Protein structure model construction method, device, storage medium and electronic device

Abstract

The present application provides a method, device, storage medium and electronic device for constructing a protein structure model, and relates to the field of computer technology. The protein structure model construction method provided in the embodiment of the present application can construct an initial structure model of the protein to be processed based on the candidate structure template corresponding to the original amino acid sequence of the protein to be processed, and for each first amino acid pair in the original amino acid sequence, respectively determine the actual distance between the two amino acids in each first amino acid pair, and then adjust the initial structure model based on the actual distance between the two amino acids in each first amino acid pair to obtain a target structure model of the protein to be processed. Compared with the related art, the accuracy of the protein structure can be improved to obtain a higher-precision protein structure.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of computer technology, and in particular to a method, device, storage medium, and electronic device for constructing a protein structure model. Background Art

[0002] The prediction and optimization of protein structure have certain application value in medical health and bioengineering. For example, in the development of new drugs, it is not only necessary to understand protein structure information, but also to continuously optimize the protein structure. Enzyme engineering also needs to design catalytic active centers based on protein structure, and so on.

[0003] Currently, protein structure can be determined through experimental methods such as X-ray crystallization, nuclear magnetic resonance, and cryo-electron microscopy, but this method requires a lot of time and high cost, and the accuracy of the protein structure obtained after prediction and optimization is not high. Summary of the Invention

[0004] To solve the technical problems existing in the related art, the embodiments of the present application provide a method, device, storage medium and electronic device for constructing a protein structure model, which can improve the accuracy of protein structure.

[0005] To achieve the above objectives, the technical solution of the embodiment of the present application is implemented as follows:

[0006] In a first aspect, the present invention provides a method for constructing a protein structure model, the method comprising:

[0007] constructing an initial structural model of the protein to be processed based on a candidate structural template corresponding to the original amino acid sequence of the protein to be processed;

[0008] For each first amino acid pair in the original amino acid sequence, determining the actual distance between the two amino acids in each first amino acid pair;

[0009] Based on the actual distance between the two amino acids in each first amino acid pair, the initial structural model is adjusted to obtain a target structural model of the protein to be processed.

[0010] In an optional embodiment, before constructing the initial structural model of the protein to be processed based on the candidate structural template corresponding to the original amino acid sequence of the protein to be processed, the method further includes:

[0011] Based on the correspondence between the preset amino acid sequence and the structural template, searching for a candidate amino acid sequence that matches the original amino acid sequence in a protein structure database;

[0012] The structural template corresponding to the candidate amino acid sequence is used as the candidate structural template corresponding to the original amino acid sequence.

[0013] In an optional embodiment, searching a protein structure database for a candidate amino acid sequence that matches the original amino acid sequence based on a preset correspondence between the amino acid sequence and the structural template includes:

[0014] Performing sequence alignment on each amino acid sequence in the protein structure database and the original amino acid sequence to determine the similarity between each amino acid sequence and the original amino acid sequence;

[0015] The amino acid sequence with the greatest similarity is taken as a candidate amino acid sequence that matches the original amino acid sequence.

[0016] In an optional embodiment, for each first amino acid pair in the original amino acid sequence, determining the actual distance between the two amino acids in each first amino acid pair comprises:

[0017] determining at least one reference amino acid sequence whose similarity to the original amino acid sequence meets a set threshold;

[0018] generating a feature matrix according to the structural features of each of the reference amino acid sequences;

[0019] The feature matrix is ​​input into a distance prediction network, and the actual distance between the two amino acids in each first amino acid pair is determined based on the distance prediction network.

[0020] In an optional embodiment, adjusting the initial structural model based on the actual distance between the two amino acids in each first amino acid pair to obtain the target structural model of the protein to be processed includes:

[0021] The following operations are performed in a cyclic iterative manner until the molecular energy of the initial structural model meets the preset cyclic iterative conditions, and the target structural model of the protein to be processed is determined based on the initial structural model output in the final round:

[0022] Comparing each second amino acid pair in the initial structural model with each first amino acid pair in the original amino acid sequence to determine a missing portion of the initial structural model relative to the original amino acid sequence;

[0023] For each second amino acid pair in the initial structure model, respectively determine the template distance between the two amino acids in each second amino acid pair, and determine the difference between each template distance and each corresponding actual distance;

[0024] Based on the difference between each template distance and each corresponding actual distance, and the actual distance of the first amino acid pair included in the missing part, the molecular energy of the initial structural model is determined, and the initial structural model is adjusted according to the molecular energy.

[0025] In an optional embodiment, determining the difference between each template distance and each corresponding actual distance includes:

[0026] The distance matrix composed of the template distances and the distance matrix composed of the actual distances are both input into the difference prediction network, and the difference value between each template distance and the corresponding actual distance is determined based on the difference prediction network.

[0027] In an optional embodiment, determining the molecular energy of the initial structural model based on the difference between each template distance and each corresponding actual distance, and the actual distance of the first amino acid pair of the missing portion, includes:

[0028] determining a distance constraint parameter for the energy of each second amino acid pair based on the difference between each template distance and each corresponding actual distance;

[0029] determining the distance energy of each second amino acid pair according to the distance constraint parameter of the energy of each second amino acid pair and the template distance between the two amino acids in each second amino acid pair;

[0030] For the missing portion of the initial structure model relative to the original amino acid sequence, determining the distance energy of each first amino acid pair according to the actual distance between two amino acids in each first amino acid pair contained in the missing portion;

[0031] The molecular energy of the initial structural model is determined based on the distance energy of each second amino acid pair and the distance energy of each first amino acid pair included in the missing portion.

[0032] In an optional embodiment, determining the target structural model of the protein to be processed based on the initial structural model output in the final round includes:

[0033] The initial structural model outputted in the final round is used as the first structural model;

[0034] Randomly changing the position of any amino acid in the first structural model to obtain a second structural model;

[0035] If the molecular energy of the second structural model is less than the molecular energy of the first structural model, the second structural model is used as the target structural model of the protein to be processed.

[0036] In an optional embodiment, determining the target structural model of the protein to be processed based on the initial structural model output in the final round includes:

[0037] The initial structural model outputted in the final round is used as the first structural model;

[0038] Adjusting the position of any amino acid in the abnormal amino acid pair in the first structural model to obtain a third structural model; the abnormal amino acid pair is an amino acid pair whose template distance does not meet the set distance range;

[0039] If the molecular energy of the third structural model is less than the molecular energy of the first structural model, the third structural model is used as the target structural model of the protein to be processed.

[0040] In a second aspect, an embodiment of the present application further provides a device for constructing a protein structure model, the device comprising:

[0041] An initial model building unit, configured to build an initial structural model of the protein to be processed based on a candidate structural template corresponding to the original amino acid sequence of the protein to be processed;

[0042] an actual distance determining unit, configured to determine, for each first amino acid pair in the original amino acid sequence, an actual distance between two amino acids in each first amino acid pair;

[0043] The target model determining unit is used to adjust the initial structure model based on the actual distance between the two amino acids in each first amino acid pair to obtain the target structure model of the protein to be processed.

[0044] In an optional embodiment, the apparatus further includes a candidate template determining unit configured to:

[0045] Based on the correspondence between the preset amino acid sequence and the structural template, searching for a candidate amino acid sequence that matches the original amino acid sequence in a protein structure database;

[0046] The structural template corresponding to the candidate amino acid sequence is used as the candidate structural template corresponding to the original amino acid sequence.

[0047] In an optional embodiment, the candidate template determining unit is specifically configured to:

[0048] Performing sequence alignment on each amino acid sequence in the protein structure database and the original amino acid sequence to determine the similarity between each amino acid sequence and the original amino acid sequence;

[0049] The amino acid sequence with the greatest similarity is taken as a candidate amino acid sequence that matches the original amino acid sequence.

[0050] In an optional embodiment, the actual distance determination unit is specifically configured to:

[0051] determining at least one reference amino acid sequence whose similarity to the original amino acid sequence meets a set threshold;

[0052] generating a feature matrix according to the structural features of each of the reference amino acid sequences;

[0053] The feature matrix is ​​input into a distance prediction network, and the actual distance between the two amino acids in each first amino acid pair is determined based on the distance prediction network.

[0054] In an optional embodiment, the target model determination unit is specifically configured to:

[0055] The following operations are performed in a cyclic iterative manner until the molecular energy of the initial structural model meets the preset cyclic iterative conditions, and the target structural model of the protein to be processed is determined based on the initial structural model output in the final round:

[0056] Comparing each second amino acid pair in the initial structural model with each first amino acid pair in the original amino acid sequence to determine a missing portion of the initial structural model relative to the original amino acid sequence;

[0057] For each second amino acid pair in the initial structure model, respectively determine the template distance between the two amino acids in each second amino acid pair, and determine the difference between each template distance and each corresponding actual distance;

[0058] Based on the difference between each template distance and each corresponding actual distance, and the actual distance of the first amino acid pair included in the missing part, the molecular energy of the initial structural model is determined, and the initial structural model is adjusted according to the molecular energy.

[0059] In an optional embodiment, the target model determination unit is further configured to:

[0060] The distance matrix composed of the template distances and the distance matrix composed of the actual distances are both input into the difference prediction network, and the difference value between each template distance and the corresponding actual distance is determined based on the difference prediction network.

[0061] In an optional embodiment, the target model determination unit is further configured to:

[0062] determining a distance constraint parameter for the energy of each second amino acid pair based on the difference between each template distance and each corresponding actual distance;

[0063] determining the distance energy of each second amino acid pair according to the distance constraint parameter of the energy of each second amino acid pair and the template distance between the two amino acids in each second amino acid pair;

[0064] For the missing portion of the initial structure model relative to the original amino acid sequence, determining the distance energy of each first amino acid pair according to the actual distance between two amino acids in each first amino acid pair contained in the missing portion;

[0065] The molecular energy of the initial structural model is determined based on the distance energy of each second amino acid pair and the distance energy of each first amino acid pair included in the missing portion.

[0066] In an optional embodiment, the target model determination unit is further configured to:

[0067] The initial structural model outputted in the final round is used as the first structural model;

[0068] Randomly changing the position of any amino acid in the first structural model to obtain a second structural model;

[0069] If the molecular energy of the second structural model is less than the molecular energy of the first structural model, the second structural model is used as the target structural model of the protein to be processed.

[0070] In an optional embodiment, the target model determination unit is further configured to:

[0071] The initial structural model outputted in the final round is used as the first structural model;

[0072] Adjusting the position of any amino acid in the abnormal amino acid pair in the first structural model to obtain a third structural model; the abnormal amino acid pair is an amino acid pair whose template distance does not meet the set distance range;

[0073] If the molecular energy of the third structural model is less than the molecular energy of the first structural model, the third structural model is used as the target structural model of the protein to be processed.

[0074] In a third aspect, an embodiment of the present application further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the protein structure model construction method of the first aspect is implemented.

[0075] In a fourth aspect, an embodiment of the present application further provides an electronic device comprising a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and when the computer program is executed by the processor, the processor implements the protein structure model construction method of the first aspect.

[0076] The protein structure model construction method, apparatus, storage medium, and electronic device provided in the embodiments of the present application can construct an initial structure model of the protein to be processed based on a candidate structure template corresponding to the original amino acid sequence of the protein to be processed. For each first amino acid pair within the original amino acid sequence, the actual distance between the two amino acids in each first amino acid pair is determined. Based on the actual distance between the two amino acids in each first amino acid pair, the initial structure model is adjusted to obtain a target structure model of the protein to be processed. Compared with related technologies, the accuracy of the protein structure can be improved to obtain a higher-precision protein structure. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0078] Figure 1 A diagram illustrating an application scenario of a method for constructing a protein structure model provided in an embodiment of the present application;

[0079] Figure 2 A schematic diagram of a process for constructing a protein structure model provided in an embodiment of the present application;

[0080] Figure 3 A schematic diagram of another method for constructing a protein structure model provided in an embodiment of the present application;

[0081] Figure 4 A schematic diagram of the structure of a protein structure model provided in an embodiment of the present application;

[0082] Figure 5 A schematic diagram of a process for constructing another protein structure model provided in an embodiment of the present application;

[0083] Figure 6 A schematic diagram of a process for constructing another protein structure model provided in an embodiment of the present application;

[0084] Figure 7 A schematic diagram of a process for constructing another protein structure model provided in an embodiment of the present application;

[0085] Figure 8 A structural block diagram of a protein structure model construction device provided in an embodiment of the present application;

[0086] Figure 9 A structural block diagram of another protein structure model construction device provided in an embodiment of the present application;

[0087] Figure 10 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0088] To make the objectives, technical solutions, and advantages of this application more clear, this application will be further described in detail below with reference to the accompanying drawings. Obviously, the embodiments described are only some of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making any creative efforts are within the scope of protection of this application.

[0089] It should be noted that the terms "including" and "having" and their variations involved in the documents of this application are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or are inherent to these processes, methods, products or devices.

[0090] The following explains some of the terms used in the embodiments of the present application to facilitate understanding by those skilled in the art.

[0091] (1) Template-Based Modeling (TBM): It is a traditional and effective protein structure prediction method, which includes two important steps: template search and template-based three-dimensional model construction. Template search is to use sequence matching algorithms to find the target protein structure template that is closest to the original amino acid sequence of the protein from the protein sequence database. Template-based three-dimensional model construction means that the distance between each amino acid pair in the target protein structure template can be obtained based on the target protein structure template, and used as a distance constraint to guide the construction of the three-dimensional model.

[0092] (2) Free Modeling (FM): Search for multiple target amino acid sequences similar to the original amino acid sequence of the protein from the sequence database of the protein structure, and obtain a feature matrix based on the multiple target amino acid sequences. The feature matrix is ​​input into the distance prediction network to obtain the distance between each amino acid pair in the original amino acid sequence of the protein.

[0093] (3) Structural optimization: The structure of a protein obtained after template modeling can be called decoy. Decoy is often not the target protein structure that is closest to the experimental structure of the protein. It is necessary to determine the molecular energy based on the distance between each amino acid pair in the target protein structure template obtained by template modeling and the distance between each amino acid pair in the original amino acid sequence obtained by free modeling, and optimize the target protein structure template based on the molecular energy to obtain the final target protein structure.

[0094] The word “exemplary” is used hereinafter to mean “serving as an example, example, or illustration.” Any embodiment described as “exemplary” is not necessarily to be construed as preferred or advantageous over other embodiments.

[0095] The terms "first" and "second" are used for descriptive purposes only and should not be construed as explicitly or implicitly indicating relative importance or the number of the technical features indicated. Therefore, features specified as "first" or "second" may explicitly or implicitly include one or more of such features. In the description of the embodiments of this application, unless otherwise specified, "plurality" means two or more.

[0096] The embodiments of the present application relate to artificial intelligence (AI) and machine learning technologies, and are designed based on computer vision (CV) technology, speech processing technology (Speech Technology) and machine learning (ML) in artificial intelligence.

[0097] Artificial intelligence (AI) refers to the theories, methods, techniques, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, to perceive the environment, acquire knowledge, and use that knowledge to achieve optimal results. In other words, AI is a comprehensive field of computer science that seeks to understand the essence of intelligence and produce new intelligent machines that can respond in a manner similar to human intelligence. AI also studies the design principles and implementation methods of various intelligent machines, enabling them to possess the capabilities of perception, reasoning, and decision-making. AI technologies primarily encompass computer vision, speech processing, and machine learning / deep learning.

[0098] With the research and advancement of artificial intelligence technology, artificial intelligence has been studied and applied in many fields, such as common smart homes, image retrieval, video surveillance, smart speakers, smart marketing, unmanned driving, automatic driving, drones, robots, smart medical care, etc. It is believed that with the development of technology, artificial intelligence will be applied in more fields and play an increasingly important role.

[0099] Machine learning is a multi-disciplinary interdisciplinary subject involving probability theory, statistics, approximation theory, convex analysis, algorithmic complexity theory and other disciplines. It specializes in studying how computers simulate or implement human learning behavior to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their own performance. Machine learning is the core of artificial intelligence and the fundamental way to make computers intelligent. Its applications are spread across all areas of artificial intelligence. Machine learning and deep learning generally include artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning and other technologies. In the embodiment of the present application, during the construction of the protein structure model, the actual distance between the two amino acids in each first amino acid pair is determined by a distance prediction network based on machine learning or deep learning, and the difference value between each template distance and the corresponding actual distance is determined by a difference prediction network based on machine learning or deep learning.

[0100] To better understand the technical solutions provided in the embodiments of this application, the following briefly introduces the application scenarios to which the technical solutions provided in the embodiments of this application are applicable. It should be noted that the application scenarios introduced below are only used to illustrate the embodiments of this application and are not limiting. In specific implementation, the technical solutions provided in the embodiments of this application can be flexibly applied according to actual needs.

[0101] The protein structure model construction method provided in the embodiments of the present application can be applied to Figure 1 See the application scenario shown. Figure 1 As shown, the application scenario includes multiple terminal devices 11 and a server 12. The terminal devices 11 and the server 12 can be connected and transmit data via a wired connection or a wireless connection. For example, the terminal devices 11 and the server 12 can be connected via a data cable or a wired network; the terminal devices 11 and the server 12 can also be connected via a radio frequency module, a WiFi module, or a wireless network.

[0102] Among them, the terminal device 11 can be a computer, a notebook, a personal digital assistant (PDA), a tablet computer, etc. The server 12 can be a server or a server cluster or cloud computing center composed of several servers, or a virtualization platform, or a personal computer, a large or medium-sized computer, or a computer cluster, etc. According to implementation needs, the application scenario in the embodiment of the present application can have any number of terminal devices and servers. This application does not make any special restrictions on this. The protein structure model construction method provided in the embodiment of the present application can be performed by the server 12, or it can be performed collaboratively by the terminal device 11 and the server 12.

[0103] For example, a medical research institution has a server 12 for constructing protein structure models. Each laboratory within the medical research institution is equipped with a terminal device 11. Researchers can use their laboratory terminal 11 to transmit the raw amino acid sequence of the protein to be processed to server 12. After receiving the raw amino acid sequence, server 12 can obtain the target structure model of the protein to be processed based on the raw amino acid sequence.

[0104] To further illustrate the technical solutions provided by the embodiments of the present application, this is described in detail below with reference to the accompanying drawings and specific implementation methods. Although the embodiments of the present application provide method operation steps as shown in the following embodiments or drawings, more or fewer operation steps may be included in the method based on routine or no creative labor. In steps where there is no necessary causal relationship logically, the execution order of these steps is not limited to the execution order provided by the embodiments of the present application. The method may be executed in the order of the methods shown in the embodiments or drawings or in parallel during the actual processing process or when the device is executed.

[0105] Figure 2 A flowchart of a protein structure model construction method provided in an embodiment of the present application is shown. The method can be Figure 1 The server 12 in the embodiment can also execute the process, which can also be executed by the terminal device 11 or other electronic devices. For example, the computer used for constructing the protein structure model is used as the execution subject to illustrate the specific implementation process of the protein structure model construction method of the embodiment of the present application. Figure 2 As shown, the protein structure model construction method includes the following steps:

[0106] Step S201 : constructing an initial structural model of the protein to be processed based on a candidate structural template corresponding to the original amino acid sequence of the protein to be processed.

[0107] The protein to be processed may be a protein of unknown structure. After obtaining the protein to be processed of unknown structure, it is necessary to obtain the amino acids corresponding to the protein through chemical experimental methods, so that the original amino acid sequence corresponding to the protein can be determined based on the amino acids corresponding to the protein.

[0108] For the original amino acid sequence of the protein to be processed, each amino acid sequence in the protein structure database can be aligned with the original amino acid sequence to determine the similarity between each amino acid sequence and the original amino acid sequence, and the amino acid sequence with the greatest similarity can be used as the candidate amino acid sequence that matches the original amino acid sequence.

[0109] For example, the Basic Local Alignment Search Tool (BLAST) can be used to align the original amino acid sequence D0 with each of the amino acid sequences D1, D2, and D3 in the protein structure database. If the similarity between the original amino acid sequence D0 and the amino acid sequence D1 is 85%, the similarity between the original amino acid sequence D0 and the amino acid sequence D2 is 95%, and the similarity between the original amino acid sequence D0 and the amino acid sequence D3 is 82%, then the amino acid sequence D2 can be used as a candidate amino acid sequence.

[0110] After finding the candidate amino acid sequence, the structural template corresponding to the candidate amino acid sequence can be used as the candidate structural template corresponding to the original amino acid sequence, and the initial structural model of the protein to be processed can be constructed based on the candidate structural template.

[0111] Step S202 : for each first amino acid pair in the original amino acid sequence, determining the actual distance between the two amino acids in each first amino acid pair.

[0112] We can first search the protein structure database for all reference amino acid sequences whose similarity with the original amino acid sequence meets a set threshold, then generate a feature matrix based on the structural features of each reference amino acid sequence, and input the feature matrix into the distance prediction network to determine the actual distance between the two amino acids in each first amino acid pair.

[0113] For example, if the length of the original amino acid sequence is L, 200 reference amino acid sequences of length L can be determined from a protein structure database. These 200 reference amino acid sequences can then be converted into an L×L×21×21 feature matrix, which is then input into a distance prediction network. This distance prediction network is a fully convolutional neural network (FCN). Because the FCN network structure consists of convolutional layers, the distance prediction network first performs multiple convolution operations on the feature matrix, and then uses a classification network to obtain an L×L×42 classification result. This L×L×42 classification result divides the distance between each amino acid pair into 42 distance intervals, each with a probability value. Based on this L×L×42 classification result, the actual distance between the two amino acids in each first amino acid pair of the original amino acid sequence can be determined.

[0114] Step S203 : adjusting the initial structural model based on the actual distance between the two amino acids in each first amino acid pair to obtain a target structural model of the protein to be processed.

[0115] After obtaining the initial structural model of the protein to be processed, each second amino acid pair within the initial structural model can be compared with each first amino acid pair within the original amino acid sequence to determine the missing portion of the initial structural model relative to the original amino acid sequence. The template distance between the two amino acids in each second amino acid pair within the initial structural model can be determined based on the obtained initial structural model. The distance matrix composed of each template distance and the distance matrix composed of each actual distance are input into the difference prediction network to obtain the difference value between each template distance and the corresponding actual distance. The difference prediction network can perform nonlinear difference calculation processing on the distance matrix composed of each template distance and the distance matrix composed of each actual distance to determine the difference value between each template distance and the corresponding actual distance.

[0116] Based on the difference between each template distance and the corresponding actual distance, the distance constraint parameter for the energy of each second amino acid pair can be first determined. Then, based on the distance constraint parameter for the energy of each second amino acid pair and the template distance between the two amino acids in each second amino acid pair, the distance energy of each second amino acid pair can be determined. For the missing portion of the initial structural model relative to the original amino acid sequence, the distance energy of each first amino acid pair can be determined based on the actual distance between the two amino acids in each first amino acid pair included in the missing portion. Finally, the molecular energy of the initial structural model can be determined based on the distance energy of each second amino acid pair and the distance energy of each first amino acid pair included in the missing portion.

[0117] The initial structural model can be iterated multiple times according to the determined molecular energy until the energy of each amino acid pair is minimized. The iterative process of the initial structural model can be ended and the target structural model of the protein to be processed can be output.

[0118] In one embodiment, after determining the target structural model of the protein to be processed, an all-atom structural jitter operation can be performed on the target structural model, that is, the target structural model can be used as the first structural model, and the position of any amino acid in the first structural model can be randomly changed to obtain a second structural model. If the molecular energy of the second structural model is less than the molecular energy of the first structural model, the second structural model is used as the target structural model of the protein to be processed.

[0119] In another embodiment, after determining the target structural model of the protein to be processed, a clash removal operation can be performed on the target structural model. That is, the target structural model can be used as the first structural model, and the position of any amino acid in the abnormal amino acid pair in the first structural model is adjusted to obtain a third structural model. The abnormal amino acid pair is an amino acid pair whose template distance does not meet a set distance range. If the molecular energy of the third structural model is less than the molecular energy of the first structural model, the third structural model is used as the target structural model of the protein to be processed.

[0120] In another embodiment, after determining the target structural model of the protein to be processed, a full-atom structural dithering operation can be performed on the target structural model, followed by a conflict resolution operation. Specifically, the target structural model can be used as the first structural model, and the position of any amino acid in the first structural model can be randomly changed to obtain a second structural model. If the molecular energy of the second structural model is less than that of the first structural model, the second structural model is used as the target structural model of the protein to be processed. The obtained target structural model is then used as the fourth structural model, and the position of any amino acid in the abnormal amino acid pair in the fourth structural model is adjusted to obtain a fifth structural model. If the molecular energy of the fifth structural model is less than that of the fourth structural model, the fifth structural model is used as the target structural model of the protein to be processed.

[0121] The protein structure model construction method provided in the embodiments of the present application can construct an initial structure model of the protein to be processed based on a candidate structure template corresponding to the original amino acid sequence of the protein to be processed. For each first amino acid pair in the original amino acid sequence, the actual distance between the two amino acids in each first amino acid pair is determined. Based on the actual distance between the two amino acids in each first amino acid pair, the initial structure model is adjusted to obtain a target structure model of the protein to be processed. Compared with related technologies, the accuracy of the protein structure can be improved, thereby obtaining a more precise protein structure.

[0122] In some embodiments, when determining the molecular energy of the initial structural model, in addition to obtaining the distance energy of each second amino acid pair and the distance energy of each first amino acid pair contained in the missing portion, it is also necessary to obtain the angle energy of each second amino acid pair and the angle energy of each first amino acid pair contained in the missing portion, as well as the molecular force field energy of the initial structural model. Among them, the process of obtaining the angle energy of each second amino acid pair and the angle energy of each first amino acid pair contained in the missing portion is similar to the process of obtaining the distance energy of each second amino acid pair and the distance energy of each first amino acid pair contained in the missing portion, and will not be repeated here. The molecular force field energy of the initial structural model is the energy generated by the interaction between atoms, which can be the energy generated by the bonding interaction between atoms constituting the same molecule, or the energy generated by the van der Waals interaction between different molecules. The molecular force field can generally include bond stretching energy, bond angle bending energy, dihedral angle distortion energy, non-bonded interaction and cross energy terms. Bond stretching energy refers to the energy change caused by the stretching and shrinking movement of each chemical bond constituting a molecule in the direction of the bond axis; bond angle bending energy refers to the molecular energy change caused by the change in bond angle; dihedral angle distortion energy refers to the energy change caused by the distortion of the molecular skeleton caused by the rotation of a single bond; non-bonded interaction refers to energy-related non-bonded interactions including van der Waals force and electrostatic interaction; cross energy term refers to the energy change caused by the coupling between various molecular interactions.

[0123] In one embodiment, in step S202 above, before the distance prediction network determines the actual distance between the two amino acids in each amino acid pair within the original amino acid sequence, the distance prediction network needs to be trained. A set of original amino acid sequences can be obtained by removing completely identical and highly similar amino acid sequences from the amino acid sequences of all known protein structures. Then, all reference amino acid sequences whose similarity to each original amino acid sequence in the set of original amino acid sequences meets a set threshold are searched from a protein structure database. A feature matrix is ​​generated based on the structural features of the reference amino acid sequence corresponding to each original amino acid sequence. The feature matrix corresponding to each original amino acid sequence serves as a training data set for the distance prediction network. After obtaining the training data set, each feature matrix can be labeled based on the distance between the two amino acids in each amino acid pair within each original amino acid sequence, and the distance prediction network can be trained based on the labeled feature matrix. For example, for an original amino acid sequence D0 of length L in the original amino acid sequence set, 100 reference amino acid sequences of length L can be determined from a protein structure database, and the 100 reference amino acid sequences can be converted into a feature matrix T0 of L×L×21×21. Based on the distance between the two amino acids in each amino acid pair in the original amino acid sequence D0, the feature matrix T0 can be annotated, that is, the distance between the amino acid pairs corresponding to the amino acid pairs in the original amino acid sequence D0 in the feature matrix T0 is annotated as 1, and the distance between other amino acid pairs is annotated as 0. After the feature matrix is ​​annotated, the distance prediction network can be trained based on the annotated feature matrix.

[0124] When training a distance prediction network, a feature matrix can be extracted from the training dataset as training sample data. This extracted feature matrix is ​​then input into the distance prediction network to be trained. The actual distance between the two amino acids in each amino acid pair within the original amino acid sequence corresponding to the feature matrix is ​​then determined. The actual distance between the two amino acids in each amino acid pair is then compared with the annotated distance in the feature matrix to determine the loss value.

[0125] When calculating the loss value, a preset loss function can be used. The loss function can be a cross-entropy loss function, such as a sigmoid function. Other loss functions used may include, but are not limited to, a multi-class cross-entropy loss function, a contrastive loss function related to metric learning, or a triplet loss function. Generally, the loss value determines how close the actual output is to the expected output. The smaller the loss value, the closer the actual output is to the expected output.

[0126] Determine whether the loss value has converged to the preset target value. If the loss value is less than or equal to the preset target value, or if the change in the loss value obtained after N consecutive training cycles is less than or equal to the preset target value, the loss value is considered to have converged to the preset target value, indicating that the loss value has converged; otherwise, it indicates that the loss value has not converged. If the loss value has not converged, the network parameters of the distance prediction network are adjusted. After adjusting the network parameters, the feature matrix from the training dataset is extracted again as training sample data and the next round of training continues. If the loss value converges, the current distance prediction network is used as the trained distance prediction network, thus completing the training of the distance prediction network.

[0127] In step S203, before the difference prediction network determines the difference between each template distance and the corresponding actual distance, the difference prediction network must be trained. The training process for the difference prediction network is similar to that for the distance prediction network and will not be described in detail here. The only difference is that the training data set obtained is different from that for the distance prediction network. The difference prediction network has two training data sets: one training data set is the template distance between the two amino acids in each second amino acid pair within the initial structural model, and the other training data set is the actual distance between the two amino acids in each first amino acid pair within the original amino acid sequence. When training the difference prediction network, the template distance or actual distance between the two amino acids in the corresponding amino acid pairs in the two training data sets can be extracted respectively.

[0128] Figure 3 The detailed implementation process of the protein structure model construction method proposed in this application is shown as follows: Figure 3 As shown, the following steps may be included:

[0129] Step S301 : constructing an initial structural model of the protein to be processed based on the candidate structural template corresponding to the original amino acid sequence of the protein to be processed.

[0130] After obtaining the original amino acid sequence of the protein to be processed, it is necessary to first determine the candidate structural template corresponding to the original amino acid sequence, and then construct the initial structural model of the protein to be processed based on the determined candidate structural template. The obtained initial structural model can be as follows: Figure 4 shown.

[0131] Step S302 : determining the actual distance between the two amino acids in each first amino acid pair in the original amino acid sequence.

[0132] At least one reference amino acid sequence whose similarity to the original amino acid sequence meets a set threshold can be determined, and a feature matrix can be generated based on the structural features of each reference amino acid sequence. The feature matrix is ​​then input into a trained distance prediction network, and the actual distance between the two amino acids in each first amino acid pair is determined based on the trained distance prediction network. The determined actual distance can be as follows Figure 4 shown.

[0133] For example, the original amino acid sequence D0 contains two amino acid pairs, amino acid pair A0 and amino acid pair A1. A protein structure database can be searched for reference amino acid sequences whose similarity to the original amino acid sequence D0 meets a set threshold. Assuming that three reference amino acid sequences, amino acid sequence D1, amino acid sequence D2, and amino acid sequence D3, are found, then amino acid sequence D1, amino acid sequence D2, and amino acid sequence D3 each contain two amino acid pairs. By converting amino acid sequence D1, amino acid sequence D2, and amino acid sequence D3 into a feature matrix and inputting it into a trained distance prediction network, it can be determined that the actual distance between amino acids a01 and a02 in amino acid pair A0 is SD0, and the actual distance between amino acids a11 and a12 in amino acid pair A1 is SD1.

[0134] Step S303 : determining the template distance between the two amino acids in each second amino acid pair in the initial structure model.

[0135] The template distance between the two amino acids in each second amino acid pair can be extracted from the constructed initial structural model. The extracted template distance can be as follows: Figure 4 shown.

[0136] For example, the initial structural model contains three amino acid pairs, namely amino acid pair A2, amino acid pair A3 and amino acid pair A4. The template distance between amino acids a21 and a22 in amino acid pair A2 can be extracted as MD2, the template distance between amino acids a31 and a32 in amino acid pair A3 can be extracted as MD3, and the template distance between amino acids a41 and a42 in amino acid pair A4 can be extracted as MD4.

[0137] Step S304 : determining the molecular energy of the initial structural model based on the actual distance between the two amino acids in each first amino acid pair and the template distance between the two amino acids in each second amino acid pair.

[0138] After determining the actual distance between the two amino acids in each first amino acid pair in the original amino acid sequence and the template distance between the two amino acids in each second amino acid pair in the initial structural model, the molecular energy of the initial structural model can be determined based on the actual distance between the two amino acids in each first amino acid pair and the template distance between the two amino acids in each second amino acid pair.

[0139] Step S305, determining whether the obtained molecular energy meets the preset loop iteration condition; if yes, executing step S306; if not, executing step S307.

[0140] An energy minimization operation is performed on the initial structure model according to the molecular energy in a cyclic iteration manner until the molecular energy of the initial structure model meets the preset cyclic iteration conditions, and the target structure model of the protein to be processed is determined based on the initial structure model output in the final round. Specifically, when the initial structure model is subjected to multiple rounds of energy minimization based on the molecular energy, if the difference between the molecular energy obtained in the current cyclic iteration and the molecular energy obtained in the previous cyclic iteration is less than a set threshold, or if a total of 2N cyclic iterations are performed and the difference between the sum of the molecular energies obtained in the last N cyclic iterations and the sum of the molecular energies obtained in the first N cyclic iterations is less than a set threshold, it can be determined that the molecular energy of the initial structure model meets the preset cyclic iteration conditions, and the cyclic iteration process can be stopped; otherwise, it can be determined that the molecular energy of the initial structure model does not meet the preset cyclic iteration conditions.

[0141] Step S306: Obtain the target structural model of the protein to be processed.

[0142] If the molecular energy of the initial structural model meets the preset cyclic iteration conditions, the cyclic iteration process of the initial structural model can be stopped to obtain the target structural model of the protein to be processed.

[0143] Step S307: adjusting the initial structural model.

[0144] If the molecular energy of the initial structural model does not meet the preset cyclic iteration conditions, the initial structural model can be adjusted, and the template distance between the two amino acids in each second amino acid pair in the adjusted initial structural model can be determined respectively, and the molecular energy can be re-determined. Then, the energy minimization operation is performed on the initial structural model according to the re-determined molecular energy until the molecular energy meets the preset cyclic iteration conditions, thereby determining the target structural model of the protein to be processed. The target structural model can be as follows: Figure 4 shown.

[0145] In one embodiment, in the above step S301, the specific process of constructing the initial structural model corresponding to the protein to be processed according to the original amino acid sequence of the protein to be processed can be as follows: Figure 5 As shown, the following steps are included:

[0146] Step S501 : Based on the preset correspondence between the amino acid sequence and the structure template, a candidate amino acid sequence matching the original amino acid sequence is searched in a protein structure database.

[0147] Each amino acid sequence in the protein structure database can be aligned with the original amino acid sequence of the protein to be processed to determine the similarity between each amino acid sequence and the original amino acid sequence, and the amino acid sequence with the greatest similarity can be used as a candidate amino acid sequence that matches the original amino acid sequence.

[0148] Step S502: The structural template corresponding to the candidate amino acid sequence is used as the candidate structural template corresponding to the original amino acid sequence.

[0149] Step S503: constructing an initial structural model of the protein to be processed based on the candidate structural template.

[0150] After determining the candidate amino acid sequence that matches the original amino acid sequence, the structural template corresponding to the candidate amino acid sequence can be used as the candidate structural template corresponding to the original amino acid sequence, and the initial structural model of the protein to be processed is constructed based on the candidate structural template.

[0151] In another embodiment, in the above step S304, the specific process of determining the molecular energy of the initial structure model based on the actual distance between the two amino acids in each first amino acid pair and the template distance between the two amino acids in each second amino acid pair can be as follows: Figure 6 As shown, the following steps are included:

[0152] Step S601 : aligning each second amino acid pair with each first amino acid pair to determine the missing portion of the initial structure model relative to the original amino acid sequence.

[0153] After obtaining the initial structural model of the protein to be processed, each second amino acid pair within the initial structural model can be compared with each first amino acid pair within the original amino acid sequence to determine the missing portion of the initial structural model relative to the original amino acid sequence. For example, the original amino acid sequence contains four amino acid pairs, namely amino acid pair A0, amino acid pair A2, amino acid pair A3, and amino acid pair A4, and the initial structural model contains six amino acid pairs, namely amino acid pair A1, amino acid pair A2, amino acid pair A4, amino acid pair A5, amino acid pair A6, and amino acid pair A7. By comparing the amino acid pairs contained in the original amino acid sequence with the amino acid pairs contained in the initial structural model, it can be determined that the initial structural model is missing amino acid pair A0 and amino acid pair A3 relative to the original amino acid sequence. Therefore, the missing portion of the initial structural model relative to the original amino acid sequence is amino acid pair A0 and amino acid pair A3.

[0154] Step S602: Determine the difference between each template distance and each corresponding actual distance.

[0155] After determining the template distance between the two amino acids in each second amino acid pair in the initial structure model and the actual distance between the two amino acids in each first amino acid pair in the original amino acid sequence, the distance matrix composed of each template distance and the distance matrix composed of each actual distance can be input into the difference prediction network, and the difference value between each template distance and the corresponding actual distance can be determined based on the difference prediction network. The difference value can be as follows Figure 5 shown.

[0156] Step S603 : determining the energy distance constraint parameter of each second amino acid pair in the initial structure model according to the difference between each template distance and each corresponding actual distance.

[0157] The distance constraint parameter can be obtained by the following formula:

[0158]

[0159] Among them, delta i,j is the distance constraint parameter for the energy of each second amino acid pair in the initial structure model, dev i,j is the difference between each template distance and the corresponding actual distance. A unit of distance.

[0160] Step S604 : determining the distance energy of each second amino acid pair in the initial structure model according to the distance constraint parameter of the energy of each second amino acid pair and the template distance between the two amino acids in each second amino acid pair.

[0161] The distance energy of each second amino acid pair in the initial structure model can be obtained according to the following formula:

[0162]

[0163] Among them, g TBM (d i,j ) is the distance energy of each second amino acid pair in the initial structure model, d i,j is the distance to be optimized between the two amino acids in each second amino acid pair, d0 is the template distance between the two amino acids in each second amino acid pair, delta i,j is the distance constraint parameter for the energy of each second amino acid pair.

[0164] Step S605 , for the missing portion of the initial structure model relative to the original amino acid sequence, determine the distance energy of each first amino acid pair according to the actual distance between the two amino acids in each first amino acid pair contained in the missing portion.

[0165] The distance energy of each first amino acid pair of the missing part of the initial structure model relative to the original amino acid sequence can be determined based on the actual distance between the two amino acids in each first amino acid pair contained in the missing part, that is, the actual distance between the two amino acids in each first amino acid pair contained in the missing part can be determined by a distance prediction network, and then the distance energy of each first amino acid pair is determined based on the actual distance between the two amino acids in each first amino acid pair contained in the missing part.

[0166] Step S606 : determining the distance energy of the initial structure model based on the distance energy of each second amino acid pair and the distance energy of each first amino acid pair included in the missing portion.

[0167] The distance energy of the initial structure model can be obtained by the following formula:

[0168]

[0169] Among them, g i,j is the distance energy of the initial structure model, and for any pair of amino acids (i, j), if the amino acid pair appears in the matching region of the sequence alignment, that is, align(i, j) = 1, the distance energy of the initial structure model can be based on the distance energy g of each second amino acid pair TBM If not, that is, align(i,j) = 0, the distance energy of the initial structure model can be based on the distance energy g of each first amino acid pair contained in the missing part. FM Determine that the distance energy g FM That is, the distance energy is determined based on the actual distance of each first amino acid pair contained in the missing part.

[0170] Step S607: determining the molecular energy of the initial structure model according to the distance energy, angle energy and molecular force field energy of the initial structure model.

[0171] The molecular energy of the initial structure model can be obtained by the following formula:

[0172] E=∑ i，j g i，j +∑ i，j f i，j +P

[0173] Where E is the molecular energy of the initial structure model, g i,j is the distance energy of the initial structure model, f i,j is the angular energy of the initial structure model, and P is the molecular force field energy of the initial structure model.

[0174] Angular energy f of the initial structural model i,j The determination process of the distance energy g from the initial structure model i,j The determination process is similar to that of , and the actual angle between the two amino acids in each first amino acid pair of the original amino acid sequence and the template angle between the two amino acids in each second amino acid pair in the initial structural model can be determined respectively. Then, the difference between each template angle and the corresponding actual angle is determined, and then the angle constraint parameter of the energy of each second amino acid pair in the initial structural model is determined. Based on the angle constraint parameter of the energy of each second amino acid pair and the template angle between the two amino acids in each second amino acid pair, the angle energy f of each second amino acid pair in the initial structural model can be determined. TBM After determining the missing portion of the initial structure model relative to the original amino acid sequence, the angle energy f of each first amino acid pair can be determined based on the actual angle between the two amino acids in each first amino acid pair included in the missing portion. FM , and then the angle energy f of each second amino acid pair can be TBM and the angle energy f of each first amino acid pair contained in the missing part FM , determine the angular energy f of the initial structural model i,j .

[0175] The molecular force field energy P of the initial structural model is the energy generated by the interaction between atoms in the molecule. It can be the energy generated by the bonding interaction between atoms that make up the same molecule, or the energy generated by the van der Waals interaction between different molecules. The molecular force field can generally include bond stretching energy, bond angle bending energy, dihedral angle distortion energy, non-bonded interaction, and cross energy terms. Bond stretching energy refers to the energy change caused by the stretching and shrinking movement of each chemical bond that constitutes the molecule in the direction of the bond axis. Bond angle bending energy refers to the molecular energy change caused by the change in bond angle. Dihedral angle distortion energy refers to the energy change caused by the distortion of the molecular skeleton caused by the rotation of a single bond. Non-bonded interactions refer to non-bonded interactions related to energy, including van der Waals forces and electrostatic interactions. Cross energy terms refer to the energy change caused by the coupling between various molecular interactions.

[0176] See Figure 7 As shown, the following uses a specific application scenario to further explain the above embodiment in detail:

[0177] Assume that there is a protein D0 with an unknown structure. Before modeling protein D0 and determining its specific structure, it is necessary to first determine the amino acids contained in protein D0, and then obtain the amino acid sequence A0 corresponding to protein D0 based on the amino acids contained in protein D0.

[0178] Step S701: Searching for an amino acid sequence A1 that matches the amino acid sequence A0 in a protein structure database.

[0179] Each amino acid sequence in the protein structure database can be aligned with the amino acid sequence A0 to determine the similarity between each amino acid sequence and the amino acid sequence A0, and the amino acid sequence with the greatest similarity is used as the amino acid sequence A1 matching the amino acid sequence A0.

[0180] Step S702: The structural template corresponding to the amino acid sequence A1 is used as the structural template H0 corresponding to the amino acid sequence A0.

[0181] Step S703: constructing an initial structural model M0 of protein D0 based on the structural template H0.

[0182] After determining amino acid sequence A1, since the protein structure database also contains a structural template corresponding to amino acid sequence A1, the structural template corresponding to amino acid sequence A1 can be directly used as the structural template H0 corresponding to amino acid sequence A0. Furthermore, the initial structural model M0 of protein D0 can be constructed based on structural template H0.

[0183] Step S704 : determining the actual distance SD0 between the two amino acids in each first amino acid pair in the amino acid sequence A0 .

[0184] We can first search the protein structure database for all reference amino acid sequences whose similarity to the amino acid sequence A0 meets a set threshold, then generate a feature matrix based on the structural features of each reference amino acid sequence, and input the feature matrix into the distance prediction network to determine the actual distance SD0 between the two amino acids in each first amino acid pair.

[0185] Step S705 : determining the template distance MD0 between the two amino acids in each second amino acid pair in the initial structure model M0 .

[0186] After constructing the initial structural model M0 corresponding to the protein D0, the template distance MD0 between the two amino acids in each second amino acid pair can be extracted from the initial structural model M0.

[0187] Step S706 , aligning each second amino acid pair in the initial structural model M0 with each first amino acid pair in the amino acid sequence A0 to determine the missing portion of the initial structural model M0 relative to the amino acid sequence A0.

[0188] Step S707 : determining the difference value C0 between each template distance MD0 and each corresponding actual distance SD0 .

[0189] The distance matrix composed of each template distance MD0 and the distance matrix composed of each actual distance SD0 are respectively input into the difference prediction network, and the difference value C0 between each template distance MD0 and the corresponding actual distance can be obtained.

[0190] Step S708 : determining the distance energy gE1 of each second amino acid pair in the initial structure model M0 according to the difference value C0 and the template distance MD0 .

[0191] According to the difference value C0, the distance constraint parameter B0 of the energy of each second amino acid pair can be determined first, and then according to the distance constraint parameter B0 and the template distance MD0, the distance energy gE1 of each second amino acid pair in the initial structure model M0 can be determined.

[0192] Step S709 , for the missing portion of the initial structure model M0 relative to the amino acid sequence A0 , determine the distance energy gE2 of each first amino acid pair according to the actual distance SD0 between the two amino acids in each first amino acid pair included in the missing portion.

[0193] Step S710 : determining the distance energy gE0 of the initial structure model M0 according to the distance energy gE1 and the distance energy gE2 .

[0194] Step S711 , determining the molecular energy E0 of the initial structure model M0 according to the distance energy gE0 , the angle energy fE0 and the molecular force field energy PE0 .

[0195] Step S712 , performing multiple rounds of energy minimization on the initial structure model M0 according to the molecular energy E0 to obtain the target structure model M1 of the protein D0.

[0196] Based on the obtained molecular energy E0, multiple rounds of energy minimization iterations can be performed on the initial structural model M0. When the molecular energy E0 satisfies a preset iteration condition, the iterations can be terminated to obtain the target structural model M1 of protein D0. If the molecular energy E0 does not satisfy the preset iteration condition, the initial structural model M0 can be adjusted and steps S705 to S712 can be repeated until the molecular energy E0 satisfies the preset iteration condition.

[0197] and Figure 2 The protein structure model construction method described above is based on the same inventive concept. The present embodiment also provides a protein structure model construction device. This protein structure model construction device can be deployed in a server or terminal device. Since this device corresponds to the protein structure model construction method of the present application and solves the problem in a similar manner to the method, the implementation of this device can be referenced to the implementation of the aforementioned method, and any repetitions will not be repeated here.

[0198] Figure 8 A schematic diagram of a protein structure model construction device provided in an embodiment of the present application is shown. Figure 8 As shown, the protein structure model building device includes an initial model building unit 801, an actual distance determination unit 802 and a target model determination unit 803.

[0199] The initial model building unit 801 is used to build an initial structural model of the protein to be processed based on the candidate structural template corresponding to the original amino acid sequence of the protein to be processed;

[0200] an actual distance determining unit 802 for determining, for each first amino acid pair in the original amino acid sequence, the actual distance between the two amino acids in each first amino acid pair;

[0201] The target model determining unit 803 is configured to adjust the initial structure model based on the actual distance between the two amino acids in each first amino acid pair to obtain a target structure model of the protein to be processed.

[0202] In an optional embodiment, as Figure 9 As shown, the above-mentioned protein structure model construction apparatus may further include a candidate template determination unit 901, which is used to:

[0203] Based on the correspondence between the preset amino acid sequence and the structural template, a candidate amino acid sequence matching the original amino acid sequence is searched in the protein structure database;

[0204] The structural template corresponding to the candidate amino acid sequence is used as the candidate structural template corresponding to the original amino acid sequence.

[0205] In an optional embodiment, the candidate template determining unit 901 is specifically configured to:

[0206] Align each amino acid sequence in the protein structure database with the original amino acid sequence to determine the similarity between each amino acid sequence and the original amino acid sequence;

[0207] The amino acid sequence with the greatest similarity is taken as the candidate amino acid sequence that matches the original amino acid sequence.

[0208] In an optional embodiment, the actual distance determining unit 802 is specifically configured to:

[0209] determining at least one reference amino acid sequence whose similarity to the original amino acid sequence satisfies a set threshold;

[0210] Generate a feature matrix based on the structural features of each reference amino acid sequence;

[0211] The feature matrix is ​​input into a distance prediction network, and the actual distance between the two amino acids in each first amino acid pair is determined based on the distance prediction network.

[0212] In an optional embodiment, the target model determining unit 803 is specifically configured to:

[0213] The following operations are performed in a cyclic iterative manner until the molecular energy of the initial structural model meets the preset cyclic iteration conditions. The target structural model of the protein to be processed is determined based on the initial structural model output in the final round:

[0214] Comparing each second amino acid pair in the initial structural model with each first amino acid pair in the original amino acid sequence to determine the missing portion of the initial structural model relative to the original amino acid sequence;

[0215] For each second amino acid pair in the initial structure model, determining the template distance between the two amino acids in each second amino acid pair, and determining the difference between each template distance and each corresponding actual distance;

[0216] Based on the difference between each template distance and each corresponding actual distance, and the actual distance of the first amino acid pair included in the missing part, the molecular energy of the initial structure model is determined, and the initial structure model is adjusted according to the molecular energy.

[0217] In an optional embodiment, the target model determining unit 803 is further configured to:

[0218] The distance matrix composed of each template distance and the distance matrix composed of each actual distance are both input into the difference prediction network, and the difference value between each template distance and the corresponding actual distance is determined based on the difference prediction network.

[0219] In an optional embodiment, the target model determining unit 803 is further configured to:

[0220] determining a distance constraint parameter for the energy of each second amino acid pair based on the difference between each template distance and each corresponding actual distance;

[0221] determining a distance energy for each second amino acid pair based on a distance constraint parameter of the energy of each second amino acid pair and a template distance between two amino acids in each second amino acid pair;

[0222] For the missing portion of the initial structure model relative to the original amino acid sequence, determining the distance energy of each first amino acid pair according to the actual distance between the two amino acids in each first amino acid pair contained in the missing portion;

[0223] The molecular energy of the initial structural model is determined based on the distance energy of each second amino acid pair and the distance energy of each first amino acid pair included in the missing portion.

[0224] In an optional embodiment, the target model determining unit 803 is further configured to:

[0225] The initial structural model outputted in the final round is used as the first structural model;

[0226] Randomly change the position of any amino acid in the first structural model to obtain the second structural model;

[0227] If the molecular energy of the second structural model is less than the molecular energy of the first structural model, the second structural model is used as the target structural model of the protein to be processed.

[0228] In an optional embodiment, the target model determining unit 803 is further configured to:

[0229] The initial structural model outputted in the final round is used as the first structural model;

[0230] Adjusting the position of any amino acid in the abnormal amino acid pair in the first structural model to obtain a third structural model; the abnormal amino acid pair is an amino acid pair whose template distance does not meet the set distance range;

[0231] If the molecular energy of the third structural model is less than the molecular energy of the first structural model, the third structural model is used as the target structural model of the protein to be processed.

[0232] Based on the same inventive concept as the above method embodiment and device embodiment, an electronic device is also provided in the embodiment of the present application. The electronic device may be a server, such as Figure 1 In this embodiment, the structure of the electronic device can be as follows: Figure 10 As shown, it includes a memory 1001 , a communication module 1003 and one or more processors 1002 .

[0233] Memory 1001 is used to store computer programs executed by processor 1002. Memory 1001 may mainly include a program storage area and a data storage area. The program storage area may store an operating system and programs required for running instant messaging functions, while the data storage area may store various instant messaging messages and operating instruction sets.

[0234] Memory 1001 may be a volatile memory, such as random-access memory (RAM); or a non-volatile memory, such as read-only memory, flash memory, a hard disk drive (HDD), or a solid-state drive (SSD); or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. Memory 1001 may be a combination of the aforementioned memories.

[0235] The processor 1002 may include one or more central processing units (CPUs) or digital processing units, etc. The processor 1002 is configured to implement the above-mentioned protein structure model construction method when calling the computer program stored in the memory 1001 .

[0236] The communication module 1003 is used to communicate with the terminal device and other electronic devices. If the electronic device is a server, the server can receive the original amino acid sequence of the protein to be processed sent by the terminal device through the communication module 1003.

[0237] The specific connection medium between the memory 1001, the communication module 1003 and the processor 1002 is not limited in the embodiment of the present application. Figure 10 In the embodiment, the memory 1001 and the processor 1002 are connected via a bus 1004. The bus 1004 is connected to the processor 1002 via a bus 1004. Figure 10 The connections between the other components are shown in bold lines, which are only for illustration and are not intended to be limiting. The bus 1004 can be divided into an address bus, a data bus, a control bus, etc. For ease of illustration, Figure 10 Only one thick line is used in the diagram, but this does not mean that there is only one bus or one type of bus.

[0238] In another embodiment, the electronic device may also be a tablet computer, a desktop computer, a laptop computer, etc.

[0239] According to one aspect of the present application, a computer program product or computer program is provided, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, causing the computer device to perform the protein structure model construction method in the above-mentioned embodiment. The program product can use any combination of one or more readable media. The readable medium can be a readable signal medium or a readable storage medium. The readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of readable storage media (a non-exhaustive list) include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0240] The above is only a specific implementation method of the present application, but the scope of protection of the present application is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed in this application, which should be covered by the scope of protection of the present application.

Claims

1. A method for constructing a protein structure model, characterized in that: The method comprises: constructing an initial structural model of the protein to be processed based on a candidate structural template corresponding to the original amino acid sequence of the protein to be processed; For each first amino acid pair in the original amino acid sequence, determining the actual distance between the two amino acids in each first amino acid pair; The following operations are performed in a cyclic iterative manner until the molecular energy of the initial structural model meets the preset cyclic iterative conditions, and the target structural model of the protein to be processed is determined based on the initial structural model output in the final round: Comparing each second amino acid pair in the initial structural model with each first amino acid pair in the original amino acid sequence to determine a missing portion of the initial structural model relative to the original amino acid sequence; For each second amino acid pair in the initial structure model, respectively determine the template distance between the two amino acids in each second amino acid pair, and determine the difference between each template distance and each corresponding actual distance; Based on the difference between each template distance and each corresponding actual distance, and the actual distance of the first amino acid pair included in the missing part, the molecular energy of the initial structural model is determined, and the initial structural model is adjusted according to the molecular energy.

2. The method according to claim 1, characterized in that Before constructing the initial structural model of the protein to be processed based on the candidate structural template corresponding to the original amino acid sequence of the protein to be processed, the method further includes: Performing sequence alignment between each amino acid sequence in a protein structure database and the original amino acid sequence to determine the similarity between each amino acid sequence and the original amino acid sequence; the protein structure database stores the corresponding relationship between each amino acid sequence and the structural template; The amino acid sequence with the greatest similarity is taken as a candidate amino acid sequence that matches the original amino acid sequence; The structural template corresponding to the candidate amino acid sequence is used as the candidate structural template corresponding to the original amino acid sequence.

3. The method according to claim 1, characterized in that The step of determining, for each first amino acid pair in the original amino acid sequence, the actual distance between the two amino acids in each first amino acid pair comprises: determining at least one reference amino acid sequence whose similarity to the original amino acid sequence meets a set threshold; generating a feature matrix according to the structural features of each of the reference amino acid sequences; The feature matrix is ​​input into a distance prediction network, and the actual distance between the two amino acids in each first amino acid pair is determined based on the distance prediction network.

4. The method according to claim 1, wherein Determining the difference between each template distance and each corresponding actual distance includes: The distance matrix composed of each template distance and the distance matrix composed of each actual distance are input into the difference prediction network, and the distance matrix composed of each template distance and the distance matrix composed of each actual distance are compared by the difference prediction network to determine the difference value between each template distance and the corresponding actual distance.

5. The method according to claim 1, characterized in that Determining the molecular energy of the initial structural model based on the difference between each template distance and each corresponding actual distance, and the actual distance of the first amino acid pair of the missing portion, comprises: determining a distance constraint parameter for the energy of each second amino acid pair based on the difference between each template distance and each corresponding actual distance; determining the distance energy of each second amino acid pair according to the distance constraint parameter of the energy of each second amino acid pair and the template distance between the two amino acids in each second amino acid pair; For the missing portion of the initial structure model relative to the original amino acid sequence, determining the distance energy of each first amino acid pair according to the actual distance between two amino acids in each first amino acid pair contained in the missing portion; The molecular energy of the initial structural model is determined based on the distance energy of each second amino acid pair and the distance energy of each first amino acid pair included in the missing portion.

6. The method according to any one of claims 1 to 5, characterized in that Determining the target structural model of the protein to be processed based on the initial structural model outputted in the final round includes: The initial structural model outputted in the final round is used as the first structural model; Randomly changing the position of any amino acid in the first structural model to obtain a second structural model; If the molecular energy of the second structural model is less than the molecular energy of the first structural model, the second structural model is used as the target structural model of the protein to be processed.

7. The method according to claim 6, characterized in that After obtaining the second structural model, determining the target structural model of the protein to be processed based on the initial structural model output in the final round further includes: Adjusting the position of any amino acid in the abnormal amino acid pair in the second structural model to obtain a third structural model; the abnormal amino acid pair is an amino acid pair whose template distance does not meet the set distance range; If the molecular energy of the third structural model is less than the molecular energy of the second structural model, the third structural model is used as the target structural model of the protein to be processed.

8. The method according to any one of claims 1 to 5, characterized in that Determining the target structural model of the protein to be processed based on the initial structural model outputted in the final round includes: The initial structural model outputted in the final round is used as the first structural model; Adjusting the position of any amino acid in the abnormal amino acid pair in the first structural model to obtain a third structural model; the abnormal amino acid pair is an amino acid pair whose template distance does not meet the set distance range; If the molecular energy of the third structural model is less than the molecular energy of the first structural model, the third structural model is used as the target structural model of the protein to be processed.

9. A protein structure model construction device, characterized in that: include: An initial model building unit, configured to build an initial structural model of the protein to be processed based on a candidate structural template corresponding to the original amino acid sequence of the protein to be processed; an actual distance determining unit, configured to determine, for each first amino acid pair in the original amino acid sequence, an actual distance between two amino acids in each first amino acid pair; a target model determination unit, configured to perform the following operations in a cyclic iterative manner until the molecular energy of the initial structure model satisfies a preset cyclic iterative condition, and determine the target structure model of the protein to be processed based on the initial structure model output in the final round: aligning each second amino acid pair in the initial structure model with each first amino acid pair in the original amino acid sequence to determine a missing portion of the initial structure model relative to the original amino acid sequence; For each second amino acid pair in the initial structure model, respectively determine the template distance between the two amino acids in each second amino acid pair, and determine the difference between each template distance and each corresponding actual distance; Based on the difference between each template distance and each corresponding actual distance, and the actual distance of the first amino acid pair included in the missing part, the molecular energy of the initial structural model is determined, and the initial structural model is adjusted according to the molecular energy.

10. A computer-readable storage medium storing a computer program, wherein: When the computer program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

11. An electronic device, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and when the computer program is executed by the processor, the method according to any one of claims 1 to 8 is implemented.

Images