Regression prediction method for ipsc pluripotency based on raman spectrum and cnn-lstm network regression model integration

Abstract

The application belongs to the technical field of stem cell identification, and particularly relates to an iPSC pluripotency regression prediction method based on a Raman spectrum and a CNN-LSTM network regression model integration, and comprises the following steps: data set acquisition, culturing iPSC in a culture medium, collecting Raman spectra of the iPSC in a culture process by using a Raman spectrometer, collecting Raman spectra of multiple cells at each time period, and grouping according to time periods; pre-processing the collected Raman spectra; constructing and dividing the data set, dividing the Raman spectra of the cell samples into six categories according to time sequences, and constructing and randomly dividing the data set; training and testing the CNN-LSTM model, and evaluating the pluripotency of the iPSC; and accurately quantitatively evaluating the discrete pluripotent state and transition ability gradient of the stem cell by using a regression method.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of stem cell identification, and particularly relates to an iPSC pluripotency regression prediction method based on Raman spectrum and a CNN-LSTM network regression model set. BACKGROUND

[0002] Induced pluripotent stem cells (iPSCs) have important scientific significance and application value in regenerative medicine, human disease modeling, tissue engineering, etc. due to their self-renewal and differentiation potential into multiple cell types, which can promote innovation in the fields of medicine, drug development, and regenerative medicine.

[0003] The pluripotency of iPSCs enables them to differentiate into various cell types such as cardiomyocytes, neural cells, and pancreatic islet cells, which can be used to treat a variety of diseases. iPSCs have a wide range of applications in disease modeling, personalized medicine, developmental biology, and basic research. Their unique pluripotency and plasticity make them an important tool in modern biomedical research and clinical applications. The clinical translation of iPSC-derived therapies is highly dependent on strict quality control of pluripotency. Pluripotency is a dynamic cellular state that requires continuous monitoring and accurate evaluation before differentiation to ensure cell consistency and treatment safety. Currently, methods for detecting iPSC pluripotency mainly rely on genomics, proteomics, and metabolomics techniques such as immunocytochemistry, flow cytometry, Western blot, and polymerase chain reaction. These methods are tedious, time-consuming, and not cost-effective, and are destructive and invasive to cells, which can compromise the integrity of stem cells. Moreover, these methods require prior knowledge of specific biomarkers, further affecting their practical application.

[0004] Raman spectroscopy is a non-invasive optical scattering spectroscopy technique based on Raman scattering effect, which can reflect the chemical composition and structural information of the sample in a non-destructive manner, and understand the molecular dynamics of basic cellular processes. Raman spectroscopy contains useful information about molecular bonds and structures, and can identify chemical components in substances, so this technology is widely used in molecular structure identification, industrial process control, life science research, etc. In the field of life sciences, Raman spectroscopy is particularly suitable for stem cell research, which can non-invasively monitor the molecular changes and pluripotency state of stem cells, providing important technical support for stem cell culture, differentiation, and application. Unlike traditional analysis methods, Raman spectroscopy does not require the use of predetermined markers or external labels, and can directly analyze the internal state of stem cells in a simpler way, providing non-invasive, real-time molecular information. In particular, in characterizing the pluripotency of iPSCs and its dynamic changes, Raman spectroscopy exhibits unique advantages, providing a powerful tool for monitoring the differentiation state of stem cells, evaluating pluripotency, and studying the cell reprogramming process, with broad application potential.

[0005] However, the existing spectrum analysis strategy is difficult to capture the subtle state changes of cells during differentiation, and cannot accurately quantify the dynamic changes of stem cells when they transition from one pluripotent state to another. On the other hand, the change of cell state is a spatiotemporal dynamic process, which needs to consider the information of both time and space dimensions to achieve a comprehensive understanding of the pluripotency state of cells. Traditional spectrum analysis methods such as principal component analysis, linear discriminant analysis, and partial least squares regression can only provide static spectral information, and it is difficult to effectively extract accurate spatiotemporal spectral patterns. Deep learning methods such as Convolutional Neural Network (CNN), Gated Recurrent Unit (GRU), and Long Short-Term Memory (LSTM) models are mainly used for discrete state classification in the analysis of stem cell pluripotency, but such methods lose the continuity information between stem cell states, cannot reflect the dynamic transition process of stem cells, and cannot quantify the ability gradient of state transition, making it difficult to reflect the gradual process of stem cell dynamic transition. CNN, LSTM have certain limitations in processing sequence data and spatiotemporal dynamic changes. Therefore, it is of great significance to develop a method that does not rely on tedious manual feature selection, can simultaneously process the spatial and temporal features of spectral data, and can accurately evaluate the pluripotency of stem cells and the transition ability gradient quantification technology without loss. SUMMARY

[0006] The purpose of the present application is to provide a regression prediction method for iPSC pluripotency based on Raman spectrum and CNN-LSTM network regression model integration, which can accurately quantify and evaluate the discrete pluripotent state and transition ability gradient of stem cells through regression method.

[0007] The technical solutions adopted by the present application are as follows:

[0008] The regression prediction method for iPSC pluripotency based on Raman spectrum and CNN-LSTM network regression model integration comprises the following steps:

[0009] Step 1: Obtain the data set, culture iPSC in the culture medium, and use the Raman spectrometer to collect the Raman spectrum of iPSC during the culture process; the collection process is divided into multiple time periods, and the spectrum data of multiple cells are collected in each time period;

[0010] Step 2: Data preprocessing, baseline removal, cosmic ray removal, noise reduction and normalization processing of Raman spectrum data;

[0011] Step 3: Construct and divide the data set, divide the Raman spectra of the cell samples into 6 stages in chronological order within a time period, and each stage of spectral data is a class, a total of six sets of spectral data, and the processed spectral data is randomly divided into a training set, a validation set and a test set;

[0012] Step 4: Construct a CNN-LSTM network regression model and train it, which is composed of CNN and LSTM two parts; the CNN layer is used to extract the key spectral features of the Raman fingerprint, and the LSTM layer learns the long-term dependence of the time sequence, combines the spatial features and time sequence dependence of the Raman spectrum, and identifies the spectrum in different time periods;

[0013] Step 5: Evaluate the pluripotency of iPSCs with the trained CNN-LSTM network regression model, and directly input the pre-reserved test set not involved in network training in step 3 into the trained model, the test set does not participate in training or hyperparameter tuning, and is used for final model performance evaluation to predict the strength of the transition of stem cells between different states. The network evaluates the pluripotency of stem cells, quantifies the pluripotent state as a continuous value, and predicts the specific value of each state through the regression model.

[0014] Preferably, in step 1, during the cultivation of iPSCs, the iPSC culture solution is replaced with RPMI1640 culture solution at 1, 3, 6, 12 and 24 hours respectively, so as to affect the differentiation of iPSCs, and the spectra of iPSC cells at five time nodes and the original iPSC cells without replacing the culture solution are collected.

[0015] In the data acquisition of the present application, the laser wavelength is 633 nm, the laser power is set to 50%, the numerical aperture is 0.5, and the L50x objective is used to collect the Raman spectrum in the range of 600-1800 cm -1 For each time point sample, at least 50 cells are selected for signal acquisition, and each cell sample is measured at least 15 times.

[0016] Preferably, in step 1, the fluorescence intensity values of stem cell markers SOX2 and OCT4 are detected by immunofluorescence staining technology; ImageJ is used to calculate the average immunofluorescence intensity of the iPSC state, and the average fluorescence intensity = the total fluorescence intensity of the colony area / the area of the colony area.

[0017] Preferably, in step 2, when the obtained data is preprocessed, WiRE4.3 software is used, including baseline removal, cosmic ray removal, noise reduction, normalization of spectral data before the data set is input into the network for training, and the data value is scaled to the range of [0, 1] to eliminate the potential influence of spectral intensity changes on the model results.

[0018] Preferably, in step 3, the Raman spectrum is divided into six stages in chronological order of processing, and the Raman spectrum data is randomly extracted into a training set, a validation set and a test set in a ratio of 8:1:1; the random division process uses stratified sampling to ensure the balance of the class distribution, and ensures that the spectrum data of the same cell will only appear in one of the training set, the validation set and the test set, ensuring the reliability of the results; the divided training set, validation set and test set are used for model training, hyperparameter tuning and performance evaluation, respectively.

[0019] Preferably, in step 4, the CNN-LSTM network regression model is composed of a CNN layer and an LSTM layer; the local features of the spectrum data are extracted by CNN, the input data dimension is the spectrum length, and after four layers of convolution with kernel size of 12, 14, 16 and 16, the output dimension is 16, 32, 64 and 128 in turn; the LSTM layer models the time sequence of the 128-dimensional feature sequence, captures the time sequence dependency through four layers of bidirectional LSTM, and then maps the features to the regression values of the six time points through the full connection layer, corresponding to the pluripotency scores of the two markers SOX2 and OCT4, to realize high-precision pluripotency evaluation.

[0020] In step 4, during the training of the CNN-LSTM network regression model, Adam is used as the optimizer to train the model, the learning rate of Adam is set to 1x10 -5 , and mean square error loss is used as the loss function; the CNN-LSTM network regression model is built by PyTorch, and the training set and the validation set after data set processing in step 3 are put into the CNN-LSTM network regression model for training, at the end of each round, the model performance is evaluated on the validation set, and the training loss (Train Loss), the validation loss (Val Loss), the root mean square error (Root Mean Square Error, RMSE) and the determination coefficient R² are recorded.

[0021] The technical effects achieved by the present application are:

[0022] The present application realizes quantitative evaluation of pluripotency. The key features of the Raman spectrum are extracted by the adaptive wide kernel convolution module, and the dynamic changes of the spectrum sequence are captured by combining the LSTM time-dependent learning module, which improves the recognition accuracy of the spectrum details and time sequence features. The CNN-LSTM model can retain the continuity information between the states of stem cells through the regression prediction of the Raman spectrum, accurately reflect the dynamic change process of the stem cells, quantitatively represent the ability gradient of the transition between states, and achieve the effect of accurately evaluating the pluripotency state and transition ability.

[0023] Compared with the traditional method for detecting the pluripotency of stem cells, the present application can obtain the biomolecular information of stem cells in different pluripotent states without the destructive and lengthy sample pretreatment process in the traditional detection method, and the spectral characteristic information in the cells is retained. Therefore, the present application can realize accurate and non-destructive evaluation of the pluripotent state during the differentiation of stem cells.

[0024] The present application forms a powerful feature extraction network with the CNN and LSTM modules, which can accurately extract local features and time sequence dependencies in complex spectral data. Moreover, through multi-layer convolution and pooling operations, the model can more comprehensively perceive the key features and dynamic change information in the spectral data, and realize accurate prediction of the pluripotent state of stem cells. BRIEF DESCRIPTION OF DRAWINGS

[0025] Figure 1 Flowchart of an embodiment of the present application;

[0026] Figure 2 Induction scheme for induced pluripotent stem cells;

[0027] Figure 3 Immunofluorescence images and fluorescence intensity column charts of the induced pluripotent stem cells at different time points after replacing the RPMI1640 medium;

[0028] Figure 4 Single-cell average Raman spectrum of six cell states during the culture of stem cells based on RPMI1640 medium;

[0029] Figure 5 Schematic diagram of the CNN-LSTM network regression model;

[0030] Figure 6 Loss curve of the CNN-LSTM network trained for 600 times;

[0031] Figure 7 Comparison chart of the predicted and actual values of the fluorescence intensity of the iPSC marker SOX2 over time;

[0032] Figure 8 Comparison chart of the predicted and actual values of the fluorescence intensity of the iPSC marker OCT4 over time. DETAILED DESCRIPTION

[0033] In order to make the purpose and advantages of the present application clearer and more apparent, the present application will be specifically described below in combination with embodiments. It should be understood that the following text is only used to describe one or several specific embodiments of the present application, and does not strictly limit the specific protection scope requested by the present application.

[0034] As Figure 1As shown, the regression prediction method of iPSC pluripotency based on Raman spectrum and CNN-LSTM network regression model set integration includes the following steps:

[0035] Step 1: Obtain the data set, culture iPSC in the culture medium, and collect the Raman spectrum of iPSC during the culture process by using a Raman spectrometer; the collection process is divided into multiple time periods, and the spectrum data of multiple cells are collected in each time period;

[0036] Step 2: Data preprocessing, baseline removal, cosmic ray removal, noise reduction and normalization processing are performed on the Raman spectrum data;

[0037] Step 3: Construct and divide the data set, in a time period, the Raman spectrum of the cell sample is divided into 6 stages according to the time sequence, the spectrum data of each stage is a class, a total of six classes of spectrum data set, and the processed spectrum data is randomly divided into a training set, a validation set and a test set;

[0038] Step 4: Construct and train the CNN-LSTM network regression model, the CNN-LSTM network regression model is composed of CNN and LSTM two parts; the CNN layer is used to extract the key spectral features of the Raman fingerprint, the LSTM layer learns the long-term dependence of the time sequence, combines the spatial features and time sequence dependence of the Raman spectrum, and identifies the spectrum of different time periods;

[0039] Step 5: Evaluate the pluripotency of iPSC by using the CNN-LSTM network regression model obtained by training, input the test set reserved in step 3 which does not participate in network training into the trained model, the test set does not participate in training or hyperparameter tuning, and is used for final model performance evaluation, to predict the ability of stem cells to transition between different states. The network evaluates the pluripotency of stem cells, quantifies the pluripotent state as a continuous value, and predicts the specific value of each state by using the regression model.

[0040] Compared with the traditional method for detecting the pluripotency of stem cells, the present application can obtain the biomolecular information of stem cells in different pluripotent states without the destructive and lengthy sample pretreatment process in the traditional detection method, and the spectral characteristic information of the cell population is retained. Therefore, the present application can realize accurate and non-destructive evaluation of the pluripotent state during the differentiation of stem cells; at the same time, the CNN and LSTM modules form a powerful feature extraction network, which can accurately extract local features and time sequence dependencies in complex spectral data; and through multi-layer convolution and pooling operations, the model can more comprehensively perceive the key features and dynamic change information in the spectral data, realizing accurate prediction of the pluripotent state of stem cells. The CNN-LSTM network regression model can retain the continuity information between the states of stem cells through regression prediction of Raman spectra, accurately reflect the dynamic transition process of stem cells, quantitatively characterize the ability gradient of the transition between states, and achieve the effect of accurately evaluating the pluripotent state and transition ability.

[0041] As Figure 2 During the cultivation of iPSCs, the iPSC culture medium was replaced with RPMI1640 culture medium at 1, 3, 6, 12 and 24 hours, respectively, so as to affect the differentiation of iPSCs, as shown in Figure 3 The iPSC cells were subjected to immunofluorescence staining at different time nodes. The characteristics of iPSCs are that they can be self-renewed indefinitely and produce all cell types of organisms. OCT4 and SOX2 are two essential transcription factors that jointly regulate target genes to maintain the pluripotency of iPSCs. The fluorescence intensity values of stem cell markers SOX2 and OCT4 at different time points were detected by immunofluorescence staining technology, and the fluorescence intensity can accurately characterize the pluripotent state, so the actual fluorescence intensity was used as the true value for model training. Olympus fluorescence microscope IX73 was used to take immunofluorescence images of iPSC cells at six time points; ImageJ was used to calculate the average immunofluorescence intensity of the iPSC colony state, and the average fluorescence intensity = the total fluorescence intensity of the colony area / the area of the colony area, and a column chart was drawn for comparison of the fluorescence intensity at different time points.

[0042] As Figure 4 The single-cell average Raman spectrum of the six cell states during the cultivation of stem cells based on RPMI1640 culture medium was drawn. When collecting Raman spectrum data using a Renishaw confocal Raman spectrometer, the laser wavelength was set to 633 nm, the laser power was set to 50%, the L50x objective with a numerical aperture of 0.5 was used, and the Raman spectrum data in the range of 600-1800 cm -1Raman spectra in the wavenumber range; for each time point of the sample, at least 50 cells were selected for signal acquisition, and each cell sample was measured at least 15 times. When pre-processing the obtained data, WiRE 4.3 software was used, including baseline removal, cosmic ray removal, and noise reduction. Before inputting the data set into the network for training, the spectral data was normalized to scale the data values to the range of [0, 1] to eliminate the potential impact of spectral intensity variation on the model results. The processed spectra were divided into six stages (iPSC, 1h, 3h, 6h, 12h, 24h) according to the time sequence of cell processing, and a spectral data set containing six types of cell states was constructed. The Raman spectral data was randomly extracted into a training set, a validation set, and a test set in a ratio of 8:1:1. The random division process used stratified sampling to ensure the balance of the class distribution, and the divided training set, validation set, and test set could be used for model training, hyperparameter tuning, and performance evaluation, respectively.

[0043] As Figure 5 , the CNN-LSTM network regression model is composed of a convolutional neural network (CNN) and a long short-term memory network (LSTM). The pre-processed Raman spectral data is used as input, and the fluorescence intensity values of stem cell markers SOX2 and OCT4 at different time points detected by immunofluorescence staining technology are used as true value output. The local features of the spectral data are extracted by CNN, the input data dimension is the spectral length, and four convolution layers with kernel sizes of 12, 14, 16, and 16 are used, and the output dimensions are 16, 32, 64, and 128, respectively. Batch Normalization (BN) and ReLU activation function are used after each layer to enhance the non-linear expression ability, and Max-Pooling layer is used to gradually reduce the data dimension. The output of the convolution layer is flattened and passed through a bidirectional LSTM layer. The LSTM layer models the 128-dimensional feature sequence in time series, and captures the time series dependency through four layers of bidirectional LSTM. The forward LSTM transmits information from the beginning to the end of the sequence, capturing past information of the sequence, and the backward LSTM transmits information from the end to the beginning of the sequence, capturing future information of the sequence. The number of hidden units is 128, and the output dimension is 256. The outputs of the forward and backward LSTM are combined to obtain the final output. The output of the LSTM is passed through a Dropout layer to prevent overfitting, and then a Fully Connected Layer (FC) is used to map the features to the regression values at the six time points, corresponding to the pluripotency scores of the two markers (SOX2 and OCT4), to realize the evaluation of stem cell pluripotency.

[0044] During the training process, Adam was used as the optimizer to train the model, and the learning rate of Adam was set to 1x10 -5and the mean square error loss is used as the loss function. The CNN-LSTM network regression model is built by PyTorch, and the divided training set and validation set are put into the network for training. At the end of each round, the model performance is evaluated on the validation set, and the training loss (Train Loss), validation loss (Val Loss), root mean square error (RMSE) and coefficient of determination R² are recorded.

[0045] As Figure 6 , the loss of the training and validation data set gradually decreases, and the accuracy steadily improves, and the curves tend to converge, indicating that no overfitting occurs during the training process. Through the fluorescence intensity column charts of (b) and (c) in Figure 3 , we observe that the fluorescence intensity of the biomarkers SOX2 and OCT4 of iPSC decreases with time during the differentiation process, which indicates that the pluripotency of iPSC gradually decreases during the differentiation process. In order to quantitatively analyze the change of fluorescence intensity with time, we use the CNN-LSTM network regression model to predict the fluorescence intensity at different time points. The prediction results of the test set are grouped according to the time points, and the mean of the 6 predicted fluorescence intensity values of the two biomarkers (SOX2 and OCT4) in each category is calculated. Finally, a 6x6 table is output, where each row corresponds to a time point, and the first three columns are the predicted mean of SOX2, and the last three columns are the predicted mean of OCT4. Subsequently, the ORIGIN software is used to calculate the average of the three predicted values at each time point and the corresponding actual fluorescence intensity mean, and based on these data, the line chart is drawn, as shown in Figure 7 and Figure 8In the figure, the gray line represents the true fluorescence intensity values at different time points measured by the immunofluorescence experiment, and the black line is the fluorescence intensity at different time points predicted by the CNN-LSTM network regression model. The coefficient of determination R² represents the degree of correlation between the predicted value and the true value, and the value close to 1 indicates that the predicted value has a high correlation with the true value. The root mean square error RMSE represents the standard deviation of the difference between the predicted value and the true value, and the smaller the value, the higher the accuracy of the prediction. During the training process, the R² and RMSE of the validation set are calculated after each round to monitor the performance of the model, and in the test stage, the coefficient of determination R² and RMSE of the test set are calculated to evaluate the final performance of the model. The CNN-LSTM network regression model achieved a coefficient of determination of 0.93256 and a root mean square error value of 0.13611 in quantitative evaluation, which indicates that the trend of the predicted value and the true value is generally consistent. The CNN-LSTM network regression model can output continuous stem cell pluripotency scores, accurately evaluate the degree of iPSC in a specific pluripotent state, and capture the dynamic changes of iPSC during the differentiation process. The combination of Raman spectroscopy and CNN-LSTM network regression model realizes high-precision and continuous evaluation of iPSC pluripotency, which shows that the CNN-LSTM network regression model overcomes the limitations of traditional classification models and better reflects the changes of stem cell pluripotency, providing a powerful tool for stem cell research.

[0046] The above only describes the preferred embodiments of the present application, and it should be noted that for ordinary skilled in the art, without departing from the principles of the present application, several improvements and refinements can be made, and these improvements and refinements should be considered as the protection scope of the present application. The structures, devices and operation methods not specifically described and explained in the present application, such as without special description and limitation, are implemented according to the conventional means in the art.

Claims

1. A regression prediction method of iPSC pluripotency based on Raman spectroscopy and CNN-LSTM network regression model integration, characterized in that: The method comprises the following steps: Step 1: obtaining a data set, culturing iPSCs in a culture medium, and collecting Raman spectra of the iPSCs in the culture process by using a Raman spectrometer; The collection process is divided into multiple time periods, and the spectral data of multiple cells are collected in each time period; Step 2: data preprocessing, baseline removal, cosmic ray removal, noise reduction and normalization processing are performed on the Raman spectrum data; Step 3: constructing and dividing the data set, in a time period, the Raman spectra of the cell samples are divided into 6 stages in chronological order, the spectral data of each stage is taken as a class, a total of six spectral data sets, and the processed spectral data is randomly divided into a training set, a validation set and a test set; Step 4: constructing and training a CNN-LSTM network regression model, the CNN-LSTM network regression model is composed of CNN and LSTM; the CNN layer is used to extract the key spectral features of the Raman fingerprint, the LSTM layer learns the long-term dependence of the time sequence, combines the spatial features and time sequence dependence of the Raman spectrum, and identifies the spectra in different time periods; In step 4, specifically, the CNN-LSTM network regression model is composed of a convolutional neural network and a long short-term memory network; the preprocessed Raman spectrum data is taken as input, the fluorescence intensity values of stem cell markers SOX2 and OCT4 at different time points are detected by immunofluorescence staining technology as true value output, the local features of the spectral data are extracted by CNN, the input data dimension is the spectral length, and the output dimension is 16, 32, 64 and 128 through four convolution layers with kernel sizes of 12, 14, 16 and 16, respectively; batch normalization and ReLU activation function are used after each layer to enhance the nonlinear expression ability, and the data dimension is gradually reduced through a maximum pooling layer, the output of the convolution layer is flattened and passed through a bidirectional LSTM layer; the LSTM layer models the 128-dimensional feature sequence in time sequence, and captures the time sequence dependence through four bidirectional LSTM layers; The forward LSTM transmits information from the beginning to the end of the sequence, captures the past information of the sequence, and the backward LSTM transmits information from the end to the beginning of the sequence, captures the future information of the sequence; the number of hidden units is 128, and the output dimension is 256; the outputs of the forward and backward LSTM are combined to obtain the final output; the output of the LSTM is subjected to a Dropout layer to prevent overfitting, and then the features are mapped to the regression values of the 6 time points through a fully connected layer, corresponding to the pluripotency scores of the two markers SOX2 and OCT4, to realize the pluripotency evaluation of stem cells; Step 5: evaluating the pluripotency of iPSCs by using the CNN-LSTM network regression model obtained by training, directly inputting the test set reserved in step 3 and not participating in network training into the trained model, the test set does not participate in training or hyperparameter tuning, and is used for final model performance evaluation; the network evaluates the pluripotency of stem cells, quantizes the pluripotent state into a continuous value, and predicts the specific value of each state by using the regression model.

2. The iPSC pluripotency regression prediction method based on Raman spectrum and CNN-LSTM network regression model integration according to claim 1, characterized in that: In step 1, the iPSC culture solution was replaced with RPMI1640 culture solution at 1, 3, 6, 12, and 24 hours, respectively, and the spectra of iPSC cells at five time nodes and the original iPSC cells without replacing the culture solution were collected.

3. The iPSC pluripotency regression prediction method based on Raman spectrum and CNN-LSTM network regression model integration according to claim 2, characterized in that: In step 1, the fluorescence intensity values of stem cell markers SOX2 and OCT4 were detected by immunofluorescence staining technology; the average immunofluorescence intensity of iPSC state was calculated using ImageJ, and the average fluorescence intensity = total fluorescence intensity of colony area / area of colony area.

4. The iPSC pluripotency regression prediction method based on Raman spectrum and CNN-LSTM network regression model integration according to claim 3, characterized in that: In step 2, when the obtained data is preprocessed, it includes baseline removal, cosmic ray removal, noise reduction, and normalization of spectral data before the data set is input to the network for training, scaling the data value to the range of [0, 1].

5. The iPSC pluripotency regression prediction method based on Raman spectrum and CNN-LSTM network regression model integration according to claim 4, characterized in that: In step 3, the Raman spectrum is divided into six stages according to the time sequence of the process, and the Raman spectrum data is randomly extracted into training set, validation set and test set according to the ratio of 8:1:1; the process of random division uses stratified sampling, and the divided training set, validation set and test set are used for model training, hyperparameter tuning and performance evaluation, respectively.

6. The iPSC pluripotency regression prediction method based on Raman spectrum and CNN-LSTM network regression model integration of claim 5, characterized in that: In step 4, during the training of the CNN-LSTM network regression model, Adam is used as the optimizer to train the model, the learning rate of Adam is set to 1x10 -5 and the mean square error loss is used as the loss function; the CNN-LSTM network regression model is built by PyTorch, and the training set and validation set processed in step 3 are put into the CNN-LSTM network regression model for training. At the end of each round, the model performance is evaluated on the validation set, and the training loss, validation loss, root mean square error and determination coefficient R² are recorded.

Images