Method and device for processing mRNA vaccine design task

Abstract

The embodiment of the invention relates to an mRNA (messenger ribonucleic acid) vaccine design task processing method and device, and the method comprises the following steps: designing a vaccine characteristic prediction model for predicting three characteristics of an mRNA vaccine by taking a Un i-RNA model as a trunk, and training the model based on an antigen vaccine data set, three objective functions are designed based on the maximum incidence relation between three characteristics of codon use frequency, secondary structure and GC content and vaccine stability, translation efficiency and immunogenicity; carrying out mRNA sequence initialization according to an antigen sequence input by a user, and carrying out iterative optimization processing on an mRNA sequence set according to a multi-objective optimization mode of an NSGA-I algorithm by virtue of a vaccine characteristic prediction model, three objective functions and corresponding in-vitro / in-vivo experiment means to obtain a corresponding optimized mRNA sequence set to be fed back to the user. Through the method, the sequence richness can be improved, the design quality is ensured, the design efficiency is improved, and the design cost is reduced.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a method and device for processing mRNA vaccine design tasks. Background Art

[0002] mRNA vaccines have shown great potential and advantages in the field of immunotherapy, especially in responding to emerging infectious diseases such as the new coronavirus (COVID-19). Compared with traditional vaccines, mRNA vaccines have the characteristics of rapid response, flexibility and efficient production. Therefore, mRNA vaccines have received widespread attention and application during the COVID-19 pandemic. However, despite the many advantages of mRNA vaccine technology, there are still many challenges in designing mRNA sequences with optimal stability (usually characterized by the half-life of mRNA in the body), translation efficiency (usually characterized by the expression of the target antigen protein) and immunogenicity (usually characterized by antibody titer). The most common challenge is the cost and efficiency issue: currently, each mRNA sequence planned for the initial design must pass the in vivo / in vitro experimental analysis process before the next screening and optimization step can be carried out, and each mRNA sequence optimized in each step also needs to pass the in vivo / in vitro experimental analysis before it can be transferred to the next step of operation; such a cycle of iteration not only requires a lot of manpower and material costs for multiple repetitive experiments, but also causes the problem of too long design cycle and inability to improve design efficiency. Summary of the invention

[0003] The purpose of the present invention is to provide a method, device, electronic device and computer-readable storage medium for processing mRNA vaccine design tasks in view of the defects of the prior art. The present invention pre-designs a working model for predicting three vaccine characteristics (stability, translation efficiency and immunogenicity) of mRNA sequences with a pre-trained Uni-RNA model as the feature extraction backbone, and records it as a vaccine characteristic prediction model, and trains this model based on an antigen vaccine data set constructed by a big data acquisition method, and identifies the maximum correlation between the three basic features (codon usage frequency, RNA secondary structure, GC content) that can be predicted by this model and the three vaccine characteristics (stability, translation efficiency and immunogenicity), and determines three objective functions (stability objective function, translation efficiency objective function, immunogenicity objective function) based on the identification results. Then, in the specific mRNA vaccine design task processing process: the present invention first uses the mRNA sequence generation tool to batch generate the transcribed mRNA sequences of the target antigen protein to obtain an initialized mRNA sequence set; then uses the vaccine characteristic prediction model and three objective functions to iteratively optimize and screen the mRNA sequence set according to the multi-objective optimization method of the NSGA-II algorithm to obtain a preferred set with a reduced number of sequences; then based on in vitro experimental means, the stability and translation efficiency of the preferred set are experimentally analyzed; if the error between the in vitro experimental analysis results and the model prediction results is too large, the model is automatically optimized based on the experimental data of this experiment, and the mRNA sequence set is iteratively optimized again after optimization, and if the error between the in vitro experimental analysis results and the model prediction results is controllable, the immunogenicity of the preferred set is further analyzed based on the in vivo experimental means; if the error between the in vivo experimental analysis results and the model prediction results is too large, the model is automatically optimized again based on the experimental data of this experiment, and the mRNA sequence set is iteratively optimized again after optimization, and if the error between the in vivo experimental analysis results and the model prediction results is controllable, the current latest preferred set is output as the final design candidate set. The mRNA sequence initialization method of the present invention can get rid of the limitations of expert experience and provide rich and diverse sequence structures; the vaccine characteristic prediction model of the present invention can reduce a large number of experimental analysis times, shorten the design cycle, and improve design efficiency; the model training method and multi-objective optimization process of the present invention can effectively guarantee the quality of the designed mRNA sequence.

[0004] To achieve the above objectives, the first aspect of the embodiments of the present invention provides a method for processing an mRNA vaccine design task, the method comprising:

[0005] A working model for predicting three vaccine characteristics of mRNA sequences is designed with the Uni-RNA model as the feature extraction backbone, which is recorded as the corresponding vaccine characteristic prediction model; the three vaccine characteristics include stability, translation efficiency and immunogenicity; the vaccine characteristic prediction model includes the Uni-RNA model, three basic feature prediction heads, a feature fusion module and three vaccine characteristic prediction heads;

[0006] A large amount of antigen protein and its corresponding mRNA vaccine information is collected by big data collection to construct a corresponding antigen vaccine data set; and the vaccine characteristic prediction model is trained based on the antigen vaccine data set; and after the model training is completed, the maximum correlation between the three output features of the three basic feature prediction heads and the three output features of the three vaccine characteristic prediction heads is identified and the corresponding three objective functions are determined based on the identification results; the three objective functions include a stability objective function, a translation efficiency objective function and an immunogenicity objective function;

[0007] Receive the molecular sequence of the target antigen protein input by the user as the corresponding first antigen sequence; and initialize the mRNA sequence according to the first antigen sequence to obtain the corresponding first mRNA sequence set; and according to the multi-objective optimization method of the NSGA-II algorithm, with the help of the vaccine characteristic prediction model, the three objective functions and the corresponding in vitro / in vivo experimental means, iteratively optimize the first mRNA sequence set to obtain the corresponding optimized mRNA sequence set to feed back to the user; the first mRNA sequence set consists of multiple first mRNA sequences; the optimized mRNA sequence set consists of one or more optimized mRNA sequences.

[0008] Preferably, the vaccine characteristic prediction model is used to predict the stability, translation efficiency and immunogenicity characteristics of the input mRNA sequence and output corresponding stability prediction data, translation efficiency prediction data and immunogenicity prediction data;

[0009] The vaccine characteristic prediction model includes the Uni-RNA model, the three basic feature prediction heads, the feature fusion module and the three vaccine characteristic prediction heads; the three basic feature prediction heads include a codon usage frequency prediction head, an RNA secondary structure prediction head, and a GC content prediction head; the three vaccine characteristic prediction heads include a stability prediction head, a translation efficiency prediction head and an immunogenicity prediction head;

[0010] The model input end of the vaccine characteristic prediction model is used to receive the corresponding mRNA sequence, and the first model output end, the second model output end and the third model output end are used to output the corresponding stability prediction data, the translation efficiency prediction data and the immunogenicity prediction data respectively;

[0011] The input end of the Uni-RNA model is connected to the input end of the model, and the output end is respectively connected to the input ends of the codon usage frequency prediction head, the RNA secondary structure prediction head and the GC content prediction head; the output ends of the codon usage frequency prediction head, the RNA secondary structure prediction head and the GC content prediction head are connected to the first, second and third input ends of the feature fusion module; the output end of the feature fusion module is respectively connected to the input ends of the stability prediction head, the translation efficiency prediction head and the immunogenicity prediction head; the output ends of the stability prediction head, the translation efficiency prediction head and the immunogenicity prediction head are respectively connected to the corresponding first, second and third model output ends;

[0012] The Uni-RNA model has completed model pre-training; the Uni-RNA model is used to perform feature encoding processing on the input mRNA sequence to obtain a corresponding feature encoding vector and send it to the codon usage frequency prediction head, the RNA secondary structure prediction head and the GC content prediction head;

[0013] The codon usage frequency prediction head is implemented based on a type of nonlinear regression model or a type of neural network model; the codon usage frequency prediction head is used to predict the codon usage frequency of each base position on the mRNA sequence according to the feature coding vector to obtain the corresponding codon usage frequency prediction vector and send it to the feature fusion module; the nonlinear regression model at least includes an MLP model, an NNR model, a GBDT model and an XGBoost model; the neural network model at least includes an MLP model, a CNN model, a ResNet model and a GNN model;

[0014] The RNA secondary structure prediction head is implemented based on another type of neural network model; the RNA secondary structure prediction head is used to predict the secondary structure features of each base position on the mRNA sequence according to the feature encoding vector to obtain the corresponding secondary structure prediction vector and send it to the feature fusion module;

[0015] The GC content prediction head is implemented based on another type of the nonlinear regression model or another type of the neural network model; the GC content prediction head is used to predict the GC content of each base position on the mRNA sequence according to the feature encoding vector to obtain a corresponding GC content prediction vector and send it to the feature fusion module;

[0016] The feature fusion module is used to perform vector splicing on the codon usage frequency prediction vector, the secondary structure prediction vector and the GC content prediction vector to obtain a corresponding splicing vector and send it to the stability prediction header, the translation efficiency prediction header and the immunogenicity prediction header;

[0017] The stability prediction head is implemented based on another type of the nonlinear regression model or another type of the neural network model; the stability prediction head is used to predict the stability characteristics of the mRNA sequence according to the splicing vector and output the corresponding stability prediction data;

[0018] The translation efficiency prediction head is implemented based on another type of the nonlinear regression model or another type of the neural network model; the translation efficiency prediction head is used to predict the translation characteristics of the mRNA sequence according to the splicing vector and output the corresponding translation efficiency prediction data;

[0019] The immunogenicity prediction head is implemented based on another type of the nonlinear regression model or another type of the neural network model; the immunogenicity prediction head is used to predict the immunogenicity characteristics of the mRNA sequence according to the splicing vector and output the corresponding immunogenicity prediction data.

[0020] Preferably, the antigen vaccine data set includes multiple antigen vaccine data records; the antigen vaccine data records include antigen molecule sequence, mRNA molecule sequence, codon usage frequency label vector, secondary structure label vector, GC content label vector, stability label data, translation efficiency label data and immunogenicity label data.

[0021] Preferably, the model training of the vaccine characteristic prediction model based on the antigen vaccine dataset specifically includes:

[0022] Extracting the mRNA molecular sequence, the codon usage frequency label vector, the secondary structure label vector and the GC content label vector of each antigen vaccine data record in the antigen vaccine data set to form a corresponding first data record; and forming a corresponding first data set from all the obtained first data records;

[0023] and extracting the mRNA molecular sequence, the stability label data, the translation efficiency label data and the immunogenicity label data of each of the antigen vaccine data records in the antigen vaccine data set to form a corresponding second data record; and all the obtained second data records form a corresponding second data set;

[0024] Based on the first data set, the Uni-RNA model and the three basic feature prediction heads of the vaccine characteristic prediction model are subjected to the first stage model training; after the first round of model training is completed, the three vaccine characteristic prediction heads of the vaccine characteristic prediction model are subjected to the second stage model training based on the second data set; and after the second round of model training is completed, the completion of this model training is confirmed.

[0025] Furthermore, the first stage model training of the Uni-RNA model and the three basic feature prediction heads of the vaccine characteristic prediction model based on the first data set specifically includes:

[0026] Step 51, based on a preset first segmentation ratio, randomly split the first data set into two sub-data sets recorded as a corresponding first training set and a first evaluation set;

[0027] Wherein, both the first training set and the first evaluation set are composed of a plurality of the first data records; the ratio of the total number of records in the first training set to the total number of records in the first evaluation set satisfies the first segmentation ratio;

[0028] Step 52, extracting the first first data record of the first training set as the corresponding current training record;

[0029] Step 53: record the codon usage frequency label vector, the secondary structure label vector and the GC content label vector of the current training record as the corresponding first label vector P tag1 , the second label vector P tag2 and the third label vector P tag3 ;

[0030] Step 54, input the mRNA molecular sequence of the current training record into the Uni-RNA model for feature encoding processing to obtain a corresponding first feature encoding vector; and input the first feature encoding vector into the codon usage frequency prediction header, the RNA secondary structure prediction header and the GC content prediction header for prediction processing to obtain a corresponding first prediction vector P pre1 , the second prediction vector P pre2 and the third prediction vector P pre3 ;

[0031] Step 55: transform the first prediction vector P pre1 And the corresponding first label vector P tag1 Substitute the preset first loss function L a , and the second prediction vector P pre2 And the corresponding second label vector P tag2 Substitute the preset second loss function L b , and the third prediction vector P pre3 And the corresponding third label vector P tag3 Bring in the preset third loss function L c , and by the first loss function L a 、The second loss function L b And the third loss function Lc Add together to form the corresponding first model loss function L M1 ; and calculate the function values ​​of the first, second, third loss functions and the first model loss function and use the calculation results as the corresponding first loss value, second loss value, third loss value and first overall loss value;

[0032] Among them, the first model loss function L M1 for:

[0033] L M1 =L a (P pre1 ,P tag1 )+L b (P pre2 ,P tag2 )+L c (P pre3 ,P tag3 );

[0034] The first loss function L a At least including L1 loss function and L2 loss function; the second loss function L b It is composed of a base type loss function and a base structure loss function, wherein the base type loss function at least includes a cross entropy loss function, and the base structure loss function at least includes an L1 loss function and an L2 loss function; the third loss function L c At least including L1 loss function and L2 loss function;

[0035] Step 56, identifying whether the first, second, third loss values ​​and the first overall loss value all satisfy the corresponding first loss value range, second loss value range, third loss value range and first overall loss value range; if the first, second, third loss values ​​and the first overall loss value all satisfy the corresponding loss value range, go to step 57; if the first loss value does not satisfy the first loss value range, then based on the preset first model optimizer, move towards making the first loss function L a The model parameters of the Uni-RNA model and the codon usage frequency prediction head are modulated in a round in the direction of reaching the minimum value, and the step 54 is returned to when the parameter modulation of this round is finished; if the second loss value does not satisfy the second loss value range, the second model optimizer is based on the preset second model to make the second loss function L b The model parameters of the Uni-RNA model and the RNA secondary structure prediction head are modulated in a round in the direction of reaching the minimum value, and the step 54 is returned to when the parameter modulation of this round is finished; if the third loss value does not satisfy the third loss value range, the third model optimizer is based on a preset third model to make the third loss function L cThe model parameters of the Uni-RNA model and the GC content prediction head are modulated in a round in the direction of reaching the minimum value, and the step 54 is returned to when the parameter modulation of this round is finished; if the first overall loss value does not satisfy the first overall loss value range, the fourth model optimizer is based on a preset fourth model optimizer in the direction of making the first model loss function L M1 Perform a round of modulation on the model parameters of the Uni-RNA model and the three basic feature prediction heads in the direction of reaching the minimum value, and return to step 54 when this round of parameter modulation is completed;

[0036] Wherein, the first, second, third and fourth model optimizers all include at least an SGD optimizer and an ADAM optimizer;

[0037] Step 57, identifying whether the current training record is the last first data record of the first training set; if so, going to step 58; if not, extracting the next first data record of the first training set as the new current training record and returning to step 53;

[0038] Step 58, perform a round of traversal on all the first data records of the first evaluation set; and in this round of traversal, use the first data record currently traversed as the corresponding current evaluation record; and input the mRNA molecule sequence of the current evaluation record into the Uni-RNA model for feature encoding processing to obtain the corresponding second feature encoding vector; and input the second feature encoding vector into the three basic feature prediction heads respectively for prediction processing to obtain the corresponding fourth, fifth and sixth prediction vectors to form the corresponding first prediction vector sequence; and record the codon usage frequency label vector, the secondary structure label vector and the GC content label vector of the current evaluation record as the corresponding fourth, fifth and sixth label vectors to form the corresponding first label vector sequence; and the first prediction vector sequence and the first label vector sequence form the corresponding first prediction-label pair; and at the end of this round of traversal, input all the obtained first prediction-label pairs into the preset first model evaluation function to calculate and obtain the corresponding first evaluation value;

[0039] Wherein, the first model evaluation function at least includes an RMSE function;

[0040] Step 59, identifying whether the first evaluation value meets the preset first evaluation value range; if not, returning to step 51 to continue training; if satisfied, confirming that the first stage model training is completed.

[0041] Furthermore, the second stage model training of the three vaccine characteristic prediction heads of the vaccine characteristic prediction model based on the second data set specifically includes:

[0042] Step 61, based on a preset second segmentation ratio, randomly split the second data set into two sub-data sets recorded as corresponding second training sets and second evaluation sets;

[0043] Wherein, both the second training set and the second evaluation set are composed of a plurality of the second data records; the ratio of the total number of records in the second training set to the total number of records in the second evaluation set satisfies the second segmentation ratio;

[0044] Step 62, extracting the first second data record of the second training set as the corresponding current training record;

[0045] Step 63: record the stability label data, the translation efficiency label data and the immunogenicity label data of the current training record as corresponding first label data D tag1 , the second label data D tag2 and the third tag data D tag3 ;

[0046] Step 64: input the mRNA molecular sequence of the current training record into the vaccine characteristic prediction model for prediction processing to obtain the corresponding first prediction data D pre1 , the second prediction data D pre2 and the third prediction vector D pre3 ;

[0047] Step 65: convert the first prediction data D pre1 And the corresponding first tag data D tag1 Substitute the preset fourth loss function L e , and the second prediction data D pre2 And the corresponding second tag data D tag2 Substitute the preset fifth loss function L f , and the third prediction vector D pre3 And the corresponding third tag data D tag3 Substitute the preset sixth loss function L g , and the fourth loss function L e , the fifth loss function L f And the sixth loss function L g Add together to form the corresponding second model loss function L M2 ; and calculate the function values ​​of the fourth, fifth, sixth loss functions and the second model loss function and use the calculation results as the corresponding fourth loss value, fifth loss value, sixth loss value and second overall loss value;

[0048] Among them, the second model loss function L M2 for:

[0049] L M2 =L e (D pre1 ,D tag2 )+L f (D pre2 ,D tag2 )+L g (D pre3 ,D tag3 );

[0050] The fourth loss function L e At least including L1 loss function, L2 loss function, and cross entropy loss function; the fifth loss function L f At least including L1 loss function, L2 loss function; the sixth loss function L g At least including L1 loss function, L2 loss function, and cross entropy loss function;

[0051] Step 66, identifying whether the fourth, fifth, sixth loss values ​​and the second overall loss value all satisfy the corresponding fourth loss value range, fifth loss value range, sixth loss value range and second overall loss value range; if the fourth, fifth, sixth loss values ​​and the second overall loss value all satisfy the corresponding loss value range, go to step 67; if the fourth loss value does not satisfy the fourth loss value range, then based on the preset fifth model optimizer, move towards making the fourth loss function L e The model parameters of the stability prediction head are modulated in a round in the direction of reaching the minimum value, and the step 64 is returned to when the parameter modulation of this round is finished; if the fifth loss value does not satisfy the fifth loss value range, the sixth model optimizer based on the preset sixth model optimizer moves toward making the fifth loss function L f The model parameters of the translation efficiency prediction head are modulated in a round in the direction of reaching the minimum value, and the step 64 is returned to when the parameter modulation of this round is finished; if the sixth loss value does not satisfy the sixth loss value range, the seventh model optimizer based on the preset seventh model optimizer moves in the direction of making the sixth loss function L g The model parameters of the immunogenicity prediction head are modulated in a round in the direction of reaching the minimum value, and the step 64 is returned to when the parameter modulation of this round ends; if the second overall loss value does not satisfy the second overall loss value range, the eighth model optimizer based on the preset eighth model optimizer moves in the direction of making the second model loss function L M2 Perform a round of modulation on the model parameters of the three vaccine characteristic prediction heads in the direction of reaching the minimum value, and return to step 64 when this round of parameter modulation is completed;

[0052] Wherein, the fifth, sixth, seventh and eighth model optimizers all include at least an SGD optimizer and an ADAM optimizer;

[0053] Step 67, identifying whether the current training record is the last second data record of the second training set; if so, going to step 68; if not, extracting the next second data record of the second training set as the new current training record and returning to step 63;

[0054] Step 68, perform a round of traversal on all the second data records of the second evaluation set; and in this round of traversal, use the second data record currently traversed as the corresponding current evaluation record; and input the mRNA molecular sequence of the current evaluation record into the vaccine characteristic prediction model for prediction processing to obtain the corresponding fourth, fifth and sixth prediction data to form the corresponding first prediction data sequence; and record the stability label data, the translation efficiency label data and the immunogenicity label data of the current evaluation record as the corresponding fourth, fifth and sixth label data to form the corresponding first label data sequence; and the first prediction data sequence and the first label data sequence form the corresponding second prediction-label pair; and at the end of this round of traversal, input all the obtained second prediction-label pairs into the preset second model evaluation function to calculate and obtain the corresponding second evaluation value;

[0055] Wherein, the second model evaluation function at least includes an RMSE function;

[0056] Step 69, identifying whether the second evaluation value meets the preset second evaluation value range; if not, returning to step 61 to continue training; if satisfied, confirming that the second stage model training is completed.

[0057] Preferably, after the model training is completed, the maximum correlation relationship between the three output features of the three basic feature prediction heads and the three output features of the three vaccine characteristic prediction heads is identified and the corresponding three objective functions are determined based on the identification results, specifically including:

[0058] Step 71, after the model training is completed, the mapping relationship between the three types of basic feature prediction vectors generated by the vaccine characteristic prediction model in the model prediction process and the three types of characteristic prediction data finally output by the model is recorded as the corresponding stability mapping relationship F A , translation efficiency mapping relationship F B and immunogenicity mapping relationship F C ;

[0059] Among them, y a =F A (x a ,x b ,x c ), y b =F B (xa ,x b ,x c ), y c =F C (x a ,x b ,x c );

[0060] Independent variable x a 、x b 、x c are the codon usage frequency prediction vector, the secondary structure prediction vector and the GC content prediction vector generated in the model prediction process, respectively, and the dependent variable y a ,y b ,y c The stability prediction data, the translation efficiency prediction data and the immunogenicity prediction data finally output by the model are respectively;

[0061] Step 72, record each antigen vaccine data record in the antigen vaccine data set as record R i , 1≤i≤N, N is the total number of records in the antigen vaccine data set; and each record R i The mRNA molecular sequence is input into the vaccine characteristic prediction model for prediction processing to obtain the corresponding six prediction results, which are x a,i 、x b,i 、x c,i ,y a,i ,y b,i ,y c,i ; and the prediction result x corresponding to each index i a,i 、x b,i 、x c,i ,y a,i Form a corresponding first independent / dependent variable data group, and form a corresponding first data group set by the obtained N first independent / dependent variable data groups; and form a corresponding first data group set by the prediction result x corresponding to each index i a,i 、x b,i 、x c,i ,y b,i Form a corresponding second independent / dependent variable data group, and form a corresponding second data group set by the obtained N second independent / dependent variable data groups; and form a corresponding second data group set by the prediction result x corresponding to each index i a,i 、x b,i 、x c,i ,y c,i Forming a corresponding third independent / dependent variable data group, and forming a corresponding third data group set from the obtained N third independent / dependent variable data groups;

[0062] Step 73, based on the SHAP evaluation algorithm, evaluate the independent variable x according to the first data set a 、x b 、x c For the dependent variable y a The contribution coefficient of the independent variable x with the largest contribution coefficient a 、x b or x c As the dependent variable y a The maximum associated variable x ma ; and based on the SHAP evaluation algorithm, evaluate the independent variable x according to the second data set a 、x b 、x c For the dependent variable y b The contribution coefficient of the independent variable x with the largest contribution coefficient a 、x b or x c As the dependent variable y b The maximum associated variable x mb ; and based on the SHAP evaluation algorithm, evaluate the independent variable x according to the third data set a 、x b 、x c For the dependent variable y c The contribution coefficient of the independent variable x with the largest contribution coefficient a 、x b or x c As the dependent variable y c The maximum associated variable x mc ;

[0063] Step 74, the dependent variable y a The maximum associated variable x corresponding to it ma Substitute the preset correlation function F corr Get the corresponding stability objective function F corr (x ma ,y a ) and the dependent variable y b The maximum associated variable x corresponding to it mb Substitute the correlation function F corr Get the corresponding translation efficiency objective function F corr (x mb ,y b ) and the dependent variable y c The maximum associated variable x corresponding to it mc Substitute the correlation function F corr Get the corresponding immunogenicity objective function F corr (x mc ,y c); and the stability objective function, the translation efficiency objective function and the immunogenicity objective function form the corresponding three objective functions; the correlation function F corr At least including Pearson correlation function, Spearman correlation function, and Kendall correlation function.

[0064] Preferably, the initializing the mRNA sequence according to the first antigen sequence to obtain the corresponding first mRNA sequence set specifically includes:

[0065] A preset mRNA sequence generation tool is used to perform sequence generation processing on the transcribed mRNA sequence of the first antigen sequence to obtain a plurality of corresponding first mRNA sequences forming the corresponding first mRNA sequence set; wherein the mRNA sequence generation tool is a type of mRNA sequence reverse translation tool or a type of mRNA sequence reverse decoding model; the mRNA sequence reverse translation tool at least includes the Translate Tool tool of ExPASy, which is used to reverse translate and output the transcribed mRNA sequence corresponding to the specified antigen sequence; the mRNA sequence reverse decoding model at least includes a reverse decoding model implemented based on a GANs model, which is used to reverse decode and output the transcribed mRNA sequence corresponding to the specified antigen sequence.

[0066] Preferably, the multi-objective optimization method according to the NSGA-II algorithm, using the vaccine characteristic prediction model, the three objective functions and the corresponding in vitro / in vivo experimental means, performs iterative optimization processing on the first mRNA sequence set to obtain the corresponding optimized mRNA sequence set and feed it back to the user, specifically including:

[0067] Step 91, initializing the first iteration counter to 1;

[0068] Step 92, input each of the first mRNA sequences of the first mRNA sequence set into the vaccine characteristic prediction model for prediction, and bring the prediction results into the three objective functions for calculation to obtain the corresponding first objective function value, second objective function value and third objective function value; and sort all the first objective function values, all the second objective function values ​​and all the third objective function values ​​obtained from high to low according to the function value to obtain the corresponding first, second and third function value sequences;

[0069] Step 93, according to the non-dominated level and crowding distance identification rules in the NSGA-II algorithm, the non-dominated level and crowding distance corresponding to each of the first mRNA sequences are confirmed based on the first, second and third function value sequences; and based on the tournament selection strategy in the NSGA-II algorithm, mRNA sequence individuals are selected according to the non-dominated level and crowding distance of each of the first mRNA sequences, and the first individual set is formed by all the selected first mRNA sequences; and based on the preset individual crossover rule, all the first mRNA sequences in the first individual set are subjected to individual crossover processing, and each of the first mRNA sequences that have completed the crossover is subjected to individual variation processing; and after all the first mRNA sequences in the first individual set have completed individual variation, the first individual set and the first mRNA sequence set are merged to obtain a new first mRNA sequence set;

[0070] Wherein, the individual crossover rules include single-point crossover and multi-point crossover;

[0071] Step 94, adding 1 to the first iteration counter; and identifying whether the first iteration counter after adding 1 exceeds a preset iteration counter threshold; if so, go to step 95; if not, return to step 92;

[0072] Step 95, inputting each of the first mRNA sequences of the latest first mRNA sequence set into the vaccine characteristic prediction model to predict and obtain corresponding first stability prediction data, first translation efficiency prediction data and first immunogenicity prediction data; and sorting all the first stability prediction data and all the first translation efficiency prediction data obtained based on the preset stability and translation efficiency sorting rules, and extracting one or more first mRNA sequences whose stability and translation efficiency are ranked before the specified ranking to form a corresponding preferred set;

[0073] Step 96, performing experimental analysis on the stability and translation efficiency of each of the first mRNA sequences in the preferred set by in vitro experimental means to obtain corresponding first stability experimental data and first translation efficiency experimental data; and calculating the relative error between the first stability prediction data and the first stability experimental data of each of the first mRNA sequences in the preferred set to obtain a corresponding first relative error, and calculating the relative error between the first translation efficiency prediction data and the first translation efficiency experimental data of each of the first mRNA sequences in the preferred set to obtain a corresponding second relative error;

[0074] Step 97, identifying whether all the first and second relative errors obtained are less than the corresponding first and second relative error thresholds; if so, go to step 98; if not, use all the first stability experimental data and the first translation efficiency experimental data obtained this time as label data, perform a round of parameter optimization processing on the stability prediction head and the translation efficiency prediction head of the vaccine characteristic prediction model according to the supervised model training method, and return to step 91 at the end of this round of optimization processing;

[0075] Step 98, performing experimental analysis on the immunogenicity characteristics of each of the first mRNA sequences in the preferred set by means of in vivo experiments to obtain corresponding first immunogenicity experimental data; and calculating the relative error between the first immunogenicity prediction data and the first immunogenicity experimental data of each of the first mRNA sequences in the preferred set to obtain corresponding third relative errors;

[0076] Step 99, identify whether all the third relative errors obtained are less than the corresponding third relative error threshold; if not, use all the first immunogenicity experimental data obtained this time as label data, and perform a round of parameter optimization processing on the immunogenicity prediction head of the vaccine characteristic prediction model according to the supervised model training method, and return to step 91 at the end of this round of optimization processing; if so, use each of the first mRNA sequences in the preferred set as the corresponding optimized mRNA sequence, and form the corresponding optimized mRNA sequence set by all the optimized mRNA sequences obtained and output it.

[0077] A second aspect of an embodiment of the present invention provides a device for implementing the method for processing the mRNA vaccine design task described in the first aspect, the device comprising: a working model construction module, a working model training module and a vaccine design task processing module;

[0078] The working model construction module is used to design a working model for predicting three vaccine characteristics of mRNA sequences using the Uni-RNA model as a feature extraction backbone, which is recorded as a corresponding vaccine characteristic prediction model; the three vaccine characteristics include stability, translation efficiency and immunogenicity; the vaccine characteristic prediction model includes the Uni-RNA model, three basic feature prediction heads, a feature fusion module and three vaccine characteristic prediction heads;

[0079] The working model training module is used to collect a large amount of antigen proteins and their corresponding mRNA vaccine information through big data collection to construct a corresponding antigen vaccine data set; and to train the vaccine characteristic prediction model based on the antigen vaccine data set; and after the model training is completed, the maximum correlation relationship between the three output features of the three basic feature prediction heads and the three output features of the three vaccine characteristic prediction heads is identified and the corresponding three objective functions are determined based on the identification results; the three objective functions include a stability objective function, a translation efficiency objective function and an immunogenicity objective function;

[0080] The vaccine design task processing module is used to receive the molecular sequence of the target antigen protein input by the user as the corresponding first antigen sequence; and initialize the mRNA sequence according to the first antigen sequence to obtain the corresponding first mRNA sequence set; and according to the multi-objective optimization method of the NSGA-II algorithm, with the help of the vaccine characteristic prediction model, the three objective functions and the corresponding in vitro / in vivo experimental means, iteratively optimize the first mRNA sequence set to obtain the corresponding optimized mRNA sequence set to feedback to the user; the first mRNA sequence set consists of multiple first mRNA sequences; the optimized mRNA sequence set consists of one or more optimized mRNA sequences.

[0081] A third aspect of an embodiment of the present invention provides an electronic device, including: a memory, a processor, and a transceiver;

[0082] The processor is used to be coupled to the memory, read and execute instructions in the memory, so as to implement the method steps described in the first aspect above;

[0083] The transceiver is coupled to the processor, and the processor controls the transceiver to send and receive messages.

[0084] A fourth aspect of an embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions. When the computer instructions are executed by a computer, the computer executes the instructions of the method described in the first aspect above.

[0085] The embodiment of the present invention provides a method, device, electronic device and computer-readable storage medium for processing mRNA vaccine design tasks. As can be seen from the above content, the embodiment of the present invention pre-designs a working model for predicting three vaccine characteristics (stability, translation efficiency and immunogenicity) of mRNA sequences with a pre-trained Uni-RNA model as the feature extraction backbone, and records it as a vaccine characteristic prediction model, and trains this model based on an antigen vaccine data set constructed by a big data acquisition method, and identifies the maximum correlation between the three basic features (codon usage frequency, RNA secondary structure, GC content) that can be predicted by this model and the three vaccine characteristics (stability, translation efficiency and immunogenicity) and determines three objective functions (stability objective function, translation efficiency objective function, immunogenicity objective function) based on the identification results. In the specific mRNA vaccine design task processing process: the embodiment of the present invention first uses the mRNA sequence generation tool to batch generate the transcribed mRNA sequences of the target antigen protein to obtain an initialized mRNA sequence set; then uses the vaccine characteristic prediction model and three objective functions to iteratively optimize and screen the mRNA sequence set according to the multi-objective optimization method of the NSGA-II algorithm to obtain a preferred set with a reduced number of sequences; then based on in vitro experimental means, the stability and translation efficiency of the preferred set are experimentally analyzed; if the error between the in vitro experimental analysis results and the model prediction results is too large, the model is automatically optimized based on the experimental data of this experiment, and the mRNA sequence set is iteratively optimized again after optimization; if the error between the in vitro experimental analysis results and the model prediction results is controllable, the immunogenicity of the preferred set is further analyzed based on the in vivo experimental means; if the error between the in vivo experimental analysis results and the model prediction results is too large, the model is automatically optimized again based on the experimental data of this experiment, and the mRNA sequence set is iteratively optimized again after optimization; if the error between the in vivo experimental analysis results and the model prediction results is controllable, the current latest preferred set is output as the final design candidate set. The mRNA sequence initialization method of the embodiment of the present invention breaks away from the limitations of expert experience and improves the structural diversity of mRNA sequences; the vaccine characteristic prediction model of the embodiment of the present invention reduces a large number of experimental analyses, shortens the design cycle, and improves design efficiency; the model training method and multi-objective optimization process of the embodiment of the present invention effectively guarantee the design quality of mRNA sequences. BRIEF DESCRIPTION OF THE DRAWINGS

[0086] Figure 1 A schematic diagram of a method for processing an mRNA vaccine design task provided in Example 1 of the present invention;

[0087] Figure 2 This is a module structure diagram of the vaccine characteristic prediction model provided in Example 1 of the present invention;

[0088] Figure 3 A module structure diagram of a processing device for an mRNA vaccine design task provided in Example 2 of the present invention;

[0089] Figure 4 A schematic diagram of the structure of an electronic device provided in Embodiment 3 of the present invention. DETAILED DESCRIPTION

[0090] In order to make the purpose, technical solution and advantages of the present invention clearer, the present invention will be further described in detail below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0091] Embodiment 1 of the present invention provides a method for processing an mRNA vaccine design task, such as Figure 1 As shown in the schematic diagram of a method for processing an mRNA vaccine design task provided in Example 1 of the present invention, the method mainly comprises the following steps:

[0092] Step 1: Using the Uni-RNA model as the feature extraction backbone, a working model for predicting the three vaccine characteristics of the mRNA sequence is designed and recorded as the corresponding vaccine characteristic prediction model.

[0093] Here, the three vaccine characteristics mentioned in the embodiments of the present invention include stability, translation efficiency and immunogenicity; stability is usually characterized by the half-life of mRNA in the body, translation efficiency is usually characterized by the expression level of the target antigen protein, and immunogenicity is usually characterized by antibody titer.

[0094] The vaccine characteristic prediction model of the embodiment of the present invention is used to predict the stability, translation efficiency and immunogenicity characteristics of the input mRNA sequence and output corresponding stability prediction data, translation efficiency prediction data and immunogenicity prediction data; the stability prediction data, translation efficiency prediction data and immunogenicity prediction data here are actually half-life prediction data, antigen protein expression level prediction data and antibody titer prediction data.

[0095] like Figure 2 As shown in the module structure diagram of the vaccine characteristic prediction model provided in Example 1 of the present invention, the model components of the vaccine characteristic prediction model include: a Uni-RNA model, three basic feature prediction heads, a feature fusion module and three vaccine characteristic prediction heads; among them, the three basic feature prediction heads include a codon usage frequency prediction head, an RNA secondary structure prediction head, and a GC content prediction head; the three vaccine characteristic prediction heads include a stability prediction head, a translation efficiency prediction head and an immunogenicity prediction head.

[0096] The input / output ends of the vaccine characteristic prediction model are described as follows: the model input end is used to receive the corresponding mRNA sequence, and the first model output end, the second model output end and the third model output end are used to output the corresponding stability prediction data, translation efficiency prediction data and immunogenicity prediction data respectively.

[0097] The connection relationship between each model component is as follows: the input end of the Uni-RNA model is connected to the model input end, and the output end is respectively connected to the input ends of the codon usage frequency prediction head, the RNA secondary structure prediction head and the GC content prediction head; the output ends of the codon usage frequency prediction head, the RNA secondary structure prediction head and the GC content prediction head are connected to the first, second and third input ends of the feature fusion module; the output end of the feature fusion module is respectively connected to the input ends of the stability prediction head, the translation efficiency prediction head and the immunogenicity prediction head; the output ends of the stability prediction head, the translation efficiency prediction head and the immunogenicity prediction head are respectively connected to the corresponding first, second and third model output ends.

[0098] The component functions of each model component are as follows:

[0099] 1) Uni-RNA model:

[0100] The Uni-RNA model of the embodiment of the present invention has completed model pre-training; the Uni-RNA model is used to perform feature encoding processing on the input mRNA sequence to obtain the corresponding feature encoding vector and send it to the codon usage frequency prediction head, RNA secondary structure prediction head and GC content prediction head;

[0101] Here, the Uni-RNA model of the embodiment of the present invention is a feature encoding model implemented based on a transformer encoder structure, and the Uni-RNA model has completed pre-training; the model structure, function and pre-training scheme of this model have been disclosed in the technical document "UNI-RNA: UNIVERSAL PRE-TRAINED MODELS REVOLUTIONIZE RNA RESEARCH", and it can be known from the technical document that the Uni-RNA model can be used to perform base-level deep learning on an RNA sequence;

[0102] 2) Codon usage frequency prediction header:

[0103] The codon usage frequency prediction head of the embodiment of the present invention is implemented based on a type of nonlinear regression model or a type of neural network model; the codon usage frequency prediction head is used to predict the codon usage frequency of each base position on the mRNA sequence according to the feature coding vector to obtain the corresponding codon usage frequency prediction vector and send it to the feature fusion module;

[0104] Here, the nonlinear regression model mentioned in the embodiment of the present invention includes at least an MLP model, an NNR model, a GBDT model and an XGBoost model; the neural network model includes at least an MLP model, a CNN model, a ResNet model and a GNN model;

[0105] 3) RNA secondary structure prediction head:

[0106] The RNA secondary structure prediction head of the embodiment of the present invention is implemented based on another type of neural network model; the RNA secondary structure prediction head is used to predict the secondary structure features of each base position on the mRNA sequence according to the feature encoding vector to obtain the corresponding secondary structure prediction vector and send it to the feature fusion module;

[0107] 4) GC content prediction head:

[0108] The GC content prediction head of the embodiment of the present invention is implemented based on another type of nonlinear regression model or another type of neural network model; the GC content prediction head is used to predict the GC content of each base position on the mRNA sequence according to the feature encoding vector to obtain a corresponding GC content prediction vector and send it to the feature fusion module;

[0109] 5) Feature fusion module:

[0110] The feature fusion module of the embodiment of the present invention is used to perform vector splicing on the codon usage frequency prediction vector, the secondary structure prediction vector and the GC content prediction vector to obtain the corresponding splicing vector and send it to the stability prediction head, the translation efficiency prediction head and the immunogenicity prediction head;

[0111] 6) Stability prediction head:

[0112] The stability prediction head of the embodiment of the present invention is implemented based on another type of nonlinear regression model or another type of neural network model; the stability prediction head is used to predict the stability characteristics of the mRNA sequence according to the splicing vector and output the corresponding stability prediction data;

[0113] 7) Translation efficiency prediction head:

[0114] The translation efficiency prediction head of the embodiment of the present invention is implemented based on another type of nonlinear regression model or another type of neural network model; the translation efficiency prediction head is used to predict the translation characteristics of the mRNA sequence according to the splicing vector and output the corresponding translation efficiency prediction data;

[0115] 8) Immunogenicity prediction head:

[0116] The immunogenicity prediction head of the embodiment of the present invention is implemented based on another type of nonlinear regression model or another type of neural network model; the immunogenicity prediction head is used to predict the immunogenicity characteristics of the mRNA sequence according to the splicing vector and output the corresponding immunogenicity prediction data.

[0117] Step 2: collect a large amount of antigen proteins and their corresponding mRNA vaccine information through big data collection to construct a corresponding antigen vaccine data set; and train the vaccine characteristic prediction model based on the antigen vaccine data set; and after the model training is completed, identify the maximum correlation between the three output features of the three basic feature prediction heads and the three output features of the three vaccine characteristic prediction heads, and determine the corresponding three objective functions based on the identification results;

[0118] Specifically comprising: step 21, collecting a large amount of antigen protein and its corresponding mRNA vaccine information through big data collection to construct a corresponding antigen vaccine data set;

[0119] Among them, the antigen vaccine data set includes multiple antigen vaccine data records; the antigen vaccine data records include antigen molecule sequence, mRNA molecule sequence, codon usage frequency label vector, secondary structure label vector, GC content label vector, stability label data, translation efficiency label data and immunogenicity label data;

[0120] Here, the embodiment of the present invention collects data from multiple public mRNA vaccine databases and multiple types of public technical documents / materials / magazines / papers through big data collection. The collection objects include various types of antigen proteins and their corresponding mRNA vaccine information. The mRNA vaccine information here at least includes the corresponding mRNA molecule sequence, codon usage frequency, secondary structure, GC content, stability parameter, i.e., the half-life of mRNA, translation efficiency parameter, i.e., the expression amount of the target antigen protein, and immunogenicity parameter, i.e., the antibody titer of the corresponding antibody; it should be noted that if the characteristic / characteristic parameters (code usage frequency, secondary structure, GC content, stability parameter, translation efficiency parameter, immunogenicity parameter) of a certain mRNA vaccine are found to be missing during the data collection process, they will be calculated and supplemented through the preset biochemical information tool;

[0121] Step 22, training the vaccine characteristic prediction model based on the antigen vaccine dataset;

[0122] Specifically comprising: step 221, extracting the mRNA molecular sequence, codon usage frequency label vector, secondary structure label vector and GC content label vector of each antigen vaccine data record in the antigen vaccine data set to form a corresponding first data record; and forming a corresponding first data set from all the obtained first data records;

[0123] Here, the obtained first data set is composed of a plurality of first data records; each first data record is composed of an mRNA molecule sequence, a codon usage frequency label vector, a secondary structure label vector, and a GC content label vector;

[0124] Step 222, extract the mRNA molecular sequence, stability label data, translation efficiency label data and immunogenicity label data of each antigen vaccine data record in the antigen vaccine data set to form a corresponding second data record; and all the obtained second data records form a corresponding second data set;

[0125] Here, the obtained second data set consists of a plurality of second data records; each second data record consists of an mRNA molecule sequence, stability label data, translation efficiency label data, and immunogenicity label data;

[0126] Step 223, performing first-stage model training on the Uni-RNA model and three basic feature prediction heads of the vaccine characteristic prediction model based on the first data set;

[0127] Specifically, it includes: step 2231, randomly dividing the first data set into two sub-data sets based on a preset first division ratio and recording them as a corresponding first training set and a first evaluation set;

[0128] The first segmentation ratio is a preset ratio value, such as 8:2; the first training set and the first evaluation set are both composed of a plurality of first data records; the ratio of the total number of records in the first training set to the total number of records in the first evaluation set satisfies the first segmentation ratio;

[0129] Step 2232, extracting the first first data record of the first training set as the corresponding current training record;

[0130] Step 2233, record the codon usage frequency label vector, secondary structure label vector and GC content label vector of the current training record as the corresponding first label vector P tag1 , the second label vector P tag2 and the third label vector P tag3 ;

[0131] Step 2234, input the mRNA molecular sequence of the current training record into the Uni-RNA model for feature encoding processing to obtain the corresponding first feature encoding vector; and input the first feature encoding vector into the codon usage frequency prediction head, RNA secondary structure prediction head and GC content prediction head for prediction processing to obtain the corresponding first prediction vector P pre1 , the second prediction vector P pre2 and the third prediction vector P pre3 ;

[0132] Step 2235: transform the first prediction vector P pre1 And the corresponding first label vector P tag1 Substitute the preset first loss function L a , and the second prediction vector P pre2 And the corresponding second label vector P tag2 Substitute the preset second loss function L b , and the third prediction vector P pre3 And the corresponding third label vector P tag3 Bring in the preset third loss function L c , and by the first loss function L a , the second loss function L b And the third loss function L c Add together to form the corresponding first model loss function L M1 ; and calculate the function values ​​of the first, second, third loss functions and the first model loss function and use the calculation results as the corresponding first loss value, second loss value, third loss value and first overall loss value;

[0133] Among them, the first model loss function L M1 for:

[0134] L M1 =L a (P pre1 ,P tag1 )+L b (P pre2 ,P tag2 )+L c (P pre3 ,P tag3 );

[0135] The first loss function L a At least including L1 loss function, L2 loss function; the second loss function L b It is composed of a base type loss function and a base structure loss function. The base type loss function at least includes a cross entropy loss function, and the base structure loss function at least includes an L1 loss function and an L2 loss function. The third loss function L c At least including L1 loss function and L2 loss function;

[0136] Step 2236, identify whether the first, second, third loss values ​​and the first overall loss value all satisfy the corresponding first loss value range, second loss value range, third loss value range and first overall loss value range; if the first, second, third loss values ​​and the first overall loss value all satisfy the corresponding loss value range, go to step 2237; if the first loss value does not satisfy the first loss value range, then based on the preset first model optimizer, move towards making the first loss function L a The model parameters of the Uni-RNA model and the codon usage frequency prediction head are modulated in a round in the direction of reaching the minimum value, and the step 2234 is returned to when the parameter modulation of this round ends; if the second loss value does not satisfy the second loss value range, the second model optimizer based on the preset second model optimizer moves in the direction of making the second loss function L b The model parameters of the Uni-RNA model and the RNA secondary structure prediction head are modulated in a round in the direction of reaching the minimum value, and the step 2234 is returned to when the parameter modulation of this round ends; if the third loss value does not satisfy the third loss value range, the third model optimizer is based on the preset third model to make the third loss function L c The model parameters of the Uni-RNA model and the GC content prediction head are modulated in a round in the direction of reaching the minimum value, and the step 2234 is returned to when the parameter modulation of this round is finished; if the first overall loss value does not satisfy the first overall loss value range, the fourth model optimizer is based on the preset fourth model optimizer to make the first model loss function L M1 Perform a round of modulation on the model parameters of the Uni-RNA model and the three basic feature prediction heads in the direction where the minimum value is reached, and return to step 2234 when this round of parameter modulation is completed;

[0137] The first, second, third loss value ranges and the first overall loss value range are four pre-set loss value ranges; the first, second, third and fourth model optimizers all include at least an SGD optimizer and an ADAM optimizer;

[0138] Step 2237, identifying whether the current training record is the last first data record of the first training set; if so, going to step 2238; if not, extracting the next first data record of the first training set as the new current training record and returning to step 2233;

[0139] Step 2238, perform a round of traversal on all the first data records of the first evaluation set; and in this round of traversal, use the currently traversed first data record as the corresponding current evaluation record; and input the mRNA molecular sequence of the current evaluation record into the Uni-RNA model for feature encoding processing to obtain the corresponding second feature encoding vector; and input the second feature encoding vector into the three basic feature prediction heads respectively for prediction processing to obtain the corresponding fourth, fifth and sixth prediction vectors to form the corresponding first prediction vector sequence; and record the codon usage frequency label vector, secondary structure label vector and GC content label vector of the current evaluation record as the corresponding fourth, fifth and sixth label vectors to form the corresponding first label vector sequence; and the first prediction vector sequence and the first label vector sequence form the corresponding first prediction-label pair; and at the end of this round of traversal, input all the obtained first prediction-label pairs into the preset first model evaluation function to calculate and obtain the corresponding first evaluation value;

[0140] Wherein, the first model evaluation function at least includes the RMSE function;

[0141] Step 2239, identifying whether the first evaluation value meets the preset first evaluation value range; if not, returning to step 2231 to continue training; if satisfied, confirming that the first stage model training is completed;

[0142] Wherein, the first evaluation value range is a preset evaluation value range;

[0143] Step 224, after the first round of model training is completed, a second stage model training is performed on the three vaccine characteristic prediction heads of the vaccine characteristic prediction model based on the second data set;

[0144] Specifically, it includes: step 2241, randomly dividing the second data set into two sub-data sets based on a preset second division ratio and recording them as corresponding second training sets and second evaluation sets;

[0145] The second segmentation ratio is a preset ratio value, such as 8:2; the second training set and the second evaluation set are both composed of a plurality of second data records; the ratio of the total number of records in the second training set to the total number of records in the second evaluation set satisfies the second segmentation ratio;

[0146] Step 2242, extracting the first second data record of the second training set as the corresponding current training record;

[0147] Step 2243, record the stability label data, translation efficiency label data and immunogenicity label data of the current training record as the corresponding first label data D tag1 , the second label data D tag2 and the third tag data D tag3 ;

[0148] Step 2244: input the mRNA molecular sequence of the current training record into the vaccine characteristic prediction model for prediction processing to obtain the corresponding first prediction data D pre1 , the second prediction data D pre2 and the third prediction vector D pre3 ;

[0149] Step 2245: The first prediction data D pre1 And the corresponding first label data D tag1 Substitute the preset fourth loss function L e , and the second prediction data D pre2 And the corresponding second label data D tag2 Substitute the preset fifth loss function L f , and the third prediction vector D pre3 And the corresponding third label data D tag3 Substitute the preset sixth loss function L g , and by the fourth loss function L e , the fifth loss function L f And the sixth loss function L g Add together to form the corresponding second model loss function L M2 ; and calculate the function values ​​of the fourth, fifth, sixth loss functions and the second model loss function and use the calculation results as the corresponding fourth loss value, fifth loss value, sixth loss value and second overall loss value;

[0150] Among them, the second model loss function L M2 for:

[0151] L M2 =L e (D pre1 ,D tag2 )+L f (D pre2 ,D tag2 )+L g (D pre3 ,D tag3 );

[0152] The fourth loss function L e At least including L1 loss function, L2 loss function, cross entropy loss function; the fifth loss function L f At least including L1 loss function, L2 loss function; the sixth loss function L g At least including L1 loss function, L2 loss function, and cross entropy loss function;

[0153] Step 2246, identify whether the fourth, fifth, sixth loss values ​​and the second overall loss value all satisfy the corresponding fourth loss value range, fifth loss value range, sixth loss value range and second overall loss value range; if the fourth, fifth, sixth loss values ​​and the second overall loss value all satisfy the corresponding loss value range, go to step 2247; if the fourth loss value does not satisfy the fourth loss value range, then based on the preset fifth model optimizer, move towards making the fourth loss function L e The model parameters of the stability prediction head are modulated in a round in the direction of reaching the minimum value, and the step 2244 is returned to when the parameter modulation of this round ends; if the fifth loss value does not satisfy the fifth loss value range, the sixth model optimizer based on the preset sixth model optimizer moves in the direction of making the fifth loss function L f The model parameters of the translation efficiency prediction head are modulated in a round in the direction of reaching the minimum value, and the step 2244 is returned to when the parameter modulation of this round ends; if the sixth loss value does not satisfy the sixth loss value range, the seventh model optimizer based on the preset seventh model optimizer moves in the direction of making the sixth loss function L g The model parameters of the immunogenicity prediction head are modulated in a round in the direction of reaching the minimum value, and when this round of parameter modulation ends, the step 2244 is returned; if the second overall loss value does not satisfy the second overall loss value range, the eighth model optimizer based on the preset eighth model optimizer moves in the direction of making the second model loss function L M2 Perform a round of modulation on the model parameters of the three vaccine characteristic prediction heads in the direction of reaching the minimum value, and return to step 2244 when this round of parameter modulation is completed;

[0154] Among them, the fourth, fifth, sixth loss value ranges and the second overall loss value range are four preset loss value ranges; the fifth, sixth, seventh and eighth model optimizers all include at least an SGD optimizer and an ADAM optimizer;

[0155] Step 2247, identifying whether the current training record is the last second data record of the second training set; if so, going to step 2248; if not, extracting the next second data record of the second training set as the new current training record and returning to step 2243;

[0156] Step 2248, perform a round of traversal on all the second data records of the second evaluation set; and in this round of traversal, use the currently traversed second data record as the corresponding current evaluation record; and input the mRNA molecular sequence of the current evaluation record into the vaccine characteristic prediction model for prediction processing to obtain the corresponding fourth, fifth and sixth prediction data to form the corresponding first prediction data sequence; and record the stability label data, translation efficiency label data and immunogenicity label data of the current evaluation record as the corresponding fourth, fifth and sixth label data to form the corresponding first label data sequence; and the first prediction data sequence and the first label data sequence form the corresponding second prediction-label pair; and at the end of this round of traversal, input all the obtained second prediction-label pairs into the preset second model evaluation function to calculate and obtain the corresponding second evaluation value;

[0157] Wherein, the second model evaluation function at least includes the RMSE function;

[0158] Step 2249, identifying whether the second evaluation value meets the preset second evaluation value range; if not, returning to step 2241 to continue training; if satisfied, confirming that the second stage model training is completed;

[0159] Wherein, the second evaluation value range is a preset evaluation value range;

[0160] Step 225, after the second round of model training is completed, confirm that the current model training is completed;

[0161] Step 23, after the model training is completed, the maximum correlation between the three output features of the three basic feature prediction heads and the three output features of the three vaccine characteristic prediction heads is identified, and the corresponding three objective functions are determined based on the identification results;

[0162] Among them, the three objective functions include stability objective function, translation efficiency objective function and immunogenicity objective function;

[0163] Specifically, step 231, after the model training is completed, the mapping relationship between the three types of basic feature prediction vectors generated by the vaccine characteristic prediction model in the model prediction process and the three types of characteristic prediction data finally output by the model is recorded as the corresponding stability mapping relationship F A , translation efficiency mapping relationship F B and immunogenicity mapping relationship F C ;

[0164] Among them, y a =F A (x a ,x b ,x c ), y b =F B (xa ,x b ,x c ), y c =F C (x a ,x b ,x c );

[0165] Independent variable x a 、x b 、x c are the codon usage frequency prediction vector, secondary structure prediction vector and GC content prediction vector generated in the model prediction process, respectively. The dependent variable y a ,y b ,y c They are the stability prediction data, translation efficiency prediction data and immunogenicity prediction data finally output by the model;

[0166] Step 232, record each antigen vaccine data record in the antigen vaccine data set as record R i , 1≤i≤N, N is the total number of records in the antigen vaccine data set; and each record R i The mRNA molecular sequence is input into the vaccine characteristic prediction model for prediction processing to obtain the corresponding six prediction results, which are x a,i 、x b,i 、x c,i ,y a,i ,y b,i ,y c,i ; and the prediction result x corresponding to each index i a,i 、x b,i 、x c,i ,y a,i Form a corresponding first independent / dependent variable data group, and form a corresponding first data group set by the obtained N first independent / dependent variable data groups; and form a corresponding first data group set by the prediction result x corresponding to each index i a,i 、x b,i 、x c,i ,y b,i Form a corresponding second independent / dependent variable data group, and form a corresponding second data group set by the obtained N second independent / dependent variable data groups; and form a corresponding second data group set by the prediction result x corresponding to each index i a,i 、x b,i 、x c,i ,y c,i Forming a corresponding third independent / dependent variable data group, and forming a corresponding third data group set from the obtained N third independent / dependent variable data groups;

[0167] Step 233, based on the SHAP evaluation algorithm, evaluate the independent variable x according to the first data set a 、xb 、x c For the dependent variable y a The contribution coefficient of the independent variable x with the largest contribution coefficient a 、x b or x c As the dependent variable y a The maximum associated variable x ma ; and based on the SHAP evaluation algorithm, evaluate the independent variable x according to the second data set a 、x b 、x c For the dependent variable y b The contribution coefficient of the independent variable x with the largest contribution coefficient a 、x b or x c As the dependent variable y b The maximum associated variable x mb ; and based on the SHAP evaluation algorithm, evaluate the independent variable x according to the third data set a 、x b 、x c For the dependent variable y c The contribution coefficient of the independent variable x with the largest contribution coefficient a 、x b or x c As the dependent variable y c The maximum associated variable x mc ;

[0168] Step 234, the dependent variable y a The maximum associated variable x corresponding to it ma Substitute the preset correlation function F corr Get the corresponding stability objective function F corr (x ma ,y a ) and the dependent variable y b The maximum associated variable x corresponding to it mb Substitute the correlation function F corr Get the corresponding translation efficiency objective function F corr (x mb ,y b ) and the dependent variable y c The maximum associated variable x corresponding to it mc Substitute the correlation function F corr Get the corresponding immunogenicity objective function F corr (x mc ,y c ); and the three corresponding objective functions are composed of a stability objective function, a translation efficiency objective function and an immunogenicity objective function;

[0169] Among them, the correlation function F corr At least including Pearson correlation function, Spearman correlation function, and Kendall correlation function.

[0170] Step 3, receiving the molecular sequence of the target antigen protein input by the user as the corresponding first antigen sequence; and initializing the mRNA sequence according to the first antigen sequence to obtain the corresponding first mRNA sequence set; and according to the multi-objective optimization method of the NSGA-II algorithm, using the vaccine characteristic prediction model, three objective functions and corresponding in vitro / in vivo experimental methods, iteratively optimizing the first mRNA sequence set to obtain the corresponding optimized mRNA sequence set and feeding it back to the user;

[0171] Specifically, it includes: step 31, receiving the molecular sequence of the target antigen protein input by the user as the corresponding first antigen sequence;

[0172] Step 32, initializing the mRNA sequence according to the first antigen sequence to obtain a corresponding first mRNA sequence set;

[0173] Wherein, the first mRNA sequence set consists of a plurality of first mRNA sequences;

[0174] Specifically, it includes: using a preset mRNA sequence generation tool to perform sequence generation processing on the transcribed mRNA sequence of the first antigen sequence to obtain a plurality of corresponding first mRNA sequences to form a corresponding first mRNA sequence set;

[0175] Here, the mRNA sequence generation tool of the embodiment of the present invention is a type of mRNA sequence reverse translation tool or a type of mRNA sequence reverse decoding model; wherein the mRNA sequence reverse translation tool at least includes the Translate Tool tool of ExPASy, which is used to reversely translate and output the transcribed mRNA sequence corresponding to the specified antigen sequence; the mRNA sequence reverse decoding model at least includes a reverse decoding model implemented based on the GANs model, which is used to reversely decode and output the transcribed mRNA sequence corresponding to the specified antigen sequence;

[0176] Step 323, according to the multi-objective optimization method of the NSGA-II algorithm, the first mRNA sequence set is iteratively optimized with the help of the vaccine characteristic prediction model, three objective functions and corresponding in vitro / in vivo experimental means to obtain the corresponding optimized mRNA sequence set and feed it back to the user;

[0177] Wherein, the optimized mRNA sequence set consists of one or more optimized mRNA sequences;

[0178] Specifically comprising: step 3231, initializing the first iteration counter to 1;

[0179] Step 3232, input each first mRNA sequence of the first mRNA sequence set into the vaccine characteristic prediction model for prediction, and bring the prediction results into the three objective functions for calculation to obtain the corresponding first objective function value, second objective function value and third objective function value; and sort all the obtained first objective function values, all the second objective function values ​​and all the third objective function values ​​from high to low according to the function value to obtain the corresponding first, second and third function value sequences;

[0180] Step 3233, according to the non-dominated level and crowding distance identification rules in the NSGA-II algorithm, the non-dominated level and crowding distance corresponding to each first mRNA sequence are confirmed based on the first, second and third function value sequences; and based on the tournament selection strategy in the NSGA-II algorithm, mRNA sequence individuals are selected according to the non-dominated level and crowding distance of each first mRNA sequence, and the corresponding first individual set is formed by all the selected first mRNA sequences; and based on the preset individual crossover rule, all the first mRNA sequences in the first individual set are subjected to individual crossover processing, and each first mRNA sequence that has completed the crossover is subjected to individual variation processing; and after all the first mRNA sequences in the first individual set have completed individual variation, the first individual set and the first mRNA sequence set are merged to obtain a new first mRNA sequence set;

[0181] Among them, individual crossover rules include single-point crossover and multi-point crossover;

[0182] Step 3234, adding 1 to the first iteration counter; and identifying whether the first iteration counter after adding 1 exceeds the preset iteration counter threshold; if so, go to step 3235; if not, return to step 3232;

[0183] Here, the iteration counter threshold is a preset threshold parameter;

[0184] Step 3235, input each first mRNA sequence of the latest first mRNA sequence set into the vaccine characteristic prediction model for prediction to obtain corresponding first stability prediction data, first translation efficiency prediction data and first immunogenicity prediction data; and sort all the obtained first stability prediction data and all the first translation efficiency prediction data respectively based on the preset stability and translation efficiency sorting rules, and extract one or more first mRNA sequences whose stability and translation efficiency are ranked before the specified sorting ranking to form a corresponding preferred set;

[0185] Here, the stability and translation efficiency sorting rule of the embodiment of the present invention is a preset sorting rule, and the sorting is performed in a descending order by default. The higher the ranking, the better the corresponding stability / translation efficiency, and vice versa. The lower the ranking, the worse the corresponding stability / translation efficiency; the designated sorting rank is a preset ranking value, and the stability and translation efficiency sorting can use the same value as the corresponding designated sorting rank, or define a different value for each of the stability and translation efficiency as the corresponding designated sorting rank;

[0186] Step 3236, experimentally analyzing the stability and translation efficiency of each first mRNA sequence in the preferred set by in vitro experimental means to obtain corresponding first stability experimental data and first translation efficiency experimental data; and calculating the relative error between the first stability prediction data and the first stability experimental data of each first mRNA sequence in the preferred set to obtain the corresponding first relative error, and calculating the relative error between the first translation efficiency prediction data and the first translation efficiency experimental data of each first mRNA sequence in the preferred set to obtain the corresponding second relative error;

[0187] Here, the in vitro experimental means mentioned in the embodiment of the present invention is the in vitro experimental means used in the design process of conventional mRNA vaccines. The experimental means synthesizes mRNA by in vitro transcription in a cell-free system and verifies its expression efficiency and stability in an in vitro expression system; for example, T7 RNA polymerase is used to perform in vitro transcription and synthesize candidate mRNA sequences, and the translation efficiency / stability characterization data of mRNA are detected by fluorescent quantitative real-time polymerase chain reaction (Quantitative Real-time Polymerase Chain Reaction, qPCR) detection method and enzyme-linked immunosorbent assay (Enzyme Linked Immunosorbent Assay, ELISA) method;

[0188] Step 3237, identify whether all the first and second relative errors obtained are less than the corresponding first and second relative error thresholds; if so, go to step 3238; if not, use all the first stability experimental data and first translation efficiency experimental data obtained this time as label data, perform a round of parameter optimization processing on the stability prediction head and translation efficiency prediction head of the vaccine characteristic prediction model according to the supervised model training method, and return to step 3231 at the end of this round of optimization processing;

[0189] Here, the first and second relative error thresholds in the embodiment of the present invention are two preset error thresholds;

[0190] Step 3238, performing experimental analysis on the immunogenicity characteristics of each first mRNA sequence in the preferred set by means of in vivo experiments to obtain corresponding first immunogenicity experimental data; and calculating the relative error between the first immunogenicity prediction data and the first immunogenicity experimental data of each first mRNA sequence in the preferred set to obtain corresponding third relative errors;

[0191] Here, the in vivo experimental means mentioned in the embodiment of the present invention is the in vivo experimental means used in the design process of conventional mRNA vaccines, which evaluates the immunogenicity of candidate mRNA vaccines in animal models; for example, mice or rabbits are selected as animal models, and the animal models are vaccinated with candidate mRNA vaccines using standard immunization procedures, and then the antibody titer corresponding to the injected vaccine is evaluated using methods such as flow cytometry (FACS), enzyme-linked immunospot assay (ELISPOT) and neutralizing antibody assay;

[0192] Step 3239, whether all the third relative errors obtained are less than the corresponding third relative error threshold is identified; if not, all the first immunogenicity experimental data obtained this time are used as label data, and a round of parameter optimization processing is performed on the immunogenicity prediction head of the vaccine characteristic prediction model according to the supervised model training method, and at the end of this round of optimization processing, it returns to step 3231; if so, each first mRNA sequence in the preferred set is used as the corresponding optimized mRNA sequence, and the corresponding optimized mRNA sequence set is composed of all the optimized mRNA sequences obtained and output.

[0193] Here, the third relative error threshold in the embodiment of the present invention is a preset error threshold.

[0194] Figure 3 The module structure diagram of a processing device for mRNA vaccine design tasks provided in the second embodiment of the present invention is a terminal device or server that implements the aforementioned method embodiment, and can also be a device that enables the aforementioned terminal device or server to implement the aforementioned method embodiment. For example, the device can be a device or chip system of the aforementioned terminal device or server. Figure 3 As shown, the device includes: a working model building module 201, a working model training module 202 and a vaccine design task processing module 203.

[0195] The working model construction module 201 is used to design a working model for predicting three vaccine characteristics of mRNA sequences with the Uni-RNA model as the feature extraction backbone, which is recorded as the corresponding vaccine characteristic prediction model; the three vaccine characteristics include stability, translation efficiency and immunogenicity; the vaccine characteristic prediction model includes the Uni-RNA model, three basic feature prediction heads, a feature fusion module and three vaccine characteristic prediction heads.

[0196] The working model training module 202 is used to collect a large amount of antigen proteins and their corresponding mRNA vaccine information through big data collection to construct a corresponding antigen vaccine data set; and to train the vaccine characteristic prediction model based on the antigen vaccine data set; and after the model training is completed, the maximum correlation between the three output features of the three basic feature prediction heads and the three output features of the three vaccine characteristic prediction heads is identified and the corresponding three objective functions are determined based on the identification results; the three objective functions include a stability objective function, a translation efficiency objective function and an immunogenicity objective function.

[0197] The vaccine design task processing module 203 is used to receive the molecular sequence of the target antigen protein input by the user as the corresponding first antigen sequence; and initialize the mRNA sequence according to the first antigen sequence to obtain the corresponding first mRNA sequence set; and according to the multi-objective optimization method of the NSGA-II algorithm, with the help of the vaccine characteristic prediction model, three objective functions and corresponding in vitro / in vivo experimental means, iteratively optimize the first mRNA sequence set to obtain the corresponding optimized mRNA sequence set and feedback it to the user; the first mRNA sequence set consists of multiple first mRNA sequences; the optimized mRNA sequence set consists of one or more optimized mRNA sequences.

[0198] An embodiment of the present invention provides a processing device for mRNA vaccine design tasks, which can execute the method steps in the above method embodiment. Its implementation principle and technical effects are similar and will not be repeated here.

[0199] It should be noted that it should be understood that the division of the various modules of the above device is only a division of logical functions. In actual implementation, they can be fully or partially integrated into one physical entity, or they can be physically separated. And these modules can all be implemented in the form of software called by processing elements; they can also be all implemented in the form of hardware; some modules can also be implemented in the form of software called by processing elements, and some modules can be implemented in the form of hardware. For example, the working model construction module can be a separately established processing element, or it can be integrated in a chip of the above device. In addition, it can also be stored in the memory of the above device in the form of program code, and called by a processing element of the above device to perform the functions of the above-mentioned determined module. The implementation of other modules is similar. In addition, these modules can be fully or partially integrated together, or they can be implemented independently. The processing element described here can be an integrated circuit with signal processing capabilities. In the implementation process, each step of the above method or each module above can be completed by an integrated logic circuit of hardware in the processor element or instructions in software form.

[0200] For example, the above modules may be one or more integrated circuits configured to implement the above methods, such as one or more application specific integrated circuits (ASIC), or one or more digital signal processors (DSP), or one or more field programmable gate arrays (FPGA). For another example, when a module above is implemented in the form of a processing element scheduling program code, the processing element may be a general-purpose processor, such as a central processing unit (CPU) or other processor that can call program code. For another example, these modules may be integrated together and implemented in the form of a system-on-a-chip (SOC).

[0201] In the above embodiments, all or part of the embodiments may be implemented by software, hardware, firmware or any combination thereof. When implemented by software, all or part of the embodiments may be implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the above method embodiments are generated. The above computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The above-mentioned computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the above-mentioned computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center through wired (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (such as infrared, wireless, Bluetooth, microwave, etc.) methods. The above-mentioned computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more available media integrated. The above-mentioned available medium can be a magnetic medium (such as a floppy disk, a hard disk, a tape), an optical medium (such as a DVD), or a semiconductor medium (such as a solid state disk (SSD)), etc.

[0202] Figure 4 This is a schematic diagram of the structure of an electronic device provided in Embodiment 3 of the present invention. The electronic device may be a terminal device or a server for implementing the method of the aforementioned embodiment, or may be a terminal device or a server for implementing the method of the aforementioned embodiment connected to the aforementioned terminal device or server. Figure 4 As shown, the electronic device may include: a processor 301 (such as a CPU), a memory 302, and a transceiver 303; the transceiver 303 is coupled to the processor 301, and the processor 301 controls the transceiver 303. Various instructions may be stored in the memory 302 to complete various processing functions and implement the processing steps described in the aforementioned embodiment method. Preferably, the electronic device involved in the embodiment of the present invention also includes: a power supply 304, a system bus 305 and a communication port 306. The system bus 305 is used to realize the communication connection between components. The above-mentioned communication port 306 is used for connecting and communicating between the electronic device and other peripherals.

[0203] exist Figure 4The system bus 305 mentioned in the figure can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The system bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus. The communication interface is used to realize the communication between the database access device and other devices (such as clients, read-write libraries, and read-only libraries). The memory may include random access memory (RAM), and may also include non-volatile memory (Non-Volatile Memory), such as at least one disk storage.

[0204] The above-mentioned processor can be a general-purpose processor, including a central processing unit CPU, a network processor (Network Processor, NP), a graphics processing unit (Graphics Processing Unit, GPU), etc.; it can also be a digital signal processor DSP, an application-specific integrated circuit ASIC, a field programmable gate array FPGA or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0205] It should be noted that an embodiment of the present invention further provides a computer-readable storage medium, in which instructions are stored. When the computer-readable storage medium is run on a computer, the computer executes the method and processing process provided in the above embodiments.

[0206] The embodiment of the present invention provides a method, device, electronic device and computer-readable storage medium for processing mRNA vaccine design tasks. As can be seen from the above content, the embodiment of the present invention pre-designs a working model for predicting three vaccine characteristics (stability, translation efficiency and immunogenicity) of mRNA sequences with a pre-trained Uni-RNA model as the feature extraction backbone, and records it as a vaccine characteristic prediction model, and trains this model based on an antigen vaccine data set constructed by a big data acquisition method, and identifies the maximum correlation between the three basic features (codon usage frequency, RNA secondary structure, GC content) that can be predicted by this model and the three vaccine characteristics (stability, translation efficiency and immunogenicity) and determines three objective functions (stability objective function, translation efficiency objective function, immunogenicity objective function) based on the identification results. In the specific mRNA vaccine design task processing process: the embodiment of the present invention first uses the mRNA sequence generation tool to batch generate the transcribed mRNA sequences of the target antigen protein to obtain an initialized mRNA sequence set; then uses the vaccine characteristic prediction model and three objective functions to iteratively optimize and screen the mRNA sequence set according to the multi-objective optimization method of the NSGA-II algorithm to obtain a preferred set with a reduced number of sequences; then based on in vitro experimental means, the stability and translation efficiency of the preferred set are experimentally analyzed; if the error between the in vitro experimental analysis results and the model prediction results is too large, the model is automatically optimized based on the experimental data of this experiment, and the mRNA sequence set is iteratively optimized again after optimization; if the error between the in vitro experimental analysis results and the model prediction results is controllable, the immunogenicity of the preferred set is further analyzed based on the in vivo experimental means; if the error between the in vivo experimental analysis results and the model prediction results is too large, the model is automatically optimized again based on the experimental data of this experiment, and the mRNA sequence set is iteratively optimized again after optimization; if the error between the in vivo experimental analysis results and the model prediction results is controllable, the current latest preferred set is output as the final design candidate set. The mRNA sequence initialization method of the embodiment of the present invention breaks away from the limitations of expert experience and improves the structural diversity of mRNA sequences; the vaccine characteristic prediction model of the embodiment of the present invention reduces a large number of experimental analyses, shortens the design cycle, and improves design efficiency; the model training method and multi-objective optimization process of the embodiment of the present invention effectively guarantee the design quality of mRNA sequences.

[0207] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0208] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for processing mRNA vaccine design tasks, characterized in that: The method comprises: A working model for predicting three vaccine characteristics of mRNA sequences is designed with the Uni-RNA model as the feature extraction backbone, and is recorded as the corresponding vaccine characteristic prediction model; the three vaccine characteristics include stability, translation efficiency and immunogenicity; the vaccine characteristic prediction model includes the Uni-RNA model, three basic feature prediction heads, a feature fusion module and three vaccine characteristic prediction heads; A large amount of antigen protein and its corresponding mRNA vaccine information is collected by big data collection to construct a corresponding antigen vaccine data set; and the vaccine characteristic prediction model is trained based on the antigen vaccine data set; and after the model training is completed, the maximum correlation between the three output features of the three basic feature prediction heads and the three output features of the three vaccine characteristic prediction heads is identified and the corresponding three objective functions are determined based on the identification results; the three objective functions include a stability objective function, a translation efficiency objective function and an immunogenicity objective function; Receive the molecular sequence of the target antigen protein input by the user as the corresponding first antigen sequence; and initialize the mRNA sequence according to the first antigen sequence to obtain the corresponding first mRNA sequence set; and according to the multi-objective optimization method of the NSGA-II algorithm, with the help of the vaccine characteristic prediction model, the three objective functions and the corresponding in vitro / in vivo experimental means, iteratively optimize the first mRNA sequence set to obtain the corresponding optimized mRNA sequence set to feed back to the user; the first mRNA sequence set consists of multiple first mRNA sequences; the optimized mRNA sequence set consists of one or more optimized mRNA sequences.

2. The method for processing the mRNA vaccine design task according to claim 1, characterized in that: The vaccine characteristic prediction model is used to predict the stability, translation efficiency and immunogenicity characteristics of the input mRNA sequence and output corresponding stability prediction data, translation efficiency prediction data and immunogenicity prediction data; The vaccine characteristic prediction model includes the Uni-RNA model, the three basic feature prediction heads, the feature fusion module and the three vaccine characteristic prediction heads; the three basic feature prediction heads include a codon usage frequency prediction head, an RNA secondary structure prediction head, and a GC content prediction head; the three vaccine characteristic prediction heads include a stability prediction head, a translation efficiency prediction head and an immunogenicity prediction head; The model input end of the vaccine characteristic prediction model is used to receive the corresponding mRNA sequence, and the first model output end, the second model output end and the third model output end are used to output the corresponding stability prediction data, the translation efficiency prediction data and the immunogenicity prediction data respectively; The input end of the Uni-RNA model is connected to the input end of the model, and the output end is respectively connected to the input ends of the codon usage frequency prediction head, the RNA secondary structure prediction head and the GC content prediction head; the output ends of the codon usage frequency prediction head, the RNA secondary structure prediction head and the GC content prediction head are connected to the first, second and third input ends of the feature fusion module; the output end of the feature fusion module is respectively connected to the input ends of the stability prediction head, the translation efficiency prediction head and the immunogenicity prediction head; the output ends of the stability prediction head, the translation efficiency prediction head and the immunogenicity prediction head are respectively connected to the corresponding first, second and third model output ends; The Uni-RNA model has completed model pre-training; the Uni-RNA model is used to perform feature encoding processing on the input mRNA sequence to obtain a corresponding feature encoding vector and send it to the codon usage frequency prediction head, the RNA secondary structure prediction head and the GC content prediction head; The codon usage frequency prediction head is implemented based on a type of nonlinear regression model or a type of neural network model; the codon usage frequency prediction head is used to predict the codon usage frequency of each base position on the mRNA sequence according to the feature coding vector to obtain the corresponding codon usage frequency prediction vector and send it to the feature fusion module; the nonlinear regression model at least includes an MLP model, an NNR model, a GBDT model and an XGBoost model; the neural network model at least includes an MLP model, a CNN model, a ResNet model and a GNN model; The RNA secondary structure prediction head is implemented based on another type of neural network model; the RNA secondary structure prediction head is used to predict the secondary structure features of each base position on the mRNA sequence according to the feature encoding vector to obtain the corresponding secondary structure prediction vector and send it to the feature fusion module; The GC content prediction head is implemented based on another type of the nonlinear regression model or another type of the neural network model; the GC content prediction head is used to predict the GC content of each base position on the mRNA sequence according to the feature encoding vector to obtain a corresponding GC content prediction vector and send it to the feature fusion module; The feature fusion module is used to perform vector splicing on the codon usage frequency prediction vector, the secondary structure prediction vector and the GC content prediction vector to obtain a corresponding splicing vector and send it to the stability prediction header, the translation efficiency prediction header and the immunogenicity prediction header; The stability prediction head is implemented based on another type of the nonlinear regression model or another type of the neural network model; the stability prediction head is used to predict the stability characteristics of the mRNA sequence according to the splicing vector and output the corresponding stability prediction data; The translation efficiency prediction head is implemented based on another type of the nonlinear regression model or another type of the neural network model; the translation efficiency prediction head is used to predict the translation characteristics of the mRNA sequence according to the splicing vector and output the corresponding translation efficiency prediction data; The immunogenicity prediction head is implemented based on another type of the nonlinear regression model or another type of the neural network model; the immunogenicity prediction head is used to predict the immunogenicity characteristics of the mRNA sequence according to the splicing vector and output the corresponding immunogenicity prediction data.

3. The method for processing the mRNA vaccine design task according to claim 2, characterized in that: The antigen vaccine data set includes multiple antigen vaccine data records; the antigen vaccine data records include antigen molecule sequences, mRNA molecule sequences, codon usage frequency label vectors, secondary structure label vectors, GC content label vectors, stability label data, translation efficiency label data and immunogenicity label data.

4. The method for processing the mRNA vaccine design task according to claim 3, characterized in that: The performing model training on the vaccine characteristic prediction model based on the antigen vaccine data set specifically includes: Extracting the mRNA molecular sequence, the codon usage frequency label vector, the secondary structure label vector and the GC content label vector of each antigen vaccine data record in the antigen vaccine data set to form a corresponding first data record; and forming a corresponding first data set from all the obtained first data records; and extracting the mRNA molecular sequence, the stability label data, the translation efficiency label data and the immunogenicity label data of each of the antigen vaccine data records in the antigen vaccine data set to form a corresponding second data record; and all the obtained second data records form a corresponding second data set; Based on the first data set, the first stage model training is performed on the Uni-RNA model and the three basic feature prediction heads of the vaccine characteristic prediction model; after the first round of model training is completed, the second stage model training is performed on the three vaccine characteristic prediction heads of the vaccine characteristic prediction model based on the second data set; and after the second round of model training is completed, the completion of this model training is confirmed.

5. The method for processing the mRNA vaccine design task according to claim 4, characterized in that: The first stage model training of the Uni-RNA model and the three basic feature prediction heads of the vaccine characteristic prediction model based on the first data set specifically includes: Step 51, based on a preset first segmentation ratio, randomly split the first data set into two sub-data sets recorded as a corresponding first training set and a first evaluation set; Wherein, both the first training set and the first evaluation set are composed of a plurality of the first data records; the ratio of the total number of records in the first training set to the total number of records in the first evaluation set satisfies the first segmentation ratio; Step 52, extracting the first first data record of the first training set as the corresponding current training record; Step 53: record the codon usage frequency label vector, the secondary structure label vector and the GC content label vector of the current training record as the corresponding first label vector P tag1 , the second label vector P tag2 and the third label vector P tag3 ; Step 54, input the mRNA molecular sequence of the current training record into the Uni-RNA model for feature encoding processing to obtain a corresponding first feature encoding vector; and input the first feature encoding vector into the codon usage frequency prediction header, the RNA secondary structure prediction header and the GC content prediction header for prediction processing to obtain a corresponding first prediction vector Ppr e1 , the second prediction vector Ppr e2 and the third prediction vector Ppr e3 ; Step 55: transform the first prediction vector P pre1 And the corresponding first label vector P tag1 Substitute the preset first loss function L a , and the second prediction vector Ppr e2 And the corresponding second label vector P tag2 Substitute the preset second loss function L b , and the third prediction vector P pre3 And the corresponding third label vector P tag3 Bring in the preset third loss function L c , and by the first loss function L a 、The second loss function L b And the third loss function L c Add together to form the corresponding first model loss function L M1 ; and calculate the function values ​​of the first, second, third loss functions and the first model loss function and use the calculation results as the corresponding first loss value, second loss value, third loss value and first overall loss value; Among them, the first model loss function L M1 for: L M1 =L a (P pre1 ,P tag1 )+L b (P pre2 ,P tag2 )+L c (P pre3 ,P tag3 ); The first loss function L a At least including L1 loss function and L2 loss function; the second loss function L b It is composed of a base type loss function and a base structure loss function, wherein the base type loss function at least includes a cross entropy loss function, and the base structure loss function at least includes an L1 loss function and an L2 loss function; the third loss function L c At least including L1 loss function and L2 loss function; Step 56, identifying whether the first, second, third loss values ​​and the first overall loss value all satisfy the corresponding first loss value range, second loss value range, third loss value range and first overall loss value range; if the first, second, third loss values ​​and the first overall loss value all satisfy the corresponding loss value range, go to step 57; if the first loss value does not satisfy the first loss value range, then based on the preset first model optimizer, move towards making the first loss function L a The model parameters of the Uni-RNA model and the codon usage frequency prediction head are modulated in a round in the direction of reaching the minimum value, and the step 54 is returned to when the parameter modulation of this round is finished; if the second loss value does not satisfy the second loss value range, the second model optimizer is based on the preset second model to make the second loss function L b The model parameters of the Uni-RNA model and the RNA secondary structure prediction head are modulated in a round in the direction of reaching the minimum value, and the step 54 is returned to when the parameter modulation of this round is finished; if the third loss value does not satisfy the third loss value range, the third model optimizer is based on the preset third model to make the third loss function L c The model parameters of the Uni-RNA model and the GC content prediction head are modulated in a round in the direction of reaching the minimum value, and the step 54 is returned to when the parameter modulation of this round is finished; if the first overall loss value does not satisfy the first overall loss value range, the fourth model optimizer is based on a preset fourth model optimizer in a direction of making the first model loss function L M1 Perform a round of modulation on the model parameters of the Uni-RNA model and the three basic feature prediction heads in the direction of reaching the minimum value, and return to step 54 when this round of parameter modulation is completed; Wherein, the first, second, third and fourth model optimizers all include at least an SGD optimizer and an ADAM optimizer; Step 57, identifying whether the current training record is the last first data record of the first training set; if so, going to step 58; if not, extracting the next first data record of the first training set as the new current training record and returning to step 53; Step 58, perform a round of traversal on all the first data records of the first evaluation set; and in this round of traversal, use the first data record currently traversed as the corresponding current evaluation record; and input the mRNA molecule sequence of the current evaluation record into the Uni-RNA model for feature encoding processing to obtain the corresponding second feature encoding vector; and input the second feature encoding vector into the three basic feature prediction heads respectively for prediction processing to obtain the corresponding fourth, fifth and sixth prediction vectors to form a corresponding first prediction vector sequence; and record the codon usage frequency label vector, the secondary structure label vector and the GC content label vector of the current evaluation record as the corresponding fourth, fifth and sixth label vectors to form a corresponding first label vector sequence; and the first prediction vector sequence and the first label vector sequence form a corresponding first prediction-label pair; and at the end of this round of traversal, input all the obtained first prediction-label pairs into a preset first model evaluation function to calculate and obtain the corresponding first evaluation value; Wherein, the first model evaluation function at least includes an RMSE function; Step 59, identifying whether the first evaluation value meets the preset first evaluation value range; if not, returning to step 51 to continue training; if satisfied, confirming that the first stage model training is completed.

6. The method for processing the mRNA vaccine design task according to claim 4, characterized in that: The performing second-stage model training on the three vaccine characteristic prediction heads of the vaccine characteristic prediction model based on the second data set specifically includes: Step 61, based on a preset second segmentation ratio, randomly split the second data set into two sub-data sets recorded as corresponding second training sets and second evaluation sets; Wherein, both the second training set and the second evaluation set are composed of a plurality of the second data records; the ratio of the total number of records in the second training set to the total number of records in the second evaluation set satisfies the second segmentation ratio; Step 62, extracting the first second data record of the second training set as the corresponding current training record; Step 63: record the stability label data, the translation efficiency label data and the immunogenicity label data of the current training record as corresponding first label data D tag1 , the second label data D tag2 and the third tag data D tag3 ; Step 64: input the mRNA molecular sequence of the current training record into the vaccine characteristic prediction model for prediction processing to obtain the corresponding first prediction data D pre1 , the second prediction data D pre2 and the third prediction vector D pre3 ; Step 65: convert the first prediction data D pre1 And the corresponding first tag data D tag1 Substitute the preset fourth loss function L e , and the second prediction data D pre2 And the corresponding second tag data D tag2 Substitute the preset fifth loss function L f , and the third prediction vector D pre3 And the corresponding third tag data D tag3 Substitute the preset sixth loss function L g , and the fourth loss function L e , the fifth loss function L f And the sixth loss function L g Add together to form the corresponding second model loss function L M2 ; and calculate the function values ​​of the fourth, fifth, sixth loss functions and the second model loss function and use the calculation results as the corresponding fourth loss value, fifth loss value, sixth loss value and second overall loss value; Among them, the second model loss function L M2 for: L M2 =L e (D pre1 ,D tag2 )+L f (D pre2 ,D tag2 )+L g (D pre3 ,D tag3 ); The fourth loss function L e At least including L1 loss function, L2 loss function, and cross entropy loss function; the fifth loss function L f At least including L1 loss function, L2 loss function; the sixth loss function L g At least including L1 loss function, L2 loss function, and cross entropy loss function; Step 66, identifying whether the fourth, fifth, sixth loss values ​​and the second overall loss value all satisfy the corresponding fourth loss value range, fifth loss value range, sixth loss value range and second overall loss value range; if the fourth, fifth, sixth loss values ​​and the second overall loss value all satisfy the corresponding loss value range, go to step 67; if the fourth loss value does not satisfy the fourth loss value range, then based on the preset fifth model optimizer, move towards making the fourth loss function L e The model parameters of the stability prediction head are modulated in a round in the direction of reaching the minimum value, and the step 64 is returned to when the parameter modulation of this round is finished; if the fifth loss value does not satisfy the fifth loss value range, the sixth model optimizer based on the preset sixth model optimizer moves toward making the fifth loss function L f The model parameters of the translation efficiency prediction head are modulated in a round in the direction of reaching the minimum value, and the step 64 is returned to when the parameter modulation of this round is finished; if the sixth loss value does not satisfy the sixth loss value range, the seventh model optimizer based on the preset seventh model optimizer moves in the direction of making the sixth loss function L g The model parameters of the immunogenicity prediction head are modulated in a round in the direction of reaching the minimum value, and the step 64 is returned to when the parameter modulation of this round ends; if the second overall loss value does not satisfy the second overall loss value range, the eighth model optimizer based on the preset eighth model optimizer moves in the direction of making the second model loss function L M2 Perform a round of modulation on the model parameters of the three vaccine characteristic prediction heads in the direction of reaching the minimum value, and return to step 64 when this round of parameter modulation is completed; Wherein, the fifth, sixth, seventh and eighth model optimizers all include at least an SGD optimizer and an ADAM optimizer; Step 67, identifying whether the current training record is the last second data record of the second training set; if so, going to step 68; if not, extracting the next second data record of the second training set as the new current training record and returning to step 63; Step 68, perform a round of traversal on all the second data records of the second evaluation set; and in this round of traversal, use the second data record currently traversed as the corresponding current evaluation record; and input the mRNA molecular sequence of the current evaluation record into the vaccine characteristic prediction model for prediction processing to obtain the corresponding fourth, fifth and sixth prediction data to form the corresponding first prediction data sequence; and record the stability label data, the translation efficiency label data and the immunogenicity label data of the current evaluation record as the corresponding fourth, fifth and sixth label data to form the corresponding first label data sequence; and the first prediction data sequence and the first label data sequence form the corresponding second prediction-label pair; and at the end of this round of traversal, input all the obtained second prediction-label pairs into the preset second model evaluation function to calculate and obtain the corresponding second evaluation value; Wherein, the second model evaluation function at least includes an RMSE function; Step 69, identifying whether the second evaluation value meets the preset second evaluation value range; if not, returning to step 61 to continue training; if satisfied, confirming that the second stage model training is completed.

7. The method for processing the mRNA vaccine design task according to claim 3, characterized in that: After the model training is completed, the maximum correlation between the three output features of the three basic feature prediction heads and the three output features of the three vaccine characteristic prediction heads is identified, and the corresponding three objective functions are determined based on the identification results, specifically including: Step 71, after the model training is completed, the mapping relationship between the three types of basic feature prediction vectors generated by the vaccine characteristic prediction model in the model prediction process and the three types of characteristic prediction data finally output by the model is recorded as the corresponding stability mapping relationship F A , translation efficiency mapping relationship F B and immunogenicity mapping relationship F C ; where y a = F A (x a , x b , x c ), y b = F B (x a , x b , x c ), y c = F C (x a , x b , x c ); Independent variable x a 、x b 、x c are the codon usage frequency prediction vector, the secondary structure prediction vector and the GC content prediction vector generated in the model prediction process, respectively, and the dependent variable y a ,y b ,y c The stability prediction data, the translation efficiency prediction data and the immunogenicity prediction data finally output by the model are respectively; Step 72, record each antigen vaccine data record in the antigen vaccine data set as record R i , 1≤i≤N, N is the total number of records in the antigen vaccine data set; and each record R i The mRNA molecular sequence is input into the vaccine characteristic prediction model for prediction processing to obtain the corresponding six prediction results, which are x a,i 、x b,i 、x c,i ,y a,i ,y b,i ,y c,i ; and the prediction result x corresponding to each index i a,i 、x b,i 、x c,i ,y a,i Form a corresponding first independent / dependent variable data group, and form a corresponding first data group set by the obtained N first independent / dependent variable data groups; and form a corresponding first data group set by the prediction result x corresponding to each index i a,i 、x b,i 、x c,i ,y b,i Form a corresponding second independent / dependent variable data group, and form a corresponding second data group set by the obtained N second independent / dependent variable data groups; and form a corresponding second data group set by the prediction result x corresponding to each index i a,i 、x b,i 、x c,i ,y c,i Forming a corresponding third independent / dependent variable data group, and forming a corresponding third data group set from the obtained N third independent / dependent variable data groups; Step 73, based on the SHAP evaluation algorithm, evaluate the independent variable x according to the first data set a 、x b 、x c For the dependent variable y a The contribution coefficient of the independent variable x with the largest contribution coefficient a 、x b or x c As the dependent variable y a The maximum associated variable x ma ; and based on the SHAP evaluation algorithm, evaluate the independent variable x according to the second data set a 、x b 、x c For the dependent variable y b The contribution coefficient of the independent variable x with the largest contribution coefficient a 、x b or x c As the dependent variable y b The maximum associated variable x mb ; and based on the SHAP evaluation algorithm, evaluate the independent variable x according to the third data set a 、x b 、x c For the dependent variable y c The contribution coefficient of the independent variable x with the largest contribution coefficient a 、x b or x c As the dependent variable y c The maximum associated variable x mc ; Step 74, the dependent variable y a The maximum associated variable x corresponding to it ma Substitute the preset correlation function F corr Get the corresponding stability objective function F corr (x ma ,y a ) and the dependent variable y b The corresponding maximum associated variable x mb Substitute the correlation function F corr Get the corresponding translation efficiency objective function F corr (x mb ,y b ) and the dependent variable y c The maximum associated variable x corresponding to it mc Substitute the correlation function F corr Get the corresponding immunogenicity objective function F corr (x mc ,y c ); and the stability objective function, the translation efficiency objective function and the immunogenicity objective function form the corresponding three objective functions; the correlation function F corr At least including Pearson correlation function, Spearman correlation function, and Kendall correlation function.

8. The method for processing the mRNA vaccine design task according to claim 1, characterized in that: Initializing the mRNA sequence according to the first antigen sequence to obtain the corresponding first mRNA sequence set specifically includes: A preset mRNA sequence generation tool is used to perform sequence generation processing on the transcribed mRNA sequence of the first antigen sequence to obtain a plurality of corresponding first mRNA sequences forming the corresponding first mRNA sequence set; wherein the mRNA sequence generation tool is a type of mRNA sequence reverse translation tool or a type of mRNA sequence reverse decoding model; the mRNA sequence reverse translation tool at least includes the Translate Tool tool of ExPASy, which is used to reverse translate and output the transcribed mRNA sequence corresponding to the specified antigen sequence; the mRNA sequence reverse decoding model at least includes a reverse decoding model implemented based on a GANs model, which is used to reverse decode and output the transcribed mRNA sequence corresponding to the specified antigen sequence.

9. The method for processing the mRNA vaccine design task according to claim 1, characterized in that: The multi-objective optimization method according to the NSGA-II algorithm, using the vaccine characteristic prediction model, the three objective functions and the corresponding in vitro / in vivo experimental means, performs iterative optimization processing on the first mRNA sequence set to obtain the corresponding optimized mRNA sequence set and feed it back to the user, specifically including: Step 91, initializing the first iteration counter to 1; Step 92, input each of the first mRNA sequences of the first mRNA sequence set into the vaccine characteristic prediction model for prediction, and bring the prediction results into the three objective functions for calculation to obtain the corresponding first objective function value, second objective function value and third objective function value; and sort all the first objective function values, all the second objective function values ​​and all the third objective function values ​​obtained from high to low according to the function value to obtain the corresponding first, second and third function value sequences; Step 93, according to the non-dominated level and crowding distance identification rules in the NSGA-II algorithm, the non-dominated level and crowding distance corresponding to each of the first mRNA sequences are confirmed based on the first, second and third function value sequences; and based on the tournament selection strategy in the NSGA-II algorithm, mRNA sequence individuals are selected according to the non-dominated level and crowding distance of each of the first mRNA sequences, and the first individual set is formed by all the selected first mRNA sequences; and based on the preset individual crossover rule, all the first mRNA sequences in the first individual set are subjected to individual crossover processing, and each of the first mRNA sequences that have completed the crossover is subjected to individual variation processing; and after all the first mRNA sequences in the first individual set have completed individual variation, the first individual set and the first mRNA sequence set are merged to obtain a new first mRNA sequence set; Wherein, the individual crossover rules include single-point crossover and multi-point crossover; Step 94, adding 1 to the first iteration counter; and identifying whether the first iteration counter after adding 1 exceeds a preset iteration counter threshold; if so, go to step 95; if not, return to step 92; Step 95, inputting each of the first mRNA sequences of the latest first mRNA sequence set into the vaccine characteristic prediction model to predict and obtain corresponding first stability prediction data, first translation efficiency prediction data and first immunogenicity prediction data; and sorting all the first stability prediction data and all the first translation efficiency prediction data obtained based on the preset stability and translation efficiency sorting rules, and extracting one or more first mRNA sequences whose stability and translation efficiency are ranked before the specified ranking to form a corresponding preferred set; Step 96, performing experimental analysis on the stability and translation efficiency of each of the first mRNA sequences in the preferred set by in vitro experimental means to obtain corresponding first stability experimental data and first translation efficiency experimental data; and calculating the relative error between the first stability prediction data and the first stability experimental data of each of the first mRNA sequences in the preferred set to obtain a corresponding first relative error, and calculating the relative error between the first translation efficiency prediction data and the first translation efficiency experimental data of each of the first mRNA sequences in the preferred set to obtain a corresponding second relative error; Step 97, identifying whether all the first and second relative errors obtained are less than the corresponding first and second relative error thresholds; if so, go to step 98; if not, use all the first stability experimental data and the first translation efficiency experimental data obtained this time as label data, perform a round of parameter optimization processing on the stability prediction head and the translation efficiency prediction head of the vaccine characteristic prediction model according to the supervised model training method, and return to step 91 at the end of this round of optimization processing; Step 98, performing experimental analysis on the immunogenicity characteristics of each of the first mRNA sequences in the preferred set by means of in vivo experiments to obtain corresponding first immunogenicity experimental data; and calculating the relative error between the first immunogenicity prediction data and the first immunogenicity experimental data of each of the first mRNA sequences in the preferred set to obtain corresponding third relative errors; Step 99, identify whether all the third relative errors obtained are less than the corresponding third relative error threshold; if not, use all the first immunogenicity experimental data obtained this time as label data, and perform a round of parameter optimization processing on the immunogenicity prediction head of the vaccine characteristic prediction model according to the supervised model training method, and return to step 91 at the end of this round of optimization processing; if so, use each of the first mRNA sequences in the preferred set as the corresponding optimized mRNA sequence, and form the corresponding optimized mRNA sequence set by all the optimized mRNA sequences obtained and output it.

10. A device for executing the processing method of the mRNA vaccine design task according to any one of claims 1 to 9, characterized in that: The device comprises: a working model building module, a working model training module and a vaccine design task processing module; The working model construction module is used to design a working model for predicting three vaccine characteristics of mRNA sequences using the Uni-RNA model as the feature extraction backbone, which is recorded as the corresponding vaccine characteristic prediction model; the three vaccine characteristics include stability, translation efficiency and immunogenicity; the vaccine characteristic prediction model includes the Uni-RNA model, three basic feature prediction heads, a feature fusion module and three vaccine characteristic prediction heads; The working model training module is used to collect a large amount of antigen proteins and their corresponding mRNA vaccine information through big data collection to construct a corresponding antigen vaccine data set; and to train the vaccine characteristic prediction model based on the antigen vaccine data set; and after the model training is completed, the maximum correlation relationship between the three output features of the three basic feature prediction heads and the three output features of the three vaccine characteristic prediction heads is identified and the corresponding three objective functions are determined based on the identification results; the three objective functions include a stability objective function, a translation efficiency objective function and an immunogenicity objective function; The vaccine design task processing module is used to receive the molecular sequence of the target antigen protein input by the user as the corresponding first antigen sequence; and initialize the mRNA sequence according to the first antigen sequence to obtain the corresponding first mRNA sequence set; and according to the multi-objective optimization method of the NSGA-II algorithm, with the help of the vaccine characteristic prediction model, the three objective functions and the corresponding in vitro / in vivo experimental means, iteratively optimize the first mRNA sequence set to obtain the corresponding optimized mRNA sequence set and feed it back to the user; the first mRNA sequence set consists of multiple first mRNA sequences; the optimized mRNA sequence set consists of one or more optimized mRNA sequences.

11. An electronic device, characterized in that: include: memory, processors, and transceivers; The processor is used to couple with the memory, read and execute instructions in the memory, so as to implement the method according to any one of claims 1 to 9; The transceiver is coupled to the processor, and the processor controls the transceiver to send and receive messages.

12. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and when the computer instructions are executed by a computer, the computer executes the method according to any one of claims 1 to 9.

Images