Molecule generation method and apparatus, molecule design method and apparatus, and electronic device

Abstract

The application relates to a molecule generation method and device, a molecule design method and device, and an electronic device. The molecule generation method comprises the following steps: obtaining a protein molecule comprising a protein pocket conformation and at least one starting fragment conformation matched with the protein pocket conformation; setting a current starting fragment conformation in the at least one starting fragment conformation to a current matching position of the protein pocket conformation, and determining a growth direction based on a starting growth site of the current starting fragment conformation; obtaining a growth fragment conformation, wherein the growth fragment conformation comprises a connection site and a first growth site; connecting the starting growth site and the connection site in the growth direction to obtain a grown starting fragment conformation; taking the grown starting fragment conformation as the current starting fragment conformation; and repeating the last two steps until a stop growth condition is met. The application can make the generated molecule structure matched with the protein pocket.

Description

Technical Field

[0001] This application relates to the fields of artificial intelligence and computer simulation technology, and in particular to a molecular generation method, apparatus, molecular design method, apparatus and electronic device. Background Technology

[0002] With the rapid development of computer technology and artificial intelligence technology, computer simulation technology is being applied to more and more scenarios, such as materials design and drug design.

[0003] However, the applicant found that the relevant technology is prone to mismatches between the simulated molecular structure and the protein pocket. Summary of the Invention

[0004] To address or partially address the problems existing in related technologies, this application provides a molecular generation method, apparatus, molecular design method, apparatus, and electronic device that enables the generated molecular structure to match a protein pocket.

[0005] The first aspect of this application provides a molecular generation method, comprising: S1, obtaining a protein molecule including a protein pocket conformation and at least one starter fragment conformation matching the protein pocket conformation, the protein pocket conformation including at least one matching position, and each of the at least one starter fragment conformation including an initiation growth site; S2, setting the current starter fragment conformation of the at least one starter fragment conformation at the current matching position of the protein pocket conformation, and determining a growth direction based on the initiation growth site of the current starter fragment conformation; S3, obtaining a growth fragment conformation, the growth fragment conformation including a connection site and a first growth site; S4, connecting the initiation growth site and the connection site in the growth direction to obtain a grown starter fragment conformation; S5, using the grown starter fragment conformation as the current starter fragment conformation, the first growth site as the initiation growth site, and the growth direction of the grown starter fragment conformation as the initiation growth direction; S6, repeating steps S3-S5. The growth process is repeated once or multiple times until the growth cessation condition is met, resulting in the initial fragment conformation after multiple growths. The initial fragments after multiple growths are the candidate molecules. Each growth fragment conformation is located within a different specific space, which includes some atoms in the protein molecule and / or some atoms in the current initial fragment conformation in the cycle in which each growth fragment is located. The current initial fragment conformation, the growth fragment conformation, the initial fragment conformation after growth, and the initial fragment conformation after multiple growths are all separated from the protein molecule. The growth cessation condition includes at least one of the following: the number of repetitions reaches a preset number, the molecular weight of the initial fragment after multiple growths is greater than or equal to a molecular weight threshold, and the number of atoms in the initial fragment after multiple growths is greater than or equal to an atomic number threshold.

[0006] The second aspect of this application provides a design method, comprising: generating a molecular structure according to the above-described molecular generation method; and designing candidate drugs or materials based on the molecular structure.

[0007] A third aspect of this application provides a molecular generation apparatus, comprising: a conformation acquisition module, a first growth module, a growth fragment conformation acquisition module, a second growth module, a starting fragment conformation update module, and a candidate molecule acquisition module. The conformation acquisition module is used to acquire a protein molecule including a protein pocket conformation and at least one starting fragment conformation matching the protein pocket conformation. The protein pocket conformation includes at least one matching position, and each of the at least one starting fragment conformation includes an initiation growth site. The first growth module is used to set the current starting fragment conformation of the at least one starting fragment conformation at the current matching position of the protein pocket conformation, based on a growth direction determined by the initiation growth site of the current starting fragment conformation. The growth fragment conformation acquisition module is used to acquire a growth fragment conformation, which includes a connection site and a first growth site. Each growth fragment conformation is located within a different specific space, which includes partial atoms of the protein molecule and / or partial atoms of the current starting fragment conformation in the cycle in which each growth fragment is located. The current starting fragment conformation, the growth fragment... The segment conformation, the starting fragment conformation after growth, and the starting fragment conformation after multiple growths are all separated from the protein molecule. The second growth module is used to connect the starting growth site and the connection site in the growth direction to obtain the starting fragment conformation after growth. The starting fragment conformation update module is used to use the starting fragment conformation after growth as the current starting fragment conformation, the first growth site as the starting growth site, and the growth direction of the starting fragment conformation after growth as the starting growth direction. If the candidate molecule acquisition module meets the growth cessation condition, it obtains the starting fragment conformation after multiple growths, and the starting fragment after multiple growths is the candidate molecule. The growth cessation condition includes at least one of the following: the number of repetitions reaches a preset number, the molecular weight of the starting fragment after multiple growths is greater than or equal to a molecular weight threshold, and the number of atoms of the starting fragment after multiple growths is greater than or equal to an atomic number threshold.

[0008] A fourth aspect of this application provides a design apparatus, comprising: a molecular structure generation module for generating a molecular structure according to the molecular generation method described above; and a design module for designing candidate drugs or materials based on the molecular structure.

[0009] The fifth aspect of this application provides an electronic device, including: a processor; and a memory having executable code stored thereon, which, when executed by the processor, causes the processor to perform the method described above.

[0010] A sixth aspect of this application also provides a computer-readable storage medium having executable code stored thereon, which, when executed by a processor of an electronic device, causes the processor to perform the above-described method.

[0011] The seventh aspect of this application also provides a computer program product including executable code that, when executed by a processor, implements the above-described method.

[0012] The molecular generation method, apparatus, molecular design method, apparatus, and electronic device provided in this application, after setting the conformation of the starting fragment at a matching position in the guarantee pocket, perform fragment growth within a specific space. The resulting fragment is then used as a new starting fragment for further fragment growth within the newly constructed specific space. This growth process is repeated until the generation cessation condition is met. Because molecules are generated using fragment conformations as the basic unit, the generated result contains information about the molecule and its conformation. This method ensures that the generated molecule's shape matches the protein pocket.

[0013] Furthermore, in some embodiments of this application, the growth fragment can be a fragment obtained by cleaving a real drug-like molecule according to specified rules. During cleavage, the neighboring atom types at the cleavage position are preserved, allowing molecules to be grown based on these neighboring atom types. This fragment preparation method effectively improves the syntheticability of the generated molecule and reduces the probability of developing features unsuitable for drug development.

[0014] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0015] The above and other objects, features and advantages of this application will become more apparent from the more detailed description of exemplary embodiments thereof in conjunction with the accompanying drawings, wherein the same reference numerals generally represent the same components in the exemplary embodiments thereof.

[0016] Figure 1 This illustration schematically shows an exemplary system architecture to which molecular generation methods, apparatuses, molecular design methods, apparatuses, and electronic devices can be applied according to embodiments of this application;

[0017] Figure 2 The illustration shows a scenario in which molecular generation methods, apparatuses, molecular design methods, apparatuses, and electronic devices can be applied according to embodiments of this application;

[0018] Figure 3 A flowchart illustrating a molecular generation method according to an embodiment of this application is shown schematically;

[0019] Figures 4A to 4C A schematic diagram of a specific space according to an embodiment of this application is shown;

[0020] Figure 5 The schematic diagram illustrates the protein molecule structure and origin fragment conformation according to embodiments of this application;

[0021] Figure 6 This schematic diagram illustrates the structure of the starting segment conformation after two growths according to an embodiment of this application.

[0022] Figure 7 A schematic diagram illustrating a virtual atom and a neighboring real atom according to an embodiment of this application is shown.

[0023] Figure 8 A schematic diagram illustrating the growth direction according to an embodiment of this application is shown.

[0024] Figures 9A-9B A schematic diagram of sampling points according to an embodiment of this application is shown;

[0025] Figure 10 This illustration schematically shows a scoring diagram based on sampling points according to an embodiment of this application;

[0026] Figures 11-13 This illustration schematically shows a diagram of molecular structure segmentation according to an embodiment of this application;

[0027] Figure 14 A schematic diagram illustrating the skeleton clustering results according to an embodiment of this application is shown.

[0028] Figure 15 A logic diagram of a molecular generation method according to an embodiment of this application is illustrated schematically;

[0029] Figure 16 A flowchart illustrating a design method according to an embodiment of this application is shown schematically.

[0030] Figure 17 A block diagram of a molecular generation apparatus according to an embodiment of this application is shown schematically;

[0031] Figure 18 A block diagram of a design device according to an embodiment of this application is schematically shown;

[0032] Figure 19 A block diagram of an electronic device according to an embodiment of this application is shown schematically. Detailed Implementation

[0033] Embodiments of this application will now be described in more detail with reference to the accompanying drawings. While embodiments of this application are shown in the drawings, it should be understood that this application may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to make this application more thorough and complete, and to fully convey the scope of this application to those skilled in the art.

[0034] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to limit the application. The terms "comprising," "including," etc., as used herein indicate the presence of features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0035] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein are to be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0036] It should be understood that although the terms "first," "second," "third," etc., may be used in this application to describe various information, this information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0037] Before describing the technical solution of this application, some of the technical terms in this field will be explained.

[0038] A molecular descriptor is a data structure that represents a molecule in a way that a computer program can process.

[0039] The three rules for drug-like organisms are three principles selected from the five rules for drug-like organisms for screening fragments (molecular weight less than 300, ClogP <= 3, and the number of hydrogen bond donors and hydrogen bond acceptors both not exceeding 3). ClogP refers to the lipid-water partition coefficient.

[0040] Collision number (also known as conflict number) refers to the number of atoms that collide or intersect with a protein pocket. Collisions or intersections occur when atoms on a fragment are in a non-disjoint state with atoms in a protein pocket. For example, two spheres are tangent to each other. Spheres are tangent but cannot overlap because there are electrons outside the atoms; it is not feasible for atoms to be too close together.

[0041] The number of adjacent pocket surfaces (also known as the close contact number) is the number of atoms in a fragment that come into contact with a pocket but do not collide.

[0042] The number of pocket spaces (also known as the volume number) is the total number of a segment minus the number of pockets clash_number and close_contact_number.

[0043] The van der Waals radius is a definition of an atomic radius.

[0044] MURCKO_SCAFFOLD is a type of compound skeleton definition designed by Murcko et al., which is often used for the shape and structure analysis of drugs.

[0045] Protein binding pockets are cavities on or inside a protein surface suitable for binding to ligands. The amino acid residues surrounding the pocket determine its shape, location, physicochemical properties, and function. The dynamics of these pockets are crucial for specific protein-protein interactions.

[0046] Angstrom An angstrom is a unit of length measurement; 1 angstrom = 0.1 nanometer.

[0047] Traditional pharmaceutical research and development is characterized by its lengthy process and high failure rate; for example, nearly three-quarters of pre-research drugs fail due to efficacy or safety issues. Compared to traditional methods such as computer-aided drug design, artificial intelligence-driven drug development (AIDD) has gradually gained recognition and application in recent years.

[0048] For example, in the early design phase of drug development, identifying small molecules with excellent activity, drug-like properties, and synthetic feasibility against a specific disease-related protein target is a crucial task. Molecular generation, as a technique that enables rapid sampling of molecules with certain characteristics in this stage, has received widespread attention in the pharmaceutical industry.

[0049] During drug development, the applicant discovered that the molecular generation methods of related technologies yield molecular results that can be SMILES strings representing two-dimensional molecular structures, lacking specific conformational information. This leads to a high likelihood of mismatches between the resulting molecules and protein pockets. Furthermore, these two-dimensional molecules sometimes exhibit characteristics unsuitable for drug development, such as toxicity, instability, poor absorption, difficulty in synthesis, or explosiveness. Additionally, even if the resulting molecules are structurally plausible, they may be unsuitable for drug design.

[0050] Generating molecules and molecular conformations that conform to the protein's α-conformation is a problem that urgently needs to be solved. Furthermore, ensuring that the generated molecules and molecular conformations are suitable for drug development is also a problem that urgently needs to be solved.

[0051] This application embodiment uses growth fragment conformations as the basic unit to perform molecular growth in a specific space located inside a protein pocket conformation. This ensures that the generated molecule includes conformational information and that the molecular conformation matches the shape of the protein pocket conformation. Furthermore, the growth fragment can be a fragment obtained by cleaving a real drug-like molecule according to specified rules, and the types of neighboring atoms at the cleavage site are preserved during cleavage. This method of obtaining growth fragments ensures the syntheticability of the generated molecule and effectively reduces the probability of unsuitable drug-like characteristics.

[0052] The following will be through Figures 1 to 19 This application provides a detailed description of a molecular generation method, apparatus, molecular design method, apparatus, and electronic device according to embodiments of the present application.

[0053] Figure 1 This illustration schematically depicts an exemplary system architecture to which molecular generation methods, apparatuses, molecular design methods, apparatuses, and electronic devices can be applied according to embodiments of this application. It should be noted that... Figure 1 The examples shown are merely examples of system architectures that can be applied to the embodiments of this application, in order to help those skilled in the art understand the technical content of this application, but do not mean that the embodiments of this application cannot be used in other devices, systems, environments or scenarios.

[0054] See Figure 1 The system architecture 100 according to this embodiment may include terminal devices 101, 102, and 103, a network 104, and a server 105. The network 104 serves as a medium for providing a communication link between the terminal devices 101, 102, and 103 and the server 105. The network 104 may include various connection types, such as wired or wireless communication links, or fiber optic cables, etc.

[0055] Users can use terminal devices 101, 102, and 103 to interact with other terminal devices and server 105 via network 104 to receive or send information, such as sending conformation requests, fragment library requests, simulated growth requests, and scoring calculation requests, and receiving fragment conformations, fragment libraries, growth results, and scoring results. Terminal devices 101, 102, and 103 can be equipped with various communication client applications, such as drug development applications, materials design applications, web browser applications, database applications, simulation applications, search applications, instant messaging tools, email clients, and social media platforms.

[0056] Terminal devices 101, 102, and 103 include, but are not limited to, smart desktop computers, tablet computers, laptop computers, and other electronic devices that can support functions such as internet access, modeling, analysis and calculation, and design.

[0057] Server 105 can receive and send configuration requests, fragment library requests, simulated growth requests, scoring calculation requests, etc., and can also be used to send fragment configurations, fragment libraries, growth results, and scoring results to terminal devices 101, 102, and 103. For example, server 105 can be a backend management server, a server cluster, etc.

[0058] It should be noted that the number of terminal devices, networks, and servers is merely illustrative. Depending on the implementation requirements, any number of terminal devices, networks, and cloud components can be included.

[0059] Figure 2 The illustration shows a scenario in which molecular generation methods, apparatuses, molecular design methods, apparatuses, and electronic devices can be applied according to embodiments of this application.

[0060] See Figure 2 In this embodiment, the molecular growth process can be simulated within the protein pocket conformation using the actual growth direction of the growth fragment conformation. Specifically, the growth fragment is used as the basic unit for successive epitaxial growth until the growth cessation condition is met. This ensures that the molecular conformation obtained through the molecular generation method is compatible with the protein pocket conformation, resulting in a suitable molecular conformation for drug development. Furthermore, when the growth fragment is a segment extracted from a real drug-like molecule, the aforementioned molecular conformation exhibits good manufacturability and drug-like characteristics.

[0061] Figure 3 A flowchart illustrating a molecular generation method according to an embodiment of this application is shown schematically.

[0062] like Figure 3 As shown, this embodiment provides a method for molecule generation, which includes operations S1 to S6.

[0063] In operation S1, a protein molecule including a protein pocket conformation and at least one starter fragment conformation matching the protein pocket conformation are obtained, the protein pocket conformation including at least one matching position and each of the at least one starter fragment conformation including an initiation growth site.

[0064] Proteins are extremely important macromolecular organic compounds in living organisms, making up as much as 54% of the dry weight of the human body. Proteins are mainly composed of amino acids, and different combinations and arrangements of these amino acids create various types of proteins. Proteins serve as receptors for many drugs.

[0065] A protein-binding pocket (hereinafter referred to as a protein pocket) is a cavity on or inside a protein surface suitable for binding to ligands. The dynamics of the protein pocket are crucial for the specific interactions of proteins. The flexibility and mobility of protein structure allow the binding pocket to open, close, and adapt, thereby regulating the ligand binding process and enabling specific protein functions. Drug molecules developed that match the protein pocket can exert their drug effects more effectively.

[0066] At the molecular level, protein pockets inhibit or activate specific biological functions by binding to small molecules. For example, a three-dimensional (3D) protein pocket conformation of a target protein can be given to determine a matching small molecule conformation. The 3D protein pocket conformation includes geometric information about the protein pocket.

[0067] Protein active sites and protein binding sites refer to the regions where proteins bind to other substances and play important roles when they are physiologically active. An active site or binding site can bind to several different origin fragment conformations. For example, at least one origin fragment conformation matching a protein pocket conformation can be determined based on detection, experimentation, or experience.

[0068] In one specific embodiment, the user may provide the protein file and the starting fragment of the protein pocket.

[0069] In operation S2, the current starter fragment conformation in at least one starter fragment conformation is set at the current matching position of the protein pocket conformation, and the growth direction is determined based on the starting growth site of the current starter fragment conformation.

[0070] In this embodiment, various simulation modeling software can be used to simulate the growth process. For example, by modeling, the current starting fragment conformation in at least one starting fragment conformation can be set to the current matching position in the protein pocket conformation.

[0071] The starting fragment conformation has a specific spatial structure, which, due to various reasons such as the fact that the distance between two atoms or groups is less than the sum of their van der Waals radii, will result in repulsion; the conformation of the molecular chain in the crystal depends on the mutual repulsion or attraction of the groups on the molecular chain. Therefore, after setting the starting fragment conformation at a specific position in the protein pocket conformation, the growth direction of the initial growth site of the current starting fragment conformation can be obtained based on the matching position of the protein pocket conformation and the starting fragment conformation.

[0072] It should be noted that the growth fragment conformation, the origin fragment conformation, and the protein pocket conformation are separated (i.e., they do not contact each other; if the distance is too small, they will be subject to greater repulsive forces), but there is a connection between the growth fragment conformation and the origin fragment conformation.

[0073] In operation S3, a growth fragment conformation is obtained, which includes a connection site and a first growth site.

[0074] In this embodiment, possible growth fragments can be selected from a fragment library. These growth fragments can be connected to the starting growth site with a specific growth fragment configuration. The growth fragment configuration can include attitude information of the growth fragment relative to a specific coordinate system, such as rotation angle.

[0075] In operation S4, in the growth direction, the starting growth site is connected to the connection site to obtain the conformation of the starting fragment after growth.

[0076] Connecting the initiation growth site with the connection site to obtain the conformation of the growth origin fragment can be achieved using a variety of related techniques, which are not limited here.

[0077] In operation S5, the grown starting segment conformation is taken as the current starting segment conformation, the first growth site is taken as the starting growth site, and the growth direction of the grown starting segment conformation is taken as the starting growth direction.

[0078] In this embodiment, the molecular conformation after two growth cycles based on the growth fragment can be obtained in the manner described above. This molecular conformation may be the final molecular conformation or an intermediate result, depending on whether the growth cessation condition is met.

[0079] In operation S6, repeat operations S3-S5 once or multiple times until the growth cessation condition is met, and obtain the starting fragment conformation after multiple growths. The starting fragment after multiple growths is the candidate molecule.

[0080] In this embodiment, the growth-stopping conditions include at least one of the following: the number of repetitions reaches a preset number, the molecular weight of the starting fragment after multiple growths is greater than or equal to a molecular weight threshold, and the number of atoms of the starting fragment after multiple growths is greater than or equal to an atomic number threshold.

[0081] Each growth fragment conformation resides within a distinct, specific space, which includes atoms from the protein molecule and / or atoms from the current starting fragment conformation within the cycle in which each growth fragment is located. The current starting fragment conformation, the growth fragment conformation, the post-growth starting fragment conformation, and the starting fragment conformation after multiple growth cycles are all separated from the protein molecule. This ensures that the generated molecular conformation matches the protein pocket conformation, reducing the likelihood of the molecular conformation failing to bind to the protein pocket due to mechanical interference or other factors.

[0082] Figures 4A to 4C A schematic diagram of a specific space according to an embodiment of this application is shown. Figure 4A A specific space within the cylinder is shown. Figure 4B The specific space of the sphere is shown. Figure 4C A specific space of the cube is shown.

[0083] It should be noted that, Figures 4A to 4C The specific spaces shown are merely illustrative and should not be construed as limiting this application. The following primarily uses a cylinder as an example; rotation of the cylinder itself does not affect the overall position and is easy to calculate. If a cube or other similar shape were used, the protein sampling information after rotation would need to be considered, increasing complexity. For example, for each fragment growth, a cylinder needs to be constructed as the specific space for molecular growth. Furthermore, from a usage perspective, the cylinder size can be a preset value, but this preset value can be adjusted (the principle of adjustment is to ensure that reference information (starting fragment, growth fragment) is included). A larger cylinder size provides more reference information and greater accuracy, but excessive size may lead to excessive unnecessary noise; the specific size may depend on the actual situation, for example, the cylinder height may be greater than or equal to 5 angstroms.

[0084] For example, the starting point can be redefined based on the entire molecule that has been screened and connected to a growth segment. Then, the above steps can be repeated to grow the molecule until the preset termination conditions are met, such as how many growth segments have been grown from the initial starting segment, or the molecular weight of the entire molecule, the number of atoms in the entire molecule, etc.

[0085] In this embodiment, fragment conformations are used as the basic unit to generate molecules within a virtual protein pocket. The simulated generation result includes the molecule and its conformational information. Compared to the SMILES string representing a two-dimensional molecular structure in related technologies, the molecular conformation generated in this embodiment ensures that the generated molecule matches the shape of the protein pocket.

[0086] In some embodiments, the growth fragment conformation is obtained by: obtaining a set of growth fragment conformations that are compatible with the starting fragment conformation from a fragment library, and obtaining the growth fragment conformation from the set of growth fragment conformations according to a preset standard, which includes: random selection, matching degree from high to low, or matching degree from low to high.

[0087] Figure 5 The schematic diagram illustrates the protein molecular structure and origin fragment conformation according to embodiments of this application.

[0088] A protein molecule's structure includes a protein pocket, which may contain multiple binding sites. Each binding site can bind to one or more origin fragment conformations. Once an origin fragment conformation binds to a binding site within the protein pocket, the growth direction of the next growth fragment is determined.

[0089] Figure 6 The schematic diagram illustrates the structure of the starting segment conformation after two growths according to an embodiment of this application.

[0090] See Figure 6 Once the starting fragment conformation is set at the matching position, its growth direction is determined. A specific space is constructed based on this growth direction, and the next growth fragment is simulated within this space to obtain a suitable growth fragment conformation. At this point, the growth direction of the next growth fragment of the previous growth fragment conformation is determined. Next, a new specific space is constructed based on this growth direction, and the next growth fragment is simulated within this space to obtain a suitable growth fragment conformation. This process is repeated until the growth cessation condition is met. At this point, the starting fragment conformation and all the grown growth fragment conformations together constitute a molecule adapted to the protein pocket.

[0091] In some embodiments, the fragments in the fragment library may be obtained by segmenting molecules from data developed by the company itself or from commercial datasets. For example, the dataset may be a dataset such as ZINC15 / ZINC20 (containing tens of millions of small molecule compounds).

[0092] The dataset can also include drug-like small molecule data, which can also be obtained through various means. For example, it could be a collection of all drug-like small molecule data that are freely available through public channels, such as the CHEMBL dataset containing biologically active drug-like small molecules.

[0093] Datasets can also include datasets from MoleculeNet and physicochemical properties from within the applicant's database. For example, the BBBP dataset contains data on molecules that have been measured to cross the blood-brain barrier. The ClinTox dataset includes data on FDA-approved drugs and those for which clinical trials failed due to toxicity issues. The Tox21 dataset includes data on the measured toxicity of compounds to 12 different targets, including experimental measurements of nuclear receptors and stress response pathways. The HIV dataset includes data on small molecules with the ability to inhibit HIV replication, as determined experimentally.

[0094] In some embodiments, for the first growth fragment connected after the starting fragment, it is necessary to traverse the fragment library (preferably by combining the specific information of each growth fragment in the fragment library (segmentation position information, neighboring atom types, etc.) to select candidate growth fragments that can be traversed), and then use a scoring function (considering only spatial position) for screening. Preferably, it can also be further screened by whether there is a preset interaction between the current proposed growth fragment position and the protein.

[0095] The following provides an exemplary description of how to obtain a set of growth fragment conformations that are compatible with the starting fragment conformation from a fragment library, and how to obtain growth fragment conformations from the set of growth fragment conformations.

[0096] In some embodiments, obtaining a set of growth fragment conformations adapted to the starting fragment conformation from a fragment library may include the following operations.

[0097] First, candidate growth fragments are selected from the fragment library based on the attribute information of the starting fragment conformation. The attribute information includes at least one of the following: the splitting position information or the type of neighboring atoms.

[0098] Then, by rotating the candidate growth fragments, the growth fragment conformations are obtained to construct a set of growth fragment conformations.

[0099] For each candidate segment connected to the starting segment, since the candidate segment is rotatable, it needs to be rotated one full revolution at a preset angle, and filtering must be performed after each rotation. Therefore, theoretically, the same candidate segment may have multiple possible connection methods that pass the filtering due to different connections (rotations) with the starting segment.

[0100] In some embodiments, after obtaining the growth fragment conformation, the above method may further include the following operation: determining a target growth fragment conformation from multiple growth fragment conformations corresponding to the current candidate growth fragment, so as to construct a growth fragment conformation set based on the target growth fragment conformation, wherein the stability of the target growth fragment conformation is higher than that of the non-target growth fragment conformations.

[0101] The different spatial arrangements of atoms in a molecule, resulting from rotation around single bonds, are called conformations. A compound can have numerous conformations, affecting its stability, reactivity, and selectivity. The stability of a compound's structure refers to its ability to maintain its original equilibrium state under load. Factors affecting conformational stability mainly include: torsional strain, nonbonded interactions, dipole-dipole interactions, stereoelectronic effects, hydrogen bonds, conjugation effects, and adjacent crossover effects.

[0102] For example, molecular mechanics (MM) can be used to analyze molecular structure and stability. Specifically, chemical bonds exist between atoms in a molecule, with standard bond lengths and bond angles. Non-bonded interactions also exist within the molecule. By adjusting the molecule's own conformation, the optimal arrangement of atomic positions is obtained, resulting in a stable conformation.

[0103] In some embodiments, the fragment library includes multiple fragment classes, each fragment class including at least one candidate growth fragment. Accordingly, obtaining a set of growth fragment conformations adapted to the starting fragment conformation from the fragment library may include the following operations.

[0104] First, candidate growth fragments with a first preset ratio are obtained from multiple fragment classes to generate a first sampled fragment class. Then, by traversing this first sampled fragment class, the target fragment class containing the candidate growth fragments is determined from among the multiple fragment classes. The first preset ratio can be determined based on experience or usage results. If rapid molecule generation is desired, the first preset ratio can be set lower, such as 1%, 5%, 10%, 15%, 20%, or 30%. If generating as many molecules as possible is desired, the first preset ratio can be set higher, such as 40%, 45%, 50%, 60%, 70%, or 80%.

[0105] Then, by traversing the target fragment class, at least one growing fragment configuration that can be successively extended on the current starting fragment configuration is obtained to generate a set of growing fragment configurations.

[0106] In this embodiment, the sampling process can effectively reduce the number of fragment classes that need to be traversed, thereby improving the efficiency of molecule generation.

[0107] In some embodiments, the fragment library includes multiple fragment classes, each fragment class including at least one candidate growth fragment. Accordingly, for each of the multiple fragment classes, obtaining the set of growth fragment conformations adapted to the starting fragment conformation from the fragment library may include the following operations.

[0108] First, candidate growth fragments with a second preset ratio are obtained from the fragment class to generate a second sampling fragment class. The second preset ratio may be the same as or different from the first preset ratio.

[0109] Then, the second sample fragment class is traversed until a third candidate growth fragment set matching the neighboring atom type of the connection site is obtained; each candidate growth fragment in the third candidate growth fragment set is rotated around the growth direction according to a preset step size to obtain multiple third candidate growth fragment conformations; the third candidate growth fragment conformations are connected to the starting growth site one by one according to the growth direction.

[0110] If at least one of the multiple third candidate growth fragment conformations is located inside a specific space, then the fragment class to which the second sampled fragment class belongs is traversed to obtain at least one set of fourth candidate growth fragments that matches the neighboring atom type of the connection site; the fourth candidate growth fragments are rotated around the growth direction according to a preset step size to obtain multiple fourth candidate growth fragment conformations; the conformation located inside the specific space among the multiple fourth candidate growth fragment conformations is taken as the growth fragment conformation.

[0111] In this embodiment, after obtaining a preset proportion of fragments according to the fragment classification and repeatedly performing growth, after the growth is stopped, it is determined whether there are any fragments that meet the requirements for growth. If so, all matching candidate fragments in each fragment class (growth fragment conformation set) are traversed, grown, filtered, and judged without setting a proportion. This effectively improves the efficiency of molecule growth and can obtain suitable molecules as comprehensively as possible.

[0112] Obtaining growth fragment conformations from the growth fragment conformation set according to preset standards may include the following operations.

[0113] First, a specific space is constructed based on the starting fragment conformation. The origin of the specific space is the spatial position of the virtual atoms of the starting fragment conformation, and the extension direction of the origin of the specific space is determined according to the growth direction of the initial growth site of the starting fragment conformation.

[0114] Figure 7 A schematic diagram illustrating a virtual atom and its adjacent real atom according to an embodiment of this application is shown.

[0115] See Figure 7 The diagram illustrates a growth segment conformation having neighboring real atoms and a growth direction based on those real atoms. Virtual atoms are atomic positions based on the growth direction that can connect to the growth segment conformation.

[0116] Figure 8 A schematic diagram illustrating the growth direction according to an embodiment of this application is shown.

[0117] See Figure 8The protein molecule structure includes a protein pocket. A starter fragment conformation can be set at the binding site (start site) of the protein pocket conformation. This allows the growth direction of the next growth fragment conformation to be determined, and a specific space can be constructed based on that growth direction.

[0118] Then, the set of growth fragment conformations is traversed to obtain candidate growth fragment conformations that match the starting fragment conformation and are located within a specific space when connected to the starting growth site. When a growth fragment conformation is located within a specific space, it has no intersection with the outer wall of that space.

[0119] In some embodiments, obtaining a growth fragment conformation may include the following operations.

[0120] First, the positional relationships between atoms of the candidate growth fragment conformation and atoms of the protein molecule and the origin fragment conformation are obtained by sampling.

[0121] Then, the growth segment conformation is determined from multiple candidate growth segment conformations based on positional relationships.

[0122] For example, the positional relationships between atoms can include the positional relationships between all atoms in the starting segment, all atoms in the growth segment, and all atoms in the protein molecule within a specific space. For instance, these positional relationships can be determined using atomic spherical coordinates and information on the interatomic distances between the protein within the protein pocket and the grown molecule for specific interactions.

[0123] In some embodiments, obtaining the positional relationship between atoms of the candidate growth fragment conformation and atoms of the protein molecule and the origin fragment conformation by sampling may include: if a first exclusion condition or a second exclusion condition is met, then obtaining the positional relationship between atoms of the candidate growth fragment conformation and atoms of the protein molecule and the origin fragment conformation by sampling.

[0124] For example, the first exclusion criterion includes: all atoms of the candidate growth fragment conformation are located inside a specific space.

[0125] For example, the second exclusion condition includes: all atoms of the candidate growth fragment conformation are located inside a specific space, and at least one of the following: the molecular weight of the molecular structure formed by the start fragment and the candidate growth fragment is less than or equal to a molecular weight threshold, and the number of atoms in the molecular structure formed by the start fragment and the candidate growth fragment is less than or equal to an atomic number threshold.

[0126] In some embodiments, obtaining the positional relationship between atoms of the candidate growth fragment conformation and atoms of the protein molecule and the origin fragment conformation by sampling may include the following operations.

[0127] First, by sampling points in a specific space, the scores of candidate growth fragment conformations relative to protein molecule and origin fragment conformations are obtained. These scores are related to collision point parameters and neighboring point parameters, or they are related to collision point parameters, neighboring point parameters, and at least one of the following: collision point weight parameters and neighboring point weight parameters. For example, a scoring function can be defined to calculate the scores of candidate growth fragment conformations relative to protein molecule and origin fragment conformations.

[0128] Then, the scoring results are used to characterize the positional relationships between the atoms of the candidate growth fragment conformation and the atoms of the protein molecule and the origin fragment conformation. For example, a higher score indicates a more suitable positional relationship between the atoms of the protein molecule and the origin fragment conformation, such as no overlap between them, or a higher filling rate within the protein pocket.

[0129] In some embodiments, obtaining a score of the candidate growth fragment conformation relative to the protein molecule and the origin fragment conformation by sampling points in a specific space may include the following operations.

[0130] First, by using sampling points in a specific space, the following are obtained: the number of adjacent points N1 where the distance between the candidate growth fragment conformation and the protein pocket conformation is less than a preset distance threshold; the number of collision points N2 between the candidate growth fragment conformation and the protein pocket conformation; and the number of non-empty points N3 among the sampling points in the specific space. The sampling points in the specific space can be preset sampling points, and each sampling point is used to sample a subspace within the specific space. For example, it is determined whether a sampling point simultaneously contains atoms of the starting growth fragment and atoms of the protein molecule; if so, the overlap between the two is determined.

[0131] If the number of collision points N2 is greater than zero, the score will be negative.

[0132] If the number of collision points N2 is zero, the score is a weighted sum of the number of adjacent points N1 and the number of non-empty points N3, where the first weight of the number of adjacent points N1 and the second weight of the number of non-empty points N3 have no common divisor other than 1.

[0133] For example, for multiple conformations of a growth segment, each is scored, and the conformation with the highest score is retained. If all scores for a growth segment are negative, then that growth segment is not selected.

[0134] For example, the scoring function for a growth segment can be shown in equation (1):

[0135] Score = w1 * N closecontact +w2*N volume Equation (1)

[0136] If N clashIf the value is greater than 0, then the Score = -50000. Where W1 and W2 are weighting coefficients, and N... clash N represents the number of collision points. closecontact N represents the number of points in the neighboring pocket. volume This represents the number of remaining non-empty points. For example, W1 and W2 are 1 and 10; 1 and 5; 2 and 1; 3 and 1, etc. In one specific embodiment, testing revealed that it is better if W1 and W2 have no common divisor other than 1. For example, W1 and W2 being 1 and 2 is better than W1 and W2 being 5 and 10.

[0137] It should be noted that before scoring, the growth fragment needs to be aligned with the coordinates of the cylindrical space, and the direction of the single bond of the growth fragment starting from the connection site should be consistent with the growth direction. After alignment, the virtual atom coordinates of the growth fragment are consistent with the real atom coordinates of the starting fragment, and vice versa.

[0138] Furthermore, the scoring function and the method for selecting the scoring results described above are merely illustrative examples. Other methods can also be used for scoring, such as pharmacophore-based scoring, pharmacophore and fragment scoring in cylindrical space, etc. Additionally, the method for selecting the scoring results can be to retain the top N or top N% of the highest scores, without limitation. N is a positive integer.

[0139] Figures 9A-9B A schematic diagram of sampling points according to an embodiment of this application is shown. Figure 9A The sampling points of a specific space in the shape of a cylinder are shown. Figure 9B The sampling points of a specific space in the cube shape are shown.

[0140] In some embodiments, the specific space includes a cylindrical space, with p sampling faces perpendicular to the growth direction. Each of the p sampling faces includes 360 / n sampling edges passing through the center of the circle, and each sampling edge includes m sampling points, where p, m, and n are integers greater than 2, and 360 is divisible by n.

[0141] In some embodiments, a particular space can be constructed in the following manner.

[0142] First, the first vector is generated based on the growth direction of the virtual atoms in the initial fragment conformation;

[0143] Then, starting from the virtual atoms of the initial fragment conformation, 360 / n second vectors are generated in a direction perpendicular to the first vector;

[0144] Next, for each second vector, m sampling points are generated along the direction of the second vector, using virtual atoms as base points;

[0145] Then, along the direction of the first vector, all sampling points are translated p times according to a preset step size to obtain sampling points in a specific space.

[0146] Figure 10 A schematic diagram illustrating scoring based on sampling points according to an embodiment of this application is shown. See also Figure 10 After setting the growth fragment conformation in a specific space, the growth fragment conformation can be scored based on the sampling results of each sampling point in the specific space.

[0147] For example, starting with virtual atoms and using the growth direction as the axis, construct a base with a radius of... Gao Wei The three-dimensional space of a cylinder, see Figure 8 As shown. This specific space is the growth location for the next segment.

[0148] First, take a base plane in this specific space. Starting from the intersection of the cylinder axis on this base plane, arbitrarily select a vector pointing towards the edge of the base plane. Rotate this vector around a 10-degree angle, resulting in 36 vectors. Specifically, rotate an initial vector 360 degrees. Here, 10 degrees is an empirical value; theoretically, any integer divisible by 360 is acceptable, such as 5 degrees, 15 degrees, 20 degrees, etc. This rotation angle determines the rotatable range of the segment. The smaller the value, the more reference segments there are, and the more accurate the calculation result. However, if the angle is too small, more computation is required.

[0149] Then, the vector group on this plane is arranged perpendicular to the axis, with... By performing step-size translation, 20 sets of vectors are obtained, ultimately yielding the radiation vectors of the test tube brush shape. Please refer to [the documentation / reference]. Figure 9A and Figure 10 A total of 36*20 radiation vectors were obtained. Figure 9A Only 5 sets of vectors are shown.

[0150] The virtual cylinder in this calculation is only a theoretical concept; in actual calculations, only the information of the 36*20*20 matrix is ​​considered. This process describes the calculation of this three-dimensional matrix. The above... The 0.5 in the step size translation is an empirical value, which is theoretically modifiable; smaller values ​​provide greater accuracy but increase computational complexity, and values ​​higher than 1 are not recommended. The unit of the step size can be the radius or diameter of a hydrogen atom, or other units can be chosen. Considering that molecules / fragments and proteins are both composed of atoms, this method uses an atomic sphere model, defining the atomic sphere radius as the atomic sphere radius to detect the spatial structure and positional relationships between proteins and fragments.

[0151] If two atoms come into contact without forming bonds, it is considered a collision. Similarly, if any two atoms (proteins, fragments, or molecules) come into contact without forming bonds, it is considered a collision. Based on these conditions, we can determine whether a collision occurs by considering whether two atomic spheres come into contact.

[0152] It should be noted that, for statistical purposes, this embodiment only considers the case along the radiation vector direction when actually detecting collisions. For example, collisions may also occur in directions other than the radiation vector direction. This collision might occur within an isosceles triangle with a vertex of 10 degrees and adjacent sides of 10 angstroms. The theoretical error within this range is detectable (a very small error is acceptable). For each radiation vector... For intervals at Take 20 points within the range (there are a total of 36*20*20 points) and consider whether these points collide.

[0153] In this embodiment, molecular generation is performed using growth fragment conformations as the basic unit, and the generated result includes the molecule and its conformational information. This method ensures that the generated molecule matches the shape of the protein pocket. Furthermore, directly generating molecular conformations corresponding to the protein pocket location eliminates the need for conformation generation and docking steps, effectively improving the efficiency of molecular design and screening.

[0154] In some embodiments, the above method may further include the following operation: if at least one matching position includes an untraversed matching position, the untraversed matching position is used as the updated matching position, and the starting fragment conformation is set at the updated matching position of the protein pocket conformation to obtain an updated growth direction for the connection site, so as to connect the starting growth site and the connection site in the updated growth direction to obtain the grown starting fragment conformation.

[0155] Please see also Figure 5 The starting fragment construct can be set in addition to being able to be set in the starting fragment construct. Figure 5 The matching positions shown can also be set at other matching positions. After completing the traversal of molecular fragments at a certain matching position for a starting fragment conformation, the starting fragment conformation can be moved to another matching position, and the traversal of molecular fragments can be performed again until all matching positions for that starting fragment conformation have been traversed.

[0156] In some embodiments, the above method may further include operations S21 to S22.

[0157] In operation S21, if at least one starter fragment conformation includes an untraversed starter fragment conformation, the current untraversed starter fragment conformation is set at the current matching position of the protein pocket conformation, and the growth direction is determined based on the starting growth site of the current untraversed starter fragment conformation.

[0158] In operation S22, repeat steps S21, S3-S5 once or more until the growth cessation condition is met, and obtain the starting fragment conformation after multiple growths. The starting fragment after multiple growths is the candidate molecule.

[0159] Please see also Figure 5 Multiple origin fragment conformations can exist that can match a protein pocket. Once a certain origin fragment conformation has been traversed, a new origin fragment conformation can be used for the next round of traversal.

[0160] Through the above embodiments, a molecular growth process targeting a specific protein pocket can be achieved, resulting in molecules that are compatible with the protein pocket.

[0161] The following provides an example of information related to the fragment library.

[0162] In some embodiments, the above method may further include constructing a fragment library. The grown fragments in the fragment library include fragments obtained by trimming from multiple candidate molecular structures, and the grown fragments have cleavage position information and neighboring atom type information corresponding to the cleavage position information.

[0163] For example, building a fragment library can include the following operations.

[0164] First, multiple candidate molecular structures are segmented based on at least one of rotatable bonds, molecular backbones, or functional groups to obtain multiple candidate growth fragments.

[0165] Then, a fragment library is constructed based on multiple candidate growth fragments. Various related techniques can be used, and no specific limitations are specified here.

[0166] Figures 11-13 The illustration shows a schematic diagram of molecular structure segmentation according to an embodiment of this application.

[0167] See Figure 11 The C-C single bond between the two benzene rings is a rotatable bond, which can be used to cleave the growth into two segments. See also Figure 12 The oxygen bonds were broken down based on functional groups, and the resulting functional groups were used as growth fragments. See also Figure 13 The C / C bonds were split based on functional groups, and the resulting two functional groups were used as growth fragments. Furthermore, Figure 13 The functional groups in the left half of the structure can also be divided.

[0168] In some embodiments, after obtaining multiple candidate growth fragments, the above method can also perform the following operation: determine the selected growth fragment from multiple candidate growth fragments based on the correspondence between the starting growth site and the atom type obtained from the molecule segmentation. The correspondence is determined based on the segmentation position information and the neighboring atom type information corresponding to the segmentation position information.

[0169] Accordingly, a fragment library is constructed based on multiple candidate growth fragments, including: constructing a fragment library based on multiple selected growth fragments.

[0170] For example, download all molecules marked as Lead-Like from the ZINC20 database, approximately 3.94 million smils. After downloading, these molecules are fragmented according to rotatable bonds. During fragmentation, the neighboring atomic information of the fragment connection sites must be preserved to ensure that no abnormal connection relationships occur during fragment growth.

[0171] See Figure 13 After the C4H single bond is broken, it can be determined that the free radical can be bonded to the C atom, and the breaking point is at C4H. 11 The single bond between the third c and the fourth c.

[0172] In some embodiments, a specific method for constructing a fragment library based on multiple candidate growth fragments includes at least one of the following operations.

[0173] First, remove candidate growth fragments that contain specific elements.

[0174] Then, remove duplicate candidate growth segments after regularization.

[0175] Next, the type of the starting atom for the connection site of the candidate growth fragment is determined based on the molecular structure information before segmentation.

[0176] Then, candidate growth fragments are filtered based on molecular weight, the number of hydrogen bond donors, and the number of hydrogen bond acceptors.

[0177] Next, candidate growth fragments are filtered based on expert experience.

[0178] In some embodiments, the conformation may also be compressed for storage.

[0179] Specifically, the above method may further include the following operations: First, for each growth fragment in the fragment library, a growth fragment conformation is generated, and the growth fragment conformation is encoded and compressed. Then, the encoded and compressed growth fragment is read and decoded to obtain the growth fragment.

[0180] In this embodiment, after screening, due to the large file size and slow reading speed, the selected candidate growth fragments are encoded and compressed to reduce storage requirements and improve computation speed. Compression effectively addresses the problem of excessive file memory usage and significantly improves the speed of reading and decoding. Decoding is then performed only when the specific file is needed.

[0181] In one specific embodiment, the following steps are performed after the segmentation is completed:

[0182] First, fragments containing elements such as P, B, Si, and Sn are removed, as these elements are not present or are uncommon in drug-like molecules.

[0183] Then, remove the duplicate segments after regularization.

[0184] Next, based on the molecular information before segmentation, the starting atom type of each connection site of each fragment is statistically analyzed.

[0185] Then, according to Ro3 filtration, specifically, filtering out molecules with a molecular weight less than 300, ClogP <= 3, and hydrogen bond donors and acceptors with a number not exceeding 3.

[0186] Next, filtering is performed based on conditions derived from human experience. For example, a seven-membered ring is used as the upper limit for the number of large rings in a segment, with a maximum of five rings, and the number of chiral centers is limited to a maximum of three, etc.

[0187] Then, the fragment's conformation is generated and compressed. Specifically, based on the atomic sphere model, conformations exceeding the spatial bounds are removed. For example, the spatial bounds are those with a base radius of... Gao Wei Cylindrical space.

[0188] It should be noted that the method of molecular segmentation is not limited to rotatable bonds (it can preserve certain types of single bonds or split certain double bonds, etc.). There is no absolutely clear standard for the scale / size of fragments, but it is necessary to ensure that the spatial volume of the grown fragment is smaller than the volume of a specific space. Fragment screening needs to be based on the specific circumstances of the molecular library, and the screening conditions rely on human experience and data analysis of the molecular library.

[0189] Taking the ZINC20 database as an example, the fragments were divided according to the atomic type of the fragment starting point, and the statistical results are shown in Table 1.

[0190] Table 1

[0191] Starting point atom type Total number of segments C 81767 N 26714 O 27 S 219

[0192] In some embodiments, the fragment library can be further categorized and organized to increase its usability.

[0193] Specifically, constructing a fragment library based on multiple candidate growth fragments can include the following operations.

[0194] First, backbone clustering is performed on multiple candidate growth fragments to obtain multiple fragment classes. See [link / reference] Figure 14 The two segments above share the same skeleton, therefore, these two growth segments can be classified into the same skeleton category. For example, MURCKO_SCAFFOLD can be used to calculate the molecular skeleton corresponding to the growth segment, calculate the 128-bit Morgan fingerprint of the molecular skeleton, and finally use the KMeans algorithm to classify the Morgan fingerprint into 24 categories.

[0195] Then, add multiple fragment classes to the fragment library, or for each of the multiple fragment classes, extract a preset proportion or a preset number of fragments from the fragment class according to a preset sampling rate and add them to the fragment library.

[0196] Similarly, in some embodiments, the fragment library can be optimized in the following ways. Specifically, after constructing the fragment library based on multiple candidate growth fragments, the above method may further include the following operations.

[0197] First, the skeletons of multiple candidate growth fragments are clustered to obtain multiple fragment classes.

[0198] Then, add multiple fragment classes to the fragment library, or for each of the multiple fragment classes, extract a preset proportion or a preset number of fragments from the fragment class according to a preset sampling rate and add them to the fragment library.

[0199] In this embodiment, the fragments are labeled and categorized during library construction. This allows for the selection of the same number / proportion of fragments by category during traversal of the fragment library. Traversing by proportion allows for a quick understanding of the basic situation, such as whether there are any growth fragments that meet the expectations. If so, a complete traversal is performed. If not, a complete traversal can be avoided by changing the starting fragment or changing the matching position of the starting fragment and the protein molecule, thereby reducing unnecessary calculations and computational resources.

[0200] It should be noted that, considering the large number of fragments starting from C and N, skeleton clustering (only for fragments starting from C and N) was ultimately used to divide the fragments. In actual use, each skeleton category is randomly sampled according to a sampling rate (e.g., 10%). This method can reduce the number of fragments (reduce the amount of computation) without missing any skeleton type.

[0201] In some embodiments, after generating candidate molecules, the above method may further include the following operation: conformational optimization of the candidate molecules. For example, conformational optimization can reduce the potential energy of the molecular conformation, making the molecular conformation more stable.

[0202] In this embodiment, for the same growth fragment, during library construction, it can be saved according to different conformations, or only the most stable conformation can be saved, or multiple conformations can be saved together for subsequent growth and connection; for the same growth fragment, different conformations may be connected to the connection point of the starting fragment for growth, rotation, and screening.

[0203] In this embodiment, the growth fragments used for growth are at least partially obtained by cleaving real drug-like molecules according to specified rules, and the types of neighboring atoms at the cleavage sites are preserved during cleavage. This fragment preparation method ensures the syntheticability of the generated molecules and avoids the presence of features unsuitable for drug development.

[0204] Figure 15 A logic diagram of a molecular generation method according to an embodiment of this application is illustrated schematically.

[0205] See Figure 15 In one specific embodiment, the user provides a protein molecule (PDB format file) and a starter fragment (PDB format file) at a matching location within the protein molecule as input. The approximate matching location information of the protein pocket is determined based on the position of the starter fragment; the growth connection site is determined based on the position of the dummy atoms in the starter fragment; the growth direction is determined based on the bonding direction of the dummy atoms (the direction in which adjacent real atoms point to the dummy atoms); and the starter type of the next fragment is determined based on the type of neighboring atoms of the dummy atoms.

[0206] Starting with virtual atoms and using the growth direction as the axis, construct a base with a radius of... Gao Wei The cylindrical space is designated as the specific space. This specific space serves as the growth space for the next growth segment. The construction process of the specific space, the radiation vector, and the sampling points determined based on the radiation vector are as described above and will not be detailed here.

[0207] After determining the cylindrical space and the radiation vector, the starting atom type of the growth segment needs to be determined based on the connection atom type of the starting segment. The starting atom type of the segments traversed in this iteration is the same.

[0208] The various growth segment conformations of this growth segment are scored individually, and the conformation with the highest score is retained. If all scores for this segment are negative, then this segment is not selected.

[0209] After traversing the fragment library, the top ten highest-scoring growth fragments are merged with the starting fragment to serve as the starting fragment for the next traversal (growing the next fragment). The new starting fragment needs to determine its new growth direction based on the remaining connection points. After determining the growth direction, the cylindrical space and radiation vector are determined using the same method as before, and the fragment library is traversed. The scores mentioned above represent the scores of the growth fragments at that angle (calculated based on the aforementioned scoring function); the top ten scores represent the top ten after the angle is determined. If 36 angles are rotated, there are 36 scores; the highest score is taken, and the angle corresponding to the highest score is the determined angle (different growth fragments may have their own determined angles, but they are only compared through scores). Besides selecting the top ten highest-scoring fragments, the selection can also be topM or Top M%, where M is a positive integer.

[0210] The growth-stopping condition is either completing N traversals (growing N fragments) or being unable to grow (no fragments are retained after traversal). Specifically, the growth-stopping condition can be that there are no fragments available for the next traversal, or that five layers have been reached (i.e., five consecutive fragments have been grown). These conditions are particularly suitable for small protein pockets (i.e., situations where the protein pocket space is small). Furthermore, the growth-stopping condition can also include the following: for example, the growth range exceeds a preset range (spatial range, such as a cube or sphere), which is in the same coordinate system as the protein and the starting fragment. For example, the molecular weight is less than 500. For example, the number of atoms is less than 100. The relationship between multiple conditions can be such that satisfying any one of these conditions satisfies the growth-stopping condition.

[0211] When molecular growth is complete, the generated molecular conformation can be further optimized, such as by using Smina software to optimize the generated molecular conformation and compensate for the error in cylindrical space detection.

[0212] This application also provides a design method. Figure 16 A flowchart illustrating a design method according to an embodiment of this application is shown schematically.

[0213] See Figure 16 The above design 1600 may include operations S1610 to S1620.

[0214] In operation S1610, the molecular structure is generated according to the above-described molecular generation method. For details, please refer to the relevant embodiments described above; further elaboration will not be repeated here.

[0215] In operation S1620, candidate drug design or material design is performed based on molecular structure. Candidate drugs include both potentially marketable drugs and already marketed drugs.

[0216] This embodiment uses molecular structures obtained through generative molecular methods for design and development, such as for pharmaceutical development, which can effectively improve design efficiency and success rate.

[0217] This application also provides a molecular generation apparatus.

[0218] Figure 17 A block diagram of a molecular generation apparatus according to an embodiment of this application is shown schematically.

[0219] See Figure 17 The molecule generation device 1700 may include: a conformation acquisition module 1710, a first growth module 1720, a growth fragment conformation acquisition module 1730, a second growth module 1740, a starting fragment conformation update module 1750, and a candidate molecule acquisition module 1760.

[0220] The conformation acquisition module is used to obtain a protein molecule including a protein pocket conformation and at least one starter fragment conformation that matches the protein pocket conformation. The protein pocket conformation includes at least one matching position, and each of the at least one starter fragment conformation includes an initiation growth site.

[0221] The first growth module is used to set the current starter fragment conformation in at least one starter fragment conformation to the current matching position of the protein pocket conformation, and to determine the growth direction based on the starting growth site of the current starter fragment conformation.

[0222] The growth fragment conformation acquisition module is used to obtain a growth fragment conformation, which includes a connection site and a first growth site. Each growth fragment conformation is located inside a different specific space, which includes some atoms in the protein molecule and / or some atoms in the current starting fragment conformation in the cycle in which each growth fragment is located. The current starting fragment conformation, the growth fragment conformation, the starting fragment conformation after growth, and the starting fragment conformation after multiple growths are all separated from the protein molecule.

[0223] The second growth module is used to connect the starting growth site and the connection site in the growth direction to obtain the conformation of the starting fragment after growth.

[0224] The starting fragment conformation update module is used to take the grown starting fragment conformation as the current starting fragment conformation, the first growth site as the starting growth site, and the growth direction of the grown starting fragment conformation as the starting growth direction.

[0225] If the candidate molecule acquisition module meets the growth cessation conditions, it obtains the starting fragment conformation after multiple growths, and the starting fragment after multiple growths is the candidate molecule. The growth cessation conditions include at least one of the following: the number of repetitions reaches a preset number, the molecular weight of the starting fragment after multiple growths is greater than or equal to a molecular weight threshold, and the number of atoms of the starting fragment after multiple growths is greater than or equal to an atomic number threshold.

[0226] In some embodiments, the growth fragment conformation acquisition module includes: a growth fragment conformation set acquisition unit and a growth fragment conformation acquisition unit.

[0227] Among them, the growth fragment conformation set acquisition unit is used to obtain a growth fragment conformation set that matches the starting fragment conformation from the fragment library;

[0228] The growth fragment conformation acquisition unit is used to obtain growth fragment conformations from the growth fragment conformation set according to preset standards. The preset standards include: random selection, matching degree from high to low, and matching degree from low to high.

[0229] In some embodiments, the growth fragment conformation set acquisition unit includes:

[0230] The filtering subunit is used to filter candidate growth fragments from the fragment library based on the attribute information of the starting fragment conformation. The attribute information includes at least one of the following: the splitting position information or the neighboring atom type.

[0231] The fragment rotation subunit is used to obtain the growth fragment conformation by rotating the candidate growth fragments, so as to construct the growth fragment conformation set.

[0232] In some embodiments, the above-described apparatus 1700 further includes a target growth fragment conformation determination module, configured to determine a target growth fragment conformation from a plurality of growth fragment conformations corresponding to the current candidate growth fragment, so as to construct a growth fragment conformation set based on the target growth fragment conformation, wherein the stability of the target growth fragment conformation is higher than that of the non-target growth fragment conformations.

[0233] In some embodiments, the growth fragment conformation set acquisition unit includes: a sampled fragment class generation subunit and a target fragment class traversal subunit.

[0234] The first sampling subunit is used to obtain candidate growth fragments of a first preset ratio from multiple fragment classes and generate a first sampling fragment class so as to determine the target fragment class that includes candidate growth fragments among the multiple fragment classes by traversing the first sampling fragment class.

[0235] The first traversal subunit is used to obtain at least one growing segment configuration that can be successively extended on the current starting segment configuration by traversing the target segment class, so as to generate a set of growing segment configurations.

[0236] In some embodiments, the growth fragment conformation acquisition unit includes: a second sampling subunit and a second traversal subunit and a growth fragment conformation determination subunit.

[0237] The second sampling subunit is used to obtain candidate growth fragments with a second preset ratio from the fragment class and generate a second sampling fragment class;

[0238] The second traversal subunit is used to traverse the second sampled fragment class until a third candidate growth fragment set matching the neighboring atom type of the connection site is obtained; each candidate growth fragment in the third candidate growth fragment set is rotated around the growth direction according to a preset step size to obtain multiple third candidate growth fragment conformations; the third candidate growth fragment conformations are connected to the starting growth site one by one according to the growth direction.

[0239] The growth fragment conformation determination subunit is used to traverse the fragment class to which the second sampled fragment class belongs if at least one of the multiple third candidate growth fragment conformations is located inside a specific space, to obtain at least one set of fourth candidate growth fragments that matches the neighboring atom types of the connection site; to rotate the fourth candidate growth fragments around the growth direction according to a preset step size to obtain multiple fourth candidate growth fragment conformations; and to take the conformation located inside the specific space among the multiple fourth candidate growth fragment conformations as the growth fragment conformation.

[0240] In some embodiments, the growth fragment conformation acquisition unit includes a specific space construction subunit and a conformation set traversal subunit.

[0241] A specific space construction subunit is used to construct a specific space based on the starting fragment conformation. The specific space takes the spatial position of the virtual atoms of the starting fragment conformation as the spatial origin, and the extension direction of the spatial origin of the specific space is determined according to the growth direction of the initial growth site of the starting fragment conformation.

[0242] The conformation set traversal sub-unit is used to traverse the growth fragment conformation set to obtain candidate growth fragment conformations that match the starting fragment conformation and are located within a specific space when connected to the starting growth site.

[0243] In some embodiments, the candidate growth segment configuration is a configuration obtained by rotating the candidate growth segment around the growth direction according to a preset step size, and the candidate growth segment configuration is located inside a specific space when connected to the current starting point segment configuration.

[0244] In some embodiments, the growth fragment conformation acquisition module includes: a positional relationship determination unit and a fragment conformation determination unit.

[0245] The positional relationship determination unit is used to obtain the positional relationships between atoms of candidate growth fragment conformations and atoms of protein molecules and origin fragment conformations through sampling.

[0246] The fragment conformation determination unit is used to determine the growth fragment conformation from multiple candidate growth fragment conformations based on positional relationships.

[0247] In some embodiments, the positional relationship determination unit includes a subunit and a positional relationship characterization subunit.

[0248] The unit is used to obtain a score of the candidate growth fragment conformation relative to the protein molecule and the origin fragment conformation through sampling points in a specific space. The score is related to the collision point parameter and the neighboring point parameter, or the score is related to the collision point parameter, the neighboring point parameter, and at least one of the following: the collision point weight parameter and the neighboring point weight parameter.

[0249] The positional relationship characterization subunit is used to characterize the positional relationships between atoms of candidate growth fragment conformations and atoms of protein molecules and origin fragment conformations based on scoring results.

[0250] In some embodiments, the specific space includes a cylindrical space, with p sampling faces perpendicular to the growth direction. Each of the p sampling faces includes 360 / n sampling edges passing through the center of the circle, and each sampling edge includes m sampling points, where p, m, and n are integers greater than 2, and 360 is divisible by n.

[0251] In some embodiments, the above-described apparatus 1800 further includes a matching position update module.

[0252] The matching position update module is used to update the matching position if at least one matching position includes an untraversed matching position, and to set the start fragment conformation at the update matching position of the protein pocket conformation to obtain the update growth direction for the connection site. In the update growth direction, the start growth site is connected to the connection site to obtain the grown start fragment conformation.

[0253] In some embodiments, the above-described apparatus 1800 further includes a fragment library construction module.

[0254] The fragment library construction module is used to build a fragment library. The grown fragments in the fragment library include fragments obtained by cutting from multiple candidate molecular structures. The grown fragments have splitting position information and neighboring atom type information corresponding to the splitting position information.

[0255] In some embodiments, the above-described apparatus 1800 further includes an encoding compression module and a decoding module.

[0256] The encoding and compression module is used to generate a growth fragment conformation for each growth fragment in the fragment library, and to encode and compress the growth fragment conformation.

[0257] The decoding module is used to read the encoded and compressed growth fragments and decode them to obtain the growth fragments.

[0258] In some embodiments, the above-described apparatus 1800 further includes a conformation optimization module for conformation optimization of candidate molecules.

[0259] This application also provides a design device.

[0260] Figure 18 A block diagram of a design device according to an embodiment of this application is shown schematically.

[0261] See Figure 18 The prediction device 1800 may include a molecular structure generation module 1810 and a design module 1820.

[0262] The molecular structure generation module 1810 is used to generate a molecular structure according to the molecular generation method described above.

[0263] Design module 1820 is used for candidate drug design or material design based on molecular structure.

[0264] Regarding the devices 1700 and 1800 in the above embodiments, the specific ways in which each module and unit performs operations have been described in detail in the embodiments related to the method, and will not be elaborated further here.

[0265] Another aspect of this application provides an electronic device.

[0266] Figure 19 A block diagram of an electronic device according to an embodiment of this application is shown schematically.

[0267] See Figure 19 The electronic device 1900 includes a memory 1910 and a processor 1920.

[0268] The processor 1920 can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor.

[0269] Memory 1910 may include various types of storage units, such as system memory, read-only memory (ROM), and permanent storage devices. ROM may store static data or instructions required by processor 1920 or other modules of the computer. Permanent storage devices may be read-write storage devices. Permanent storage devices may be non-volatile storage devices that retain stored instructions and data even when the computer is powered off. In some embodiments, permanent storage devices use mass storage devices (e.g., magnetic or optical disks, flash memory) as permanent storage devices. In other embodiments, permanent storage devices may be removable storage devices (e.g., floppy disks, optical drives). System memory may be a read-write storage device or a volatile read-write storage device, such as dynamic random access memory. System memory may store some or all of the instructions and data required by the processor during operation. Furthermore, memory 1910 may include any combination of computer-readable storage media, including various types of semiconductor memory chips (e.g., DRAM, SRAM, SDRAM, flash memory, programmable read-only memory), and disks and / or optical disks may also be used. In some embodiments, the memory 1910 may include a removable storage device that is readable and / or writable, such as a laser disc (CD), a read-only digital versatile optical disc (e.g., DVD-ROM, dual-layer DVD-ROM), a read-only Blu-ray disc, an ultra-high density optical disc, a flash memory card (e.g., SD card, mini SD card, Micro-SD card, etc.), a magnetic floppy disk, etc. Computer-readable storage media do not contain carrier waves or transient electronic signals transmitted wirelessly or via wired connections.

[0270] The memory 1910 stores executable code, which, when processed by the processor 1920, can cause the processor 1920 to execute some or all of the methods described above.

[0271] Furthermore, the method according to this application can also be implemented as a computer program or computer program product, which includes computer program code instructions for performing some or all of the steps in the method described above.

[0272] Alternatively, this application may be implemented as a computer-readable storage medium (or a non-transitory machine-readable storage medium or a machine-readable storage medium) storing executable code (or computer program or computer instruction code) thereon, which, when executed by a processor of an electronic device (or server, etc.), causes the processor to perform part or all of the steps of the methods described above according to this application.

[0273] The various embodiments of this application have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or improvement of the technology in the market, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. A method for generating molecules, characterized in that, include: S1. Obtain a protein molecule including a protein pocket conformation and at least one starter fragment conformation matching the protein pocket conformation, wherein the protein pocket conformation includes at least one matching position and each of the at least one starter fragment conformation includes an initiation growth site. S2. Set the current start-up fragment conformation in the at least one start-up fragment conformation to the current matching position of the protein pocket conformation, and determine the growth direction based on the starting growth site of the current start-up fragment conformation; S3. Obtain a growth fragment conformation, wherein the growth fragment conformation includes a connection site and a first growth site; S4. In the growth direction, connect the starting growth site to the connection site to obtain the growing starting segment conformation; S5. The grown starting segment conformation is taken as the current starting segment conformation, the first growth site is taken as the starting growth site, and the growth direction of the grown starting segment conformation is taken as the starting growth direction. S6. Repeat steps S3-S5 once or more until the growth stop condition is met to obtain the starting fragment conformation after multiple growths. The starting fragment after multiple growths is the candidate molecule. Each of the growth fragment conformations is located within a different specific space, which includes some atoms in the protein molecule and / or some atoms in the current starting fragment conformation in the cycle in which each growth fragment is located; the current starting fragment conformation, the growth fragment conformation, the starting fragment conformation after growth, and the starting fragment conformation after multiple growths are all separated from the protein molecule. The conditions for stopping growth include at least one of the following: the number of repetitions reaches a preset number, the molecular weight of the starting fragment after multiple growths is greater than or equal to a molecular weight threshold, and the number of atoms in the starting fragment after multiple growths is greater than or equal to an atomic number threshold.

2. The molecular generation method according to claim 1, characterized in that, The conformation of the growth segment was obtained in the following manner: Obtain a set of growth fragment conformations that match the starting fragment conformation from the fragment library, and obtain growth fragment conformations from the set of growth fragment conformations according to preset standards. The preset standards include any one of the following: random selection, matching degree from high to low, or matching degree from low to high.

3. The molecular generation method according to claim 2, characterized in that, Obtaining a set of growth fragment conformations adapted to the starting fragment conformation from the fragment library includes: Candidate growth fragments are selected from the fragment library based on the attribute information of the starting fragment conformation, wherein the attribute information includes at least one of the following: splitting position information or neighboring atom type; By rotating the candidate growth fragments, growth fragment conformations are obtained to construct the growth fragment conformation set.

4. The molecular generation method according to claim 3, characterized in that, After obtaining the conformation of the growth segment, the method further includes: A target growth fragment conformation is determined from multiple growth fragment conformations corresponding to the current candidate growth fragment, and the growth fragment conformation set is constructed based on the target growth fragment conformation, wherein the stability of the target growth fragment conformation is higher than that of the non-target growth fragment conformations.

5. The molecular generation method according to claim 2, characterized in that, The fragment library includes multiple fragment classes, and each fragment class includes at least one candidate growth fragment; Obtaining a set of growth fragment conformations adapted to the starting fragment conformation from the fragment library includes: Candidate growth segments with a first preset ratio are obtained from multiple segment classes to generate a first sampled segment class, so as to determine the target segment class that includes the candidate growth segments among the multiple segment classes by traversing the first sampled segment class; By traversing the target fragment class, at least one growing fragment configuration that can be successively extended and grown on the current starting fragment configuration is obtained, so as to generate the set of growing fragment configurations.

6. The molecular generation method according to claim 2, characterized in that, The fragment library includes multiple fragment classes, and each fragment class includes at least one candidate growth fragment; Obtaining the set of growth fragment conformations adapted to the starting fragment conformation from the fragment library includes: for each of the plurality of fragment classes, From the fragment class, a second preset ratio of candidate growth fragments is obtained to generate a second sampling fragment class; Traverse the second sampled fragment class until a third candidate growth fragment set matching the neighboring atom type of the connection site is obtained; rotate each candidate growth fragment in the third candidate growth fragment set around the growth direction according to a preset step size to obtain multiple third candidate growth fragment conformations; connect the third candidate growth fragment conformations one by one to the starting growth site according to the growth direction; If at least one of the plurality of third candidate growth fragment conformations is located inside the specific space, then the fragment class to which the second sampled fragment class belongs is traversed to obtain at least one fourth candidate growth fragment set that matches the neighboring atom type of the connection site; the fourth candidate growth fragments are rotated around the growth direction according to a preset step size to obtain a plurality of fourth candidate growth fragment conformations; the conformation located inside the specific space among the plurality of fourth candidate growth fragment conformations is taken as the growth fragment conformation.

7. The molecular generation method according to claim 2, characterized in that, Obtaining growth fragment conformations from the growth fragment conformation set according to a preset standard includes: A specific space is constructed based on the starting fragment conformation, wherein the spatial origin of the specific space is the spatial position of the virtual atoms of the starting fragment conformation, and the extension direction of the spatial origin of the specific space is determined according to the growth direction of the initial growth site of the starting fragment conformation. Traverse the set of growth fragment conformations to obtain candidate growth fragment conformations that match the starting fragment conformation and are located within the specific space when connected to the starting growth site.

8. The molecular generation method according to claim 7, characterized in that, The candidate growth segment configuration is a configuration obtained by rotating the candidate growth segment around the growth direction according to a preset step size, and the candidate growth segment configuration is located inside the specific space when connected to the current starting point segment configuration.

9. The molecular generation method according to claim 1, characterized in that, Obtaining a growth segment conformation includes: The positional relationships between atoms of the candidate growth fragment conformation and atoms of the protein molecule and the starting fragment conformation are obtained by sampling. The growth segment conformation is determined from a plurality of candidate growth segment conformations based on the positional relationship.

10. The molecular generation method according to claim 9, characterized in that, The method of obtaining the positional relationship between atoms of the candidate growth fragment conformation and atoms of the protein molecule and the origin fragment conformation through sampling includes: If the first exclusion condition is met, or the second exclusion condition is met, the positional relationship between the atoms of the candidate growth fragment conformation and the atoms of the protein molecule and the starting fragment conformation is obtained by sampling. The first exclusion condition includes: all atoms of the candidate growth fragment conformation are located inside the specific space; The second exclusion condition includes: all atoms of the candidate growth fragment conformation are located inside the specific space, and at least one of the following: the molecular weight of the molecular structure formed by the starting fragment and the candidate growth fragment is less than or equal to a molecular weight threshold, and the number of atoms in the molecular structure formed by the starting fragment and the candidate growth fragment is less than or equal to an atomic number threshold.

11. The molecular generation method according to claim 9, characterized in that, The method of obtaining the positional relationship between atoms of the candidate growth fragment conformation and atoms of the protein molecule and the origin fragment conformation through sampling includes: By sampling points in the specific space, a scoring result is obtained of the candidate growth fragment conformation relative to the protein molecule and the origin fragment conformation. The scoring result is related to the collision point parameter and the neighboring point parameter, or the scoring result is related to the collision point parameter, the neighboring point parameter, and at least one of the following: the collision point weight parameter and the neighboring point weight parameter. The scoring results characterize the positional relationship between the atoms of the candidate growth fragment conformation and the atoms of the protein molecule and the origin fragment conformation.

12. The molecular generation method according to claim 11, characterized in that, The step of obtaining a score for the candidate growth fragment conformation relative to the protein molecule and the origin fragment conformation through sampling points in the specific space includes: By sampling points in the specific space, the number of adjacent points N1 where the distance between the candidate growth fragment conformation and the protein pocket conformation is less than a preset distance threshold, the number of collision points N2 between the candidate growth fragment conformation and the protein pocket conformation, and the number of non-empty points N3 in the sampling points in the specific space are obtained. If the number of collision points N2 is greater than zero, the scoring result is negative. If the number of collision points N2 is equal to zero, the scoring result is a weighted sum of the number of adjacent points N1 and the number of non-empty points N3, wherein the first weight of the number of adjacent points N1 and the second weight of the number of non-empty points N3 have no common divisor other than 1.

13. The molecular generation method according to claim 1, characterized in that, The specific space includes a cylindrical space, and the p sampling surfaces of the cylindrical space are perpendicular to the growth direction. Each of the p sampling surfaces includes 360 / n sampling edges passing through the center of the circle. Each sampling edge includes m sampling points, where p, m, and n are integers greater than 2, and 360 is divisible by n.

14. The molecular generation method according to claim 13, characterized in that, The specific space is constructed in the following manner: A first vector is generated based on the growth direction of the virtual atoms in the starting fragment conformation; Starting from the virtual atoms of the aforementioned starting fragment conformation, 360 / n second vectors are generated in a direction perpendicular to the first vector; For each second vector, m sampling points are generated along the direction of the second vector, using the virtual atom as the base point; Along the direction of the first vector, all sampling points are translated p times with a preset step size to obtain the sampling points in the specific space.

15. The molecular generation method according to any one of claims 1 to 14, characterized in that, Also includes: If the at least one matching position includes an untraversed matching position, the untraversed matching position is used as the updated matching position, and the starting fragment conformation is set at the updated matching position of the protein pocket conformation to obtain an updated growth direction for the connection site, so that the starting growth site is connected to the connection site in the updated growth direction to obtain the grown starting fragment conformation.

16. The molecular generation method according to any one of claims 1 to 14, characterized in that, Also includes: S21. If the at least one starter fragment conformation includes an untraversed starter fragment conformation, then the current untraversed starter fragment conformation is set at the current matching position of the protein pocket conformation, and the growth direction is determined based on the starting growth site of the current untraversed starter fragment conformation. S22. Repeat steps S21, S3-S5 one or more times until the growth cessation condition is met, and obtain the starting fragment conformation after multiple growths. The starting fragment after multiple growths is the candidate molecule.

17. The molecular generation method according to any one of claims 2 to 12, characterized in that, Also includes: A fragment library is constructed, wherein the growth fragments in the fragment library include fragments obtained by cutting from multiple candidate molecular structures, and the growth fragments have cutting position information and neighboring atom type information corresponding to the cutting position information.

18. The molecular generation method according to claim 17, characterized in that, The constructed fragment library includes: Multiple candidate molecular structures are segmented based on at least one of rotatable bonds, molecular backbones, or functional groups to obtain multiple candidate growth fragments; The fragment library is constructed based on multiple candidate growth fragments.

19. The molecular generation method according to claim 18, characterized in that, After obtaining multiple candidate growth fragments, the method further includes: Based on the correspondence between the initial growth sites and atom types obtained from molecule segmentation, selected growth segments are determined from multiple candidate growth segments. The correspondence is determined based on the segmentation position information and the neighboring atom type information corresponding to the segmentation position information. The step of constructing the fragment library based on multiple candidate growth fragments includes: constructing the fragment library based on multiple selected growth fragments.

20. The molecular generation method according to claim 18, characterized in that, The specific method for constructing the fragment library based on multiple candidate growth fragments includes at least one of the following: Remove candidate growth fragments containing specific elements; Remove duplicate candidate growth segments after regularization; The type of the starting atom at the connection site of the candidate growth fragment is determined based on the molecular structure information before segmentation; The candidate growth fragments were filtered based on molecular weight, the number of hydrogen bond donors, and the number of hydrogen bond acceptors. The candidate growth fragments are filtered based on expert experience.

21. The molecular generation method according to claim 20, characterized in that, Also includes: For each growth fragment in the fragment library, a growth fragment conformation is generated for the growth fragment, and the growth fragment conformation is encoded and compressed. The encoded and compressed growth segment is read and decoded to obtain the growth segment.

22. The molecular generation method according to claim 18, characterized in that, The construction of the fragment library based on multiple candidate growth fragments includes: The candidate growth segments are clustered into multiple segments to obtain multiple segment classes. Add multiple fragment classes to the fragment library, or for each of the multiple fragment classes, extract a preset proportion or a preset number of fragments from the fragment class according to a preset sampling rate and add them to the fragment library.

23. The molecular generation method according to claim 18, characterized in that, After constructing the fragment library based on the multiple candidate growth fragments, the method further includes: The candidate growth segments are clustered into multiple segments to obtain multiple segment classes. Add multiple fragment classes to the fragment library, or for each of the multiple fragment classes, extract a preset proportion or a preset number of fragments from the fragment class according to a preset sampling rate and add them to the fragment library.

24. The molecular generation method according to any one of claims 1 to 14, characterized in that, After generating the candidate molecule, the method further includes: The candidate molecules were conformationally optimized.

25. A design method, characterized in that, The method includes: The molecular structure is generated according to any one of claims 1 to 24; Candidate drug design or material design can be carried out based on the molecular structure.

26. A molecular generating apparatus, characterized in that, include: An information acquisition module is used to acquire a protein molecule including a protein pocket conformation and at least one starter fragment conformation matching the protein pocket conformation, wherein the protein pocket conformation includes at least one matching position and each of the at least one starter fragment conformation includes an initiation growth site. A first growth module is configured to set the current start-up fragment conformation in the at least one start-up fragment conformation to the current matching position of the protein pocket conformation, and to determine the growth direction based on the start-up growth site of the current start-up fragment conformation. A fragment conformation acquisition module is used to obtain a growth fragment conformation, wherein the growth fragment conformation includes a connection site and a first growth site, wherein each growth fragment conformation is located inside a different specific space, the specific space including some atoms in the protein molecule and / or some atoms in the current starting fragment conformation in the cycle in which each growth fragment is located; the current starting fragment conformation, the growth fragment conformation, the grown starting fragment conformation, and the starting fragment conformation after multiple growths are all separated from the protein molecule; The second growth module is used to connect the starting growth site and the connection site in the growth direction to obtain the growing starting fragment conformation. The fragment conformation update module is used to take the grown starting fragment conformation as the current starting fragment conformation, the first growth site as the starting growth site, and the growth direction of the grown starting fragment conformation as the starting growth direction. An iterative module is used to repeat the growth process until the growth cessation condition is met, and obtain the conformation of the starting fragment after multiple growths. The starting fragment after multiple growths is the candidate molecule. The growth cessation condition includes at least one of the following: the number of repetitions reaches a preset number, the molecular weight of the starting fragment after multiple growths is greater than or equal to a molecular weight threshold, and the number of atoms of the starting fragment after multiple growths is greater than or equal to an atomic number threshold.

27. A design device, characterized in that, include: A molecular generation module, used to generate molecular structures using the molecular generation apparatus according to claim 26; The design module is used for drug design or material design based on the molecular structure.

28. An electronic device, characterized in that, include: processor; as well as A memory having executable code stored thereon, which, when executed by the processor, causes the processor to perform the method according to any one of claims 1-25.

29. A computer-readable storage medium, characterized in that, It stores executable code that, when executed by a processor of an electronic device, causes the processor to perform the method according to any one of claims 1-25.

Images