Affinity prediction method, device, equipment and storage medium

Abstract

The present application discloses an affinity prediction method, device, equipment and storage medium, which belongs to the field of computer technology. Through the technical solution provided in the embodiment of the present application, the affinity prediction model extracts features of the gene information and sequence of the immune cell receptor to obtain the gene features and sequence features of the immune cell receptor. In the process of obtaining the receptor features of the immune cell receptor, the gene features, sequence features and three-dimensional structural features are integrated. The introduction of three-dimensional structural features enriches the content of the receptor features and improves the expressiveness of the receptor features, so that when affinity prediction is performed based on the receptor features, the accuracy of the affinity between the immune cell receptor and the target antigen is high.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to an affinity prediction method, device, equipment and storage medium. Background Art

[0002] The human immune system is composed of innate immunity and adaptive immunity. The adaptive immune system is realized by a variety of immune cells, which can respond specifically to specific pathogens. The immune cell receptor is the region of the immune cell that recognizes antigens. Successful recognition of antigens can activate the immune system to eliminate pathogens and play an important role in maintaining human health. Studying the affinity of immune cell receptors for specific antigens is crucial to understanding the immune system, and can further promote the design and development of immunotherapy and vaccines. Based on this, a method to predict the affinity of immune cell receptors for specific antigens is urgently needed. Summary of the invention

[0003] The present application embodiment provides an affinity prediction method, device, equipment and storage medium, which can predict the affinity of immune cell receptors to specific antigens. The technical solution is as follows:

[0004] In one aspect, a method for affinity prediction is provided, the method comprising:

[0005] Input the genetic information, sequence information, and three-dimensional structural features of immune cell receptors into the affinity prediction model;

[0006] By using the affinity prediction model, feature extraction is performed on the gene information and sequence information of the immune cell receptor to obtain the gene features and sequence features of the immune cell receptor;

[0007] The gene characteristics, sequence characteristics and three-dimensional structural characteristics of the immune cell receptor are integrated through the affinity prediction model to obtain the receptor characteristics of the immune cell receptor;

[0008] The affinity prediction model is used to perform multiple full connections on the receptor features of the immune cell receptor, and the affinity of the immune cell receptor for the target antigen is output, where the target antigen is an antigen that can specifically bind to the immune cell receptor.

[0009] In one aspect, a method for training an affinity prediction model is provided, the method comprising:

[0010] Inputting the gene information, sequence information and three-dimensional structural characteristics of the sample immune cell receptor into the affinity prediction model;

[0011] By using the affinity prediction model, feature extraction is performed on the gene information and sequence information of the sample immune cell receptor to obtain the gene features and sequence features of the sample immune cell receptor;

[0012] The gene features, sequence features and three-dimensional structural features of the sample immune cell receptor are integrated through the affinity prediction model to obtain the receptor features of the sample immune cell receptor;

[0013] The affinity prediction model is used to perform multiple full connections on the receptor features of the sample immune cell receptor, and the predicted affinity of the sample immune cell receptor for the sample antigen is output, where the sample antigen is an antigen that can specifically bind to the sample immune cell receptor;

[0014] Classifying the receptor characteristics of the sample immune cell receptors by using the affinity prediction model, and outputting the predicted affinity level of the sample immune cell receptors for the sample antigen;

[0015] The affinity prediction model is trained based on first difference information and second difference information, wherein the first difference information is the difference information between the predicted affinity of the sample immune cell receptor for the sample antigen and the labeled affinity of the sample immune cell receptor for the sample antigen, and the second difference information is the difference information between the predicted affinity level of the sample immune cell receptor for the sample antigen and the labeled affinity level of the sample immune cell receptor for the sample antigen.

[0016] In one aspect, a device for predicting affinity is provided, the device comprising:

[0017] An input unit, used to input the gene information, sequence information and three-dimensional structural characteristics of the immune cell receptor into the affinity prediction model;

[0018] A feature extraction unit, used to extract features from the gene information and sequence information of the immune cell receptor using the affinity prediction model to obtain the gene features and sequence features of the immune cell receptor;

[0019] A feature fusion unit, used to fuse the gene features, sequence features and three-dimensional structural features of the immune cell receptor through the affinity prediction model to obtain the receptor features of the immune cell receptor;

[0020] The affinity prediction unit is used to perform affinity prediction based on the receptor characteristics of the immune cell receptor through the affinity prediction model, and output the affinity of the immune cell receptor to the target antigen, where the target antigen is an antigen that can specifically bind to the immune cell receptor.

[0021] In a possible embodiment, the feature extraction unit is used to encode the VDJ information of the immune cell receptor through the gene encoder of the affinity prediction model to obtain the gene characteristics of the immune cell receptor, wherein V is the encoding variable region, D is the encoding hypervariable region, and J is the encoding cross-linking region; and encode the amino acid sequence of the immune cell receptor through the sequence encoder of the affinity prediction model to obtain the sequence characteristics of the immune cell receptor.

[0022] In a possible implementation, the feature extraction unit is configured to perform any of the following:

[0023] When the immune cell receptor is a B cell receptor, encoding the VJ information of the light chain and the VDJ information of the heavy chain of the immune cell receptor to obtain the gene characteristics of the immune cell receptor;

[0024] When the immune cell receptor is a T cell receptor, the VJ information of the α chain and the VDJ information of the β chain of the immune cell receptor are encoded to obtain the gene characteristics of the immune cell receptor.

[0025] In a possible embodiment, the feature extraction unit is used to fully connect the VJ information of the light chain and the VDJ information of the heavy chain of the immune cell receptor to obtain the gene characteristics of the immune cell receptor, and the gene characteristics of the immune cell receptor include the light chain gene characteristics of the immune cell receptor and the heavy chain gene characteristics of the immune cell receptor; encoding the VJ information of the α chain and the VDJ information of the β chain of the immune cell receptor to obtain the gene characteristics of the immune cell receptor includes: fully connecting the VJ information of the α chain and the VDJ information of the β chain of the immune cell receptor to obtain the gene characteristics of the immune cell receptor, and the gene characteristics of the immune cell receptor include the α chain gene characteristics of the immune cell receptor and the β chain gene characteristics of the immune cell receptor.

[0026] In a possible implementation, the feature extraction unit is configured to perform any of the following:

[0027] In the case where the immune cell receptor is a B cell receptor, the amino acid sequence of the light chain and the amino acid sequence of the heavy chain of the immune cell receptor are encoded by the sequence encoder of the affinity prediction model based on the attention mechanism to obtain the sequence features of the immune cell receptor, wherein the sequence features of the immune cell receptor include the sequence features of the light chain and the heavy chain of the immune cell receptor;

[0028] In the case where the immune cell receptor is a T cell receptor, the amino acid sequence of the α chain and the amino acid sequence of the β chain of the immune cell receptor are encoded by the sequence encoder of the affinity prediction model based on the attention mechanism to obtain the sequence characteristics of the immune cell receptor, and the sequence characteristics of the immune cell receptor include the α chain sequence characteristics and the β chain sequence characteristics of the immune cell receptor.

[0029] In a possible embodiment, the feature fusion unit is used to splice the gene features and sequence features of the immune cell receptor through the feature fusion device of the affinity prediction model to obtain the gene sequence fusion features of the immune cell receptor; based on the gated attention mechanism, the gene sequence fusion features and three-dimensional structure features of the immune cell receptor are weightedly fused to obtain the receptor features of the immune cell receptor.

[0030] In a possible implementation, the device further includes:

[0031] A three-dimensional structural feature acquisition unit is used to acquire the amino acid sequence of the CDR3 region of the immune cell receptor; perform a multiple sequence alignment on the amino acid sequence of the CDR3 region of the immune cell receptor to obtain at least one reference amino acid sequence, wherein the similarity between the reference amino acid sequence and the amino acid sequence of the CDR3 region of the immune cell receptor meets a similarity condition; acquire a homologous template corresponding to the amino acid sequence of the CDR3 region of the immune cell receptor, wherein the homologous template includes structural information of the homologous sequence of the amino acid sequence of the CDR3 region of the immune cell receptor; perform multiple rounds of iterations based on the amino acid sequence of the CDR3 region of the immune cell receptor, the at least one reference amino acid sequence and the homologous template to obtain the three-dimensional structural feature of the immune cell receptor.

[0032] In a possible implementation, the device further includes:

[0033] A three-dimensional structural feature acquisition unit, used to acquire the three-dimensional structural information of the immune cell receptor, wherein the three-dimensional structural information includes the three-dimensional coordinates of a plurality of amino acids in the immune cell receptor;

[0034] The three-dimensional structure feature acquisition unit is used to perform any of the following:

[0035] Performing graph convolution on the three-dimensional structural information of the immune cell receptor to obtain the three-dimensional structural features of the immune cell receptor;

[0036] The three-dimensional structural information of the immune cell receptor is encoded based on the attention mechanism to obtain the three-dimensional structural features of the immune cell receptor.

[0037] In a possible embodiment, the feature fusion unit is also used to fuse the gene features, sequence features, three-dimensional structural features of the immune cell receptor and the physicochemical information of amino acids in the immune cell receptor through the affinity prediction model to obtain the receptor features of the immune cell receptor.

[0038] In one aspect, a training device for an affinity prediction model is provided, the device comprising:

[0039] A training information input unit, used to input the gene information, sequence information and three-dimensional structural characteristics of the sample immune cell receptor into the affinity prediction model;

[0040] A training feature extraction unit is used to extract features from the gene information and sequence information of the sample immune cell receptor by using the affinity prediction model to obtain the gene features and sequence features of the sample immune cell receptor;

[0041] A training feature fusion unit is used to fuse the gene features, sequence features and three-dimensional structural features of the sample immune cell receptor through the affinity prediction model to obtain the receptor features of the sample immune cell receptor;

[0042] A predicted affinity output unit, used to perform multiple full connections on the receptor features of the sample immune cell receptor through the affinity prediction model, and output the predicted affinity of the sample immune cell receptor for the sample antigen, wherein the sample antigen is an antigen that can specifically bind to the sample immune cell receptor;

[0043] A training classification unit is used to classify the receptor characteristics of the sample immune cell receptor by using the affinity prediction model, and output the predicted affinity level of the sample immune cell receptor for the sample antigen;

[0044] A training unit is used to train the affinity prediction model based on first difference information and second difference information, wherein the first difference information is the difference information between the predicted affinity of the sample immune cell receptor for the sample antigen and the labeled affinity of the sample immune cell receptor for the sample antigen, and the second difference information is the difference information between the predicted affinity level of the sample immune cell receptor for the sample antigen and the labeled affinity level of the sample immune cell receptor for the sample antigen.

[0045] In one possible embodiment, the training classification unit is used to fully connect and normalize the receptor features of the sample immune cell receptor through the affinity prediction model to obtain the probability that the sample immune cell receptor corresponds to multiple candidate affinity levels; based on the probability that the sample immune cell receptor corresponds to multiple candidate affinity levels, determine the predicted affinity level from the multiple candidate affinity levels.

[0046] On the one hand, a computer device is provided, comprising one or more processors and one or more memories, wherein at least one computer program is stored in the one or more memories, and the computer program is loaded and executed by the one or more processors to implement the affinity prediction method or the affinity prediction model training method.

[0047] On the one hand, a computer-readable storage medium is provided, wherein at least one computer program is stored in the computer-readable storage medium, and the computer program is loaded and executed by a processor to implement the affinity prediction method or the affinity prediction model training method.

[0048] On the one hand, a computer program product or a computer program is provided, which includes a program code, and the program code is stored in a computer-readable storage medium. A processor of a computer device reads the program code from the computer-readable storage medium, and the processor executes the program code, so that the computer device executes the above-mentioned affinity prediction method or affinity prediction model training method.

[0049] Through the technical solution provided in the embodiment of the present application, the affinity prediction model extracts features from the gene information and sequence of the immune cell receptor to obtain the gene features and sequence features of the immune cell receptor. In the process of obtaining the receptor features of the immune cell receptor, the gene features, sequence features and three-dimensional structural features are integrated. The introduction of three-dimensional structural features enriches the content of the receptor features and improves the expressiveness of the receptor features, so that when affinity prediction is performed based on the receptor features, the accuracy of the affinity between the obtained immune cell receptor and the target antigen is high. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0051] Figure 1 is a schematic diagram of an implementation environment of an affinity prediction method provided in an embodiment of the present application;

[0052] Figure 2 is a flow chart of an affinity prediction method provided in an embodiment of the present application;

[0053] Figure 3 is a flow chart of another affinity prediction method provided in an embodiment of the present application;

[0054] Figure 4 is a flow chart for determining three-dimensional structural features provided by an embodiment of the present application;

[0055] Figure 5 is a flow chart of another affinity prediction method provided in an embodiment of the present application;

[0056] Figure 6 It is a schematic diagram of an experimental result provided in an embodiment of the present application;

[0057] Figure 7 is a flow chart of a training method for an affinity prediction model provided in an embodiment of the present application;

[0058] Figure 8 is a schematic diagram of an affinity prediction interface provided in an embodiment of the present application;

[0059] Fig. 9 is a schematic diagram of an affinity display interface provided in an embodiment of the present application;

[0060] Fig.10 is a schematic structural diagram of an affinity prediction device provided in an embodiment of the present application;

[0061] Fig.11 It is a structural schematic diagram of a training device for an affinity prediction model provided in an embodiment of the present application;

[0062] Fig.12 is a schematic diagram of the structure of a terminal provided in an embodiment of the present application;

[0063] Fig.13 It is a structural diagram of a server provided in an embodiment of the present application. DETAILED DESCRIPTION

[0064] In order to make the objectives, technical solutions and advantages of the present application clearer, the implementation methods of the present application will be further described in detail below in conjunction with the accompanying drawings.

[0065] In this application, the terms "first", "second", etc. are used to distinguish identical or similar items with basically the same effects and functions. It should be understood that there is no logical or temporal dependency between "first", "second", and "nth", nor is there any limitation on quantity and execution order.

[0066] Artificial Intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technology in computer science that attempts to understand the essence of intelligence and produce a new intelligent machine that can respond in a similar way to human intelligence. Artificial intelligence is to study the design principles and implementation methods of various intelligent machines so that machines have the functions of perception, reasoning and decision-making.

[0067] Artificial intelligence technology is a comprehensive discipline that covers a wide range of fields, including both hardware-level and software-level technologies. The basic technologies of artificial intelligence generally include sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics and other technologies. Artificial intelligence software technology mainly includes computer vision technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0068] Machine Learning (ML) is a multi-disciplinary subject that involves probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory and other disciplines. It specializes in studying how computers simulate or implement human learning behavior to acquire new knowledge or skills and reorganize existing knowledge sub-models to continuously improve their performance. Machine learning is the core of artificial intelligence and the fundamental way to make computers intelligent. Its applications are spread across all areas of artificial intelligence. Machine learning and deep learning usually include artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and teaching learning.

[0069] Embedded Coding: Embedded coding mathematically represents a correspondence, that is, mapping data in X space to Y space through a function F, where function F is an injective function, and the mapping result is structural preservation. The injective function indicates that the data after mapping is uniquely corresponding to the data before mapping, and structural preservation indicates that the size relationship of the data before mapping is the same as the size relationship of the data after mapping. For example, before mapping, there are data X1 and X2, and after mapping, Y1 corresponding to X1 and Y2 corresponding to X2 are obtained. If the data X1 before mapping is greater than X2, then correspondingly, the data Y1 after mapping is greater than Y2. For words, it means mapping the words to another space to facilitate subsequent machine learning and processing.

[0070] Attention weight: It can indicate the importance of a certain data in the training or prediction process. Importance indicates the influence of input data on output data. Data with high importance has a higher corresponding attention weight value, and data with low importance has a lower corresponding attention weight value. In different scenarios, the importance of data is different. The process of training the attention weight of the model is also the process of determining the importance of data.

[0071] Immune cells: commonly known as white blood cells, including innate lymphocytes, various phagocytes, and lymphocytes that can recognize antigens and produce specific immune responses.

[0072] T cells: full name T lymphocytes (T-lymphocyte), derived from pluripotent stem cells in the bone marrow (or from the yolk sac and liver in the embryonic period). During the embryonic and neonatal period of the human body, a portion of pluripotent stem cells or pre-T cells in the bone marrow migrate to the thymus, differentiate and mature under the induction of thymic hormones, and become immune-active T cells.

[0073] TCR: T cell antigen receptor (T cell receptor, TCR) is a characteristic marker on the surface of all T cells. The function of TCR is to recognize antigens.

[0074] B cells: The full name is B lymphocytes, which are derived from pluripotent stem cells in the bone marrow. The progenitor cells of B lymphocytes exist in the hematopoietic cell islands of the fetal liver (14 days in embryonic mice or 8-9 weeks in infants), after which the production and differentiation sites of B lymphocytes are gradually replaced by the bone marrow. Mature B cells mainly reside in the lymph nodules in the superficial layer of the lymph node cortex and the lymph nodules in the red and white pulp of the spleen. B cells can differentiate into plasma cells under antigen stimulation. Plasma cells can synthesize and secrete antibodies (immunoglobulins), and mainly perform the body's humoral immunity.

[0075] BCR: B-cell antigen receptor (BCR) is a molecule located on the surface of B cells that is responsible for specific recognition and binding to antigens. Its essence is a membrane surface immunoglobulin. BCR has antigen binding specificity.

[0076] Antigen: refers to all substances that can stimulate the body to produce specific immune responses (humoral immunity and cellular immunity).

[0077] Cloud Technology refers to a hosting technology that unifies hardware, software, network and other resources within a wide area network or local area network to achieve data computing, storage, processing and sharing.

[0078] The technical solution provided in the embodiment of the present application can also be combined with cloud technology, for example, the trained affinity prediction model is deployed on a cloud server. Among them, the medical cloud in cloud technology refers to the use of "cloud computing" to create a medical and health service cloud platform based on new technologies such as cloud computing, mobile technology, multimedia, 4G communication, big data, and the Internet of Things, combined with medical technology, to achieve the sharing of medical resources and the expansion of medical scope.

[0079] It should be noted that the information (including but not limited to user device information, user personal information, etc.), data (including but not limited to data used for analysis, stored data, displayed data, etc.) and signals involved in this application are all authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards of relevant countries and regions. For example, the genetic information involved in this application is obtained with full authorization.

[0080] Figure 1 is a schematic diagram of an implementation environment of an affinity prediction method provided in an embodiment of the present application, see Figure 1 , the implementation environment may include a terminal 110 and a server 140.

[0081] The terminal 110 is connected to the server 140 via a wireless network or a wired network. Optionally, the terminal 110 is a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart watch, etc., but is not limited thereto. The terminal 110 has an application installed and running that supports affinity prediction.

[0082] Server 140 is an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, distribution networks (Content Delivery Network, CDN), as well as big data and artificial intelligence platforms.

[0083] Optionally, terminal 110 generally refers to one of multiple terminals, and the embodiment of the present application only takes terminal 110 as an example.

[0084] Those skilled in the art will appreciate that the number of the above terminals may be more or less. For example, there may be only one terminal, or there may be dozens or hundreds of terminals, or a greater number, and the above implementation environment may also include other terminals. The embodiment of the present application does not limit the number and device type of the terminals.

[0085] After introducing the implementation environment of the embodiment of the present application, the technical solution provided by the embodiment of the present application will be described in combination with the above-mentioned implementation environment. In the following description, the terminal is the terminal 110 in the above-mentioned implementation environment, and the server is the server 140 in the above-mentioned implementation environment.

[0086] The affinity prediction method provided in the embodiment of the present application can be applied in the fields of scientific research and vaccine design, that is, to determine the affinity scenario of the immune cell receptor and the specific antigen, wherein the specific antigen refers to an antigen that can specifically bind to the immune cell receptor. Through the technical solution provided in the embodiment of the present application, the technician uploads the gene information, sequence information and three-dimensional structural features of the immune cell receptor to the server through the terminal, and the server processes the gene information, sequence information and three-dimensional structural features of the immune cell receptor through the trained affinity prediction model to obtain the receptor features of the immune cell receptor, wherein the gene information of the immune cell receptor includes the VDJ information of the immune cell receptor, the sequence information is the amino acid sequence of the immune cell receptor, and the three-dimensional structural features are used to represent the three-dimensional structure of the immune cell receptor. The server performs affinity prediction based on the receptor features of the immune cell receptor through the affinity prediction model, and outputs the affinity between the immune cell receptor and the target antigen, which is also an antigen that can specifically bind to the immune cell receptor, and the technician can conduct further scientific research or vaccine design based on the affinity. The use of the technical solution provided in the embodiments of the present application can reduce the number of experiments conducted by technicians based on immune cell receptors and improve the efficiency of scientific research and vaccine design.

[0087] After introducing the implementation environment and application scenarios of the embodiment of the present application, the affinity prediction method provided by the embodiment of the present application is described below. The technical solution provided by the embodiment of the present application can be executed by a terminal or a server, or by a terminal and a server together. In the following description, the execution subject is taken as an example of the server, see Figure 2 , the method comprises the following steps.

[0088] 201. The server inputs the genetic information, sequence information and three-dimensional structural characteristics of the immune cell receptor into the affinity prediction model.

[0089] Wherein, the immune cell receptor is a T cell receptor or a B cell receptor. In some embodiments, the genetic information of the immune cell receptor includes the VDJ information of the immune cell receptor, wherein V encodes the variable region, D encodes the hypervariable region, and J encodes the cross-linking region. The sequence information of the immune cell receptor is the amino acid sequence of the immune cell receptor. The three-dimensional structural characteristics of the immune cell receptor are determined based on the three-dimensional structure of the immune cell receptor, wherein the three-dimensional structure is used to represent the positions of multiple amino acids in the immune cell receptor, and the three-dimensional structural characteristics can reflect the three-dimensional structure of the immune cell receptor as a whole. The affinity prediction model is a model trained based on the genetic information, sequence information and three-dimensional structural characteristics of the sample immune cell receptor, and has the function of predicting the affinity between the immune cell receptor and the target antigen.

[0090] 202. The server extracts features of the gene information and sequence information of the immune cell receptor through the affinity prediction model to obtain the gene features and sequence features of the immune cell receptor.

[0091] Among them, the process of extracting features from the genetic information and sequence information of the immune cell receptor is, that is, the process of abstractly expressing the genetic information and sequence information of the immune cell receptor. The obtained genetic features and sequence features can not only represent the genetic information and sequence information of the immune cell receptor, but also facilitate subsequent processing by the server.

[0092] 203. The server fuses the gene features, sequence features, and three-dimensional structural features of the immune cell receptor through the affinity prediction model to obtain the receptor features of the immune cell receptor.

[0093] Among them, the receptor characteristics of the immune cell receptor are obtained by integrating gene characteristics, sequence characteristics and three-dimensional structural characteristics, which means that the immune cell receptor can be represented from three aspects: gene, sequence and structure. Therefore, the receptor characteristics have a strong expression ability.

[0094] 204. The server performs multiple full connections on the receptor features of the immune cell receptor through the affinity prediction model, and outputs the affinity of the immune cell receptor for the target antigen, where the target antigen is an antigen that can specifically bind to the immune cell receptor.

[0095] Among them, the process of multiple full connections of the receptor characteristics of the immune cell receptor is the process of affinity prediction based on the receptor characteristics of the immune cell receptor, that is, the process of regression based on the receptor characteristics to obtain the affinity between the immune cell receptor and the target antigen.

[0096] Through the technical solution provided in the embodiment of the present application, the affinity prediction model extracts features from the gene information and sequence of the immune cell receptor to obtain the gene features and sequence features of the immune cell receptor. In the process of obtaining the receptor features of the immune cell receptor, the gene features, sequence features and three-dimensional structural features are integrated. The introduction of three-dimensional structural features enriches the content of the receptor features and improves the expressiveness of the receptor features, so that when affinity prediction is performed based on the receptor features, the accuracy of the affinity between the obtained immune cell receptor and the target antigen is high.

[0097] The above steps 201-204 are a brief description of the affinity prediction method provided in the embodiment of the present application. The affinity prediction method provided in the embodiment of the present application will be further described below with reference to some examples. Figure 3 Taking the execution subject as a server as an example, the method includes the following steps.

[0098] 301. The server obtains the three-dimensional structural characteristics of the immune cell receptor.

[0099] Among them, the immune cell receptor is a T cell receptor or a B cell receptor, which is used to recognize antigens and specifically bind to antigens, thereby activating the immune system. The immune cell receptor is a protein, which includes multiple amino acids. The three-dimensional structural characteristics of the immune cell receptor are used to represent the positions of multiple amino acids of the immune cell receptor in space.

[0100] In a possible embodiment, the server obtains the target amino acid sequence of the immune cell receptor, and the target amino acid sequence includes the CDR3 region of the immune cell receptor. The server performs a multiple sequence alignment on the target amino acid sequence of the immune cell receptor to obtain at least one reference amino acid sequence, and the similarity between the reference amino acid sequence and the target amino acid sequence meets the similarity condition. The server obtains a homologous template corresponding to the target amino acid sequence, and the homologous template includes structural information of the homologous sequence of the target amino acid sequence. The server performs multiple rounds of iterations based on the target amino acid sequence, the at least one reference amino acid sequence, and the homologous template to obtain the three-dimensional structural characteristics of the immune cell receptor.

[0101] Among them, there is a complementary determining region (CDR) on the immune cell receptor, which includes three sub-regions CDR1, CDR2 and CDR3. Among them, CDR3 is the most variable and plays a key role in antigen recognition.

[0102] Under this embodiment, the server can determine the three-dimensional structural characteristics of the immune cell receptor based on the target amino acid sequence of the immune cell receptor without the need for observation through other equipment such as a cryo-electron microscope, thereby improving the efficiency of acquiring the three-dimensional structural characteristics and reducing the cost of acquiring the three-dimensional structural characteristics.

[0103] For example, the server obtains the sequencing data of the immune cell receptor, and the sequencing data includes multiple amino acids of the immune cell receptor and the arrangement order of the multiple amino acids. The sequencing data is obtained by the technician through the gene sequencing equipment test, and the embodiment of the present application does not limit this. The server preprocesses the sequencing data of the immune cell receptor (Data Preprocessing) to obtain the reference sequencing data of the immune cell receptor, wherein the preprocessing of the sequencing data includes eliminating the error data in the sequencing data and converting the sequencing data into a format that is convenient for the server to process, etc. The preprocessing rules are set by the technician according to the actual situation, and the embodiment of the present application does not limit this. The server performs quality control (Quality Control) on the reference sequencing data to obtain the target sequencing data of the immune cell receptor, wherein the quality control of the reference sequencing data includes filtering out dead cells, background estimation, paired chains, signal correction, Log-rank test, and receptor gene aggregation, etc. The server intercepts the amino acid sequence containing the CDR3 region of the target length from the target sequencing data, and the amino acid sequence containing the CDR3 region of the target length is also the target amino acid sequence, wherein the target length is set by the technician according to the actual situation, such as being set to be greater than 50 amino acids, etc., and the embodiment of the present application does not limit this. The server searches in the gene database based on the target amino acid sequence to obtain at least one reference amino acid sequence, and the at least one reference amino acid sequence is also an amino acid sequence whose similarity with the target amino acid sequence is greater than or equal to the similarity threshold. The similarity between the amino acid sequences is determined by comparing the type and arrangement order of the amino acids in the amino acid sequence. Multiple sequence alignment is also called multiple sequence alignment, which is used to extract and input amino acid sequences from a large database. Similar sequences, and alignment is performed by the way. Since amino acid sequences with similar sequences generally have similar folding modes, multiple sequence alignment can add similar sequence structure information to the features. The server searches in the structure database based on the target amino acid sequence to obtain a homologous template corresponding to the target amino acid sequence, and the homologous template includes the structural information of the homologous sequence of the target amino acid sequence. Based on the attention mechanism, the server performs multiple rounds of iterative encoding on the target amino acid sequence, the at least one reference amino acid sequence and the homologous template to obtain the distance distribution between each pair of amino acids in the target amino acid sequence and the angle of the chemical bond connecting them.The server uses the attention mechanism to encode the distance distribution between each pair of amino acids in the target amino acid sequence and the angle of the chemical bond connecting them, and outputs the three-dimensional structural information of the immune cell receptor, wherein the three-dimensional structural information of the immune cell receptor includes the three-dimensional positions of multiple amino acids in the immune cell receptor. The server extracts features of the three-dimensional structure of the immune cell receptor, such as processing the immune cell receptor using a graph network to obtain the three-dimensional structural features of the immune cell receptor.

[0104] In order to explain the above-mentioned embodiment more clearly, Figure 4 The above-mentioned embodiment is described.

[0105] See also Figure 4 , the server preprocesses the sequencing data of the immune cell receptor 401 to obtain the reference sequencing data of the immune cell receptor. The server performs quality control 402 on the reference sequencing data to obtain the target sequencing data of the immune cell receptor, wherein the quality control 402 includes dead cell removal 4021, background estimation 4022, chain pairing 4023, signal correction 4024, Log-rank test 4025 and receptor gene aggregation 4026. The server performs sequence interception 403 on the target sequencing data to obtain the target amino acid sequence. The server performs multiple sequence alignment 404 based on the target amino acid sequence to obtain at least one reference amino acid sequence. The server searches in a structural database based on the target amino acid sequence to obtain a homologous template corresponding to the target amino acid sequence. Based on the attention mechanism, the server performs multiple rounds of iterative encoding 405 on the target amino acid sequence, the at least one reference amino acid sequence and the homologous template to obtain the three-dimensional structural information of the immune cell receptor.

[0106] The above implementation is a method for the server to determine the three-dimensional structural characteristics of the immune cell receptor based on the target amino acid sequence of the immune cell receptor. In other possible implementations, the server can use the trained structure prediction model to obtain the three-dimensional structural characteristics based on the amino acid sequence, wherein the structure prediction model includes RoseTTAFold, AlphaFold and AlphaFold2 models. Of course, with the development of science and technology, other structure prediction models can also be used, and the embodiments of the present application are not limited to this.

[0107] The following describes a method for a server to obtain the three-dimensional structural features of the immune cell receptor based on the three-dimensional structural information of the immune cell receptor, wherein the three-dimensional structural information includes the three-dimensional positions of multiple amino acids in the immune cell receptor.

[0108] In a possible embodiment, the server obtains the three-dimensional structural information of the immune cell receptor, the three-dimensional structural information including the three-dimensional coordinates of multiple amino acids in the immune cell receptor, and performs graph convolution on the three-dimensional structural information of the immune cell receptor to obtain the three-dimensional structural features of the immune cell receptor.

[0109] Among them, the three-dimensional structural information is a three-dimensional structural file of the immune cell receptor. In some embodiments, the three-dimensional structural information is obtained by images taken by a cryo-electron microscope, or is obtained by a structural prediction model based on the amino acid sequence of the immune cell receptor, and the embodiments of the present application are not limited to this. The full name of graph convolution is graph convolutional neural network (Graph Convolutional Network, GCN), which is used to extract the features of a graph (Graph). In the embodiments of the present application, the nodes in the graph are the amino acids in the immune cell receptor, and the lines in the graph are used to represent the relative position relationship between the amino acids.

[0110] In this implementation, the server can obtain the three-dimensional structural features of the immune cell receptor by directly performing graph convolution on the three-dimensional structural information of the immune cell receptor, without first determining the three-dimensional structural information of the immune cell receptor, and the efficiency of determining the three-dimensional structural features is high.

[0111] For example, the server obtains the three-dimensional structural information of the immune cell receptor. The server generates a three-dimensional structural graph of the immune cell receptor based on the three-dimensional structural information, wherein the nodes in the three-dimensional structural graph correspond to the amino acids of the immune cell receptor, and the lines in the three-dimensional structural graph are used to represent the connection relationship between the amino acids. The node features of the nodes in the three-dimensional structural graph include the type of the corresponding amino acid and the three-dimensional coordinates. The server performs graph convolution on the three-dimensional structural graph to obtain the three-dimensional structural features of the immune cell receptor.

[0112] In a possible embodiment, the server obtains the three-dimensional structural information of the immune cell receptor, and the three-dimensional structural information includes the three-dimensional coordinates of multiple amino acids in the immune cell receptor. The server encodes the three-dimensional structural information of the immune cell receptor based on the attention mechanism to obtain the three-dimensional structural features of the immune cell receptor.

[0113] In this implementation, the server can obtain the three-dimensional structural features of the immune cell receptor by directly encoding the three-dimensional structural information of the immune cell receptor based on the attention mechanism, without first determining the three-dimensional structural information of the immune cell receptor, and the efficiency of determining the three-dimensional structural features is high.

[0114] For example, the server obtains the three-dimensional structural information of the immune cell receptor. The server embeds and encodes multiple amino acids in the three-dimensional structural information to obtain multiple amino acid embedding features, wherein the process of embedding and encoding multiple amino acids is to represent multiple amino acids in a discretized form, which is convenient for subsequent processing by the server. The server uses an attention mechanism to encode the multiple amino acid embedding features based on the three-dimensional structural information to obtain the attention weights of multiple amino acids. Based on the attention weights of the multiple amino acids, the server fuses the multiple amino acid embedding features to obtain the three-dimensional structural features of the immune cell receptor. In some embodiments, the server can use the encoder of the Transformer model to encode the three-dimensional structural information of the immune cell receptor to obtain the three-dimensional structural features of the immune cell receptor.

[0115] It should be noted that the above two implementations are described by taking the server using graph convolution and attention mechanism to encode the three-dimensional structural information of the immune cell receptor to obtain three-dimensional structural features as an example. In other possible implementations, the server can also use other models to encode the three-dimensional structural information of the immune cell receptor, which is not limited to the embodiments of the present application.

[0116] It should be noted that the above step 301 is an optional step.

[0117] 302. The server inputs the gene information, sequence information and three-dimensional structural characteristics of the immune cell receptor into the affinity prediction model.

[0118] Among them, the genetic information of the immune cell receptor includes the VDJ information of the immune cell receptor, wherein V is for encoding the variable region, D is for encoding the hypervariable region, and J is for encoding the cross-linking region. The sequence information of the immune cell receptor is the amino acid sequence of the immune cell receptor. For example, AEGAL is an amino acid sequence, wherein A represents alanine, E represents glutamic acid, G represents glycine, and L represents leucine. The immune cell receptor is a protein, and the amino acid sequence is also called the one-dimensional structure of the protein. The affinity prediction model is a model trained based on the genetic information, sequence information, and three-dimensional structural characteristics of the sample immune cell receptor, and has the function of predicting the corresponding antigen of the immune cell receptor.

[0119] In a possible embodiment, the affinity prediction model includes three information encoding channels, wherein the first information encoding channel is a gene information encoding channel, the gene information encoding channel includes a gene encoder, and the gene encoder is used to encode gene information; the second information encoding channel is a sequence information encoding channel, the sequence information encoding channel includes a sequence encoder, and the sequence encoder is used to encode sequence information; the third information encoding channel is a structural feature encoding channel, the structural feature encoding channel includes a structural encoder, and the structural encoder is used to encode structural features. The server inputs the gene information of the immune cell receptor into the gene information encoding channel of the antigen prediction model, and subsequently encodes the gene information through the gene encoder in the gene information encoding channel. The server inputs the sequence information of the immune cell receptor into the sequence information encoding channel of the antigen prediction model, and subsequently encodes the sequence information through the sequence encoder in the sequence information encoding channel. The server inputs the three-dimensional structural features of the immune cell receptor into the structural feature encoding channel, and subsequently encodes the three-dimensional structural features through the structural encoder in the structural feature encoding channel.

[0120] In some embodiments, before the sequence information of the immune cell receptor is input into the affinity prediction model, the server can also pre-process the sequence information of the immune cell receptor to ensure that the length of the sequence information input into the affinity prediction model is the same. In the case where the length of the sequence information of the immune cell receptor is greater than the length threshold, the server truncates the portion of the sequence information of the immune cell receptor whose length is greater than or equal to the length threshold to obtain sequence information with a length of the length threshold, and subsequently inputs the truncated sequence information into the affinity prediction model. In the case where the length of the sequence information of the immune cell receptor is less than the length threshold, the server fills the target symbol in the sequence information of the immune cell receptor to obtain sequence information with a length of the length threshold, and subsequently inputs the truncated sequence information into the affinity prediction model, wherein the target symbol is set by the technician according to the actual situation, such as 0.

[0121] It should be noted that the above steps 301-302 are described by taking the example of the server obtaining the three-dimensional structural characteristics of the immune cell receptor in advance. In other possible implementations, the server may also obtain the three-dimensional structural information of the immune cell receptor in advance, input the three-dimensional structural information into the structural characteristic encoding channel of the affinity prediction model, and subsequently obtain the three-dimensional structural characteristics of the immune cell receptor through the structural encoder of the structural characteristic encoding channel. The embodiment of the present application does not limit this.

[0122] In addition, the above steps 301-302 are explained by taking the example of the server obtaining the three-dimensional structural characteristics of the immune cell receptor and inputting the genetic information, sequence information and three-dimensional structural characteristics of the immune cell receptor into the affinity prediction model. In other possible embodiments, when the server does not obtain the three-dimensional structural characteristics of the immune cell receptor, only the genetic information and sequence information of the immune cell receptor may be input into the affinity prediction model.

[0123] 303. The server extracts features of the gene information and sequence information of the immune cell receptor through the affinity prediction model to obtain the gene features and sequence features of the immune cell receptor.

[0124] Among them, the process of extracting features from the genetic information and sequence information of the immune cell receptor is, that is, the process of abstractly expressing the genetic information and sequence information of the immune cell receptor. The obtained genetic features and sequence features can not only represent the genetic information and sequence information of the immune cell receptor, but also facilitate subsequent processing by the server.

[0125] In a possible embodiment, the affinity prediction model includes a gene encoder and a sequence encoder. The server encodes the VDJ information of the immune cell receptor through the gene encoder of the affinity prediction model to obtain the gene characteristics of the immune cell receptor, wherein V is the encoding variable region, D is the encoding hypervariable region, and J is the encoding cross-linking region. The server encodes the amino acid sequence of the immune cell receptor through the sequence encoder of the affinity prediction model to obtain the sequence characteristics of the immune cell receptor.

[0126] In this implementation, the server can encode the gene information and sequence information of the immune cell receptor respectively through the gene encoder and sequence encoder of the affinity prediction model, that is, perform feature extraction on the gene information and sequence information, and the obtained gene features and sequence features can represent the immune cell receptor from different dimensions.

[0127] In order to explain the above implementation more clearly, the above implementation will be explained in two parts below.

[0128] In the first part, the server encodes the VDJ information of the immune cell receptor through the gene encoder of the affinity prediction model to obtain the gene characteristics of the immune cell receptor.

[0129] In a possible embodiment, when the immune cell receptor is a B cell receptor, the server encodes the VJ information of the light chain and the VDJ information of the heavy chain of the immune cell receptor through the gene encoder of the affinity prediction model to obtain the gene characteristics of the immune cell receptor.

[0130] Among them, the B cell receptor includes two identical heavy chains (Heavy Chain, H chain) and two identical light chains (Light Chain, L chain), which are connected by interchain disulfide bonds to form a tetrapeptide chain structure. The molecular weight of the heavy chain is about 50-75kD, and it is composed of 450-550 amino acid residues. The molecular weight of the light chain is about 25kD, and it is composed of 214 amino acid residues.

[0131] In order to explain the above embodiment more clearly, the above embodiment will be explained through three examples below.

[0132] Example 1: The server fully connects the VJ information of the light chain and the VDJ information of the heavy chain of the immune cell receptor through the gene encoder of the affinity prediction model to obtain the gene characteristics of the immune cell receptor, which include the gene characteristics of the light chain of the immune cell receptor and the gene characteristics of the heavy chain of the immune cell receptor.

[0133] In a possible embodiment, the affinity prediction model includes two gene encoders, and the server splices the VJ information of the light chain of the B cell receptor through the first gene encoder of the affinity prediction model to obtain the light chain gene information of the B cell receptor. The server splices the VDJ information of the light chain of the B cell receptor through the second gene encoder of the affinity prediction model to obtain the heavy chain gene information of the B cell receptor. The server fully connects the light chain gene information of the B cell receptor twice through the first gene encoder of the affinity prediction model to obtain the light chain gene feature of the B cell receptor. The server fully connects the heavy chain gene information of the B cell receptor twice through the second gene encoder of the affinity prediction model to obtain the heavy chain gene feature of the B cell receptor. The light chain gene feature and the heavy chain gene feature of the B cell receptor constitute the gene feature of the B cell receptor.

[0134] Example 2: The server convolves the VJ information of the light chain and the VDJ information of the heavy chain of the immune cell receptor through the gene encoder of the affinity prediction model to obtain the gene characteristics of the immune cell receptor, and the gene characteristics of the immune cell receptor include the light chain gene characteristics of the immune cell receptor and the heavy chain gene characteristics of the immune cell receptor.

[0135] In a possible embodiment, the affinity prediction model includes two gene encoders, and the server splices the VJ information of the light chain of the B cell receptor through the first gene encoder of the affinity prediction model to obtain the light chain gene information of the B cell receptor. The server splices the VDJ information of the light chain of the B cell receptor through the second gene encoder of the affinity prediction model to obtain the heavy chain gene information of the B cell receptor. The server convolves the light chain gene information of the B cell receptor twice through the first gene encoder of the affinity prediction model to obtain the light chain gene feature of the B cell receptor. The server convolves the heavy chain gene information of the B cell receptor twice through the second gene encoder of the affinity prediction model to obtain the heavy chain gene feature of the B cell receptor. The light chain gene feature and the heavy chain gene feature of the B cell constitute the gene feature of the B cell receptor.

[0136] Example 3: The server encodes the VJ information of the light chain and the VDJ information of the heavy chain of the immune cell receptor through the gene encoder of the affinity prediction model based on the attention mechanism to obtain the gene characteristics of the immune cell receptor, and the gene characteristics of the immune cell receptor include the light chain gene characteristics of the immune cell receptor and the heavy chain gene characteristics of the immune cell receptor.

[0137] In a possible embodiment, the affinity prediction model includes two gene encoders, and the server splices the VJ information of the light chain of the B cell receptor through the first gene encoder of the affinity prediction model to obtain the light chain gene information of the B cell receptor. The server splices the VDJ information of the light chain of the B cell receptor through the second gene encoder of the affinity prediction model to obtain the heavy chain gene information of the B cell receptor. The server encodes the light chain gene information of the B cell receptor based on the attention mechanism through the first gene encoder of the affinity prediction model to obtain the light chain gene feature of the B cell receptor. The server encodes the heavy chain gene information of the B cell receptor based on the attention mechanism through the second gene encoder of the affinity prediction model to obtain the heavy chain gene feature of the B cell receptor. The light chain gene feature and heavy chain gene feature of the B cell receptor constitute the gene feature of the B cell receptor.

[0138] The above description is made by taking the immune cell receptor as a B cell receptor as an example, and the following description is made by taking the immune cell receptor as a T cell receptor as an example.

[0139] In a possible embodiment, when the immune cell receptor is a T cell receptor, the server encodes the VJ information of the α chain and the VDJ information of the β chain of the immune cell receptor through the gene encoder of the affinity prediction model to obtain the gene characteristics of the immune cell receptor.

[0140] Among them, some T cell receptors include α chains and β chains, and such T cell receptors are also called αβ-TCRs. Other T cell receptors include γ chains and δ chains, and such T cell receptors are also called γδ-TCRs. Since the number of αβ-TCRs in the human body is far greater than the number of γδ-TCRs, the T cell receptor is αβ-TCR as an example in the following description. For γδ-TCR, its structure is similar to that of αβ-TCR, both of which are double-chain structures, and the processing method belongs to the same inventive concept. The implementation process is described below.

[0141] In order to explain the above embodiment more clearly, the above embodiment will be explained through three examples below.

[0142] Example 1: The server fully connects the VJ information of the α chain and the VDJ information of the β chain of the immune cell receptor through the gene encoder of the affinity prediction model to obtain the gene characteristics of the immune cell receptor, and the gene characteristics of the immune cell receptor include the gene characteristics of the α chain of the immune cell receptor and the gene characteristics of the β chain of the immune cell receptor.

[0143] In a possible embodiment, the affinity prediction model includes two gene encoders, and the server splices the VJ information of the α chain of the T cell receptor through the first gene encoder of the affinity prediction model to obtain the α chain gene information of the T cell receptor. The server splices the VDJ information of the α chain of the T cell receptor through the second gene encoder of the affinity prediction model to obtain the β chain gene information of the T cell receptor. The server fully connects the α chain gene information of the T cell receptor twice through the first gene encoder of the affinity prediction model to obtain the α chain gene feature of the T cell receptor. The server fully connects the β chain gene information of the T cell receptor twice through the second gene encoder of the affinity prediction model to obtain the β chain gene feature of the T cell receptor. The α chain gene feature and the β chain gene feature of the T cell receptor constitute the gene feature of the T cell receptor.

[0144] Example 2: The server convolves the VJ information of the α chain and the VDJ information of the β chain of the immune cell receptor through the gene encoder of the affinity prediction model to obtain the gene characteristics of the immune cell receptor, and the gene characteristics of the immune cell receptor include the gene characteristics of the α chain of the immune cell receptor and the gene characteristics of the β chain of the immune cell receptor.

[0145] In a possible embodiment, the affinity prediction model includes two gene encoders, and the server splices the VJ information of the α chain of the T cell receptor through the first gene encoder of the affinity prediction model to obtain the α chain gene information of the T cell receptor. The server splices the VDJ information of the α chain of the T cell receptor through the second gene encoder of the affinity prediction model to obtain the β chain gene information of the T cell receptor. The server convolves the α chain gene information of the T cell receptor twice through the first gene encoder of the affinity prediction model to obtain the α chain gene feature of the T cell receptor. The server convolves the β chain gene information of the T cell receptor twice through the second gene encoder of the affinity prediction model to obtain the β chain gene feature of the T cell receptor. The α chain gene feature and the β chain gene feature of the T cell receptor constitute the gene feature of the T cell receptor.

[0146] Example 3: The server encodes the VJ information of the α chain and the VDJ information of the β chain of the immune cell receptor through the gene encoder of the affinity prediction model based on the attention mechanism to obtain the gene characteristics of the immune cell receptor. The gene characteristics of the immune cell receptor include the gene characteristics of the α chain of the immune cell receptor and the gene characteristics of the β chain of the immune cell receptor.

[0147] In a possible embodiment, the affinity prediction model includes two gene encoders, and the server splices the VJ information of the α chain of the T cell receptor through the first gene encoder of the affinity prediction model to obtain the α chain gene information of the T cell receptor. The server splices the VDJ information of the α chain of the T cell receptor through the second gene encoder of the affinity prediction model to obtain the β chain gene information of the T cell receptor. The server encodes the α chain gene information of the T cell receptor based on the attention mechanism through the first gene encoder of the affinity prediction model to obtain the α chain gene feature of the T cell receptor. The server encodes the β chain gene information of the T cell receptor based on the attention mechanism through the second gene encoder of the affinity prediction model to obtain the β chain gene feature of the T cell receptor. The α chain gene feature and the β chain gene feature of the T cell receptor constitute the gene feature of the T cell receptor.

[0148] In the second part, the server encodes the amino acid sequence of the immune cell receptor through the sequence encoder of the affinity prediction model to obtain the sequence characteristics of the immune cell receptor.

[0149] In a possible embodiment, when the immune cell receptor is a B cell receptor, the server encodes the amino acid sequence of the light chain and the amino acid sequence of the heavy chain of the immune cell receptor through the sequence encoder of the affinity prediction model based on the attention mechanism to obtain the sequence features of the immune cell receptor, and the sequence features of the immune cell receptor include the sequence features of the light chain and the heavy chain of the immune cell receptor. In some embodiments, the sequence encoder is an encoder of a Transformer model.

[0150] For example, the affinity prediction model includes two sequence encoders. When the immune cell receptor is a B cell receptor, the server embeds the amino acid sequence of the light chain of the B cell receptor through the first sequence encoder of the affinity prediction model to obtain the light chain embedding feature of the B cell receptor, and one light chain embedding feature corresponds to one amino acid on the light chain. The server encodes multiple light chain embedding features based on the order of multiple amino acids in the amino acid sequence of the B cell receptor through the first sequence encoder to obtain the attention weight corresponding to each light chain embedding feature. The server weightedly fuses multiple light chain embedding features based on the attention weight corresponding to each light chain embedding feature through the first sequence encoder to obtain the light chain sequence feature of the B cell receptor. The server embeds the amino acid sequence of the heavy chain of the B cell receptor through the second sequence encoder of the affinity prediction model to obtain the heavy chain embedding feature of the B cell receptor, and one heavy chain embedding feature corresponds to one amino acid on the heavy chain. The server encodes multiple heavy chain embedding features based on the order of multiple amino acids in the amino acid sequence of the B cell receptor through the second sequence encoder to obtain the attention weight corresponding to each heavy chain embedding feature. The server uses the second sequence encoder to weightedly fuse multiple heavy chain embedded features based on the attention weights corresponding to each heavy chain embedded feature to obtain the heavy chain sequence feature of the B cell receptor. The light chain sequence feature of the B cell receptor and the heavy chain sequence feature of the B cell receptor constitute the sequence feature of the B cell receptor. In some embodiments, embedded coding can adopt one-hot mode or other modes, which are not limited in the embodiments of the present application.

[0151] In a possible embodiment, when the immune cell receptor is a T cell receptor, the server encodes the amino acid sequence of the α chain and the amino acid sequence of the β chain of the immune cell receptor through the sequence encoder of the affinity prediction model based on the attention mechanism to obtain the sequence characteristics of the immune cell receptor, and the sequence characteristics of the immune cell receptor include the α chain sequence characteristics and the β chain sequence characteristics of the immune cell receptor.

[0152] For example, the affinity prediction model includes two sequence encoders. When the immune cell receptor is a T cell receptor, the server embeds the amino acid sequence of the α chain of the T cell receptor through the first sequence encoder of the affinity prediction model to obtain the α chain embedding feature of the T cell receptor, and one α chain embedding feature corresponds to one amino acid on the α chain. The server encodes multiple α chain embedding features based on the order of multiple amino acids in the amino acid sequence of the T cell receptor through the first sequence encoder to obtain the attention weight corresponding to each α chain embedding feature. The server weightedly fuses multiple α chain embedding features based on the attention weight corresponding to each α chain embedding feature through the first sequence encoder to obtain the α chain sequence feature of the T cell receptor. The server embeds the amino acid sequence of the β chain of the T cell receptor through the second sequence encoder of the affinity prediction model to obtain the β chain embedding feature of the T cell receptor, and one β chain embedding feature corresponds to one amino acid on the β chain. The server encodes multiple β chain embedding features based on the order of multiple amino acids in the amino acid sequence of the T cell receptor through the second sequence encoder to obtain the attention weight corresponding to each β chain embedding feature. The server uses the second sequence encoder to weightedly fuse multiple β-chain embedding features based on the attention weights corresponding to each β-chain embedding feature to obtain the β-chain sequence feature of the T cell receptor. The light chain sequence feature of the T cell receptor and the heavy chain sequence feature of the T cell receptor constitute the sequence feature of the T cell receptor.

[0153] 304. The server fuses the gene features, sequence features, and three-dimensional structural features of the immune cell receptor through the affinity prediction model to obtain the receptor features of the immune cell receptor.

[0154] Among them, the receptor characteristics of the immune cell receptor are obtained by fusing gene characteristics, sequence characteristics and three-dimensional structural characteristics, which means that the immune cell receptor can be represented from three aspects: gene, sequence and structure. The receptor characteristics can represent the immune cell receptor more completely.

[0155] In a possible implementation, the server splices the gene features and sequence features of the immune cell receptor through the feature fusion module of the affinity prediction model to obtain the gene sequence fusion features of the immune cell receptor. The server performs weighted fusion of the gene sequence fusion features and three-dimensional structural features of the immune cell receptor through the feature fusion module of the affinity prediction model based on the gated attention mechanism to obtain the receptor features of the immune cell receptor.

[0156] In this implementation, the server can first fuse the gene features and sequence features of the immune cell receptor through the feature fusion module to obtain the gene sequence fusion features of the immune cell receptor. The server then uses the gated attention mechanism to fuse the sequence fusion features and the three-dimensional structure features to finally obtain the receptor features of the immune cell receptor. The introduction of the gated attention mechanism enables the model to pay more attention to content with higher importance. Through the feature fusion method provided by the above implementation, the gene features, sequence features and three-dimensional structure features can be organically combined, and the obtained receptor features have stronger expression capabilities.

[0157] In the case where the immune cell receptor is a B cell receptor, the gene feature of the B cell receptor includes the light chain gene feature and the heavy chain gene feature of the B cell receptor, and the sequence feature of the B cell receptor includes the light chain sequence feature of the B cell receptor and the heavy chain sequence feature of the B cell receptor. The server adds the light chain gene feature of the B cell receptor and the light chain sequence feature of the B cell receptor through the feature fusion module to obtain the light chain gene sequence feature of the B cell receptor. The server adds the heavy chain gene feature of the B cell receptor and the heavy chain sequence feature of the B cell receptor through the feature fusion module to obtain the heavy chain gene sequence feature of the B cell receptor. The server splices the light chain gene sequence feature and the heavy chain gene sequence feature of the B cell receptor through the feature fusion module to obtain the gene sequence fusion feature of the B cell receptor. The server encodes the gene sequence fusion feature and the three-dimensional structure feature of the B cell receptor using the attention mechanism through the feature fusion module, and obtains the first attention weight of the gene sequence fusion feature encoding the three-dimensional structure feature and the second attention weight of the three-dimensional structure feature encoding the gene sequence fusion feature. The server processes the first attention weight and the second attention weight using a gating function through the feature fusion module to obtain a first gating weight and a second gating weight, and the first gating weight and the second gating weight are used to control the flow of information during feature fusion. The server uses the first gating weight to weightedly fuse the gene sequence fusion feature and the three-dimensional structure feature of the B cell receptor through the feature fusion module to obtain the target gene sequence fusion feature of the B cell receptor. In some embodiments, the first gating weight is multiplied by the three-dimensional structure feature and then added to the gene sequence fusion feature to obtain the target gene sequence fusion feature. The server uses the second gating weight to weightedly fuse the gene sequence fusion feature and the three-dimensional structure feature of the B cell receptor through the feature fusion module to obtain the target three-dimensional structure feature of the B cell receptor. In some embodiments, the second gating weight is multiplied by the gene sequence fusion feature and then added to the three-dimensional structure feature to obtain the target three-dimensional structure feature. The server uses the feature fusion module to perform tensor fusion of the target gene sequence fusion feature and the target three-dimensional structure feature, such as multiplying the target gene sequence fusion feature and the target three-dimensional structure to obtain the initial receptor feature of the B cell receptor. The server uses the feature fusion module to perform at least two full connections on the initial receptor feature of the B cell receptor to obtain the receptor feature of the B cell receptor.

[0158] In the case where the immune cell receptor is a T cell receptor, the gene feature of the T cell receptor includes the α chain gene feature and the β chain gene feature of the T cell receptor, and the sequence feature of the T cell receptor includes the α chain sequence feature of the T cell receptor and the β chain sequence feature of the T cell receptor. The server adds the α chain gene feature of the T cell receptor and the α chain sequence feature of the T cell receptor through the feature fusion module to obtain the α chain gene sequence feature of the T cell receptor. The server adds the β chain gene feature of the T cell receptor and the β chain sequence feature of the T cell receptor through the feature fusion module to obtain the β chain gene sequence feature of the T cell receptor. The server splices the α chain gene sequence feature and the β chain gene sequence feature of the T cell receptor through the feature fusion module to obtain the gene sequence fusion feature of the T cell receptor. The server encodes the gene sequence fusion feature and the three-dimensional structure feature of the T cell receptor using the attention mechanism through the feature fusion module, and obtains the third attention weight of the gene sequence fusion feature encoding the three-dimensional structure feature and the fourth attention weight of the three-dimensional structure feature encoding the gene sequence fusion feature. The server processes the third attention weight and the fourth attention weight by the feature fusion module using a gating function to obtain a third gating weight and a fourth gating weight, and the third gating weight and the fourth gating weight are used to control the flow of information during feature fusion. The server uses the third gating weight to weightedly fuse the gene sequence fusion feature and the three-dimensional structure feature of the T cell receptor through the feature fusion module to obtain the target gene sequence fusion feature of the T cell receptor. In some embodiments, the third gating weight is multiplied by the three-dimensional structure feature and then added to the gene sequence fusion feature to obtain the target gene sequence fusion feature. The server uses the fourth gating weight to weightedly fuse the gene sequence fusion feature and the three-dimensional structure feature of the T cell receptor through the feature fusion module to obtain the target three-dimensional structure feature of the T cell receptor. In some embodiments, the fourth gating weight is multiplied by the gene sequence fusion feature and then added to the three-dimensional structure feature to obtain the target three-dimensional structure feature. The server uses the feature fusion module to perform tensor fusion of the target gene sequence fusion feature and the target three-dimensional structure feature, such as multiplying the target gene sequence fusion feature and the target three-dimensional structure to obtain the initial receptor feature of the T cell receptor. The server uses the feature fusion Moaqua to perform at least two full connections on the initial receptor feature of the T cell receptor to obtain the receptor feature of the T cell receptor.

[0159] In a possible embodiment, the server adds the gene feature and sequence feature of the immune cell receptor through the feature fusion module of the affinity prediction model to obtain the gene sequence fusion feature of the immune cell receptor. The server splices and fully connects the sequence feature and three-dimensional structure feature of the immune cell receptor at least once through the feature fusion module to obtain the receptor feature of the immune cell receptor.

[0160] In this implementation, the server uses the feature fusion module to quickly fuse the gene features, sequence features and three-dimensional structural features of the immune cell receptor by addition, splicing and full connection, thereby obtaining the receptor features of the immune cell receptor with high efficiency.

[0161] In the case where the immune cell receptor is a B cell receptor, the gene feature of the B cell receptor includes the light chain gene feature and the heavy chain gene feature of the B cell receptor, and the sequence feature of the B cell receptor includes the light chain sequence feature of the B cell receptor and the heavy chain sequence feature of the B cell receptor. The server adds the light chain gene feature of the B cell receptor and the light chain sequence feature of the B cell receptor through the feature fusion module to obtain the light chain gene sequence feature of the B cell receptor. The server adds the heavy chain gene feature of the B cell receptor and the heavy chain sequence feature of the B cell receptor through the feature fusion module to obtain the heavy chain gene sequence feature of the B cell receptor. The light chain gene sequence feature and the heavy chain gene sequence feature of the B cell receptor constitute the gene sequence fusion feature of the B cell receptor. The server splices the gene sequence fusion feature and the three-dimensional structure feature of the B cell receptor through the feature fusion module to obtain the initial receptor feature of the B cell receptor. The server performs at least one full connection on the initial receptor feature of the B cell receptor through the feature fusion module to obtain the receptor feature of the B cell receptor.

[0162] In the case where the immune cell receptor is a T cell receptor, the gene feature of the T cell receptor includes the α chain gene feature and the β chain gene feature of the T cell receptor, and the sequence feature of the T cell receptor includes the α chain sequence feature of the T cell receptor and the β chain sequence feature of the T cell receptor. The server adds the α chain gene feature of the T cell receptor and the α chain sequence feature of the T cell receptor through the feature fusion module to obtain the α chain gene sequence feature of the T cell receptor. The server adds the β chain gene feature of the T cell receptor and the β chain sequence feature of the T cell receptor through the feature fusion module to obtain the β chain gene sequence feature of the T cell receptor. The α chain gene sequence feature and the β chain gene sequence feature of the T cell receptor constitute the gene sequence fusion feature of the T cell receptor. The server splices the gene sequence fusion feature and the three-dimensional structure feature of the T cell receptor through the feature fusion module to obtain the initial receptor feature of the T cell receptor. The server performs at least one full connection on the initial receptor feature of the T cell receptor through the feature fusion module to obtain the receptor feature of the T cell receptor.

[0163] It should be noted that the above is explained by taking the example of the server fusing the gene characteristics, sequence characteristics and three-dimensional structural characteristics of the immune cell receptor to obtain the receptor characteristics of the immune cell receptor. In other possible implementations, in addition to fusing the gene characteristics, sequence characteristics and three-dimensional structural characteristics of the immune cell receptor, the server can also fuse other information to obtain the receptor characteristics of the immune cell receptor, see the following implementation.

[0164] In a possible embodiment, the server fuses the gene features, sequence features, three-dimensional structural features of the immune cell receptor and the physicochemical information of amino acids in the immune cell receptor through the feature fusion module of the affinity prediction model to obtain the receptor features of the immune cell receptor.

[0165] The physicochemical information of amino acids in the immune cell receptor includes the physical and chemical properties of amino acids, wherein the physical properties include basic composition and structure, solubility, melting point, boiling point, optical behavior and optical rotation, etc. The chemical properties include acidity and alkalinity and hydrophobicity, etc. Introducing the physicochemical information of amino acids into the receptor characteristics of the immune cell receptor can improve the expression ability of the receptor characteristics, so that the receptor characteristics can more completely represent the immune cell receptor.

[0166] For example, the server uses the feature fusion module to splice the gene features and sequence features of the immune cell receptor to obtain the gene sequence fusion features of the immune cell receptor. The server uses the feature fusion module of the affinity prediction model to weightedly fuse the gene sequence fusion features and three-dimensional structural features of the immune cell receptor based on the gated attention mechanism to obtain the initial receptor features of the immune cell receptor. The server uses the feature fusion module to add the initial receptor features of the immune cell receptor and the physicochemical information of the amino acids in the immune cell receptor to obtain the receptor features of the immune cell receptor.

[0167] 305. The server performs multiple full connections on the receptor features of the immune cell receptor through the affinity prediction model, and outputs the affinity of the immune cell receptor for the target antigen, where the target antigen is an antigen that can specifically bind to the immune cell receptor.

[0168] In a possible embodiment, the server performs multiple full connections on the receptor features of the immune cell receptor through the regression module of the affinity prediction model to obtain the affinity of the immune cell receptor for the target antigen. Through multiple full connections, the receptor features can be mapped to a numerical value, which is also the affinity of the immune cell receptor for the target antigen. In some embodiments, the regression module includes a multilayer perceptron (Multilayer Perception, MLP). The regression module is also called a regression head.

[0169] In this implementation, the server regresses the receptor characteristics of the immune cell receptor through the regression module of the affinity prediction model, and outputs the affinity of the immune cell receptor for the target antigen, without the need for repeated experiments and measurements, and is highly efficient.

[0170] The following will be combined Figure 5 The above steps 301-305 are explained.

[0171] See also Figure 5, the server inputs the gene information, sequence information and three-dimensional structure information of the immune cell receptor into the affinity prediction model, and the affinity prediction model includes a gene encoder 501, a sequence encoder 502 and a structure encoder 503. The server encodes the gene information of the immune cell receptor through the gene encoder 501 to obtain the gene characteristics of the immune cell receptor. The server encodes the sequence information of the immune cell receptor through the sequence encoder 502 to obtain the sequence characteristics of the immune cell receptor. The server encodes the three-dimensional structure information of the immune cell receptor through the structure encoder 503 to obtain the three-dimensional structure characteristics of the immune cell receptor. The affinity prediction model also includes a feature fusion module 504, and the server splices the gene characteristics and sequence characteristics of the immune cell receptor through the feature fusion module 504 to obtain the gene sequence fusion characteristics of the immune cell receptor. bio The server uses the feature fusion module of the affinity prediction model to fuse the gene sequence of the immune cell receptor with the feature h based on the gated attention mechanism. bio and three-dimensional structural features h stru Perform weighted fusion to obtain the immune cell receptor target gene sequence fusion feature h / bio and the target 3D structural feature h / stru The server uses the feature fusion module 504 to fuse the target gene sequence with the feature h / bio Multiply h by the target three-dimensional structure / stru , get the initial receptor characteristics of the B cell receptor h fusion The server uses the feature fusion module 504 to combine the initial receptor feature h fusion . Perform two full connections (FC1, FC2) to obtain the receptor feature Representation of the B cell receptor. The affinity prediction model also includes a regression module. The server performs affinity prediction based on the receptor feature of the immune cell receptor through the regression module of the affinity prediction model, and outputs the affinity of the immune cell receptor to the target antigen.

[0172] It should be noted that the above description process is based on an example in which the server executes the above steps 301-305. In other possible implementations, the above steps 301-305 may also be executed by the terminal, and the embodiment of the present application does not limit this.

[0173] All the above optional technical solutions can be arbitrarily combined to form optional embodiments of the present application, which will not be described one by one here.

[0174] Figure 6The results of testing the affinity prediction method provided in the examples of the present application on a public data set are shown.

[0175] See also Figure 6 , the results of the affinity prediction model provided by the affinity prediction method provided in the embodiment of the present application when tested on a public data set, and the rightmost column of each data set corresponds to the affinity prediction model provided in the embodiment of the present application. Figure 6 It can be seen that the mean square error (RMSE) of the affinity prediction model provided in the embodiment of the present application is smaller than that of other models in the related art on multiple public data sets, and the R2 score is higher than that of other models in the related art on multiple public data sets.

[0176] Through the technical solution provided in the embodiment of the present application, the affinity prediction model extracts features from the gene information and sequence of the immune cell receptor to obtain the gene features and sequence features of the immune cell receptor. In the process of obtaining the receptor features of the immune cell receptor, the gene features, sequence features and three-dimensional structural features are integrated. The introduction of three-dimensional structural features enriches the content of the receptor features and improves the expressiveness of the receptor features, so that when affinity prediction is performed based on the receptor features, the accuracy of the affinity between the obtained immune cell receptor and the target antigen is high.

[0177] In order to more clearly illustrate the affinity prediction method provided in the embodiment of the present application, the training method of the affinity prediction model provided in the embodiment of the present application is described below, see Figure 7 Taking the execution subject as a server as an example, the method includes the following steps.

[0178] 701. The server inputs the gene information, sequence information and three-dimensional structural characteristics of the sample immune cell receptor into the affinity prediction model.

[0179] Step 701 and the above-mentioned step 302 belong to the same inventive concept. For the implementation process, please refer to the relevant description of the above-mentioned step 302, which will not be repeated here.

[0180] 702. The server extracts features of the gene information and sequence information of the sample immune cell receptor through the affinity prediction model to obtain the gene features and sequence features of the sample immune cell receptor.

[0181] Step 702 and the above-mentioned step 303 belong to the same inventive concept. For the implementation process, please refer to the relevant description of the above-mentioned step 303, which will not be repeated here.

[0182] 703. The server fuses the gene features, sequence features, and three-dimensional structural features of the sample immune cell receptor through the affinity prediction model to obtain the receptor features of the sample immune cell receptor.

[0183] Step 703 and the above-mentioned step 304 belong to the same inventive concept. For the implementation process, please refer to the relevant description of the above-mentioned step 304, which will not be repeated here.

[0184] 704. The server performs multiple full connections on the receptor features of the sample immune cell receptor through the affinity prediction model, and outputs the predicted affinity of the sample immune cell receptor for the sample antigen, where the sample antigen is an antigen that can specifically bind to the sample immune cell receptor.

[0185] Step 704 and the above-mentioned step 305 belong to the same inventive concept. For the implementation process, please refer to the relevant description of the above-mentioned step 305, which will not be repeated here.

[0186] 705. The server classifies the receptor characteristics of the sample immune cell receptor through the affinity prediction model, and outputs the predicted affinity level of the sample immune cell receptor for the sample antigen.

[0187] In a possible embodiment, the server fully connects and normalizes the receptor features of the immune cell receptor through the classification module of the affinity prediction model, and outputs the probability of the immune cell receptor corresponding to multiple candidate affinity levels with the target antigen. The server determines the predicted affinity level from the multiple candidate affinity levels based on the probability of the immune cell receptor corresponding to the target antigen.

[0188] Among them, the classification model of the affinity prediction model is only used in the model training process, and works together with the regression module of the affinity prediction model to train the affinity prediction model, that is, it constitutes a multi-task training system. The regression module corresponds to the task of predicting affinity, and the classification module corresponds to the task of predicting affinity level. When the affinity prediction model is used for affinity prediction, the classification model does not take effect.

[0189] For example, the server fully connects and normalizes the receptor features of the sample immune cell receptor through the affinity prediction model to obtain the probability that the sample immune cell receptor corresponds to multiple candidate affinity levels. The server determines the predicted affinity level from the multiple candidate affinity levels based on the probability that the sample immune cell receptor corresponds to multiple candidate affinity levels. For example, the server fully connects the receptor features of the immune cell receptor through the classification module of the affinity prediction model to obtain the grade classification matrix of the immune cell receptor. The server normalizes the grade classification matrix of the immune cell receptor through the classification module to obtain the probability set corresponding to the immune cell receptor, which includes multiple probabilities, each probability corresponding to a candidate affinity level. The server determines the candidate affinity level corresponding to the probability that meets the target condition in the probability set as the predicted affinity level through the classification model. In some embodiments, the probability that meets the target condition refers to the highest probability in the probability set, or the probability that the probability in the probability set is greater than or equal to the probability threshold, and the probability threshold is set by the technician according to the actual situation, and the embodiment of the present application does not limit this.

[0190] 706. The server trains the affinity prediction model based on the first difference information and the second difference information, wherein the first difference information is the difference information between the predicted affinity of the sample immune cell receptor for the sample antigen and the labeled affinity of the sample immune cell receptor for the sample antigen, and the second difference information is the difference information between the predicted affinity level of the sample immune cell receptor for the sample antigen and the labeled affinity level of the sample immune cell receptor for the sample antigen.

[0191] In a possible implementation, the server constructs a mean square error loss function (MSE Loss) based on the first difference information. The server constructs a cross entropy loss function (Categorical Cross-entropy (CE) Loss) based on the second difference information. The server constructs a joint loss function based on the mean square error loss function and the cross entropy loss function. The server uses the gradient descent method and the joint loss function to train the affinity prediction model, that is, to adjust the model parameters of the affinity prediction model.

[0192] Among them, the mean square error loss function and the cross entropy loss function are regular term constraints on each other, and jointly promote affinity prediction to learn information-rich and highly generalizable features.

[0193] For example, the server constructs a mean square error loss function as shown in formula (1) based on the first difference information. The server constructs a cross entropy loss function as shown in formula (2) based on the second difference information. The server constructs a joint loss function as shown in formula (3) based on the mean square error loss function and the cross entropy loss function. The server uses the gradient descent method to train the affinity prediction model using the joint loss function.

[0194]

[0195]

[0196]

[0197] in, is the mean square error loss function, is the cross entropy loss function, is the joint loss function, N is the number of sample immune cell receptors, a positive integer, and y i is the annotated affinity of the sample immune cell receptor numbered i, is the predicted affinity of the sample immune cell receptor, is the first difference information, is the annotated affinity rating of the sample immune cell receptor, p c is the probability that the predicted affinity level of the sample immune cell receptor is c, C is the set of candidate affinity levels, and λ is a hyperparameter.

[0198] In a possible implementation, the affinity prediction method provided in the embodiment of the present application can be provided as a cloud service, that is, the user transmits relevant data to the cloud server through the terminal, the cloud server executes the affinity prediction method provided in the embodiment of the present application, returns the affinity to the terminal, and the terminal displays the affinity to the user.

[0199] For example, see Figure 8, the terminal displays an affinity prediction interface 800, which is used to obtain the gene information, sequence information and three-dimensional structural characteristics of the immune cell receptor. Taking the immune cell receptor as a T cell receptor as an example, the affinity prediction interface includes a first sequence information input area 801, a second sequence information input area 802, a first gene information input area 803, a second gene information input area 804 and a three-dimensional structural characteristic input area 805. Among them, the first sequence information input area 801 is used to input the amino acid sequence of the α chain of the T cell receptor, the second sequence information input area 802 is used to input the amino acid sequence of the β chain of the T cell receptor, the first gene information input area 803 is used to input the VJ information of the α chain of the T cell receptor, the second gene information input area 804 is used to input the VJ information of the β chain of the T cell receptor, and the three-dimensional structural characteristic input area 805 is used to input the three-dimensional structural characteristics of the T cell receptor. The affinity prediction interface 800 also includes a submit control 806. In response to a click operation on the submit control 806, the terminal uploads the gene information, sequence information and three-dimensional structural characteristics of the immune cell receptor to the cloud server, and the cloud server performs calculations based on the gene information, sequence information and three-dimensional structural characteristics of the immune cell receptor to obtain the affinity of the immune cell receptor for the target antigen. For example, see Fig. 9 The terminal displays an affinity display interface 900 , in which the affinity of the immune cell receptor to the target antigen is displayed.

[0200] Fig.10 is a schematic diagram of the structure of an affinity prediction device provided in an embodiment of the present application, see Fig.10 The device includes: an input unit 1001, a feature extraction unit 1002, a feature fusion unit 1003 and an affinity prediction unit 1004.

[0201] The input unit 1001 is used to input the gene information, sequence information and three-dimensional structural characteristics of the immune cell receptor into the affinity prediction model.

[0202] The feature extraction unit 1002 is used to extract features from the gene information and sequence information of the immune cell receptor through the affinity prediction model to obtain the gene features and sequence features of the immune cell receptor.

[0203] The feature fusion unit 1003 is used to fuse the gene features, sequence features and three-dimensional structure features of the immune cell receptor through the affinity prediction model to obtain the receptor features of the immune cell receptor.

[0204] The affinity prediction unit 1004 is used to perform multiple full connections on the receptor features of the immune cell receptor through the affinity prediction model, and output the affinity of the immune cell receptor to the target antigen, where the target antigen is an antigen that can specifically bind to the immune cell receptor.

[0205] In a possible embodiment, the feature extraction unit 1002 is used to encode the VDJ information of the immune cell receptor through the gene encoder of the affinity prediction model to obtain the gene feature of the immune cell receptor, wherein V is the encoding variable region, D is the encoding hypervariable region, and J is the encoding cross-linking region. The amino acid sequence of the immune cell receptor is encoded through the sequence encoder of the affinity prediction model to obtain the sequence feature of the immune cell receptor.

[0206] In a possible implementation, the feature extraction unit 1002 is configured to perform any of the following:

[0207] When the immune cell receptor is a B cell receptor, the VJ information of the light chain and the VDJ information of the heavy chain of the immune cell receptor are encoded to obtain the gene characteristics of the immune cell receptor.

[0208] When the immune cell receptor is a T cell receptor, the VJ information of the α chain and the VDJ information of the β chain of the immune cell receptor are encoded to obtain the gene characteristics of the immune cell receptor.

[0209] In a possible embodiment, the feature extraction unit 1002 is used to fully connect the VJ information of the light chain and the VDJ information of the heavy chain of the immune cell receptor to obtain the gene characteristics of the immune cell receptor, and the gene characteristics of the immune cell receptor include the gene characteristics of the light chain of the immune cell receptor and the gene characteristics of the heavy chain of the immune cell receptor. The VJ information of the α chain and the VDJ information of the β chain of the immune cell receptor are encoded to obtain the gene characteristics of the immune cell receptor, including: fully connecting the VJ information of the α chain and the VDJ information of the β chain of the immune cell receptor to obtain the gene characteristics of the immune cell receptor, and the gene characteristics of the immune cell receptor include the gene characteristics of the α chain of the immune cell receptor and the gene characteristics of the β chain of the immune cell receptor.

[0210] In a possible implementation, the feature extraction unit 1002 is configured to perform any of the following:

[0211] When the immune cell receptor is a B cell receptor, the amino acid sequence of the light chain and the amino acid sequence of the heavy chain of the immune cell receptor are encoded by the sequence encoder of the affinity prediction model based on the attention mechanism to obtain the sequence characteristics of the immune cell receptor. The sequence characteristics of the immune cell receptor include the light chain sequence characteristics and the heavy chain sequence characteristics of the immune cell receptor.

[0212] When the immune cell receptor is a T cell receptor, the amino acid sequence of the α chain and the amino acid sequence of the β chain of the immune cell receptor are encoded by the sequence encoder of the affinity prediction model based on the attention mechanism to obtain the sequence characteristics of the immune cell receptor. The sequence characteristics of the immune cell receptor include the α chain sequence characteristics and the β chain sequence characteristics of the immune cell receptor.

[0213] In a possible implementation, the feature fusion unit 1003 is used to splice the gene feature and sequence feature of the immune cell receptor through the feature fusion device of the affinity prediction model to obtain the gene sequence fusion feature of the immune cell receptor. Based on the gated attention mechanism, the gene sequence fusion feature and the three-dimensional structure feature of the immune cell receptor are weightedly fused to obtain the receptor feature of the immune cell receptor.

[0214] In a possible implementation, the device further includes:

[0215] A three-dimensional structural feature acquisition unit is used to obtain the amino acid sequence of the CDR3 region of the immune cell receptor. A multiple sequence alignment is performed on the amino acid sequence of the CDR3 region of the immune cell receptor to obtain at least one reference amino acid sequence, and the similarity between the reference amino acid sequence and the amino acid sequence of the CDR3 region of the immune cell receptor meets the similarity condition. A homologous template corresponding to the amino acid sequence of the CDR3 region of the immune cell receptor is obtained, and the homologous template includes the structural information of the homologous sequence of the amino acid sequence of the CDR3 region of the immune cell receptor. Multiple rounds of iterations are performed based on the amino acid sequence of the CDR3 region of the immune cell receptor, the at least one reference amino acid sequence and the homologous template to obtain the three-dimensional structural features of the immune cell receptor.

[0216] In a possible implementation, the device further includes:

[0217] The three-dimensional structure feature acquisition unit is used to acquire the three-dimensional structure information of the immune cell receptor, and the three-dimensional structure information includes the three-dimensional coordinates of multiple amino acids in the immune cell receptor.

[0218] The three-dimensional structure feature acquisition unit is used to perform any of the following:

[0219] Graph convolution is performed on the three-dimensional structural information of the immune cell receptor to obtain the three-dimensional structural characteristics of the immune cell receptor.

[0220] Based on the attention mechanism, the three-dimensional structural information of the immune cell receptor is encoded to obtain the three-dimensional structural characteristics of the immune cell receptor.

[0221] In a possible embodiment, the feature fusion unit 1003 is also used to fuse the gene features, sequence features, three-dimensional structural features of the immune cell receptor and the physicochemical information of amino acids in the immune cell receptor through the affinity prediction model to obtain the receptor features of the immune cell receptor.

[0222] It should be noted that: the affinity prediction device provided in the above embodiment only uses the division of the above functional modules as an example to illustrate when predicting affinity. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the computer device is divided into different functional modules to complete all or part of the functions described above. In addition, the affinity prediction device provided in the above embodiment and the affinity prediction method embodiment belong to the same concept, and the specific implementation process is detailed in the method embodiment, which will not be repeated here.

[0223] Through the technical solution provided in the embodiment of the present application, the affinity prediction model extracts features from the gene information and sequence of the immune cell receptor to obtain the gene features and sequence features of the immune cell receptor. In the process of obtaining the receptor features of the immune cell receptor, the gene features, sequence features and three-dimensional structural features are integrated. The introduction of three-dimensional structural features enriches the content of the receptor features and improves the expressiveness of the receptor features, so that when affinity prediction is performed based on the receptor features, the accuracy of the affinity between the obtained immune cell receptor and the target antigen is high.

[0224] Fig.11 is a schematic diagram of the structure of a training device for an affinity prediction model provided in an embodiment of the present application, see Fig.11 The device includes: a training information input unit 1101, a training feature extraction unit 1102, a training feature fusion unit 1103, a predicted affinity output unit 1104, a training classification unit 1105 and a training unit 1106.

[0225] The training information input unit 1101 is used to input the gene information, sequence information and three-dimensional structural characteristics of the sample immune cell receptor into the affinity prediction model.

[0226] The training feature extraction unit 1102 is used to extract features from the gene information and sequence information of the sample immune cell receptor through the affinity prediction model to obtain the gene features and sequence features of the sample immune cell receptor.

[0227] The training feature fusion unit 1103 is used to fuse the gene features, sequence features and three-dimensional structure features of the sample immune cell receptor through the affinity prediction model to obtain the receptor features of the sample immune cell receptor.

[0228] The predicted affinity output unit 1104 is used to perform multiple full connections on the receptor features of the sample immune cell receptor through the affinity prediction model, and output the predicted affinity of the sample immune cell receptor to the sample antigen, where the sample antigen is an antigen that can specifically bind to the sample immune cell receptor.

[0229] The training classification unit 1105 is used to classify the receptor characteristics of the sample immune cell receptor through the affinity prediction model, and output the predicted affinity level of the sample immune cell receptor to the sample antigen.

[0230] The training unit 1106 is used to train the affinity prediction model based on the first difference information and the second difference information, wherein the first difference information is the difference information between the predicted affinity of the sample immune cell receptor for the sample antigen and the labeled affinity of the sample immune cell receptor for the sample antigen, and the second difference information is the difference information between the predicted affinity level of the sample immune cell receptor for the sample antigen and the labeled affinity level of the sample immune cell receptor for the sample antigen.

[0231] In a possible implementation, the training classification unit 1105 is used to fully connect and normalize the receptor features of the sample immune cell receptor through the affinity prediction model to obtain the probability that the sample immune cell receptor corresponds to multiple candidate affinity levels. Based on the probability that the sample immune cell receptor corresponds to multiple candidate affinity levels, the predicted affinity level is determined from the multiple candidate affinity levels.

[0232] It should be noted that: the affinity prediction model training device provided in the above embodiment only uses the division of the above functional modules as an example when training the affinity prediction model. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the computer device is divided into different functional modules to complete all or part of the functions described above. In addition, the affinity prediction device provided in the above embodiment belongs to the same concept as the affinity prediction method embodiment. The specific implementation process is detailed in the method embodiment and will not be repeated here.

[0233] The embodiment of the present application provides a computer device for executing the above method. The computer device can be implemented as a terminal or a server. The structure of the terminal is first introduced below:

[0234] Fig.12 1 is a schematic diagram of a terminal structure provided in an embodiment of the present application. The terminal 1200 may be: a smart phone, a tablet computer, a laptop computer or a desktop computer. The terminal 1200 may also be called a user equipment, a portable terminal, a laptop terminal, a desktop terminal or other names.

[0235] Typically, the terminal 1200 includes: one or more processors 1201 and one or more memories 1202 .

[0236] The processor 1201 may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor 1201 may be implemented in at least one hardware form of DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), and PLA (Programmable Logic Array). The processor 1201 may also include a main processor and a coprocessor. The main processor is a processor for processing data in the awake state, also known as a CPU (Central Processing Unit); the coprocessor is a low-power processor for processing data in the standby state. In some embodiments, the processor 1201 may be integrated with a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content to be displayed on the display screen. In some embodiments, the processor 1201 may also include an AI (Artificial Intelligence) processor, which is used to process computing operations related to machine learning.

[0237] The memory 1202 may include one or more computer-readable storage media, which may be non-transitory. The memory 1202 may also include a high-speed random access memory, and a non-volatile memory, such as one or more disk storage devices, flash memory storage devices. In some embodiments, the non-transitory computer-readable storage medium in the memory 1202 is used to store at least one computer program, which is used to be executed by the processor 1201 to implement the affinity prediction method or affinity prediction model training method provided in the method embodiment of the present application.

[0238] In some embodiments, the terminal 1200 may further optionally include: a peripheral device interface 1203 and at least one peripheral device. The processor 1201, the memory 1202 and the peripheral device interface 1203 may be connected via a bus or a signal line. Each peripheral device may be connected to the peripheral device interface 1203 via a bus, a signal line or a circuit board. Specifically, the peripheral device includes: at least one of a radio frequency circuit 1204, a display screen 1205, a camera assembly 1206, an audio circuit 1207 and a power supply 1208.

[0239] The peripheral device interface 1203 may be used to connect at least one peripheral device related to I / O (Input / Output) to the processor 1201 and the memory 1202. In some embodiments, the processor 1201, the memory 1202, and the peripheral device interface 1203 are integrated on the same chip or circuit board; in some other embodiments, any one or two of the processor 1201, the memory 1202, and the peripheral device interface 1203 may be implemented on a separate chip or circuit board, which is not limited in this embodiment.

[0240] The radio frequency circuit 1204 is used to receive and transmit RF (Radio Frequency) signals, also known as electromagnetic signals. The radio frequency circuit 1204 communicates with a communication network and other communication devices via electromagnetic signals. The radio frequency circuit 1204 converts electrical signals into electromagnetic signals for transmission, or converts received electromagnetic signals into electrical signals. Optionally, the radio frequency circuit 1204 includes: an antenna system, an RF transceiver, one or more amplifiers, a tuner, an oscillator, a digital signal processor, a codec chipset, a user identity module card, and the like.

[0241] The display screen 1205 is used to display a UI (User Interface). The UI may include graphics, text, icons, videos, and any combination thereof. When the display screen 1205 is a touch display screen, the display screen 1205 also has the ability to collect touch signals on the surface or above the surface of the display screen 1205. The touch signal may be input as a control signal to the processor 1201 for processing. At this time, the display screen 1205 may also be used to provide virtual buttons and / or virtual keyboards, also known as soft buttons and / or soft keyboards.

[0242] The camera assembly 1206 is used to collect images or videos. Optionally, the camera assembly 1206 includes a front camera and a rear camera. Usually, the front camera is set on the front panel of the terminal, and the rear camera is set on the back of the terminal.

[0243] The audio circuit 1207 may include a microphone and a speaker. The microphone is used to collect sound waves from the user and the environment, and convert the sound waves into electrical signals and input them to the processor 1201 for processing, or input them to the RF circuit 1204 to achieve voice communication.

[0244] The power supply 1208 is used to supply power to various components in the terminal 1200. The power supply 1208 may be alternating current, direct current, a disposable battery, or a rechargeable battery.

[0245] In some embodiments, the terminal 1200 further includes one or more sensors 1209 . The one or more sensors 1209 include, but are not limited to: an acceleration sensor 1210 , a gyroscope sensor 1211 , a pressure sensor 1212 , an optical sensor 1213 , and a proximity sensor 1214 .

[0246] The acceleration sensor 1210 can detect the magnitude of acceleration on three coordinate axes of the coordinate system established with the terminal 1200 .

[0247] The gyro sensor 1211 can detect the body direction and rotation angle of the terminal 1200 . The gyro sensor 1211 can cooperate with the acceleration sensor 1210 to collect the user's 3D actions on the terminal 1200 .

[0248] The pressure sensor 1212 may be provided at the side frame of the terminal 1200 and / or the lower layer of the display screen 1205. When the pressure sensor 1212 is provided at the side frame of the terminal 1200, it may detect the user's holding signal of the terminal 1200, and the processor 1201 may perform left and right hand recognition or shortcut operation according to the holding signal collected by the pressure sensor 1212. When the pressure sensor 1212 is provided at the lower layer of the display screen 1205, the processor 1201 may control the operability controls on the UI interface according to the user's pressure operation on the display screen 1205.

[0249] The optical sensor 1213 is used to collect the ambient light intensity. In one embodiment, the processor 1201 can control the display brightness of the display screen 1205 according to the ambient light intensity collected by the optical sensor 1213.

[0250] The proximity sensor 1214 is used to collect the distance between the user and the front of the terminal 1200 .

[0251] Those skilled in the art will understand that Fig.12 The structure shown in the figure does not constitute a limitation on the terminal 1200, and the terminal 1200 may include more or fewer components than those shown in the figure, or combine certain components, or adopt a different component arrangement.

[0252] The above-mentioned computer device can also be implemented as a server. The structure of the server is introduced below:

[0253] Fig.13 It is a structural diagram of a server provided in an embodiment of the present application. The server 1300 may have relatively large differences due to different configurations or performances, and may include one or more processors (Central Processing Units, CPU) 1301 and one or more memories 1302, wherein the one or more memories 1302 store at least one computer program, and the at least one computer program is loaded and executed by the one or more processors 1301 to implement the methods provided in the above-mentioned various method embodiments. Of course, the server 1300 may also have components such as a wired or wireless network interface, a keyboard, and an input and output interface for input and output. The server 1300 may also include other components for implementing device functions, which will not be described in detail here.

[0254] In an exemplary embodiment, a computer-readable storage medium is also provided, such as a memory including a computer program, and the computer program can be executed by a processor to complete the affinity prediction method or the affinity prediction model training method in the above embodiment. For example, the computer-readable storage medium can be a read-only memory (ROM), a random access memory (RAM), a compact disc (CD-ROM), a magnetic tape, a floppy disk, an optical data storage device, etc.

[0255] In an exemplary embodiment, a computer program product or a computer program is also provided, which includes a program code, and the program code is stored in a computer-readable storage medium. A processor of a computer device reads the program code from the computer-readable storage medium, and the processor executes the program code, so that the computer device executes the above-mentioned affinity prediction method or affinity prediction model training method.

[0256] In some embodiments, the computer program involved in the embodiments of the present application may be deployed and executed on a computer device, or on multiple computer devices located at one location, or on multiple computer devices distributed at multiple locations and interconnected by a communication network. Multiple computer devices distributed at multiple locations and interconnected by a communication network may constitute a blockchain system.

[0257] A person skilled in the art will understand that all or part of the steps to implement the above embodiments may be accomplished by hardware or by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a disk or an optical disk, etc.

[0258] The above are only optional embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims

1. A method for predicting affinity, characterized in that: The method comprises: Input the genetic information, sequence information, and three-dimensional structural features of immune cell receptors into the affinity prediction model; By using the affinity prediction model, feature extraction is performed on the gene information and sequence information of the immune cell receptor to obtain the gene features and sequence features of the immune cell receptor; The gene characteristics, sequence characteristics and three-dimensional structural characteristics of the immune cell receptor are integrated through the affinity prediction model to obtain the receptor characteristics of the immune cell receptor; The affinity prediction model is used to perform multiple full connections on the receptor features of the immune cell receptor, and the affinity of the immune cell receptor for the target antigen is output, where the target antigen is an antigen that can specifically bind to the immune cell receptor.

2. The method according to claim 1, characterized in that The feature extraction of the gene information and sequence information of the immune cell receptor by the affinity prediction model to obtain the gene features and sequence features of the immune cell receptor includes: The VDJ information of the immune cell receptor is encoded by the gene encoder of the affinity prediction model to obtain the gene characteristics of the immune cell receptor, wherein V is the encoding variable region, D is the encoding hypervariable region, and J is the encoding cross-linking region; The amino acid sequence of the immune cell receptor is encoded by the sequence encoder of the affinity prediction model to obtain the sequence characteristics of the immune cell receptor.

3. The method according to claim 2, characterized in that The encoding of the VDJ information of the immune cell receptor to obtain the gene characteristics of the immune cell receptor includes any of the following: When the immune cell receptor is a B cell receptor, encoding the VJ information of the light chain and the VDJ information of the heavy chain of the immune cell receptor to obtain the gene characteristics of the immune cell receptor; When the immune cell receptor is a T cell receptor, the VJ information of the α chain and the VDJ information of the β chain of the immune cell receptor are encoded to obtain the gene characteristics of the immune cell receptor.

4. The method according to claim 3, characterized in that When the immune cell receptor is a B cell receptor, encoding the VJ information of the light chain and the VDJ information of the heavy chain of the immune cell receptor to obtain the gene characteristics of the immune cell receptor includes: When the immune cell receptor is a B cell receptor, fully connect the VJ information of the light chain and the VDJ information of the heavy chain of the immune cell receptor to obtain the gene characteristics of the immune cell receptor, wherein the gene characteristics of the immune cell receptor include the gene characteristics of the light chain of the immune cell receptor and the gene characteristics of the heavy chain of the immune cell receptor; When the immune cell receptor is a T cell receptor, encoding the VJ information of the α chain and the VDJ information of the β chain of the immune cell receptor to obtain the gene characteristics of the immune cell receptor includes: When the immune cell receptor is a T cell receptor, the VJ information of the α chain and the VDJ information of the β chain of the immune cell receptor are fully connected to obtain the gene characteristics of the immune cell receptor, and the gene characteristics of the immune cell receptor include the gene characteristics of the α chain of the immune cell receptor and the gene characteristics of the β chain of the immune cell receptor.

5. The method according to claim 2, characterized in that: The amino acid sequence of the immune cell receptor is encoded by the sequence encoder of the affinity prediction model to obtain the sequence features of the immune cell receptor including any of the following: In the case where the immune cell receptor is a B cell receptor, the amino acid sequence of the light chain and the amino acid sequence of the heavy chain of the immune cell receptor are encoded by the sequence encoder of the affinity prediction model based on the attention mechanism to obtain the sequence features of the immune cell receptor, wherein the sequence features of the immune cell receptor include the sequence features of the light chain and the heavy chain of the immune cell receptor; In the case where the immune cell receptor is a T cell receptor, the amino acid sequence of the α chain and the amino acid sequence of the β chain of the immune cell receptor are encoded by the sequence encoder of the affinity prediction model based on the attention mechanism to obtain the sequence characteristics of the immune cell receptor, and the sequence characteristics of the immune cell receptor include the α chain sequence characteristics and the β chain sequence characteristics of the immune cell receptor.

6. The method according to claim 1, characterized in that The affinity prediction model is used to fuse the gene features, sequence features and three-dimensional structural features of the immune cell receptor to obtain the receptor features of the immune cell receptor, including: The gene features and sequence features of the immune cell receptor are spliced ​​by the feature fusion device of the affinity prediction model to obtain the gene sequence fusion features of the immune cell receptor; Based on the gated attention mechanism, the gene sequence fusion features and the three-dimensional structure features of the immune cell receptor are weightedly fused to obtain the receptor features of the immune cell receptor.

7. The method according to claim 1, characterized in that Before inputting the gene information, sequence information and three-dimensional structural features of the immune cell receptor into the affinity prediction model, the method includes: Obtaining the amino acid sequence of the CDR3 region of the immune cell receptor; Performing a multiple sequence alignment on the amino acid sequence of the CDR3 region of the immune cell receptor to obtain at least one reference amino acid sequence, wherein the similarity between the reference amino acid sequence and the amino acid sequence of the CDR3 region of the immune cell receptor meets a similarity condition; Obtaining a homologous template corresponding to the amino acid sequence of the CDR3 region of the immune cell receptor, wherein the homologous template includes structural information of a homologous sequence of the amino acid sequence of the CDR3 region of the immune cell receptor; Multiple rounds of iterations are performed based on the amino acid sequence of the CDR3 region of the immune cell receptor, the at least one reference amino acid sequence and the homologous template to obtain the three-dimensional structural characteristics of the immune cell receptor.

8. The method according to claim 1, characterized in that Before inputting the gene information, sequence information and three-dimensional structural features of the immune cell receptor into the affinity prediction model, the method includes: Acquiring three-dimensional structural information of the immune cell receptor, wherein the three-dimensional structural information includes three-dimensional coordinates of multiple amino acids in the immune cell receptor; The method further comprises any of the following: Performing graph convolution on the three-dimensional structural information of the immune cell receptor to obtain the three-dimensional structural features of the immune cell receptor; The three-dimensional structural information of the immune cell receptor is encoded based on the attention mechanism to obtain the three-dimensional structural features of the immune cell receptor.

9. The method according to claim 1, characterized in that: After extracting features from the gene information and sequence information of the immune cell receptor using the affinity prediction model to obtain the gene features and sequence features of the immune cell receptor, the method further includes: Through the affinity prediction model, the gene characteristics, sequence characteristics, three-dimensional structural characteristics of the immune cell receptor and the physicochemical information of amino acids in the immune cell receptor are integrated to obtain the receptor characteristics of the immune cell receptor.

10. A method for training an affinity prediction model, characterized in that: The method comprises: Inputting the gene information, sequence information and three-dimensional structural characteristics of the sample immune cell receptor into the affinity prediction model; By using the affinity prediction model, feature extraction is performed on the gene information and sequence information of the sample immune cell receptor to obtain the gene features and sequence features of the sample immune cell receptor; The gene features, sequence features and three-dimensional structural features of the sample immune cell receptor are integrated through the affinity prediction model to obtain the receptor features of the sample immune cell receptor; The affinity prediction model is used to perform multiple full connections on the receptor features of the sample immune cell receptor, and the predicted affinity of the sample immune cell receptor for the sample antigen is output, where the sample antigen is an antigen that can specifically bind to the sample immune cell receptor; Classifying the receptor characteristics of the sample immune cell receptors by using the affinity prediction model, and outputting the predicted affinity level of the sample immune cell receptors for the sample antigen; The affinity prediction model is trained based on first difference information and second difference information, wherein the first difference information is the difference information between the predicted affinity of the sample immune cell receptor for the sample antigen and the labeled affinity of the sample immune cell receptor for the sample antigen, and the second difference information is the difference information between the predicted affinity level of the sample immune cell receptor for the sample antigen and the labeled affinity level of the sample immune cell receptor for the sample antigen.

11. The method according to claim 10, characterized in that The method of classifying the receptor characteristics of the sample immune cell receptor by the affinity prediction model and outputting the predicted affinity level of the sample immune cell receptor for the sample antigen includes: Fully connecting and normalizing the receptor features of the sample immune cell receptors through the affinity prediction model to obtain the probability that the sample immune cell receptors correspond to multiple candidate affinity levels; The predicted affinity rating is determined from the plurality of candidate affinity ratings based on a probability that the sample immune cell receptor corresponds to the plurality of candidate affinity ratings.

12. An affinity prediction device, characterized in that: The device comprises: An input unit, used to input the gene information, sequence information and three-dimensional structural characteristics of the immune cell receptor into the affinity prediction model; A feature extraction unit, used to extract features from the gene information and sequence information of the immune cell receptor using the affinity prediction model to obtain the gene features and sequence features of the immune cell receptor; A feature fusion unit, used to fuse the gene features, sequence features and three-dimensional structural features of the immune cell receptor through the affinity prediction model to obtain the receptor features of the immune cell receptor; The affinity prediction unit is used to perform multiple full connections on the receptor characteristics of the immune cell receptor through the affinity prediction model, and output the affinity of the immune cell receptor to the target antigen, where the target antigen is an antigen that can specifically bind to the immune cell receptor.

13. A training device for an affinity prediction model, characterized in that: The device comprises: A training information input unit, used to input the gene information, sequence information and three-dimensional structural characteristics of the sample immune cell receptor into the affinity prediction model; A training feature extraction unit is used to extract features from the gene information and sequence information of the sample immune cell receptor by using the affinity prediction model to obtain the gene features and sequence features of the sample immune cell receptor; A training feature fusion unit is used to fuse the gene features, sequence features and three-dimensional structural features of the sample immune cell receptor through the affinity prediction model to obtain the receptor features of the sample immune cell receptor; A predicted affinity output unit, used to perform multiple full connections on the receptor features of the sample immune cell receptor through the affinity prediction model, and output the predicted affinity of the sample immune cell receptor for the sample antigen, wherein the sample antigen is an antigen that can specifically bind to the sample immune cell receptor; A training classification unit is used to classify the receptor characteristics of the sample immune cell receptor by using the affinity prediction model, and output the predicted affinity level of the sample immune cell receptor for the sample antigen; A training unit is used to train the affinity prediction model based on first difference information and second difference information, wherein the first difference information is the difference information between the predicted affinity of the sample immune cell receptor for the sample antigen and the labeled affinity of the sample immune cell receptor for the sample antigen, and the second difference information is the difference information between the predicted affinity level of the sample immune cell receptor for the sample antigen and the labeled affinity level of the sample immune cell receptor for the sample antigen.

14. A computer device, characterized in that: The computer device includes one or more processors and one or more memories, and at least one computer program is stored in the one or more memories. The computer program is loaded and executed by the one or more processors to implement the affinity prediction method according to any one of claims 1 to 9, or to implement the affinity prediction model training method according to claim 10 or claim 11.

15. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores at least one computer program, which is loaded and executed by a processor to implement the affinity prediction method according to any one of claims 1 to 9, or to implement the affinity prediction model training method according to any one of claims 10 or 11.

16. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, it implements the affinity prediction method according to any one of claims 1 to 9, or implements the affinity prediction model training method according to any one of claim 10 or claim 11.

Images