Antibody sequence generation method, related apparatus, novel coronavirus antibody and use

Abstract

Disclosed are an antibody sequence generation method and apparatus, a computer device, and a storage medium, relating to the technical field of computers. The method comprises: acquiring antigen sequence information, wherein the antigen sequence information represents amino acids in an antigen sequence; performing feature extraction on the antigen sequence information to obtain antigen sequence features of the antigen sequence information; performing feature decoding on the antigen sequence features to obtain predicted probabilities corresponding to each position in an antibody sequence, wherein the predicted probabilities comprise a predicted probability that an amino acid at the position is each of a variety of amino acids; and determining sequence information formed by amino acids corresponding to the maximum predicted probabilities at different positions as antibody sequence information, wherein the antibody sequence indicated by the antibody sequence information can bind to the antigen sequence. An antibody sequence capable of binding to a specific antigen sequence is generated, thereby improving the specificity of the generated antibody sequence.

Description

Antibody sequence generation method, related device, anti-new coronavirus antibody and application

[0001] This application claims priority to the Chinese patent application filed with the China Patent Office on May 18, 2023, with application number 202310566319X and application name “Antibody sequence generation method, device, computer equipment and storage medium”, and claims priority to the Chinese patent application filed with the China Patent Office on December 6, 2023, with application number 2023116658138 and application name “Anti-new coronavirus antibodies and their applications”, the entire contents of which are incorporated by reference into this application. Technical Field

[0002] The embodiments of the present application relate to the fields of computer technology and biopharmaceutical technology, and in particular to antibody sequence generation, anti-novel coronavirus antibodies and their applications. Background Art

[0003] Antigen sequences are substances that the immune system needs to eliminate. They can induce an immune response and bind to antibody sequences produced by the immune response, generating immune defenses that play a vital role in maintaining human health. Besides being produced from animals or humans, antibody sequences can also be artificially generated to enable the development of drugs such as antibody vaccines.

[0004] In the related art, when developing new antibody sequences, naturally occurring antibody sequences are typically modified and optimized, and then antibody sequences with high affinity for the antigen sequence are screened. However, this method makes it difficult to generate antibody sequences that can bind to specific antigen sequences. Therefore, a method for generating antibody sequences that are specific to specific antigen sequences is urgently needed.

[0005] Summary of the Invention

[0006] The embodiments of the present application provide a method, apparatus, device, storage medium, and program product for generating an antibody sequence that can bind to a specific antigen sequence, thereby improving the specificity of the generated antibody sequence. The technical solution is as follows:

[0007] In one aspect, a method for generating an antibody sequence is provided, the method comprising:

[0008] Obtaining antigen sequence information, wherein the antigen sequence information represents amino acids in the antigen sequence;

[0009] performing feature extraction on the antigen sequence information to obtain antigen sequence features of the antigen sequence information;

[0010] Performing feature decoding on the antigen sequence feature to obtain a predicted probability for each position in the antibody sequence, the predicted probability including a predicted probability that the amino acid at the position belongs to each of a plurality of amino acids;

[0011] The sequence information composed of the amino acids with the highest predicted probability at each position is determined as the antibody sequence information. The antibody sequence indicated by the antibody sequence information is capable of binding to the antigen sequence.

[0012] In another aspect, an antibody sequence generation device is provided, comprising:

[0013] An information acquisition module is used to obtain antigen sequence information, wherein the antigen sequence information represents amino acids in the antigen sequence;

[0014] A feature extraction module is used to extract features from the antigen sequence information to obtain antigen sequence features of the antigen sequence information;

[0015] a feature decoding module, configured to perform feature decoding on the antigen sequence feature to obtain a predicted probability for each position in the antibody sequence, wherein the predicted probability includes a predicted probability that the amino acid at the position belongs to each of a plurality of amino acids;

[0016] The information determination module is used to determine the sequence information composed of the amino acids with the highest predicted probability at each position as antibody sequence information, and the antibody sequence indicated by the antibody sequence information is capable of binding to the antigen sequence.

[0017] The present application also provides an anti-novel coronavirus antibody, which has a high binding affinity to the S protein of the novel coronavirus.

[0018] In the first aspect of the present application, the present application provides an antibody or antigen-binding fragment. According to an embodiment of the present application, the antibody or antigen-binding fragment comprises an HCDR3 having an amino acid sequence as shown in SEQ ID NO: 3 or a conservatively modified form thereof. Compared with natural antibodies, the antibody containing the HCDR3 of the present application has better affinity and neutralization ability for the new coronavirus, and can effectively detect the new coronavirus, diagnose related diseases caused by the new coronavirus, or prevent and / or treat related diseases caused by the new coronavirus.

[0019] According to an embodiment of the present application, the antibody or antigen-binding fragment may further include at least one of the following technical features:

[0020] According to an embodiment of the present application, the antibody or antigen-binding fragment further includes a CDR selected from at least one of the following: heavy chain variable region CDR: amino acid sequence of SEQ ID NO: 1-2 or a conservative modification thereof; light chain variable region CDR: amino acid sequence of SEQ ID NO: 4-6 or a conservative modification thereof.

[0021] According to an embodiment of the present application, the antibody or antigen-binding fragment comprises: HCDR1, HCDR2, HCDR3 as shown in the amino acid sequences of SEQ ID NOs: 1, 2 and 3, respectively; and LCDR1, LCDR2, LCDR3 as shown in the amino acid sequences of SEQ ID NOs: 4, 5 and 6, respectively.

[0022] According to an embodiment of the present application, the antibody or antigen-binding fragment includes a heavy chain framework region.

[0023] According to an embodiment of the present application, at least a portion of the heavy chain framework region is derived from at least one of a mouse antibody, a primate antibody, a bovine antibody, a horse antibody, a cow antibody, a porcine antibody, a sheep antibody, a goat antibody, a dog antibody, a cat antibody, a rabbit antibody, a camel antibody, a donkey antibody, a deer antibody, a mink antibody, a chicken antibody, a duck antibody, a goose antibody, a turkey antibody, a fighting cock antibody, or a mutant thereof.

[0024] According to an embodiment of the present application, at least a portion of the heavy chain framework region is derived from at least one of a murine antibody and a human antibody.

[0025] According to an embodiment of the present application, the antibody or antigen-binding fragment includes a light chain framework region.

[0026] According to an embodiment of the present application, at least a portion of the light chain framework region is derived from at least one of a mouse antibody, a primate antibody, a bovine antibody, a horse antibody, a cow antibody, a porcine antibody, a sheep antibody, a goat antibody, a dog antibody, a cat antibody, a rabbit antibody, a camel antibody, a donkey antibody, a deer antibody, a mink antibody, a chicken antibody, a duck antibody, a goose antibody, a turkey antibody, a fighting cock antibody, or a mutant thereof.

[0027] According to an embodiment of the present application, at least a portion of the light chain framework region is derived from at least one of a murine antibody and a human antibody.

[0028] According to an embodiment of the present application, the antibody or antigen-binding fragment comprises: a heavy chain variable region having an amino acid sequence as shown in SEQ ID NO: 7 or an amino acid sequence having at least 90% homology thereto; and a light chain variable region having an amino acid sequence as shown in SEQ ID NO: 8 or an amino acid sequence having at least 90% homology thereto.

[0029] According to an embodiment of the present application, the antibody or antigen-binding fragment further includes a heavy chain constant region.

[0030] According to an embodiment of the present application, the antibody or antigen-binding fragment further includes a light chain constant region.

[0031] According to an embodiment of the present application, at least a portion of at least one of the heavy chain constant region and the light chain constant region is derived from at least one of a mouse antibody, a primate antibody, a bovine antibody, a horse antibody, a cow antibody, a porcine antibody, a sheep antibody, a goat antibody, a dog antibody, a cat antibody, a rabbit antibody, a camel antibody, a donkey antibody, a deer antibody, a mink antibody, a chicken antibody, a duck antibody, a goose antibody, a turkey antibody, a fighting cock antibody, or a mutant thereof.

[0032] According to an embodiment of the present application, the heavy chain constant region includes a heavy chain constant region selected from IgG1, IgG2, IgG3, IgG4, IgA, IgM, IgE or IgD; or the light chain constant region includes a light chain constant region selected from κ or λ type.

[0033] According to an embodiment of the present application, the light chain constant region and the heavy chain constant region are both derived from a murine antibody or a mutant thereof, or a human antibody or a mutant thereof.

[0034] According to an embodiment of the present application, the N-terminus of the heavy chain constant region is connected to the C-terminus of the heavy chain variable region.

[0035] According to an embodiment of the present application, the N-terminus of the light chain constant region is connected to the C-terminus of the light chain variable region.

[0036] According to an embodiment of the present application, the heavy chain constant region includes the heavy chain constant region shown in SEQ ID NO: 9 or an amino acid sequence having at least 80% identity thereto.

[0037] According to an embodiment of the present application, the light chain constant region includes the light chain constant region shown in SEQ ID NO: 10 or an amino acid sequence having at least 80% identity thereto.

[0038] According to an embodiment of the present invention, the antibody includes at least one of a polyclonal antibody, a full-length monoclonal antibody, a Fab antibody, a Fab' antibody, a F(ab')2 antibody, a Fv antibody, a single-chain antibody, a single-domain antibody and a minimum recognition unit; or the antigen-binding fragment includes at least one of a F(ab')2 fragment, a Fab' fragment, a Fab fragment, a F(ab)2 fragment, a Fv fragment, a scFv fragment, a scFv-Fc fusion protein, a scFv-Fv fusion protein and a minimum recognition unit.

[0039] According to an embodiment of the present application, the antibody or antigen-binding fragment comprises: a heavy chain having an amino acid sequence as shown in SEQ ID NO: 11 or an amino acid sequence having at least 80% homology thereto; and a light chain having an amino acid sequence as shown in SEQ ID NO: 12 or an amino acid sequence having at least 80% homology thereto.

[0040] In the second aspect of this application, a recombinant protein is proposed. According to an embodiment of this application, the recombinant protein comprises: the antibody or antigen-binding fragment described in the first aspect. The recombinant protein of this application has improved affinity for and neutralization ability against the novel coronavirus, and can effectively detect the novel coronavirus, diagnose diseases related to the novel coronavirus, or prevent and / or treat diseases related to the novel coronavirus.

[0041] According to an embodiment of the present application, the recombinant protein further includes at least one of a biologically active protein or a fragment thereof, or a biologically active polypeptide or a fragment thereof.

[0042] According to an embodiment of the present application, the biologically active protein or fragment thereof includes at least one selected from a protein tag, a protein toxin or a fragment thereof, an interferon or a fragment thereof, a biological response modifier or a fragment thereof, and an Fc fragment.

[0043] In the third aspect of this application, a multispecific antibody is proposed. According to an embodiment of this application, the multispecific antibody comprises: a first binding region comprising the antibody or antigen-binding fragment described in the first aspect; and a second binding region having a first molecule binding activity. The multispecific antibody of this application has improved affinity for and neutralization ability against the novel coronavirus, and can effectively detect the novel coronavirus, diagnose related diseases caused by the novel coronavirus, or prevent and / or treat related diseases caused by the novel coronavirus.

[0044] According to an embodiment of the present application, the first molecule includes at least one of an antigen, a virus, a bacterium, an endotoxin, a cytokine, and a cytokine receptor.

[0045] In a fourth aspect of the present application, a nucleic acid molecule is provided. According to embodiments of the present application, the nucleic acid molecule encodes the antibody or antigen-binding fragment described in the first aspect, the recombinant protein described in the second aspect, or the multispecific antibody described in the third aspect. The nucleic acid molecule of the present application may encode the antibody or antigen-binding fragment described in the first aspect, the recombinant protein described in the second aspect, or the multispecific antibody described in the third aspect.

[0046] According to an embodiment of the present application, the nucleic acid molecule is DNA.

[0047] In a fifth aspect of the present application, an expression vector is provided. According to an embodiment of the present application, the expression vector carries the nucleic acid molecule described in the fourth aspect. Thus, the expression vector of the present application can effectively achieve expression of the antibody or antigen-binding fragment of the first aspect, the recombinant protein of the second aspect, or the multispecific antibody of the third aspect, thereby achieving large-scale in vitro production of the antibody or antigen-binding fragment, recombinant protein, or multispecific antibody.

[0048] According to an embodiment of the present application, the expression vector is selected from a eukaryotic expression vector or a prokaryotic expression vector.

[0049] In the sixth aspect of this application, a recombinant cell is provided. According to embodiments of this application, the recombinant cell comprises a nucleic acid molecule as described in the fourth aspect or an expression vector as described in the fifth aspect; or expresses the antibody or antigen-binding fragment as described in the first aspect, the recombinant protein as described in the second aspect, or the multispecific antibody as described in the third aspect. Under suitable conditions, the recombinant cell can effectively express the aforementioned antibody or antigen-binding fragment, recombinant protein, or multispecific antibody within the cell.

[0050] According to an embodiment of the present application, the recombinant cell is obtained by introducing the expression vector described in the fifth aspect into a host cell.

[0051] In the seventh aspect of the present application, the present application proposes a conjugate. According to the embodiments of the present application, the conjugate comprises the antibody or antigen-binding fragment described in the first aspect, the recombinant protein described in the second aspect, or the multispecific antibody described in the third aspect; and a coupling portion, which is connected to the antibody or antigen-binding fragment, recombinant protein or multispecific antibody. The conjugate of the present application can bind to the new coronavirus, has better affinity for the new coronavirus and neutralization ability of the new coronavirus, and can effectively detect the new coronavirus, diagnose related diseases caused by the new coronavirus, or prevent and / or treat related diseases caused by the new coronavirus.

[0052] According to an embodiment of the present application, the coupling portion includes at least one selected from a carrier, a drug, a toxin, a cytokine, a protein tag and a modifier.

[0053] In the eighth aspect of the present application, the present application proposes a pharmaceutical composition. According to the embodiments of the present application, the pharmaceutical composition includes the antibody or antigen-binding fragment described in the first aspect, the recombinant protein described in the second aspect, the multispecific antibody described in the third aspect, the nucleic acid molecule described in the fourth aspect, the expression vector described in the fifth aspect, the recombinant cell described in the sixth aspect, or the conjugate described in the seventh aspect. The pharmaceutical composition of the present application can bind to the new coronavirus, has better affinity for the new coronavirus and neutralization ability of the new coronavirus, and can effectively prevent and / or treat related diseases caused by the new coronavirus.

[0054] According to an embodiment of the present application, the pharmaceutical composition further includes a pharmaceutically acceptable excipient.

[0055] In the ninth aspect of the present application, the present application proposes a kit. According to the embodiments of the present application, the kit includes the antibody or antigen-binding fragment described in the first aspect, the recombinant protein described in the second aspect, the multispecific antibody described in the third aspect, the nucleic acid molecule described in the fourth aspect, the expression vector described in the fifth aspect, the recombinant cell described in the sixth aspect, or the conjugate described in the seventh aspect. The kit of the present application can bind to the new coronavirus and has a strong affinity for the new coronavirus, and can effectively detect the new coronavirus or diagnose related diseases caused by the new coronavirus.

[0056] In the tenth aspect of the present application, the present application proposes a use of the antibody or antigen-binding fragment described in the first aspect, the recombinant protein described in the second aspect, the multispecific antibody described in the third aspect, the nucleic acid molecule described in the fourth aspect, the expression vector described in the fifth aspect, the recombinant cell described in the sixth aspect, or the conjugate described in the seventh aspect in the preparation of a drug, wherein the drug is used to prevent and / or treat related diseases caused by the new coronavirus.

[0057] In the eleventh aspect of the present application, the present application proposes a use of the antibody or antigen-binding fragment described in the first aspect, the recombinant protein described in the second aspect, the multispecific antibody described in the third aspect, the nucleic acid molecule described in the fourth aspect, the expression vector described in the fifth aspect, the recombinant cell described in the sixth aspect, or the conjugate described in the seventh aspect in the preparation of a kit, wherein the kit is used to detect the new coronavirus, diagnose related diseases caused by the new coronavirus, or evaluate the prognosis of related diseases caused by the new coronavirus.

[0058] In the twelfth aspect of this application, the present application proposes a method for detecting the novel coronavirus. According to embodiments of the present application, the method comprises: contacting the sample to be tested with the antibody or antigen-binding fragment described in the first aspect, the recombinant protein described in the second aspect, the multispecific antibody described in the third aspect, or the conjugate described in the seventh aspect to form an immune complex. The method of the present application has the advantages of detecting the novel coronavirus, especially for in vitro detection, and has high detection accuracy.

[0059] According to an embodiment of the present application, whether the sample to be tested contains the content of the new coronavirus is determined based on the signal of the immune complex.

[0060] According to an embodiment of the present application, the immune complex further includes a second antibody, which binds to the antibody or antigen-binding fragment.

[0061] According to an embodiment of the present application, the immune complex further includes a second antibody, which binds to the new coronavirus.

[0062] On the other hand, a computer device is provided, comprising a processor and a memory, wherein the memory stores at least one computer program, and the at least one computer program is loaded and executed by the processor to implement the method described in the above aspect.

[0063] On the other hand, a computer-readable storage medium is provided, in which at least one computer program is stored. The at least one computer program is loaded and executed by a processor to implement the method described in the above aspect.

[0064] On the other hand, a computer program product is provided, comprising a computer program, wherein the computer program is loaded and executed by a processor to implement the method according to the above aspect.

[0065] The solution provided in the embodiments of the present application takes into account that when the antibody sequence is able to bind to the antigen sequence, the amino acids in the antigen sequence should have a certain correlation with the amino acids in the antibody sequence. Therefore, from the perspective of antigen sequence information, based on the antigen sequence information that can represent the amino acids in the antigen sequence, the antigen sequence characteristics are extracted, and based on the antigen sequence characteristics, the antibody sequence information corresponding to the antibody sequence that can bind to the antigen sequence is directly decoded, thereby generating an antibody sequence that can bind to a specific antigen sequence, thereby improving the specificity of the generated antibody sequence. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] FIG1 is a schematic diagram of an implementation environment provided by an embodiment of the present application;

[0067] FIG2 is a flow chart of a method for generating an antibody sequence provided in an embodiment of the present application;

[0068] FIG3 is a flow chart of a method for generating an antibody sequence provided in an embodiment of the present application;

[0069] FIG4 is a flow chart of a method for training an antibody production model provided in an embodiment of the present application;

[0070] FIG5 is a schematic diagram of generating antibody sequence information provided in an embodiment of the present application;

[0071] FIG6 is a flowchart of a pre-training method for a decoding sub-model provided in an embodiment of the present application;

[0072] FIG7 is a schematic diagram of a pre-training process of a decoding sub-model provided in an embodiment of the present application;

[0073] FIG8 is a flowchart of a pre-training method for an encoding sub-model provided in an embodiment of the present application;

[0074] FIG9 is a schematic diagram of a complex structure of an antibody sequence and an antigen sequence provided in an embodiment of the present application;

[0075] FIG10 is a comparative diagram of antibody performance provided in Examples of the present application;

[0076] FIG11 is a schematic structural diagram of an antibody sequence generation device provided in an embodiment of the present application;

[0077] FIG12 is a schematic structural diagram of another antibody sequence generation device provided in an embodiment of the present application;

[0078] FIG13 is a schematic structural diagram of a terminal provided in an embodiment of the present application;

[0079] FIG14 is a schematic structural diagram of a server provided in an embodiment of the present application;

[0080] Figure 15 is a schematic diagram of the process of generating an antigen epitope sequence into an antibody sequence in Example 1 of the present application;

[0081] Figure 16 is a schematic diagram of the structural docking of the complex composed of the artificial intelligence-generated antibody and the novel coronavirus S antigen in Example 1 of the present application;

[0082] Figure 17 is a graph showing the results of a protein blotting experiment of natural antibodies and artificial intelligence-generated antibodies binding to the novel coronavirus S protein in Example 2 of the present application. DETAILED DESCRIPTION

[0083] In order to make the objectives, technical solutions and advantages of the embodiments of the present application clearer, the implementation methods of the present application will be further described in detail below with reference to the accompanying drawings.

[0084] It is understood that the terms "first," "second," and the like used herein may be used to describe various concepts herein, but unless otherwise specified, these concepts are not limited by these terms. These terms are merely used to distinguish one concept from another. For example, without departing from the scope of this application, first antibody sequence information may be referred to as second antibody sequence information, and similarly, second antibody sequence information may be referred to as first antibody sequence information.

[0085] Here, "at least one" refers to one or more than one. For example, at least one amino acid can be one amino acid, two amino acids, three amino acids, or any other integer greater than one. "Plurality" refers to two or more than two. For example, "plurality" can be two amino acids, three amino acids, or any other integer greater than two. "Each" refers to each of at least one. For example, "each amino acid" refers to each amino acid in the plurality of amino acids. If the plurality of amino acids is three amino acids, "each amino acid" refers to each of the three amino acids.

[0086] It is understandable that in the embodiments of this application, related data such as antigen sequence information and antibody sequence information are involved. When the above embodiments of this application are applied to specific products or technologies, user permission or consent is required, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards of relevant countries and regions.

[0087] Artificial Intelligence (AI) refers to the theories, methods, techniques, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, to perceive the environment, acquire knowledge, and use that knowledge to achieve optimal results. In other words, AI is a comprehensive technology within computer science that seeks to understand the essence of intelligence and produce new intelligent machines that can respond in a manner similar to human intelligence. AI also studies the design principles and implementation methods of various intelligent machines, enabling them to possess the capabilities of perception, reasoning, and decision-making.

[0088] Artificial intelligence (AI) technology is a comprehensive discipline encompassing a wide range of fields, encompassing both hardware and software technologies. Foundational AI technologies generally include sensors, specialized AI chips, cloud computing, distributed storage, big data processing, operating / interaction systems, and mechatronics. AI software technologies primarily encompass computer vision, speech processing, natural language processing, as well as machine learning / deep learning, autonomous driving, and smart transportation.

[0089] Natural language processing (NLP) is a key area of ​​research in computer science and artificial intelligence. It studies the theories and methods that enable effective communication between humans and computers using natural language. Natural language processing integrates linguistics, computer science, and mathematics. Therefore, research in this field involves natural language—the language we use in everyday life—and is closely linked to the study of linguistics.

[0090] Machine learning (ML) is a multidisciplinary field that encompasses probability theory, statistics, approximation theory, convex analysis, and algorithmic complexity theory. It specifically studies how computers can simulate or implement human learning behaviors to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their performance. Machine learning is at the core of artificial intelligence and the fundamental way to make computers intelligent. Its applications span all areas of AI. Machine learning and deep learning typically include techniques such as artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and learning through demonstration.

[0091] The following describes the method for generating antibody sequences provided in the embodiments of the present application based on artificial intelligence technology and machine learning technology.

[0092] The antibody sequence generation method provided in the embodiment of the present application can be used in a computer device. Optionally, the computer device is a terminal or a server. Optionally, the server is an independent physical server, or a server cluster or a 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, CDN (Content Delivery Network, content distribution network) and big data and artificial intelligence platforms. Optionally, the terminal is a smart phone, tablet computer, laptop computer, desktop computer, smart speaker, smart watch, etc., but is not limited to this.

[0093] In one possible implementation, the computer program involved in the embodiments of the present application can 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 through a communication network. Multiple computer devices distributed at multiple locations and interconnected through a communication network can constitute a blockchain system.

[0094] FIG1 is a schematic diagram of an implementation environment provided by an embodiment of the present application. Referring to FIG1 , the implementation environment includes: a terminal 101 and a server 102. The terminal 101 and the server 102 are connected via a wireless or wired network. Optionally, the server 102 is used to train an antibody generation model, which is used to generate antibody sequence information of antibody sequences that can bind to antigen sequences. The server 102 sends the trained antibody generation model to the terminal 101, and the terminal 101 can generate antibody sequence information of antibody sequences that can bind to antigen sequences through the antibody generation model.

[0095] It should be noted that Figure 1 only uses the example of the server 102 training the antibody generation model and sending it to the terminal 101. In another embodiment, the server can also generate antibody sequence information of the antibody sequence that can bind to the antigen sequence through the antibody generation model, and then send the generated antibody sequence information to the terminal 101.

[0096] Antigen sequences are substances that the immune system needs to eliminate. They can induce an immune response and bind to antibody sequences produced by the immune response, triggering an immune defense (specific response). Antigen sequences are composed of amino acids. Antibody sequences are proteins specifically synthesized by the immune system, also known as immunoglobulins. Natural antibody sequences are molecules produced by B lymphocytes. Antibody sequences can recognize and bind to invading pathogens such as bacteria, viruses, and fungi, and play a key role in the immune response. Antibody sequences are composed of heavy chains and light chains. The heavy chain is the large polypeptide subunit of the antibody sequence, while the light chain is the small polypeptide subunit of the antibody sequence. Antibodies are typically composed of four polypeptide chains: two identical light chains with a smaller relative molecular mass and two identical heavy chains with a larger relative molecular mass. Antibodies are symmetrical structures connected by disulfide bonds. Both light and heavy chains are composed of a series of variable regions and constant regions. The variable regions of the heavy and light chains are capable of specifically binding to the antigen sequence. The variable regions encompass the Complementarity Determining Region (CDR), which is crucial for antibody recognition of specific antigens. This includes the crucial CDR3 (Complementarity Determining Region). CDR3 is the most diverse and important region in the antibody sequence, playing a key role in antigen recognition. The CDR3 of the heavy chain is called CDRH3, and different amino acid combinations are the primary cause of varying antibody specificity.

[0097] The antibody sequence generation method provided in the embodiments of the present application can be applied to any scenario where it is necessary to generate an antibody sequence that can bind to an antigen sequence.

[0098] For example, it is applied to the research and development scenario of antibody vaccines. Antibody vaccines are vaccines that use prepared antibody sequences to protect the human body from pathogen infection. Compared with traditional vaccines, antibody vaccines do not directly guide the human immune system to produce its own antibody sequences, but directly administer prepared antibody sequences to the human body to prevent pathogen infection. Antibody vaccines are commonly used to treat or prevent infectious diseases, including viral infections, bacterial infections, and parasitic infections. Usually, these antibody sequences can be prepared from animals or humans, and the preparation method includes collecting animal immune serum or using recombinant DNA (Deoxyribo Nucleic Acid) technology to produce artificial antibodies. Antibody vaccines can not only treat diseases after infection, but also prevent diseases before infection. However, the antibody sequence generation method provided in the embodiments of the present application makes it possible for the research and development of antibody vaccines no longer rely solely on naturally occurring antibody sequences, but can generate antibody sequences that can bind to specific antigen sequences, thereby accelerating the development speed of antibody vaccines and improving the effectiveness of antibody vaccines.

[0099] FIG2 is a flow chart of a method for generating an antibody sequence provided in an embodiment of the present application. The embodiment of the present application is executed by a computer device. Referring to FIG2 , the method includes:

[0100] 201. A computer device obtains antigen sequence information, where the antigen sequence information represents amino acids in the antigen sequence.

[0101] In the embodiment of the present application, the computer device obtains antigen sequence information of an antigen sequence, which is the antigen sequence of the antibody sequence to be generated and capable of binding thereto. The antigen sequence is composed of multiple amino acids, and the antigen sequence information represents the amino acids in the antigen sequence.

[0102] The antigen sequence can be a complete antigen or a partial sequence of a complete antigen. For example, the antigen sequence can be a partial sequence composed of the amino acids of the antigenic epitope of the antigen. The antigenic epitope is also called the antigenic determinant, which refers to the site on the antigen that can specifically bind to the antibody binding site and determine the specificity of the antigen.

[0103] 202. The computer device extracts features from the antigen sequence information to obtain antigen sequence features of the antigen sequence information.

[0104] After acquiring the antigen sequence information, the computer device performs feature extraction on the antigen sequence information to extract antigen sequence features of the antigen sequence information. The antigen sequence features represent the types of amino acids in the antigen sequence and the connection structure of multiple amino acids. The antigen sequence features can be features in the form of vectors or matrices, etc., which are not limited in the present embodiment.

[0105] 203. The computer device decodes the antigen sequence features to obtain the predicted probability of each position in the antibody sequence, including the predicted probability of the amino acid at the position belonging to each of the multiple amino acids.

[0106] Taking into account that when the antibody sequence can bind to the antigen sequence, the type and structure of the amino acids in the antigen sequence have a certain correlation with the type and structure of the amino acids in the antibody sequence, after the computer equipment obtains the antigen sequence characteristics, it decodes the antigen sequence characteristics to obtain the predicted probability that the amino acid at each position in the antibody sequence belongs to each of the multiple amino acids, thereby completing the prediction of the antibody sequence, which is equivalent to generating an antibody sequence that can bind to the antigen sequence based on the antigen sequence.

[0107] 204. The computer device determines the sequence information composed of the amino acids with the highest predicted probability at each position as the antibody sequence information. The antibody sequence indicated by the antibody sequence information is capable of binding to the antigen sequence.

[0108] Each position corresponds to a predicted probability of each amino acid among multiple amino acids. For a certain position, the maximum predicted probability is determined among the predicted probabilities of each amino acid among multiple amino acids, and the amino acid with the maximum predicted probability is determined as the amino acid at that position. This method is used to determine the amino acid at each position, and the sequence information composed of the amino acids at multiple positions is the antibody sequence information.

[0109] The method provided in the embodiments of the present application takes into account that when the antibody sequence is able to bind to the antigen sequence, the amino acids in the antigen sequence should have a certain correlation with the amino acids in the antibody sequence. Therefore, from the perspective of antigen sequence information, based on the antigen sequence information that can represent the amino acids in the antigen sequence, the antigen sequence characteristics are extracted, and based on the antigen sequence characteristics, the antibody sequence information corresponding to the antibody sequence that can bind to the antigen sequence is directly decoded, thereby generating an antibody sequence that can bind to a specific antigen sequence, thereby improving the specificity of the generated antibody sequence.

[0110] In another embodiment, a computer device stores an antibody generation model, which is used to generate an antibody sequence that can bind to an antigen sequence. The antibody generation model includes an encoding sub-model and a decoding sub-model. The computer device generates an antibody sequence using the antibody generation model. For detailed process, see the embodiment shown in Figure 3 below. Figure 3 is a flow chart of an antibody sequence generation method provided in an embodiment of the present application. The embodiment of the present application is executed by a computer device. Referring to Figure 3, the method includes:

[0111] 301. A computer device obtains antigen sequence information, where the antigen sequence information represents amino acids in the antigen sequence.

[0112] The computer device obtains antigen sequence information of an antigen sequence, which is the antigen sequence of the antibody sequence to be generated and capable of binding thereto. The antigen sequence is composed of multiple amino acids, and the antigen sequence information represents the amino acids in the antigen sequence.

[0113] Among them, the antigen sequence information can be any type of sequence information. For example, the antigen sequence information can include amino acid identifiers of multiple amino acids in the antigen sequence. Alternatively, the antigen sequence information can be tokens of multiple amino acids in the antigen sequence. In natural language processing, tokens usually refer to basic units in text, such as words or symbols. In the BERT model, tokens usually refer to the form in which the basic units in the text are encoded. In the embodiment of the present application, amino acids are the basic units in the antigen sequence, and tokens refer to the form in which the amino acids in the antigen sequence are encoded.

[0114] 302. The computer device extracts features of the antigen sequence information through the encoding sub-model in the antibody generation model to obtain antigen sequence features of the antigen sequence information.

[0115] The computer device inputs the antigen sequence information into the encoding sub-model of the antibody generation model, and the encoding sub-model extracts features of the antigen sequence information and outputs antigen sequence features of the antigen sequence information.

[0116] In one possible implementation, during the feature extraction process of the encoding sub-model on the antigen sequence information, multiple antigen key features, antigen value features, and antigen query features are generated. The antigen sequence features in the embodiment of the present application include target antigen key features and target antigen value features. Wherein, the encoding sub-model includes m encoding networks, where m is a positive integer. The process of the computer device obtaining the target antigen key features and target antigen value features through the encoding sub-model includes:

[0117] The computer device performs feature extraction on the antigen sequence information using the first encoding network in the encoding sub-model to obtain a first candidate antigen bond feature, a first candidate antigen value feature, and a first candidate antigen query feature. The first candidate antigen bond feature, the first candidate antigen value feature, and the first candidate antigen query feature are fused to obtain a first antigen fusion feature. The computer device performs feature extraction on the nth antigen fusion feature using the n+1th encoding network in the encoding sub-model to obtain the n+1th candidate antigen bond feature, the n+1th candidate antigen value feature, and the n+1th candidate antigen query feature, where n is a positive integer less than m. The computer device determines the mth candidate antigen bond feature and the mth candidate antigen value feature as the target antigen bond feature and the target antigen value feature.

[0118] Wherein, n can be any positive integer less than m. The above is only explained by taking the processing of the first encoding network and the n+1th encoding network in the encoding sub-model as an example. The complete processing process of the encoding sub-model is to input the antigen sequence information into the first encoding network for processing, and then process it through each encoding network in turn until the last encoding network completes the processing. The detailed process is as follows: the computer device performs feature extraction on the antigen sequence information through the first encoding network in the encoding sub-model to obtain the first candidate antigen bond feature, the first candidate antigen value feature and the first candidate antigen query feature, and fuses the first candidate antigen bond feature, the first candidate antigen value feature and the first candidate antigen query feature to obtain the first antigen fusion feature. The computer device performs feature extraction on the first antigen fusion feature through the second encoding network in the encoding sub-model to obtain the second candidate antigen bond feature, the second candidate antigen value feature and the second candidate antigen query feature, and fuses the second candidate antigen bond feature, the second candidate antigen value feature and the second candidate antigen query feature to obtain the second antigen fusion feature. The computer device continues to extract features from the second antigen fusion feature through the third encoding network in the encoding sub-model to obtain a third candidate antigen bond feature, a third candidate antigen value feature, and a third candidate antigen query feature, and fuses the third candidate antigen bond feature, the third candidate antigen value feature, and the third candidate antigen query feature to obtain a third antigen fusion feature, and so on, until the computer device obtains the last candidate antigen bond feature, the last candidate antigen value feature, and the last candidate antigen query feature through the last encoding network in the encoding sub-model. The processing is then completed, and the last candidate antigen bond feature is determined as the target antigen bond feature, and the last candidate antigen value feature is determined as the target antigen value feature.

[0119] The antigen sequence features are encoded sequentially through m encoding networks. Each encoding network will fuse the extracted candidate antigen key features, candidate antigen value features and candidate antigen query features, and use the obtained antigen fusion features as the input of the next encoding network, so that the extraction results of one encoding network will be inherited by the next encoding network, effectively improving the encoding accuracy.

[0120] In the examples of this application, the antigen query feature, antigen key feature, and antigen value feature belong to different feature spaces. Each amino acid in the antigen sequence has its own antigen key feature, antigen value feature, and antigen query feature. The antigen value feature represents the characteristics of the amino acid itself, while the antigen query feature and antigen key feature are used to determine the feature vector of the attention weight.

[0121] In one possible implementation, each encoding network includes a self-attention layer and a feedforward neural network layer. Taking the first encoding network as an example, the computer device uses the self-attention layer in the encoding network to extract features from the antigen sequence information to obtain candidate antigen key features, candidate antigen value features, and candidate antigen query features. The candidate antigen key features, candidate antigen value features, and candidate antigen query features are then fused to obtain an initial antigen fusion feature. The computer device uses the feedforward neural network layer in the encoding network to perform feature transformation on the initial antigen fusion feature to obtain the antigen fusion feature.

[0122] Optionally, the process of a computer device obtaining candidate antigen query features, candidate antigen key features and candidate antigen value features includes: determining the product of the antigen sequence information and the parameter matrix as an antigen sequence matrix, and obtaining the candidate antigen query features, candidate antigen key features and candidate antigen value features corresponding to the antigen sequence information based on the antigen sequence matrix.

[0123] Among them, the parameter matrix is ​​used for spatial transformation, and the parameter matrix is ​​the model parameter obtained by training. After obtaining the antigen sequence matrix, the computer device splits the antigen sequence matrix into candidate antigen query features, candidate antigen key features, and candidate antigen value features corresponding to the antigen sequence information. For example, the parameter matrix is ​​a 3-dimensional matrix, and the antigen sequence matrix obtained by multiplying the antigen sequence matrix and the parameter matrix is ​​a 3-dimensional matrix. The computer device uses each dimension of the antigen sequence matrix as a candidate antigen query feature, candidate antigen key feature, and candidate antigen value feature corresponding to the antigen sequence information.

[0124] Optionally, the computer device fuses the candidate antigen key feature, the candidate antigen value feature, and the candidate antigen query feature to obtain an initial antigen fusion feature, including:

[0125] The product of the candidate antigen query feature, the candidate antigen key feature and the scaling factor is normalized to obtain a normalized feature, and the product of the normalized feature and the candidate antigen value feature is determined as the initial antigen fusion feature. The computer device obtains the scaling factor, and the scaling factor represents the normalized scaling multiple. In the embodiment of the present application, the computer device uses the scaling factor as a normalization parameter to normalize the product to obtain a normalized feature, which characterizes the correlation between the candidate antigen query feature and the candidate antigen key feature. The computer device can use the normalized feature as the weight of the candidate antigen value feature, so the computer device determines the product between the normalized feature and the candidate antigen value feature as the initial antigen fusion feature.

[0126] The computer device uses the following formula to fuse the candidate antigen key feature, the candidate antigen value feature, and the candidate antigen query feature to obtain the initial antigen fusion feature:

[0127] Among them, A represents the initial antigen fusion feature, Q represents the candidate antigen query feature, K represents the candidate antigen key feature, and T represents transposition. represents the scaling factor, V represents the candidate antigen value feature, and softmax(·) represents the normalization function.

[0128] Optionally, the feedforward neural network layer can be expressed by the following formula: B = FFN (A) = max (0, AW1 + b1) W ％ +b ％

[0129] Among them, B represents the antigen fusion feature, FFN(·) represents the feedforward neural network, A represents the initial antigen fusion feature, max(·) represents the maximum value, W1, b1, W ％ and b ％ Represents model parameters in a feed-forward neural network layer.

[0130] In one possible implementation, the encoder model is a BERT (Bidirectional Encoder Representations from Transformers) model or a Roformer (Rotary Transformers) model. BERT is obtained by pre-training a deep bidirectional Transformer encoder, and Roformer is a transformer with rotational position embeddings.

[0131] 303. The computer device decodes the antigen sequence features through the decoding sub-model in the antibody generation model to obtain the predicted probability of each position in the antibody sequence.

[0132] The computer device inputs the antigen sequence features into the decoding sub-model of the antibody generation model. The decoding sub-model decodes the antigen sequence features and outputs the predicted probability of each position.

[0133] By introducing an antibody generation model based on antigen sequence information, the encoding sub-model included therein can be used to accurately extract the antigen-related features, and then the decoding sub-model can be used to accurately determine the predicted probability of possible amino acids at each position in the antigen sequence based on these features.

[0134] In one possible implementation, the antigen sequence features include target antigen bond features and target antigen value features. The computer device decodes the target antigen bond features, target antigen value features, and reference amino acid features at each position using a decoding sub-model to obtain a predicted probability for each position. The reference amino acid feature at a position refers to the feature of the amino acid preceding that position.

[0135] That is, when predicting the amino acid at the current position, in addition to utilizing the target antigen bond feature and the target antigen value feature, the predicted amino acid features at other positions prior to the current position are also utilized. The predicted amino acid features at other positions prior to the current position are used as reference amino acid features to achieve reference-based amino acid prediction, which is beneficial for improving prediction accuracy. The amino acid feature may be an amino acid identifier or an amino acid token, etc., which is not limited in the present embodiment.

[0136] When the antigen sequence features determine the accurate target antigen key features and target antigen value features through the encoding sub-model, the decoding sub-model can perform feature decoding on the reference amino acid features at each position based on this, effectively improving the decoding quality.

[0137] In the embodiment of the present application, the prediction of the amino acid at the first position and the prediction of the amino acid at a non-first position are divided into the following two cases.

[0138] The first case: the computer device determines the preset amino acid feature as the reference amino acid feature of the first position, and through the decoding sub-model, performs feature decoding on the target antigen bond feature, the target antigen value feature and the reference amino acid feature of the first position to obtain the predicted probability at the first position.

[0139] For the first position, the computer device uses a preset amino acid feature as a reference amino acid feature for the first position. The preset amino acid feature can be regarded as an initial value. The preset amino acid feature can be freely set by the R&D personnel, or it can be obtained by training as a model parameter of the antibody generation model. The embodiments of the present application do not limit this.

[0140] The second case: the computer device determines the amino acid feature of the amino acid with the highest predicted probability at the first i positions as the reference amino acid feature at the i+1th position, and performs feature decoding on the target antigen bond feature, the target antigen value feature and the reference amino acid feature at the i+1th position through the decoding sub-model to obtain the predicted probability at the i+1th position, where i is a positive integer.

[0141] For each position after the first, the computer determines the predicted amino acid signature as the reference amino acid signature for that position. Therefore, when predicting the amino acid at each position, the amino acid at at least one position before it is referenced, which helps improve prediction accuracy.

[0142] The above description uses the amino acid prediction process at the first and i+1 positions as examples. The complete processing of the decoding sub-model sequentially obtains the prediction probability for each position, that is, predicts the amino acid at each position, until the prediction probability for the last position is obtained. The detailed process is as follows: the computer device determines the preset amino acid signature as the reference amino acid signature for the first position. The decoding sub-model performs feature decoding on the target antigen bond signature, target antigen value signature, and the reference amino acid signature at the first position to obtain the prediction probability for the first position. The amino acid signature of the amino acid with the highest prediction probability at the first position is determined as the reference amino acid signature for the second position. The decoding sub-model performs feature decoding on the target antigen bond signature, target antigen value signature, and the reference amino acid signature at the second position to obtain the prediction probability for the second position. The computer device determines the amino acid signature of the amino acids with the highest prediction probability at the first two positions as the reference amino acid signature for the third position. The decoding sub-model performs feature decoding on the target antigen bond signature, target antigen value signature, and the reference amino acid signature at the third position to obtain the prediction probability for the third position. The computer device determines the amino acid signature of the amino acid with the highest predicted probability at the first three positions as the reference amino acid signature for the fourth position. Using the decoding sub-model, the computer device decodes the target antigen bond signature, target antigen value signature, and the reference amino acid signature at the fourth position to obtain the predicted probability for the fourth position. This process continues until the computer device obtains the predicted probability for the last position using the decoding sub-model.

[0143] Since the combination of amino acids has certain rules, when decoding the features of each position in sequence, the amino acid feature with the highest predicted probability at the previously decoded position will be used as the reference amino acid feature for this position, thereby improving the rationality and accuracy of the predicted probability.

[0144] In one possible implementation, the decoding submodel includes h decoding networks and an output network, where h is a positive integer. For each position, the computer device performs feature decoding on the target antigen bond feature, the target antigen value feature, and the reference amino acid feature of the position through the first decoding network in the decoding submodel to obtain the first antibody sequence feature. The target antigen bond feature, the target antigen value feature, and the k-1th antibody sequence feature are feature decoded through the kth decoding network in the decoding submodel to obtain the kth antibody sequence feature, where k is a positive integer greater than 1 and not greater than h. The computer device normalizes the hth antibody sequence feature through the output network in the decoding submodel to obtain the predicted probability of the position.

[0145] The above description only uses the processing of the first and kth decoding networks in the decoding sub-model as an example. The complete processing process of the decoding sub-model is to input the target antigen bond feature, target antigen value feature, and the reference amino acid feature of the position into the first decoding network for processing. Subsequently, each decoding network is processed in sequence until the last decoding network completes the processing. The detailed process is as follows: the computer device decodes the target antigen bond feature, target antigen value feature, and the reference amino acid feature of the position through the first decoding network in the decoding sub-model to obtain the first antibody sequence feature. The computer device extracts the target antigen bond feature, target antigen value feature, and the first antibody sequence feature through the second decoding network in the decoding sub-model to obtain the second antibody sequence feature. The computer device continues to extract the target antigen bond feature, target antigen value feature, and the second antibody sequence feature through the third decoding network in the decoding sub-model to obtain the third antibody sequence feature. And so on, until the computer device obtains the last antibody sequence feature through the last decoding network in the decoding sub-model. Then, the output network in the decoding sub-model normalizes the last antibody sequence feature to obtain the predicted probability of the position.

[0146] The target antigen key features, target antigen value features and reference amino acid features at each position are feature decoded through h encoding networks. Each decoding network will perform its own feature decoding based on the decoding results of the previous decoding network, so that the extraction results of one decoding network will be inherited by the next decoding network, effectively improving the decoding accuracy.

[0147] In one possible implementation, the decoding network includes a self-attention layer and a cross-attention layer. Taking the first decoding network as an example, the computer device, through the first decoding network in the decoding sub-model, performs feature decoding on the target antigen bond feature, the target antigen value feature, and the reference amino acid feature at the position to obtain a first antibody sequence feature. This includes: performing feature decoding on the reference amino acid feature at the position through the self-attention layer to obtain an antibody query feature; and fusing the target antigen bond feature, the target antigen value feature, and the antibody query feature through the cross-attention layer to obtain the first antibody sequence feature.

[0148] In an embodiment of the present application, when decoding is performed in the decoding network, the antigen bond features and antigen value features output by the encoding network are cross-fused with the antigen query features output by the decoding network, which is conducive to learning the correlation between the antigen sequence and the antibody sequence, and improves the accuracy of the acquired antibody sequence features.

[0149] Optionally, the computer device decodes the reference amino acid feature of the position through a self-attention layer to obtain the antibody query feature, including: determining the product of the reference amino acid feature of the position and the parameter matrix as an antibody sequence matrix through the self-attention layer, and obtaining the antibody query feature, antibody bond feature and antibody value feature based on the antibody sequence matrix.

[0150] Among them, the antibody query feature, antibody bond feature, and antibody value feature in the embodiment of the present application belong to different feature spaces respectively. Each amino acid in the antibody sequence has its own antibody bond feature, antibody value feature, and antibody query feature. The antibody value feature represents the characteristics of the amino acid itself. The antibody query feature and antibody bond feature are used to determine the feature vector of the attention weight. The antibody query feature obtained by the self-attention layer is used to cross-fuse with the antigen bond feature and antigen value feature output by the encoding network.

[0151] The parameter matrix is ​​used for spatial transformation, and the parameter matrix is ​​the model parameter obtained through training. After obtaining the antibody sequence matrix, the computer device splits the antibody sequence matrix into antibody query features, antibody bond features, and antibody value features. For example, the parameter matrix is ​​a 3-dimensional matrix, and the antibody sequence matrix obtained by multiplying the antibody sequence matrix and the parameter matrix is ​​a 3-dimensional matrix. The computer device uses each dimension of the antibody sequence matrix as an antibody query feature, an antibody bond feature, and an antibody value feature.

[0152] Optionally, the computer device fuses the target antigen bond feature, the target antigen value feature and the antibody query feature through a cross attention layer to obtain an antibody sequence feature, including: the computer device normalizes the product of the antibody query feature, the target antigen bond feature and the scaling factor through a cross attention layer to obtain a normalized feature, and determines the product of the normalized feature and the target antigen value feature as the antibody sequence feature. The computer device obtains a scaling factor, which represents a normalized scaling factor. In an embodiment of the present application, the computer device uses a scaling factor as a normalization parameter to normalize the product to obtain a normalized feature, which characterizes the correlation between the antibody query feature and the target antigen bond feature. The computer device can use the normalized feature as the weight of the target antigen value feature, so the computer device determines the product between the normalized feature and the target antigen value feature as the antibody sequence feature.

[0153] In one possible implementation, the decoding sub-model is a BERT (Bidirectional Encoder Representations from Transformers) model or a Roformer (Rotary Transformers) model. BERT is obtained by pre-training a deep bidirectional Transformer encoder, and Roformer is a transformer with rotational position embeddings.

[0154] 304. The computer device determines the sequence information composed of the amino acids with the highest predicted probability at each position as the antibody sequence information. The antibody sequence indicated by the antibody sequence information is capable of binding to the antigen sequence.

[0155] Each position corresponds to a predicted probability of each amino acid among multiple amino acids. For a certain position, the maximum predicted probability is determined among the predicted probabilities of each amino acid among multiple amino acids, and the amino acid with the maximum predicted probability is determined as the amino acid at that position. This method is used to determine the amino acid at each position, and the sequence information composed of the amino acids at multiple positions is the antibody sequence information.

[0156] In the examples of the present application, the relationship between antigen sequences and antibody sequences is analogized to the relationship between language translation. It is assumed that the antigen sequence and the antibody sequence have similar semantic structures, just like the vocabulary and grammatical rules between different languages. Therefore, the process of converting the antigen sequence into the antibody sequence can be regarded as the process of translating a text in one language into another. Therefore, we can learn the semantic rules of antigen sequences and antibody sequences through the antibody generation model, so as to realize the generation of antibody sequences that can be combined with them using the antibody generation model.

[0157] The method provided in the embodiments of the present application takes into account that when the antibody sequence is able to bind to the antigen sequence, the amino acids in the antigen sequence should have a certain correlation with the amino acids in the antibody sequence. Therefore, from the perspective of antigen sequence information, based on the antigen sequence information that can represent the amino acids in the antigen sequence, the antigen sequence characteristics are extracted, and based on the antigen sequence characteristics, the antibody sequence information corresponding to the antibody sequence that can bind to the antigen sequence is directly decoded, thereby generating an antibody sequence that can bind to a specific antigen sequence, thereby improving the specificity of the generated antibody sequence.

[0158] FIG4 is a flow chart of a method for training an antibody production model provided in an embodiment of the present application. The embodiment of the present application is executed by a computer device. Referring to FIG4 , the method includes:

[0159] 401. A computer device obtains sample antigen sequence information and a true probability at each position in a sample antibody sequence, where the sample antigen sequence information represents amino acids in the sample antigen sequence, and the true probability includes a true probability that the amino acid at the position belongs to each of a plurality of amino acids, and the sample antibody sequence is able to bind to the sample antigen sequence.

[0160] The true probability of each position in the sample antibody sequence includes the true probability of the amino acid at each position belonging to each of the multiple amino acids. For example, if the amino acid at a position in the sample antibody sequence is A, the true probability of amino acid A at that position is 1, and the true probability of any other amino acid at that position is 0.

[0161] The process of step 401 is the same as that of step 301 above, and will not be repeated here.

[0162] 402. The computer device extracts features of the sample antigen sequence information using the encoding sub-model in the antibody generation model to obtain sample antigen sequence features.

[0163] The computer device inputs the sample antigen sequence information into the encoding sub-model of the antibody generation model, and the encoding sub-model extracts the features of the sample antigen sequence information and outputs the sample antigen sequence features.

[0164] In one possible implementation, during the feature extraction process of the sample antigen sequence information by the encoding sub-model, multiple sample antigen key features, sample antigen value features, and sample antigen query features are generated. The sample antigen sequence features in the embodiment of the present application include target sample antigen key features and target sample antigen value features. The encoding sub-model includes m encoding networks, where m is a positive integer. The process of the computer device obtaining the target sample antigen key features and target sample antigen value features through the encoding sub-model includes:

[0165] The computer device performs feature extraction on the sample antigen sequence information using the first encoding network in the encoding sub-model to obtain a first candidate sample antigen key feature, a first candidate sample antigen value feature, and a first candidate sample antigen query feature. The first candidate sample antigen key feature, the first candidate sample antigen value feature, and the first candidate sample antigen query feature are fused to obtain a first sample antigen fusion feature. The computer device performs feature extraction on the nth sample antigen fusion feature using the n+1th encoding network in the encoding sub-model to obtain the n+1th candidate sample antigen key feature, the n+1th candidate sample antigen value feature, and the n+1th candidate sample antigen query feature, where n is a positive integer less than m. The computer device determines the mth candidate sample antigen key feature and the mth candidate sample antigen value feature as the target sample antigen key feature and the target sample antigen value feature.

[0166] In one possible implementation, each encoding network includes a self-attention layer and a feedforward neural network layer. Taking the first encoding network as an example, the computer device uses the self-attention layer in the encoding network to extract features from the sample antigen sequence information to obtain candidate sample antigen key features, candidate sample antigen value features, and candidate sample antigen query features. The candidate sample antigen key features, candidate sample antigen value features, and candidate sample antigen query features are then fused to obtain an initial sample antigen fusion feature. The computer device uses the feedforward neural network layer in the encoding network to perform feature transformation on the initial sample antigen fusion feature to obtain a sample antigen fusion feature.

[0167] Optionally, the process of the computer device obtaining the candidate sample antigen query feature, the candidate sample antigen key feature and the candidate sample antigen value feature includes: determining the product of the sample antigen sequence information and the parameter matrix as the sample antigen sequence matrix, and obtaining the candidate sample antigen query feature, the candidate sample antigen key feature and the candidate sample antigen value feature corresponding to the sample antigen sequence information based on the sample antigen sequence matrix.

[0168] Optionally, the computer device fuses the candidate sample antigen bond feature, the candidate sample antigen value feature and the candidate sample antigen query feature to obtain the initial sample antigen fusion feature, including: normalizing the product of the candidate sample antigen query feature, the candidate sample antigen bond feature and the scaling factor to obtain a normalized feature, and determining the product of the normalized feature and the candidate sample antigen value feature as the initial sample antigen fusion feature.

[0169] The process of determining the sample antigen sequence characteristics in step 402 is similar to the process of determining the antigen sequence characteristics in step 302 above, and each optional solution in step 402 will not be explained in detail here.

[0170] 403. The computer device decodes the sample antigen sequence features through the decoding sub-model in the antibody generation model to obtain the sample prediction probability at each position, and the sample prediction probability includes the prediction probability of the amino acid at the position belonging to each of the multiple amino acids.

[0171] The computer device inputs the sample antigen sequence features into the decoding sub-model of the antibody generation model. The decoding sub-model decodes the sample antigen sequence features and outputs the sample prediction probability at each position.

[0172] In one possible implementation, the sample antigen sequence features include target sample antigen bond features and target sample antigen value features. The computer device, through a decoding sub-model, decodes the target sample antigen bond features, target sample antigen value features, and the sample reference amino acid features at each position to obtain the sample prediction probability at each position. The sample reference amino acid feature at a position refers to the feature of the amino acid preceding that position.

[0173] Optionally, the computer device determines a preset amino acid feature as the sample reference amino acid feature at the first position, and performs feature decoding on the target sample antigen bond feature, the target sample antigen value feature, and the sample reference amino acid feature at the first position through a decoding sub-model to obtain a sample prediction probability at the first position. The computer device determines the amino acid feature of the amino acid with the highest sample prediction probability at the first i positions as the sample reference amino acid feature at the i+1th position, and performs feature decoding on the target sample antigen bond feature, the target sample antigen value feature, and the sample reference amino acid feature at the i+1th position through a decoding sub-model to obtain a sample prediction probability at the i+1th position, where i is a positive integer.

[0174] In one possible implementation, the decoding submodel includes h decoding networks and an output network, where h is a positive integer. For each position, the computer device performs feature decoding on the target sample antigen bond feature, the target sample antigen value feature, and the sample reference amino acid feature at that position through the first decoding network in the decoding submodel to obtain the first sample antibody sequence feature. The target sample antigen bond feature, the target sample antigen value feature, and the k-1th sample antibody sequence feature are feature decoded through the kth decoding network in the decoding submodel to obtain the kth sample antibody sequence feature, where k is a positive integer greater than 1 and not greater than h. The computer device normalizes the hth sample antibody sequence feature through the output network in the decoding submodel to obtain the sample prediction probability at that position.

[0175] In one possible implementation, the decoding network includes a self-attention layer and a cross-attention layer. Taking the first decoding network as an example, the computer device uses the self-attention layer to decode the sample reference amino acid feature at that position to obtain the sample antibody query feature. The cross-attention layer then fuses the target sample antigen bond feature, target sample antigen value feature, and sample antibody query feature to obtain the first sample antibody sequence feature.

[0176] The process of determining the sample prediction probability of each position in step 403 is the same as the process of determining the prediction probability of each position in step 303 above, and each optional scheme in step 403 will not be explained in detail here.

[0177] 404. The computer device trains the antibody generation model based on the predicted probability and the true probability of the sample at each position.

[0178] The computer device obtains the true probability at each position, which represents the true probability of each amino acid at each position in the sample antibody sequence. The sample predicted probability obtained by the computer device is predicted by the antibody generation model. The computer device trains the antibody generation model based on the difference between the sample predicted probability and the true probability.

[0179] Since the goal of an antibody generation model is to predict the true probability at each position in a sample antibody sequence, the closer the sample predicted probability is to the true probability, the more accurate the antibody generation model. The computer device then trains the antibody generation model based on the difference between the sample predicted probability and the true probability, reducing the difference between the sample predicted probability and the true probability obtained by the trained antibody generation model. This improves the predictive ability of the antibody generation model, thereby increasing the accuracy of the antibody generation model.

[0180] In one possible implementation, the computer device repeats steps 401 to 404 to iteratively train the antibody generation model. In response to the number of iterations reaching a first threshold, the computer device stops training the antibody generation model. Alternatively, in response to the loss value obtained in the current iteration being no greater than a second threshold, the computer device stops training the antibody generation model. The first and second thresholds are both arbitrary values, for example, the first threshold is 1000 or 1500, and the second threshold is 0.004 or 0.003.

[0181] Fig. 5 is a schematic diagram of a method for generating antibody sequence information provided by an embodiment of the present application, wherein the antibody generation model includes an encoding sub-model and a decoding sub-model as shown in Fig. 5, and the encoding sub-model includes multiple encoding networks (taking 3 as an example in the figure), and the decoding sub-model includes multiple decoding networks (taking 3 as an example in the figure), and the input of the encoding sub-model is antigen sequence information. The computer device inputs the antigen bond feature (K) and antigen value feature (V) output by the last encoding network into each decoding network of the decoding sub-model, and the last decoding network outputs the predicted probability of each position, and the antibody sequence information can be obtained according to the predicted probability of each position. As shown in Fig. 5, each decoding network includes a self-attention layer and a cross attention layer, and the computer device inputs the antibody query feature (Q) output by the self-attention layer into the cross attention layer, and the antigen bond feature (K) and antigen value feature (V) output by the encoding sub-model are input into the cross attention layer, and the cross attention layer cross-fuses the antibody query feature (Q), antigen bond feature (K) and antigen value feature (V). Among them, the model parameters of the cross-attention layer can be randomly initialized model parameters, and the model parameters in the remaining network structures can be pre-trained model parameters. The pre-training process can be referred to the embodiments shown in Figures 6 and 8 below.

[0182] The method provided in the embodiment of the present application compares the relationship between the antigen sequence and the antibody sequence to the relationship between language translation. When training the antibody generation model, taking into account the case where the antibody sequence can be combined with the antigen sequence, the amino acids in the antigen sequence and the amino acids in the antibody sequence should have a certain correlation. Therefore, from the perspective of the antigen sequence information, according to the antigen sequence information that can represent the amino acids in the antigen sequence, the antigen sequence features are extracted, and based on the antigen sequence features, the antibody sequence information corresponding to the antibody sequence that can be combined with the antigen sequence is directly decoded, thereby learning the semantic rules of the antigen sequence and the antibody sequence through the antibody generation model, and realizing the generation of the antibody sequence that can be combined with it using the antibody generation model.

[0183] The antibody generation model includes an encoding sub-model and a decoding sub-model. Prior to training, the encoding sub-model and the decoding sub-model can be untrained or pre-trained. After pre-training, the pre-trained encoding sub-model and the decoding sub-model are used to construct the antibody generation model, which is then trained.

[0184] Pre-training is the process of unsupervised learning using large datasets. By training on large amounts of unlabeled data, the model automatically learns patterns and regularities in the data, enabling it to better handle various tasks. Pre-training is typically performed using neural networks, which extract high-level feature representations of the data through a layer-by-layer learning process.

[0185] In an embodiment of the present application, the encoding sub-model is pre-trained on a data set of large-scale antigen sequences (or protein sequences), and the decoding sub-model is pre-trained on a data set of large-scale antibody sequences (or heavy chains of antibody sequences). The pre-trained encoding sub-model and decoding sub-model are applied to the construction of the antibody generation model. For the encoding sub-model, it includes a plurality of encoding networks, and the model parameters of the encoding network obtained by pre-training can be directly applied to the construction of the antibody generation model. For the decoding sub-model, it includes a plurality of decoding networks, and the model parameters of the self-attention layer of each decoding network obtained by pre-training can be directly applied to the construction of the antibody generation model, and the randomly initialized model parameters of the cross-attention layer of each decoding network are applied to the construction of the antibody generation model.

[0186] Among them, the pre-training process of the encoding sub-model and the decoding sub-model refers to the embodiments shown in Figures 6 and 8 below.

[0187] FIG6 is a flowchart of a pre-training method for a decoding sub-model provided in an embodiment of the present application. The embodiment of the present application is executed by a computer device. Referring to FIG6 , the method includes:

[0188] 601. A computer device obtains first antibody sequence information, masks amino acid information at some positions in the first antibody sequence information, and obtains second antibody sequence information.

[0189] The first antibody sequence information may be antibody sequence information of any antibody sequence, wherein the first antibody sequence information represents amino acids in the antibody sequence, and the antibody sequence includes multiple amino acids. The first antibody sequence information includes amino acid information at multiple positions, wherein the amino acid information at each position represents the amino acid at that position. Alternatively, the antibody sequence information may include sequence information of a heavy chain in the antibody sequence, etc.

[0190] The computer device randomly masks the amino acid information at some positions in the first antibody sequence information to obtain the second antibody sequence information. The amino acid information at some positions in the second antibody sequence information is unknown. For example, the computer device may randomly mask 15% of the amino acid information in the first antibody sequence information.

[0191] In one possible implementation, the computer device replaces amino acid information at certain positions in the first antibody sequence information with preset characters, such as [mask]. For example, if the first antibody sequence information is "MDVLY," the "V" and "Y" in the first antibody sequence information are masked, that is, "V" and "Y" are replaced with [mask], resulting in the second antibody sequence information being "MD[mask]L[mask]."

[0192] 602. The computer device predicts the amino acid information at the masked position in the second antibody sequence information through the decoding sub-model to obtain a first prediction probability, where the first prediction probability represents the predicted probability that the amino acid at the masked position in the first antibody sequence information belongs to each amino acid.

[0193] The computer device inputs the second antibody sequence information into the decoding sub-model, and the decoding sub-model predicts the amino acid information at the masked position in the second antibody sequence information based on the unmasked amino acid information in the second antibody sequence information and the context semantics, and outputs a first prediction probability.

[0194] In one possible implementation, the computer device determines the antibody sequence features of the second antibody sequence information using the decoding sub-model, and performs a prediction based on the antibody sequence features to obtain a first prediction probability. The process of extracting the antibody sequence features of the second antibody sequence information by the decoding sub-model is similar to the process of extracting the antibody sequence features in the embodiment shown in FIG. 4 , and is not further described here.

[0195] 603. The computer device trains a decoding sub-model based on the first predicted probability and the first true probability, where the first true probability represents the true probability that the amino acid at the masked position in the first antibody sequence information belongs to each amino acid.

[0196] The computer device obtains a first true probability, which represents the true probability that the amino acid at the masked position in the first antibody sequence information belongs to each amino acid. The first predicted probability obtained by the computer device is predicted by the decoding sub-model. The computer device trains the decoding sub-model based on the difference between the first predicted probability and the first true probability.

[0197] Since the goal of the decoding sub-model is to predict the true probability of the amino acid at the masked position in the first antibody sequence information belonging to each amino acid, the more similar the first predicted probability is to the first true probability, the stronger the decoding sub-model's ability to extract antibody sequence features. The computer device then trains the first predicted probability and the first true probability based on the difference between the first predicted probability and the first true probability, thereby reducing the difference between the first predicted probability and the first true probability obtained by the trained decoding sub-model, thereby improving the feature extraction capability of the decoding sub-model, and thus enabling the decoding sub-model to initially learn to extract antibody sequence features.

[0198] In one possible implementation, the computer device determines a first loss parameter based on the first predicted probability and the first true probability, and trains the decoding sub-model based on the first loss parameter to reduce the first loss parameter obtained by the trained decoding sub-model. The greater the difference between the first predicted probability and the first true probability, the greater the first loss parameter, and the smaller the difference between the first predicted probability and the first true probability, the smaller the first loss parameter.

[0199] Optionally, the computer device determines the first loss parameter using the following formula:

[0200] in, represents the first loss parameter, y represents the first true probability, represents the first predicted probability.

[0201] FIG7 is a schematic diagram of a pre-training process for a decoding sub-model provided in an embodiment of the present application. As shown in FIG7 , the first antibody sequence information is “[cls]MDVLY…[eos],” and the second antibody sequence information obtained after masking is “[cls]MD[mask]L[mask]…[eos].” The computer device inputs the second antibody sequence information “[cls]MD[mask]L[mask]…[eos]” into the decoding sub-model, and trains the decoding sub-model so that the first antibody sequence information “[cls]MDVLY…[eos]” can be obtained based on the predicted probability output by the trained decoding sub-model.

[0202] The method provided in the embodiment of the present application pre-trains the decoding sub-model in a self-supervised manner based on the sample antibody sequence information, thereby improving the ability of the decoding sub-model to extract features based on the antibody sequence information. The pre-trained decoding sub-model is then used to construct an antibody generation model, which is beneficial to reducing the training pressure of the antibody generation model and ensuring the accuracy of the antibody generation model.

[0203] FIG8 is a flowchart of a pre-training method for an encoding sub-model provided in an embodiment of the present application. The embodiment of the present application is executed by a computer device. Referring to FIG8 , the method includes:

[0204] 801. A computer device obtains first antigen sequence information, masks amino acid information at some positions in the first antigen sequence information, and obtains second antigen sequence information.

[0205] The first antigen sequence information can be the antigen sequence information of any antigen sequence, the first antigen sequence information represents the amino acids in the antigen sequence, and the antigen sequence includes multiple amino acids. The first antigen sequence information includes amino acid information at multiple positions, and the amino acid information at each position represents the amino acid at that position.

[0206] The computer randomly masks the amino acid information at some positions in the first antigen sequence information to obtain the second antigen sequence information. The amino acid information at some positions in the second antigen sequence information is unknown. The masking method for the first antigen sequence information is similar to the masking method for the first antibody sequence information in the embodiment shown in FIG. 6 , and will not be further described here.

[0207] 802. The computer device predicts the amino acid information at the masked position in the second antigen sequence information through the encoding sub-model to obtain a second prediction probability, where the second prediction probability represents the predicted probability that the amino acid at the masked position in the first antigen sequence information belongs to each amino acid.

[0208] The computer device inputs the second antigen sequence information into the encoding sub-model, which predicts the amino acid information at the masked position in the second antigen sequence information based on the unmasked amino acid information in the second antigen sequence information and combined with the context semantics, and outputs a second prediction probability.

[0209] In one possible implementation, during the pre-training process, the encoding sub-model further includes an output network compared to the antibody generation model. After completing the pre-training of the encoding sub-model, the output network in the encoding sub-model is removed, and the encoding sub-model after removing the output network is used to construct the antibody generation model. The computer device determines the antigen sequence features of the second antigen sequence information through the encoding sub-model, and makes a prediction based on the antigen sequence features to obtain a second prediction probability. The process of extracting the antigen sequence features of the second antigen sequence information by the encoding sub-model is the same as the process of extracting the antigen sequence features in the embodiment shown in FIG4 above, and will not be repeated here.

[0210] 803. The computer device trains the encoding sub-model based on the second predicted probability and the second true probability, where the second true probability represents the true probability that the amino acid at the masked position in the first antigen sequence information belongs to each amino acid.

[0211] The computer device obtains a second true probability, which represents the true probability that the amino acid at the masked position in the first antigen sequence information belongs to each amino acid. The second predicted probability obtained by the computer device is predicted by the encoding sub-model. The computer device trains the encoding sub-model based on the difference between the second predicted probability and the second true probability.

[0212] Since the goal of the encoding sub-model is to predict the true probability of each amino acid at the masked position in the first antigen sequence information, the more similar the second predicted probability is to the second true probability, the stronger the encoding sub-model's ability to extract antigen sequence features. The computer device then trains the encoding sub-model based on the difference between the second predicted probability and the second true probability, thereby reducing the difference between the second predicted probability and the second true probability obtained by the trained encoding sub-model and improving the encoding sub-model's feature extraction ability, thereby enabling the encoding sub-model to initially learn to extract antigen sequence features.

[0213] In one possible implementation, the computer device determines a second loss parameter based on the second predicted probability and the second true probability, and trains the encoding sub-model based on the second loss parameter to reduce the second loss parameter obtained by the trained encoding sub-model. The greater the difference between the second predicted probability and the second true probability, the greater the second loss parameter, and the smaller the difference between the second predicted probability and the second true probability, the smaller the second loss parameter. The process of determining the second loss parameter by the computer device is similar to the process of determining the first loss parameter in step 603 above and is not further described here.

[0214] The method provided in the embodiment of the present application pre-trains the encoding sub-model in a self-supervised manner based on the sample antigen sequence information, thereby improving the ability of the encoding sub-model to extract features based on the antigen sequence information. The pre-trained encoding sub-model is then used to construct an antibody generation model, which is beneficial to reducing the training pressure of the antibody generation model and ensuring the accuracy of the antibody generation model.

[0215] In order to verify the effectiveness of the embodiments of the present application, the method provided in the embodiments of the present application was subjected to a performance test. The stable peptide segment of the HR2 region of the antigen in the 14H data set was used as the antigen sequence information input into the antibody generation model, and a CDRH3 antibody sequence that can bind to the stable peptide segment in the HR2 region was generated. The complex structure of the generated antibody sequence bound to the antigen sequence and the complex structure of the natural antibody sequence bound to the antigen sequence were obtained. The results are shown in Figure 9. The upper half of Figure 9 is the complex structure of the generated antibody sequence bound to the antigen sequence, and the lower half of Figure 9 is the complex structure of the natural antibody sequence bound to the antigen sequence. As can be seen from Figure 9, the generated antibody sequence can bind to the CDRH3 region and the stable region of the antigen sequence. The natural antibody sequence will target other sites of the antigen sequence. The antibody sequence binding to the stable region of the antigen sequence represents a more powerful applicability and can avoid the situation where the new variant causes the antibody sequence to fail.

[0216] In addition, the antibody generation model is trained on the Cov-AbDab dataset, and then attempts are made to generate antibody sequences that do not exist in the Cov-AbDab dataset. The posture of the complex structure bound by the antibody sequence and antigen sequence is evaluated to obtain the interface energy (Interface Energy) and the root mean square deviation (RMSD) of the interface for each complex structure. The interface energy is a key indicator for evaluating the interaction between molecular ligands (small molecules) and target proteins during molecular docking. The interface energy can be decomposed into the sum of various interaction energies, including Coulomb interaction energy, van der Waals interaction energy, hydrogen bond energy, etc. between molecules. Among them, the interface energy is an important component of describing the non-covalent interaction energy between ligands and proteins. The root mean square error is used to measure the degree of difference between two structures. The smaller the root mean square error, the smaller the distance between the two structures after overlapping, that is, the more similar the structures are.

[0217] Figure 10 is a comparison diagram of a kind of antibody performance that the present application embodiment provides, as shown in Figure 10, the curve graph a on the left is the binding surface energy of the antibody sequence (Artifical) generated and the natural antibody sequence (Natural), and the curve graph b on the right is the root mean square error of the antibody sequence generated and the natural antibody sequence.The binding surface energy and the root mean square error of the antibody sequence generated are both better than the binding surface energy and the root mean square error of the natural antibody sequence, which shows that the antibody sequence generated has a higher affinity for the antigen sequence.The structural diagram c above in Figure 10 is the complex structure that the antibody sequence generated is combined with the antigen sequence, and the structural diagram d below is the complex structure that the natural antibody sequence is combined with the antigen sequence, and the antibody sequence generated is mainly combined with the antigen sequence in a continuous region of the light chain, and the natural antibody sequence has discontinuous partial binding sites on the light chain and the heavy chain.The results show that the antibody generation model provided in the present application embodiment can quickly generate potential high-affinity antibody sequences for the antigen sequence of new variants.

[0218] FIG11 is a schematic diagram of the structure of an antibody sequence generation device provided in an embodiment of the present application. Referring to FIG11 , the device includes:

[0219] The information acquisition module 1101 is used to obtain antigen sequence information, where the antigen sequence information represents amino acids in the antigen sequence;

[0220] A feature extraction module 1102 is used to extract features from the antigen sequence information to obtain antigen sequence features of the antigen sequence information;

[0221] Feature decoding module 1103, for decoding the antigen sequence features to obtain the predicted probability of each position in the antibody sequence, including the predicted probability of the amino acid at the position belonging to each of the multiple amino acids;

[0222] The information determination module 1104 is configured to determine the sequence information composed of the amino acids with the highest predicted probability at each position as the antibody sequence information. The antibody sequence indicated by the antibody sequence information is capable of binding to the antigen sequence.

[0223] The antibody sequence generation device provided in the embodiments of the present application takes into account that when the antibody sequence is capable of binding to the antigen sequence, the amino acids in the antigen sequence should have a certain correlation with the amino acids in the antibody sequence. Therefore, from the perspective of antigen sequence information, based on the antigen sequence information that can represent the amino acids in the antigen sequence, the antigen sequence features are extracted, and based on the antigen sequence features, the antibody sequence information corresponding to the antibody sequence that can bind to the antigen sequence is directly decoded, thereby generating an antibody sequence that can bind to a specific antigen sequence, thereby improving the targetedness of the generated antibody sequence.

[0224] Optionally, the predicted probability is obtained by an antibody generation model, which includes an encoding sub-model and a decoding sub-model;

[0225] A feature extraction module 1102 is used to extract features from the antigen sequence information using the encoding sub-model to obtain antigen sequence features of the antigen sequence information;

[0226] The feature decoding module 1103 is used to perform feature decoding on the antigen sequence features through the decoding sub-model to obtain the predicted probability of each position.

[0227] Optionally, the antigen sequence features include target antigen key features and target antigen value features, the encoding sub-model includes m encoding networks, where m is a positive integer; the feature extraction module 1102 is used to:

[0228] The first encoding network in the encoding sub-model is used to extract features of the antigen sequence information to obtain a first candidate antigen bond feature, a first candidate antigen value feature, and a first candidate antigen query feature. The first candidate antigen bond feature, the first candidate antigen value feature, and the first candidate antigen query feature are fused to obtain a first antigen fusion feature.

[0229] The nth antigen fusion feature is extracted through the n+1th encoding network in the encoding sub-model to obtain the n+1th candidate antigen key feature, the n+1th candidate antigen value feature and the n+1th candidate antigen query feature, where n is a positive integer less than m;

[0230] The mth candidate antigen bond feature and the mth candidate antigen value feature are determined as the target antigen bond feature and the target antigen value feature.

[0231] Optionally, the feature decoding module 1103 is configured to:

[0232] Through the decoding sub-model, the target antigen bond feature, target antigen value feature and the reference amino acid feature of each position are feature decoded to obtain the predicted probability of each position. The reference amino acid feature refers to the feature of the amino acid located before the position.

[0233] Optionally, for the first position in the antibody sequence, the feature decoding module 1103 is configured to:

[0234] determining the preset amino acid feature as the reference amino acid feature at the first position;

[0235] The target antigen bond feature, target antigen value feature and reference amino acid feature at the first position are feature decoded by the decoding sub-model to obtain the predicted probability at the first position.

[0236] Optionally, for the i+1th position in the antibody sequence, the feature decoding module 1103 is configured to:

[0237] The amino acid feature of the amino acid with the highest predicted probability at the first i positions is determined as the reference amino acid feature at the i+1th position. The target antigen bond feature, target antigen value feature and the reference amino acid feature at the i+1th position are feature decoded through the decoding sub-model to obtain the predicted probability at the i+1th position, where i is a positive integer.

[0238] Optionally, the decoding sub-model includes h decoding networks and an output network, where h is a positive integer. The feature decoding module 1103 is configured to:

[0239] For each position, the first decoding network in the decoding sub-model is used to perform feature decoding on the target antigen bond feature, the target antigen value feature, and the reference amino acid feature of the position to obtain the first antibody sequence feature;

[0240] The target antigen key feature, the target antigen value feature and the k-1th antibody sequence feature are decoded by the k-th decoding network in the decoding sub-model to obtain the k-th antibody sequence feature, where k is a positive integer greater than 1 and not greater than h;

[0241] The h-th antibody sequence feature is normalized through the output network in the decoding sub-model to obtain the predicted probability of the position.

[0242] Optionally, the decoding network includes a self-attention layer and a cross-attention layer, and a feature decoding module 1103 for:

[0243] Through the self-attention layer, the reference amino acid features of the position are feature decoded to obtain the antibody query features;

[0244] Through the cross-attention layer, the target antigen bond features, target antigen value features and antibody query features are fused to obtain the first antibody sequence feature.

[0245] Optionally, referring to FIG12 , the apparatus further includes a first training module 1105 configured to:

[0246] Obtaining the true probability at each position in the sample antigen sequence information and the sample antibody sequence, where the sample antigen sequence information represents the amino acids in the sample antigen sequence, and the true probability includes the true probability that the amino acid at the position belongs to each of the multiple amino acids, and the sample antibody sequence can bind to the sample antigen sequence;

[0247] Through the encoding sub-model in the antibody generation model, the sample antigen sequence information is extracted to obtain the sample antigen sequence characteristics;

[0248] The decoding sub-model in the antibody generation model decodes the sample antigen sequence features to obtain the sample prediction probability at each position. The sample prediction probability includes the predicted probability of the amino acid at the position belonging to each of the multiple amino acids.

[0249] The antibody generation model is trained based on the sample predicted probability and true probability at each position.

[0250] Optionally, referring to FIG12 , the decoding sub-model is a trained sub-model, and the apparatus further includes a second training module 1106 for:

[0251] Obtaining first antibody sequence information, masking amino acid information at some positions in the first antibody sequence information, and obtaining second antibody sequence information;

[0252] The decoding sub-model is used to predict the amino acid information at the masked position in the second antibody sequence information to obtain a first prediction probability, where the first prediction probability represents the predicted probability that the amino acid at the masked position in the first antibody sequence information belongs to each amino acid;

[0253] The decoding sub-model is trained based on the first predicted probability and the first true probability, where the first true probability represents the true probability that the amino acid at the masked position in the first antibody sequence information belongs to each amino acid.

[0254] Optionally, referring to FIG12 , the encoding sub-model is a trained sub-model, and the apparatus further includes a third training module 1107 for:

[0255] Obtaining first antigen sequence information, masking amino acid information at some positions in the first antigen sequence information, and obtaining second antigen sequence information;

[0256] The amino acid information at the masked position in the second antigen sequence information is predicted using the encoding sub-model to obtain a second prediction probability, where the second prediction probability represents the predicted probability that the amino acid at the masked position in the first antigen sequence information belongs to each amino acid;

[0257] The encoding sub-model is trained based on the second predicted probability and the second true probability, where the second true probability represents the true probability that the amino acid at the masked position in the first antigen sequence information belongs to each amino acid.

[0258] It should be noted that the antibody sequence generation apparatus provided in the above embodiment is merely exemplified by the division of the aforementioned functional modules. In actual applications, the aforementioned functions can be distributed among different functional modules as needed, i.e., the internal structure of a computer device can be divided into different functional modules to perform all or part of the functions described above. Furthermore, the antibody sequence generation apparatus provided in the above embodiment and the antibody sequence generation method embodiment are based on the same concept. The specific implementation process is detailed in the method embodiment and will not be further described here.

[0259] An embodiment of the present application further provides a computer device, comprising a processor and a memory, wherein the memory stores at least one computer program, which is loaded and executed by the processor to implement the operations performed in the antibody sequence generation method of the above embodiment.

[0260] Optionally, the computer device is provided as a terminal. FIG13 shows a schematic structural diagram of a terminal 1300 provided in an exemplary embodiment of the present application.

[0261] The terminal 1300 includes a processor 1301 and a memory 1302 .

[0262] Processor 1301 may include one or more processing cores, such as a quad-core processor or an octa-core processor. In some embodiments, processor 1301 may be integrated with a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content required to be displayed on the display screen. In some embodiments, processor 1301 may also include an AI (Artificial Intelligence) processor, which is used to handle computing operations related to machine learning.

[0263] The memory 1302 may include one or more computer-readable storage media, which may be non-transitory. The memory 1302 may also include high-speed random access memory, and 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 1302 is used to store at least one computer program, which is used to be executed by the processor 1301 to implement the antibody sequence generation method provided in the method embodiment of the present application.

[0264] In some embodiments, terminal 1300 may optionally include a peripheral device interface 1303 and at least one peripheral device. Processor 1301, memory 1302, and peripheral device interface 1303 may be connected via a bus or signal lines. Each peripheral device may be connected to peripheral device interface 1303 via a bus, signal lines, or circuit boards. Optionally, the peripheral device includes at least one of a radio frequency circuit 1304, a display screen 1305, a camera assembly 1306, an audio circuit 1307, and a power supply 1308.

[0265] The peripheral device interface 1303 may be used to connect at least one I / O (Input / Output)-related peripheral device to the processor 1301 and the memory 1302 .

[0266] The radio frequency circuit 1304 is used to receive and transmit RF (Radio Frequency) signals, also known as electromagnetic signals.

[0267] The display screen 1305 is used to display a UI (User Interface). The UI may include graphics, text, icons, videos, or any combination thereof.

[0268] The camera assembly 1306 is used to collect images or videos. Optionally, the camera assembly 1306 includes a front camera and a rear camera.

[0269] Audio circuitry 1307 may include a microphone and a speaker.

[0270] The power supply 1308 is used to supply power to various components in the terminal 1300 .

[0271] Those skilled in the art will understand that the structure shown in FIG13 does not constitute a limitation on the terminal 1300 , and the terminal 1300 may include more or fewer components than shown in the figure, or combine certain components, or adopt a different component arrangement.

[0272] Optionally, the computer device is provided as a server. Figure 14 is a structural diagram of a server provided in an embodiment of the present application. The server 1400 may have relatively large differences due to different configurations or performances, and may include one or more processors (Central Processing Units, CPU) 1401 and one or more memories 1402, wherein the memory 1402 stores at least one computer program, and the at least one computer program is loaded and executed by the processor 1401 to implement the methods provided in the above-mentioned various method embodiments. Of course, the server 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 may also include other components for implementing device functions, which will not be described in detail here.

[0273] Severe acute respiratory syndrome (SARS) is an acute respiratory infectious disease caused by the SARS coronavirus (SARS-CoV). SARS is a respiratory infectious disease primarily transmitted through close-range droplet transmission or contact with respiratory secretions from patients.

[0274] SARS-CoV-2 is a single-stranded RNA virus belonging to the genus Betacoronavirus. It shares 79.6% genome identity with SARS-CoV and 96.2% with the genome of the bat coronavirus RaTG13. Similar to other Betacoronaviruses, SARS-CoV-2 infection is mediated by the binding of the spike (S) glycoprotein to its receptor, human angiotensin-converting enzyme 2 (hACE2). The S protein is a trimeric fusion protein on the virion surface. Upon binding to host cells, it is cleaved by the cellular serine protease TMPRSS2 and the lysosomal protease cathepsin into the receptor-binding fragment S1 subunit and the fusion fragment S2 subunit. The S1 subunit interacts with hACE2 via its C-terminal receptor-binding domain (RBD), then switches its conformation from "sit-down" to "stand-up" to dissociate and expose the S2 subunit, which drives fusion between the virus and the cell membrane. Currently, among the natural antibody sequences targeting SARS-CoV-2, the antigenic affinity and antigen-neutralizing capacity of many novel coronavirus natural antibodies need to be improved, making them difficult to meet clinical needs.

[0275] To this end, the present application aims to solve at least one of the technical problems existing in the related art to a certain extent. To this end, the present invention provides an anti-novel coronavirus antibody having a high binding affinity to the S protein of the novel coronavirus.

[0276] This application details

[0277] Definitions and General Terms

[0278] In this document, the terms "include" or "comprising" are open expressions, that is, including the contents specified in this application, but not excluding other contents.

[0279] As used herein, the terms "optionally," "optional," or "optionally" generally mean that the subsequently described event or circumstance may but need not occur, and that the description includes instances where the event or circumstance occurs and instances where it does not.

[0280] As used herein, the term "fragment" refers to a target protein or polypeptide, as well as a target protein or polypeptide with N-terminal (N-terminus) or C-terminal (C-terminus) truncation, and / or internal deletion.

[0281] As used herein, the terms "identity," "homology," or "similarity" are used to describe an amino acid sequence or nucleic acid sequence relative to a reference sequence, and the percentage of identical amino acids or nucleotides between two amino acid sequences or nucleic acid sequences is determined by conventional methods, for example, see Ausubel et al., eds. (1995), Current Protocols in Molecular Biology, Chapter 19 (Greene Publishing and Wiley-Interscience, New York); and the ALIGN program (Dayhoff (1978), Atlas of Protein Sequence and Structure 5: Suppl. 3 (National Biomedical Research Foundation, Washington, DC). There are many algorithms for aligning sequences and determining sequence identity, including the homology alignment algorithm of Needleman et al. (1970) J. Mol. Biol. 48:443; the local homology algorithm of Smith et al. (1981) Adv. Appl. Math. 2:482; the similarity search method of Pearson et al. (1988) Proc. Natl. Acad. Sci. 85:2444; the Smith-Waterman algorithm (Meth. Mol. Biol. 10:106); the similarity search method of Pearson et al. (1988) Proc. Natl. Acad. Sci. 10:116; the similarity search method of Pearson et al. (1990) Proc. Natl. Acad. Sci. 10:117; the similarity search method of Pearson et al. (1990 ... .70:173-187 (1997); and BLASTP, BLASTN, and BLASTX algorithms (see Altschul et al. (1990) J. Mol. Biol. 215:403-410). Computer programs that utilize these algorithms are also available and include, but are not limited to, ALIGN or Megalign (DNASTAR) software, or WU-BLAST-2 (Altschul et al., Meth. Enzym., 266:460-480 (1996)); or GAP, BESTFIT, BLAST Altschul et al., supra, FASTA, and TFASTA, available in the Genetics Computing Group (GCG) package, Version 8, Madison, Wisconsin, USA; and CLUSTAL in the PC / Gene program provided by Intelligenetics, Mountain View, California.

[0282] As used herein, the term "at least 80% identity" refers to at least 80%, and may be 80%, 81%, 82%, 83%, 84%, 85%, 86%, 87%, 88%, 89%, 90%, 91%, 92%, 93%, 94%, 95%, 96%, 97%, 98%, 99%, 99.5%, or 99.9% identity to the respective reference sequence.

[0283] As used herein, the term "at least 90% identity" refers to at least 90%, and may be 90%, 91%, 92%, 93%, 94%, 95%, 96%, 97%, 98%, 99%, 99.5%, or 99.9% identity to the respective reference sequence.

[0284] In this article, the term "expression vector" generally refers to a nucleic acid molecule that can be inserted into a suitable host and replicates itself, and the nucleic acid molecule contains a nucleotide sequence that can express a target protein, and the nucleic acid molecule can be transferred into a host cell and / or between host cells. The expression vector may include a vector mainly used to insert DNA or RNA into a cell, a vector mainly used to replicate DNA or RNA, and a vector mainly used for expression of transcription and / or translation of DNA or RNA. The expression vector also includes vectors with multiple of the above functions. The expression vector can be a polynucleotide that can be transcribed and translated into a polypeptide when introduced into a suitable host cell. Generally, the expression vector can produce a desired expression product by culturing a suitable host cell containing the expression vector.

[0285] In this article, the term "recombinant cell" generally refers to the use of genetic engineering technology or cell fusion technology to modify or reorganize the genetic material of the host cell to obtain a cell with a unique characteristic of stable inheritance. Wherein, the term "host cell" refers to a prokaryotic cell or eukaryotic cell that can be imported into a recombinant expression vector. The terms "transformed" or "transfected" as used herein refer to the introduction of nucleic acid (such as an expression vector) into a cell by various techniques known in the art. Suitable host cells can be transformed or transfected with the DNA sequence of the application and can be used for the expression and / or secretion of the target protein. Examples of suitable host cells that can be used for the application include immortalized hybridoma cells, NS / 0 myeloma cells, 293 cells, Chinese hamster ovary (CHO) cells, HeLa cells, Cap cells (cells derived from human amniotic fluid) and CoS cells.

[0286] As used herein, the term "pharmaceutical composition" generally refers to a unit dosage form and can be prepared by any of the methods well known in the pharmaceutical art. All methods include the step of bringing the active ingredient into association with a carrier which constitutes one or more accessory ingredients. Generally, the composition is prepared by uniformly and thoroughly combining the active antibody or antigen-binding fragment with a liquid carrier, a finely divided solid carrier, or both.

[0287] As used herein, the term "pharmaceutically acceptable excipient" may include any solvent, solid excipient, diluent, or other liquid excipient, etc., suitable for the specific target dosage form. Except to the extent that any conventional excipient is incompatible with the antibody or antigen-binding fragment of the present application, such as by producing any adverse biological effect or interacting in a deleterious manner with any other component of the pharmaceutically acceptable composition, their use is also contemplated by the present application.

[0288] In this article, the term "administration" refers to introducing a predetermined amount of material into a patient by a certain suitable manner. The antibody or antigen-binding fragment, recombinant protein, multispecific antibody or pharmaceutical composition of the present application can be administered by any common approach, as long as it can reach the expected tissue. The various modes of administration are expected to include peritoneal, intravenous, intramuscular, subcutaneous injections, etc., but the application is not limited to these exemplified modes of administration. Preferably, the composition of the present application adopts intravenous or subcutaneous injection mode to be administered.

[0289] As used herein, the term "treatment" refers to any agent used to obtain a desired pharmacological and / or physiological effect. The effect may be preventive in terms of completely or partially preventing a disease or its symptoms, and / or therapeutic in terms of partially or completely curing a disease and / or the adverse effects caused by the disease. "Treatment" as used herein covers diseases in mammals, particularly humans, and includes: (a) preventing the occurrence of a disease or condition in an individual who is susceptible to the disease but has not yet been diagnosed with the disease; (b) inhibiting the disease, such as arresting the progression of the disease; or (c) alleviating the disease, such as alleviating the symptoms associated with the disease. "Treatment" as used herein covers any medication that administers a drug or antibody or antigen-binding fragment to an individual to treat, cure, alleviate, improve, reduce or inhibit the individual's disease, including but not limited to administering a drug containing the antibody or antigen-binding fragment described herein to an individual in need.

[0290] Detailed description of the anti-novel coronavirus antibodies and their applications

[0291] The present application proposes an antibody or antigen-binding fragment, a recombinant protein, a multispecific antibody, a nucleic acid molecule, an expression vector, a recombinant cell, a conjugate, a pharmaceutical composition, a kit, and their uses, and a method for detecting the novel coronavirus, which will be described in detail below.

[0292] Antibodies or antigen-binding fragments

[0293] In one aspect of the present application, the present application provides an antibody or antigen-binding fragment. According to an embodiment of the present application, the antibody or antigen-binding fragment comprises an HCDR3 having an amino acid sequence as shown in SEQ ID NO: 3 or a conservatively modified form thereof. Compared with natural antibodies, the antibody containing the HCDR3 of the present application has better affinity and neutralization ability for the new coronavirus, and can effectively detect the new coronavirus, diagnose related diseases caused by the new coronavirus, or prevent and / or treat related diseases caused by the new coronavirus.

[0294] In this article, the term "antibody" is used in the broadest sense, which can include full-length monoclonal antibodies, multispecific antibodies, and chimeric antibodies. The specific structure is not limited as long as they exhibit the desired biological activity. It usually includes a light chain with a lighter molecular weight and a heavy chain with a heavier molecular weight. The heavy chain (H chain) and the light chain (L chain) are connected by a disulfide bond to form an antibody molecule. Among them, the amino acid sequence of the amino terminal (N-terminal) of the peptide chain varies greatly and is called the variable region (V region); the carboxyl terminal (C-terminal) is relatively stable and varies very little, and is called the constant region (C region). The V regions of the L chain and H chain are called VL and VH, respectively.

[0295] Herein, the heavy chain complementary determining regions (heavy chain variable region CDRs) are referred to as "HCDRs" or "HCDRs," which include HCDR1 (also known as CDR-H1), HCDR2 (also known as CDR-H2), and HCDR3 (also known as CDR-H3); the light chain complementary determining regions (light chain variable region CDRs) are referred to as "LCDRs" or "LCDRs," which include LCDR1 (also known as CDR-L1), LCDR2 (also known as CDR-L2), and LCDR3 (also known as CDR-L3). Commonly used CDR definition schemes in the art include: Kabat definition, Chothia definition, IMGT definition, Contact definition, and AbM definition.

[0296] As used herein, the "Kabat definition" refers to the definition system described in Kabat et al., U. S. Patent No. of Health and Human Services, "Sequence of Proteins of Immunological Interest" (1983). For the "Chothia definition", see Chothia et al., J Mol Biol 196:901-917 (1987). Exemplary defined CDRs are listed in Table A below. The definitions in different documents vary slightly. Given the variable region amino acid sequence of an antibody, those skilled in the art can routinely determine which residues comprise a specific CDR. It should be noted that the CDRs in this application include but are not limited to CDRs defined by other methods in Table A. CDRs determined based on the heavy chain variable region and light chain variable region disclosed in this application using other rules disclosed in the art also fall within the scope of protection of this disclosure.

[0297] Table A: CDR Definition 1

[0298] 1 The numbering of all CDR definitions in Table A is according to the Kabat numbering system (see below), with amino acid numbers on the heavy chain indicated by "H+numbers" and amino acid numbers on the light chain indicated by "L+numbers".

[0299] 2 "AbM" as used in Table A with a lowercase "b" refers to CDRs defined by Oxford Molecular's "AbM" antibody modeling software.

[0300] 3 If both H35A and H35B are absent, CDR-H1 ends at position 35; if only H35A is present, CDR-H1 ends at position 35A; if both H35A and H35B are present, CDR-H1 ends at position 35B.

[0301] 4 If both H35A and H35B are absent, CDR-H1 ends at position 32; if only H35A is present, CDR-H1 ends at position 33; if both H35A and H35B are present, CDR-H1 ends at position 34.

[0302] 5 If both H35A and H35B are absent, CDR-H1 ends at position 33; if only H35A is present, CDR-H1 ends at position 34; if both H35A and H35B are present, CDR-H1 ends at position 35.

[0303] Kabat et al. also defined a numbering system applicable to the variable region sequence of any antibody. Those skilled in the art can clearly map the Kabat numbering system to any variable region sequence without relying on any experimental data other than the sequence itself. As described herein, "Kabat numbering" refers to numbering using the numbering system described in Kabat et al., U.S. Patent No. 2002, "Sequence of Proteins of Immunological Interest" (1983). The HCDRs and LCDRs of the antibodies or antigen-binding fragments of the present application are numbered using the above numbering system. The specific numbering results are shown in Table A. It should be noted that the polypeptide sequences in the sequence listing and Table B of the present application are numbered according to the Kabat numbering system. However, those skilled in the art are fully capable of converting the sequence Kabat numbering in the sequence listing to "HCDRs" and / or "LCDRs" under other numbering systems, all of which are within the scope of protection of the present application.

[0304] As used herein, the term "antigen-binding fragment" refers to a fragment comprising a portion or all of an antibody that lacks at least some of the amino acids present in the full-length chain but is still capable of specifically binding to an antigen. For example, the fragment may comprise a portion or all of an antibody CDR. Such fragments are biologically active because they bind to an antigen and can compete with other antigen-binding molecules (including intact antibodies) for binding to a given epitope. Such fragments are selected from Fab, Fv, scFv, or single-domain antibodies. Such fragments can be produced by recombinant nucleic acid technology or can be produced by enzymatic or chemical cleavage of antigen-binding molecules (including intact antibodies).

[0305] As used herein, "conservatively modified amino acid sequences" refer to amino acid modifications that do not significantly affect or alter the binding properties of an antibody comprising the amino acid sequence, including amino acid substitutions, additions, and deletions. Modifications can be introduced into the antibodies of the present application by standard techniques such as site-directed mutagenesis and PCR-mediated mutagenesis. Conservative amino acid substitutions are substitutions in which an amino acid residue is replaced with an amino acid residue having a similar side chain. Families of amino acid residues with similar side chains have been identified in the art. These families include amino acids with basic side chains (such as lysine, arginine, histidine), amino acids with acidic side chains (such as aspartic acid, glutamic acid), amino acids with uncharged polar side chains (such as glycine, asparagine, glutamine, serine, threonine, tyrosine, cysteine, tryptophan), amino acids with non-polar side chains (such as alanine, valine, leucine, isoleucine, proline, phenylalanine, methionine), amino acids with β-branched side chains (such as threonine, valine, isoleucine), and amino acids with aromatic side chains (such as tyrosine, phenylalanine, tryptophan, histidine). Therefore, one or more amino acid residues in the CDR region of the antibody in the present application can be replaced by other amino acid residues from the same side chain family, and the retained function of the modified antibody can be tested using the functional assay method described herein. Preferably, the conservative modifications do not exceed 1 or 2 in number.

[0306] According to an embodiment of the present application, the antibody or antigen-binding fragment may further include at least one of the following technical features:

[0307] According to an embodiment of the present application, the antibody or antigen-binding fragment further includes a CDR selected from at least one of the following: heavy chain variable region CDR: amino acid sequence of SEQ ID NO: 1-2 or a conservative modification thereof; light chain variable region CDR: amino acid sequence of SEQ ID NO: 4-6 or a conservative modification thereof.

[0308] According to an embodiment of the present application, the antibody or antigen-binding fragment comprises: HCDR1, HCDR2, HCDR3 as shown in the amino acid sequences of SEQ ID NOs: 1, 2 and 3, respectively; and LCDR1, LCDR2, LCDR3 as shown in the amino acid sequences of SEQ ID NOs: 4, 5 and 6, respectively.

[0309] According to an embodiment of the present application, the antibody or antigen-binding fragment includes a heavy chain framework region.

[0310] According to an embodiment of the present application, at least a portion of the heavy chain framework region is derived from at least one of a mouse antibody, a primate antibody, a bovine antibody, a horse antibody, a cow antibody, a porcine antibody, a sheep antibody, a goat antibody, a dog antibody, a cat antibody, a rabbit antibody, a camel antibody, a donkey antibody, a deer antibody, a mink antibody, a chicken antibody, a duck antibody, a goose antibody, a turkey antibody, a fighting cock antibody, or a mutant thereof.

[0311] According to an embodiment of the present application, at least a portion of the heavy chain framework region is derived from at least one of a murine antibody and a human antibody.

[0312] According to an embodiment of the present application, the antibody or antigen-binding fragment includes a light chain framework region.

[0313] According to an embodiment of the present application, at least a portion of the light chain framework region is derived from at least one of a mouse antibody, a primate antibody, a bovine antibody, a horse antibody, a cow antibody, a porcine antibody, a sheep antibody, a goat antibody, a dog antibody, a cat antibody, a rabbit antibody, a camel antibody, a donkey antibody, a deer antibody, a mink antibody, a chicken antibody, a duck antibody, a goose antibody, a turkey antibody, a fighting cock antibody, or a mutant thereof.

[0314] According to an embodiment of the present application, at least a portion of the light chain framework region is derived from at least one of a murine antibody and a human antibody.

[0315] According to an embodiment of the present application, the antibody or antigen-binding fragment comprises: a heavy chain variable region having an amino acid sequence as shown in SEQ ID NO: 7 or an amino acid sequence having at least 90% homology thereto; and a light chain variable region having an amino acid sequence as shown in SEQ ID NO: 8 or an amino acid sequence having at least 90% homology thereto.

[0316] According to an embodiment of the present application, the antibody or antigen-binding fragment further includes a heavy chain constant region.

[0317] According to an embodiment of the present application, the antibody or antigen-binding fragment further includes a light chain constant region.

[0318] According to an embodiment of the present application, at least a portion of at least one of the heavy chain constant region and the light chain constant region is derived from at least one of a mouse antibody, a primate antibody, a bovine antibody, a horse antibody, a cow antibody, a porcine antibody, a sheep antibody, a goat antibody, a dog antibody, a cat antibody, a rabbit antibody, a camel antibody, a donkey antibody, a deer antibody, a mink antibody, a chicken antibody, a duck antibody, a goose antibody, a turkey antibody, a fighting cock antibody, or a mutant thereof.

[0319] According to an embodiment of the present application, the heavy chain constant region includes a heavy chain constant region selected from IgG1, IgG2, IgG3, IgG4, IgA, IgM, IgE or IgD.

[0320] According to an embodiment of the present application, the light chain constant region comprises a light chain constant region selected from a κ-type or a λ-type.

[0321] According to an embodiment of the present application, the light chain constant region and the heavy chain constant region are both derived from a murine antibody or a mutant thereof, or a human antibody or a mutant thereof.

[0322] According to an embodiment of the present application, the N-terminus of the heavy chain constant region is connected to the C-terminus of the heavy chain variable region.

[0323] According to an embodiment of the present application, the N-terminus of the light chain constant region is connected to the C-terminus of the light chain variable region.

[0324] According to an embodiment of the present application, the heavy chain constant region includes the heavy chain constant region shown in SEQ ID NO: 9 or an amino acid sequence having at least 80% identity thereto.

[0325] According to an embodiment of the present application, the light chain constant region includes the light chain constant region shown in SEQ ID NO: 10 or an amino acid sequence having at least 80% identity thereto.

[0326] According to an embodiment of the present application, the antibody includes at least one of a polyclonal antibody, a full-length monoclonal antibody, a Fab antibody, a Fab' antibody, a F(ab')2 antibody, a Fv antibody, a single-chain antibody, a single-domain antibody and a minimum recognition unit; or the antigen-binding fragment includes at least one of a F(ab')2 fragment, a Fab' fragment, a Fab fragment, a F(ab)2 fragment, a Fv fragment, a scFv fragment, a scFv-Fc fusion protein, a scFv-Fv fusion protein and a minimum recognition unit.

[0327] As used herein, the terms "full-length antibody," "full-length monoclonal antibody," or "full-length monoclonal antibody" are composed of at least two identical light chains and at least two identical heavy chains connected by interchain disulfide bonds, such as immunoglobulin G (IgG), immunoglobulin A (IgA), immunoglobulin M (IgM), immunoglobulin D (IgD), or immunoglobulin E (IgE).

[0328] In this article, the terms "polyantibody" and "multispecific antibody" are synonymous, both referring to antibodies that can recognize multiple antigenic epitopes, such as antibodies that can recognize two antigenic epitopes (bispecific antibodies, referred to as bispecific antibodies), antibodies that recognize three antigenic epitopes, or antibodies that recognize four antigenic epitopes. They are understood in a broad sense, and the specific structure is not limited, as long as they can recognize multiple antigenic epitopes. In this application, at least one of the multiple antigenic epitopes is derived from a new coronavirus.

[0329] Herein, the terms "single-domain antibody", "nanoantibody" and "VHH antibody" are used interchangeably, which were originally described as antigen-binding immunoglobulin (variable) domains of "heavy chain antibodies" (i.e., "antibodies lacking light chains") (Hamers-Casterman C, Atarhouch T, Muyldermans S, Robinson G, Hamers C, Songa EB, Bendahman N, Hamers R.: "Naturally occurring antibodies devoid of light chains"; Nature 363, 446-448 (1993)), comprising a heavy chain variable region (VH) and conventional CH2 and CH3 regions, which specifically bind to antigen proteins (e.g., S protein of the new coronavirus) through the heavy chain variable region.

[0330] As used herein, the term "Fab antibody" or "Fab fragment" generally refers to an antibody or fragment containing only the Fab molecule, which is composed of the VH and CH1 of the heavy chain and a complete light chain, with the light chain and heavy chain connected by a disulfide bond.

[0331] As used herein, the term "F(ab')2 antibody" or "F(ab')2 fragment" has two antigen-binding F(ab') portions linked together by a disulfide bond.

[0332] As used herein, the term "Fv antibody" or "Fv fragment" generally refers to an antibody or fragment consisting solely of a light chain variable region (VL) and a heavy chain variable region (VH) linked by non-covalent bonds. It is the smallest functional fragment of an antibody that retains a complete antigen-binding site.

[0333] As used herein, the terms "single-chain antibody" and "scFv fragment" refer to antibodies or fragments formed by connecting the heavy chain variable region and the light chain variable region of an antibody via a short peptide.

[0334] In this article, the terms "minimum recognition unit" and "MRU" both refer to antibodies or fragments consisting of only one CDR, which has a very small molecular weight of only about 1% of the complete antibody.

[0335] According to an embodiment of the present application, the antibody or antigen-binding fragment comprises: a heavy chain having an amino acid sequence as shown in SEQ ID NO: 11 or an amino acid sequence having at least 80% homology thereto; and a light chain having an amino acid sequence as shown in SEQ ID NO: 12 or an amino acid sequence having at least 80% homology thereto.

[0336] Recombinant proteins and multispecific antibodies

[0337] In the second aspect of this application, a recombinant protein is proposed. According to an embodiment of this application, the recombinant protein comprises: the antibody or antigen-binding fragment described in the first aspect. The recombinant protein of this application has improved affinity for and neutralization ability against the novel coronavirus, and can effectively detect the novel coronavirus, diagnose diseases related to the novel coronavirus, or prevent and / or treat diseases related to the novel coronavirus.

[0338] According to an embodiment of the present application, the recombinant protein further includes at least one of a biologically active protein or a fragment thereof, or a biologically active polypeptide or a fragment thereof.

[0339] According to an embodiment of the present application, the biologically active protein or fragment thereof includes at least one selected from a protein tag, a protein toxin or a fragment thereof, an interferon or a fragment thereof, a biological response modifier or a fragment thereof, and an Fc fragment.

[0340] In this article, "protein tag" generally refers to a polypeptide or protein that is fused with a target protein (antibody or antigen-binding fragment) and can be used for the expression, detection, disappearance or purification of the target protein. It includes but is not limited to the His tag (also known as His-Tag, sequence HHHHHH), Flag tag (also known as Flag-Tag, sequence DYKDDDDK), GST tag (also known as GST-Tag, glutathione S-transferase tag), MBP tag (also known as MBP-Tag, maltose binding protein tag), SUMO tag and C-Myc tag.

[0341] As used herein, "toxin" generally refers to a substance toxic to the host, including protein toxins and non-protein toxins. Protein toxins include, but are not limited to, abrin, ricin A, Pseudomonas exotoxin, and diphtheria toxin. In this application, protein toxins are preferably enzymatically active.

[0342] In this article, "interferon" generally refers to a glycoprotein that has the ability to directly kill or inhibit viruses, including but not limited to IFN-α, INF-β and IFN-γ.

[0343] In this article, "biological response modifiers" generally refer to a class of protein substances that directly or indirectly enhance the body's anti-tumor response through the immune system, including but not limited to lymphokines, IL-2, IL-6, IL-10, and GM-CSF.

[0344] As used herein, "Fc fragment" generally refers to an Fc region from IgG (e.g., IgG1, IgG2, IgG3, or IgG4 subtype), IgA1, IgA2, IgD, IgE, or IgM, including CH2, CH3, and optionally, a hinge region. Preferably, the IgG, IgA1, IgA2, IgD, IgE, or IgM is derived from mouse, human, primate, or alpaca.

[0345] In the third aspect of this application, a multispecific antibody is proposed. According to an embodiment of this application, the multispecific antibody comprises: a first binding region comprising the antibody or antigen-binding fragment described in the first aspect; and a second binding region having a first molecule binding activity. The multispecific antibody of this application has better affinity for and neutralization ability against the novel coronavirus, and can effectively detect the novel coronavirus, diagnose related diseases caused by the novel coronavirus, or effectively prevent and / or treat related diseases caused by the novel coronavirus.

[0346] According to an embodiment of the present application, the first molecule includes at least one of an antigen, a virus, a bacterium, an endotoxin, a cytokine, and a cytokine receptor.

[0347] Nucleic acid molecules, expression vectors and recombinant cells

[0348] In a fourth aspect of the present application, a nucleic acid molecule is provided. According to embodiments of the present application, the nucleic acid molecule encodes the antibody or antigen-binding fragment described in the first aspect, the recombinant protein described in the second aspect, or the multispecific antibody described in the third aspect. The nucleic acid molecule of the present application may encode the antibody or antigen-binding fragment described in the first aspect, the recombinant protein described in the second aspect, or the multispecific antibody described in the third aspect.

[0349] According to an embodiment of the present application, the nucleic acid molecule is DNA.

[0350] It should be noted that, for nucleic acid molecules mentioned herein, those skilled in the art will understand that they actually include any one or both of the complementary double strands. For convenience, in this specification and claims, although only one strand is provided in most cases, the other complementary strand is also disclosed. In addition, the nucleic acid sequences in this application include DNA or RNA forms, and disclosure of one of them implies disclosure of the other.

[0351] In the fifth aspect of the present application, an expression vector is provided. According to embodiments of the present application, the expression vector carries the nucleic acid molecule described in the fourth aspect. When the nucleic acid molecule is connected to the expression vector, the nucleic acid molecule can be directly or indirectly connected to the control elements on the expression vector, as long as these control elements are capable of controlling the translation and expression of the nucleic acid molecule. Of course, these control elements can be directly derived from the expression vector itself, or they can be exogenous, that is, not derived from the expression vector itself. Of course, the nucleic acid molecule and the control elements can be operably connected.

[0352] As used herein, "operably linked" means that a foreign gene is linked to an expression vector so that the control elements within the expression vector, such as transcriptional control sequences and translational control sequences, can function as intended to regulate the transcription and translation of the foreign gene. Commonly used expression vectors include plasmids, bacteriophages, and the like.

[0353] According to some specific embodiments of the present application, after the expression vector is introduced into suitable recipient cells, it can effectively achieve the expression of the aforementioned antibody or antigen-binding fragment, the aforementioned recombinant protein or the aforementioned multispecific antibody under the mediation of the regulatory system, thereby achieving large-scale in vitro production of the antibody or antigen-binding fragment, recombinant protein or multispecific antibody.

[0354] According to an embodiment of the present application, the expression vector is selected from a eukaryotic expression vector or a prokaryotic expression vector.

[0355] In an optional embodiment of the present application, the expression vector is a plasmid expression vector, a viral expression vector, such as a lentiviral expression vector.

[0356] In the sixth aspect of this application, a recombinant cell is provided. According to embodiments of this application, the recombinant cell comprises a nucleic acid molecule as described in the fourth aspect or an expression vector as described in the fifth aspect; or expresses the antibody or antigen-binding fragment as described in the first aspect, the recombinant protein as described in the second aspect, or the multispecific antibody as described in the third aspect. Under suitable conditions, the recombinant cell can effectively express the aforementioned antibody or antigen-binding fragment, recombinant protein, or multispecific antibody within the cell.

[0357] It should be noted that "suitable conditions" refer to conditions suitable for the expression of the antibodies or antigen-binding fragments, recombinant proteins or multispecific antibodies described herein. It will be readily understood by those skilled in the art that conditions suitable for the expression of the antibodies or antigen-binding fragments, recombinant proteins or multispecific antibodies include, but are not limited to, suitable transformation or transfection methods, suitable transformation or transfection conditions, healthy host cell status, suitable host cell density, suitable cell culture environment, and suitable cell culture time. "Suitable conditions" are not particularly limited, and those skilled in the art can optimize the most suitable conditions for the expression of the above-mentioned antibodies or antigen-binding fragments, recombinant proteins or multispecific antibodies based on the specific environment of the laboratory.

[0358] According to an embodiment of the present application, the recombinant cell is obtained by introducing the expression vector described in the fifth aspect into a host cell.

[0359] According to an embodiment of the present application, the recombinant cell is a eukaryotic cell, preferably a mammalian cell.

[0360] Conjugates, pharmaceutical compositions and kits

[0361] In the seventh aspect of the present application, the present application proposes a conjugate. According to the embodiments of the present application, the conjugate comprises the antibody or antigen-binding fragment described in the first aspect, the recombinant protein described in the second aspect, or the multispecific antibody described in the third aspect; and a coupling portion, which is connected to the antibody or antigen-binding fragment, recombinant protein or multispecific antibody. The conjugate of the present application can bind to the new coronavirus, has better affinity for the new coronavirus and neutralization ability of the new coronavirus, and can effectively detect the new coronavirus, diagnose related diseases caused by the new coronavirus, or prevent and / or treat related diseases caused by the new coronavirus.

[0362] According to an embodiment of the present application, the coupling portion includes at least one selected from a carrier, a drug, a toxin, a cytokine, a protein tag and a modifier.

[0363] Herein, the carrier can be a substance that can be suspended or dispersed in a liquid phase (e.g., solid phase carriers such as particles and magnetic beads), or a solid phase that can contain or carry a liquid phase (e.g., supports such as plates, membranes, test tubes, and containers such as well plates, microfluidics, glass capillaries, nanocolumns, and monolithic columns); it can also be a labeling carrier for labeling antibodies or antigen-binding fragments, recombinant proteins, or multispecific antibodies, such as an enzyme (e.g., peroxidase, alkaline phosphatase, luciferase, β-galactosidase), an affinity substance (e.g., one of streptavidin and biotin, one of mutually complementary sense and antisense nucleic acids), a fluorescent substance (e.g., fluorescein, fluorescein isothiocyanate, rhodamine, green fluorescent protein, red fluorescent protein), a luminescent substance (e.g., luciferin, aequorin, acridinium ester, tris(2,2'-bipyridyl)ruthenium, luminol), a radioactive isotope (e.g., 3 H. 14 C. 32 P. 35 S. 125 I) and gold colloid, etc.

[0364] According to an embodiment of the present application, the drug is a small molecule drug that can bind to an antibody or antigen-binding fragment, a recombinant protein or a multispecific antibody.

[0365] According to the embodiments of the present application, the protein tags include but are not limited to His tags, Flag tags, GST tags, MBP tags, SUMO tags and C-Myc tags.

[0366] According to the embodiments of the present application, the modifier should be understood in a broad sense and may refer to a substance used to modify a protein. For example, it may be polyethylene glycol or a derivative thereof.

[0367] It should be noted that methods known in the art can be used to bind the conjugated moiety to the antibody, antigen-binding fragment, recombinant protein, or multispecific antibody. Examples include physical adsorption, covalent binding, methods using affinity substances (e.g., biotin, streptavidin), and ion binding.

[0368] In the eighth aspect of the present application, the present application proposes a pharmaceutical composition. According to the embodiments of the present application, the pharmaceutical composition includes the antibody or antigen-binding fragment described in the first aspect, the recombinant protein described in the second aspect, the multispecific antibody described in the third aspect, the nucleic acid molecule described in the fourth aspect, the expression vector described in the fifth aspect, the recombinant cell described in the sixth aspect, or the conjugate described in the seventh aspect. The pharmaceutical composition of the present application can bind to the new coronavirus, has better affinity for the new coronavirus and neutralization ability of the new coronavirus, and can effectively prevent and / or treat related diseases caused by the new coronavirus.

[0369] According to an embodiment of the present application, the pharmaceutical composition further includes a pharmaceutically acceptable excipient.

[0370] The administration of the pharmaceutical composition of the present application can be carried out by any acceptable mode of administration. The pharmaceutical composition of the present application can be formulated into a preparation in solid, semisolid, liquid or gaseous form, such as an injection, a lyophilized powder, and the current methods for preparing these dosage forms are known or will be apparent to those skilled in the art. Typical routes of administering such pharmaceutical compositions include, but are not limited to, subcutaneous injection, intravenous, intramuscular, intradermal, intrasternal injection or infusion techniques. The pharmaceutical composition of the present application is formulated so as to allow the biologically active ingredients contained therein to be bioavailable after the composition is administered to a patient.

[0371] In the ninth aspect of the present application, the present application proposes a kit. According to the embodiments of the present application, the kit includes the antibody or antigen-binding fragment described in the first aspect, the recombinant protein described in the second aspect, the multispecific antibody described in the third aspect, the nucleic acid molecule described in the fourth aspect, the expression vector described in the fifth aspect, the recombinant cell described in the sixth aspect, or the conjugate described in the seventh aspect. The kit of the present application can bind to the new coronavirus and has a strong affinity for the new coronavirus, and can effectively detect the new coronavirus or diagnose related diseases caused by the new coronavirus.

[0372] use

[0373] In the tenth aspect of the present application, the present application proposes a use of the antibody or antigen-binding fragment described in the first aspect, the recombinant protein described in the second aspect, the multispecific antibody described in the third aspect, the nucleic acid molecule described in the fourth aspect, the expression vector described in the fifth aspect, the recombinant cell described in the sixth aspect, or the conjugate described in the seventh aspect in the preparation of a drug, wherein the drug is used to prevent and / or treat related diseases caused by the new coronavirus.

[0374] In the eleventh aspect of the present application, the present application proposes a use of the antibody or antigen-binding fragment described in the first aspect, the recombinant protein described in the second aspect, the multispecific antibody described in the third aspect, the nucleic acid molecule described in the fourth aspect, the expression vector described in the fifth aspect, the recombinant cell described in the sixth aspect, or the conjugate described in the seventh aspect in the preparation of a kit, wherein the kit is used to detect the new coronavirus, diagnose related diseases caused by the new coronavirus, or evaluate the prognosis of related diseases caused by the new coronavirus.

[0375] method

[0376] In the twelfth aspect of this application, this application proposes a method for detecting the novel coronavirus, which can be implemented by a computer device. According to an embodiment of the present application, the method comprises: contacting the sample to be tested with the antibody or antigen-binding fragment described in the first aspect, the recombinant protein described in the second aspect, the multispecific antibody described in the third aspect, or the conjugate described in the seventh aspect to form an immune complex. The method of the present application has the advantages of detecting the novel coronavirus, especially for in vitro detection, and has high detection accuracy.

[0377] According to an embodiment of the present application, whether the sample to be tested contains the content of the new coronavirus is determined based on the signal of the immune complex.

[0378] According to an embodiment of the present application, the immune complex further includes a second antibody, which binds to the antibody or antigen-binding fragment.

[0379] According to an embodiment of the present application, the immune complex further includes a second antibody, which binds to the new coronavirus.

[0380] In the thirteenth aspect of this application, the present application proposes a method for preventing and / or treating diseases related to the novel coronavirus. According to an embodiment of the present application, the method comprises administering to a subject a pharmaceutically acceptable amount of the antibody or antigen-binding fragment of the first aspect, the recombinant protein of the second aspect, the multispecific antibody of the third aspect, or the conjugate of the seventh aspect. The method of the present application can effectively treat or prevent the novel coronavirus.

[0381] The effective amount of the antibody or antigen-binding fragment, recombinant protein, multispecific antibody, pharmaceutical composition or conjugate described herein may vary depending on the mode of administration and the severity of the disease to be treated. The selection of the preferred effective amount can be determined by a person of ordinary skill in the art based on various factors (e.g., through clinical trials). The factors include, but are not limited to: pharmacokinetic parameters of the active ingredient, such as bioavailability, metabolism, half-life, etc.; the severity of the disease to be treated, the patient's weight, the patient's immune status, the route of administration, etc. For example, depending on the urgency of the treatment condition, several divided doses may be administered daily, or the dose may be reduced proportionally.

[0382] The antibody or antigen-binding fragment, recombinant protein, multispecific antibody, pharmaceutical composition or conjugate of the present application can be incorporated into a drug suitable for parenteral administration (e.g., intravenous, subcutaneous, intraperitoneal, intramuscular). These drugs can be prepared into various forms. For example, liquid, semisolid and solid dosage forms, etc., include but are not limited to liquid solutions (e.g., injection solutions and infusion solutions) or lyophilized powders. Typical drugs are in the form of injection solutions or infusion solutions. The aforementioned antibody or antigen-binding fragment, recombinant protein, multispecific antibody, pharmaceutical composition or conjugate can be administered by intravenous infusion or injection or intramuscular or subcutaneous injection.

[0383] As used herein, the term "subject" refers to a vertebrate, preferably a mammal, most preferably a human. Mammals include, but are not limited to, rodents, apes, humans, livestock, athletic animals, and pets. Tissues, cells, and progeny of biological entities obtained in vivo or cultured in vitro are also included.

[0384] According to an embodiment of the present application, the administration route of the method is subcutaneous injection or intravenous injection.

[0385] In the fourteenth aspect of this application, this application proposes a method for diagnosing diseases related to the novel coronavirus. According to embodiments of this application, the method comprises: using the antibody or antigen-binding fragment described in the first aspect, the recombinant protein described in the second aspect, the multispecific antibody described in the third aspect, or the conjugate described in the seventh aspect to test a sample to be tested, and based on the test results, determining the level of the novel coronavirus in the sample to be tested. The method of this application can effectively diagnose diseases related to the novel coronavirus.

[0386] According to an embodiment of the present application, the sample to be tested is blood.

[0387] In the fifteenth aspect of this application, this application proposes a method for assessing the prognosis of a disease related to the novel coronavirus. According to an embodiment of this application, the method comprises: using the antibody or antigen-binding fragment described in the first aspect, the recombinant protein described in the second aspect, the multispecific antibody described in the third aspect, or the conjugate described in the seventh aspect to test a sample to be tested, and based on the test results, determining the content of the novel coronavirus in the sample to be tested. The method of this application can effectively assess the prognosis of a disease related to the novel coronavirus.

[0388] According to an embodiment of the present application, the sample to be tested is derived from a patient suffering from a disease related to the novel coronavirus before or after treatment.

[0389] According to an embodiment of the present application, the prognosis of a disease caused by the novel coronavirus is determined based on the content of the novel coronavirus in a test sample of a patient suffering from a disease caused by the novel coronavirus before or after treatment.

[0390] According to an embodiment of the present application, a decrease in the content of the novel coronavirus in a test sample of a patient suffering from a disease related to the novel coronavirus after treatment is an indication of a good prognosis for the patient.

[0391] According to an embodiment of the present application, the sample to be tested is blood.

[0392] The amino acid sequences involved in this article are shown in Table B:

[0393] Table B: Amino acid sequences

[0394] Below in conjunction with embodiment, the scheme of the application will be explained. Those skilled in the art will appreciate that the following examples are merely for illustration of the application and should not be considered as limiting the scope of the application. Where specific techniques or conditions are not indicated in the examples, they are carried out according to the techniques or conditions described in the literature in this area or according to the product specifications. Reagents used or instruments not indicated by the manufacturer are conventional products that can be obtained commercially.

[0395] The present application will be described in detail below through examples. In the examples or test examples, if the experimental methods do not specify specific conditions, they are all carried out under conventional conditions.

[0396] Practice of the present disclosure will employ, unless otherwise indicated, conventional techniques of cell biology, molecular biology (including recombinant techniques), microbiology, biochemistry, and immunology, which are within the capabilities of a skilled artisan.

[0397] In the examples of the present application, the nucleotide sequence used to prepare the expression vector can be obtained according to its amino acid sequence using conventional methods or conventional software (such as the online program Vectorbuilder (website: https: / / www.vectorbuilder.cn / tool / codon-optimization.html), GeneOptimizer online program, etc.).

[0398] Example 1: Artificial Intelligence-Based Generation of Novel Coronavirus Antibodies

[0399] The specific screening method of this embodiment is described in patent application publication number CN116959576A and application number 202310566319.X, wherein the main steps are as follows:

[0400] The Roformer (Su, Jianlin, Yu Lu, Shengfeng Pan, Ahmed Murtadha, Bo Wen, and Yunfeng Liu. 2021. “RoFormer: Enhanced Transformer with Rotary Position Embedding.” ArXiv[Cs.CL].arXiv.http: / / arxiv.org / abs / 2104.09864.) model was pre-trained on over 100 million antibody sequences from the Antibody Observation Space (OAS) dataset to learn the sequence features of the antibody light and heavy chains, and to establish the Antibody Roformer pre-training model. The specific pre-training process is shown in Figure 7, which includes the following steps:

[0401] 1) Randomly select some words or characters for masking: Randomly select some words or characters in the training data for masking. Specifically, each amino acid in the antibody sequence is used as a token, and 15% of the tokens are randomly selected for masking.

[0402] 2) Replace masked words or characters: The masked words or characters are replaced with a special marker [MASK]. For example, the sequence "MDVLY" will become "MD[MASK]L[MASK]" after the masked characters are segmented.

[0403] 3) Input into the model: The masked and replaced text sequence is input into the model for training. The model needs to combine the context of the surrounding text to predict the masked words or characters.

[0404] 4) Calculate the loss function: The model outputs the predicted word or character probability distribution, uses the cross-entropy loss function to calculate the difference between the model's predicted result and the actual result, and performs backpropagation to update the model parameters. The cross-entropy loss function formula is as follows:

[0405] The Roformer model is a variant of the BERT (Devlin, Jacob, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2018. “BERT: Pre-Training of Deep Bidirectional Transformers for Language Understanding.” ArXiv [Cs.CL]. arXiv. http: / / arxiv.org / abs / 1810.04805.) model that uses rotational position encoding. The BERT model architecture is mainly composed of a multi-layer Transformer (Vaswani, Ashish, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Kaiser, and Illia Polosukhin. 2017. “Attention Is All You Need.” Advances in Neural Information Processing Systems 30. https: / / proceedings.neurips.cc / paper / 7181-attention-is-all.) encoder, each Transformer block contains two parts: self-attention and feed-forward neural network.

[0406] The self-attention mechanism is used to learn the dependencies between positions in the input sequence. In BERT, each word is represented as a vector, and the self-attention mechanism can calculate the weight of each word with respect to other words, and then use these weights to calculate the weighted average vector representation, thereby achieving the effect of paying attention to each part of the sequence. By obtaining the Queue (Q), Key (K) and Value (V) matrices, and then calculating the attention weight matrix A:

[0407] Feedforward neural network layer is a basic neural network structure in deep learning. Input Z into a feedforward neural network layer as follows: FFN(A)=max(0,AW1+b1)W ％ +b ％ , (3)

[0408] Then, as shown in Figure 15, ESM2 (Lin, Zeming, Halil Akin, Roshan Rao, Brian Hie, Zhongkai Zhu, Wenting Lu, Nikita Smetanin, et al. 2022. “Evolutionary-Scale Prediction of Atomic Level Protein Structure with a Language Model.” BioRxiv. https: / / doi.org / 10.1101 / 2022.07.20.500902.) pre-trained on large-scale protein sequence data is used as an encoder to encode the input antigen epitope sequence, which is called the Antigen model; the pre-trained Heavy Roformer is used as a decoder to decode the encoding input by the encoder to generate an antibody sequence, ultimately realizing the generation process from antigen epitope sequence to antibody sequence. Specifically, the parameters of the previously pre-trained Antigen model and Antibody Roformer are directly used in the construction of the encoder and decoder. The encoder consists of multiple antigen layers, each with its own self-attention parameter weights derived from the ESM2 model. Similarly, the decoder's antigen layer's self-attention parameter weights are inherited from the antigen Roformer. The decoder's cross-attention layer is randomly initialized. The generative model is then trained using antigen-CDRH3 sequences from the Cov-AbDab and 14H datasets.

[0409] This application uses the 14H dataset (Engelhart, Emily, Ryan Emerson, Leslie Shing, Chelsea Lennartz, Daniel Guion, Mary Kelley, Charles Lin, Randolph Lopez, David Younger, and Matthew E. Walsh. 2022. “A Dataset Comprised of Binding Interactions for 104,972 Antibodies against a SARS-CoV-2 Peptide.” Scientific Data 9(1): 653.) The stable peptide segment of the HR2 region of the new coronavirus antigen S protein corresponding to the 14H dataset was used as the input of the model to generate the antibody CDRH3 sequence that can bind to this stable epitope.Using Alpha Fold2 (Jumper, John, Richard Evans, Alexander Pritzel, Tim Green, Michael Figurnov, Olaf Ronneberger, Kathryn Tunyasuvunakool, et al. 2021. "Highly Accurate Protein Structure Prediction with AlphaFold." Nature 596(7873):583-89.), Clus Pro (Kozakov, Dima, David R. Hall, Bing Xia, Kathryn A. Porter, Dzmitry Padhorny, Christine Yueh, Dmitri Beglov, and Sandor Vajda. 2017. "The ClusPro Web Server for Protein-Protein Docking." Nature Protocols 12(2):255-78.) and SnugDock (Sircar, Aroop, and Jeffrey J.Gray. 2010. "SnugDock: Paratope Structural Optimization during The complex structures of generated antibodies and natural antibodies with antigens were obtained using tools such as "Antibody-Antigen Docking Compensates for Errors in Antibody Homology Models." PLoS Computational Biology 6(1):e1000644.), and the results are shown in Figure 16. The AI-generated antibodies are able to bind to the CDRH3 region of the antibody and the stable region of the antigen. Antibody binding to the stable region represents a stronger applicability and can avoid the situation where the antibody fails due to new variants. This is of great significance for the design of antibodies for antigens with fast mutation rates such as the new coronavirus.

[0410] Example 2: Preparation of antibodies

[0411] 1. Using genetic engineering techniques, the nucleotide sequences encoding the heavy chain (amino acid sequence shown in SEQ ID NO: 11) and the light chain (amino acid sequence shown in SEQ ID NO: 12) were cloned into the pCMV3 expression vector, and the expression vectors were amplified to obtain the corresponding expression vectors. The resulting expression vectors were then transfected into HEK-293 cells using the chemical transfection reagent PEI, and cell lines that stably expressed the antibody sequences were obtained through pressure screening.

[0412] 2. Use protein purification technology to extract and purify the new coronavirus antibody sequence from the above cell lines.

[0413] Example 3: Detection of Antibody Binding Ability to SARS-CoV-2 S Protein and Neutralization Potency against SARS-CoV-2

[0414] In order to determine the binding ability of the artificial intelligence-generated antibody prepared in Example 2 to the novel coronavirus S protein (amino acid sequence as shown in SEQ ID NO: 15), in vitro experimental methods such as Western blot, surface plasmon resonance analysis, and pseudovirus neutralization assay in novel coronavirus antibody evaluation studies were used (for details, see He, Qingwen, Lili Wu, Zepeng Xu, Xiaoyun Wang, Yufeng Xie, Yan Chai, Anqi Zheng, et al. 2023. “An Updated Atlas of Antibody Evasion by SARS-CoV-2 Omicron Sub-Variants Including BQ.1.1and XBB.” Cell Reports. Medicine 4(4): 100991.; Qu, Panke, Julia N. Faraone, John P. Evans, Yi-Min Zheng, Claire Carlin, Mirela Anghelina, Patrick Stevens, et al. 2023. “Enhanced Evasion of Neutralizing Antibody Response by Omicron XBB.1.5, CH.1.1, and CA.3.1 Variants." Cell Reports 42(5):112443.) The artificial intelligence-generated antibody (i.e., artificial) and two natural antibodies (i.e., natural, all purchased from ACROBiosystems, wherein the amino acid sequence of the heavy chain of natural 1 is shown in SEQ ID NO:16, the amino acid sequence of the light chain is shown in SEQ ID NO:12, and the amino acid sequence of the heavy chain of natural 2 is shown in SEQ ID NO:17, and the amino acid sequence of the light chain is shown in SEQ ID NO:12) were evaluated. Among them, the protein blotting results of the natural antibodies (natural 1 and natural 2) and the artificial intelligence-generated antibody (artificial) binding to the new coronavirus S protein are shown in Figure 17, and the surface plasmon resonance analysis results of the natural antibodies (natural 1 and natural 2) and the artificial intelligence-generated antibody (artificial) binding to the new coronavirus S protein are shown in Table 1.The results of the protein blotting experiment showed that the AI-generated antibodies and two natural antibodies used as controls could bind to the 2019-nCoV S protein; compared with the two natural antibodies, the AI-generated antibodies exhibited high binding affinity to the 2019-nCoV S protein and superior neutralizing potency against the 2019-nCoV. The equilibrium dissociation constant (KD) of the AI-generated antibodies was 0.05 nm SARS-CoV-2, and the half-maximal inhibitory concentration (IC50) was 0.023 μg / ml.

[0415] .

[0416] Example 4: Application of artificial antibodies in the treatment, prevention and diagnosis of diseases related to the new coronavirus

[0417] The purified antibody sequences were used in experimental animal models for treatment and prevention studies of the novel coronavirus to observe their inhibitory effects on viral replication and infection processes, as well as their effects on normal tissues.

[0418] 1) Selection of experimental animal models: Rhesus monkeys (Cynomolgus macaque), mice, and rats were used as experimental animal models.

[0419] 2) Design of experimental and control groups: Four groups were designed for different species, namely: control group A: not infected with the new coronavirus and not given antibody treatment; control group B: infected with the new coronavirus and not given antibody treatment; experimental group C: infected with the new coronavirus and given the artificial antibody prepared in Example 2; experimental group D: not infected with the new coronavirus and given the artificial antibody prepared in Example 2.

[0420] The infection was carried out using the novel coronavirus S protein (amino acid sequence shown in SEQ ID NO: 15). Each group received the same dosage, and blood and tissue samples were collected from the animals at regular intervals (e.g., 24, 48, and 72 hours). The collected samples were subjected to real-time fluorescence quantitative PCR, viral titer determination, histological examination, ELISA, and inflammatory factor testing.

[0421] 3) Administration: a) Control groups A and B: no drug was given; b) Experimental group C: treated with purified antibody sequences within 24 hours after infection with the new coronavirus; c) Experimental group D: treated with purified antibody sequences and then infected with the new coronavirus 24 hours later.

[0422] 4) Detection indicators and methods: a) Viral load: Real-time fluorescence quantitative PCR (RT-qPCR) was used to detect the viral load in the respiratory secretions of each group of animals; b) Viral infection progression: The extent of viral infection was detected by viral titer determination; c) Tissue effects: The effects of antibodies on normal tissues were evaluated by histological examination; d) Immune response: The antibody levels in animals were detected by ELISA; e) Inflammatory response: The extent of inflammation was evaluated by detecting the levels of inflammatory factors.

[0423] The test results showed that by observing the differences in clinical symptoms, viral replication and infection level between the experimental group C and the control group B, it was found that artificial antibodies can effectively treat novel coronavirus infection; by observing the differences in clinical symptoms, viral replication and infection level between the experimental group D and the control group A, it was found that artificial antibodies have a preventive effect on novel coronavirus infection; through histological examination to evaluate the effects of antibodies on normal tissues, as well as changes in immune and inflammatory responses, it was found that the treatment with artificial antibodies is safe.

[0424] Based on the experimental data, the effectiveness and safety of the novel coronavirus antibody sequence of this application in the treatment and prevention of novel coronavirus will be evaluated, and a theoretical basis will be provided for further clinical trials.

[0425] An embodiment of the present application further provides a computer-readable storage medium, which stores at least one computer program, and the at least one computer program is loaded and executed by a processor to implement the operations performed by the antibody sequence generation method of the above embodiment.

[0426] An embodiment of the present application further provides a computer program product, comprising a computer program, which is loaded and executed by a processor to implement the operations performed by the antibody sequence generation method of the above embodiment.

[0427] Those 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 a program to instruct the relevant hardware, and the program may be stored in a computer-readable storage medium, which may be a read-only memory, a disk, or an optical disk, etc.

[0428] The above description is merely an optional embodiment of the embodiments of the present application and is not intended to limit the embodiments of the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the embodiments of the present application shall be included in the scope of protection of the present application.

Claims

1. A method for generating an antibody sequence, the method being executed by a computer device, the method comprising: Acquiring antigen sequence information, wherein the antigen sequence information represents amino acids in the antigen sequence; Extracting features from the antigen sequence information to obtain antigen sequence features of the antigen sequence information; Decoding the antigen sequence features to obtain a predicted probability for each position in the antibody sequence, wherein the predicted probability includes a predicted probability that the amino acid at the position belongs to each of the multiple amino acids; The sequence information composed of the amino acids with the highest predicted probability at each position is determined as the antibody sequence information, and the antibody sequence indicated by the antibody sequence information is capable of binding to the antigen sequence.

2. The method according to claim 1, wherein the predicted probability is obtained by an antibody generation model, wherein the antibody generation model comprises an encoding sub-model and a decoding sub-model; The step of extracting features from the antigen sequence information to obtain antigen sequence features of the antigen sequence information includes: By using the encoding sub-model, feature extraction is performed on the antigen sequence information to obtain antigen sequence features of the antigen sequence information; The feature decoding of the antigen sequence feature to obtain the predicted probability of each position in the antibody sequence includes: The antigen sequence features are decoded by the decoding sub-model to obtain the predicted probability of each position in the antibody sequence.

3. The method according to claim 2, wherein the antigen sequence features include target antigen key features and target antigen value features, and the encoding sub-model includes m encoding networks, where m is a positive integer; and extracting features from the antigen sequence information through the encoding sub-model to obtain the antigen sequence features of the antigen sequence information comprises: Through the first encoding network in the encoding sub-model, feature extraction is performed on the antigen sequence information to obtain a first candidate antigen key feature, a first candidate antigen value feature and a first candidate antigen query feature, and the first candidate antigen key feature, the first candidate antigen value feature and the first candidate antigen query feature are fused to obtain a first antigen fusion feature; By using the n+1th encoding network in the encoding sub-model, feature extraction is performed on the n+1th antigen fusion feature to obtain the n+1th candidate antigen key feature, the n+1th candidate antigen value feature and the n+1th candidate antigen query feature, where n is a positive integer less than m; The mth candidate antigen bond feature and the mth candidate antigen value feature are determined as the target antigen bond feature and the target antigen value feature.

4. The method according to claim 2, wherein the antigen sequence features include target antigen key features and target antigen value features, and wherein the antigen sequence features are feature decoded by the decoding submodel to obtain the predicted probability of each position in the antibody sequence, comprising: The target antigen bond feature, the target antigen value feature and the reference amino acid feature at each position are feature decoded by the decoding sub-model to obtain the predicted probability of each position in the antibody sequence, wherein the reference amino acid feature refers to the feature of the amino acid located before the position.

5. The method according to claim 4, for the first position in the antibody sequence, performing feature decoding on the target antigen bond feature, the target antigen value feature and the reference amino acid feature of each position through the decoding sub-model to obtain the predicted probability of each position in the antibody sequence, comprising: determining a preset amino acid feature as a reference amino acid feature at a first position; The target antigen bond feature, the target antigen value feature and the reference amino acid feature of the first position are feature decoded by the decoding sub-model to obtain the predicted probability at the first position.

6. The method according to claim 5, for the i+1th position in the antibody sequence, performing feature decoding on the target antigen bond feature, the target antigen value feature and the reference amino acid feature of each position through the decoding sub-model to obtain the predicted probability of each position in the antibody sequence, comprising: The amino acid feature of the amino acid with the highest predicted probability at the first i positions is determined as the reference amino acid feature at the i+1th position. The target antigen bond feature, the target antigen value feature and the reference amino acid feature at the i+1th position are feature decoded through the decoding sub-model to obtain the predicted probability at the i+1th position, where i is a positive integer.

7. The method according to claim 4, wherein the decoding sub-model comprises h decoding networks and an output network, where h is a positive integer, and wherein the target antigen key feature, the target antigen value feature and the reference amino acid feature at each position are feature decoded by the decoding sub-model to obtain a predicted probability for each position in the antibody sequence, comprising: For each position in the antibody sequence, the target antigen bond feature, the target antigen value feature and the reference amino acid feature of the position are feature decoded by the first decoding network in the decoding sub-model to obtain a first antibody sequence feature; The target antigen key feature, the target antigen value feature and the k-1th antibody sequence feature are feature decoded by the kth decoding network in the decoding sub-model to obtain the kth antibody sequence feature, where k is a positive integer greater than 1 and not greater than h; The h-th antibody sequence feature is normalized through the output network in the decoding sub-model to obtain the predicted probability of the position.

8. The method according to claim 7, wherein the decoding network comprises a self-attention layer and a cross-attention layer, and wherein the first decoding network in the decoding sub-model is used to perform feature decoding on the target antigen key feature, the target antigen value feature, and the reference amino acid feature of the position to obtain a first antibody sequence feature, comprising: By means of the self-attention layer, feature decoding is performed on the reference amino acid feature of the position to obtain an antibody query feature; Through the cross attention layer, the target antigen bond feature, the target antigen value feature and the antibody query feature are fused to obtain a first antibody sequence feature.

9. The method according to any one of claims 2 to 8, wherein the training process of the antibody production model comprises: Obtaining the true probability at each position in the sample antigen sequence information and the sample antibody sequence, wherein the sample antigen sequence information represents the amino acids in the sample antigen sequence, the true probability includes the true probability that the amino acid at the position belongs to each of the plurality of amino acids, and the sample antibody sequence can bind to the sample antigen sequence; Extracting features of the sample antigen sequence information through the encoding sub-model in the antibody generation model to obtain sample antigen sequence features; The sample antigen sequence features are decoded by a decoding sub-model in the antibody generation model to obtain a sample prediction probability at each position, wherein the sample prediction probability includes a prediction probability that the amino acid at the position belongs to each of the multiple amino acids; The antibody generation model is trained based on the sample prediction probability and the true probability at each position.

10. The method according to claim 9, wherein the decoding sub-model is a trained sub-model, and the training process of the decoding sub-model comprises: Acquire first antibody sequence information, mask amino acid information at some positions in the first antibody sequence information, and obtain second antibody sequence information; Using the decoding sub-model, predicting the amino acid information at the masked position in the second antibody sequence information to obtain a first prediction probability, wherein the first prediction probability indicates the predicted probability that the amino acid at the masked position in the first antibody sequence information belongs to each amino acid; The decoding sub-model is trained based on the first predicted probability and the first true probability, wherein the first true probability represents the true probability that the amino acid at the masked position in the first antibody sequence information belongs to each amino acid.

11. The method according to claim 9, wherein the encoding sub-model is a trained sub-model, and the training process of the encoding sub-model comprises: Acquire first antigen sequence information, mask amino acid information at some positions in the first antigen sequence information, and obtain second antigen sequence information; Using the coding sub-model, the amino acid information at the masked position in the second antigen sequence information is predicted to obtain a second prediction probability, wherein the second prediction probability indicates the predicted probability that the amino acid at the masked position in the first antigen sequence information belongs to each amino acid; The encoding sub-model is trained based on the second predicted probability and the second true probability, wherein the second true probability represents the true probability that the amino acid at the masked position in the first antigen sequence information belongs to each amino acid.

12. A device for generating an antibody sequence, the device comprising: An information acquisition module, used to acquire antigen sequence information, wherein the antigen sequence information represents amino acids in the antigen sequence; A feature extraction module, used to extract features from the antigen sequence information to obtain antigen sequence features of the antigen sequence information; A feature decoding module, used to perform feature decoding on the antigen sequence feature to obtain a predicted probability of each position in the antibody sequence, wherein the predicted probability includes a predicted probability that the amino acid at the position belongs to each of the multiple amino acids; The information determination module is used to determine the sequence information composed of the amino acids with the highest prediction probability at each position as antibody sequence information, and the antibody sequence indicated by the antibody sequence information can bind to the antigen sequence.

13. An antibody or antigen-binding fragment comprising a HCDR3 having an amino acid sequence as shown in SEQ ID NO: 3 or a conservatively modified form thereof.

14. The antibody or antigen-binding fragment according to claim 13, further comprising a CDR selected from at least one of the following: Heavy chain variable region CDR: amino acid sequence of SEQ ID NO: 1-2 or conservatively modified forms thereof; Light chain variable region CDR: amino acid sequence of SEQ ID NO: 4-6 or conservatively modified forms thereof; Optionally, the antibody or antigen-binding fragment comprises: HCDR1, HCDR2, HCDR3 as shown in the amino acid sequences of SEQ ID NOs: 1, 2 and 3, respectively; and LCDR1, LCDR2, LCDR3 are shown in the amino acid sequences of SEQ ID NO: 4, 5 and 6 respectively.

15. The antibody or antigen-binding fragment of claim 14, comprising a heavy chain framework region; Optionally, at least a portion of the heavy chain framework region is from at least one of a mouse antibody, a primate antibody, a bovine antibody, a horse antibody, a dairy cow antibody, a porcine antibody, a sheep antibody, a goat antibody, a dog antibody, a cat antibody, a rabbit antibody, a camel antibody, a donkey antibody, a deer antibody, a mink antibody, a chicken antibody, a duck antibody, a goose antibody, a turkey antibody, a fighting cock antibody, or a mutant thereof; Optionally, at least a portion of the heavy chain framework region is derived from at least one of a murine antibody and a human antibody; Optionally, the antibody or antigen-binding fragment comprises a light chain framework region; Optionally, at least a portion of the light chain framework region is derived from at least one of a mouse antibody, a primate antibody, a bovine antibody, a horse antibody, a dairy cow antibody, a porcine antibody, a sheep antibody, a goat antibody, a dog antibody, a cat antibody, a rabbit antibody, a camel antibody, a donkey antibody, a deer antibody, a mink antibody, a chicken antibody, a duck antibody, a goose antibody, a turkey antibody, a fighting cock antibody, or a mutant thereof; Optionally, at least a portion of the light chain framework region is from at least one of a murine antibody and a human antibody.

16. The antibody or antigen-binding fragment according to claim 13, comprising: A heavy chain variable region having an amino acid sequence as shown in SEQ ID NO: 7, or an amino acid sequence having at least 90% homology thereto; and The light chain variable region has an amino acid sequence as shown in SEQ ID NO: 8, or an amino acid sequence having at least 90% homology thereto.

17. The antibody or antigen-binding fragment according to any one of claims 13 to 16, further comprising a heavy chain constant region; Optionally, the antibody or antigen-binding fragment further comprises a light chain constant region; Optionally, at least a portion of at least one of the heavy chain constant region and the light chain constant region is from at least one of a mouse antibody, a primate antibody, a bovine antibody, a horse antibody, a dairy cow antibody, a porcine antibody, a sheep antibody, a goat antibody, a dog antibody, a cat antibody, a rabbit antibody, a camel antibody, a donkey antibody, a deer antibody, a mink antibody, a chicken antibody, a duck antibody, a goose antibody, a turkey antibody, a fighting cock antibody, or a mutant thereof; Optionally, the heavy chain constant region comprises a heavy chain constant region selected from IgG1, IgG2, IgG3, IgG4, IgA, IgM, IgE or IgD; or The light chain constant region comprises a light chain constant region selected from a κ type or a λ type; Optionally, the light chain constant region and the heavy chain constant region are both from a murine antibody or a mutant thereof, or a human antibody or a mutant thereof; Optionally, the N-terminus of the heavy chain constant region is linked to the C-terminus of the heavy chain variable region; Optionally, the N-terminus of the light chain constant region is linked to the C-terminus of the light chain variable region; Optionally, the heavy chain constant region comprises the heavy chain constant region set forth in SEQ ID NO:9, or an amino acid sequence having at least 80% identity thereto; Optionally, the light chain constant region comprises the light chain constant region set forth in SEQ ID NO: 10, or an amino acid sequence at least 80% identical thereto.

18. The antibody or antigen-binding fragment according to any one of claims 13 to 16, wherein the antibody comprises at least one of a polyclonal antibody, a full-length monoclonal antibody, a Fab antibody, a Fab' antibody, a F(ab')2 antibody, a Fv antibody, a single-chain antibody, a single-domain antibody, and a minimum recognition unit; or The antigen binding fragment comprises at least one of a F(ab')2 fragment, a Fab' fragment, a Fab fragment, a F(ab)2 fragment, a Fv fragment, a scFv fragment, a scFv-Fc fusion protein, a scFv-Fv fusion protein and a minimum recognition unit; Optionally, the antibody or antigen-binding fragment comprises: The heavy chain of an amino acid sequence as shown in SEQ ID NO: 11, or an amino acid sequence having at least 80% homology thereto; and The light chain of the amino acid sequence shown in SEQ ID NO: 12 or an amino acid sequence having at least 80% homology thereto.

19. A recombinant protein comprising: The antibody or antigen-binding fragment according to any one of claims 13 to 18; Optionally, further comprising at least one of a biologically active protein or a fragment thereof, a biologically active polypeptide or a fragment thereof; Optionally, the biologically active protein or fragment thereof comprises at least one selected from the group consisting of a protein tag, a protein toxin or a fragment thereof, an interferon or a fragment thereof, a biological response modifier or a fragment thereof, and an Fc fragment.

20. A multispecific antibody comprising: A first binding region, the first binding region comprising the antibody or antigen-binding fragment of any one of claims 13 to 18; a second binding region, the second binding region having a first molecule binding activity; Optionally, the first molecule comprises at least one of an antigen, a virus, a bacterium, an endotoxin, a cytokine, and a cytokine receptor.

21. A nucleic acid molecule encoding the antibody or antigen-binding fragment of any one of claims 13 to 18, the recombinant protein of claim 19, or the multispecific antibody of claim 20; Optionally, the nucleic acid molecule is DNA.

22. An expression vector carrying the nucleic acid molecule of claim 21; Optionally, the expression vector comprises a vector selected from a eukaryotic expression vector or a prokaryotic expression vector.

23. A recombinant cell, comprising: Carrying the nucleic acid molecule according to claim 21 or the expression vector according to claim 22; or, Expressing the antibody or antigen-binding fragment of any one of claims 13 to 18, the recombinant protein of claim 19, or the multispecific antibody of claim 20; Optionally, the recombinant cell is obtained by introducing the expression vector of claim 10 into a host cell.

24. A conjugate comprising: The antibody or antigen-binding fragment of any one of claims 13 to 18, the recombinant protein of claim 19, or the multispecific antibody of claim 20; and a coupling moiety, which is connected to the antibody or antigen-binding fragment, recombinant protein or multispecific antibody; Optionally, the coupling moiety comprises at least one selected from a carrier, a drug, a toxin, a cytokine, a protein tag and a modifier.

25. A pharmaceutical composition comprising: The antibody or antigen-binding fragment of any one of claims 13 to 19, the recombinant protein of claim 20, the multispecific antibody of claim 21, the nucleic acid molecule of claim 22, the expression vector of claim 23, the recombinant cell of claim 24, or the conjugate of claim 25; Optionally, a pharmaceutically acceptable excipient is further included.

26. A kit comprising: The antibody or antigen-binding fragment of any one of claims 13 to 18, the recombinant protein of claim 19, the multispecific antibody of claim 20, the nucleic acid molecule of claim 21, the expression vector of claim 22, the recombinant cell of claim 23, or the conjugate of claim 24.

27. Use of the antibody or antigen-binding fragment of any one of claims 13 to 18, the recombinant protein of claim 19, the multispecific antibody of claim 20, the nucleic acid molecule of claim 21, the expression vector of claim 22, the recombinant cell of claim 23 or the conjugate of claim 24 in the preparation of a medicament for preventing and treating diseases related to the novel coronavirus.

28. Use of the antibody or antigen-binding fragment of any one of claims 13 to 18, the recombinant protein of claim 19, the multispecific antibody of claim 20, the nucleic acid molecule of claim 21, the expression vector of claim 22, the recombinant cell of claim 23, or the conjugate of claim 24 in the preparation of a kit for detecting the novel coronavirus, diagnosing diseases related to the novel coronavirus, or evaluating the prognosis of diseases related to the novel coronavirus.

29. A method for detecting a novel coronavirus, the method being implemented by a computer device, the method comprising: The antibody or antigen-binding fragment of any one of claims 13 to 18, the recombinant protein of claim 19, the multispecific antibody of claim 20 or the conjugate of claim 24 is contacted with a sample to be detected to form an immune complex; Optionally, based on the signal of the immune complex, determining whether the sample to be tested contains the content of the new coronavirus; Optionally, the immune complex further comprises a second antibody that binds to the antibody or antigen-binding fragment; Optionally, the immune complex also includes a second antibody that binds to the new coronavirus.

30. A computer device, comprising a processor and a memory, wherein at least one computer program is stored in the memory, and the at least one computer program is loaded and executed by the processor to implement the operations performed by the antibody sequence generation method according to any one of claims 1 to 11, or to implement the operations performed by the method for detecting a new coronavirus according to claim 29.

31. A computer-readable storage medium, wherein at least one computer program is stored in the computer-readable storage medium, and the at least one computer program is loaded and executed by a processor to implement the operations performed by the antibody sequence generation method as described in any one of claims 1 to 11, or to implement the operations performed by the method for detecting a new coronavirus as described in claim 29.

32. A computer program product, comprising a computer program, which is loaded and executed by a processor to implement the operations performed by the antibody sequence generation method as described in any one of claims 1 to 11, or to implement the operations performed by a method for detecting a new coronavirus as described in claim 29.

Images