Protein stability prediction method and device, computer equipment and storage medium

Abstract

The embodiment of the invention provides a protein stability prediction method and device, computer equipment and a storage medium. The method comprises the following steps: acquiring a first to-be-detected sequence corresponding to an original protein and a second to-be-detected sequence corresponding to a mutant protein; determining a plurality of first target residue features from the first to-be-detected sequence according to an attention mechanism of a protein language model, and extracting global features of the first target residue features to obtain first virtual structure microenvironment features; similarly, obtaining second virtual structure microenvironment features corresponding to a plurality of second target residue features; determining a first difference value based on the first virtual structure microenvironment feature and the second virtual structure microenvironment feature, and determining a second difference value based on a first global sequence feature corresponding to the first to-be-detected sequence and a second global sequence feature corresponding to the second to-be-detected sequence; and performing prediction according to the first difference value and the second difference value to obtain a protein stability prediction result. Therefore, the accuracy of predicting the stability of the protein can be improved.

Description

Technical Field

[0001] The present application relates to the field of bioinformatics, and in particular to a protein stability prediction method, device, computer equipment and storage medium. Background Art

[0002] Protein stability refers to the ability of a protein to maintain its three-dimensional structure under specific conditions. Protein stability can provide important insights into protein folding and function and is an important part of protein evolutionary fitness. Since the replacement of single or multiple amino acids in a protein sequence can affect the structure of the protein, it can also affect the stability and function of its overall structure. For example, certain amino acids such as proline and glycine may significantly change the conformation and stability of a protein after mutation due to their special chemical properties.

[0003] In the related art, sequence methods are generally used to predict the stability of proteins. However, the sequence method can only roughly capture the overall structure of the protein and analyze it, resulting in the problem of being unable to accurately predict the stability of the protein. Summary of the invention

[0004] The main purpose of the embodiments of the present application is to propose a protein stability prediction method, apparatus, computer equipment and storage medium, which can improve the accuracy of predicting protein stability.

[0005] To achieve the above objectives, the first aspect of the embodiments of the present application proposes a protein stability prediction method, the method comprising:

[0006] Obtaining a first sequence to be tested corresponding to the original protein and a second sequence to be tested of a mutant protein corresponding to the original protein;

[0007] Determine a plurality of first target residue features from the first sequence to be tested according to the attention mechanism of the pre-trained protein language model, and extract global features of the plurality of first target residue features to obtain a first virtual structure microenvironment feature;

[0008] Determining a plurality of second target residue features from the second sequence to be tested according to the attention mechanism of the protein language model, and extracting global features of the plurality of second target residue features to obtain a second virtual structure microenvironment feature;

[0009] Acquire a first global sequence feature of the first sequence to be tested and a second global sequence feature of the second sequence to be tested by using the protein language model;

[0010] Determining a first difference value based on the first virtual structural microenvironment feature and the second virtual structural microenvironment feature, and determining a second difference value based on the first global sequence feature and the second global sequence feature;

[0011] The protein stability of the mutant protein is predicted according to the first difference and the second difference to obtain a protein stability prediction result.

[0012] Accordingly, a second aspect of the embodiments of the present application proposes a protein stability prediction device, the device comprising:

[0013] A first acquisition module is used to acquire a first sequence to be tested corresponding to the original protein and a second sequence to be tested of a mutant protein corresponding to the original protein;

[0014] A first extraction module is used to determine a plurality of first target residue features from the first sequence to be tested according to the attention mechanism of the pre-trained protein language model, and extract global features of the plurality of first target residue features to obtain a first virtual structure microenvironment feature;

[0015] A second extraction module is used to determine a plurality of second target residue features from the second sequence to be tested according to the attention mechanism of the protein language model, and extract global features of the plurality of second target residue features to obtain a second virtual structure microenvironment feature;

[0016] A second acquisition module, configured to acquire a first global sequence feature of the first sequence to be tested and a second global sequence feature of the second sequence to be tested by using the protein language model;

[0017] A determination module, configured to determine a first difference based on the first virtual structure microenvironment feature and the second virtual structure microenvironment feature, and to determine a second difference based on the first global sequence feature and the second global sequence feature;

[0018] A prediction module is used to predict the protein stability of the mutant protein according to the first difference and the second difference to obtain a protein stability prediction result.

[0019] In some embodiments, the prediction module is further used to:

[0020] Inputting the first difference value corresponding to the first target residue feature and the second target residue feature into a pre-trained prediction model to predict the protein folding free energy in the residue feature change dimension, and obtaining a first prediction result;

[0021] Inputting the second difference value corresponding to the first sequence to be tested and the second sequence to be tested into the prediction model to predict the protein folding free energy of the global change dimension, and obtaining a second prediction result;

[0022] Based on the average of the first prediction result and the second prediction result, a protein stability prediction result is obtained.

[0023] In some embodiments, the first extraction module is further used to:

[0024] Generate a first overall attention map corresponding to the first sequence to be tested according to the attention mechanism of the pre-trained protein language model;

[0025] Performing global pooling processing on the first overall attention map to obtain a first target overall attention map;

[0026] Determining a plurality of first positions of a plurality of mutant residue features from the first test sequence;

[0027] Based on the matching of the multiple first positions in the first target overall attention map, obtaining multiple first attention score vectors corresponding to the multiple mutation residue features;

[0028] The number of selected first residue features is obtained, and based on the multiple first attention score vectors, the same number of first target residue features as the number of selected first residue features is determined from the multiple mutant residue features.

[0029] In some embodiments, the first extraction module is further used to:

[0030] Obtaining a preset first residue feature initial vector, and concatenating the first residue feature initial vector with the plurality of first target residue features to obtain a first target feature sequence; wherein the first residue feature initial vector is located at the first position of the first target feature sequence;

[0031] Based on the first residue feature initial vector, feature aggregation is performed on each first target residue feature corresponding to the first target feature sequence and the relationship between the multiple first target residue features to obtain a first virtual structural microenvironment feature for characterizing the global information of the first target feature sequence.

[0032] In some embodiments, the second extraction module is further used to:

[0033] Generate a second overall attention graph corresponding to the second sequence to be tested according to the attention mechanism of the protein language model;

[0034] Performing global pooling processing on the second overall attention map to obtain a second target overall attention map;

[0035] Determining a plurality of second positions of a plurality of mutant residue features from the second test sequence;

[0036] Based on the matching of the multiple second positions in the second target overall attention map, obtain multiple second attention score vectors corresponding to the multiple mutation residue features;

[0037] A second residue feature selection quantity is obtained, and based on the multiple second attention score vectors, a number of second target residue features equal to the second residue feature selection quantity is determined from the multiple mutation residue features.

[0038] In some embodiments, the second extraction module is further used to:

[0039] Obtaining a preset second residue feature initial vector, and concatenating the second residue feature initial vector with the plurality of second target residue features to obtain a second target feature sequence; wherein the second residue feature initial vector is located at the first position of the second target feature sequence;

[0040] Based on the second residue feature initial vector, feature aggregation is performed on each second target residue feature corresponding to the second target feature sequence and the relationship between the multiple second target residue features to obtain a second virtual structural microenvironment feature for characterizing the global information of the second target feature sequence.

[0041] In some implementations, the second acquisition module is further configured to:

[0042] Obtaining a preset first global initial vector, and performing feature aggregation on a plurality of first features in the first sequence to be tested based on the first global initial vector by using the protein language model to obtain a first global sequence feature;

[0043] A preset second global initial vector is obtained, and a plurality of second features in the second sequence to be tested are aggregated based on the second global initial vector by using the protein language model to obtain a second global sequence feature.

[0044] Accordingly, the third aspect of the embodiments of the present application proposes a computer device, which includes a memory and a processor, the memory stores a computer program, and the processor implements the protein stability prediction method described in any one of the embodiments of the first aspect of the present application when executing the computer program.

[0045] Correspondingly, the fourth aspect of the embodiments of the present application proposes a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the protein stability prediction method described in any one of the embodiments of the first aspect of the present application.

[0046] The present application obtains a first test sequence of an original protein and a second test sequence of a mutant protein corresponding to the original protein; determines a plurality of first target residue features from the first test sequence according to the attention mechanism of a pre-trained protein language model, and extracts global features of the plurality of first target residue features to obtain a first virtual structural microenvironment feature; determines a plurality of second target residue features from the second test sequence according to the attention mechanism of the protein language model, and extracts global features of the plurality of second target residue features to obtain a second virtual structural microenvironment feature; obtains a first global sequence feature of the first test sequence and a second global sequence feature of the second test sequence through a protein language model; determines a first difference based on the first virtual structural microenvironment feature and the second virtual structural microenvironment feature, and determines a second difference based on the first global sequence feature and the second global sequence feature; predicts protein stability of the mutant protein according to the first difference and the second difference to obtain a protein stability prediction result. In this way, the stability of the protein can be predicted by combining the local environmental characteristics around the mutated residues and the global characteristics of the entire protein sequence, so that the protein language model can understand the impact of the mutation on the local structure of the protein and the impact of the mutation on the global structure, thereby improving the accuracy of predicting the stability of the protein. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 Schematic diagram of the structure of the protein stability prediction system provided in the embodiment of the present application;

[0048] Figure 2 is a flow chart of the protein stability prediction method provided in the embodiments of the present application;

[0049] Figure 3 is an overall flow chart of the protein stability prediction method provided in the embodiments of the present application;

[0050] Figure 4 is a diagram of the construction process of the virtual structure microenvironment provided in the embodiment of the present application;

[0051] Figure 5 It is a schematic diagram of the functional modules of the protein stability prediction device provided in the embodiment of the present application;

[0052] Figure 6 It is a schematic diagram of the hardware structure of the computer device provided in the embodiment of the present application. DETAILED DESCRIPTION

[0053] 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.

[0054] It should be noted that, although the functional modules are divided in the device schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first", "second", etc. in the specification, claims and the above drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0055] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0056] Protein stability refers to the ability of a protein to maintain its three-dimensional structure under specific conditions. Protein stability can provide important insights into protein folding and function and is an important part of protein evolutionary fitness. Since the replacement of single or multiple amino acids in a protein sequence can affect the structure of the protein, it can also affect the stability and function of its overall structure. For example, certain amino acids such as proline and glycine may significantly change the conformation and stability of a protein after mutation due to their special chemical properties.

[0057] In the related art, the sequence method is generally used to predict the stability of proteins. However, the sequence method can only roughly capture the overall structure of the protein and analyze it, resulting in the problem of not being able to accurately predict the stability of the protein. Alternatively, the structure-dependent method is used to predict the stability of proteins. However, the structure-dependent method requires high-resolution laboratory analysis of the structure. These data are severely scarce or even missing in some protein families. When the data is difficult to obtain comprehensively, it will also affect the accurate prediction of protein stability.

[0058] Based on this, the embodiments of the present application provide a protein stability prediction method, apparatus, computer equipment and storage medium, which can improve the accuracy of predicting protein stability.

[0059] The protein stability prediction method, apparatus, computer device and storage medium provided in the embodiments of the present application are specifically described through the following embodiments. First, the protein stability prediction system in the embodiments of the present application is described.

[0060] Please refer to Figure 1In some embodiments, the present application provides a protein stability prediction system, including a terminal 11 and a server 12.

[0061] Exemplarily, the terminal 11 may be a device with which the user directly interacts, and the user may access and operate the protein stability prediction system through the terminal 11. For example, the terminal 11 may be a client of a personal computer, a smart phone or a tablet computer, a workstation or a high-performance computing cluster, and the like.

[0062] Furthermore, the server 12 may be a backend server or cloud service facility for providing protein stability prediction services, responsible for processing requests from the terminal 11, executing the prediction algorithm, and returning the prediction results. For example, the server 12 may be a Web server, an application server, a database server, a high-performance computing cluster, a cloud computing platform, and the like.

[0063] The server 12 can efficiently process requests from different terminals 11 and provide accurate and reliable protein stability prediction services.

[0064] The protein stability prediction method in the embodiments of the present application can be illustrated by the following examples.

[0065] It should be noted that in each specific implementation of the present application, when it comes to the need to perform relevant processing based on data related to user identity or characteristics such as user information, user behavior data, user historical data, and user location information, the user's permission or consent will be obtained first. Moreover, the collection, use, and processing of these data will comply with relevant laws, regulations, and standards. In addition, when the embodiment of the present application needs to obtain the user's sensitive personal information, the user's separate permission or separate consent will be obtained through a pop-up window or by jumping to a confirmation page. After clearly obtaining the user's separate permission or separate consent, the necessary user-related data for enabling the normal operation of the embodiment of the present application will be obtained.

[0066] In the embodiments of the present application, the description will be made from the perspective of a protein stability prediction device, which can be integrated into a computer device. Figure 2 , Figure 2 This is a flowchart of the steps of the protein stability prediction method provided in the embodiment of the present application. In the embodiment of the present application, the protein stability prediction device is specifically integrated on a terminal or a server as an example. When the processor on the terminal or the server executes the program instructions corresponding to the protein stability prediction method, the specific process is as follows:

[0067] Step 101, obtaining a first sequence to be tested corresponding to the original protein and a second sequence to be tested corresponding to the mutant protein of the original protein.

[0068] In some embodiments, in order to accurately predict the effect of mutation on protein stability, a first test sequence corresponding to the original protein and a second test sequence of the mutant protein corresponding to the original protein can be obtained for comparative analysis.

[0069] Among them, the original protein can be the normal form of the protein in nature, which is usually used as a reference benchmark to compare the properties of the mutant protein.

[0070] The first sequence to be tested may be the amino acid sequence of the original protein, which is a reference sequence for predicting protein stability.

[0071] The mutant protein may be a protein in which one or more mutations occur in the original protein, and the mutation may be a replacement, insertion or deletion of one or more amino acids.

[0072] The second sequence to be tested may be the amino acid sequence of a mutant protein, which is a compiled sequence used for predicting protein stability.

[0073] For example, suppose there is a natural original protein that has not been subjected to any mutation treatment, and it is known that a certain position of the original protein has mutated to form a mutant protein. At this time, the amino acid sequence of the original protein can be obtained as the first sequence to be tested, assuming that the first sequence to be tested is MAGF (only as an example, the first sequence to be tested may be more complex in actual situations). And the amino acid sequence corresponding to the mutant protein obtained after the mutation of the original protein is obtained as the second sequence to be tested, assuming that the amino acid sequence of the second sequence to be tested is MAGV.

[0074] By obtaining the sequences of the original protein and the mutant protein, it is convenient to conduct subsequent accurate comparative analysis and prediction of protein stability.

[0075] Step 102, determining a plurality of first target residue features from the first sequence to be tested according to the attention mechanism of the pre-trained protein language model, and extracting global features of the plurality of first target residue features to obtain a first virtual structure microenvironment feature.

[0076] In some embodiments, in order to capture the local environmental information around the mutant residue, multiple first target residue features can be determined through an attention mechanism, and the first virtual structural microenvironment feature can be obtained by integrating the multiple first target residue features and the surrounding local environmental information to improve the accuracy of protein stability prediction.

[0077] The protein language model can be a deep learning model obtained through pre-training. The initial protein language model can be obtained by learning a large number of protein sequences in advance. The trained protein language model can capture the association rules between amino acids. Furthermore, the protein language model is usually based on the Transformer architecture, which can process protein sequence data and extract useful features from it.

[0078] Among them, the attention mechanism can be a mechanism in the protein language model that determines the interaction strength between different positions in the first test sequence by learning different representations of the first test sequence, and assigns different attention scores to different positions based on the importance of different positions.

[0079] The first target residue feature may be a position-related residue feature of the mutant protein before mutation determined from the first sequence to be tested through an attention mechanism.

[0080] The first virtual structure microenvironment feature may be a comprehensive feature representation obtained by performing global feature extraction on the first target residue feature, which includes local environment information around the mutant residue before mutation.

[0081] Exemplarily, the attention mechanism of the pre-trained protein language model can be used to output the overall attention map corresponding to the first sequence to be tested, and the overall attention map can be pooled to obtain the pooled overall attention map. Further, the position of the mutated residue feature in the first sequence to be tested can be determined, and the position can be searched in the pooled overall attention map to determine the attention score vector of the position.

[0082] Furthermore, after obtaining the attention score vector of the position where each residue feature is located, the residue feature with the largest attention score can be determined according to the number of first target residue features to be selected. For example, if 10 first target residue features need to be selected, the 10 residue features with the highest attention scores are selected as the first target residue features according to the attention scores of multiple residue features. In this way, it is ensured that the residue information with the greatest impact on the mutation position is captured, so that the protein language model can focus more on important information, while removing unnecessary noise and avoiding being affected by secondary information.

[0083] Furthermore, global features can be extracted from multiple first target residue features to obtain corresponding virtual structural microenvironment features, so as to reflect the impact of mutations on local structures by generating a comprehensive feature representation.

[0084] By determining multiple first target residue features through the attention mechanism and performing global feature extraction on the first target residue features, the first virtual structure microenvironment features are obtained, which can capture and screen important feature information and improve the accuracy of protein language model prediction.

[0085] In some embodiments, in order to improve the accuracy of protein stability prediction, multiple first target residue features can be determined from the first sequence to be tested according to the attention mechanism of the pre-trained protein language model to capture the residue information that has the greatest impact on the mutation position, and at the same time, the computational efficiency of the protein language model can be improved through screening. For example, "determining multiple first target residue features from the first sequence to be tested according to the attention mechanism of the protein language model" in step 102 may include:

[0086] (102.a1) generating a first overall attention map corresponding to the first sequence to be tested according to the attention mechanism of the pre-trained protein language model;

[0087] (102.a2) performing global pooling processing on the first overall attention map to obtain a first target overall attention map;

[0088] (102.a3) determining a plurality of first positions of a plurality of mutant residue features from the first test sequence;

[0089] (102.a4) matching the first target overall attention map based on the multiple first positions to obtain multiple first attention score vectors corresponding to the multiple mutation residue features;

[0090] (102.a5) Obtain the number of first residue feature selections, and determine the same number of first target residue features as the number of first residue feature selections from multiple mutant residue features based on multiple first attention score vectors.

[0091] Among them, the first overall attention map can be used to reflect the interaction strength between different positions in the first sequence to be tested, and the first overall attention map can be a two-dimensional matrix, each element of which represents the attention score corresponding to the feature of the first sequence to be tested.

[0092] The first target overall attention map may be an attention map obtained by performing global pooling on the first overall attention map. The pooling method may be maximum pooling, average pooling, adaptive pooling, global pooling, and the like.

[0093] The mutated residue feature may be the position of the feature where the mutation occurs in the first sequence to be tested. The mutated residue feature may be determined based on the comparison between the first sequence to be tested and the second sequence to be tested, or may be calibrated by the user or automatically identified by an algorithm.

[0094] The first position may be a specific position where the mutant residue feature is located in the first sequence to be tested.

[0095] Among them, the first attention score vector may refer to the attention score vector of the position corresponding to the mutation position in the first target overall attention map.

[0096] Among them, the number of first residue features selected can be the number of mutant residue features that need to be selected from multiple mutant residue features. For example, the number of first residue features selected can be 10, 20, etc. The specific number of selections can be flexibly determined according to actual conditions.

[0097] For example, assume that the first sequence to be tested is MAGF (for example only, in fact, the sequence structure may have multiple forms), and it is known that the fourth amino acid position has mutated (F becomes V). The first sequence to be tested is input into the pre-trained protein language model to obtain the first sequence feature Z corresponding to the first sequence to be tested. 1 seq The protein language model can generate the first overall attention map corresponding to the first sequence to be tested based on the features of the first sequence to be tested through the last Transformer layer. Here, H refers to the number of attention heads in the Transformer layer, and L1 refers to the sequence length of the first sequence to be tested.

[0098] Furthermore, the first overall attention map can be globally pooled to obtain the first target overall attention map, so as to retain the most critical information while reducing the data dimension. The formula is as follows:

[0099]

[0100] in, Represents the first overall attention map A last1 The first sub-attention map of the i-th attention head in , H refers to the number of attention heads in the Transformer layer, Represents the overall attention map of the first target.

[0101] Furthermore, the first position pos of each mutated residue feature can be determined by automatic system calibration (or by comparing the sequence before and after the protein mutation, etc.) mut1 , and the first position is the first target overall attention map after pooling to obtain the first attention score vector associated with the mutated residue features. The formula is as follows:

[0102]

[0103] Among them, Amut1 Represents the first attention score vector.

[0104] Furthermore, the number of first residue feature selections is set to k (the value of k can be set according to actual conditions), and then the features of several residues with the highest first attention scores need to be selected from the mutant residue features:

[0105] (pos 1 ,pos 2 ,…,pos k1 )=argmax k1 (A mut1 )

[0106]

[0107] where argmax k1 (·) represents the function of retrieving the mutation residue features with the largest attention scores according to the first attention score vector as the first target residue feature, Represents the top K residues with the largest first attention scores for the mutated residue feature.

[0108] Through the above methods, the local environmental information around the mutated residues is effectively captured, which improves the accuracy and robustness of protein stability prediction. At the same time, redundant information is reduced through feature selection, which improves the computational efficiency of the protein language model.

[0109] In some embodiments, in order to capture the local environment information around the mutated residue to determine the interaction strength between the mutated residue and other amino acids, a global feature extraction may be performed on multiple first target residue features to obtain a first virtual structural microenvironment feature to more accurately assess the effect of the mutation on protein stability. For example, "extracting a global feature of multiple first target residue features to obtain a first virtual structural microenvironment feature" in step 102 may include:

[0110] (102.b1) obtaining a preset first residue feature initial vector, and concatenating the first residue feature initial vector with a plurality of first target residue features to obtain a first target feature sequence; wherein the first residue feature initial vector is located at the first position of the first target feature sequence;

[0111] (102.b2) Based on the first residue feature initial vector, feature aggregation is performed on each first target residue feature corresponding to the first target feature sequence, as well as the relationship between multiple first target residue features, to obtain a first virtual structural microenvironment feature for characterizing the global information of the first target feature sequence.

[0112] The first residue feature initial vector may be a vector obtained through a pre-trained protein language model and used to capture information of the entire sequence.

[0113] The first target feature sequence may be a feature sequence obtained by concatenating the first residue feature initial vector with a plurality of first target residue features.

[0114] Exemplarily, the first residue feature initial vector Z CLS1 can be a learnable vector, After obtaining multiple first target residue features Afterwards, you can Send it to a transformer layer to aggregate these K first target residue features into a first virtual structure microenvironment feature, the formula is as follows:

[0115]

[0116] in, represents the final first virtual structure microenvironment feature, ExtractCLS represents the operation of extracting CLS tokens (global vectors) in the protein language model, represents a transformer layer with trainable parameters θ, Represents the concatenation of the first residue feature initial vector and multiple first target residue features.

[0117] Through the first residue feature initial vector, the protein language model can pass on additional information when processing sequence data and better understand the interaction between the mutant residue features and their adjacent mutant residue features, so that the first virtual structural microenvironment feature can achieve the optimal feature integration effect, thereby more accurately evaluating the impact of mutations on protein stability.

[0118] Step 103, determining a plurality of second target residue features from the second sequence to be tested according to the attention mechanism of the protein language model, and extracting global features of the plurality of second target residue features to obtain a second virtual structure microenvironment feature.

[0119] In some embodiments, in order to capture the local environmental information around the mutated residues after the protein mutates, a plurality of second target residue features can be determined through an attention mechanism, and a second virtual structural microenvironment feature can be obtained by integrating the plurality of second target residue features and the surrounding local environmental information to improve the accuracy of protein stability prediction.

[0120] The protein language model can be a deep learning model obtained through pre-training. The initial protein language model can be obtained by learning a large number of protein sequences in advance. The trained protein language model can capture the association rules between amino acids. Furthermore, the protein language model is usually based on the Transformer architecture, which can process protein sequence data and extract useful features from it.

[0121] Among them, the attention mechanism can be a mechanism in the protein language model that determines the interaction strength between different positions in the second test sequence by learning different representations of the second test sequence, and assigns different attention scores to different positions based on the importance of different positions.

[0122] The second target residue feature may be a residue feature related to the mutation position of the mutant protein after mutation determined from the second test sequence through the attention mechanism.

[0123] The second virtual structure microenvironment feature may be a comprehensive feature representation obtained by performing global feature extraction on the second target residue feature, which includes local environment information around the mutant residue.

[0124] Exemplarily, the attention mechanism of the pre-trained protein language model can be used to output the overall attention map corresponding to the second sequence to be tested, and the overall attention map can be pooled to obtain the pooled overall attention map. Further, the position of the mutated residue feature in the second sequence to be tested can be determined, and the position can be searched in the pooled overall attention map to determine the attention score vector of the position.

[0125] Furthermore, after obtaining the attention score vector of the position where each residue feature is located, the residue feature with the largest attention score can be determined according to the number of second target residue features to be selected. For example, if 10 second target residue features need to be selected, the 10 residue features with the highest attention scores are selected as the second target residue features according to the attention scores of multiple residue features. In this way, it is ensured that the residue information with the greatest impact on the mutation position is captured, so that the protein language model can focus more on important information, while removing unnecessary noise and avoiding being affected by secondary information.

[0126] Furthermore, global features can be extracted for multiple second target residue features to obtain corresponding virtual structural microenvironment features, so as to reflect the impact of mutations on local structures by generating a comprehensive feature representation.

[0127] By determining multiple second target residue features through the attention mechanism and performing global feature extraction on the second target residue features, the second virtual structure microenvironment features are obtained, which can capture and screen important feature information and improve the accuracy of protein language model prediction.

[0128] In some embodiments, in order to improve the accuracy of protein stability prediction, multiple second target residue features can be determined from the second sequence to be tested according to the attention mechanism of the pre-trained protein language model to capture the residue information that has the greatest impact on the mutation position, and at the same time, the computational efficiency of the protein language model can be improved through screening. For example, "determining multiple second target residue features from the second sequence to be tested according to the attention mechanism of the protein language model" in step 103 may include:

[0129] (103.a1) generating a second overall attention map corresponding to the second sequence to be tested according to the attention mechanism of the protein language model;

[0130] (103.a2) performing global pooling processing on the second overall attention map to obtain a second target overall attention map;

[0131] (103.a3) determining a plurality of second positions of a plurality of mutant residue features from a second test sequence;

[0132] (103.a4) matching the second target overall attention map based on the multiple second positions to obtain multiple second attention score vectors corresponding to the multiple mutation residue features;

[0133] (103.a5) Obtain the number of second residue feature selections, and determine the same number of second target residue features as the number of second residue feature selections from multiple mutant residue features based on multiple second attention score vectors.

[0134] The second overall attention map can be used to reflect the interaction strength between different positions in the second sequence to be tested, and the second overall attention map can be a two-dimensional matrix, each element of which represents the attention score corresponding to the feature of the second sequence to be tested.

[0135] The second target overall attention map may be an attention map obtained by performing global pooling on the second overall attention map. The pooling method may be maximum pooling, average pooling, adaptive pooling, global pooling, or the like.

[0136] The mutated residue feature may be the position of the feature where the mutation occurs in the second sequence to be tested. The mutated residue feature may be determined based on the comparison between the second sequence to be tested and the second sequence to be tested, or may be calibrated by the user or automatically identified by an algorithm.

[0137] The second position may be the specific position where the mutant residue feature is located in the second sequence to be tested.

[0138] Among them, the second attention score vector may refer to the attention score vector of the position corresponding to the mutation position in the second target overall attention map.

[0139] Among them, the number of second residue features selected can be the number of mutant residue features that need to be selected from multiple mutant residue features. For example, the number of second residue features selected can be 10, 20, etc. The specific number of selections can be flexibly determined according to actual conditions.

[0140] For example, assume that the second sequence to be tested is MAGF (for example only, in fact, the sequence structure may have multiple forms), and it is known that the fourth amino acid position has mutated (F becomes V). The second sequence to be tested is input into the pre-trained protein language model to obtain the second sequence feature Z corresponding to the second sequence to be tested. 2 seq The protein language model can generate a second overall attention map corresponding to the second sequence to be tested based on the features of the second sequence to be tested through the last Transformer layer. Here, H refers to the number of attention heads in the Transformer layer, and L2 refers to the sequence length of the second sequence to be tested.

[0141] Furthermore, the second overall attention map can be globally pooled to obtain the second target overall attention map, so as to reduce the data dimension while retaining the most critical information. The formula is as follows:

[0142]

[0143] in, Represents the second overall attention map A 1ast2 The second sub-attention graph of the jth attention head in , H refers to the number of attention heads in the Transformer layer, Represents the overall attention map of the second target.

[0144] Furthermore, the second position pos of each mutated residue feature can be determined by automatic system calibration (or by comparing the sequence before and after the protein mutation, etc.) mut2 , and the second target overall attention map after pooling at the second position to obtain the second attention score vector associated with the mutated residue features. The formula is as follows:

[0145]

[0146] Among them, Amut2 Represents the second attention score vector.

[0147] Furthermore, the number of selected second residue features is set to k (the value of k can be set according to actual conditions), and then the features of several residues with the highest second attention scores need to be selected from the mutant residue features:

[0148] (pos 1 ,pos 2 ,…,pos k2 )=argmax k2 (A mut2 )

[0149]

[0150] where argmax k2 (·) represents the function of retrieving the mutation residue features with the largest attention scores according to the second attention score vector as the second target residue feature, Represents the top K residues with the largest second attention scores for the mutated residue feature.

[0151] Through the above methods, the local environmental information around the mutated residues is effectively captured, which improves the accuracy and robustness of protein stability prediction. At the same time, redundant information is reduced through feature selection, which improves the computational efficiency of the protein language model.

[0152] In some embodiments, in order to capture the local environment information around the mutated residue to determine the interaction strength between the mutated residue and other amino acids, the global features of multiple second target residue features can be extracted to obtain the second virtual structural microenvironment features to more accurately evaluate the impact of the mutation on protein stability. For example, "extracting the global features of multiple second target residue features to obtain the second virtual structural microenvironment features" in step 103 may include:

[0153] (103.b1) obtaining a preset second residue feature initial vector, and concatenating the second residue feature initial vector with a plurality of second target residue features to obtain a second target feature sequence; wherein the second residue feature initial vector is located at the first position of the second target feature sequence;

[0154] (103.b2) Based on the second residue feature initial vector, feature aggregation is performed on each second target residue feature corresponding to the second target feature sequence, as well as the relationship between multiple second target residue features, to obtain a second virtual structural microenvironment feature for characterizing the global information of the second target feature sequence.

[0155] The second residue feature initial vector may be a vector obtained through a pre-trained protein language model and used to capture information of the entire sequence.

[0156] The second target feature sequence may be a feature sequence obtained by concatenating the second residue feature initial vector with a plurality of second target residue features.

[0157] Exemplarily, the second residue feature initial vector Z CLS2 can be a learnable vector, After obtaining multiple second target residue features Afterwards, you can Send it to a transformer layer to aggregate these K second target residue features into a second virtual structure microenvironment feature, the formula is as follows:

[0158]

[0159] in, represents the final second virtual structure microenvironment feature, ExtractCLS represents the operation of extracting CLS tokens (global vectors) in the protein language model, represents a transformer layer with trainable parameters θ, Represents the concatenation of the second residue feature initial vector and multiple second target residue features.

[0160] Through the second residue feature initial vector, the protein language model can pass on additional information when processing sequence data and better understand the interaction between the mutant residue features and their adjacent mutant residue features, so that the second virtual structural microenvironment feature can achieve the optimal feature integration effect, thereby more accurately evaluating the impact of mutations on protein stability.

[0161] Step 104: obtaining a first global sequence feature of the first sequence to be tested and a second global sequence feature of the second sequence to be tested by using a protein language model.

[0162] In some embodiments, in order to better understand the overall structure and function of the protein, the information of the mutated residue and its surrounding environment is integrated to obtain a first global sequence feature of the first test sequence and a second global sequence feature of the second test sequence, so as to compare and analyze the effect of the mutation on the protein stability at the global level.

[0163] The first global sequence feature may be a global feature extracted by the protein language model from the first sequence to be tested. The first global sequence feature may be used to understand the overall structure and function of the original protein.

[0164] The second global sequence feature may be a global feature extracted by the protein language model from the second sequence to be tested. The second global sequence feature may be used to understand the overall structure and function of the mutant protein.

[0165] Exemplarily, the first global initial vector can be concatenated with the first sequence to be tested, and feature aggregation is performed on each first sequence to be tested feature in the first sequence to be tested and the relationship between the first sequence to be tested features according to the first global initial vector to obtain the first global sequence feature for characterizing the global information of the first sequence to be tested.

[0166] Similarly, the second global initial vector can be concatenated with the second sequence to be tested, and feature aggregation can be performed on each second sequence to be tested feature in the second sequence to be tested and the relationship between the second sequence to be tested features according to the second global initial vector to obtain the second global sequence feature for characterizing the global information of the second sequence to be tested.

[0167] By obtaining the first global sequence features and the second global sequence features, a global perspective can be provided to better understand the overall structure and function of the original protein and the mutant protein, which helps to integrate the information of the mutation site and its surrounding environment, and compare and analyze the effects of mutations on protein stability at the global level, thereby improving the accuracy and robustness of the prediction.

[0168] In some embodiments, in order to comprehensively evaluate the effect of mutation on protein stability, the first global sequence feature corresponding to the original protein can be obtained to understand the state of the protein before the mutation, and the second global sequence feature corresponding to the mutant protein can be obtained to understand the state of the protein after the mutation, so as to accurately predict the stability of the protein. For example, step 104 may include:

[0169] (104.1) obtaining a preset first global initial vector, and performing feature aggregation on a plurality of first features in a first sequence to be tested based on the first global initial vector by using a protein language model to obtain a first global sequence feature;

[0170] (104.2) Obtain a preset second global initial vector, and perform feature aggregation on multiple second features in the second sequence to be tested based on the second global initial vector using a protein language model to obtain a second global sequence feature.

[0171] The first global initial vector may be the first vector used to initialize the feature sequence before feature aggregation. The first global initial vector is used to help the protein language model better understand the overall structure and function of the original protein.

[0172] The second global initial vector may be the first vector used to initialize the feature sequence before feature aggregation, and the second global initial vector is used to help the protein language model better understand the overall structure and function of the mutant protein.

[0173] Exemplarily, a preset first global initial vector Z may be obtained first. q1 CLS , and the first global initial vector and the first sequence to be tested (Z 11 ,…,Z 1END ) to obtain Z seq1 =(Z q1 CLS ,Z 11 ,…,Z 1END ). Further, based on the first global learning vector, the first features in the first sequence to be tested can be aggregated through the Transformer layer in the protein language model or other feature aggregation methods to obtain the first global sequence feature Z Q1 CLS , to effectively capture the overall structural and functional information of the original protein, so as to better understand the state of the protein before mutation.

[0174] In some embodiments, a pre-trained timing model, such as a long short-term memory network, can be obtained, and the first test sequence can be input into the timing model to obtain a first test sequence with enhanced timing information. Thereafter, the first test sequence with enhanced timing information is further processed as described above to generate a corresponding first global sequence feature. This can better capture the temporal evolution of protein structure and function after mutation, further enhance the model's understanding of the time dimension, and provide more refined prediction results.

[0175] Exemplarily, the preset second global initial vector Z can be obtained first q2 CLS , and the second global initial vector and the second sequence to be tested (Z 21 ,…,Z 2END ) to obtain Z seq2 =(Z q2 CLS ,Z 21 ,…,Z 2END ). Furthermore, based on the second global learning vector, the second features in the second sequence to be tested can be aggregated through the Transformer layer in the protein language model or other feature aggregation methods to obtain the second global sequence feature Z Q2 CLS . To effectively capture the overall structural and functional information of mutant proteins, so as to better understand the state of proteins after mutation.

[0176] In some embodiments, a pre-trained timing model, such as a long short-term memory network, can be obtained, and the second test sequence can be input into the timing model to obtain a second test sequence with enhanced timing information. Thereafter, the second test sequence with enhanced timing information is further processed as described above to generate a corresponding second global sequence feature. This can better capture the temporal evolution of protein structure and function after mutation, further enhance the model's understanding of the time dimension, and provide more refined prediction results.

[0177] Through the above methods, the overall structural and functional information of the original protein and the mutant protein can be effectively captured, so as to better understand the state of the protein before and after mutation, so as to improve the accuracy and robustness of subsequent predictions of protein stability.

[0178] Step 105: determining a first difference based on the first virtual structure microenvironment feature and the second virtual structure microenvironment feature, and determining a second difference based on the first global sequence feature and the second global sequence feature.

[0179] In some embodiments, in order to effectively capture the impact of mutations at local and global scales, a first difference can be determined based on the first virtual structural microenvironment feature and the second virtual structural microenvironment feature, and a second difference can be determined based on the first global sequence feature and the second global sequence feature to assess whether the mutation destroys important local domains and protein functions, thereby improving the accuracy and robustness of the prediction.

[0180] The first difference may be a value determined by comparing the difference between the first virtual structure microenvironment feature and the second virtual structure microenvironment feature. The first difference may be used to represent the change of the protein in the local microenvironment before and after the mutation, so as to intuitively reflect the influence of the mutation on the local microenvironment feature.

[0181] The second difference can be a numerical value determined by comparing the difference between the first global sequence feature and the second global sequence feature. The second difference can be used to represent the change in the overall structure of the protein before and after the mutation, so as to comprehensively evaluate the impact of the mutation on the global characteristics of the protein.

[0182] Exemplarily, if the first virtual structure microenvironment characteristic is The second virtual structure microenvironment is characterized by Then the first difference is the difference between the first virtual structure microenvironment characteristic and the second virtual structure microenvironment characteristic, that is,

[0183] Exemplarily, if the first global sequence feature is The second global sequence feature is Then the second difference is the difference between the first global sequence feature and the second global sequence feature, that is,

[0184] In the above way, multi-scale information can be integrated to effectively capture the impact of protein mutations at local and global scales, thereby improving the accuracy and robustness of predictions.

[0185] Step 106, predicting the protein stability of the mutant protein according to the first difference and the second difference to obtain a protein stability prediction result.

[0186] In some embodiments, the first difference provides the change in the local microenvironment of the protein before and after mutation, and the second difference provides the change in the overall structure of the protein before and after mutation. Therefore, the protein stability of the mutant protein can be predicted based on the first difference and the second difference to ensure that the prediction takes into account not only local changes but also global changes, making the prediction results more reliable and accurate.

[0187] The protein stability prediction result can be a result of predicting the protein stability by analyzing the differences in the local microenvironment and overall structure of the protein before and after the mutation. The protein stability prediction result can be used to evaluate the impact of the mutation on the protein stability.

[0188] Exemplarily, the first difference and the second difference can be respectively input into a pre-trained prediction model to obtain a first prediction result corresponding to the first difference and a second prediction result corresponding to the second difference. Further, the first prediction result and the second prediction result can be added and averaged to obtain a protein stability prediction result. Specifically, the prediction model can be a multi-layer perceptron, a deep residual network, and the like.

[0189] Exemplarily, when the prediction model is a multi-layer perceptron, the data processing process is as follows:

[0190]

[0191] ΔΔG=(ΔΔG CLS +ΔΔG ENV ) / 2

[0192] in, is the first virtual structural microenvironment feature, is the second virtual structural microenvironment characteristic, ΔΔG ENV is a first prediction result obtained according to the first difference prediction; is the first global sequence feature, is the second global sequence feature, ΔΔG CLSis the second prediction result obtained according to the second difference prediction.

[0193] Furthermore, the first prediction result and the second prediction result can be adjusted according to their importance and then averaged to obtain the protein stability prediction result. Specifically, a first weight can be set for the first prediction result, and a second weight can be set for the second prediction result. For example, protein stability prediction result = (a*first prediction result + b*second prediction result) / 2. Wherein, a represents the first weight, b represents the second weight, a and b are both values ​​between 0 and 1, the sum of a and b is 1, and the values ​​of a and b can be adjusted according to actual conditions.

[0194] The protein stability is predicted by analyzing the differences in the local microenvironment and overall structure of the protein before and after the mutation, and the first difference and the second difference are input into the pre-trained prediction model to obtain the protein stability prediction result. This method can effectively capture the impact of mutations at the local and global scales, thereby improving the accuracy and robustness of the prediction.

[0195] In some embodiments, in order to effectively capture the impact of mutations at local and global scales, protein stability prediction can be performed on the mutant protein according to the first difference and the second difference, respectively, thereby improving the accuracy and robustness of the prediction. For example, step 106 can include:

[0196] (106.1) inputting a first difference value corresponding to the first target residue feature and the second target residue feature into a pre-trained prediction model to predict the protein folding free energy in the residue feature change dimension, and obtaining a first prediction result;

[0197] (106.2) inputting the second difference value corresponding to the first sequence to be tested and the second sequence to be tested into the prediction model to predict the protein folding free energy of the global change dimension, and obtaining a second prediction result;

[0198] (106.3) Based on the average of the first prediction result and the second prediction result, the protein stability prediction result is obtained.

[0199] The prediction model may be a machine learning model or a deep learning model, which can predict the stability change of the protein after mutation according to the input first difference after training. Exemplarily, the prediction model may be a multi-layer perceptron, a convolutional neural network, a deep residual network, a recurrent neural network, etc. This application does not impose any specific restrictions on this, as long as the selected prediction model can accurately predict the first prediction result according to the first difference and accurately predict the second prediction result according to the second difference after pre-training.

[0200] The protein folding free energy may refer to the free energy released or absorbed by a protein in the process of folding from its linear amino acid sequence into a three-dimensional structure. It is understood that since a protein mutation will change the protein's folding free energy, the stability of the protein before and after the mutation can be determined by predicting the protein's folding free energy.

[0201] The first prediction result may be a predicted value of the change in folding free energy of the protein in the local microenvironment before and after mutation, obtained by inputting the first difference into a pre-trained prediction model.

[0202] The second prediction result may be a predicted value of the change in folding free energy of the protein in the global structure before and after the mutation, obtained by inputting the second difference into a pre-trained prediction model.

[0203] Specifically, the first difference corresponding to the first target residue feature and the second target residue feature, that is, the first difference determined based on the first virtual structural microenvironment feature and the second virtual structural microenvironment feature, can be directly input into the prediction model after obtaining the first difference to obtain the first prediction result.

[0204] Specifically, the second difference between the first sequence to be tested and the second sequence to be tested, that is, the second difference determined based on the first global sequence feature and the second global sequence feature, can be directly input into the prediction model to obtain the second prediction result after obtaining the second difference.

[0205] Exemplarily, if the first difference is calculated to be [-0.1, -0.1, 0.2, 0.1, 0.1], the first difference is input into the prediction model, and the prediction result of protein folding free energy under the dimension of virtual structural microenvironment characteristics is -1.5 kcal / mol. The second difference is calculated to be [0.1, 0.1, 0.1, 0.1, 0.1], and the second difference is input into the prediction model, and the prediction result of protein folding free energy under the dimension of global sequence characteristics is -1.0 kcal / mol. Then, the final protein stability prediction result is (-1.5-1.0) / 2=-1.25 kcal / mol.

[0206] Furthermore, in order to determine the degree of influence of the mutation on protein stability, the protein stability prediction result can be compared with a preset threshold value, and a classification decision can be made accordingly. For example, when the protein stability prediction result is less than a preset threshold value (e.g., 1.0 kcal / mol), it is considered that the mutation has little influence on protein stability; when the protein stability prediction result is greater than a positive threshold value (e.g., 1.0 kcal / mol), it is considered that the mutation reduces protein stability, etc. The preset threshold value can be flexibly set according to the actual situation, and no specific limitation is made here.

[0207] In some embodiments, the first virtual structural microenvironment, the second virtual structural microenvironment, the first global sequence feature, and the second global sequence feature can be mapped to a graph structure, and the topological structure can be modeled by a graph neural network to further capture the relationship between protein residues. Furthermore, the difference between the first virtual graph structure corresponding to the first virtual structural microenvironment and the second virtual graph structure corresponding to the second virtual structural microenvironment can be identified by a graph neural network to obtain a first difference, and the difference between the first global graph structure corresponding to the first global sequence feature and the second global graph structure corresponding to the second global sequence feature can be identified by a graph neural network to obtain a second difference, and the first difference and the second difference are predicted based on a prediction model (which can be a multi-layer perceptron or a graph neural network), and finally a protein stability prediction result is obtained, so that the structural relationship between protein residues can be captured more accurately.

[0208] Please refer to Figure 3 , Figure 3 This is an embodiment of the overall process of the present application. Exemplarily, the original protein and the mutant protein corresponding to the original protein can be respectively input into the pre-trained protein language model to extract sequence features, and the first virtual structural microenvironment feature and the first global sequence feature corresponding to the original protein, as well as the second virtual structural microenvironment feature and the second global sequence feature corresponding to the mutant protein can be obtained. Further, the first virtual structural microenvironment feature corresponding to the original protein can be subtracted from the second virtual structural microenvironment feature corresponding to the mutant protein to obtain a first difference, and the first global sequence feature corresponding to the original protein can be subtracted from the second global sequence feature corresponding to the mutant protein to obtain a second difference. The first difference and the second difference are respectively input into the pre-trained prediction model to obtain the first prediction result corresponding to the first difference, and the second prediction result corresponding to the second difference. Further, the first prediction result and the second prediction result can be averaged to obtain a protein stability prediction result.

[0209] Further, combined with Figure 4, an example is given of the construction of the first virtual structure microenvironment characteristic and the second virtual structure microenvironment characteristic. It should be noted that since the construction process of the first virtual structure microenvironment characteristic and the second virtual structure microenvironment characteristic are the same, for the convenience of introduction, the construction of the virtual structure microenvironment characteristic is introduced below, and the following process is applicable to the construction of the first virtual structure microenvironment characteristic and the second virtual structure microenvironment characteristic.

[0210] like Figure 4 As shown, the sequence S to be tested is input into the protein language model to obtain the sequence feature Z corresponding to the sequence to be tested. seq , the protein language model can output the overall attention map of the last Transformer layer Where H refers to the number of attention heads of the transformer, and L refers to the length of the sequence to be tested. First, the output overall attention map can be pooled to obtain the target overall attention map:

[0211]

[0212] in, Represents the overall attention map A last The overall attention map of the i-th attention head in .

[0213] Then, the position pos of the mutated residue in the given sequence is mut , through the overall attention map of the target after pooling to obtain the attention score vector associated with the mutated residue

[0214]

[0215] Select K (the value of K can be set according to actual conditions) features of several residues with the highest attention scores:

[0216] (pos 1 ,pos 2 ,…,pos k )=argmax k (A mut )

[0217]

[0218] Among them, argmax k (·) represents the function of retrieving the top K residues with the largest attention scores according to the attention score vector, Represents the top K residues with the largest attention scores of the mutated residues.

[0219] Furthermore, a learnable initial vector of residue features is set Will Send it to a transformer layer to integrate the features of these K residues into a virtual structural microenvironment feature:

[0220]

[0221] in, represents the final virtual structure microenvironment feature, ExtractCLS represents the operation of extracting CLS tokens in the protein language model, represents a transformer layer with trainable parameters θ, Represents the concatenation of features.

[0222] The present application obtains a first test sequence of an original protein and a second test sequence of a mutant protein corresponding to the original protein; determines a plurality of first target residue features from the first test sequence according to the attention mechanism of a pre-trained protein language model, and extracts global features of the plurality of first target residue features to obtain a first virtual structural microenvironment feature; determines a plurality of second target residue features from the second test sequence according to the attention mechanism of the protein language model, and extracts global features of the plurality of second target residue features to obtain a second virtual structural microenvironment feature; obtains a first global sequence feature of the first test sequence and a second global sequence feature of the second test sequence through a protein language model; determines a first difference based on the first virtual structural microenvironment feature and the second virtual structural microenvironment feature, and determines a second difference based on the first global sequence feature and the second global sequence feature; predicts protein stability of the mutant protein according to the first difference and the second difference to obtain a protein stability prediction result. In this way, the stability of the protein can be predicted by combining the local environmental characteristics around the mutated residues and the global characteristics of the entire protein sequence, so that the protein language model can understand the impact of the mutation on the local structure of the protein and the impact of the mutation on the global structure, thereby improving the accuracy of predicting the stability of the protein.

[0223] See also Figure 5 The present application also provides a protein stability prediction device, which can implement the above protein stability prediction method. The protein stability prediction device includes:

[0224] A first acquisition module 51 is used to acquire a first sequence to be tested corresponding to the original protein and a second sequence to be tested of the mutant protein corresponding to the original protein;

[0225] A first extraction module 52 is used to determine a plurality of first target residue features from the first sequence to be tested according to the attention mechanism of the pre-trained protein language model, and extract global features of the plurality of first target residue features to obtain a first virtual structure microenvironment feature;

[0226] A second extraction module 53 is used to determine a plurality of second target residue features from the second sequence to be tested according to the attention mechanism of the protein language model, and extract global features of the plurality of second target residue features to obtain a second virtual structure microenvironment feature;

[0227] A second acquisition module 54 is used to acquire a first global sequence feature of the first sequence to be tested and a second global sequence feature of the second sequence to be tested through a protein language model;

[0228] A determination module 55, configured to determine a first difference based on the first virtual structure microenvironment feature and the second virtual structure microenvironment feature, and to determine a second difference based on the first global sequence feature and the second global sequence feature;

[0229] The prediction module 56 is used to predict the protein stability of the mutant protein according to the first difference and the second difference to obtain a protein stability prediction result.

[0230] The specific implementation of the protein stability prediction device is basically the same as the specific implementation of the protein stability prediction method described above, and will not be repeated here. On the premise of meeting the requirements of the embodiments of this application, the protein stability prediction device can also be provided with other functional modules to implement the protein stability prediction method in the above embodiments.

[0231] The embodiment of the present application also provides a computer device, the computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the above protein stability prediction method when executing the computer program. The computer device can be any intelligent terminal including a tablet computer, a car computer, etc.

[0232] See also Figure 6 , Figure 6 The hardware structure of a computer device according to another embodiment is shown, and the computer device includes:

[0233] The processor 61 may be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present application;

[0234] The memory 62 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device or a random access memory (RAM). The memory 62 can store an operating system and other application programs. When the technical solution provided in the embodiment of this specification is implemented by software or firmware, the relevant program code is stored in the memory 62, and the processor 61 is called to execute the protein stability prediction method of the embodiment of this application;

[0235] Input / output interface 63, used to implement information input and output;

[0236] Communication interface 64, used to realize communication interaction between the device and other devices, which can be realized by wired mode (such as USB, network cable, etc.) or wireless mode (such as mobile network, WIFI, Bluetooth, etc.);

[0237] A bus 65 that transmits information between the various components of the device (e.g., the processor 61, the memory 62, the input / output interface 63, and the communication interface 64);

[0238] The processor 61 , the memory 62 , the input / output interface 63 and the communication interface 64 are connected to each other in communication within the device via a bus 65 .

[0239] An embodiment of the present application also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the above-mentioned protein stability prediction method is implemented.

[0240] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0241] The embodiments described in the embodiments of the present application are intended to more clearly illustrate the technical solutions of the embodiments of the present application and do not constitute a limitation on the technical solutions provided in the embodiments of the present application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0242] Those skilled in the art will appreciate that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

[0243] The device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0244] Those skilled in the art will appreciate that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices may be implemented as software, firmware, hardware, or a suitable combination thereof.

[0245] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0246] It should be understood that in the present application, "at least one (item)" and "several" refer to one or more, and "plurality" refers to two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the previous and subsequent associated objects are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0247] In the several embodiments provided in the present application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are merely schematic. For example, the division of the above units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0248] The units described above as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0249] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0250] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including multiple instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, referred to as ROM), random access memory (Random Access Memory, referred to as RAM), disk or optical disk and other media that can store programs.

[0251] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but the scope of the rights of the present invention is not limited thereto. Any modification, equivalent substitution and improvement made by a person skilled in the art without departing from the scope and essence of the present invention should be within the scope of the rights of the present invention.

Claims

1. A method for predicting protein stability, characterized in that: The method comprises: Obtaining a first sequence to be tested corresponding to the original protein and a second sequence to be tested of a mutant protein corresponding to the original protein; Determine a plurality of first target residue features from the first sequence to be tested according to the attention mechanism of the pre-trained protein language model, and extract global features of the plurality of first target residue features to obtain a first virtual structure microenvironment feature; Determining a plurality of second target residue features from the second sequence to be tested according to the attention mechanism of the protein language model, and extracting global features of the plurality of second target residue features to obtain a second virtual structure microenvironment feature; Acquire a first global sequence feature of the first sequence to be tested and a second global sequence feature of the second sequence to be tested by using the protein language model; Determining a first difference based on the first virtual structural microenvironment feature and the second virtual structural microenvironment feature, and determining a second difference based on the first global sequence feature and the second global sequence feature; The protein stability of the mutant protein is predicted according to the first difference and the second difference to obtain a protein stability prediction result.

2. The protein stability prediction method according to claim 1, characterized in that: The step of performing protein stability prediction on the mutant protein according to the first difference and the second difference to obtain a protein stability prediction result comprises: Inputting the first difference corresponding to the first target residue feature and the second target residue feature into a pre-trained prediction model to predict the protein folding free energy in the residue feature change dimension, and obtaining a first prediction result; Inputting the second difference value corresponding to the first sequence to be tested and the second sequence to be tested into the prediction model to predict the protein folding free energy of the global change dimension, and obtaining a second prediction result; Based on the average of the first prediction result and the second prediction result, a protein stability prediction result is obtained.

3. The protein stability prediction method according to claim 1, characterized in that: The determining of a plurality of first target residue features from the first sequence to be tested according to the attention mechanism of the pre-trained protein language model comprises: Generate a first overall attention map corresponding to the first sequence to be tested according to the attention mechanism of the pre-trained protein language model; Performing global pooling processing on the first overall attention map to obtain a first target overall attention map; Determining a plurality of first positions of a plurality of mutant residue features from the first test sequence; Based on the matching of the multiple first positions in the first target overall attention map, obtaining multiple first attention score vectors corresponding to the multiple mutation residue features; The number of first residue feature selections is obtained, and based on the multiple first attention score vectors, the same number of first target residue features as the number of the first residue feature selections is determined from the multiple mutation residue features.

4. The protein stability prediction method according to claim 3, characterized in that: The extracting of global features from the plurality of first target residue features to obtain a first virtual structure microenvironment feature comprises: Obtaining a preset first residue feature initial vector, and concatenating the first residue feature initial vector with the plurality of first target residue features to obtain a first target feature sequence; wherein the first residue feature initial vector is located at the first position of the first target feature sequence; Based on the first residue feature initial vector, feature aggregation is performed on each first target residue feature corresponding to the first target feature sequence and the relationship between the multiple first target residue features to obtain a first virtual structural microenvironment feature for characterizing the global information of the first target feature sequence.

5. The protein stability prediction method according to claim 1, characterized in that: The step of determining a plurality of second target residue features from the second sequence to be tested according to the attention mechanism of the protein language model comprises: Generate a second overall attention graph corresponding to the second sequence to be tested according to the attention mechanism of the protein language model; Performing global pooling processing on the second overall attention map to obtain a second target overall attention map; Determining a plurality of second positions of a plurality of mutant residue features from the second test sequence; Based on the matching of the multiple second positions in the second target overall attention map, obtain multiple second attention score vectors corresponding to the multiple mutation residue features; A second residue feature selection quantity is obtained, and based on the multiple second attention score vectors, a number of second target residue features equal to the second residue feature selection quantity is determined from the multiple mutation residue features.

6. The protein stability prediction method according to claim 5, characterized in that: The extracting of global features from the plurality of second target residue features to obtain a second virtual structure microenvironment feature comprises: Obtaining a preset second residue feature initial vector, and concatenating the second residue feature initial vector with the plurality of second target residue features to obtain a second target feature sequence; wherein the second residue feature initial vector is located at the first position of the second target feature sequence; Based on the second residue feature initial vector, feature aggregation is performed on each second target residue feature corresponding to the second target feature sequence and the relationship between the multiple second target residue features to obtain a second virtual structural microenvironment feature for characterizing the global information of the second target feature sequence.

7. The protein stability prediction method according to claim 1, characterized in that: The step of obtaining a first global sequence feature of the first sequence to be tested and a second global sequence feature of the second sequence to be tested by using the protein language model includes: Obtaining a preset first global initial vector, and performing feature aggregation on a plurality of first features in the first sequence to be tested based on the first global initial vector by using the protein language model to obtain a first global sequence feature; A preset second global initial vector is obtained, and a plurality of second features in the second sequence to be tested are aggregated based on the second global initial vector by using the protein language model to obtain a second global sequence feature.

8. A protein stability prediction device, characterized in that: The device comprises: A first acquisition module is used to acquire a first sequence to be tested corresponding to the original protein and a second sequence to be tested of a mutant protein corresponding to the original protein; A first extraction module is used to determine a plurality of first target residue features from the first sequence to be tested according to the attention mechanism of the pre-trained protein language model, and extract global features of the plurality of first target residue features to obtain a first virtual structure microenvironment feature; A second extraction module is used to determine a plurality of second target residue features from the second sequence to be tested according to the attention mechanism of the protein language model, and extract global features of the plurality of second target residue features to obtain a second virtual structure microenvironment feature; A second acquisition module, configured to acquire a first global sequence feature of the first sequence to be tested and a second global sequence feature of the second sequence to be tested by using the protein language model; A determination module, configured to determine a first difference based on the first virtual structure microenvironment feature and the second virtual structure microenvironment feature, and to determine a second difference based on the first global sequence feature and the second global sequence feature; A prediction module is used to predict the protein stability of the mutant protein according to the first difference and the second difference to obtain a protein stability prediction result.

9. A computer device, characterized in that: The computer device comprises a memory and a processor, the memory stores a computer program, and the processor implements the protein stability prediction method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the protein stability prediction method according to any one of claims 1 to 7 is implemented.

Images