Method and system for predicting binding between T cell receptor and antigen epitope
By matrix encoding and feature fusion of T cell receptors and antigen epitopes based on attention mechanism, the problem of low accuracy of binding of T cell receptors and antigen epitopes in the prior art is solved, and the data is insufficient, thus achieving the problem of low prediction accuracy of T cell receptors and antigen epitopes. Efficient prediction results.
Patent Information
- Application Number
- CN202111667711.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-12-30
- Publication Date
- 2024-12-17
- Estimated Expiration
- 2041-12-30
AI Technical Summary
The existing deep learning methods are not accurate when predicting the binding of T cell receptors to antigen epitopes, and due to insufficient data, the training network performance is poor, and data processing is complicated, which consumes a lot of manpower and time.
T cell receptor and antigen epitope sequences were matrix-encoded, features were extracted and fused, and processed using a pre-trained prediction network to predict whether the two could be combined. The method includes feature extraction network and feature fusion network, extracting and fusion features through components such as multi-head attention layer, regularization layer, and residual connection layer, and using the prediction network to make the final prediction.
Through this method, the characteristics of TCR and antigen epitope can be extracted in a small number of samples, reducing the amount of data and calculations, and improving the accuracy and efficiency of prediction.
Smart Images
Figure CN114360644B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of biological information processing technology, and in particular to a method and system for predicting the binding of a T cell receptor to an antigen epitope based on an attention mechanism. Background Art
[0002] Bioinformatics analysis is a key task and important research direction in the intersection of computer science and biology. It refers to the process of analyzing and predicting based on acquired biological sequence data. From a mathematical point of view, it is the process of multivariate statistics and spatial mapping of sequence data.
[0003] Immune response is an immune effector mechanism produced by the immune system during its evolution and development. It usually refers to adaptive immunity, which is the defense function of the body to recognize and respond to specific pathogens (antigens) during long-term contact with exogenous pathogens, and eventually eliminate them from the body. Most lymphocytes can specifically recognize specific pathogens. According to the different components and functions involved in immunity, they can be divided into humoral immunity and cellular immunity. The special part of the antigen molecule that is recognized by individual lymphocytes is called an antigenic determinant or epitope. Lymphocyte antigen receptors exist on the surface of the lymphocyte membrane, which can specifically recognize antigenic determinants.
[0004] In the process of cellular immunity, the binding of antigens to T cell receptors (TCR) requires two processes. First, protein antigens are processed into polypeptides by antigen presenting cells (APCs), and then bind to major histocompatibility complexes (MHC) and move to the surface of APCs to generate activation TCR signals. Second, the antigens bound to MHC bind to TCRs to generate a second membrane signal, a synergistic stimulation signal. Only under dual signal stimulation can T lymphocytes be activated, becoming a dual signal mode. After being activated, T lymphocytes are transformed into lymphoblasts, and rapidly proliferate and differentiate to exert immune effects.
[0005] Immune cells usually only recognize a specific part of the antigen peptide molecule, namely the epitope, also known as the antigenic determinant. It is a special chemical group with a certain composition and structure located on the surface of the antigen molecule or other parts. It is the basic structural unit for the immunogen to cause an immune response and produce a specific reaction with antibodies. The study of antigen epitopes is of great significance for the diagnosis and prognosis of diseases, the site-specific modification of protein molecules to reduce the immunogenicity of protein drugs, the design of vaccine molecular structure and immune intervention therapy.
[0006] In recent years, with the in-depth development of deep learning technology, bioinformatics analysis technology has achieved good results, among which the prediction of the binding of T cells and antigen epitopes has been widely developed. Although the existing deep learning methods have achieved good results, the existing deep learning methods focus more on the prediction of the binding of antigens and MHC molecules, that is, the first process of immune response, and there are fewer studies on the second process, that is, the prediction of the binding of antigens and TCR molecules, resulting in low prediction accuracy in this direction. Moreover, the existing deep learning methods require large-scale data sets for training network models, but there is a lack of sufficient data support in this field, resulting in poor performance of the trained network models, and the data in this field are mostly data pairs that can bind TCR and antigens, and in the face of complicated sequencing work, a lot of manpower and time are consumed. Summary of the invention
[0007] The object of the present invention is to provide a method and system for predicting the binding of T cell receptors and antigen epitopes, which can extract effective TCR sequence features and antigen epitope sequence features, interactively fuse the extracted features, and predict the fused features, thereby obtaining a binding prediction result with a higher accuracy, so as to solve at least one technical problem existing in the above-mentioned background technology.
[0008] In order to achieve the above object, the present invention adopts the following technical solutions:
[0009] In one aspect, the present invention provides a method for predicting the binding of a T cell receptor to an antigen epitope, comprising:
[0010] Matrix encoding of T cell receptor sequences and antigen epitope sequences to be predicted;
[0011] Extracting encoded T cell receptor sequence features;
[0012] Extracting the encoded antigen epitope sequence features, and fusing the T cell receptor sequence features with the antigen epitope sequence features to obtain fusion features;
[0013] The fusion features are processed using a pre-trained prediction network to obtain a result of whether the T cell receptor to be predicted and the antigen epitope can bind.
[0014] Optionally, the pre-trained prediction network is obtained by training a training set, wherein the training set includes a plurality of sequence pairs consisting of T cell receptor sequences and antigen epitope sequences and labels indicating whether the T cell receptor sequences and antigen epitope sequences in the sequence pairs can bind.
[0015] Optionally, matrix encoding of the predicted T cell receptor sequence and antigen epitope sequence includes: using the BLOSUM62 matrix to encode the TCR sequence and the antigen epitope sequence, the BLOSUM matrix is a matrix formed by selecting protein families with good conservation from the BLOCKS database, and counting the relative frequency and probability of amino acid substitutions, and BLOSUM62 is used to compare sequences with 62% similarity.
[0016] Optionally, a feature extraction network is used to extract features of the encoded T cell receptor sequence, wherein the feature extraction network is composed of 6 attention modules, each of which includes 1 multi-head attention layer, 1 regularization layer and residual connection layer, 1 fully connected layer, and 1 regularization layer and residual connection layer.
[0017] Optionally, a feature fusion network is used to extract antigen epitope features, and at the same time, the extracted TCR features are fused with the antigen epitope features. The feature fusion network is composed of 6 attention modules, each of which includes 2 multi-head attention layers with regularization layers and residual connection layers, and 1 fully connected layer with regularization layers and residual connection layers; wherein the first attention layer of the first attention module extracts feature information of the antigen epitope, and the second attention layer fuses the extracted T cell receptor sequence features with the antigen epitope features, and subsequent attention modules continue to fuse features and fuse again to obtain the final fused features.
[0018] Optionally, the prediction network consists of 3 fully connected layers with dropout layers, 1 pooling layer, 1 ReLU layer, 1 dropout layer and a SoftMax layer.
[0019] In a second aspect, the present invention provides a T cell receptor and antigen epitope binding prediction system, comprising:
[0020] An encoding module, used for matrix encoding of T cell receptor sequences and antigen epitope sequences to be predicted;
[0021] A feature extraction module, used to extract the features of the encoded T cell receptor sequence;
[0022] A feature fusion module is used to extract the encoded antigen epitope sequence features and fuse the T cell receptor sequence features with the antigen epitope sequence features to obtain a fusion feature;
[0023] The prediction module is used to process the fusion features using a pre-trained prediction network to obtain a result of whether the T cell receptor to be predicted and the antigen epitope can bind.
[0024] In a third aspect, the present invention provides a non-transitory computer-readable storage medium, which is used to store computer instructions. When the computer instructions are executed by a processor, the above-mentioned method for predicting the binding of a T cell receptor to an antigen epitope is implemented.
[0025] In a fourth aspect, the present invention provides a computer program product, comprising a computer program, which, when executed on one or more processors, is used to implement the method for predicting the binding between a T cell receptor and an antigen epitope as described above.
[0026] In a fifth aspect, the present invention provides an electronic device comprising: a processor, a memory and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory so that the electronic device executes instructions for implementing the method for predicting the binding of a T cell receptor to an antigen epitope as described above.
[0027] The beneficial effects of the present invention are as follows: T cell receptor and antigen epitope binding prediction based on the attention mechanism improves the prediction network, inputs a set of sequence pairs of TCR sequences and antigen epitope sequences, extracts TCR sequence features, extracts antigen epitope sequence features and fuses the extracted TCR sequence features to obtain fusion features, and uses the prediction network to predict the possibility of TCR sequence and antigen epitope sequence binding through the fusion features; in this way, the features of different TCR sequences and antigen epitope sequences can be extracted through a small number of samples and the binding of the sequences can be predicted, which reduces the amount of data used for model training, reduces the amount of calculation data, and improves calculation efficiency and prediction accuracy.
[0028] Additional aspects and advantages of the present invention will be given in part in the following description, which will become obvious from the following description, or may be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS
[0029] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.
[0030] Figure 1 This is a functional principle block diagram of the T cell receptor and antigen epitope binding prediction system based on the attention mechanism described in Example 1 of the present invention.
[0031] Figure 2 This is a flow chart of the method for predicting the binding between a T cell receptor and an antigen epitope based on the attention mechanism described in Example 1 of the present invention.
[0032] Figure 3 This is a flow chart of the method for predicting the binding between a T cell receptor and an antigen epitope based on the attention mechanism described in Example 2 of the present invention.
[0033] Figure 4 This is a functional principle block diagram of the T cell receptor and antigen epitope binding prediction system based on the attention mechanism described in Example 2 of the present invention. DETAILED DESCRIPTION
[0034] The embodiments of the present invention are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements with the same or similar functions. The embodiments described below by the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be interpreted as limiting the present invention.
[0035] It should be understood by those skilled in the art that unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by those skilled in the art to which this invention belongs.
[0036] It should also be understood that terms, such as those defined in commonly used dictionaries, should be understood to have a meaning consistent with that in the context of the prior art and will not be interpreted in an idealized or overly formal sense unless as defined herein.
[0037] Those skilled in the art will appreciate that, unless otherwise stated, the singular forms "a", "an", "said" and "the" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present invention refers to the presence of the features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements and / or groups thereof.
[0038] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine different embodiments or examples described in this specification and the features of different embodiments or examples, unless they are contradictory.
[0039] To facilitate understanding of the present invention, the present invention is further explained below with reference to specific embodiments in conjunction with the accompanying drawings, and the specific embodiments do not constitute a limitation on the embodiments of the present invention.
[0040] Those skilled in the art should understand that the drawings are merely schematic diagrams of embodiments, and the components in the drawings are not necessarily necessary for implementing the present invention.
[0041] Example 1
[0042] like Figure 1 As shown, this embodiment 1 provides a T cell receptor and antigen epitope binding prediction system, the system comprising:
[0043] An encoding module, used for matrix encoding of T cell receptor sequences and antigen epitope sequences to be predicted;
[0044] A feature extraction module, used to extract the features of the encoded T cell receptor sequence;
[0045] A feature fusion module is used to extract the encoded antigen epitope sequence features and fuse the T cell receptor sequence features with the antigen epitope sequence features to obtain a fusion feature;
[0046] The prediction module is used to process the fusion features using a pre-trained prediction network to obtain a result of whether the T cell receptor to be predicted and the antigen epitope can bind.
[0047] like Figure 2 As shown, in this embodiment 1, the above system is used to implement a method for predicting the binding between a T cell receptor and an antigen epitope, including:
[0048] Matrix encoding of T cell receptor sequences and antigen epitope sequences to be predicted;
[0049] Extracting encoded T cell receptor sequence features;
[0050] Extracting the encoded antigen epitope sequence features, and fusing the T cell receptor sequence features with the antigen epitope sequence features to obtain fusion features;
[0051] The fusion features are processed using a pre-trained prediction network to obtain a result of whether the T cell receptor to be predicted and the antigen epitope can bind.
[0052] The pre-trained prediction network is obtained by training a training set, wherein the training set includes a plurality of sequence pairs consisting of T cell receptor sequences and antigen epitope sequences and labels indicating whether the T cell receptor sequences and antigen epitope sequences in the sequence pairs can bind.
[0053] Matrix encoding of the predicted T cell receptor sequence and antigen epitope sequence includes: using the BLOSUM62 matrix to encode the TCR sequence and antigen epitope sequence. The specific method is as follows: set the maximum length of the TCR sequence and antigen epitope sequence to 25, use '*' to fill in the sequence with a length of less than 25, and use the BLOSUM62 matrix to encode the TCR sequence and antigen epitope sequence. The BLOSUM62 matrix is calculated based on the relationship sequence with a consistency of ≥62%. The rows and columns of the matrix represent the 20 common amino acids in the human body, and the matrix elements are calculated based on the relative frequency and probability of amino acid substitutions. The TCR sequence and antigen epitope sequence are self-encoded by the BLOSUM62 matrix.
[0054] The feature extraction network is used to extract the features of the encoded T cell receptor sequence. The feature extraction network consists of 6 attention modules, each of which contains 1 multi-head attention layer, 1 regularization layer and residual connection layer, 1 fully connected layer, and 1 regularization layer and residual connection layer. The specific method is as follows: The feature extraction network is used to extract the TCR sequence features, and the output of the last fully connected layer of the model is used as the extracted deep features. Specifically, the feature extraction network consists of 6 attention modules, each of which contains 1 multi-head attention layer, 1 regularization layer and residual connection layer, 1 fully connected layer, and 1 regularization layer and residual connection layer. The first few modules of the network learn relatively global features, and as the modules deepen, more local features will be learned. Take the initial encoding of the TCR sequence as input and output the TCR features extracted from the TCR sequence.
[0055] The feature fusion network is used to extract antigen epitope features. At the same time, the extracted TCR features are fused with the antigen epitope features. The feature fusion network consists of 6 attention modules, each of which includes 2 multi-head attention layers with regularization layers and residual connection layers, and 1 fully connected layer with regularization layers and residual connection layers. Among them, the first attention layer of the first attention module extracts the feature information of the antigen epitope, and the second attention layer fuses the extracted T cell receptor sequence features and antigen epitope features. Subsequent attention modules continue to fuse features and fuse them again to obtain the final fused features.
[0056] The prediction network consists of 3 fully connected layers with a dropout layer, 1 pooling layer, 1 ReLU layer, 1 dropout layer, and a SoftMax layer. The fused features are used as input, and after the prediction network output, a threshold is set for the output result. If it is higher than the threshold, it is considered that the TCR is more likely to bind to the antigen epitope. If it is lower than the threshold, it is considered that the TCR is more likely not to bind to the antigen epitope, and the pixel loss is back-propagated to continuously improve the learning ability.
[0057] Example 2
[0058] In this embodiment 2, a method for predicting the binding of a T cell receptor to an antigen epitope based on an attention mechanism is provided, such as Figure 3 As shown, the method for predicting the binding between T cell receptor and antigen epitope based on attention mechanism includes the following steps:
[0059] Step (1): receiving the collected sequence data;
[0060] Step (2): classify the received sequence data into categories that can be combined;
[0061] Step (3): screening the classified sequence data;
[0062] Step (4): performing data enhancement processing on the screened sequence data;
[0063] Step (5): Encoding the enhanced sequence data;
[0064] Step (6): input the initially encoded sequence data into the feature extraction network for training and learning, and extract features according to the model;
[0065] Step (7): The extracted sequence features are input into the feature fusion network to fuse the features;
[0066] Step (8): The fusion features are input into the prediction network to obtain the final combined prediction result.
[0067] The specific steps include the following:
[0068] Step (1): Receive the collected image data. The received image data includes TCR sequences and antigen epitope sequences. The sequences have been labeled and the images are classified according to the labels. The TCR sequences and antigen epitope sequences can be pre-stored in the memory of the computer device. When they need to be processed, the processor reads the data directly from the memory of the computer device. Of course, the processor can also obtain the TCR and antigen epitope sequences from an external device. For example, the TCR and antigen epitope sequences are stored in the cloud, and when processing operations need to be performed, the processor obtains them from the cloud. This embodiment does not limit the specific way in which the processor obtains the TCR and antigen epitope sequences.
[0069] Step (2): Classify the received sequence data according to whether they can bind. After receiving the collected TCR sequences and antigen sequences, the data used for training and testing should be divided into two categories according to whether the TCR can bind to the antigen. The binding label of the sequence pairs that can bind is set to 1, and the binding label of the sequence pairs that cannot bind is set to 0. In addition, each antigen epitope is divided into a set.
[0070] Step (3): Screen the classified sequence data. For TCR sequences and antigen epitope sequences, the length of the data to be trained is not fixed, and the distribution of data of different lengths is also uneven. Therefore, in order to facilitate the training of subsequent models, the lengths of TCR sequences and antigen epitope sequences need to be counted during preprocessing. After statistical analysis, sequence pairs with a length of more than 25 are eliminated.
[0071] Step (4): Perform data enhancement processing on the screened sequence data. Data enhancement processing can purposefully adjust the amount of data, prevent the network from overfitting, and enhance the robustness of the network. For the existing data, the ratio of positive and negative samples is unbalanced, and most of them are data that can bind to TCR and antigen epitopes. For the actual situation of biological species, the data that TCR and antigen epitopes cannot bind far exceeds the data that can bind. Therefore, more negative sample pairs, that is, sequence pairs that TCR and antigen epitopes cannot bind, are generated by random mismatching, so that the ratio of positive and negative sample pairs for each type of antigen epitope is 1:10. Specifically, as shown in Table 1, part of the data used to train the T cell receptor and antigen epitope binding prediction model of the attention mechanism.
[0072] Table 1
[0073]
[0074] Step (5): Preliminary encoding of the enhanced sequence data, the specific method is as follows: set the maximum length of the TCR sequence and the antigen epitope sequence to 25, use '*' to fill the sequence with a length less than 25, and use the BLOSUM62 matrix to encode the TCR sequence and the antigen epitope sequence. The BLOSUM62 matrix is calculated based on the relationship sequence with a consistency of ≥62%. The rows and columns of the matrix represent the 20 common amino acids in the human body, and the matrix elements are calculated based on the relative frequency and probability of amino acid substitution. The TCR sequence and the antigen epitope sequence are self-encoded by the BLOSUM62 matrix.
[0075] Step (6): The preliminarily encoded sequence data is input into the feature extraction network for training and learning, and features are extracted according to the model. The specific method is as follows: Use the feature extraction network to extract TCR sequence features, and use the output of the last fully connected layer of the model as the extracted deep features. Specifically, the feature extraction network consists of 6 attention modules, each of which contains 1 multi-head attention layer, 1 regularization layer and residual connection layer, 1 fully connected layer and 1 regularization layer and residual connection layer. The first few modules of the network learn relatively global features, and as the modules deepen, more local features will be learned. Take the initial encoding of the TCR sequence as input and output the TCR features extracted from the TCR sequence.
[0076] Step (7): The extracted sequence features are input into the feature fusion network to fuse the features. The specific method is as follows: the feature fusion network is used to extract the antigen epitope features, and at the same time, the extracted TCR features are fused with the antigen epitope features. Specifically, the feature fusion network is composed of 6 attention modules, each of which contains 2 multi-head attention layers with regularization layers and residual connection layers, and 1 fully connected layer with regularization layers and residual connection layers, with the initial encoding of the TCR sequence features and the antigen epitope sequence as input. The first attention layer of the first attention module of the network extracts the feature information of the antigen epitope, and the second attention layer fuses the extracted TCR features with the antigen epitope features. The subsequent attention modules continue to fuse the features at a deeper level and fuse them again. After the deepening of the modules, the feature information of the TCR sequence and the antigen epitope sequence is better fused.
[0077] Step (8), the fused features are input into the prediction network to obtain the final binding prediction result. The specific method is as follows: the fused features are input into the prediction network, and the TCR is predicted to bind to the antigen epitope based on the extracted fusion features. Specifically, the prediction network consists of 3 fully connected layers with a dropout layer, 1 pooling layer, 1 ReLU layer, 1 dropout layer and a SoftMax layer.
[0078] The fused features are used as input. After the prediction network output, a threshold is set for the output result. If it is higher than the threshold, it is considered that the TCR is more likely to bind to the antigen epitope. If it is lower than the threshold, it is considered that the TCR is more likely not to bind to the antigen epitope. The pixel loss is back-propagated to continuously improve the learning ability. The loss function is expressed as follows:
[0079]
[0080] Among them, y i represents the true label of the i-th sequence pair, p i Denotes the corresponding predicted binding probability.
[0081] Table 2 shows the accuracy of binding prediction for different TCR sequences and antigen epitope sequences using the method of this embodiment.
[0082] Table 2
[0083]
[0084] In this Example 2, based on the above method, a T cell receptor and antigen epitope binding prediction system based on the attention mechanism was constructed, such as Figure 4 As shown, the system includes:
[0085] The preprocessing module is configured to preprocess the TCR sequence and antigen epitope sequence, remove the sequence with a length of more than 25, and generate the encoding of the TCR sequence and antigen epitope sequence using BLOSUM62 encoding;
[0086] An input module is configured to input the TCR sequence and the antigen epitope sequence into a feature extraction network and a feature fusion network respectively;
[0087] A feature extraction module is configured to extract features of the TCR sequence based on a preset deep neural feature extraction network;
[0088] The feature fusion module is configured to extract features of the antigen epitope sequence based on a preset deep neural network, and fuse it with the extracted TCR features to obtain a fusion feature;
[0089] The result output module is configured to combine and predict the fusion features through a multi-layer perceptron and output the prediction results.
[0090] Example 3
[0091] Embodiment 3 of the present invention provides a non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium is used to store computer instructions. When the computer instructions are executed by a processor, the above-mentioned method for predicting the binding between a T cell receptor and an antigen epitope is implemented. The method includes:
[0092] Matrix encoding of T cell receptor sequences and antigen epitope sequences to be predicted;
[0093] Extracting encoded T cell receptor sequence features;
[0094] Extracting the encoded antigen epitope sequence features, and fusing the T cell receptor sequence features with the antigen epitope sequence features to obtain fusion features;
[0095] The fusion features are processed using a pre-trained prediction network to obtain a result of whether the T cell receptor to be predicted and the antigen epitope can bind.
[0096] Example 4
[0097] Embodiment 4 of the present invention provides a computer program (product), including a computer program, which, when executed on one or more processors, is used to implement the above-mentioned method for predicting the binding of a T cell receptor to an antigen epitope, the method comprising:
[0098] Matrix encoding of T cell receptor sequences and antigen epitope sequences to be predicted;
[0099] Extracting encoded T cell receptor sequence features;
[0100] Extracting the encoded antigen epitope sequence features, and fusing the T cell receptor sequence features with the antigen epitope sequence features to obtain fusion features;
[0101] The fusion features are processed using a pre-trained prediction network to obtain a result of whether the T cell receptor to be predicted and the antigen epitope can bind.
[0102] Example 5
[0103] Embodiment 5 of the present invention provides an electronic device, comprising: a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory, so that the electronic device executes instructions for implementing the above-mentioned method for predicting the binding of a T cell receptor to an antigen epitope, the method comprising:
[0104] Matrix encoding of T cell receptor sequences and antigen epitope sequences to be predicted;
[0105] Extracting encoded T cell receptor sequence features;
[0106] Extracting the encoded antigen epitope sequence features, and fusing the T cell receptor sequence features with the antigen epitope sequence features to obtain fusion features;
[0107] The fusion features are processed using a pre-trained prediction network to obtain a result of whether the T cell receptor to be predicted and the antigen epitope can bind.
[0108] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0109] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0110] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0111] These computer program instructions can also be loaded onto a computer or other programmable data processing device, and a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0112] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative work on the basis of the technical solution disclosed in the present invention should be included in the scope of protection of the present invention.
Claims
1. A method for predicting the binding of a T cell receptor to an antigen epitope, characterized in that: include: Matrix encoding of T cell receptor sequences and antigen epitope sequences to be predicted; Extracting encoded T cell receptor sequence features; Extracting the encoded antigen epitope sequence features, and fusing the T cell receptor sequence features with the antigen epitope sequence features to obtain fusion features; Using a pre-trained prediction network, the fusion feature is processed to obtain a result of whether the T cell receptor to be predicted and the antigen epitope can bind; The specific steps are as follows: step (1): receiving the collected sequence data; step (2): classifying the received sequence data to determine whether they can be combined; step (3): screening the classified sequence data; step (4): performing data enhancement processing on the screened sequence data; step (5): encoding the enhanced sequence data; step (6): inputting the initially encoded sequence data into the feature extraction network for training and learning, and extracting features according to the model; step (7): inputting the extracted sequence features into the feature fusion network for fusion; step (8): inputting the fusion features into the prediction network to obtain the final combined prediction result; The feature fusion network is used to extract antigen epitope features. At the same time, the extracted TCR features are fused with the antigen epitope features. The feature fusion network consists of 6 attention modules, each of which includes 2 multi-head attention layers with regularization layers and residual connection layers, and 1 fully connected layer with regularization layers and residual connection layers. Among them, the first attention layer of the first attention module extracts the feature information of the antigen epitope, and the second attention layer fuses the extracted T cell receptor sequence features and antigen epitope features. Subsequent attention modules continue to fuse features and fuse them again to obtain the final fused features.
2. The method for predicting the binding of a T cell receptor to an antigen epitope according to claim 1, characterized in that: The pre-trained prediction network is obtained by training a training set, wherein the training set includes a plurality of sequence pairs consisting of a T cell receptor sequence and an antigen epitope sequence and a label indicating whether the T cell receptor sequence and the antigen epitope sequence in the sequence pair can bind.
3. The method for predicting the binding of a T cell receptor to an antigen epitope according to claim 1, characterized in that: Matrix encoding of the predicted T cell receptor sequence and antigen epitope sequence includes: using the BLOSUM62 matrix to encode the TCR sequence and the antigen epitope sequence, the rows and columns of the BLOSUM62 matrix represent the 20 common amino acids in the human body, and the matrix elements are calculated based on the relative frequency and probability of amino acid substitutions.
4. The method for predicting the binding of a T cell receptor to an antigen epitope according to claim 1, characterized in that: The feature extraction network is used to extract the encoded T cell receptor sequence features, where the feature extraction network consists of 6 attention modules, each of which contains 1 multi-head attention layer, 1 regularization layer and residual connection layer, 1 fully connected layer and 1 regularization layer and residual connection layer.
5. The method for predicting the binding between a T cell receptor and an antigen epitope according to claim 1, characterized in that: The prediction network consists of 3 fully connected layers with dropout layers, 1 pooling layer, 1 ReLU layer, 1 dropout layer and a SoftMax layer.
6. A system for predicting the binding of a T cell receptor to an antigen epitope, using the method for predicting the binding of a T cell receptor to an antigen epitope as claimed in claim 1, characterized in that: include: An encoding module, used for matrix encoding of T cell receptor sequences and antigen epitope sequences to be predicted; A feature extraction module, used to extract the features of the encoded T cell receptor sequence; A feature fusion module is used to extract the encoded antigen epitope sequence features and fuse the T cell receptor sequence features with the antigen epitope sequence features to obtain a fusion feature; The prediction module is used to process the fusion features using a pre-trained prediction network to obtain a result of whether the T cell receptor to be predicted and the antigen epitope can bind.
7. A non-transitory computer-readable storage medium, characterized in that: The non-transitory computer-readable storage medium is used to store computer instructions, and when the computer instructions are executed by the processor, the method for predicting the binding between a T cell receptor and an antigen epitope as described in any one of claims 1 to 5 is implemented.
8. A computer program product, characterized in that The invention comprises a computer program, which, when running on one or more processors, is used to implement the method for predicting the binding between a T cell receptor and an antigen epitope according to any one of claims 1 to 5.
9. An electronic device, characterized in that: include: A processor, a memory and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory so that the electronic device executes instructions for implementing the method for predicting the binding of a T cell receptor to an antigen epitope as described in any one of claims 1 to 5.