] ;

[0099] in, Indicated in i The first j Image data of cell samples of different types;

[0100] Set up cell senescence characterization type set , Indicates the setting i cell aging characterization type; extract the cell aging characterization data of each cell image data in the cell image data matrix using the CNN neural network model according to the cell aging characterization type set to obtain an aging characterization data set matrix ;as follows,

[0101] ;

[0102] in, express The corresponding aging characterization dataset, , express Middle k Types of aging characterization data;

[0103] S14, calculating the difference between the aging characterization data of each type of cell sample at each culture time point in the aging characterization data set matrix and the corresponding aging characterization data at the first culture time point in the culture time point set, to obtain a first aging characterization difference degree data set matrix ;as follows,

[0104] ;

[0105] in, express The corresponding aging representation difference degree dataset, , express Middle k The calculation formula is as follows:

[0106] ;

[0107] In the formula, express The corresponding aging characterization dataset k Types of aging characterization data;

[0108] S15, constructing the first final multi-layer perceptron classification model;

[0109] The S15 comprises the following steps:

[0110] S151, set the cell aging level set , Indicates the setting i The level of cell senescence, represents the total number of set cell aging levels; collecting multiple types of cell samples and corresponding cell aging level data according to the first cell type data set to obtain a second cell sample set , Indicates the number of i Cell samples;

[0111] According to the cell aging characterization type set, the cell aging characterization difference degree data and the corresponding cell aging grade data of each cell in the second cell sample set at each stage are obtained to obtain the second aging characterization difference degree data set matrix And the first cell senescence level data matrix ; They are as follows,

[0112] ; ;

[0113] in, and Respectively represent the second cell sample set i The cell samples were j The data set of differences in the degree of cell aging characterization in different periods and the corresponding cell aging grade data; , express The k Data on the degree of differences in the types of aging manifestations; e represents the total number of data collection periods for each cell sample in the second cell sample set;

[0114] S152, constructing a first initial multi-layer perceptron classification model; using the second aging representation difference degree data set matrix and the first cell aging level data matrix to train and test the first initial multi-layer perceptron classification model; after the training and testing are completed, a first final multi-layer perceptron classification model is obtained;

[0115] The S152 comprises the following steps:

[0116] S1521, setting a first training data ratio; dividing the second aging representation difference degree data set matrix and the first cell aging level data matrix according to the first training data ratio to obtain an aging representation difference degree training data set matrix, an aging representation difference degree test data set matrix, a cell aging level training data matrix, and a cell aging level test data matrix;

[0117] S1522, setting a first training error threshold; inputting the aging representation difference degree training data set matrix and the cell aging level training data matrix into the first initial multi-layer perceptron classification model for training; during the training process, when the training error is less than the first training error threshold, stopping the training; otherwise, continuing the training; after the training is completed, obtaining a trained multi-layer perceptron classification model;

[0118] S1523, setting a first accuracy threshold; inputting the aging characterization difference degree test data set matrix and the cell aging level test data matrix into the trained multi-layer perceptron classification model for testing; after the test is completed, obtaining a first accuracy rate;

[0119] When the first accuracy rate is greater than or equal to the first accuracy rate threshold, the trained multilayer perceptron classification model is used as the first final multilayer perceptron classification model; otherwise, return to S1522 to continue training until the first accuracy rate is greater than or equal to the first accuracy rate threshold, and obtain the first final multilayer perceptron classification model;

[0120] S2, constructing a first classification model, wherein the first classification model uses the first final multi-layer perceptron classification model to classify the first aging representation difference degree data set matrix to obtain a second cell aging level data matrix; using multiple metabolite data and the second cell aging level data matrix to construct a second final multi-layer perceptron classification model, a third final multi-layer perceptron classification model, a fourth final multi-layer perceptron classification model and a fifth final multi-layer perceptron classification model;

[0121] The S2 comprises the following steps:

[0122] S21, using the first final multi-layer perceptron classification model to perform a classification operation on each aging representation difference degree data in the first aging representation difference degree data set matrix to obtain a second cell aging level data matrix ;as follows,

[0123] ;

[0124] in, express The corresponding cellular senescence grade data of the senescence characterization difference degree dataset;

[0125] S22, constructing a second initial multilayer perceptron classification model, a third initial multilayer perceptron classification model, a fourth initial multilayer perceptron classification model, and a fifth initial multilayer perceptron classification model;

[0126] The second initial multi-layer perceptron classification model is trained and tested using the first metabolite data matrix and the second cell aging level data matrix; after the training and testing are completed, a second final multi-layer perceptron classification model is obtained;

[0127] The third initial multi-layer perceptron classification model is trained and tested using the second metabolite data matrix and the second cell aging level data matrix; after the training and testing are completed, a third final multi-layer perceptron classification model is obtained;

[0128] The fourth initial multi-layer perceptron classification model is trained and tested using the third metabolite data matrix and the second cell aging level data matrix; after the training and testing are completed, a fourth final multi-layer perceptron classification model is obtained;

[0129] The fifth initial multi-layer perceptron classification model is trained and tested using the fourth metabolite data matrix and the second cell aging level data matrix; after the training and testing are completed, a fifth final multi-layer perceptron classification model is obtained;

[0130] S3, constructing a first final BP prediction model, a second final BP prediction model, a third final BP prediction model, and a fourth final BP prediction model using multiple metabolite data;

[0131] The S3 comprises the following steps:

[0132] S31, constructing a first initial BP prediction model, a second initial BP prediction model, a third initial BP prediction model and a fourth initial BP prediction model;

[0133] S32, setting a second training error threshold, a third training error threshold, a fourth training error threshold, and a fifth training error threshold;

[0134] Then set the second training data ratio, the third training data ratio, the fourth training data ratio and the fifth training data ratio; divide the first metabolite data matrix, the second metabolite data matrix, the third metabolite data matrix and the fourth metabolite data matrix according to the second training data ratio, the third training data ratio, the fourth training data ratio and the fifth training data ratio, respectively, to obtain the first metabolite training data matrix, the first metabolite test data matrix, the second metabolite training data matrix, the second metabolite test data matrix, the third metabolite training data matrix, the third metabolite test data matrix, the fourth metabolite training data matrix and the fourth metabolite test data matrix;

[0135] S33, inputting the first metabolite training data matrix into the first initial BP prediction model for training; during the training process, when the training error is less than the second training error threshold, stopping the training to obtain the first trained BP prediction model; otherwise, continuing the training;

[0136] Inputting the second metabolite training data matrix into the second initial BP prediction model for training; during the training process, when the training error is less than a third training error threshold, stopping the training to obtain a second trained BP prediction model; otherwise, continuing the training;

[0137] Inputting the third metabolite training data matrix into the third initial BP prediction model for training; during the training process, when the training error is less than a fourth training error threshold, stopping the training to obtain the third trained BP prediction model; otherwise, continuing the training;

[0138] Inputting the fourth metabolite training data matrix into the fourth initial BP prediction model for training; during the training process, when the training error is less than the fifth training error threshold, stopping the training to obtain the fourth trained BP prediction model; otherwise, continuing the training;

[0139] Set the initial parameters of the first trained BP prediction model, the second trained BP prediction model, the third trained BP prediction model and the fourth trained BP prediction model to obtain an initial parameter matrix ;as follows,

[0140] ;

[0141] in, , , Respectively represent the number of nodes in the hidden layer, the number of layers in the hidden layer, and the learning rate of the first trained BP prediction model; , , Respectively represent the number of nodes in the hidden layer, the number of layers in the hidden layer, and the learning rate of the second trained BP prediction model; , , Respectively represent the number of nodes in the hidden layer, the number of layers in the hidden layer and the learning rate of the third trained BP prediction model; , , Respectively represent the number of nodes in the hidden layer, the number of layers in the hidden layer, and the learning rate of the fourth trained BP prediction model;

[0142] S34, setting an initial parameter set to be optimized; setting a second accuracy threshold , the third accuracy threshold , the fourth accuracy threshold and the fifth accuracy threshold ;

[0143] Inputting the first metabolite test data matrix into a first trained BP prediction model for testing, inputting the second metabolite test data matrix into a second trained BP prediction model for testing, inputting the third metabolite test data matrix into a third trained BP prediction model for testing, and inputting the fourth metabolite test data matrix into a fourth trained BP prediction model for testing;

[0144] After the test is completed, the second accuracy is obtained , the third accuracy , the fourth accuracy and fifth accuracy ;

[0145] when When the first trained BP prediction model is used as the first final BP prediction model; otherwise, the number of nodes in the hidden layer, the number of layers in the hidden layer, and the learning rate of the first trained BP prediction model are added to the initial set of parameters to be optimized;

[0146] when When the second trained BP prediction model is used as the second final BP prediction model; otherwise, the number of nodes in the hidden layer, the number of layers in the hidden layer, and the learning rate of the second trained BP prediction model are added to the initial parameter set to be optimized;

[0147] when When the third trained BP prediction model is used as the third final BP prediction model; otherwise, the number of nodes in the hidden layer, the number of layers in the hidden layer, and the learning rate of the third trained BP prediction model are added to the initial parameter set to be optimized;

[0148] when When the fourth trained BP prediction model is used as the fourth final BP prediction model; otherwise, the number of nodes in the hidden layer, the number of layers in the hidden layer, and the learning rate of the fourth trained BP prediction model are added to the initial parameter set to be optimized;

[0149] After all tests are completed, the final set of parameters to be optimized is obtained , Indicates i The parameters that need to be optimized are Indicates the total number of parameters that need to be optimized;

[0150] S35, iteratively optimize each parameter to be optimized in the final parameter set to be optimized to obtain an optimized parameter set , Indicates i optimized parameters; substituting the optimized parameter set into the corresponding trained BP prediction model to obtain a first final BP prediction model, a second final BP prediction model, a third final BP prediction model and a fourth final BP prediction model;

[0151] In S35, each parameter to be optimized in the final parameter set to be optimized is iteratively optimized to obtain an optimized parameter set. The following steps are involved:

[0152] S351. Establishing Harris Hawk Population , The Harris Hawk population i Harris Hawk, represents the size of the Harris Hawk population; sets the maximum number of iterations of the Harris Hawk population to , the current number of iterations is And the search space dimension is ;

[0153] S352: setting the initial position of each Harris hawk in the Harris hawk population according to the final parameter set to be optimized, and obtaining an initial position matrix ;as follows,

[0154] ;

[0155] in, The Harris Hawk population i The initial position of the Harris Hawk is j The search space dimension is calculated as follows:

[0156] ;

[0157] In the formula, Indicates for Generate a random number between 0 and 1; , Respectively The upper and lower limits of the value of ;

[0158] S353, setting the fitness function of the Harris hawk population ;as follows,

[0159] ;

[0160] In the formula,β A positive number indicates a correction parameter;

[0161] S354, start iterating; in each round of iteration, according to the fitness function Calculate the fitness value of each Harris hawk in the Harris hawk population to obtain a fitness value set; select the maximum fitness value in the fitness value set as the prey, and the position of the prey is ; Set the prey escape energy of the current iteration to ; The calculation formula is as follows,

[0162] ;

[0163] In the formula, Represents a random number between -1 and 1 generated during the current round of iteration;

[0164] when When , the Harris hawk population executes a global search strategy; otherwise, it executes a local development strategy;

[0165] S355, when When , stop the iteration operation; output the final global optimal position; otherwise, continue to iterate;

[0166] Substituting the final global optimal position into the corresponding trained BP prediction model to obtain the BP prediction model after substitution; testing the BP prediction model after substitution in the manner of S34;

[0167] when , , , If all of them are satisfied, the final global optimal position is used as the optimized parameter set; otherwise, return to S354 and continue to iterate until , , , until all are satisfied, the optimized parameter set is obtained;

[0168] S4, constructing a prediction model, wherein the prediction model uses the first final BP prediction model, the second final BP prediction model, the third final BP prediction model, and the fourth final BP prediction model to predict multiple metabolite data of the cell sample to be predicted, and obtain a metabolite future data matrix; constructing a second classification model, wherein the second classification model uses the multiple final multi-layer perceptron classification models to classify the metabolite future data matrix, and obtain a third cell aging level data matrix;

[0169] The S4 comprises the following steps:

[0170] S41, collecting a cell sample to be predicted; setting a historical culture time point set for the cell sample to be predicted , represents the first i A historical training time point, represents the total number of historical culture time points set for the cell sample to be predicted;

[0171] S42, placing the cell sample to be predicted in a culture dish for culture; during the culture process, at each historical culture time point in the historical culture time point set, measuring the content data of each aging metabolite in the aging metabolite type set of the cell sample to be predicted, to obtain a metabolite historical data matrix;

[0172] S43, setting a set of future culture time points; inputting each row of data in the metabolite historical data matrix into the first final BP prediction model, the second final BP prediction model, the third final BP prediction model, and the fourth final BP prediction model to predict the SDHA antibody content data, the oxidized lipid content data, the pyruvate dehydrogenase kinase 4 content data, and the phosphoglycerate dehydrogenase content data at each future culture time point to obtain a metabolite future data matrix;

[0173] S44, using the second final multi-layer perceptron classification model, the third final multi-layer perceptron classification model, the fourth final multi-layer perceptron classification model and the fifth final multi-layer perceptron classification model to classify each row of data in the metabolite future data matrix respectively, to obtain a third cell aging level data matrix;

[0174] S5, constructing an aging determination model, wherein the aging determination model determines whether the cell sample to be predicted will age in the future according to the third cell aging level data matrix;

[0175] The S5 comprises the following steps:

[0176] S51, setting the aging level data threshold;

[0177] S52. When there is cell aging level data in the third cell aging level data matrix that is greater than or equal to the aging level data threshold, it is determined that the cell sample to be predicted will age at the future culture time point corresponding to the cell aging level data; otherwise, no aging will occur.

[0178] Embodiment 2

[0179] This embodiment discloses a cell aging prediction model construction system based on cell metabolism, the system can implement the method of the above embodiment, including a first metabolite data acquisition module, a characterization data processing module, an aging level classification model construction module, a first classification module, a metabolite data classification module, a metabolite data prediction model construction module, a metabolite data prediction model construction module, a second metabolite data acquisition module, a prediction module and an aging determination module;

[0180] The first metabolite data collection module is used to collect multiple metabolite data of multiple types of cell samples;

[0181] The characterization data processing module is used to calculate the difference degree of aging characterization data of various types of cell samples to obtain a first aging characterization difference degree data set matrix;

[0182] The aging level classification model building module is used to build a first final multi-layer perceptron classification model;

[0183] The first classification module is used to classify the first aging representation difference degree data set matrix using the first final multi-layer perceptron classification model to obtain a second cell aging level data matrix;

[0184] The metabolite data classification module is used to construct a second final multi-layer perceptron classification model, a third final multi-layer perceptron classification model, a fourth final multi-layer perceptron classification model and a fourth final multi-layer perceptron classification model using a plurality of metabolite data and a second cell aging level data matrix;

[0185] The metabolite data prediction model construction module is used to construct a first final BP prediction model, a second final BP prediction model, a third final BP prediction model, and a fourth final BP prediction model using multiple metabolite data;

[0186] The second metabolite data collection module is used to collect multiple metabolite data of the cell sample to be predicted;

[0187] The prediction module is used to predict multiple metabolite data of the cell sample to be predicted using the first final BP prediction model, the second final BP prediction model, the third final BP prediction model, and the fourth final BP prediction model to obtain a metabolite future data matrix;

[0188] The aging determination module is used to determine whether the cell sample to be predicted will age in the future according to the future data matrix of the metabolites.

[0189] In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0190] The preferred embodiments of the invention disclosed above are only used to help explain the invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to the specific implementation methods described. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the invention, so that those skilled in the art can understand and use the invention well.

Claims

1. A method for constructing a cell aging prediction model based on cell metabolism, characterized in that: The following steps are involved: S1. Collect multiple metabolite data from multiple types of cell samples; Constructing a characterization data processing model, wherein the characterization data processing model calculates the degree of difference of aging characterization data of various types of cell samples and constructs a first final multi-layer perceptron classification model to obtain a first aging characterization difference degree data set matrix; S2. constructing a first classification model, wherein the first classification model uses the first final multi-layer perceptron classification model to classify the first aging representation difference degree data set matrix to obtain a second cell aging level data matrix; A variety of final multi-layer perceptron classification models were constructed using a variety of metabolite data and the second cell aging level data matrix; S3, using multiple metabolite data to construct multiple final BP prediction models; The S3 comprises the following steps: S31, constructing multiple initial BP prediction models; S32, performing data partitioning on the multiple metabolite data to obtain a first metabolite training data matrix, a first metabolite test data matrix, a second metabolite training data matrix, a second metabolite test data matrix, a third metabolite training data matrix, a third metabolite test data matrix, a fourth metabolite training data matrix, and a fourth metabolite test data matrix; S33, using the first metabolite training data matrix, the second metabolite training data matrix, the third metabolite training data matrix and the fourth metabolite training data matrix to train multiple initial BP prediction models respectively and set corresponding parameters to obtain a first trained BP prediction model, a second trained BP prediction model, a third trained BP prediction model, a fourth trained BP prediction model and an initial parameter matrix; S34, setting an initial parameter set to be optimized; setting a second accuracy threshold , the third accuracy threshold , the fourth accuracy threshold and the fifth accuracy threshold ; The first metabolite test data matrix, the second metabolite test data matrix, the third metabolite test data matrix and the fourth metabolite test data matrix are respectively input into a plurality of trained BP prediction models for testing; after the test is completed, the second accuracy rates are respectively obtained , the third accuracy , the fourth accuracy and fifth accuracy ;when When the first trained BP prediction model is used as the first final BP prediction model; when When the second trained BP prediction model is used as the second final BP prediction model; when When , the third trained BP prediction model is used as the third final BP prediction model; when When the fourth trained BP prediction model is used as the fourth final BP prediction model; when , , as well as If the conditions are not met at the same time, the number of nodes in the hidden layer, the number of layers in the hidden layer and the learning rate of the corresponding trained BP prediction model are added to the initial parameter set to be optimized; after all tests are completed, the final parameter set to be optimized is obtained; S35, iteratively optimizing each parameter to be optimized in the final parameter set to be optimized to obtain an optimized parameter set; substituting the optimized parameter set into the corresponding trained BP prediction model to obtain a first final BP prediction model, a second final BP prediction model, a third final BP prediction model and a fourth final BP prediction model; S4, constructing a prediction model, wherein the prediction model uses multiple final BP prediction models to predict multiple metabolite data of the cell sample to be predicted, and obtains a metabolite future data matrix; Constructing a second classification model, wherein the second classification model uses the multiple final multi-layer perceptron classification models to classify the metabolite future data matrix to obtain a third cell aging level data matrix; S5. Constructing an aging determination model, wherein the aging determination model determines whether the cell sample to be predicted will age in the future according to the third cell aging level data matrix.

2. The method for constructing a cell aging prediction model based on cell metabolism according to claim 1, characterized in that: The S1 comprises the following steps: S11, measuring the content data of multiple metabolites of multiple cell samples at multiple culture time points respectively to obtain multiple metabolite data matrices; S12, extracting cell aging characterization data from cell image data of multiple cell samples to obtain an aging characterization data set matrix; S13, calculating the difference between the aging characterization data of each type of cell sample at each culture time point in the aging characterization data set matrix and the corresponding aging characterization data at the first culture time point in the culture time point set, to obtain a first aging characterization difference degree data set matrix; S14. Construct the first final multi-layer perceptron classification model.

3. The method for constructing a cell aging prediction model based on cell metabolism according to claim 2, characterized in that: The S14 comprises the following steps: S141, setting a cell senescence level set; collecting and calculating cell senescence characteristic difference degree data of multiple cells at various stages and corresponding cell senescence level data according to the cell senescence level set, to obtain a second senescence characteristic difference degree data set matrix and a first cell senescence level data matrix; S142, constructing a first initial multi-layer perceptron classification model; using the second aging characterization difference degree data set matrix and the first cell aging level data matrix to train and test the first initial multi-layer perceptron classification model; after the training and testing are completed, obtaining the first final multi-layer perceptron classification model.

4. The method for constructing a cell aging prediction model based on cell metabolism according to claim 3, characterized in that: The S2 comprises the following steps: S21, using the first final multi-layer perceptron classification model to perform a classification operation on the first aging representation difference degree data set matrix to obtain a second cell aging level data matrix; S22. Construct multiple initial multi-layer perceptron classification models; use the multiple metabolite data matrices and the second cell aging level data matrix to train and test the multiple initial multi-layer perceptron classification models respectively; obtain the second final multi-layer perceptron classification model, the third final multi-layer perceptron classification model, the fourth final multi-layer perceptron classification model and the fifth final multi-layer perceptron classification model.

5. The method for constructing a cell aging prediction model based on cell metabolism according to claim 4, characterized in that: In S35, each parameter to be optimized in the final parameter set to be optimized is iteratively optimized to obtain the optimized parameter set, which includes the following steps: S351, constructing a Harris Hawk population; setting the maximum number of iterations of the Harris Hawk population to , the current number of iterations is ; S352, setting the initial position of each Harris hawk in the Harris hawk population according to the final parameter set to be optimized to obtain an initial position matrix; S353, setting the fitness function of the Harris hawk population; S354, start the iteration operation; in each round of iteration, calculate the fitness value of each Harris hawk in the Harris hawk population according to the fitness function to obtain a fitness value set; select the maximum fitness value in the fitness value set as the prey, and set the prey escape energy of the current round of iteration to ;when When , the Harris hawk population executes a global search strategy; otherwise, it executes a local development strategy; S355, when When , stop the iteration operation; output the final global optimal position; otherwise, continue to iterate; Substituting the final global optimal position into the corresponding trained BP prediction model to obtain the BP prediction model after substitution; testing the BP prediction model after substitution in the manner of S34; when , , , If all of them are satisfied, the final global optimal position is used as the optimized parameter set; otherwise, return to S354 and continue to iterate until , , , Until all are satisfied, the optimized parameter set is obtained.

6. The method for constructing a cell aging prediction model based on cell metabolism according to claim 5, characterized in that: The S4 comprises the following steps: S41, collecting a cell sample to be predicted; and setting a historical culture time point set for the cell sample to be predicted; S42, measuring the content data of each aging metabolite of the cell sample to be predicted at each historical culture time point in the historical culture time point set, to obtain a metabolite historical data matrix; S43, setting a future culture time point set; predicting the future metabolite content data of the cell sample to be predicted according to the metabolite historical data matrix and in conjunction with the first final BP prediction model, the second final BP prediction model, the third final BP prediction model and the fourth final BP prediction model to obtain a metabolite future data matrix; S44. Use the second final multi-layer perceptron classification model, the third final multi-layer perceptron classification model, the fourth final multi-layer perceptron classification model and the fifth final multi-layer perceptron classification model to classify each row of data in the metabolite future data matrix respectively to obtain the third cell aging level data matrix.

7. The method for constructing a cell aging prediction model based on cell metabolism according to claim 6, characterized in that: The S5 comprises the following steps: S51, setting the aging level data threshold; S52. When there is cell aging level data in the third cell aging level data matrix that is greater than or equal to the aging level data threshold, it is determined that the cell sample to be predicted will age at the future culture time point corresponding to the cell aging level data; otherwise, no aging will occur.

8. A system for implementing the method for constructing a cell aging prediction model based on cell metabolism as described in any one of claims 1 to 7.

Images