A method and system for predicting plasmid fermentation temperature change process based on machine learning

Abstract

The present invention relates to a machine learning-based method and system for predicting the temperature-variable process of plasmid fermentation. The method comprises the following steps: preprocessing fermentation data to construct a feature matrix; selecting a prediction model algorithm based on the constructed feature matrix; optimizing hyperparameters of the selected prediction model algorithm; performing characteristic factor analysis on the feature matrix and establishing biological constraints; and, in conjunction with the biological constraints, using the selected prediction model to predict the temperature of the entire fermentation process using temperature as a prediction parameter. Compared with existing technologies, the present invention achieves intelligent plasmid fermentation temperature-variable process, eliminating reliance on manual fermentation experience and avoiding human errors, thereby improving production efficiency and reducing production costs.

Description

Technical Field

[0001] The present invention relates to the field of biological fermentation, and in particular to a method and system for predicting temperature changes in a plasmid fermentation process based on a machine learning model. Background Art

[0002] In recent years, many viral infectious diseases have become unpreventable with traditional vaccines, such as inactivated and attenuated vaccines. The advent of DNA vaccines has improved this situation. Research on nucleic acid vaccines has become a hot topic in vaccine development over the past decade. Their ability to elicit both humoral and cellular immune responses, while also exhibiting advantages in productivity, stability, and storage, holds far-reaching significance. As research on nucleic acid vaccines continues to deepen, plasmid DNA has also gained attention as a carrier for these vaccines.

[0003] Unlike protein expression and amino acid fermentation, plasmid DNA fermentation yields do not increase gradually over time, but rather fluctuate with the physiological state of the host bacteria. Plasmids are intracellular expression products, and their yield and quality are influenced by both the microbial metabolic state and the macroscopic fermentation environment. If metabolic changes caused by environmental stimuli lead to plasmid loss or replication blockage, thus affecting plasmid production efficiency. Therefore, to stabilize the physiological state of the host bacteria, plasmid fermentation processes require greater precision. Currently, specific growth rate and temperature are considered key factors influencing plasmid yield. Cultivation at a lower specific growth rate can yield higher plasmid yields. However, a higher specific growth rate results in excessive cell division, leaving free plasmids with insufficient time to replicate, resulting in lower plasmid content per cell or even plasmid loss. However, plasmid production through controlled specific growth rate imposes stringent nutrient requirements: insufficient nutrient intake leads to decreased metabolic activity, while excessive nutrient intake leads to increased metabolic activity. Currently, most plasmid fermentations utilize a variable temperature process to achieve high plasmid yields, with low temperatures used to accumulate biomass and high temperatures used to induce plasmid synthesis. However, the speed of the temperature ramp-up determines the stability of plasmid synthesis and the formation of its supercoiled structure. Only plasmids with supercoiled structures can achieve effective therapeutic effects. Therefore, precise temperature control during the plasmid fermentation process is crucial for high plasmid yield and quality.

[0004] The application of machine learning in biology has made intelligent fermentation possible, reducing the reliance on human experience in fermentation process control. To date, no method has been reported for predicting the process of variable-temperature fermentation using machine learning. If machine learning is combined with biology and existing fermentation data is used as a dataset, a reasonable process model for variable-temperature fermentation could be predicted, enabling machines to replace manual labor and move towards intelligent fermentation. This would avoid losses caused by human error and significantly reduce production costs. Summary of the Invention

[0005] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and provide a method and system for predicting the temperature change process of plasmid fermentation based on machine learning.

[0006] The purpose of the present invention can be achieved by the following technical solutions:

[0007] As a first aspect of the present invention, a method for predicting a plasmid fermentation temperature change process based on machine learning is provided, characterized in that the method comprises the following steps:

[0008] The fermentation data were preprocessed to construct the feature matrix;

[0009] Select the prediction model algorithm based on the constructed feature matrix;

[0010] Optimize hyperparameters of the selected prediction model algorithm;

[0011] Performing eigenfactor analysis on the eigenvalue matrix and establishing biological constraints;

[0012] Combined with biological constraints, the selected prediction model was used to predict the temperature of the entire plasmid fermentation process using temperature as a prediction parameter.

[0013] Furthermore, the preprocessing includes screening, cleaning, serialization and integration of data; the specific steps of constructing the feature matrix include:

[0014] Collect fermentation data of the temperature change process to be predicted;

[0015] Select the required characteristic parameters in combination with biological analysis;

[0016] Fitting the fermentation kinetics model to the selected characteristic parameters to make the characteristic parameters continuous and remove the errors caused by measurement;

[0017] Integrate relevant parameters and construct a feature matrix.

[0018] Furthermore, the fermentation data includes biological characteristic parameters: temperature, time, biomass, cumulative substrate consumption, specific growth rate, ventilation flow, volume, stirring, oxygen consumption rate, carbon dioxide release rate, respiratory quotient, unit cell carbon dioxide release rate, unit cell oxygen consumption rate and unit cell plasmid yield.

[0019] Furthermore, the prediction model algorithm selection specific steps include:

[0020] Applying a variety of machine learning algorithms to model the matrix as a data set to predict the warming pattern;

[0021] Each machine learning algorithm model performs K-Fold cross validation under the default model parameters;

[0022] R 2 The quality of the models was evaluated as the result score, and the model with the highest quality was selected as the prediction model.

[0023] Furthermore, the machine learning algorithm includes:

[0024] Gradient boosting regression, partial least squares regression, random forest, and SVM with three different kernel functions: support vector machine-linear kernel; support vector machine-polynomial kernel and support vector machine radial-basis kernel function.

[0025] Furthermore, the characteristic factor analysis uses Pearson correlation coefficient analysis to establish biological constraints based on the normal physiological metabolic level of the bacteria.

[0026] Furthermore, after the prediction is completed, the obtained prediction model will be verified, and the fermentation data of the verification experiment will continue to be used as a sample set for further training.

[0027] As a second aspect of the present invention, a system for predicting a plasmid fermentation temperature-varying process based on machine learning is provided, wherein the system applies the plasmid fermentation temperature-varying process prediction method described above to predict the temperature-varying process, comprising:

[0028] The fermentation data preprocessing module is used to collect, screen, and clean fermentation data, use fermentation kinetics model fitting to make the data continuous, and integrate the data to construct a feature matrix;

[0029] The prediction model algorithm selection module uses a variety of machine learning algorithms to build models based on the feature matrix, evaluates the quality of the models under default parameters, and selects the model with the highest quality as the prediction model;

[0030] Hyperparameter optimization module, used to optimize the hyperparameters of the selected prediction model;

[0031] A characteristic factor analysis module, used for analyzing the characteristic factors of the characteristic matrix using the Pearson correlation coefficient and establishing biological constraints;

[0032] The temperature prediction module is used to combine biological constraints and use the selected prediction model to take temperature as a prediction parameter to predict the temperature of the entire fermentation process.

[0033] As a third aspect of the present invention, there is provided an electronic device, comprising:

[0034] one or more processors;

[0035] a memory having one or more programs stored thereon,

[0036] When the one or more programs are executed by the one or more processors, the one or more processors implement the plasmid fermentation temperature change process prediction method as described above.

[0037] As a fourth aspect of the present invention, there is provided a computer-readable medium having a computer program stored thereon.

[0038] When the program is executed by a processor, the steps of the method for predicting the temperature-varying process of plasmid fermentation are implemented as described above.

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] (1) This invention combines machine learning with biofermentation, using fermentation data from high-yield plasmid batches as training samples. The method analyzes the correlation between each characteristic parameter and accurately predicts the fermentation process temperature using a supervised learning model. This method intelligently controls the temperature variation of plasmid fermentation, eliminating reliance on manual fermentation experience while avoiding human errors, improving production efficiency, and reducing production costs.

[0041] (2) The present invention fits the fermentation kinetics model to the fermentation data. On the one hand, it can make the discrete fermentation data continuous and supplement the default values ​​in accordance with biology. On the other hand, it can analyze the relationship between process parameters and the physiological state of the bacteria in more depth through the calculation of kinetic equations.

[0042] (3) The present invention optimizes the hyperparameters of the selected model, which can improve the robustness of the model for low unknown samples. By adjusting the parameters, the quality of the prediction model can be further improved, and the prediction of unknown plasmid yield can be achieved.

[0043] (4) The present invention continues to use the fermentation data of subsequent verification experiments as a sample set for training and model building, further improving the quality of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 Schematic diagram of a method for predicting a plasmid fermentation temperature change process based on machine learning according to the present invention;

[0045] Figure 2 Schematic diagram of fitting of three kinetic equations;

[0046] Figure 3Schematic diagram of 5-fold cross-validation verification results of 6 machine learning algorithms under default parameters in an embodiment of the present invention; wherein: (A) 5-fold cross-validation results of 6 algorithms under default parameters, (B) change of GBR algorithm score with increasing sample amount, (C) change of PLS ​​algorithm score with increasing sample amount, (D) change of RF algorithm score with increasing sample amount, (E) change of SVM_Linear algorithm score with increasing sample amount, (F) change of SVM_Poly algorithm score with increasing sample amount, (G) change of SVM_RBF algorithm score with increasing sample amount;

[0047] Figure 4 The heat map of correlation coefficients between characteristic parameters in the embodiment of the present invention; wherein: the Pearson correlation coefficient on the left is Y P / X The correlation between (unit bacterial plasmid yield) and characteristic factors. The horizontal axis represents the factor value under different samples, and the color is the value of the normalized factor. The right figure shows the correlation between the color and the factors.

[0048] Figure 5 Schematic diagram of the contribution of characteristic parameters to the model in an embodiment of the present invention;

[0049] Figure 6 Schematic diagram of temperature prediction and experimental verification results in an embodiment of the present invention. DETAILED DESCRIPTION

[0050] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0051] Example 1

[0052] like Figure 1 FIG. 1 is a specific embodiment of a method for predicting a plasmid fermentation temperature change process based on machine learning according to the present invention, and the specific steps include:

[0053] (1) The fermentation data were preprocessed by screening, cleaning, serialization, and integration to construct a feature matrix.

[0054] Some fermentation parameters may not be the key to affecting yield, and participating in modeling may reduce the quality of the model, so it is necessary to screen the characteristic parameters reasonably. Some characteristic parameters are not continuous functions, and there may be errors caused by human operation during experimental operation, so equation fitting is required to make them continuous and data cleaning is required. For kinetic equation fitting, please refer to Figure 2By fitting the kinetic model, on the one hand, discrete fermentation data can be processed continuously and the default values ​​can be supplemented in a biologically consistent manner; on the other hand, the calculation of the kinetic equation can provide a more in-depth analysis of the relationship between process parameters and the physiological state of the bacteria. The specific steps of this step are as follows:

[0055] (1.1) Collect fermentation data of the temperature-changing process to be predicted;

[0056] The parameters involved in this experiment are:

[0057] Conventional parameters: pH, dissolved oxygen, aeration rate (F), stirring rate (Agit), temperature (Temp), tank pressure (Pa), volume (V);

[0058] Physiological parameters: oxygen uptake rate (OUR), carbon dioxide evolution rate (CER), respiratory quotient (RQ);

[0059] Discrete parameters: specific growth rate (μ), oxygen consumption rate per unit cell Carbon dioxide release rate per unit cell Cumulative substrate consumption (S), bacterial biomass (X), and plasmid production per unit cell (Y P / X ).

[0060] (1.2) Select the required characteristic parameters in combination with biological analysis;

[0061] The process parameters were screened: the fluctuations of dissolved oxygen and RQ processes significantly affected the model quality, so they were removed and not used as characteristic parameters for modeling.

[0062] (1.3) The selected characteristic parameters may have discrete and error parameters, and equation fitting is required to continuously remove the errors caused by human measurement in order to construct the characteristic matrix.

[0063] Figure 2 The fermentation kinetics expression is as follows:

[0064] Bacterial growth dynamics:

[0065]

[0066] Where: X0 is the initial biomass (OD 600 ); t is time (h). μ m is the maximum specific growth rate (h -1 );X m The maximum biomass (OD 600 );

[0067] Cumulative substrate consumption kinetics:

[0068] Batch fermentation stage:

[0069]

[0070] Fed-batch fermentation stage:

[0071] S (t) =kt+b (3)

[0072] Where: S represents the cumulative consumption of substrate (g); k represents the substrate consumption rate (g / h); b is a coefficient added to better fit the substrate consumption rate in the later stage.

[0073] Synthesis kinetics of unit plasmid product:

[0074]

[0075] Where: Y P / X The yield of plasmid per unit cell (mg / L / OD 600 ); P0 is the initial plasmid yield (mg / L); α represents a parameter related to bacterial growth and product synthesis; β represents a parameter independent of product synthesis. When α≠0 and β=0, product synthesis is growth-coupled; when α=0 and β≠0, product synthesis is non-growth-coupled; and when α≠0 and β≠0, product synthesis is partially growth-coupled. Because P0 corresponds to a small biomass, minimal plasmid yield, and significant instrumental error, P0 is ignored in subsequent product synthesis kinetics.

[0076] By fitting the fermentation kinetics, the process parameters (temperature (Temp), time (t), biomass (X), cumulative substrate consumption (S), specific growth rate (μ), ventilation flow (F), volume (V), pH, oxygen consumption rate (OUR), carbon dioxide release rate (CER), agitation (Agit), unit cell carbon dioxide release rate) were obtained. Oxygen consumption rate per unit cell Plasmid yield per cell (Y P / X )) are relevant data, which are integrated as characteristic parameters to establish the fermentation process parameters (t, X, S, μ, Temperature, Flow rate, Volume, pH, OUR, CER, Agitation rate, Y P / X )’s model matrix.

[0077] (2) Select the prediction model algorithm for the matrix.

[0078] In this example, 2500 samples of fermentation processes with different temperature increase strategies were selected to establish a data set. Six different machine learning methods were used to model the selected fermentation process characteristic parameters to predict the temperature increase method. 2Model quality was evaluated using the resulting score. Methods included gradient boosting regression (GBR), partial least squares regression (PLS), random forest (RF), and three different SVM kernel functions: support vector machine-linear kernel (SVM-Linear); support vector machine-polynomial kernel (SVM-Polynomial, Ploy); and support vector machine-radial basis function (SVM-RBF). A dataset of 2,500 samples from seven fermentation batches with different temperature ramping strategies was established. Six algorithms were subjected to 5-fold cross-validation using the default model parameters (Table 1).

[0079] Figure 3 Results A show that RF and GBR have higher scores than SVM and PLS. Learning curves were plotted to assess model overfitting and underfitting under different data sizes. Learning curve scores were compared (5-fold cross-validation) to assess model overfitting and underfitting under different data sizes. Comparing learning curve scores revealed that all model scores increased with increasing sample size, with the SVM model showing underfitting. When the models were constructed with the maximum sample size, no overfitting occurred. Figure 3 Results B-G show that among these machine learning algorithms, the RF model that predicts unit yield using fermentation parameters achieved the highest score (training sample = 2000, training set R 2 =0.99986, test set R 2 =0.99736).

[0080] Table 1 Default hyperparameters of the algorithms

[0081]

[0082] (3) Optimizing the hyperparameters of the matrix model algorithm.

[0083] Hyperparameter optimization can improve the robustness of the model to unknown samples. Therefore, by adjusting the parameters, the quality of the RF can be further improved, and the prediction of unknown plasmid yield can be achieved. In this embodiment, the adjustment of the parameters is guided by judging the complexity of the model. The main hyperparameters are optimized by drawing a learning curve: n_estimators, max_depth min_sample_leaf, min_sample_split, and max_features. The model score training set R is modeled on a 2500 sample data set under the default parameters (before optimization). 2 =0.99976, test set R2 =0.99879, computing time 6.68e-05 seconds.

[0084] The hyperparameters are adjusted under random_state=1 and n_estimators are adjusted in the range of 0 to 200. The optimal parameter n_estimators=80 and the test set R 2 =0.9996. max_depth (default value is the maximum value) has been adjusted to a range of 1 to 50. This value is positively correlated with model complexity. When the optimal max_depth is 19, the test set R 2 =0.99811, the complexity moves towards the simple direction. After adjusting the model parameters (min_samples_split and min_samples_leaf) in the simple direction, it is found that R 2 Decreases, indicating that the model in the simple direction has moved to the limit. Therefore, these parameters are skipped. The larger the max_features, the more complex the model (default value = 3). When max_features = 9, the best test model R 2 =0.99883 At this point, the model has reached the limit of parameter optimization. The optimal parameters are as follows: when random_state=1, n_estimators=80, max_depth=19, max_features=8. The optimized model scores the training set R 2 =0.99983, test set R 2 =0.99933, computing time 2.07e-05 seconds, after parameter optimization, the training set score increased by 0.07%, the validation set score increased by 0.05%, and the computing efficiency increased by 310%.

[0085] (4) Perform characteristic factor analysis and biological constraints on the matrix.

[0086] In this example, the relative time (t), temperature (Temp), ventilation flow (F), pH (pH), volume (V), tank pressure (Pa), biomass (X), oxygen consumption rate (OUR), carbon dioxide release rate (CER), agitation (Agit), unit cell oxygen consumption rate (q O2 ), carbon dioxide release rate per unit cell (q CO2 ), cumulative substrate consumption (S) and specific growth rate (μ) as factors affecting the plasmid yield per unit cell (Y P / X )'s characteristic factors and the heat map of the data set and correlation coefficient of the 14 characteristic factors and unit plasmid yield were plotted. Figure 4The trend of the factors involved in the model is significantly correlated with the unit plasmid yield. Through Pearson correlation coefficient analysis, the default correlation coefficient is 1 when the plasmid yield is the highest. At this time, the correlation between each characteristic parameter and the plasmid yield is observed. The factors with strong correlation (|R| ≥ 0.8) are t, X, μ, S and Temp, and the factors with moderate correlation (0.5 ≤ |R| < 0.8) are F, Agit, CER, OUR, Pa, q O2 and q CO2 , the uncorrelated factors (|R| < 0.3) were pH and V. The weak collinearity among the factors indicated that these factors were representative. Figure 5 The contribution of the characteristic factors is shown by the importance of the function characteristic factors. Among them, Temp, X, and μ (|R| ≥ 0.8) factors contribute significantly to the model. It is important to note that OUR, as a horizontal physiological parameter of bacterial metabolism, shows a relatively small contribution, which is inconsistent with biological characteristics. Reasonable biological constraints can improve the robustness of the model, so further biological constraints are needed for the prediction model. The biological constraint equation is as follows:

[0087]

[0088] Constraints:

[0089]

[0090] (5) This example combines biological constraints, integrating temperature rise with biological mechanisms to establish a relationship between conventional control parameters and physiological parameters. The model uses temperature as the characteristic parameter to be predicted to predict the temperature of the entire fermentation process. The established prediction model is verified, and subsequent fermentation data is used as a sample set for further training. Figure 6 The temperature change process predicted by the model was demonstrated. The fermentation process was verified based on the temperature change process predicted by machine learning. The model was evaluated by comparing the predicted value of plasmid yield with the actual value, and the score R 2 =0.9677, RMSE =0.4535. Higher R 2 This shows that the model has high reliability.

[0091] Example 2

[0092] As a second aspect of the present invention, a plasmid fermentation temperature-varying process prediction system based on machine learning is also provided. The system uses the plasmid fermentation temperature-varying process prediction method described in the above embodiment to predict the temperature-varying process, and comprises:

[0093] The fermentation data preprocessing module is used to collect, screen, and clean fermentation data, use fermentation kinetics model fitting to make the data continuous, and integrate the data to construct a feature matrix;

[0094] The prediction model algorithm selection module uses a variety of machine learning algorithms to build models based on the feature matrix, evaluates the quality of the models under default parameters, and selects the model with the highest quality as the prediction model;

[0095] Hyperparameter optimization module, used to optimize the hyperparameters of the selected prediction model;

[0096] A characteristic factor analysis module, used for analyzing the characteristic factors of the characteristic matrix using the Pearson correlation coefficient and establishing biological constraints;

[0097] The temperature prediction module is used to combine biological constraints and use the selected prediction model to take temperature as a prediction parameter to predict the temperature of the entire fermentation process.

[0098] Example 3

[0099] As a third aspect of the present invention, the present application further provides an electronic device comprising: one or more processors; a memory for storing one or more programs; and when the one or more programs are executed by the one or more processors, the one or more processors implement the above-mentioned method for predicting the temperature-varying process of plasmid fermentation. In addition to the processor, memory, and interface described above, any device with data processing capabilities in which the method for predicting the temperature-varying process of plasmid fermentation provided in the embodiments of the present invention is located may also include other hardware, typically based on the actual functions of the device, which will not be described in detail.

[0100] Example 4

[0101] As a fourth aspect of the present invention, the present application also provides a computer-readable storage medium having computer instructions stored thereon, which, when executed by a processor, implement the plasmid fermentation temperature-changing process prediction method as described in the above embodiments. The computer-readable storage medium can be the internal storage unit of any device with data processing capabilities described in any of the aforementioned embodiments, such as a hard disk or memory. The computer-readable storage medium can also be an external storage device, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), an SD card, a flash card (FlashCard), etc. equipped on the device. Further, the computer-readable storage medium can also include both the internal storage unit of any device with data processing capabilities and an external storage device. The computer-readable storage medium is used to store the computer program and other programs and data required by any device with data processing capabilities, and can also be used to temporarily store data that has been output or is to be output.

[0102] The above describes in detail the preferred embodiments of the present invention. It should be understood that those skilled in the art can make numerous modifications and variations based on the concepts of the present invention without inventive effort. Therefore, any technical solutions that can be derived by those skilled in the art through logical analysis, reasoning, or limited experimentation based on the concepts of the present invention and the prior art should be within the scope of protection defined by the claims.

Claims

1. A method for predicting the temperature change process of plasmid fermentation based on machine learning, characterized in that: The method comprises the following steps: The fermentation data were preprocessed to construct the feature matrix; Select the prediction model algorithm based on the constructed feature matrix; Optimize hyperparameters of the selected prediction model algorithm; Performing eigenfactor analysis on the eigenvalue matrix and establishing biological constraints; Combined with biological constraints, the selected prediction model is used to predict the temperature of the entire plasmid fermentation process using temperature as a prediction parameter; The characteristic factor analysis adopts Pearson correlation coefficient analysis, and establishes biological constraints according to the normal physiological metabolic level of the bacteria.

2. The method for predicting a plasmid fermentation temperature change process based on machine learning according to claim 1, characterized in that: The preprocessing includes screening, cleaning, serialization and integration of data; The specific steps of constructing the feature matrix include: Collect fermentation data of the temperature change process to be predicted; Select the required characteristic parameters in combination with biological analysis; Fitting the fermentation kinetics model to the selected characteristic parameters to make the characteristic parameters continuous and remove the errors caused by measurement; Integrate relevant parameters and construct a feature matrix.

3. The method for predicting plasmid fermentation temperature change process based on machine learning according to claim 1, characterized in that: The fermentation data include biological characteristic parameters: temperature, time, biomass, cumulative substrate consumption, specific growth rate, ventilation flow, volume, stirring, oxygen consumption rate, carbon dioxide release rate, respiratory quotient, carbon dioxide release rate per unit cell, oxygen consumption rate per unit cell and plasmid yield per unit cell.

4. The method for predicting a plasmid fermentation temperature change process based on machine learning according to claim 1, wherein: The specific steps of selecting the prediction model algorithm include: Applying a variety of machine learning algorithms to model the matrix as a data set to predict the warming pattern; Each machine learning algorithm model performs K-Fold cross validation under the default model parameters; R 2 The quality of the models was evaluated as the result score, and the model with the highest quality was selected as the prediction model.

5. The method for predicting plasmid fermentation temperature change process based on machine learning according to claim 4, characterized in that: The machine learning algorithm includes: Gradient boosting regression, partial least squares regression, random forest, and SVM with three different kernel functions: support vector machine-linear kernel; support vector machine-polynomial kernel and support vector machine radial-basis kernel function.

6. The method for predicting plasmid fermentation temperature change process based on machine learning according to claim 1, characterized in that: The obtained prediction model was verified, and the fermentation data of the verification experiment was continued as the sample set for training.

7. A plasmid fermentation temperature change process prediction system based on machine learning, characterized in that: The system uses the plasmid fermentation temperature change process prediction method according to any one of claims 1 to 6 to predict the temperature change process, including: The fermentation data preprocessing module is used to collect, screen, and clean fermentation data, use fermentation kinetics model fitting to make the data continuous, and integrate the data to construct a feature matrix; The prediction model algorithm selection module uses a variety of machine learning algorithms to build models based on the feature matrix, evaluates the quality of the models under default parameters, and selects the model with the highest quality as the prediction model; Hyperparameter optimization module, used to optimize the hyperparameters of the selected prediction model; A characteristic factor analysis module, used for analyzing the characteristic factors of the characteristic matrix using the Pearson correlation coefficient and establishing biological constraints; The temperature prediction module is used to combine biological constraints and use the selected prediction model to take temperature as a prediction parameter to predict the temperature of the entire fermentation process.

8. An electronic device, characterized in that: include: one or more processors; a memory having one or more programs stored therein; When the one or more programs are executed by the one or more processors, the one or more processors implement the method for predicting a plasmid fermentation temperature-varying process as described in any one of claims 1 to 6.

9. A computer-readable medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method for predicting the temperature-varying process of plasmid fermentation as described in any one of claims 1 to 6 are implemented.

Images