A method, device and medium for predicting transfection efficiency of an mRNA vaccine

Abstract

The application relates to the technical field of machine learning, and discloses a prediction method for mRNA vaccine transfection efficiency, which comprises the following steps: acquiring to-be-tested vaccine formula data and prediction requirements; determining a corresponding prediction model according to the prediction requirements; wherein the prediction model is a relationship model between the vaccine formula data and the transfection efficiency, which is constructed based on any one of a kernel ridge regression model, a linear regression model and a lightGBM model; and the to-be-tested vaccine formula data is subjected to transfection efficiency prediction by using the determined prediction model. The application can quantize the transfection efficiency of the vaccine formula by predicting the transfection efficiency of the vaccine, thereby providing support for researchers to optimize the design of mRNA vaccines.

Description

Technical Field

[0001] This application relates to the field of machine learning technology, and in particular to a method, apparatus and medium for predicting the transfection efficiency of mRNA vaccines. Background Technology

[0002] Nanoliposomes (LNPs) are a common delivery system for mRNA vaccines. The design and optimization of this delivery system depends primarily on the selection and design of ionizable lipids, as well as other factors such as experimental conditions.

[0003] Currently, literature review indicates that the aforementioned factors have a certain impact on mRNA vaccine transfection efficiency. However, there is still no technical solution for predicting mRNA vaccine transfection efficiency.

[0004] Therefore, how to predict the transfection efficiency of mRNA vaccines is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] The purpose of this application is to provide a method, device, and medium for predicting the transfection efficiency of mRNA vaccines, to predict the transfection efficiency of mRNA vaccines, and to quantify the transfection efficiency of vaccine formulations, thereby providing support for researchers to optimize the design of mRNA vaccines.

[0006] To address the aforementioned technical problems, this application provides a method for predicting the transfection efficiency of mRNA vaccines, comprising:

[0007] Obtain vaccine formulation data and predict demand;

[0008] The corresponding prediction model is determined based on the prediction requirements; wherein, the prediction model is a relationship model between vaccine formulation data and transfection efficiency constructed based on any one of the following models: nuclear ridge regression model, linear regression model, and lightGBM model.

[0009] The transfection efficiency of the vaccine formulation data to be tested is predicted using a defined prediction model.

[0010] Preferably, it further includes:

[0011] Retrieve multiple vaccine formula data from the database using preset keywords;

[0012] Multiple vaccine formulation data are filtered according to preset standards to obtain multiple target vaccine formulation data;

[0013] The antibody titer corresponding to each target vaccine formulation data is determined as a prediction object, and the characteristic variables in the target vaccine formulation data are determined; wherein, the characteristic variables include mRNA characteristics, experimental conditions, ionizable lipid characteristics, and liposome formulation characteristics; the antibody titer is used to measure vaccine transfection efficiency;

[0014] The prediction model is constructed based on the feature variables and the prediction object.

[0015] Preferably, before constructing the prediction model based on the feature variables and the prediction object, the method further includes:

[0016] Identify the continuous variables among the feature variables;

[0017] Data standardization is performed on the continuous variables and the prediction objects respectively.

[0018] Preferably, constructing the prediction model based on the feature variables and the prediction object includes:

[0019] Map the feature variables and the prediction object to a nonlinear space;

[0020] In the nonlinear space, a nonlinear function is obtained by fitting the feature variables and the predicted object using a fifth-order polynomial kernel;

[0021] The target parameters of the nonlinear function are determined based on the minimum loss function of the kernel ridge regression model to obtain the prediction model.

[0022] Preferably, constructing the prediction model based on the feature variables and the prediction object includes:

[0023] The target feature variables are obtained by screening the feature variables using a lasso regression model;

[0024] The linear regression model is used to perform linear regression fitting on the target feature variable and the prediction object corresponding to the target feature variable, and the parameters corresponding to each feature variable are determined by the minimum loss function of the linear regression model to obtain the prediction model.

[0025] Preferably, constructing the prediction model based on the feature variables and the prediction object includes:

[0026] By randomly selecting several feature variables at each bifurcation point of the lightGBM model and iterating through them to obtain their values, the optimal split point is determined.

[0027] The importance ranking of the feature variables is determined based on the number of times the feature variables are selected.

[0028] The prediction model is constructed based on the importance ranking of the feature variables and the optimal split point.

[0029] Preferably, after obtaining the prediction model, the method further includes:

[0030] The root mean square error and coefficient of determination are calculated based on the predicted and actual values ​​of the prediction model.

[0031] The accuracy of the prediction model is determined based on the root mean square error and / or the coefficient of determination.

[0032] Preferably, after obtaining the prediction model, the method further includes:

[0033] The prediction results of the prediction model were verified using wet experiments.

[0034] This application also provides a device for predicting mRNA vaccine transfection efficiency, comprising:

[0035] The acquisition module is used to acquire vaccine formula data and predict demand.

[0036] The determination module is used to determine the corresponding prediction model according to the prediction requirements; wherein, the prediction model is a relationship model between vaccine formulation data and transfection efficiency constructed based on any one of the following models: nuclear ridge regression model, linear regression model, and lightGBM model.

[0037] The prediction module is used to predict the transfection efficiency of the vaccine formulation data to be tested using a defined prediction model.

[0038] This application also provides a device for predicting the transfection efficiency of an mRNA vaccine, including a memory for storing a computer program;

[0039] A processor for executing the computer program to implement the steps of the method for predicting the transfection efficiency of the mRNA vaccine.

[0040] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the method for predicting mRNA vaccine transfection efficiency.

[0041] This application provides a method for predicting the transfection efficiency of mRNA vaccines, comprising: acquiring vaccine formulation data and prediction requirements; determining a corresponding prediction model based on the prediction requirements; wherein the prediction model is a relationship model between vaccine formulation data and transfection efficiency constructed based on any one of the following models: cerebellopontine regression model, linear regression model, and lightGBM model; and using the determined prediction model to predict the transfection efficiency of the vaccine formulation data. This application predicts vaccine transfection efficiency through modeling, which can quantify the transfection efficiency of vaccine formulations and provide support for researchers in the optimized design of mRNA vaccines.

[0042] The device, medium, and method for predicting mRNA vaccine transfection efficiency provided in this application correspond to the above-mentioned results. Attached Figure Description

[0043] To more clearly illustrate the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0044] Figure 1 A flowchart illustrating a method for predicting mRNA vaccine transfection efficiency provided in this application embodiment;

[0045] Figure 2 An importance ranking chart of feature variables provided in an embodiment of this application;

[0046] Figure 3 A distribution diagram of the absolute value of the difference in actual transfection efficiency between any two formulations in a test set, provided for an embodiment of this application;

[0047] Figure 4 A comparative image showing the transfection of SM-102 and ALC-0315 ionizable lipids in 293T cells for 24 hours, as provided in the embodiments of this application.

[0048] Figure 5 A comparative image showing the transfection of SM-102 and ALC-0315 ionizable lipids in 293T cells for 48 hours, as provided in the embodiments of this application.

[0049] Figure 6 A comparative image of SM-102 and ALC-0315 ionizable lipids transfected in Hepa cells for 24 hours, provided for embodiments of this application;

[0050] Figure 7 A comparative image of SM-102 and ALC-0315 ionizable lipids transfected in Hepa cells for 48 hours, provided for an embodiment of this application;

[0051] Figure 8 A comparative image of DlinDMA and ALC-0315 ionizable lipids transfected in 293T cells for 24 hours, provided for an embodiment of this application;

[0052] Figure 9 A comparative image of DlinDMA and ALC-0315 ionizable lipids transfected in 293T cells for 48 hours, provided for an embodiment of this application;

[0053] Figure 10 A comparative image of DlinDMA and ALC-0315 ionizable lipids transfected in Hepa cells for 24 hours, provided for an embodiment of this application;

[0054] Figure 11 A comparative image of DlinDMA and ALC-0315 ionizable lipids transfected in Hepa cells for 48 hours, provided for an embodiment of this application;

[0055] Figure 12 A comparative image showing the transfection of SM-102 and DlinDMA ionizable lipids in 293T cells for 6 hours, as provided in an embodiment of this application.

[0056] Figure 13 A comparative image of SM-102 and DlinDMA ionizable lipids transfected in 293T cells for 12 hours, provided for an embodiment of this application;

[0057] Figure 14 A comparative image of SM-102 and DlinDMA ionizable lipids transfected in 293T cells for 24 hours, provided for an embodiment of this application;

[0058] Figure 15 A comparative image of SM-102 and DlinDMA ionizable lipids transfected in 293T cells for 48 hours, provided for an embodiment of this application;

[0059] Figure 16 A comparative image of SM-102 and DlinDMA ionizable lipids transfected in Hepa cells for 6 hours, provided as an embodiment of this application;

[0060] Figure 17 A comparative image of SM-102 and DlinDMA ionizable lipids transfected in Hepa cells for 12 hours, provided as an embodiment of this application;

[0061] Figure 18 A comparative image of SM-102 and DlinDMA ionizable lipids transfected in Hepa cells for 24 hours, provided as an embodiment of this application;

[0062] Figure 19A comparative image of SM-102 and DlinDMA ionizable lipids transfected in Hepa cells for 48 hours, provided as an embodiment of this application;

[0063] Figure 20 A structural diagram of a device for predicting mRNA vaccine transfection efficiency provided in an embodiment of this application;

[0064] Figure 21 A structural diagram of another mRNA vaccine transfection efficiency prediction device provided in an embodiment of this application. Detailed Implementation

[0065] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the protection scope of this application.

[0066] The core of this application is to provide a method, device, and medium for predicting the transfection efficiency of mRNA vaccines, which can be used to predict the transfection efficiency of mRNA vaccines and quantify the transfection efficiency of vaccine formulations, providing support for researchers to optimize the design of mRNA vaccines.

[0067] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0068] Figure 1 A flowchart illustrating a method for predicting mRNA vaccine transfection efficiency provided in this application embodiment is shown below. Figure 1 As shown, methods for predicting mRNA vaccine transfection efficiency include:

[0069] S10: Obtain vaccine formulation data and predict demand.

[0070] S11: Determine the corresponding prediction model based on the prediction requirements; wherein, the prediction model is a relationship model between vaccine formulation data and transfection efficiency constructed based on any one of the following models: nuclear ridge regression model, linear regression model, and lightGBM model.

[0071] S12: Predict transfection efficiency using a defined prediction model based on the vaccine formulation data to be tested.

[0072] In step S10, the vaccine formulation data includes mRNA characteristics, experimental conditions, ionizable lipid characteristics, and liposome formulation characteristics. The prediction requirements, in addition to predicting vaccine transfection efficiency, also include understanding the impact of each characteristic variable in the vaccine formulation data on transfection efficiency, and ranking the importance of each characteristic variable.

[0073] In step S11, the prediction model includes a relationship model between vaccine formulation data and transfection efficiency constructed based on any one of the following models: ridge-nuclear regression model, linear regression model, and lightGBM model. While the linear regression model performs slightly worse in predicting vaccine transfection efficiency compared to the ridge-nuclear regression model and the lightGBM model, it can describe the trend of transfection efficiency as the characteristic variables in the vaccine formulation data change within a specific range, thus possessing good interpretability. The lightGBM model can describe the importance of the characteristic variables in the vaccine formulation data.

[0074] In step S12, the characteristic variables in the vaccine formulation data are specifically input into the determined prediction model, and the corresponding transfection efficiency can be output. In this embodiment, antibody titer (IgG) is used as a measure of vaccine transfection efficiency.

[0075] This application provides a method for predicting the transfection efficiency of mRNA vaccines, comprising: acquiring vaccine formulation data and prediction requirements; determining a corresponding prediction model based on the prediction requirements; wherein the prediction model is a relationship model between vaccine formulation data and transfection efficiency constructed based on any one of the following models: cerebellopontine regression model, linear regression model, and lightGBM model; and using the determined prediction model to predict the transfection efficiency of the vaccine formulation data. This application predicts vaccine transfection efficiency through a model, which can quantify the transfection efficiency of vaccine formulations and provide support for researchers in the optimized design of mRNA vaccines.

[0076] Based on the above embodiments, this application embodiment obtains multiple vaccine formulation data from a database using preset keywords; filters the multiple vaccine formulation data according to preset standards to obtain multiple target vaccine formulation data; determines the antibody titer corresponding to each target vaccine formulation data as a prediction object, and determines the feature variables in the target vaccine formulation data; wherein, the feature variables include mRNA characteristics, experimental conditions, ionizable lipid characteristics, and liposome formulation characteristics; antibody titer is used to measure vaccine transfection efficiency; and constructs the prediction model based on the feature variables and the prediction object. Before constructing the prediction model based on the feature variables and the prediction object, the method further includes: determining the continuous variables among the feature variables; and performing data standardization processing on the continuous variables and the prediction object respectively. For ease of understanding, this is described in detail below.

[0077] Using "mRNA," "vaccine," and "virus" as preset keywords, we searched the Web of Science website for literature related to mRNA vaccine formulations (mRNA-LNP formulations) using nanoliposomes as a delivery system. To control for variables, we used preset criteria to screen and record vaccine formulation data from the literature. The specific criteria are as follows:

[0078] 1. The mRNA in the formula contains only a single antigen (encoding single antigen).

[0079] 2. The nanoliposome formulation consists of the following four lipid components: distearate phosphatidylcholine (DSPC), cholesterol, dimyristoyl glycerol-polyethylene glycol 2000 (DMG-PEG2000), and ionizable lipids.

[0080] 3. All experimental testing time spans do not exceed one year.

[0081] 4. The experiment only includes no more than two vaccine injections.

[0082] 5. No virus challenge test was conducted.

[0083] Following this standard, 20 papers meeting the criteria were selected, yielding 365 sets of target vaccine formulation data. These data were randomly divided into training and testing sets at an 8:2 ratio. The training set, containing 292 formulations, was used to train the model and select parameters, while the testing set, containing 73 formulations, was used to measure the model's generalization ability and predictive performance.

[0084] We collected data on 365 target vaccine formulations from 20 papers, describing the formulations from four perspectives: mRNA characteristics, experimental conditions, ionizable lipid characteristics, and liposome formulation characteristics. A total of 37 feature variables were identified, and their names and meanings are shown in Table 1. The characteristics of ionizable lipids mainly stemmed from descriptions of their head, junction, and tail structures. Simultaneously, we collected antibody titers (IgG) from the experimental records of each target vaccine formulation, and used the logarithm (logIgG) as a measure of vaccine transfection efficiency. Predicting the transfection efficiency of mRNA-LNP formulations using the collected data and extracted feature variables allows us to further analyze the impact of ionizable lipid structural characteristics on transfection efficiency, providing insights and guidance for the selection and design of ionizable lipids in subsequent mRNA-LNP formulations.

[0085] Before fitting the model, all continuous variables in the prediction object and feature variables are standardized to have a mean of 0 and a variance of 1, eliminating the scaling of the influence coefficients caused by different units. This allows us to directly judge the importance of variables from the absolute value of the influence coefficients, and the predictive ability of the model can be directly observed from the model evaluation metrics.

[0086] Table 1 Description of Feature Variables and Predicted Objects

[0087]

[0088]

[0089] Based on the above embodiments, this application constructs a prediction model based on a kernel-ridge regression model, including: mapping feature variables and prediction objects to a nonlinear space; fitting the feature variables and prediction objects to a fifth-order polynomial kernel in the nonlinear space to obtain a nonlinear function; and determining the target parameters of the nonlinear function based on the minimum loss function of the kernel-ridge regression model to obtain the prediction model.

[0090] Kernel-ridge regression is an efficient nonlinear prediction model. Essentially, it maps feature variables to a nonlinear space and uses kernel tricks to achieve linear fitting. The result corresponds to a nonlinear function in the original space. A fifth-order polynomial kernel can be used to fit the prediction of transfection efficiency using the above 37 feature variables, and the model's performance on the test set can be evaluated. Let y i To predict the logIgG titer of the target, x i ∈R 37 Let x be the input vector, i = 1, ..., n, n = 317. i →Φ i =Φ(x i ) is the variable x i A mapping. Kernel ridge regression finds the optimal parameter w, that is, the parameter w corresponding to the minimum value of the following loss function (C1), the loss function is:

[0091]

[0092] Here, λ is the model penalty coefficient, and we selected the optimal value through five-fold cross-validation. The ribocytic regression model has a relatively accurate predictive ability for mRNA transfection efficiency. Despite obtaining relatively accurate predictions, the model's shortcomings lie in its poor interpretability and inability to measure the relative importance of feature variables to the predicted object.

[0093] Based on the above embodiments, this application embodiment constructs a prediction model based on the Lasso Regression Model (LASSO) and the Linear Regression Model, including: using the Lasso Regression Model to select target feature variables from feature variables; using the Linear Regression Model to perform linear regression fitting on the target feature variables and the prediction objects corresponding to the target feature variables, and determining the parameters corresponding to each feature variable through the minimum loss function of the Linear Regression Model to obtain the prediction model.

[0094] To further filter relatively important target feature variables, we use the Least Absolute Shrinkage and Selection Operator (LASSO) model for feature selection. LASSO forces some feature variables to have their parameters reduced to zero by applying an L1 penalty term. Feature variables with non-zero parameters are considered to have relatively higher explanatory power for the predicted object, thus achieving the effect of feature selection. Specifically, the goal of the LASSO model is to minimize the following loss function:

[0095]

[0096] Here, γ represents the penalty term for model complexity, and its value was selected through five-fold cross-validation. The final model selected 22 target feature variables from 37 feature variables, including 5 mRNA-related features, 6 experimental condition features, 7 ionizable lipid structure features, and 4 liposome formulation features. Due to the bias in the parameter w obtained from the LASSO regression, the target feature variables selected by the model were further fitted with linear regression. The parameters of the linear regression were used to determine the relative magnitude of the impact of each target feature variable on transfection efficiency, quantifying the positive or negative correlation between the target feature variables and transfection efficiency.

[0097] Let z i ∈R 22 The goal of linear regression is to minimize the following loss function:

[0098]

[0099] Table 2 presents the parameters related to the formulation characteristics of ionizable lipids and nanoliposomes in the linear regression model established using these 22 target feature variables. The impact of these features on transfection efficiency is also discussed. Taking the variable `tail_number` as an example, the coefficient -0.52 is generally interpreted as follows: With other feature variables equal, for every standard deviation increase in the number of tails of the ionizable lipid, the transfection efficiency (logIgG titer) decreases by 0.52. Of course, due to the inevitable collinearity between variables, for example, an increase in the number of tails may affect the lipophilicity of the ionizable lipid (logP, corresponding to the variable name XlogP3), and changes in logP will also affect transfection efficiency. Here, it is assumed that the change caused by the change in the number of tails themselves is greater, thus the relative impact of this feature variable remains unchanged. In actual prediction, multiple feature variables will change simultaneously with changes in ionizable lipids. The effects of these feature variables can be superimposed; for example, with mRNA-related features and experimental condition-related variables remaining constant, the impact of different ionizable lipids on transfection efficiency can be compared.

[0100] Table 2. Parameters of ionizable lipid and nanoliposome formulation-related characteristics in linear regression.

[0101]

[0102]

[0103] The results in Table 2 indicate that the longest carbon chain length, the carbon chain length from the head nitrogen atom to the linker, the nitrogen-to-phosphorus ratio, the number of tails, the proportion of ionizable lipids, and the ratio of ionizable lipids to DSPC and PEG are the most important factors affecting transfection efficiency.

[0104] Linear regression model results indicate that, from the perspective of ionizable lipid structure, the carbon chain length from the head nitrogen atom to the linker and the maximum carbon chain length have a positive impact on transfection efficiency. Lipophilicity, the carbon chain length from the head nitrogen atom to the head hydroxyl oxygen atom, the presence of a carboxyl tail, and the number of tails have a negative impact on transfection efficiency. From the perspective of nanoliposome formulation, the proportion of ionizable lipids, the nitrogen-to-phosphorus ratio, and the ratio of ionizable lipids to DSPC and PEG have a positive impact on transfection efficiency.

[0105] This application uses a model to predict vaccine transfection efficiency, quantifying the transfection efficiency of vaccine formulations and providing support for researchers in optimizing the design of mRNA vaccines. It attempts to use the model to screen key characteristic variables and quantify their impact on transmissibility, thereby providing more specific optimization directions for the design of future mRNA delivery systems.

[0106] Based on the above embodiments, this application embodiment constructs a prediction model based on the lightGBM model, including: randomly selecting several feature variables at each bifurcation point of the lightGBM model and iterating through them to determine the optimal split point; determining the importance ranking of the feature variables based on the number of times the feature variables are selected; and constructing a prediction model based on the importance ranking of the feature variables and the optimal split point.

[0107] Tree models are based on a greedy algorithm, randomly selecting a subset of features at each branch point and iterating through them to find the optimal split point. Based on this characteristic, the advantage of tree models lies in their ability to rank features by the number of times they are selected. We employ the lightGBM model within tree models, using the aforementioned 37 feature variables to predict transfection efficiency. Figure 2 This application provides an example of a feature variable importance ranking chart, which can be compared and analyzed with the parameters of a linear regression model. Unlike the parameters of a linear regression model, the tree model fits a non-linear relationship. Figure 2It can be concluded that characteristic variables related to experimental conditions are the most important factors affecting transfection efficiency, such as testing time, dosage, and route of administration. Secondly, characteristic variables related to mRNA, such as cap type and pseudouridine modification, also have a significant impact on transfection efficiency. Characteristic variables related to ionizable lipids and formulations, such as lipophilicity, pKa, and nitrogen-to-phosphorus ratio, also rank high in importance among characteristic variables. Among the characteristics of ionizable lipids, lipophilicity, pKa, longest carbon chain length, number of carbon atoms between the nitrogen atom and hydroxyl group in the ionizable lipid head, number of linkers, number of carbon atoms between the nitrogen atom in the head and the linker, and whether the ionizable lipid head has a hydroxyl group are relatively important influencing factors.

[0108] Based on the above embodiments, after obtaining the prediction model, the present application further includes: calculating the root mean square error and the coefficient of determination based on the predicted value and the true value of the prediction model; and determining the accuracy of the prediction model based on the root mean square error and the coefficient of determination.

[0109] To measure the predictive performance of the prediction model, the root mean square error (RMSE) and the coefficient of determination (R²) are used. 2 R is used as a metric. RMSE measures the error between the predicted and actual values; a smaller RMSE value indicates a better predictive performance. 2 RMSE measures the correlation coefficient between predicted and actual values; a higher value indicates a better predictive performance. 2 The calculation method is as follows:

[0110]

[0111]

[0112] Among them, y i This is the true value of the logIgG titer. These are the model's predicted values. This refers to the mean of the true values.

[0113] Table 3 compares the prediction performance of the three prediction models on the test set. The information in the table shows that the ridge-nuclear regression and the littleGBM model are similar in predictive ability and outperform the linear regression model. Both ridge-nuclear regression and littleGBM models are non-linear models, which can better describe the complex relationship between feature variables and the predicted object. Therefore, the advantage of these two models lies in accurately predicting the transfection efficiency of mRNA vaccines.

[0114] Table 3 Predictive Performance of Three Models

[0115]

[0116]

[0117] Although the predictive power of linear regression models is weaker than that of nonlinear models such as the ribocytic regression model and the LighGBM model, their conclusions are more interpretable. Furthermore, the performance and consistency of nonlinear and linear regression models in relative conclusions are compared. That is, given two mRNA-LNP formulations, are the conclusions of the two models correct or consistent in judging the relative magnitude of transfection efficiency?

[0118] The results showed that the accuracy of the kernel-ridge regression model in determining relative magnitudes was 87.7%, while the accuracy of the linear regression model was 81.5%. The consistency between the conclusions of the kernel-ridge regression model and the linear regression model was 81.5%. Figure 3 This application provides an embodiment of a distribution chart showing the absolute value distribution of the difference in actual transfection efficiency between any two formulations in a test set. Figure 3 This reflects the absolute value distribution of the difference in actual transfection efficiency between any two formulations in the test set, and is classified according to whether the prediction conclusions of the two models are consistent. The horizontal axis represents the absolute value of the logIgG difference; the vertical axis represents the frequency; the black part represents the part with consistent conclusions; the gray part represents the part with inconsistent conclusions. Figure 3 We can see that when the conclusions of the two models are inconsistent, the predicted transfection efficiencies of the two formulations are relatively close. Based on the above results, although the linear regression model cannot provide a very accurate prediction of transfection efficiency, its performance in predicting relative conclusions is close to that of the nonlinear model. In summary, the linear regression model describes the trend of the predicted object as the relatively important characteristic variables change within a specific range. Its conclusions can be used for comparisons between ionizable lipids or formulations and have good interpretability. Therefore, medicinal chemists can choose the appropriate model from the three models according to different purposes and needs.

[0119] Based on the above conclusions, we searched the literature for new mRNA-LNP formulations to validate our findings. Experimental data from these formulations showed that, while ensuring consistent mRNA-related characteristics and experimental conditions, different ionizable lipids resulted in varying vaccine transfection efficiencies. We used a linear regression model to predict the transfection efficiency difference between the two liposome formulations. Our linear regression model results were consistent with the experimental findings of the four articles. Table 4 describes the mRNA-LNP formulations we used for comparison. Table 5 presents our validation results.

[0120] Table 4 Comparison of mRNA-LNP formulation experimental results

[0121]

[0122]

[0123] Table 5 Model Prediction Results

[0124]

[0125] Taking the first formulation as an example, lipid5 has a longer head nitrogen atom to linker carbon chain length and lower lipophilicity compared to MC3. These factors positively influence transfection efficiency. Meanwhile, although the larger head nitrogen atom to hydroxyl carbon number, more tails, and shorter maximum carbon chain length have some negative impacts, the positive contributions outweigh the negative ones. Therefore, overall, under the same mRNA and experimental conditions, lipid5 performs better than MC3.

[0126] Based on the above conclusions and validation results, we used a linear model to compare the transfection efficiencies of three ionizable lipids—SM-102, ALC-0315, and DlinDMA—under the same mRNA and experimental conditions. We then used a t-test to demonstrate the significance of the differences, followed by wet-lab validation. The model prediction results are shown in Table 6.

[0127] Table 6. Model Predictions and t-test Results

[0128] formula Poor predicted transfection efficiency T-statistic p-value ALC0315 vs SM102 +2.11 0.01 ALC0315 vs DLinDMA +1.29 0.07 DLinDMA vs SM102 +0.55 0.28

[0129] The results showed that SM102 and DLinDMA had higher transfection efficiencies than ALC0315, while SM102 had a higher transfection efficiency than DLinDMA. Significant differences in transfection efficiency (90% significance) were found between ALC0315 and SM102, and between ALC0315 and DLinDMA, while the difference between DLinDMA and SM102 was not significant. Note that the linear regression model used here discards some relatively unimportant features compared to a nonlinear model; therefore, we adopted a significance level slightly lower than the usual 95% statistical level, but we still consider these results valid.

[0130] Wet assay results showed that the transfection efficiency of the mRNA delivery systems encapsulated with DLinDMA and SM102 was significantly higher than that of ALC0315. Meanwhile, the transfection efficiency of SM102 was higher than that of DLinDMA, but the trend was not significant. These results are consistent with the model predictions. An example of the wet assay is described below.

[0131] Example 1: According to the formulation table in Table 7, the transfection efficiency of SM-102 and ALC-0315 ionizable lipids was compared. The model was used to predict the transfection efficiency of LNP-mRNA vaccine, and relevant experiments were conducted to verify the results.

[0132] Table 7 Prescription Table

[0133] Prescription 1 Prescription 2 effect mole percentage % SM-102 ALC-0315 Ionizable lipids 50 DSPC DSPC assist lipids 10 Cholesterol Cholesterol cholesterol 38.5 PEG2000-DMG PEG2000-DMG PEGylated phospholipids 1.5

[0134] Preparation of LNP solution: Dissolve all lipid materials (ionizable lipids: DSPC: cholesterol: DMG-PEG2000 = 50:10:38.5:1.5 mol%) in 1 ml of anhydrous ethanol solution (99.7%).

[0135] Preparation of mRNA solution: Dissolve mRNA in 50mM citrate buffer at (N / P=6) to prepare a volume of 3mL.

[0136] LNP-mRNA preparation method: LNP-mRNA was prepared using the ethanol injection method. The ethanol solution of lipids was drawn into a syringe and slowly added dropwise to rapidly stirred mRNA-citrate buffer. The prepared LNP-mRNA was then added to a dialysis bag, which was placed in PBS and dialyzed for 12 hours. After sample collection, ultrafiltration was performed using a Milpore 30KD ultrafiltration tube, centrifuged at 3000g for 20 minutes, and then stored in pH 7.4 PBS solution at 4°C.

[0137] Formulation evaluation of LNP-mRNA: Quant-iT TM RNA reagent is a highly sensitive fluorescent nucleic acid staining agent capable of detecting 1-200 ng of nucleic acid in solution. This nucleic acid dye cannot pass through LNPs, therefore only free nucleic acids not encapsulated by LNPs can be bound. Triton-100, as a surfactant, is often used as a demulsifier. Treating LNP-mRNA with 2% Triton-100 can release the encapsulated nucleic acid, yielding the total nucleic acid amount. The drug loading is obtained by calculating the difference in nucleic acid amount before and after demulsification, and then divided by the total nucleic acid amount to obtain the encapsulation efficiency, i.e.: Encapsulation efficiency (%) = (quantitative amount after demulsification - quantitative amount before demulsification) / quantitative amount after demulsification.

[0138] Particle size and potential determination: The sample was diluted with 10 mM phosphate buffer (pH 7.4), and the particle size and potential were measured using a Malvern laser particle size analyzer.

[0139] Cellular evaluation of LNP-mRNA: Cationic polymers can form complexes with negatively charged mRNA. These complexes cross the membrane (either through membrane fusion or endocytosis) and are "eaten" by the cell, thus carrying the target mRNA into the cell. Approximately 24 hours after transfection, transfection efficiency and protein expression can be assessed. Two cell lines, 293T cells and mouse hepatocellular carcinoma cells (Hepa), were selected for comparison of transfection efficiency and the fluorescence intensity of expressed proteins. LNP-mRNA was administered at a dose of 0.5 μg mRNA / dose. For detection, transfection efficiency was calculated as (number of positive cells / total number of cells at detection) * 100%, and the fluorescence intensity of expressed proteins was also measured. 24 hours before transfection, cells were seeded in 12-well plates at a density of 5 x 10⁴ cells, and the medium was changed before transfection. A certain volume of LNP-mRNA was diluted with serum-containing culture medium, and after incubation for 24 and 48 hours, the expressed fluorescent protein was detected by flow cytometry.

[0140] Experimental results:

[0141] Particle size potential results: see Figure 8 .

[0142] Table 8. Particle size potential results

[0143]

[0144] The prepared samples had a particle size of approximately 150 nm, a PDI of less than 0.1, and good dispersibility. The potential results were in the range of -5 mV.

[0145] Cell transfection results:

[0146] Figure 4 The image shows a comparison of SM-102 and ALC-0315 ionizable lipids transfected in 293T cells for 24 hours, as provided in the embodiments of this application. The three curves from left to right represent: 293T mRNA-24h, 293T ALC-0315-24h, and 293TSM-102-24h, respectively.

[0147] Figure 5 The image shows a comparison of SM-102 and ALC-0315 ionizable lipids transfected in 293T cells for 48 hours, as provided in the embodiments of this application. The three curves from left to right represent: 293T mRNA-48h, 293T ALC-0315-48h, and 293TSM-102-48h, respectively.

[0148] Figure 6The image shows a comparison of SM-102 and ALC-0315 ionizable lipids transfected in Hepa cells for 24 hours, as provided in the embodiments of this application. The three curves from left to right represent Hepa mRNA-24h, Hepa ALC-0315-24h, and HepaSM-102-24h, respectively.

[0149] Figure 7 The image shows a comparison of SM-102 and ALC-0315 ionizable lipids transfected in Hepa cells for 48 hours, as provided in the embodiments of this application. The three curves from left to right represent: Hepa mRNA-48h, Hepa ALC-0315-48h, and HepaSM-102-48h, respectively.

[0150] The experimental results show that in both 293T and Hepa cells, the transfection results at 24h and 48h were that SM-102 was superior to ALC-0315. The machine learning model predicted that SM-102 was superior to ALC-0315, and the experimental results were consistent with the predictions.

[0151] Example 2: According to the formulation table in Table 9, the transfection efficiency of DlinDMA and ALC-0315 ionizable lipids was compared. The transfection efficiency of the LNP-mRNA vaccine was predicted using a model, and relevant experiments were conducted to verify the results. The machine model prediction showed that DlinDMA had a better transfection effect than ALC-0315.

[0152] Table 9 Prescription Table

[0153] Prescription 1 Prescription 2 effect mole percentage % DlinDMA ALC-0315 Ionizable lipids 50 DSPC DSPC assist lipids 10 Cholesterol Cholesterol cholesterol 38.5 PEG2000-DMG PEG2000-DMG PEGylated phospholipids 1.5

[0154] The preparation method and evaluation are the same as above.

[0155] Experimental results:

[0156] Particle size potential results: see Table 10.

[0157] Table 10. Particle size and potential results

[0158]

[0159] After preparation, the LNP-mRNA particle size is about 150 nm and the potential is about -5 mV.

[0160] Fluorescent protein expression detected by flow cytometry.

[0161] The dosage was 0.5 μg / mL, and the groups were: control group, naked mRNA group, LNP-mRNA (ALC-0315 group), and LNP-mRNA (DLinDMA group).

[0162] Cell transfection results:

[0163] Figure 8 The image shows a comparison of DlinDMA and ALC-0315 ionizable lipids transfected in 293T cells for 24 hours, as provided in the embodiments of this application. The three curves from left to right represent: 293T mRNA-24h, 293T ALC-0315-24h, and 293T DlinDMA-24h, respectively.

[0164] Figure 9 The image shows a comparison of DlinDMA and ALC-0315 ionizable lipids in 293T cells transfected for 48 hours, as provided in the embodiments of this application. The three curves from left to right represent: 293T mRNA-48h, 293T ALC-0315-48h, and 293T DlinDMA-48h, respectively.

[0165] Figure 10 The image shows a comparison of DlinDMA and ALC-0315 ionizable lipids in Hepa cells transfected for 24 hours, as provided in the embodiments of this application. The three curves from left to right represent Hepa mRNA-24h, Hepa ALC-0315-24h, and HepaDlinDMA-24h, respectively.

[0166] Figure 11 The image shows a comparison of DlinDMA and ALC-0315 ionizable lipids in Hepa cells after 48 hours of transfection, as provided in the embodiments of this application. The three curves from left to right represent: Hepa mRNA-48h, Hepa ALC-0315-48h, and Hepa DlinDMA-48h, respectively.

[0167] The experimental results show that, in 293T cells and Hepa cells, after 24h and 48h of transfection, the fluorescent protein intensity of DLinDMA was higher than that of ALC-0315, indicating higher transfection efficiency, which is consistent with the predicted results.

[0168] Example 3, according to the formulation table in Table 11, consists of two ionizable lipids, SM-102 and DlinDMA. The transfection efficiency of the LNP-mRNA vaccine was predicted using a model, and relevant experiments were conducted to verify this.

[0169] Table 11 Prescription Table

[0170] Prescription 1 Prescription 2 effect mole percentage % SM-102 DlinDMA Ionizable lipids 50 DSPC DSPC assist lipids 10 Cholesterol Cholesterol cholesterol 38.5 PEG2000-DMG PEG2000-DMG PEGylated phospholipids 1.5

[0171] The preparation method is the same as above.

[0172] Cellular evaluation of LNP-mRNA:

[0173] Two cell lines, 293T cells and mouse hepatocellular carcinoma cells (Hepa), were selected for comparison of transfection efficiency and the fluorescence intensity of expressed proteins. LNP-mRNA was administered at a dose of 0.5 μg mRNA / dose. Transfection efficiency was calculated as (number of positive cells / total number of cells at detection) * 100%, and the fluorescence intensity of expressed proteins was also measured.

[0174] 24 hours before transfection, cells were seeded in 12-well plates with a cell count of 5 x 10⁴. The medium was changed before transfection. A certain volume of LNP-mRNA was diluted with serum-containing culture medium and incubated for 6 h, 12 h, 24 h, and 48 h. The expression of fluorescent protein was detected by flow cytometry.

[0175] Cell transfection results:

[0176] Figure 12 A comparative diagram of SM-102 and DlinDMA ionizable lipids transfected in 293T cells for 6 hours, provided as an embodiment of this application.

[0177] Figure 13 A comparison diagram of SM-102 and DlinDMA ionizable lipids transfected in 293T cells for 12 hours, provided as an embodiment of this application.

[0178] Figure 14 The image shows a comparison of SM-102 and DlinDMA ionizable lipids transfected in 293T cells for 24 hours, as provided in the embodiments of this application. The three curves from left to right represent: 293T mRNA-24h, 293T DlinDMA-24h, and 293TSM-102-24h, respectively.

[0179] Figure 15 The image shows a comparison of SM-102 and DlinDMA ionizable lipids transfected in 293T cells for 48 hours, as provided in the embodiments of this application. The three curves from left to right represent: 293T mRNA-48h, 293T DlinDMA-48h, and 293TSM-102-48h, respectively.

[0180] The short-term transfection effects in 293T cells are shown in the figure. At 6h and 12h time points, SM-102 showed slightly higher fluorescence values ​​than DLinDMA, but the effect was not significant. At 24h and 48h, SM-102 was superior to DLinDMA in cell transfection results.

[0181] Figure 16 A comparison of SM-102 and DlinDMA ionizable lipids transfected in Hepa cells for 6 hours, as provided in an embodiment of this application.

[0182] Figure 17A comparison of SM-102 and DlinDMA ionizable lipids transfected in Hepa cells for 12 hours, as provided in an embodiment of this application.

[0183] Figure 18 The image shows a comparison of SM-102 and DlinDMA ionizable lipids in Hepa cells transfected for 24 hours, as provided in the embodiments of this application. The three curves from left to right represent Hepa mRNA-24h, Hepa DlinDMA-24h, and HepaSM-102-24h, respectively.

[0184] Figure 19 The image shows a comparison of SM-102 and DlinDMA ionizable lipids in Hepa cells after 48 hours of transfection, as provided in the embodiments of this application. The three curves from left to right represent Hepa mRNA-48h, Hepa DlinDMA-48h, and Hepa SM-102-48h, respectively.

[0185] In Hepa cells, the transfection efficiency was lower than that in 293T cells. At 6 h, SM-102 expression was slightly higher than DLinDMA expression. At 12 h, 24 h, and 48 h, there was no significant difference between the two lipid materials.

[0186] In the above embodiments, the method for predicting mRNA vaccine transfection efficiency has been described in detail. This application also provides embodiments corresponding to the device for predicting mRNA vaccine transfection efficiency. It should be noted that this application describes the embodiments of the device from two perspectives: one is based on functional modules, and the other is based on hardware.

[0187] Figure 20 A structural diagram of a device for predicting mRNA vaccine transfection efficiency provided in this application embodiment is shown below. Figure 20 As shown, the device for predicting mRNA vaccine transfection efficiency includes:

[0188] Module 10 is used to acquire vaccine formula data and predict demand.

[0189] The determination module 11 is used to determine the corresponding prediction model according to the prediction requirements; wherein, the prediction model is a relationship model between vaccine formulation data and transfection efficiency constructed based on any one of the following models: nuclear ridge regression model, linear regression model, and lightGBM model.

[0190] The prediction module 12 is used to predict the transfection efficiency of the vaccine formulation data under test using a defined prediction model.

[0191] Based on the above embodiments, as a preferred embodiment, it further includes:

[0192] The vaccine formula data acquisition module is used to retrieve multiple vaccine formula data from the database using preset keywords;

[0193] The filtering module is used to filter multiple vaccine formula data according to preset criteria to obtain multiple target vaccine formula data;

[0194] The prediction object and feature variable determination module is used to determine the antibody titer corresponding to each target vaccine formulation data as the prediction object, and to determine the feature variables in the target vaccine formulation data; among which, the feature variables include mRNA characteristics, experimental conditions, ionizable lipid characteristics, and liposome formulation characteristics; antibody titer is used to measure vaccine transfection efficiency;

[0195] The building block is used to construct a prediction model based on feature variables and the object to be predicted.

[0196] Based on the above embodiments, as a preferred embodiment, it further includes:

[0197] The continuous variable determination module is used to determine continuous variables among the feature variables;

[0198] The data standardization module is used to perform data standardization processing on continuous variables and prediction objects respectively.

[0199] Based on the above embodiments, as a preferred embodiment, the construction module includes:

[0200] Mapping unit, used to map feature variables and prediction objects to a nonlinear space;

[0201] The fitting unit is used to fit the feature variables and the object to be predicted in a nonlinear space using a cubic polynomial kernel to obtain a nonlinear function.

[0202] The first determining unit is used to determine the target parameters of the nonlinear function based on the minimum loss function of the kernel ridge regression model, so as to obtain the prediction model.

[0203] Based on the above embodiments, as a preferred embodiment, the construction module includes:

[0204] The filtering unit is used to filter feature variables using a lasso regression model to obtain target feature variables;

[0205] The second determining unit is used to perform linear regression fitting on the target feature variable and the prediction object corresponding to the target feature variable using a linear regression model, and to determine the parameters corresponding to each feature variable through the minimum loss function of the linear regression model in order to obtain the prediction model.

[0206] Based on the above embodiments, as a preferred embodiment, the construction module includes:

[0207] The third determining unit is used to randomly select several feature variables at each bifurcation point of the lightGBM model and iterate through the values ​​to determine the optimal split point.

[0208] The fourth determining unit is used to rank the importance of feature variables based on the number of times they are selected.

[0209] The building blocks are used to construct a prediction model based on the importance ranking of feature variables and the optimal split point.

[0210] Based on the above embodiments, as a preferred embodiment, it further includes:

[0211] The calculation module is used to calculate the root mean square error and coefficient of determination based on the predicted values ​​and the actual values ​​of the prediction model.

[0212] The accuracy determination module is used to determine the accuracy of the prediction model based on the root mean square error and / or the coefficient of determination.

[0213] Based on the above embodiments, as a preferred embodiment, it further includes:

[0214] The validation module is used to validate the prediction results of the prediction model using wet experiments.

[0215] Since the embodiments of the apparatus and the embodiments of the method correspond to each other, please refer to the description of the embodiments of the method for the embodiments of the apparatus, which will not be repeated here.

[0216] This application provides an apparatus for predicting the transfection efficiency of mRNA vaccines. It acquires vaccine formulation data and prediction requirements; determines a corresponding prediction model based on the prediction requirements; wherein the prediction model is a relationship model between vaccine formulation data and transfection efficiency constructed based on any one of the following models: cerebellopontine regression model, linear regression model, and lightGBM model; and uses the determined prediction model to predict the transfection efficiency of the vaccine formulation data. This application predicts vaccine transfection efficiency through a model, enabling the quantification of vaccine formulation transfection efficiency and providing support for researchers in the optimized design of mRNA vaccines.

[0217] Figure 21 A structural diagram of another mRNA vaccine transfection efficiency prediction device provided in this application embodiment is shown below. Figure 21 As shown, the device for predicting mRNA vaccine transfection efficiency includes: a memory 20 for storing a computer program;

[0218] The processor 21 is used to execute a computer program to implement the steps of the method for predicting the transfection efficiency of mRNA vaccines as described in the above embodiments.

[0219] The device for predicting mRNA vaccine transfection efficiency provided in this embodiment may include, but is not limited to, smartphones, tablets, laptops, or desktop computers.

[0220] The processor 21 may include one or more processing cores, such as a quad-core processor or an octa-core processor. The processor 21 may be implemented using at least one of the following hardware forms: Digital Signal Processor (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). The processor 21 may also include a main processor and a coprocessor. The main processor, also known as the Central Processing Unit (CPU), is used to process data in the wake-up state; the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, the processor 21 may integrate a Graphics Processing Unit (GPU), which is responsible for rendering and drawing the content to be displayed on the screen. In some embodiments, the processor 21 may also include an Artificial Intelligence (AI) processor, which is used to handle computational operations related to machine learning.

[0221] The memory 20 may include one or more computer-readable storage media, which may be non-transitory. The memory 20 may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices or flash memory devices. In this embodiment, the memory 20 is used to store at least the following computer program 201, which, after being loaded and executed by the processor 21, is capable of implementing the relevant steps of the method for predicting mRNA vaccine transfection efficiency disclosed in any of the foregoing embodiments. In addition, the resources stored in the memory 20 may also include an operating system 202 and data 203, and the storage method may be temporary or permanent storage. The operating system 202 may include Windows, Unix, Linux, etc. The data 203 may include, but is not limited to, vaccine formulation data and predicted demand.

[0222] In some embodiments, the device for predicting mRNA vaccine transfection efficiency may further include a display screen 22, an input / output interface 23, a communication interface 24, a power supply 25, and a communication bus 26.

[0223] Those skilled in the art will understand that Figure 21The structure shown does not constitute a limitation on the predictive device for mRNA vaccine transfection efficiency and may include more or fewer components than illustrated.

[0224] This application provides an apparatus for predicting the transfection efficiency of mRNA vaccines, including a memory and a processor. When the processor executes a program stored in the memory, it can perform the following methods: acquiring vaccine formulation data to be tested and prediction requirements; determining a corresponding prediction model based on the prediction requirements; wherein the prediction model is a relationship model between vaccine formulation data and transfection efficiency constructed based on any one of the following models: nuclear ridge regression model, linear regression model, and lightGBM model; and using the determined prediction model to predict the transfection efficiency of the vaccine formulation data to be tested.

[0225] Finally, this application also provides an embodiment corresponding to a computer-readable storage medium. The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps described in the method embodiment for predicting mRNA vaccine transfection efficiency.

[0226] It is understood that if the methods in the above embodiments are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and executes all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0227] The foregoing provides a detailed description of a method, apparatus, and medium for predicting mRNA vaccine transfection efficiency. The various embodiments in the specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section. It should be noted that those skilled in the art can make several improvements and modifications to this application without departing from the principles of this application, and these improvements and modifications also fall within the protection scope of the claims of this application.

[0228] It should also be noted that, in this specification, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

Claims

1. A method for predicting the transfection efficiency of an mRNA vaccine, characterized in that, include: Obtain vaccine formulation data and predict demand; The corresponding prediction model is determined based on the prediction requirements; wherein, the prediction model is a relationship model between vaccine formulation data and transfection efficiency constructed based on any one of the following models: nuclear ridge regression model, linear regression model, and lightGBM model. The transfection efficiency of the vaccine formulation data to be tested is predicted using a defined prediction model. Also includes: Retrieve multiple vaccine formula data from the database using preset keywords; Multiple vaccine formulation data are filtered according to preset standards to obtain multiple target vaccine formulation data; The antibody titer corresponding to each target vaccine formulation data is determined as a prediction object, and the characteristic variables in the target vaccine formulation data are determined; wherein, the characteristic variables include mRNA characteristics, experimental conditions, ionizable lipid characteristics, and liposome formulation characteristics; the antibody titer is used to measure vaccine transfection efficiency; The prediction model is constructed based on the feature variables and the prediction object; The step of constructing the prediction model based on the feature variables and the prediction object includes: Map the feature variables and the prediction object to a nonlinear space; In the nonlinear space, a nonlinear function is obtained by fitting the feature variables and the predicted object using a fifth-order polynomial kernel; The target parameters of the nonlinear function are determined based on the minimum loss function of the kernel ridge regression model to obtain the prediction model; Alternatively, the target feature variables can be obtained by using a lasso regression model to screen the feature variables; The linear regression model is used to perform linear regression fitting on the target feature variable and the prediction object corresponding to the target feature variable, and the parameters corresponding to each feature variable are determined by the minimum loss function of the linear regression model to obtain the prediction model; Alternatively, at each bifurcation point of the lightGBM model, several feature variables can be randomly selected and their values ​​iterated to determine the optimal split point. The importance ranking of the feature variables is determined based on the number of times the feature variables are selected. The prediction model is constructed based on the importance ranking of the feature variables and the optimal split point.

2. The method for predicting mRNA vaccine transfection efficiency according to claim 1, characterized in that, Before constructing the prediction model based on the feature variables and the prediction object, the method further includes: Identify the continuous variables among the feature variables; Data standardization is performed on the continuous variables and the prediction objects respectively.

3. The method for predicting mRNA vaccine transfection efficiency according to claim 1, characterized in that, After obtaining the prediction model, the process also includes: The root mean square error and coefficient of determination are calculated based on the predicted and actual values ​​of the prediction model. The accuracy of the prediction model is determined based on the root mean square error and / or the coefficient of determination.

4. The method for predicting mRNA vaccine transfection efficiency according to claim 1, characterized in that, After obtaining the prediction model, the process also includes: The prediction results of the prediction model were verified using wet experiments.

5. A device for predicting mRNA vaccine transfection efficiency, characterized in that, include: The acquisition module is used to acquire vaccine formula data and predict demand. The determination module is used to determine the corresponding prediction model according to the prediction requirements; wherein, the prediction model is a relationship model between vaccine formulation data and transfection efficiency constructed based on any one of the following models: nuclear ridge regression model, linear regression model, and lightGBM model. The prediction module is used to predict the transfection efficiency of the vaccine formulation data to be tested using a defined prediction model. Also includes: The vaccine formula data acquisition module is used to retrieve multiple vaccine formula data from the database using preset keywords; The filtering module is used to filter multiple vaccine formula data according to preset criteria to obtain multiple target vaccine formula data; The prediction object and feature variable determination module is used to determine the antibody titer corresponding to each target vaccine formulation data as the prediction object, and to determine the feature variables in the target vaccine formulation data; among which, the feature variables include mRNA characteristics, experimental conditions, ionizable lipid characteristics, and liposome formulation characteristics; antibody titer is used to measure vaccine transfection efficiency; Builder modules are used to build predictive models based on feature variables and the objects to be predicted; The building blocks include: Mapping unit, used to map feature variables and prediction objects to a nonlinear space; The fitting unit is used to fit the feature variables and the object to be predicted in a nonlinear space using a cubic polynomial kernel to obtain a nonlinear function. The first determining unit is used to determine the target parameters of the nonlinear function based on the minimum loss function of the kernel ridge regression model, so as to obtain the prediction model; Alternatively, the building blocks include: The filtering unit is used to filter feature variables using a lasso regression model to obtain target feature variables; The second determining unit is used to perform linear regression fitting on the target feature variable and the prediction object corresponding to the target feature variable using a linear regression model, and to determine the parameters corresponding to each feature variable through the minimum loss function of the linear regression model in order to obtain the prediction model. Alternatively, the building blocks include: The third determining unit is used to randomly select several feature variables at each bifurcation point of the lightGBM model and iterate through the values ​​to determine the optimal split point. The fourth determining unit is used to rank the importance of feature variables based on the number of times they are selected. The building blocks are used to construct a prediction model based on the importance ranking of feature variables and the optimal split point.

6. A device for predicting mRNA vaccine transfection efficiency, characterized in that, Includes memory used to store computer programs; A processor, configured to execute the computer program to implement the steps of the method for predicting mRNA vaccine transfection efficiency as described in any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the method for predicting mRNA vaccine transfection efficiency as described in any one of claims 1 to 4.

Images