Screening method and device of low-temperature protective agent, electronic equipment and storage medium

Abstract

The invention provides a screening method of a low-temperature protective agent, which comprises the following steps: constructing a solution system containing a plurality of single-component low-temperature protective agents, respectively carrying out molecular dynamics simulation to calculate a plurality of microscopic characteristics and target macroscopic characteristics of each solution, and constructing a total microscopic characteristic data set and each target macroscopic characteristic data set; based on clustering and importance analysis and in combination with the corresponding target macroscopic feature data set and machine learning, determining evaluation features for each target macroscopic feature from the total microscopic feature data set, and obtaining a prediction model of each target macroscopic feature at the same time; calculating evaluation characteristics of the low-temperature protective agent to be screened, and respectively inputting the evaluation characteristics into the corresponding target macroscopic characteristic prediction models to obtain corresponding prediction values; predicting the toxicity of the low-temperature protective agent to be screened; and taking the to-be-screened low-temperature protective agent of which all the target macroscopic feature predicted values and toxicity predicted values meet the requirements as a screening result. The method has the characteristics of rapidness, accuracy and low cost.

Description

Technical Field

[0001] This invention relates to the field of cryopreservation of biological materials, and particularly to a method, apparatus, electronic device, and storage medium for screening mixed cryoprotectants during the vitrification preservation of biological materials. Background Technology

[0002] Long-term in vitro preservation of biomaterials is of great significance for fields such as organ transplantation, cell therapy, and the construction of biobanks. Cryopreservation technology can effectively extend preservation time by reducing the rate of cellular metabolism. Among them, vitrification preservation at liquid nitrogen temperature can transform water into an amorphous state, avoiding structural damage and osmotic stress caused by the formation of crystal ice, and achieving almost complete metabolic cessation, showing the potential for preservation for decades or even indefinitely.

[0003] The biggest challenge in vitrification preservation lies in the formation of ice crystals during cooling and rewarming. For large tissues and organs, due to limitations in heat transfer efficiency, rapid cooling alone is insufficient for vitrification in engineering applications. Therefore, cryoprotectants are needed to reduce the critical cooling rate required for vitrification. Cryoprotectant molecules inhibit ice crystal formation by disrupting the local structure of water molecules and slowing their kinetics. To date, numerous cryoprotectant molecules, such as dimethyl sulfoxide, glycerol, and formamide, have been used in biopreservation. However, high concentrations of cryoprotectants are often accompanied by high cytotoxicity, making it difficult to guarantee survival rates after rewarming. Therefore, current techniques often employ a combination of multiple cryoprotectants to neutralize toxicity. Existing mixed cryoprotectant components and ratios rely on experimental trial and error, lacking theoretical guidance, resulting in high screening costs and low efficiency. Summary of the Invention

[0004] The purpose of this invention is to overcome the limitations of existing experimental trial-and-error methods for preparing cryoprotectant solutions, and to provide a method, apparatus, electronic device, and storage medium for screening cryoprotectants based on molecular dynamics simulations and machine learning. This method can efficiently screen mixed cryoprotectant solutions that simultaneously meet target macroscopic characteristics (such as low nucleation temperature and high glass transition temperature) and low toxicity. This invention simulates the microscopic characteristics of cryoprotectant solutions, predicts the target macroscopic characteristics of the solutions based on machine learning models, and determines the components and their proportions in the screened mixed cryoprotectant solution by combining a solution toxicity model.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] The first aspect of this invention provides a method for screening cryoprotectants, the method being used to screen from a variety of cryoprotectants to be screened those whose target macroscopic characteristics and toxicity prediction values ​​both meet the requirements, wherein the number of target macroscopic characteristics is at least one; the screening method includes:

[0007] Step S100: Construct a single-component cryoprotectant solution system containing different types and concentrations, perform molecular dynamics simulations to calculate various microscopic features of each solution and the target macroscopic features, construct a total microscopic feature dataset using all microscopic features, and construct a corresponding target macroscopic feature dataset for each target macroscopic feature.

[0008] Step S200: Based on clustering and importance analysis, and combined with the corresponding target macro feature dataset and machine learning, determine the evaluation features for each target macro feature from the total micro feature dataset, and obtain the prediction model for each target macro feature.

[0009] Step S300: Calculate the evaluation characteristics of all cryoprotectants to be screened;

[0010] Step S400: Input the evaluation characteristics of all cryoprotectants to be screened into the corresponding target macroscopic feature prediction model to obtain the predicted value of each target macroscopic feature; predict the toxicity of all cryoprotectants to be screened to obtain the toxicity prediction value.

[0011] Step S500: Select the cryoprotectants that meet the requirements for all target macroscopic characteristic prediction values ​​and toxicity prediction values ​​as the screening results.

[0012] In some embodiments, the target macroscopic characteristics of the cryoprotectant include at least one of water activity, glass transition temperature, ice nucleation rate, and ice growth rate; and / or

[0013] The water activity was obtained by isothermal simulation of the configurational evolution of the solution system, and the glass transition temperature was obtained by simulating the enthalpy of the solution system at different temperatures through cooling; and / or

[0014] The total microscopic feature dataset contains multiple microscopic features, including radial distribution function information between water and oxygen atoms, radial distribution function information between water and oxygen atoms and hydrophilic atoms of the cryoprotectant, angular distribution function between water and oxygen atoms, water bond order parameter, water fifth nearest neighbor distance, water orientation tetrahedral ordinal number, horizontally shifted tetrahedral ordinal number, water local structure index, translational order of water second shell, water tetrahedral entropy, water surface area, water volume, water sphericity, solution density, cryoprotectant cluster ratio, water diffusion coefficient, water hydroxyl rotation time, water dipole moment rotation time, water relaxation time, number of hydrogen bonds between water molecules, length of hydrogen bonds between water molecules, angle of hydrogen bonds between water molecules, lifetime of hydrogen bonds between water molecules, number of hydrogen bonds between water and cryoprotectant molecules, length of hydrogen bonds between water and cryoprotectant molecules, angle of hydrogen bonds between water and cryoprotectant molecules, lifetime of hydrogen bonds between water and cryoprotectant molecules, and the total number of hydrogen bonds in the solution.

[0015] In some embodiments, step S200 includes:

[0016] Step S210: Use clustering methods to analyze the micro-features contained in the total dataset. Classifying by microscopic features, traversing all classification cases, a total of The classification results will The classification results are sorted in ascending order according to the number of categories.

[0017] Step S220: Use a machine learning model combined with the additivity feature interpretation method to... The importance of micro-features to the macro-features of each target is analyzed to obtain the importance analysis results;

[0018] Step S230: For the Nth classification result obtained in step S210, initialize... Select the feature class that affects the first The most important micro-feature among the macro-features of the target is the first... Microscopic features, all the l-th microscopic features constitute the l-th microscopic features. Microscopic feature dataset ;use and the first constructed in step S100 A target macroscopic feature dataset For the first A machine learning model based on the macroscopic features of the target is trained to obtain a preliminary prediction model. ,like and If the performance difference exceeds a set threshold, then... Repeat this step until... and The performance difference did not exceed the set threshold, and the current preliminary prediction model was used. As the first Target Macro-characteristic Prediction Model And will input the first Target Macro-characteristic Prediction Model The microscopic features as a target for the first Evaluation characteristics of the macroscopic features of the target.

[0019] In some embodiments, step S210 includes:

[0020] Step S211: Traverse and calculate microscopic features and Correlation coefficient between , , ,based on Obtain the correlation matrix among all micro-features ,by As an assessment of micro-characteristics and Distance coefficient representing the degree of correlation between elements;

[0021] Step S212, to Each micro-feature in the set of micro-features is initialized into a class, i.e., initialized... ;

[0022] Step S213: Based on the distance coefficient, merge the two micro-feature categories with the smallest values ​​into a new micro-feature category; recalculate the distance between the new micro-feature category and other micro-feature categories using the connection criterion, and update the distance coefficient;

[0023] Step S214, Judgment Is it equal to 1? Then let Return to step S213, if Then the classification analysis ends.

[0024] In some embodiments, step S220 includes:

[0025] Step S221: For each target macroscopic feature, construct a machine learning model respectively, using the total microscopic feature dataset constructed in step S100. and the A target macroscopic feature dataset For the first Several machine learning models are trained, with each iteration using a solution... The micro-feature is the first The input of a machine learning model, with The corresponding macroscopic feature data of the solution are used as labels, with the solution's first... The predicted value of the target macroscopic characteristic is used as the first... The output of a machine learning model;

[0026] Step S222: Calculate each step separately. The importance of each micro-feature to the output of each machine learning model is determined, and all micro-features are ranked from highest to lowest importance to the output of each machine learning model; where, let the micro-feature be denoted as . For the first The importance of the output of each machine learning model is The calculation is performed using the additivity feature interpretation method:

[0027]

[0028] Among them, subscript This indicates the construction in step S100. The first single-component cryoprotectant solution system A solution, , It is the first A solution A collection of micro-features It is the first The solution does not contain microscopic features. The set of all microscopic features It is the first The machine learning model is for the first The output results of each solution Microscopic features For the first The machine learning model is for the first The marginal contribution of the output of a solution, This is the operator for counting the number of elements in a set.

[0029] In some embodiments, step S230 includes:

[0030] Step S231, Initialization ;

[0031] Step S232, for the first For each classification result, select the feature class that influences the first feature. The micro-feature with the highest importance among the macro-features of the target is used as the first... Microscopic features, totaling The first Microscopic features, which constitute the first Microscopic feature dataset Z represents the number of single-component cryoprotectant solution systems constructed in step S100;

[0032] Construct the first A machine learning model for macroscopic features of a target, in order to and For the first A machine learning model based on the macroscopic features of the target is trained to obtain a preliminary prediction model. Among them, Each microscopic feature in the middle is used as the first Input to a machine learning model for a target macroscopic feature The corresponding macroscopic features of the target are used as labels to output the first... The predicted value of each target macroscopic characteristic; if and If the performance difference exceeds a set threshold, then step S233 is executed; if and The performance difference did not exceed the set threshold, or At that time, the preliminary prediction model obtained will be used. As the first Target Macro-characteristic Prediction Model And will input the first Target Macro-characteristic Prediction Model The microscopic features as a target for the first Evaluation characteristics of the macro-level features of the target;

[0033] Step S233, let Return to step S232.

[0034] In some embodiments, the prediction of toxicity for all cryoprotectants to be screened, to obtain predicted toxicity values, includes:

[0035] By inputting the component and concentration information of the cryoprotectant solution to be screened into the toxicity prediction model, a toxicity prediction value is obtained. The toxicity prediction model is obtained by mathematically modeling the cytotoxicity value of the cryoprotectant solution based on experimental measurements.

[0036] A second aspect of the present invention provides a screening device for cryoprotectants, the screening device being used to screen from a variety of cryoprotectants to be screened those whose target macroscopic characteristics and toxicity prediction values ​​both meet the requirements, wherein the number of target macroscopic characteristics is at least one; the screening device includes:

[0037] The dataset construction module is configured to construct single-component cryoprotectant solution systems containing different types and concentrations, perform molecular dynamics simulations to calculate various microscopic features of each solution and the target macroscopic features, construct a total microscopic feature dataset using all microscopic features, and construct corresponding target macroscopic feature datasets for each target macroscopic feature.

[0038] The evaluation feature and prediction model acquisition module is configured to determine the evaluation features for each target macro feature from the total micro feature dataset based on clustering and importance analysis, combined with the corresponding target macro feature dataset and machine learning, and obtain the prediction model for each target macro feature.

[0039] The calculation module is configured to calculate the evaluation characteristics of all cryoprotectants to be screened;

[0040] The prediction module is configured to input the evaluation characteristics of all cryoprotectants to be screened into the corresponding target macroscopic feature prediction model to obtain the predicted value of each target macroscopic feature; and to predict the toxicity of all cryoprotectants to be screened to obtain the toxicity prediction value.

[0041] The screening module is configured to select cryoprotectants that meet the requirements for all target macroscopic characteristic prediction values ​​and toxicity prediction values ​​as screening results.

[0042] A third aspect of the present invention provides an electronic device comprising:

[0043] At least one processor, and a memory communicatively connected to said at least one processor;

[0044] The memory stores instructions executable by the at least one processor, the instructions being configured to perform the screening method described in any embodiment of the first aspect of the present invention.

[0045] A fourth aspect of the present invention provides a computer-readable storage medium storing computer instructions for causing the computer to perform the screening method described in any embodiment of the first aspect of the present invention.

[0046] Compared with existing trial-and-error methods, the present invention has the following characteristics and beneficial effects:

[0047] This invention utilizes a machine learning model to efficiently and accurately predict the target macroscopic characteristics of a cryoprotectant solution based on a few microscopic evaluation features, thereby screening out cryoprotectant formulations that meet requirements for water activity, glass transition temperature, and toxicity. This method is characterized by wide applicability, high efficiency, accuracy, and low cost. The screened cryoprotectants exhibit strong ice crystal inhibition capabilities and low toxicity. Attached Figure Description

[0048] Figure 1 This is a flowchart of a method for screening cryogenic protectant components provided in the first aspect of the present invention.

[0049] Figure 2 This is a comparison between the predicted and experimentally measured values ​​of the water activity of the cryoprotectant solution to be screened in this embodiment of the invention.

[0050] Figure 3 This is a comparison between the predicted and simulated values ​​of the glass transition temperature of the cryoprotectant solution to be screened in this invention embodiment.

[0051] Figure 4 This is a schematic diagram of the structure of an electronic device provided in the third aspect of the present invention. Detailed Implementation

[0052] To make the objectives, technical solutions, and advantages of this application clearer, the application will be described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely for explaining this application and are not intended to limit this application.

[0053] Conversely, this application covers any alternatives, modifications, equivalent methods, and schemes made within the spirit and scope of this application as defined by the claims. Furthermore, to provide the public with a better understanding of this application, certain specific details are described in detail below. However, this application can be fully understood by those skilled in the art even without these detailed descriptions.

[0054] The first aspect of this invention provides a method for screening cryoprotectants, which are used in the cryopreservation of biological materials. The screening method for the cryoprotectant is as follows: Figure 1 As shown, this method is used to rapidly and accurately screen cryoprotectants (which can be single-component or multi-component solutions) from a variety of cryoprotectants to be screened, selecting those whose target macroscopic characteristics and toxicity prediction values ​​both meet the requirements. The number of target macroscopic characteristics of the cryoprotectant can be one or more, determined according to the actual application scenario. These mainly include the water activity, glass transition temperature, ice nucleation rate, and ice growth rate of the cryoprotectant solution. In the following description, water activity and glass transition temperature are selected as target macroscopic characteristics. This screening method includes the following steps:

[0055] Step S100: Construct single-component cryoprotectant solution systems containing different types and concentrations, perform molecular dynamics simulations to calculate various microscopic features and target macroscopic features of each solution, and construct a total microscopic feature dataset using all microscopic features. For each target macro-feature, a corresponding target macro-feature dataset is constructed, denoted as the first... The target macroscopic feature dataset is as follows: , , The number of target macroscopic features. ;

[0056] Step S200: Based on clustering and importance analysis, and combined with the corresponding target macro feature dataset and machine learning, determine the evaluation features for each target macro feature from the total micro feature dataset, and obtain the prediction model for each target macro feature.

[0057] Step S300: Calculate the evaluation characteristics of all cryoprotectants to be screened;

[0058] Step S400: Input the evaluation characteristics of all cryoprotectants to be screened into the corresponding target macroscopic feature prediction model to obtain the predicted value of each target macroscopic feature; predict the toxicity of all cryoprotectants to be screened to obtain the toxicity prediction value.

[0059] Step S500: Select the cryoprotectants that meet the requirements for all target macroscopic characteristic prediction values ​​and toxicity prediction values ​​as the screening results.

[0060] In some embodiments, step S100 specifically includes:

[0061] Step S110, Construction (In this embodiment) ( ) Different types and concentrations of single-component cryoprotectant solution systems, specifically, the selected The single-component cryoprotectant solution system contains 41 kinds of small organic molecules, including alcohols (e.g., ethylene glycol), acids (e.g., acetic acid), aldehydes (e.g., acetaldehyde), sulfoxides (e.g., dimethyl sulfoxide), amines (e.g., ethylamine), amides (e.g., formamide), and ketones (e.g., acetone). Each type of cryoprotectant is set with four concentrations of 5%, 10%, 15%, and 20% molar fraction.

[0062] Step S120, Molecular Dynamics Simulation

[0063] Molecular dynamics simulations were performed on each single-component cryoprotectant solution system. Specifically, the force fields of water molecules and cryoprotectant molecules in each system were described using SPC / E (Extended Simple Point Charge) and CHARMM (Chemistry at HARvard Macromolecular Mechanics) models, respectively. The simulations yielded the water activity, glass transition temperature, and various microscopic characteristics of the solution. Water activity showed a quantitative relationship with melting temperature and nucleation temperature, serving as an indicator of the cryoprotectant solution's ability to inhibit nucleation. This was achieved by setting a specific temperature... (In this embodiment) The isothermal simulation of solution configuration evolution was performed, and the density fluctuations of water molecules and cryoprotectant molecules in the simulation space were statistically analyzed. Based on the Kirkwood-Buff theory (KB theory), the water activity of the solution can be calculated, as shown below:

[0064]

[0065]

[0066] in, Indicates two KB integral value between different types of molecules express Types of molecules and KB integral value between different types of molecules Indicates two KB integral values ​​between different types of molecules; This represents the volume of the solution within the selected simulation space; and It is volume internal Types of molecules and The number of types of molecules, and Both can represent water and cryoprotectants; It is the Kronecker function, when and If they are the same, the Kronecker function value is 1; otherwise, it is 0. yes The activity of different types of molecules, such as when When representing water, it indicates water activity. yes Mole fraction of type of molecule; yes Number density of different types of molecules; This represents the average value operator. Considering that the molecular coordinates will change at different times in the simulation, and the number of the two types of molecules will fluctuate within the same volume, it is necessary to calculate the average value.

[0067] Using cooling simulation to obtain the solution from temperature (In this embodiment) ) at cooling rate (In this embodiment) Cool down to temperature (In this embodiment) Different temperatures during the process Enthalpy of the simulated solution system The glass transition temperature was obtained by fitting the trend of enthalpy change with temperature. The specific fitting equation is as follows:

[0068]

[0069] in, and These are the fitting parameters.

[0070] By setting the temperature The isothermal simulation system configuration was used to calculate the cryoprotectant solution for each single component. (In this embodiment) The system utilizes several microscopic features, including radial distribution function information between water and oxygen atoms, radial distribution function information between water and oxygen atoms and hydrophilic atoms of the cryoprotectant, angular distribution function between water and oxygen atoms, water bond order parameter, water fifth nearest neighbor distance, water orientation tetrahedral number, horizontally shifted tetrahedral number, water local structure index, translational order of the water second shell, water tetrahedral entropy, water surface area, water volume, water sphericity, solution density, cryoprotectant cluster ratio, water diffusion coefficient, water hydroxyl rotation time, water dipole moment rotation time, water relaxation time, number of hydrogen bonds between water molecules, length of hydrogen bonds between water molecules, angle of hydrogen bonds between water molecules, lifetime of hydrogen bonds between water molecules, number of hydrogen bonds between water and cryoprotectant molecules, length of hydrogen bonds between water and cryoprotectant molecules, angle of hydrogen bonds between water and cryoprotectant molecules, lifetime of hydrogen bonds between water and cryoprotectant molecules, and total number of hydrogen bonds in the solution. These microscopic parameters comprehensively describe the structure, dynamics, and intermolecular interactions of the solution.

[0071] Step S130: Dataset Construction

[0072] A dataset of water activity for all single-component cryoprotectant solutions was constructed based on the results of molecular dynamics simulations. (This dataset contains) Water activity data for a cryoprotectant solution system, namely The dimension is Glass transition temperature dataset (This dataset contains) The glass transition temperature data for a cryoprotectant solution system, namely The dimension is ) and the total micro-feature dataset (This dataset contains) A cryoprotectant solution system Micro-feature data, namely The dimension is ).

[0073] Understandably, step S100 uses molecular dynamics simulations to establish a comprehensive dataset of organic small molecule cryoprotectants. This not only efficiently obtains macroscopic features describing the performance of solutions in inhibiting ice crystal formation, namely water activity and glass transition temperature, but also calculates microscopic structural, dynamic, and interaction features that are difficult to obtain at the experimental level. These comprehensive microscopic features help to establish an intrinsic link with macroscopic ice-inhibiting performance and achieve efficient and accurate prediction of water activity and glass transition temperature.

[0074] In some embodiments, step S200 includes:

[0075] Step S210: Use clustering methods to analyze the overall micro-feature dataset. Contains Classifying by microscopic features, traversing all classification cases, a total of The classification results will The classification results are sorted in ascending order according to the number of categories.

[0076] Step S220: Use a machine learning model combined with the additivity feature interpretation method to... The importance of micro-features to the macro-features of each target is analyzed to obtain the importance analysis results;

[0077] Step S230, regarding the first obtained in step S210 Classification results, initialization Select the feature class that affects the first The most important micro-feature among the macro-features of the target is the first... Microscopic features, all the first Microscopic features constitute the first Microscopic feature dataset ;use and the first constructed in step S100 A target macroscopic feature dataset For the first A machine learning model based on the macroscopic features of the target is trained to obtain a preliminary prediction model. ,like and If the performance difference exceeds a set threshold, then... Repeat this step until... and The performance difference did not exceed the set threshold, and the current preliminary prediction model was used. As the first Target Macro-characteristic Prediction Model And will input the first Target Macro-characteristic Prediction Model The microscopic features as a target for the first Evaluation characteristics of the macroscopic features of the target.

[0078] Furthermore, in step S210, due to the calculation Among the micro-features, some features are highly correlated, resulting in feature redundancy. Therefore, all micro-features need to be classified according to their degree of correlation. Preferably, hierarchical clustering is used. Classify based on microscopic features, traversing all classification cases, thus obtaining the results from... arrive The total Classification results ,Will According to the classification results Sort the values ​​in ascending order; specifically:

[0079] Step S211: Traverse and calculate microscopic features and Correlation coefficient between , , In this embodiment, the maximum information coefficient is used as the correlation coefficient. ,based on Obtain the correlation matrix among all micro-features ,by As an assessment of micro-characteristics and Distance coefficient representing the degree of correlation between elements;

[0080] Step S212, to Each micro-feature in the set of micro-features is initialized into a class, i.e., initialized... ;

[0081] Step S213: Based on the distance coefficient, merge the two micro-feature categories with the smallest values ​​into a new micro-feature category; recalculate the distance between the new micro-feature category and other micro-feature categories using the connection criterion (Ward method is used as the connection criterion in this embodiment), and update the distance coefficient;

[0082] Step S214, Judgment Is it equal to 1? Then let Return to step S213, if If the classification analysis ends, proceed to step S220.

[0083] It should be noted that hierarchical clustering methods can obtain arbitrary clustering results. The classification results of each micro-feature category are useful for subsequent steps.

[0084] Furthermore, in step S220, due to the calculation The influence of each microscopic feature on the water activity and glass transition temperature of the solution varies. A machine learning model combined with an additivity feature interpretation method is used to analyze the importance of these microscopic features; specifically:

[0085] Step S221: Construct the first and second machine learning models respectively. Each machine learning model preferably uses a random forest model, including an input layer, a sampling layer, a decision tree layer, and an ensemble output layer. Random forest will also be used for machine learning models in subsequent embodiments. Utilize the total micro-feature dataset constructed in step S100. and water activity dataset The first machine learning model is trained, where each iteration uses a solution... These micro-features were used as input to the first machine learning model, using the water activity dataset. The corresponding solution water activity is used as the label, and the predicted water activity value of the solution is used as the output of the first machine learning model; the total microscopic feature dataset constructed in step S100 is used. Glass transition temperature dataset The second machine learning model is trained, where each iteration uses a solution... The microscopic feature data is used as input to the second machine learning model, with the glass transition temperature dataset as the input. The glass transition temperature of the solution is used as the label, and the predicted glass transition temperature of the solution is used as the output of the second machine learning model.

[0086] Step S222: Calculate each step separately. The importance of each micro-feature to the output of the first and second machine learning models is calculated, and all micro-features are ranked from highest to lowest importance to the outputs of the first and second machine learning models, respectively. Here, denoted as _micro_feature_, _i_, _j_, _j_, and _j_, respectively, are considered. For the first The importance of the output of each machine learning model is The calculation, using the additivity feature interpretation method, is specifically represented as follows:

[0087]

[0088] Among them, subscript express The first of the cryoprotectant solutions A solution, , It is the first A solution A collection of micro-features It is the first The solution does not contain microscopic features. The set of all microscopic features It is the first The machine learning model is for the first The output results of each solution Microscopic features For the first The machine learning model is for the first The marginal contribution of the output of a solution, The abs operator represents the number of elements in a set. This indicates taking the absolute value.

[0089] It should be noted that there is no order in which steps S210 and S220 are executed; they can be performed simultaneously or in reverse order.

[0090] Further, in step S230, based on the results of hierarchical clustering and feature importance analysis, a few important but weakly correlated microscopic features that determine the water activity and glass transition temperature of the solution are selected as evaluation features, and a prediction model for each target macroscopic feature is obtained simultaneously. ;Specifically:

[0091] Step S231, Initialization This corresponds to the first classification result obtained in step S210;

[0092] Step S232, for the first From the classification results, the micro-feature with the highest importance to water activity in each feature class was selected as the first micro-feature, totaling [number missing]. Each of the first micro-features is used to construct the first micro-feature dataset. ; and select the first In each feature class of the classification results, the micro-feature with the highest importance affecting the glass transition temperature is selected as the second micro-feature, totaling [number missing]. Each of the second micro-features constitutes a second micro-feature dataset. ;

[0093] Construct a machine learning model for water activity using the first microscopic feature dataset. and water activity dataset The machine learning model for water activity was trained and tested (the training and testing sets were divided in a 7:3 ratio) to obtain a preliminary prediction model. Among them, the water activity dataset is used as the input to the machine learning model for water activity, which uses the first microscopic feature as the input. The corresponding solution water activity is used as a label, and the predicted water activity value is output; calculation performance In this embodiment, it is used in The predicted water activity value of a single-component cryoprotectant solution system and its root mean square error relative to its label are used as... ;like and The difference is greater than the first set threshold (for At that time, take If ), then proceed to step S233; if and The difference is less than or equal to the first set threshold, or At that time, the preliminary prediction model obtained will be used. As a water activity prediction model And input into the prediction model The microscopic characteristics are used as assessment features for water activity;

[0094] Similarly, a machine learning model for the glass transition temperature was constructed using a second microscopic feature dataset. Glass transition temperature dataset A machine learning model for glass transition temperature was trained and tested (the training and testing sets were divided in a 7:3 ratio) to obtain a preliminary prediction model. Among them, the machine learning model for glass transition temperature uses the second microscopic feature as input, and the glass transition temperature dataset... The corresponding glass transition temperature of the solution is used as a label, and the predicted glass transition temperature is output; calculation performance In this embodiment, it is used in The root mean square error between the predicted glass transition temperature and its label in a single-component cryoprotectant solution system is used as... ;like and The difference is greater than the second set threshold (for At that time, take If ), then proceed to step S233; if and The difference is less than or equal to the second set threshold, or At that time, the preliminary prediction model obtained will be used. As a model for predicting glass transition temperature And input into the prediction model The microscopic characteristics are used as evaluation features for glass transition temperature;

[0095] Step S233, let Return to step S232.

[0096] In this embodiment, after step S200, the first evaluation features finally determined are the horizontally shifted tetrahedral ordinal number, water diffusion coefficient, and the first peak value of the radial distribution function of water oxygen atoms and hydrophilic atoms of the cryoprotectant. The model with these three microscopic features as input in step S232 is used to predict the water activity of the cryoprotectant solution to be screened. The second evaluation features finally determined are the hydrogen bond lifetime between water and cryoprotectant molecules, water diffusion coefficient, and water sphericity. The model with these three microscopic features as input in step S232 is used to predict the glass transition temperature of the cryoprotectant solution to be screened.

[0097] It should be noted that the number of evaluation features for each target's macroscopic characteristics determined in step S200 may be equal or unequal. The specific values ​​are determined based on the number of microscopic features input to each prediction model when the performance of the water activity prediction model and the glass transition temperature prediction model meets the requirements.

[0098] Understandably, step S200, which combines hierarchical clustering and feature importance, reveals the structure-activity relationship between microscopic features and macroscopic anti-icing performance, and clarifies three important but weakly correlated evaluation features that affect solution water activity and glass transition temperature. This reduces the computational cost of the microscopic feature dataset while maintaining the accuracy of machine learning models in predicting water activity and glass transition temperature.

[0099] In some embodiments, step S300 specifically includes:

[0100] Step S310: Construct a cryoprotectant solution system to be screened. Specifically, in this embodiment, the solution to be screened is a mixture system of three cryoprotectants, dimethyl sulfoxide, formamide and glycerol, and water in different proportions. A single-component cryoprotectant solution system can also be used.

[0101] Step S320: Perform molecular dynamics simulation on the cryoprotectant solution system to be screened; wherein, by isothermally simulating the system configuration at a set temperature T0, the microscopic features to be evaluated are obtained. For the water activity prediction model, the features to be evaluated are the horizontally shifted tetrahedral number, water diffusion coefficient, and the first peak value of the radial distribution function of water oxygen atoms and hydrophilic atoms of cryoprotectant. For the glass transition temperature prediction model, the features to be evaluated are the hydrogen bond lifetime between water and cryoprotectant molecules, water diffusion coefficient, and water sphericity.

[0102] In some embodiments, step S400 specifically includes:

[0103] Step S410: Predict the water activity of the cryoprotectant solution to be screened; specifically, the water activity-related evaluation features of the cryoprotectant solution to be screened calculated in step S320, namely the horizontally shifted tetrahedral ordinal number, water diffusion coefficient, and the first peak value of the radial distribution function of water oxygen atoms and hydrophilic atoms of the cryoprotectant, are input into step S232. The predicted water activity value is obtained from the well-trained water activity prediction model.

[0104] Prediction of the glass transition temperature of the cryoprotectant solution to be screened; specifically, the evaluation characteristics related to the glass transition temperature of the cryoprotectant solution to be screened, calculated in step S320, namely the hydrogen bond lifetime between water and cryoprotectant molecules, water diffusion coefficient, and water sphericity, are input into step S232. The predicted glass transition temperature is obtained from the trained glass transition temperature prediction model.

[0105] Step S420: Toxicity prediction of the cryoprotectant solution to be screened; specifically, the component and concentration information of the cryoprotectant solution to be screened are input into the toxicity prediction model (Warner RM et al. Multiple cryoprotectant toxicity model for vitrification solution optimization. Cryobiology, 2022,108: 1-9, which is a mathematical model based on experimentally measured cytotoxicity values ​​of cryoprotectant solutions) to obtain the toxicity prediction value.

[0106] Understandably, step S400 can simultaneously and efficiently predict the water activity, glass transition temperature, and toxicity of unknown cryoprotectant solutions, enabling high-throughput calculations of a large number of cryoprotectant formulations and effectively reducing costs compared to experimental methods for measuring the ice-suppressing performance and toxicity of solutions.

[0107] In some embodiments, step S500 specifically includes:

[0108] Step S510: Screening out cryoprotectant solutions that meet the requirements for predicted water activity value, glass transition temperature value, and toxicity value; wherein, the predicted water activity value... It needs to be lower than the set value Glass transition temperature value It needs to be higher than the set value toxicity value It needs to be lower than the set value In the examples, the performance of the existing hybrid cryoprotectant VMP (6.44 m dimethyl sulfoxide + 6.44 m formamide + 6.12 m ethylene glycol) was used as a reference standard. , , .

[0109] Understandably, step S500 can quickly screen out cryogenic protectant components and ratios with low water activity, high glass transition temperature and low toxicity, which is simple and efficient compared to the traditional empirical trial and error method.

[0110] The following is a verification of the effectiveness of the embodiments of the present invention:

[0111] In this embodiment of the invention, the screening target is selected to be superior to the performance of existing cryoprotectant solutions VMP, namely, a water activity value below 0.678, a glass transition temperature above 201.9 K, and a toxicity value below 30.14 min. -1 In this embodiment, five cryoprotectant solutions to be screened were calculated, with the following formulations:

[0112] (1) TH-CPA1: 8 m dimethyl sulfoxide + 4 m formamide + 6 m glycerol;

[0113] (2) TH-CPA2: 7 m dimethyl sulfoxide + 4 m formamide + 7 m glycerol;

[0114] (3) TH-CPA3: 6 m dimethyl sulfoxide + 4 m formamide + 8 m glycerol;

[0115] (4) TH-CPA4: 8 m dimethyl sulfoxide + 3 m formamide + 7 m glycerol;

[0116] (5) TH-CPA5: 7 m dimethyl sulfoxide + 3 m formamide + 8 m glycerol;

[0117] The evaluation characteristics of the cryoprotectant solution to be screened are simulated by molecular dynamics. Three evaluation characteristics—the horizontal shift tetrahedral number of the solution, the water diffusion coefficient, and the first peak value of the radial distribution function of water and oxygen atoms and the hydrophilic atoms of the cryoprotectant—are input into step S232. The predicted water activity value is obtained from the trained water activity prediction model. The comparison between the predicted value and the experimental value is as follows: Figure 2 As shown, all solutions except solution (3) meet the water activity requirements; the three evaluation characteristics of the solution, namely the hydrogen bond lifetime between water and cryoprotectant molecules, water diffusion coefficient, and water sphericity, are input into step S232. The predicted glass transition temperature (GLT) value is obtained from the trained glass transition temperature prediction model. The comparison between the predicted and simulated values ​​is as follows: Figure 3 As shown, all solutions meet the glass transition temperature requirement; the components and concentration parameters of the solutions to be screened are input into the toxicity prediction model to obtain the predicted toxicity values ​​of the solutions. The toxicity values ​​of solutions (1)-(5) are 1.55 min. -1 1.13 min -1 0.86 min -1 1.93 min -1 and 1.53 min -1 All solutions met the toxicity requirements. It can be seen that this screening method can obtain cryoprotectant formulations that meet the requirements for water activity, glass transition temperature, and toxicity. This result demonstrates the effectiveness of the screening method of this invention.

[0118] In summary, this invention proposes a method for screening cryoprotectants, which can efficiently and accurately screen cryoprotectants that meet the requirements for water activity, glass transition temperature, and predicted toxicity.

[0119] A second aspect of the present invention provides a screening device for cryoprotectants, comprising:

[0120] The dataset construction module is configured to construct single-component cryoprotectant solution systems containing different types and concentrations, perform molecular dynamics simulations to calculate various microscopic features of each solution and the target macroscopic features, construct a total microscopic feature dataset using all microscopic features, and construct corresponding target macroscopic feature datasets for each target macroscopic feature.

[0121] The evaluation feature and prediction model acquisition module is configured to determine the evaluation features for each target macro feature from the total micro feature dataset based on clustering and importance analysis, combined with the corresponding target macro feature dataset and machine learning, and obtain the prediction model for each target macro feature.

[0122] The calculation module is configured to calculate the evaluation characteristics of all cryoprotectants to be screened;

[0123] The prediction module is configured to input the evaluation characteristics of all cryoprotectants to be screened into the corresponding target macroscopic feature prediction model to obtain the predicted value of each target macroscopic feature; and to predict the toxicity of all cryoprotectants to be screened to obtain the toxicity prediction value.

[0124] The screening module is configured to select cryoprotectants that meet the requirements for all target macroscopic characteristic prediction values ​​and toxicity prediction values ​​as screening results.

[0125] It should be noted that the foregoing explanation of the screening method for cryoprotectants also applies to the screening device for cryoprotectants in this embodiment, and will not be repeated here.

[0126] To implement the above embodiments, this invention also proposes a computer-readable storage medium storing a computer program that is executed by a processor to perform the filtering method described in the above embodiments.

[0127] The following is for reference. Figure 4 The diagram illustrates a structural schematic of an electronic device suitable for implementing embodiments of the present invention. It should be noted that the electronic device in the embodiments of the present invention may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), and in-vehicle terminals (e.g., in-vehicle navigation terminals), as well as fixed terminals such as digital TVs, desktop computers, and servers. Figure 4 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of the present invention.

[0128] like Figure 4As shown, the electronic device may include a processing unit (e.g., a central processing unit, a graphics processing unit, etc.) 101, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 102 or a program loaded from a storage device 108 into a random access memory (RAM) 103. The RAM 103 also stores various programs and data required for the operation of the electronic device. The processing unit 101, ROM 102, and RAM 103 are interconnected via a bus 104. An input / output (I / O) interface 105 is also connected to the bus 104.

[0129] Typically, the following devices can be connected to I / O interface 105: input devices 106 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, etc.; output devices 107 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 108 including, for example, magnetic tapes, hard disks, etc.; and communication devices 109. Communication device 109 allows electronic devices to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 4 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown. More or fewer devices may be implemented or have alternatively.

[0130] In particular, according to embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, this embodiment includes a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such an embodiment, the computer program can be downloaded and installed from a network via communication device 109, or installed from storage device 108, or installed from ROM 102. When the computer program is executed by processing device 101, it performs the functions defined above in the methods of embodiments of this disclosure.

[0131] It should be noted that the computer-readable medium described above in this invention can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this disclosure, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this disclosure, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0132] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device.

[0133] The aforementioned computer-readable medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to perform the aforementioned screening method.

[0134] Computer program code for performing the operations of this disclosure can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and Python, as well as conventional procedural programming languages ​​such as the "C-" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0135] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0136] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0137] Any process or method described in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing a particular logical function or process, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the function involved, as will be understood by those skilled in the art to which embodiments of this application pertain.

[0138] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which programs can be printed, because programs can be obtained electronically, for example, by optically scanning the paper or other media, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0139] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0140] Those skilled in the art will understand that implementing all or part of the steps of the methods in the above embodiments can be accomplished by instructing related hardware through a program. The developed program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0141] Furthermore, the functional units in the various embodiments of this application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0142] The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc. Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of this application.

Claims

1. A method for screening cryoprotectants, characterized in that, The screening method is used to select target macroscopic characteristics from a variety of cryoprotectants to be screened. The cryoprotectant whose toxicity prediction values ​​all meet the requirements, and the number of the target macroscopic characteristics is at least one; the screening method includes: Step S100: Construct a single-component cryoprotectant solution system containing different types and concentrations, perform molecular dynamics simulations to calculate various microscopic features of each solution and the target macroscopic features, construct a total microscopic feature dataset using all microscopic features, and construct a corresponding target macroscopic feature dataset for each target macroscopic feature. Step S200: Based on clustering and importance analysis, and combined with the corresponding target macro feature dataset and machine learning, determine the evaluation features for each target macro feature from the total micro feature dataset, and obtain the prediction model for each target macro feature. Step S300: Calculate the evaluation characteristics of all cryoprotectants to be screened; Step S400: Input the evaluation characteristics of all cryoprotectants to be screened into the corresponding target macroscopic feature prediction model to obtain the predicted value of each target macroscopic feature; predict the toxicity of all cryoprotectants to be screened to obtain the toxicity prediction value. Step S500: Select the cryoprotectants that meet the requirements for all target macroscopic characteristic prediction values ​​and toxicity prediction values ​​as the screening results.

2. The screening method according to claim 1, characterized in that, The target macroscopic characteristics of the cryoprotectant include at least one of water activity, glass transition temperature, ice nucleation rate, and ice growth rate; and / or The water activity was obtained by isothermal simulation of the configurational evolution of the solution system, and the glass transition temperature was obtained by simulating the enthalpy of the solution system at different temperatures through cooling; and / or The total microscopic feature dataset contains multiple microscopic features, including radial distribution function information between water and oxygen atoms, radial distribution function information between water and oxygen atoms and hydrophilic atoms of the cryoprotectant, angular distribution function between water and oxygen atoms, water bond order parameter, water fifth nearest neighbor distance, water orientation tetrahedral ordinal number, horizontally shifted tetrahedral ordinal number, water local structure index, translational order of water second shell, water tetrahedral entropy, water surface area, water volume, water sphericity, solution density, cryoprotectant cluster ratio, water diffusion coefficient, water hydroxyl rotation time, water dipole moment rotation time, water relaxation time, number of hydrogen bonds between water molecules, length of hydrogen bonds between water molecules, angle of hydrogen bonds between water molecules, lifetime of hydrogen bonds between water molecules, number of hydrogen bonds between water and cryoprotectant molecules, length of hydrogen bonds between water and cryoprotectant molecules, angle of hydrogen bonds between water and cryoprotectant molecules, lifetime of hydrogen bonds between water and cryoprotectant molecules, and the total number of hydrogen bonds in the solution.

3. The screening method according to claim 1, characterized in that, Step S200 includes: Step S210: Use clustering methods to analyze the micro-features contained in the total dataset. Classifying by microscopic features, traversing all classification cases, a total of The classification results will The classification results are sorted in ascending order according to the number of categories. Step S220: Use a machine learning model combined with the additivity feature interpretation method to... The importance of micro-features to the macro-features of each target is analyzed to obtain the importance analysis results; Step S230: For the Nth classification result obtained in step S210, initialize... Select the feature class that affects the first The most important micro-feature among the macro-features of the target is the first... Microscopic features, all the l-th microscopic features constitute the l-th microscopic features. Microscopic feature dataset ;use and the first constructed in step S100 A target macroscopic feature dataset For the A machine learning model based on the macroscopic features of the target is trained to obtain a preliminary prediction model. ,like and If the performance difference exceeds a set threshold, then... Repeat this step until... and The performance difference did not exceed the set threshold, and the current preliminary prediction model was used. As the first Target Macro-characteristic Prediction Model And will input the first Target Macro-characteristic Prediction Model The microscopic features as a target for the first Evaluation characteristics of the macroscopic features of the target.

4. The screening method according to claim 3, characterized in that, Step S210 includes: Step S211: Traverse and calculate microscopic features and Correlation coefficient between , , ,based on Obtain the correlation matrix among all micro-features ,by As an assessment of micro-characteristics and Distance coefficient representing the degree of correlation between elements; Step S212, to Each micro-feature in the set of micro-features is initialized into a class, i.e., initialized... ; Step S213: Based on the distance coefficient, merge the two micro-feature categories with the smallest values ​​into a new micro-feature category; recalculate the distance between the new micro-feature category and other micro-feature categories using the connection criterion, and update the distance coefficient; Step S214, Judgment Is it equal to 1? Then let Return to step S213, if Then the classification analysis ends.

5. The screening method according to claim 3, characterized in that, Step S220 includes: Step S221: For each target macroscopic feature, construct a machine learning model respectively, using the total microscopic feature dataset constructed in step S100. and the A target macroscopic feature dataset For the Several machine learning models are trained, with each iteration using a solution... The micro-feature is the first The input of a machine learning model, with The corresponding macroscopic feature data of the solution are used as labels, with the solution's first... The predicted value of the target macroscopic characteristic is used as the first... The output of a machine learning model; Step S222: Calculate each step separately. The importance of each micro-feature to the output of each machine learning model is calculated, and all micro-features are ranked from highest to lowest importance to the output of each machine learning model; where, let the micro-feature be denoted as . For the The importance of the output of each machine learning model is The calculation is performed using the additivity feature interpretation method: Among them, subscript This indicates the construction in step S100. The first single-component cryoprotectant solution system A solution, , It is the first A solution A collection of micro-features It is the first The solution does not contain microscopic features. The set of all microscopic features It is the first The machine learning model is for the first The output results of each solution Microscopic features For the The machine learning model is for the first The marginal contribution of the output of each solution. This is the operator for counting the number of elements in a set.

6. The screening method according to claim 3, characterized in that, Step S230 includes: Step S231, Initialization ; Step S232, for the first For each classification result, select the feature class that influences the first feature. The micro-feature with the highest importance among the macro-features of the target is used as the first... Microscopic features, totaling The first Microscopic features, which constitute the first Microscopic feature dataset Z represents the number of single-component cryoprotectant solution systems constructed in step S100; Construct the first A machine learning model for macroscopic features of a target, in order to and For the A machine learning model based on the macroscopic features of the target is trained to obtain a preliminary prediction model. Among them, Each microscopic feature in the middle is used as the first Input to a machine learning model for a target macroscopic feature The corresponding macroscopic features of the target are used as labels to output the first... The predicted value of each target macroscopic characteristic; if and If the performance difference exceeds a set threshold, then step S233 is executed; if and The performance difference did not exceed the set threshold, or At that time, the preliminary prediction model obtained will be used. As the first Target Macro-characteristic Prediction Model And will input the first Target Macro-characteristic Prediction Model The microscopic features as a target for the first Evaluation characteristics of the macro-level features of the target; Step S233, let Return to step S232.

7. The screening method according to claim 1, characterized in that, The toxicity prediction of all the cryoprotectants to be screened, and the resulting toxicity prediction values, include: By inputting the component and concentration information of the cryoprotectant solution to be screened into the toxicity prediction model, a toxicity prediction value is obtained. The toxicity prediction model is obtained by mathematically modeling the cytotoxicity value of the cryoprotectant solution based on experimental measurements.

8. A screening device for cryogenic protectants, characterized in that, The screening device is used to screen out target macroscopic characteristics from a variety of cryoprotectants to be screened. The cryoprotectant whose toxicity prediction values ​​all meet the requirements, and the number of the target macroscopic characteristics is at least one; the screening device includes: The dataset construction module is configured to construct single-component cryoprotectant solution systems containing different types and concentrations, perform molecular dynamics simulations to calculate various microscopic features of each solution and the target macroscopic features, construct a total microscopic feature dataset using all microscopic features, and construct corresponding target macroscopic feature datasets for each target macroscopic feature. The evaluation feature and prediction model acquisition module is configured to determine the evaluation features for each target macro feature from the total micro feature dataset based on clustering and importance analysis, combined with the corresponding target macro feature dataset and machine learning, and obtain the prediction model for each target macro feature. The calculation module is configured to calculate the evaluation characteristics of all cryoprotectants to be screened; The prediction module is configured to input the evaluation characteristics of all cryoprotectants to be screened into the corresponding target macroscopic feature prediction model to obtain the predicted value of each target macroscopic feature; and to predict the toxicity of all cryoprotectants to be screened to obtain the toxicity prediction value. The screening module is configured to select cryoprotectants that meet the requirements for all target macroscopic characteristic prediction values ​​and toxicity prediction values ​​as screening results.

9. An electronic device, characterized in that, include: At least one processor, and a memory communicatively connected to said at least one processor; The memory stores instructions executable by the at least one processor, the instructions being configured to perform the screening method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to perform the screening method according to any one of claims 1 to 7.

Images