Antigen prediction model processing method, device and equipment and readable storage medium

Abstract

The invention relates to an antigen prediction model processing method and device, equipment and a readable storage medium. The method comprises the following steps: acquiring a training sample set; the training sample set comprises a protein sequence feature matrix and an antigen distance tag vector corresponding to the protein sequence feature matrix; the protein sequence feature matrix comprises a plurality of protein sequence feature vectors; aiming at any one protein sequence feature vector of the plurality of protein sequence feature vectors, determining a key protein sequence feature vector from the plurality of protein sequence feature vectors according to an inner product of an antigen distance tag vector and a normalized feature vector corresponding to the any one protein sequence feature vector; determining a preset antigen prediction model according to the key protein sequence feature vector; the preset antigen prediction model is used for outputting an antigen prediction result of the to-be-detected protein sequence feature matrix. By adopting the method, the accuracy of the antigen prediction model can be improved.

Description

Technical Field

[0001] The present application relates to the field of machine learning technology, and in particular to an antigen prediction model processing method, device, equipment and readable storage medium. Background Art

[0002] Influenza is an infectious disease caused by respiratory viruses, which causes large-scale spread and impact around the world every year. The selection of influenza vaccine strains each year is based on close monitoring of global pandemic influenza viruses. Viral antigens are predicted months before the flu season to ensure that all aspects of vaccine production can proceed smoothly. However, in the face of some sudden influenza pandemics, it is extremely challenging to produce and distribute vaccines against new viruses in a short period of time. Therefore, through more accurate antigen prediction methods, more accurate prediction of the evolutionary trend of influenza virus antigens is of great significance for vaccine research and development and production, which can not only improve the timeliness of vaccines, but also better provide effective immune protection for the public.

[0003] In traditional technology, antigens can be predicted through artificial intelligence models such as support vector machines, but these models are prone to overfitting when processing high-dimensional data sets, resulting in low accuracy of the antigen prediction model. Summary of the invention

[0004] Based on this, it is necessary to provide an antigen prediction model processing method, device, computer equipment, computer-readable storage medium and computer program product that can improve the accuracy of the antigen prediction model in response to the above technical problems.

[0005] In a first aspect, the present application provides an antigen prediction model processing method, comprising:

[0006] Acquire a training sample set; the training sample set includes a protein sequence feature matrix and an antigen distance label vector corresponding to the protein sequence feature matrix; the protein sequence feature matrix includes a plurality of protein sequence feature vectors;

[0007] For any protein sequence feature vector of the multiple protein sequence feature vectors, determining a key protein sequence feature vector from the multiple protein sequence feature vectors according to the inner product of the antigen distance label vector and the normalized feature vector corresponding to the any protein sequence feature vector;

[0008] According to the key protein sequence feature vector, a preset antigen prediction model is determined; the preset antigen prediction model is used to output the antigen prediction result of the protein sequence feature matrix to be detected.

[0009] In one embodiment, for any protein sequence feature vector of the multiple protein sequence feature vectors, determining a key protein sequence feature vector from the multiple protein sequence feature vectors according to the inner product of the antigen distance label vector and the normalized feature vector corresponding to the any protein sequence feature vector comprises:

[0010] For any protein sequence feature vector of the multiple protein sequence feature vectors, determine the square modulus value of the inner product corresponding to any protein sequence feature vector according to the square modulus of the inner product of the antigen distance label vector and the normalized feature vector corresponding to the any protein sequence feature vector;

[0011] According to the inner product modulus square values ​​corresponding to the multiple protein sequence feature vectors, determining the protein sequence feature vector corresponding to the largest inner product modulus square value as the key protein sequence feature vector;

[0012] The step of taking the other protein sequence feature vectors among the multiple protein sequence feature vectors except the key protein sequence feature vector as a new multiple protein sequence feature vectors, returning any protein sequence feature vector for the multiple protein sequence feature vectors, and determining the modulus square value of the inner product corresponding to any protein sequence feature vector according to the modulus square of the inner product of the normalized feature vector corresponding to the antigen distance label vector and the any protein sequence feature vector.

[0013] In one embodiment, before determining the square modulus value of the inner product corresponding to any protein sequence feature vector according to the square modulus of the inner product of the antigen distance tag vector and the normalized feature vector corresponding to any protein sequence feature vector, the method further comprises:

[0014] When the first key protein sequence feature vector is determined, dividing any protein sequence feature vector by its corresponding norm to obtain a normalized feature vector corresponding to any protein sequence feature vector;

[0015] When the Nth key protein sequence feature vector is determined, the projection vector of any protein sequence feature vector on the last determined key protein sequence feature vector is subtracted from the any protein sequence feature vector to obtain a normalized feature vector corresponding to the any protein sequence feature vector; N is an integer greater than 1; and the projection vector is equal to the product of the inner product value of the any protein sequence feature vector and the last determined key protein sequence feature vector and the last determined key protein sequence feature vector.

[0016] In one embodiment, after determining the protein sequence feature vector corresponding to the largest inner product modulus square value as the key protein sequence feature vector, the method further comprises:

[0017] Determine whether the maximum inner product modulus square value is less than or equal to a convergence threshold;

[0018] If the maximum inner product modulus square value is less than or equal to the convergence threshold, return to the step of using the other protein sequence feature vectors among the multiple protein sequence feature vectors except the key protein sequence feature vector as the new multiple protein sequence feature vectors until the maximum inner product modulus square value is greater than the convergence threshold.

[0019] In one embodiment, determining a preset antigen prediction model according to the key protein sequence feature vector comprises:

[0020] Determining model prediction parameters according to the inner product of the key protein sequence feature vector and the antigen distance label vector;

[0021] The preset antigen prediction model is constructed according to the model prediction parameters.

[0022] In one of the embodiments, determining the model prediction parameters according to the inner product of the key protein sequence feature vector and the antigen distance label vector includes: decomposing a key protein sequence feature matrix composed of a plurality of the key protein sequence feature vectors into an orthogonal matrix and an upper triangular matrix;

[0023] The model prediction parameters are determined according to the key protein sequence feature matrix and the upper triangular matrix.

[0024] In a second aspect, the present application also provides an antigen prediction model processing device, comprising:

[0025] A sample acquisition module is used to acquire a training sample set; the training sample set includes a protein sequence feature matrix and an antigen distance label vector corresponding to the protein sequence feature matrix; the protein sequence feature matrix includes a plurality of protein sequence feature vectors;

[0026] A feature determination module, for determining a key protein sequence feature vector from any of the multiple protein sequence feature vectors according to the inner product of the antigen distance label vector and the normalized feature vector corresponding to any of the multiple protein sequence feature vectors;

[0027] The model determination module is used to determine a preset antigen prediction model according to the key protein sequence feature vector; the preset antigen prediction model is used to output the antigen prediction result of the protein sequence feature matrix to be detected.

[0028] In a third aspect, the present application further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above method when executing the computer program.

[0029] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the steps of the above method when executed by a processor.

[0030] In a fifth aspect, the present application also provides a computer program product, including a computer program, which implements the steps of the above method when executed by a processor.

[0031] The above-mentioned antigen prediction model processing method, device, equipment and readable storage medium obtain a training sample set, wherein the training sample set includes a protein sequence feature matrix composed of multiple protein sequence feature vectors and an antigen distance label vector corresponding to the protein sequence feature matrix; for any protein sequence feature vector of the protein sequence feature vectors, by calculating the inner product of the antigen distance label vector and the normalized feature vector corresponding to any protein sequence feature vector, a key protein sequence feature vector is determined from multiple protein sequence feature vectors; according to the key protein sequence feature vector, a preset antigen prediction model is determined, and the preset antigen prediction model is used to output the antigen prediction result of the protein sequence feature matrix to be detected, and the key protein sequence feature vector is determined by the inner product of the antigen distance label vector and the normalized feature vector corresponding to the protein sequence feature vector. Even when processing high-dimensional data sets, it can ensure that the selected feature vector is the most representative, and an accurate antigen prediction model is constructed based on the key protein sequence feature vector, thereby improving the accuracy of the antigen prediction model. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0033] FIG1 is an application environment diagram of an antigen prediction model processing method in one embodiment;

[0034] FIG2 is a schematic flow chart of an antigen prediction model processing method in one embodiment;

[0035] FIG3 is a logic diagram of an antigen prediction model process in one embodiment;

[0036] FIG4 is a logic diagram of an antigen prediction model process in another embodiment;

[0037] FIG5 is a schematic diagram of performance evaluation of an antigen prediction model in one embodiment;

[0038] FIG6 is a schematic flow chart of an antigen prediction model processing method in another embodiment;

[0039] FIG7 is a block diagram of an antigen prediction model processing device in one embodiment;

[0040] FIG. 8 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0041] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0042] The antigen prediction model processing method provided in the embodiment of the present application can be applied in the application environment shown in Figure 1. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be integrated on the server 104, or it can be placed on the cloud or other network servers. The terminal 102 obtains a training sample set; the training sample set includes a protein sequence feature matrix and an antigen distance label vector corresponding to the protein sequence feature matrix; the protein sequence feature matrix contains multiple protein sequence feature vectors; the terminal 102 determines the key protein sequence feature vector from the multiple protein sequence feature vectors according to the inner product of the antigen distance label vector and the normalized feature vector corresponding to any protein sequence feature vector for any protein sequence feature vector; the terminal 102 determines the preset antigen prediction model based on the key protein sequence feature vector; the preset antigen prediction model is used to output the antigen prediction result of the protein sequence feature matrix to be detected. The terminal 102 may be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, IoT devices, and portable wearable devices. The IoT devices may be smart speakers, smart TVs, smart air conditioners, smart car devices, projection devices, etc. Portable wearable devices may be smart watches, smart bracelets, head-mounted devices, etc. Head-mounted devices may be virtual reality (VR) devices, augmented reality (AR) devices, smart glasses, etc. The server 104 may be an independent physical server, a server cluster or distributed system consisting of multiple physical servers, or a cloud server that provides cloud computing services.

[0043] In an exemplary embodiment, as shown in FIG2 , a method for processing an antigen prediction model is provided, and the method is described by taking the application of the method to the terminal 102 in FIG1 as an example, including:

[0044] Step S202, obtaining a training sample set.

[0045] The training sample set includes a protein sequence feature matrix and an antigen distance label vector corresponding to the protein sequence feature matrix.

[0046] Among them, the protein sequence feature matrix contains multiple protein sequence feature vectors.

[0047] In the specific implementation, the value measured by the hemagglutination inhibition test (HI) of the influenza virus can be obtained, which is referred to as the HI titer value. The HI titer value is used to measure the hemagglutination inhibition test reaction value of different viruses and each serum. Each HI titer value can correspond to a virus-antiserum pair. The virus can refer to a circulating or tested virus isolate, and the antiserum can be produced based on a certain organism after infection. The antigenic difference of each virus-antiserum pair can be calculated by the HI titer value. The antigenic difference can be defined as the 2-fold dilution difference between the homologous titer value and the heterologous titer value. The antigenic difference can be used as a quantitative indicator to reflect the change in the antigenicity of the virus. The HI titer value can be used to calculate the antigenic distance between each virus. First, the matrix of the hemagglutination inhibition test reaction values ​​of different viruses and each serum can be supplemented with missing data and low reaction values ​​by the low-rank matrix completion method, and then the antigenic distance vector Y between each virus is calculated by the following formula.

[0048] ;

[0049] Where Y is the antigen distance vector, is a matrix of the hemagglutination inhibition test reaction values ​​(HI titer values) of different viruses and each serum; this formula can be used to determine the antigen distance label vector in the training sample set.

[0050] For each virus-antiserum pair, the amino acid sequence of the hemagglutinin (HA) protein corresponding to each virus (abbreviated as "HA sequence") can be obtained. Hemagglutinin is an important glycoprotein on the surface of influenza virus and one of the main antigens. It is often used in influenza vaccine development and antigen prediction. The amino acid sequence of the HA protein of each virus-antiserum pair can be used to generate the protein sequence feature matrix in the training sample set.

[0051] Optionally, a portion containing antigenic epitopes may be cut out from the HA protein sequence as a protein sequence feature matrix in the training sample set. For example, the protein sequence feature matrix is ​​determined based on the HA1 subunit in the HA sequence, with the signal peptide in the HA1 subunit fragment as the starting position and the cleavage site between the HA1 subunit and the HA2 subunit as the ending position, to determine the sequence fragment cut out from the HA1 subunit, because the HA1 subunit forms the globular head of the HA protein sequence and may contain antigenically important key epitopes of the influenza virus.

[0052] Optionally, the pattern-induced multiple sequence alignment (PIMA coding) function can be used to generate a similarity matrix X from the aligned HA sequences, where each element X[i][j] represents the similarity between the i-th and j-th sequences, and the similarity matrix X is used as the protein sequence feature matrix in the training sample set.

[0053] Step S204, for any protein sequence feature vector of the multiple protein sequence feature vectors, according to the inner product of the antigen distance label vector and the normalized feature vector corresponding to any protein sequence feature vector, a key protein sequence feature vector is determined from the multiple protein sequence feature vectors.

[0054] In a specific implementation, any protein sequence feature vector can be normalized to obtain a normalized feature vector corresponding to the protein sequence feature vector, so that each feature vector is within the same range, thereby improving the accuracy of subsequent calculation of the inner product.

[0055] Among them, the key protein sequence feature vector is the protein sequence feature vector with the highest correlation with the antigen distance label vector in the protein sequence feature matrix, and is the most representative protein sequence feature vector for capturing antigen evolution information.

[0056] For any protein sequence feature vector, the inner product of the antigen distance label vector and the normalized feature vector corresponding to the protein sequence feature vector is calculated. The inner product calculation can be used to measure the correlation between the two vectors. Therefore, based on the calculated inner product, the protein sequence feature vector with the highest correlation with the antigen distance label vector can be determined from multiple protein sequence feature vectors as the key protein sequence feature vectors. These key protein sequence feature vectors can capture the complex patterns of antigen evolution and are the most representative.

[0057] Step S206, determining a preset antigen prediction model according to the key protein sequence feature vector.

[0058] In a specific implementation, the preset antigen prediction model is determined according to the key protein sequence feature vector, and the model prediction parameters are determined according to the key protein sequence feature vector, and the preset antigen prediction model is determined according to the training sample set and the model prediction parameters. Optionally, the antigen prediction model can be a machine learning model such as a random forest model, which is not specifically limited.

[0059] Among them, the preset antigen prediction model is used to output the antigen prediction results of the protein sequence feature matrix to be detected. Optionally, the protein sequence feature matrix to be detected can be a protein sequence feature matrix determined according to the HA protein sequence of the virus to be detected, and the antigen prediction results can be used to output the key protein sequence feature vector of the protein sequence feature matrix to be detected and the antigen distance vector between each virus. The key protein sequence feature vector can be used to reflect the key points related to antigenicity in the viral HA protein sequence, that is, the important amino acid positions in the viral HA protein sequence; the antigen distance can be used to reflect the antigenic differences between different viruses, and to identify the progressive changes (antigenic drift) and mutations (antigenic shift) of viral antigenicity, that is, the antigenic characteristics of viral HA protein, which are important for predicting the trend of viral antigen variation and selecting appropriate vaccine strains.

[0060] In the above-mentioned antigen prediction model processing method, a training sample set is obtained, wherein the training sample set includes a protein sequence feature matrix composed of multiple protein sequence feature vectors and an antigen distance label vector corresponding to the protein sequence feature matrix; for any protein sequence feature vector of the protein sequence feature vectors, a key protein sequence feature vector is determined from multiple protein sequence feature vectors by calculating the inner product of the antigen distance label vector and the normalized feature vector corresponding to any protein sequence feature vector; according to the key protein sequence feature vector, a preset antigen prediction model is determined, and the preset antigen prediction model is used to output the antigen prediction result of the protein sequence feature matrix to be detected, and the key protein sequence feature vector is determined by the inner product of the antigen distance label vector and the normalized feature vector corresponding to the protein sequence feature vector, so that even when processing high-dimensional data sets, the selected feature vector can be ensured to be the most representative, and an accurate antigen prediction model is constructed based on the key protein sequence feature vector, thereby improving the accuracy of the antigen prediction model.

[0061] In an exemplary embodiment, for any protein sequence feature vector of multiple protein sequence feature vectors, a key protein sequence feature vector is determined from the multiple protein sequence feature vectors according to the inner product of the antigen distance label vector and the normalized feature vector corresponding to any protein sequence feature vector, including: for any protein sequence feature vector of multiple protein sequence feature vectors, according to the modulus square of the inner product of the antigen distance label vector and the normalized feature vector corresponding to any protein sequence feature vector, the inner product modulus square value corresponding to any protein sequence feature vector is determined; according to the inner product modulus square values ​​corresponding to the multiple protein sequence feature vectors, the protein sequence feature vector corresponding to the largest inner product modulus square value is determined as the key protein sequence feature vector; the protein sequence feature vectors other than the key protein sequence feature vector in the multiple protein sequence feature vectors are used as new multiple protein sequence feature vectors, and the step of returning the inner product modulus square value corresponding to any protein sequence feature vector of the multiple protein sequence feature vectors according to the modulus square of the inner product of the antigen distance label vector and the normalized feature vector corresponding to any protein sequence feature vector is determined according to the modulus square of the inner product.

[0062] Assume that the order of the protein sequence feature matrix X is , the order of the antigen distance label vector Y is .

[0063] Next, the model relationship between the protein sequence feature matrix X and the antigen distance label vector Y can be established based on the maximum selection principle. The process of establishing the antigen prediction model is as follows:

[0064] The protein sequence feature matrix X is expressed as , according to the boundary zeroing property, we have the following formula (1):

[0065] ; (1)

[0066] Among them, Y is the antigen distance label vector, is any protein sequence feature vector in the protein sequence feature matrix, I=1, 2, …, k; is the norm of any protein sequence feature vector; is the normalized feature vector corresponding to any protein sequence feature vector. Any protein sequence feature vector is normalized, that is, divided by its norm to obtain a normalized feature vector; is the inner product of the normalized feature vector corresponding to the antigen distance label vector and any protein sequence feature vector; It is the square of the modulus of the inner product of the normalized feature vector corresponding to the antigen distance label vector and any protein sequence feature vector, that is, the square of the modulus of the inner product corresponding to any protein sequence feature vector.

[0067] According to the property of the boundary tending to zero, the maximum value of the inner product modulus square value is determined, that is, the protein sequence feature vector corresponding to I=1, 2, ..., k is found in X, so that the inner product modulus square value of the normalized feature vector corresponding to the protein sequence feature vector Y reaches the maximum. I1 can be used to represent the number of columns in X of the protein sequence feature vector corresponding to the maximum inner product modulus square value, that is, the number of columns of the first key protein sequence feature vector in X. Therefore, the I1 site is one of the key sites, that is, one of the key sites related to antigenicity in the viral HA protein sequence. is the protein sequence feature vector corresponding to the largest inner product modulus square value, that is, the first key protein sequence feature vector determined.

[0068] The first key protein sequence feature vector can be normalized to obtain the normalized feature vector corresponding to the first key protein sequence feature vector .make , x1 is the square value of the inner product modulus corresponding to the first key protein sequence feature vector.

[0069] Can Exclude from multiple protein sequence feature vectors, use the remaining other protein sequence feature vectors as new multiple protein sequence feature vectors, and determine the next key protein sequence feature vector from the new multiple protein sequence feature vectors, and so on, multiple key protein sequence feature vectors can be determined.

[0070] Similarly, the boundary tendency to zero property is used to determine the next key protein sequence feature vector, as shown in formula (2):

[0071] ; (2)

[0072] Among them, the I2 site is the second key site selected. is the second key protein sequence feature vector identified.

[0073] The second key protein sequence feature vector can be normalized to obtain the normalized feature vector corresponding to the second key protein sequence feature vector .make , x2 is the square value of the inner product modulus corresponding to the first key protein sequence feature vector.

[0074] By analogy, multiple key protein sequence feature vectors can be determined, as shown in formula (3):

[0075] ; (3)

[0076] Among them, I j The site is the second key site selected. The second key protein sequence feature vector is determined. The key point here is I j Refers to the key protein sequence feature vector The position in the protein sequence feature matrix X, is the normalized feature vector corresponding to the previous key protein sequence feature vector.

[0077] make , ,u j is the normalized feature vector corresponding to the jth key protein sequence feature vector. j is the square value of the inner product modulus corresponding to the j-th key protein sequence feature vector.

[0078] In an exemplary embodiment, before determining the inner product modulus square value corresponding to any protein sequence feature vector according to the modulus square of the inner product of the antigen distance tag vector and the normalized feature vector corresponding to any protein sequence feature vector, the method further includes: when the first key protein sequence feature vector is determined, dividing any protein sequence feature vector by its corresponding norm to obtain a normalized feature vector corresponding to any protein sequence feature vector;

[0079] When the Nth key protein sequence feature vector is determined, the projection vector of any protein sequence feature vector on the last determined key protein sequence feature vector is subtracted from any protein sequence feature vector to obtain a normalized feature vector corresponding to any protein sequence feature vector; N is an integer greater than 1; the projection vector is equal to the product of the inner product value of any protein sequence feature vector and the last determined key protein sequence feature vector and the last determined key protein sequence feature vector.

[0080] It can be seen that in formula (3), is any protein sequence feature vector; is the normalized feature vector corresponding to any protein sequence feature vector; for In the previously determined key protein sequence feature vector u j-1 The projection vector on ; To subtract the projection vector of any protein sequence feature vector on the last determined key protein sequence feature vector from any protein sequence feature vector, and then Divide by its corresponding norm to obtain the normalized feature vector corresponding to any protein sequence feature vector. Here, It is an orthogonal process, ensuring the key sites ,This pre-orthogonalization processing can eliminate the redundancy and noise in the data, allowing the ,antigen prediction model to focus more on extracting key ,features and relationships.

[0081] In an exemplary embodiment, after the protein sequence feature vector corresponding to the largest inner product modulus square value is determined as the key protein sequence feature vector, it also includes: judging whether the largest inner product modulus square value is less than or equal to the convergence threshold; if the largest inner product modulus square value is less than or equal to the convergence threshold, returning to the step of using other protein sequence feature vectors in the multiple protein sequence feature vectors except the key protein sequence feature vector as new multiple protein sequence feature vectors, until the largest inner product modulus square value is greater than the convergence threshold.

[0082] As shown in formula (3), I j where j is equal to 1, 2, ..., ,in, For the maximum selection principle termination symbol, you can choose , so that , is the convergence threshold, that is, each time a key protein sequence feature vector is determined, if the square value of the inner product modulus corresponding to the key protein sequence feature vector is equal to When , no key sites are selected, that is, no key protein sequence feature vectors are selected, and the above cycle process ends. In summary, the convergence level after adaptive maximum selection can be obtained .

[0083] By setting the convergence threshold, the antigen prediction model has great adaptability and can accurately screen out features that are highly correlated with the target variable (antigen distance label vector Y), thereby avoiding unnecessary feature redundancy and effectively preventing overfitting. The adaptability in the above process is reflected in: , by setting the minimum convergence threshold This ensures that key points are selected at each step to avoid overfitting and to better characterize the relationship between features and target variables.

[0084] In an exemplary embodiment, a preset antigen prediction model is determined based on a key protein sequence feature vector, including: determining model prediction parameters based on the inner product of the key protein sequence feature vector and the antigen distance label vector; and constructing a preset antigen prediction model based on the model prediction parameters.

[0085] The inner product can be used to reflect the correlation between the key protein sequence feature vector and the antigen distance label vector. In other words, it is the degree of influence of the key protein sequence feature vector on the antigen distance label vector. Therefore, the inner product of the key protein sequence feature vector and the antigen distance label vector can be used as a model prediction parameter.

[0086] In an exemplary embodiment, the model prediction parameters are determined based on the inner product of the key protein sequence feature vector and the antigen distance label vector, including: decomposing the key protein sequence feature matrix composed of multiple key protein sequence feature vectors into an orthogonal matrix and an upper triangular matrix; determining the model prediction parameters based on the key protein sequence feature matrix and the upper triangular matrix.

[0087] In the specific implementation, the key protein sequence matrix composed of multiple key protein sequence feature vectors is: , the key protein sequence matrix V is subjected to autoregressive decomposition (AR decomposition) to obtain an orthogonal matrix Q and an upper triangular matrix R. Among them, the orthogonal matrix Q is a square matrix whose column vectors are orthogonal to each other and have a norm of 1; the upper triangular matrix R is a matrix in which all elements except the elements on the main diagonal and above are 0.

[0088] Then the model prediction parameter w is as follows: , so that the preset antigen prediction model can be constructed based on the model prediction parameters.

[0089] After obtaining the model prediction parameter w, the preset antigen prediction model can be tested using a test sample set to obtain a test result of the preset antigen prediction model, which can be used to evaluate the performance of the preset antigen prediction model.

[0090] The test sample set is used to evaluate the performance of the antigen prediction model on unseen data and provide a measure of the predictive ability of the antigen prediction model in practical applications. The test results may include accuracy, recall, mean square error, etc. These test results can be used to evaluate the generalization ability, accuracy, robustness, overfitting or underfitting of the antigen prediction model.

[0091] Assume that the protein sequence feature matrix included in the test sample set is , then according to the model prediction parameter w and the protein sequence feature matrix included in the test sample set , determine the antigen distance prediction vector as , so the test sample set includes the antigen distance label vector Distance prediction vector The difference between (test set error), and obtain multiple test results of the preset antigen prediction model until Meet the threshold condition and average multiple test results.

[0092] The preset antigen prediction model is used to output the antigen prediction results of the protein sequence feature matrix to be detected. Assume that the protein sequence feature matrix to be predicted is ,pass A predicted antigenic distance Y can be obtainedt , and important amino acid positions (key positions) .

[0093] It can be seen that the antigen prediction model has strong adaptability. The convergence threshold in , which can accurately screen out features that are highly correlated with the target variable (Y), thereby avoiding unnecessary feature redundancy and effectively preventing overfitting. Secondly, the maximum selection principle in the form of inner product is used for feature selection to ensure that the most representative features are selected at each step, further improving the prediction accuracy and stability of the model. These advantages make the new model more potential and advantageous in the prediction of influenza virus antigen evolution, so that it can be better applied to the monitoring and prevention of influenza epidemics.

[0094] To facilitate understanding by those skilled in the art, FIG. 3 exemplarily provides a logic diagram of an antigen prediction model processing and FIG. 4 exemplarily provides another logic diagram of an antigen prediction model processing.

[0095] First, the HI titer value and HA sequence are preprocessed. The HA1 amino acid sequence in the HA sequence can be subjected to the PIMA function to obtain the protein sequence feature matrix X of the training sample set, and the antigenic distance between viruses is calculated according to the HI titer value to obtain the antigenic distance label vector Y. First, l=1 is used to screen the first key point (the key point here refers to the position of the key protein sequence feature vector in the protein sequence feature matrix X), using the maximum selection principle: After obtaining the key point I1, we then perform l=2 to screen the second key point (the second key protein sequence feature vector). ,here It is an orthogonalization process, which ensures The orthogonalized result can be obtained, that is, the normalized feature vector corresponding to the second key protein sequence feature vector , the second key point vector (the second key protein sequence feature vector) is , select in this way The key point is when the convergence threshold When , stop selecting, and select in sequence to get , , .use Represented by a column vector Formed Matrix, that is, a key protein sequence matrix composed of multiple key protein sequence feature vectors, and U represents a matrix composed of normalized feature vectors corresponding to multiple key protein sequence feature vectors , and get the transformation matrix , then the formula for the model prediction parameters of the preset antigen prediction model is: , thereby obtaining a preset antigen prediction model. Optionally, the antigen prediction model can be an adaptive maximum selection forest model, which is obtained by introducing an adaptive algorithm and the maximum selection principle and combining it with a random forest model for improvement. The above process is the calculation result of a tree. The algorithm is repeated 100 times to obtain the adaptive maximum selection forest result. Set < , and obtain multiple prediction results.

[0096] For example, the training set can be subjected to feature splitting using a random method. For example, assuming that the data set includes 331 protein sequence feature vectors, the feature splitting dimension K can be set to ,here Indicates that x is rounded down, which can be an integer not exceeding x. K features are randomly selected from the 331 protein sequence feature vectors as a randomly selected feature subset to form a new matrix as the protein sequence feature matrix X. Assuming that the antigen prediction model is a random forest model, the protein sequence feature matrix X composed of K features randomly selected each time can be used to build the first tree. This process is repeated many times to build the entire random forest. This process can reduce computational costs and reduce the risk of overfitting while ensuring a certain sample diversity. By randomly selecting a feature subset (K features) to build each tree, the complexity of the model is effectively reduced and the generalization ability of the model is improved, which is especially suitable for processing high-dimensional data sets.

[0097] In summary, this method can be used to predict antigen variation based on the adaptive maximum selection forest method in the sense of the Euclidean modulus. The antigen prediction model is a nonlinear adaptive regression model with the following advantages: strong adaptability, by setting , set the minimum The value makes each step select key points to avoid overfitting, adaptively screen out features that are highly correlated with the target variable, accurately calculate parameters to capture the relationship between variables, accurately control errors, and control the training set error within 1%; resist overfitting, use the maximum selection principle in the form of inner product to select features to ensure model stability; high generalization ability, by randomly selecting feature subsets (K features) to build each tree, effectively reducing model complexity, improving generalization ability, and improving prediction accuracy by integrating multiple trees, especially for high-dimensional data sets; good interpretability, based on the principle of orthogonal projection, easy to understand, suitable for various regression problems; high robustness, can identify and ignore outliers, insensitive to noise; low computational cost, intuitive and simple implementation process, compared with some complex machine learning algorithms, such as neural networks or support vector machines, the implementation process is relatively intuitive and simple. The adaptive maximum selection forest method model integrates multiple basic models and uses an average mechanism for integration, which significantly improves prediction accuracy and brings new breakthroughs in the field of antigen variation prediction.

[0098] To facilitate the understanding of those skilled in the art, FIG5 exemplarily provides a schematic diagram of the performance evaluation of an antigen prediction model. The preset antigen prediction model is tested by a test sample set to obtain the test result of the preset antigen prediction model, thereby evaluating the model performance of the antigen prediction model, and comparing and analyzing the preset antigen prediction model with four machine learning algorithms: Ridge regression, least absolute shrinkage and selection operator (LASSO), gradient boosting regression (GBR), and elastic network regression (ENR). The model performance is evaluated by correlation coefficient (CORR), root mean square error (RMSE), precision (Precision) and F1 score (F1-Score). CORR (correlation coefficient) quantifies the degree of linear correlation between the model prediction value and the actual observation value, which not only reveals the strength of the correlation between the predicted value and the true value, but also points out the directional relationship between them; RMSE (root mean square error) is used to evaluate the degree of difference between the predicted value and the actual value. The lower the RMSE, the closer the prediction result of the model is to the true value, indicating that the model has a stronger predictive ability. In the evaluation of specific tasks (such as identifying the similarity of two influenza virus antigens), the Precision metric is introduced. Precision focuses on the proportion of samples that are actually positive when the model identifies similar antigens, helping to understand the accuracy of the model in specific tasks. Finally, in order to more comprehensively evaluate the performance of the model in identifying antigens, the F1 value metric is used. F1-Score is the harmonic mean of Precision and Recall, which not only considers the accuracy of the model in identifying positive examples, but also the model's ability to find all positive examples.

[0099] As shown in Figure 5, compared with other machine learning models, the preset antigen prediction model has the lowest root mean square error RMSE value, while the precision and F1 scores are above 0.9, and the correlation coefficient is 0.69. This shows that the antigen prediction model has a better performance in predicting the antigenicity of influenza virus than other machine learning models.

[0100] FIG. 6 is another antigen prediction model processing method according to an exemplary embodiment. As shown in FIG. 6 , the method is applied to the terminal 102 in FIG. 1 as an example for explanation, and includes the following steps:

[0101] Step S602: Obtain a training sample set.

[0102] The training sample set includes a protein sequence feature matrix and an antigen distance label vector corresponding to the protein sequence feature matrix.

[0103] The protein sequence feature matrix contains multiple protein sequence feature vectors.

[0104] Step S604, when the Nth key protein sequence feature vector is determined, determine whether the positive integer N is greater than 1, if not, execute step S606; if yes, execute step S608.

[0105] Step S606, for any protein sequence feature vector of the multiple protein sequence feature vectors, divide any protein sequence feature vector by its corresponding norm to obtain a normalized feature vector corresponding to any protein sequence feature vector, and execute step S610.

[0106] Step S608, for any protein sequence feature vector of the multiple protein sequence feature vectors, subtract the projection vector of any protein sequence feature vector on the last determined key protein sequence feature vector from any protein sequence feature vector to obtain a normalized feature vector corresponding to any protein sequence feature vector, and execute step S610.

[0107] The projection vector is equal to the product of the inner product value of any protein sequence feature vector and the last determined key protein sequence feature vector and the last determined key protein sequence feature vector.

[0108] Step S610, determining the square modulus value of the inner product corresponding to any protein sequence feature vector according to the square modulus of the inner product of the antigen distance tag vector and the normalized feature vector corresponding to any protein sequence feature vector.

[0109] Step S612: According to the inner product modulus square values ​​corresponding to the plurality of protein sequence feature vectors, the protein sequence feature vector corresponding to the largest inner product modulus square value is determined as the key protein sequence feature vector.

[0110] Step S614, determining whether the maximum inner product modulus square value is less than or equal to the convergence threshold, if so, executing step S616; if not, executing step S618.

[0111] Step S616, taking the other protein sequence feature vectors in the multiple protein sequence feature vectors except the key protein sequence feature vector as the new multiple protein sequence feature vectors, and executing step S604.

[0112] Step S618, decomposing the key protein sequence feature matrix composed of multiple key protein sequence feature vectors into an orthogonal matrix and an upper triangular matrix.

[0113] Step S620, determining model prediction parameters according to the key protein sequence feature matrix and the upper triangular matrix.

[0114] Step S622, constructing a preset antigen prediction model according to the model prediction parameters.

[0115] The preset antigen prediction model is used to output the antigen prediction results of the protein sequence feature matrix to be detected.

[0116] It should be noted that the specific limitations of the above steps can be found in the specific limitations of an antigen prediction model processing method mentioned above, which will not be repeated here.

[0117] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0118] Based on the same inventive concept, the embodiment of the present application also provides an antigen prediction model processing device for implementing the antigen prediction model processing method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in the one or more antigen prediction model processing device embodiments provided below can refer to the limitations of the antigen prediction model processing method above, and will not be repeated here.

[0119] In an exemplary embodiment, as shown in FIG7 , an antigen prediction model processing device is provided, comprising:

[0120] The sample acquisition module 710 is used to acquire a training sample set; the training sample set includes a protein sequence feature matrix and an antigen distance label vector corresponding to the protein sequence feature matrix; the protein sequence feature matrix includes multiple protein sequence feature vectors.

[0121] The feature determination module 720 is used to determine a key protein sequence feature vector from any protein sequence feature vector of the multiple protein sequence feature vectors according to the inner product of the antigen distance label vector and the normalized feature vector corresponding to any protein sequence feature vector.

[0122] The model determination module 730 is used to determine a preset antigen prediction model according to the key protein sequence feature vector; the preset antigen prediction model is used to output the antigen prediction result of the protein sequence feature matrix to be detected.

[0123] In one embodiment, the feature determination module 720 is also used to determine the square modulus of the inner product corresponding to any protein sequence feature vector of the multiple protein sequence feature vectors according to the square modulus of the inner product of the antigen distance label vector and the normalized feature vector corresponding to the any protein sequence feature vector; determine the protein sequence feature vector corresponding to the largest square modulus of the inner product as the key protein sequence feature vector according to the square modulus of the inner product corresponding to the multiple protein sequence feature vectors; use the other protein sequence feature vectors in the multiple protein sequence feature vectors except the key protein sequence feature vector as new multiple protein sequence feature vectors, and return the step of determining the square modulus of the inner product corresponding to any protein sequence feature vector of the multiple protein sequence feature vectors according to the square modulus of the inner product of the antigen distance label vector and the normalized feature vector corresponding to the any protein sequence feature vector.

[0124] In one embodiment, the feature determination module 720 is also used to, when determining the first key protein sequence feature vector, divide any protein sequence feature vector by its corresponding norm to obtain a normalized feature vector corresponding to any protein sequence feature vector; when determining the Nth key protein sequence feature vector, subtract the projection vector of any protein sequence feature vector on the last determined key protein sequence feature vector from any protein sequence feature vector to obtain a normalized feature vector corresponding to any protein sequence feature vector; N is an integer greater than 1; and the projection vector is equal to the product of the inner product value of any protein sequence feature vector and the last determined key protein sequence feature vector and the last determined key protein sequence feature vector.

[0125] In one embodiment, the feature determination module 720 is also used to determine whether the maximum inner product modulus square value is less than or equal to the convergence threshold; if the maximum inner product modulus square value is less than or equal to the convergence threshold, then return to the step of using the other protein sequence feature vectors in the multiple protein sequence feature vectors except the key protein sequence feature vector as new multiple protein sequence feature vectors until the maximum inner product modulus square value is greater than the convergence threshold.

[0126] In one embodiment, the model determination module 730 is also used to determine the model prediction parameters based on the inner product of the key protein sequence feature vector and the antigen distance label vector; train the antigen prediction model to be trained based on the model prediction parameters and the training sample set to obtain the preset antigen prediction model.

[0127] In one embodiment, the model determination module 730 is further used to decompose the key protein sequence feature matrix composed of the plurality of key protein sequence feature vectors into an orthogonal matrix and an upper triangular matrix;

[0128] The model prediction parameters are determined according to the key protein sequence feature matrix and the upper triangular matrix.

[0129] Each module in the above-mentioned antigen prediction model processing device can be implemented in whole or in part by software, hardware and a combination thereof. Each of the above-mentioned modules can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0130] In an exemplary embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be shown in FIG8. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit, and an input device. The processor, the memory, and the input / output interface are connected via a system bus, and the communication interface, the display unit, and the input device are connected to the system bus via the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a mobile cellular network, a near field communication (NFC) or other technologies. When the computer program is executed by the processor, an antigen prediction model processing method is implemented. The display unit of the computer device is used to form a visually visible picture, which may be a display screen, a projection device, or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device shell, or an external keyboard, touchpad or mouse.

[0131] Those skilled in the art will understand that the structure shown in FIG. 8 is merely a block diagram of a partial structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different arrangement of components.

[0132] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the above-mentioned antigen prediction model processing method. The steps of the antigen prediction model processing method here can be the steps of the antigen prediction model processing method in each of the above-mentioned embodiments.

[0133] In one embodiment, a computer readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the processor executes the steps of the above-mentioned antigen prediction model processing method. The steps of the antigen prediction model processing method here can be the steps of the antigen prediction model processing method in each of the above-mentioned embodiments.

[0134] In one embodiment, a computer program product is provided, including a computer program, which, when executed by a processor, causes the processor to perform the steps of the above-mentioned antigen prediction model processing method. The steps of the antigen prediction model processing method here can be the steps of the antigen prediction model processing method in each of the above-mentioned embodiments.

[0135] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0136] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0137] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0138] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for processing an antigen prediction model, characterized in that: The method comprises: Acquire a training sample set; the training sample set includes a protein sequence feature matrix and an antigen distance label vector corresponding to the protein sequence feature matrix; the protein sequence feature matrix includes a plurality of protein sequence feature vectors; For any protein sequence feature vector of the multiple protein sequence feature vectors, determining a key protein sequence feature vector from the multiple protein sequence feature vectors according to the inner product of the antigen distance label vector and the normalized feature vector corresponding to the any protein sequence feature vector; According to the key protein sequence feature vector, a preset antigen prediction model is determined; the preset antigen prediction model is used to output the antigen prediction result of the protein sequence feature matrix to be detected.

2. The method according to claim 1, characterized in that The step of determining a key protein sequence feature vector from any one of the multiple protein sequence feature vectors according to the inner product of the antigen distance label vector and the normalized feature vector corresponding to any one of the multiple protein sequence feature vectors comprises: For any protein sequence feature vector of the multiple protein sequence feature vectors, determine the square modulus value of the inner product corresponding to any protein sequence feature vector according to the square modulus of the inner product of the antigen distance label vector and the normalized feature vector corresponding to the any protein sequence feature vector; According to the inner product modulus square values ​​corresponding to the multiple protein sequence feature vectors, determining the protein sequence feature vector corresponding to the largest inner product modulus square value as the key protein sequence feature vector; The step of taking the other protein sequence feature vectors among the multiple protein sequence feature vectors except the key protein sequence feature vector as a new multiple protein sequence feature vectors, returning any protein sequence feature vector for the multiple protein sequence feature vectors, and determining the modulus square value of the inner product corresponding to any protein sequence feature vector according to the modulus square of the inner product of the normalized feature vector corresponding to the antigen distance label vector and the any protein sequence feature vector.

3. The method according to claim 2, characterized in that Before determining the square modulus value of the inner product corresponding to any protein sequence feature vector according to the square modulus of the inner product of the antigen distance tag vector and the normalized feature vector corresponding to any protein sequence feature vector, the method further comprises: When the first key protein sequence feature vector is determined, dividing any protein sequence feature vector by its corresponding norm to obtain a normalized feature vector corresponding to any protein sequence feature vector; When the Nth key protein sequence feature vector is determined, the projection vector of any protein sequence feature vector on the last determined key protein sequence feature vector is subtracted from the any protein sequence feature vector to obtain a normalized feature vector corresponding to the any protein sequence feature vector; N is an integer greater than 1; and the projection vector is equal to the product of the inner product value of the any protein sequence feature vector and the last determined key protein sequence feature vector and the last determined key protein sequence feature vector.

4. The method according to claim 2, characterized in that: After determining the protein sequence feature vector corresponding to the largest inner product modulus square value as the key protein sequence feature vector, the method further comprises: Determine whether the maximum inner product modulus square value is less than or equal to a convergence threshold; If the maximum inner product modulus square value is less than or equal to the convergence threshold, return to the step of using the other protein sequence feature vectors among the multiple protein sequence feature vectors except the key protein sequence feature vector as the new multiple protein sequence feature vectors until the maximum inner product modulus square value is greater than the convergence threshold.

5. The method according to claim 1, characterized in that Determining a preset antigen prediction model according to the key protein sequence feature vector includes: Determining model prediction parameters according to the inner product of the key protein sequence feature vector and the antigen distance label vector; The preset antigen prediction model is constructed according to the model prediction parameters.

6. The method according to claim 1, characterized in that Determining the model prediction parameters according to the inner product of the key protein sequence feature vector and the antigen distance label vector includes: Decomposing a key protein sequence feature matrix composed of a plurality of key protein sequence feature vectors into an orthogonal matrix and an upper triangular matrix; The model prediction parameters are determined according to the key protein sequence feature matrix and the upper triangular matrix.

7. An antigen prediction model processing device, characterized in that: The device comprises: A sample acquisition module is used to acquire a training sample set; the training sample set includes a protein sequence feature matrix and an antigen distance label vector corresponding to the protein sequence feature matrix; the protein sequence feature matrix includes a plurality of protein sequence feature vectors; A feature determination module, for determining a key protein sequence feature vector from any of the multiple protein sequence feature vectors according to the inner product of the antigen distance label vector and the normalized feature vector corresponding to any of the multiple protein sequence feature vectors; The model determination module is used to determine a preset antigen prediction model according to the key protein sequence feature vector; the preset antigen prediction model is used to output the antigen prediction result of the protein sequence feature matrix to be detected.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images