Protein affinity prediction method and device, electronic equipment and computer readable storage medium

Abstract

The invention relates to a computer-aided drug design technology, and provides a protein affinity prediction method and device, electronic equipment and a computer readable storage medium, the method comprises the following steps: obtaining a three-dimensional structure of a protein, the three-dimensional structure comprises a plurality of protein chains, and each protein chain comprises a plurality of amino acids; determining a feature vector of each amino acid according to a connection relationship between the amino acids, wherein the connection relationship comprises intra-chain connection between the amino acids belonging to the same protein chain and inter-chain connection between the amino acids belonging to different protein chains; and inputting the feature vectors of all amino acids into a pre-trained prediction model to obtain binding free energy of the protein, and calculating the affinity of the protein according to the binding free energy. According to the method, through intra-chain connection and inter-chain connection between the amino acids, the determined characteristics of the amino acids are more comprehensive, and accurate prediction of the protein-protein binding affinity is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of computer-aided drug design, and in particular to a protein affinity prediction method, device, electronic device and computer-readable storage medium. Background Art

[0002] Protein-protein interactions refer to the interactions between two or more proteins through physical contact, which play a key role in many biological processes, such as signal transduction, immune response and metabolic control.

[0003] Affinity refers to the strength or stability of protein-protein interaction. Affinity reflects the tightness of the binding between two proteins. With the improvement of data and computing power, artificial intelligence methods have shown great potential in the prediction of protein-protein binding affinity.

[0004] How to improve the accuracy of protein-protein binding affinity prediction is a technical problem that needs to be solved urgently by those skilled in the art. Summary of the Invention

[0005] The present invention aims to provide a protein affinity prediction method, apparatus, electronic device and computer-readable storage medium, which, through intrachain and interchain connections between amino acids, enable more comprehensive determination of amino acid characteristics, thereby utilizing the more comprehensive amino acid characteristics to achieve accurate prediction of protein binding free energy, and ultimately achieve accurate prediction of protein-protein binding affinity.

[0006] The embodiments of the present invention can be implemented as follows:

[0007] In a first aspect, the present invention provides a method for predicting protein affinity, comprising:

[0008] Obtaining a three-dimensional structure of a protein, wherein the three-dimensional structure includes a plurality of protein chains, each of the protein chains including a plurality of amino acids;

[0009] Determining a feature vector for each amino acid based on a connection relationship between the amino acids, wherein the connection relationship includes intra-chain connections between amino acids belonging to the same protein chain and inter-chain connections between amino acids belonging to different protein chains;

[0010] The feature vectors of all the amino acids are input into a pre-trained prediction model to obtain the binding free energy of the protein, so as to calculate the affinity of the protein according to the binding free energy.

[0011] In an optional embodiment, the step of determining the feature vector of each amino acid according to the connection relationship between the amino acids includes:

[0012] Taking the amino acids as nodes and determining edges and edge feature vectors between corresponding nodes according to the connection relationship between the amino acids, thereby generating a graph structure;

[0013] The feature vector of each amino acid is updated according to the graph structure.

[0014] In an optional embodiment, the step of generating a graph structure by taking the amino acids as nodes and determining edges and edge feature vectors between corresponding nodes according to the connection relationship between the amino acids includes:

[0015] For the connection relationship between any two amino acids in the three-dimensional structure, obtaining the spatial distance between the two amino acids in the three-dimensional structure;

[0016] Determining a weight vector of the connection relationship between the two amino acids based on the spatial distance, the preset interval segmentation, and the atomic types of the two amino acids;

[0017] The graph structure is generated by taking the amino acids as nodes, the connection relationships between the amino acids as edges, and the weight vectors of the connection relationships as edge feature vectors.

[0018] In an optional embodiment, before the step of generating a graph structure by using the amino acids as nodes and determining edges and edge feature vectors between corresponding nodes according to the connection relationship between the amino acids, the step includes:

[0019] For any two amino acids in the three-dimensional structure, calculating the spatial distance between the two amino acids in the three-dimensional structure;

[0020] If the two amino acids belong to the same protein chain, determining the intra-chain connection of the two amino acids based on the spatial distance and a first preset value;

[0021] If the two amino acids belong to different protein chains, the interchain connection of the two amino acids is determined based on the spatial distance and a second preset value, where the first preset value is less than the second preset value.

[0022] In an optional embodiment, if the two amino acids belong to the same protein chain, the step of determining the intra-chain connection of the two amino acids based on the spatial distance and the first preset value includes:

[0023] If the spatial distance is less than or equal to the first preset value, it is determined that the two amino acids have an intra-chain connection; otherwise, it is determined that the two amino acids do not have an intra-chain connection;

[0024] If the two amino acids belong to different protein chains, the step of determining the interchain connection of the two amino acids according to the spatial distance and the second preset value includes:

[0025] If the spatial distance is less than or equal to the second preset value, it is determined that the interchain connection exists between the two amino acids; otherwise, it is determined that the interchain connection does not exist between the two amino acids.

[0026] In an optional embodiment, each amino acid corresponding to each node in the graph structure has an initial vector, and the step of updating the feature vector of each amino acid according to the graph structure includes:

[0027] For each node in the graph structure, the initial vector of the amino acid corresponding to each node is updated according to the nodes associated with each node in the graph structure to obtain the feature vector of the amino acid corresponding to each node.

[0028] In an optional embodiment, the step of updating, for each node in the graph structure, the initial vector of the amino acid corresponding to each node according to the nodes associated with each node in the graph structure to obtain the feature vector of the amino acid corresponding to each node includes:

[0029] For any target node, the target node is used as the current node, and the initial vector of amino acids corresponding to the target node is used as the current feature vector;

[0030] Taking nodes in the graph structure that are directly connected to the current node as neighbor nodes;

[0031] Selecting a preset number of nodes from the neighboring nodes as selected nodes;

[0032] updating the current feature vector according to the initial vector of each selected node and the connection relationship between the current node and each selected node to obtain an updated current feature vector;

[0033] Re-using the updated feature vector as the current feature vector, re-using all the selected nodes as the current node, and returning to the step of using nodes directly connected to the current node in the graph structure as neighbor nodes until a preset condition is met;

[0034] The current feature vector obtained last time is used as the feature vector of the amino acid corresponding to the target node.

[0035] In an optional embodiment, after the step of inputting the feature vectors of all the amino acids into a pre-trained prediction model to obtain the binding free energy of the protein, the step further includes:

[0036] The predicted value of the affinity of the protein is calculated based on the binding free energy, a preset constant and temperature.

[0037] In a second aspect, the present invention provides a protein affinity prediction device, comprising:

[0038] an acquisition module, configured to acquire a three-dimensional structure of a protein, wherein the three-dimensional structure includes a plurality of protein chains, and each of the protein chains includes a plurality of amino acids;

[0039] a determination module, configured to determine a feature vector for each of the amino acids based on a connection relationship between the amino acids, wherein the connection relationship includes intra-chain connections between amino acids belonging to the same protein chain and inter-chain connections between amino acids belonging to different protein chains;

[0040] The prediction module is used to input the feature vectors of all the amino acids into a pre-trained prediction model to obtain the binding free energy of the protein, so as to calculate the affinity of the protein according to the binding free energy.

[0041] In a third aspect, the present invention provides an electronic device comprising a processor and a memory, wherein the memory is used to store a program, and the processor is used to implement the protein affinity prediction method as described in any one of the aforementioned embodiments when executing the program.

[0042] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the protein affinity prediction method as described in any one of the aforementioned embodiments.

[0043] Compared with the prior art, the present invention has the following beneficial effects:

[0044] This embodiment determines a feature vector for each amino acid based on the connectivity relationships between them. The feature vectors for all amino acids are then input into a pre-trained prediction model to obtain the protein's binding free energy, which is then used to calculate the protein's affinity. Because connectivity relationships include both intrachain connections between amino acids belonging to the same protein chain and interchain connections between amino acids belonging to different protein chains, the amino acid characteristics determined are more comprehensive, reflecting both the characteristics of amino acid interactions within a protein chain and those between amino acids in different protein chains. This more comprehensive set of amino acid characteristics allows for accurate prediction of protein binding free energy, ultimately leading to accurate prediction of protein-protein binding affinity. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without paying any creative work.

[0046] Figure 1 This is an example flowchart of the protein affinity prediction method provided in this embodiment.

[0047] Figure 2 This is an example diagram of the feature vector update process provided in this embodiment.

[0048] Figure 3 This is an example diagram of the overall process of the algorithm framework provided in this embodiment.

[0049] Figure 4 This is an example diagram comparing the training and test results of different models provided in this embodiment.

[0050] Figure 5 This is an example block diagram of the protein affinity prediction device provided in this embodiment.

[0051] Figure 6 This is a block diagram of an example of an electronic device provided in this embodiment.

[0052] Icons: 10 - electronic device; 11 - processor; 12 - memory; 13 - bus; 100 - protein affinity prediction device; 110 - acquisition module; 120 - determination module; 130 - prediction module. DETAILED DESCRIPTION

[0053] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.

[0054] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention as claimed, but rather merely represents selected embodiments of the present invention. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without creative effort are intended to fall within the scope of protection of the present invention.

[0055] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, it does not need to be further defined or explained in subsequent drawings.

[0056] In the description of the present invention, it should be noted that if the terms "upper", "lower", "inside", "outside", etc. appear, the orientation or position relationship indicated is based on the orientation or position relationship shown in the accompanying drawings, or is the orientation or position relationship in which the product of the invention is usually placed when in use. It is only for the convenience of describing the present invention and simplifying the description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it should not be understood as a limitation on the present invention.

[0057] In addition, the terms "first", "second", etc., if used, are merely used to distinguish and describe, and should not be understood as indicating or implying relative importance.

[0058] It should be noted that, in the absence of conflict, the features in the embodiments of the present invention may be combined with each other.

[0059] Machine learning is a typical method in artificial intelligence. When using machine learning models to predict protein-protein binding affinity, there are usually three methods, depending on the type of features used: sequence-driven methods, structure-driven methods, and hybrid methods.

[0060] Sequence-driven protein-protein binding affinity methods take amino acid sequences as input and directly predict binding affinity. They employ characterization methods such as SMR (Substitution Matrix Representation), PSSM (Position-Specific Scoring Matrix), and other embedding methods developed for natural language processing.

[0061] Structure-driven protein-protein binding affinity methods use the 3D structure of a protein as model input to predict binding affinity. Since the 3D structure of a protein contains a wealth of information about the protein-protein interface, the essential information of each protein's 3D structure can be captured through the careful design of the feature vector.

[0062] Hybrid protein-protein binding affinity methods are designed from multiple aspects, including protein amino acid sequence and protein three-dimensional structure. For example, sequence, structure, and other functional clues are integrated with naive Bayesian networks, and all this information is combined into a naive Bayesian classifier for PPI (Protein Protein Interaction) prediction.

[0063] Due to the limited size and high noise of existing protein-protein complex datasets, the results of PPI prediction using the above methods are not ideal.

[0064] In view of this, this embodiment provides a protein affinity prediction method, apparatus, electronic device, and computer-readable storage medium, which can improve the accuracy of PPI prediction, and will be described in detail below.

[0065] Please refer to Figure 1 , Figure 1 This is a flowchart of a protein affinity prediction method provided in this embodiment, which includes the following steps:

[0066] Step S101 : obtaining a three-dimensional structure of a protein, where the three-dimensional structure includes multiple protein chains, and each protein chain includes multiple amino acids.

[0067] In this context, the three-dimensional structure of a protein refers to the specific arrangement of protein molecules in space, which determines its biological function. Amino acids are the basic building blocks of proteins. The amino acid sequence determines protein structure, and protein structure determines protein function. Protein chains are biological macromolecules formed by linear linkages of amino acids through peptide bonds and are the foundation of protein structure and function.

[0068] Step S102 , determining a feature vector for each amino acid based on the connection relationship between amino acids, where the connection relationship includes intra-chain connections between amino acids belonging to the same protein chain and inter-chain connections between amino acids belonging to different protein chains.

[0069] In this embodiment, the feature vector of an amino acid is used to represent the context information of the amino acid in the amino acid sequence, the spatial information of the amino acid in the three-dimensional structure, and the information of the interaction between the amino acid and other amino acids.

[0070] In this embodiment, the connection relationship between amino acids characterizes the interaction between amino acids. The interaction between amino acids determines the structure, stability, and function of the protein. In this embodiment, when a connection relationship exists between two amino acids, it means that the interaction between the two amino acids cannot be ignored when predicting the affinity of the protein. Otherwise, it means that there is no interaction between the two amino acids or the interaction can be ignored. From the perspective of the protein chain as a biological macromolecule, intrachain connection can be considered as the connection within the molecule, and interchain connection can be considered as the connection between molecules.

[0071] Step S103 : inputting the feature vectors of all amino acids into a pre-trained prediction model to obtain the binding free energy of the protein, so as to calculate the affinity of the protein based on the binding free energy.

[0072] In this embodiment, the process of training the prediction model and obtaining the binding free energy of a protein using the prediction model are similar. The input data both include amino acid feature vectors, and the method for obtaining the amino acid feature vectors is the same. However, the difference is that when training the prediction model, the input data includes the feature vectors and labels of the amino acids of the protein sample, where the labels are the binding free energy of the protein sample. When using the prediction model to obtain a prediction result for the protein sample, the prediction model parameters need to be adjusted based on the prediction result, the labels, and a preset loss function. Once the preset training end conditions are met, the trained prediction model is obtained. However, when the prediction model is actually used, the input data includes the feature vectors of the amino acids of the protein to be predicted, and no labels are input. The prediction result obtained serves as the binding free energy of the protein to be predicted, and no parameter adjustment is required.

[0073] It should be noted that since protein binding free energy prediction is a typical regression task, the mean squared error (MSE) can be used as a loss function for loss assessment, and parameter adjustment can be performed based on the estimated loss value. In addition to MSE, other loss functions such as MAE (mean absolute error) and Huber loss can also be used.

[0074] It should also be noted that since the existing protein-protein complex dataset used for training is limited and the dataset has high noise, the accuracy of direct prediction of protein affinity is greatly affected. This embodiment utilizes the close correlation between protein affinity and protein binding free energy to convert the prediction of protein affinity into a prediction of protein binding free energy, and then obtains the protein affinity based on the protein binding free energy, thereby realizing indirect prediction of protein affinity and achieving accurate prediction of affinity even in scenarios where data is scarce and data noise is high.

[0075] The above method provided in this embodiment utilizes the intrachain connections between amino acids belonging to the same protein chain and the interchain connections between amino acids belonging to different protein chains, so that the determined amino acid characteristics can reflect the characteristics of amino acid interactions within the protein chain and cover the characteristics of amino acid interactions between protein chains. Furthermore, the more comprehensive amino acid characteristics are used to achieve accurate prediction of protein binding free energy, and ultimately achieve accurate prediction of protein-protein binding affinity.

[0076] In this embodiment, in order to effectively remove noise from the input data, simplify the amount of input data that needs to be processed during model training, and ensure the high quality and reliability of the input data, after obtaining the original protein data, this embodiment also preprocesses the original protein data and uses the three-dimensional structure of the preprocessed protein as the processing data of the method provided in this embodiment.

[0077] Preprocessing mainly includes: (1) removing water molecules, small molecules, and metal ions in the structure of the original protein data; (2) finding similar sequences in the structure of the original protein data, deduplicating similar sequences, and retaining the protein chain with the shortest center-of-mass distance among non-identical protein chains in the similar sequences; (3) reconstructing the data obtained in (2) so that the docking results between amino acids in the protein chain are closer to the real physiological state. The reconstruction can be done using the AutoDockFR tool. AutoDockFR is mainly used to consider the flexibility of the receptor when predicting the interaction between ligands and receptors (usually proteins), allowing the flexibility of the receptor to be simulated during the docking process, thereby providing more realistic docking results. During the reconstruction process, AutoDockFR will add polar hydrogen atoms and calculate partial charges, which is crucial to improving the accuracy of molecular representation.

[0078] In an optional embodiment, the connection relationship between amino acids is closely related to the spatial position of amino acids in the three-dimensional structure and the spatial distance between two amino acids. In order to determine the connection relationship between amino acids and accurately distinguish between intra-chain connections and inter-chain connections, this embodiment provides an implementation method:

[0079] First, for any two amino acids in the three-dimensional structure, the spatial distance between the two amino acids in the three-dimensional structure is calculated;

[0080] Secondly, if the two amino acids belong to the same protein chain, the intrachain connection of the two amino acids is determined based on the spatial distance and the first preset value;

[0081] A specific method for determining the intra-chain connection may be: if the spatial distance is less than or equal to a first preset value, it is determined that there is an intra-chain connection between the two amino acids; otherwise, it is determined that there is no intra-chain connection between the two amino acids;

[0082] Third, if the two amino acids belong to different protein chains, the interchain connection of the two amino acids is determined based on the spatial distance and a second preset value, where the first preset value is less than the second preset value.

[0083] The specific method for determining the interchain connection can be: if the spatial distance is less than or equal to the second preset value, it is determined that there is an interchain connection between the two amino acids; otherwise, it is determined that there is no interchain connection between the two amino acids.

[0084] In this embodiment, the step of determining the intra-chain connection and the step of determining the inter-chain connection may be executed simultaneously or one after the other. This embodiment does not limit the execution order of the two.

[0085] The first preset value and the second preset value can be set according to the actual scenario. For example, the first preset value is set to The second preset value is set to It is a unit used to represent extremely small lengths, especially widely used in atomic scale, molecular structure and optical wavelength measurement, also known as angstrom.

[0086] After obtaining the connection relationship between amino acids, in order to simplify the process of determining the amino acid feature vector, this embodiment introduces a graph expression method, using a graph to represent the connection relationship between amino acids, and then determining the amino acid feature vector based on the graph. One implementation method is:

[0087] First, amino acids are used as nodes, and the edges and edge feature vectors between corresponding nodes are determined according to the connection relationship between amino acids to generate a graph structure.

[0088] In this embodiment, amino acids are used as nodes in the graph. If there is a connection between two amino acids, there is an edge between the corresponding two nodes, otherwise there is no edge. As an implementation method, a one-dimensional structure can be extracted from a three-dimensional structure, that is, a one-dimensional protein sequence including multiple amino acids. A preset protein language model is used to generate a 1280-dimensional vector embedding for each amino acid in the protein sequence. The embedding vector combines the local and global information of the protein sequence, protein structure, function and evolutionary characteristics, including secondary structure, three-dimensional spatial relationship, solvent accessibility, catalytic activity, conserved regions and evolution-related characteristics, which can capture the contextual information of amino acids in the sequence and their potential functions and interactions in the three-dimensional structure. The embedding vector of the amino acid is used as the initial vector of the amino acid. The protein language model can be an Evolutionary Scale Modeling-2 (ESM-2) model, an AlphaFold2 model, etc.

[0089] Secondly, the feature vector of each amino acid is updated according to the graph structure.

[0090] In this embodiment, the graph structure includes nodes, edges, and edge feature vectors. This embodiment provides an implementation method for generating the graph structure:

[0091] First, for the connection relationship between any two amino acids in the three-dimensional structure, the spatial distance between the two amino acids in the three-dimensional structure is obtained;

[0092] Secondly, the weight vector of the connection relationship between the two amino acids is determined based on the spatial distance, the preset interval segmentation and the atomic types of the two amino acids;

[0093] In this embodiment, the weight vector of the link relationship includes a spatial distance weight vector and an atom type weight vector.

[0094] To more finely distinguish the effects of different distances on molecular interactions and improve prediction accuracy, inter-chain connections and intra-chain connections can correspond to different preset intervals. For example, the preset interval corresponding to inter-chain connections can be [0,15], and the preset interval corresponding to intra-chain connections can be [0,3.5]. The corresponding preset interval is segmented according to the preset number of segments to obtain each preset interval segment. The preset number of segments can be set according to the actual accuracy requirements. If the accuracy requirement is high, the preset number of segments is set to a larger number, and vice versa. Each interval segment corresponds to a segment identifier. The spatial distance weight vector is determined based on the segment identifier of the interval segment within which the spatial distance falls. The segment identifier of the interval segment within which the spatial distance falls can be directly used as the spatial distance weight vector, or it can be prefixed, suffixed, or encrypted based on the segment identifier of the interval segment within which the spatial distance falls, and the result of the expansion or conversion is used as the spatial distance weight vector.

[0095] Taking inter-chain connections as an example, the preset number of segments can be 10. [0, 15] is evenly divided into 10 segments, and each segment corresponds to an interval segment. For example, 0 to 9 represent segments 1 to 10, or the letters aj represent segments 1 to 10. If the spatial distance is 0.5, the corresponding spatial distance weight vector can be represented as 0.

[0096] In order to distinguish the influence of different atom types on molecular interactions in a finer granularity and improve the prediction accuracy, this embodiment sets 28 forms of intentional combinations according to seven atom types (A, C, OA, N, SA, HD, NA), including: 'A_A', 'A_C', 'A_OA', 'A_N', 'A_NA', 'A_SA', 'A_HD', 'C_C', 'C_OA', 'C_N', 'C_NA', 'C_SA', 'C_HD', 'OA_OA', 'OA_N', 'OA_NA', 'OA_SA', 'OA_HD', 'N_N', 'N_NA', 'N_SA', 'N_HD', 'NA_NA', 'NA_SA', 'NA_HD', 'SA_SA', 'SA_HD', 'HD_HD'. Where A represents an aromatic carbon atom, C represents an aliphatic carbon atom, OA represents a hydrogen bond acceptor oxygen, N represents a nitrogen atom, SA represents a sulfur atom, HD represents a hydrogen bond donor hydrogen, and NA represents a hydrogen bond acceptor nitrogen. Each combination corresponds to a combination identifier, which can be used as an atom type weight vector.

[0097] The spatial distance weight vector and the atom type weight vector are combined or concatenated to obtain the weight vector of the link relationship.

[0098] Finally, a graph structure is generated with amino acids as nodes, the connection relationships between amino acids as edges, and the weight vectors of the connection relationships as edge feature vectors.

[0099] In this embodiment, in order to ensure that the feature vector of each amino acid can reflect both its local characteristics within the protein chain and its global characteristics between protein chains, an intra-chain graph structure representing the local characteristics of the amino acid can be generated for each amino acid, and an inter-chain graph structure representing the global characteristics can be generated based on all inter-chain connections. This embodiment provides an implementation method:

[0100] (1) Generate the interchain graph structure of the protein using the amino acids associated with the interchain connections as nodes and all interchain connections as edges;

[0101] (2) Generate the intrachain graph structure of each protein chain using amino acids belonging to the same protein chain as nodes and intrachain connections between amino acids belonging to the same protein chain as edges;

[0102] (3) The inter-chain graph structure and the intra-chain graph structure of all protein chains are taken as the graph structure.

[0103] In this embodiment, after obtaining the graph structure and the initial vector of the amino acid corresponding to each node in the graph, the initial vector of each amino acid can be updated according to the graph structure to finally obtain the feature vector of each amino acid. One implementation method is:

[0104] For each node in the graph structure, the initial vector of the amino acid corresponding to each node is updated according to the nodes associated with each node in the graph structure to obtain the feature vector of the amino acid corresponding to each node.

[0105] In this embodiment, the update of each node in the graph structure can be performed sequentially or simultaneously. For the sake of ease of description, this embodiment is introduced for any target node. The implementation method can be:

[0106] (1) For any target node, the target node is taken as the current node, and the initial vector of amino acids corresponding to the target node is taken as the current feature vector;

[0107] (2) The nodes directly connected to the current node in the graph structure are regarded as neighbor nodes;

[0108] (3) Selecting a preset number of nodes from neighboring nodes as selected nodes;

[0109] In this embodiment, the preset number can be adjusted according to the actual training situation during the prediction model training. For example, the preset number is set to 3. It should be noted that if the total number of neighbor nodes is less than the preset number, all neighbor nodes are selected as nodes.

[0110] (4) updating the current feature vector according to the initial vector of each selected node and the edge feature vector between the current node and each selected node to obtain an updated current feature vector;

[0111] In this embodiment, if the graph structure includes an intra-chain graph structure and an inter-chain graph structure, this embodiment uses the same method to determine the corresponding feature vector for each amino acid corresponding to each node in the intra-chain graph structure and each amino acid corresponding to each node in the inter-chain graph structure. Specifically, the initial vector of the selected node, the edge feature vectors between the current node and each selected node, and the current feature vector are concatenated or pooled to obtain an updated current feature vector.

[0112] (5) The updated feature vector is used as the current feature vector again, all selected nodes are used as the current nodes again, and the process of returning to the step of using the nodes directly connected to the current node in the graph structure as neighbor nodes is repeated until the preset conditions are met;

[0113] In this embodiment, the preset condition can be set according to actual needs. For example, the preset condition is set to a preset number of times. If the current feature vector has been updated the preset number of times, no further updates will be made. As an implementation method, a network structure can be set for each update to aggregate the vectors of the nodes involved in the update. Aggregation can be performed using an aggregation function, which can be implemented through mean aggregation, max pooling, or long short-term memory networks. This aggregation not only ensures that the current feature vector includes the initial features of neighboring nodes but also incorporates the relationships between nodes, ensuring that the final amino acid feature vector has good generalization capabilities.

[0114] (6) The current feature vector obtained last time is used as the feature vector of the amino acid corresponding to the target node.

[0115] The above process of updating the current feature vector controls the computational cost and captures the information of multi-level neighbor nodes by progressively selecting the selected node from the neighbor nodes, thus achieving a balance between computational efficiency and comprehensive feature information.

[0116] To more intuitively illustrate the process of updating the current feature vector, please refer to Figure 2 , Figure 2 This is an example diagram of the updating process of the feature vector provided in this embodiment. Figure 2 In the example, node 1 is the target node, and the current eigenvector is the initial vector of node 1. First, node 1 is taken as the current node. Node 1 has 4 neighbor nodes. Nodes 2 and 3 are selected from the 4 neighbor nodes as selected nodes. The current eigenvector is updated according to the initial vectors of nodes 2 and 3. Nodes 2 and 3 are taken as the current nodes again. Both nodes 2 and 3 have two neighbor nodes. Nodes 4 and 5, the neighbor nodes of node 2, and nodes 6 and 7, the neighbor nodes of node 3, are taken as their respective selected nodes. The initial vector of node 2 is updated according to the initial vector of node 4 and the initial vector of node 5. The initial vector of node 3 is updated according to the initial vector of node 6 and the initial vector of node 7. The current eigenvector of node 1 is updated according to the updated initial vector of node 2 and the updated initial vector of node 3.

[0117] It should be noted that the above feature vector update process can be implemented based on the GraphSAGE algorithm. For any graph structure, it is expressed as G = (V, E), and the initial vector of node v is expressed as x v , v∈V. The update process of the GraphSAGE algorithm is as follows: First, input includes G, the initial vector x of the node v , weight matrix W k, k∈{1,…,K}, nonlinear activation function σ, aggregator function and neighbor function N. Initially, the node features is set as the initial vector of node v. In each iteration, for each node, the GraphSAGE algorithm first aggregates the features of its neighboring nodes in the previous round, then concatenates the aggregated results with the features of the current node and processes them through the activation function. After K rounds of iterations, the feature vector of the node is finally represented as It is obtained through L2 normalization to form a low-dimensional node embedding representation.

[0118] In order to introduce the protein binding free energy prediction process provided in this example from the overall process, please refer to Figure 3 , Figure 3 This is an example diagram of the overall process of the algorithm framework provided in this embodiment. Figure 3 In the figure, the purple and green graphs represent the intrachain graph structures corresponding to different protein chains, respectively, and the mixed purple and green graphs represent the interchain graph structures. For any intrachain graph structure or interchain graph structure, feature aggregation is performed through multiple updates to determine the feature vector of each amino acid. The feature vectors of all amino acids in the same intrachain graph structure are aggregated to obtain an aggregated feature vector. The feature vectors of all amino acids in the interchain graph structure are aggregated to obtain an aggregated feature vector. The aggregation method can use an average pooling layer or a maximum pooling layer. Each aggregated feature vector is used as an input to the MLP (Multi-Layer Perceptron) in the prediction model, and is input into the multilayer perceptron for prediction, ultimately obtaining the binding free energy of the protein.

[0119] In an optional embodiment, since there is a preset relationship between the binding free energy of the protein and the affinity of the protein, in the step of obtaining the binding free energy of the protein, the binding free energy of the protein can also be converted into the affinity of the protein according to the preset relationship. The specific implementation method can be:

[0120] Calculate the predicted affinity of a protein based on binding free energy, preset constants, and temperature.

[0121] In this embodiment, the preset relationship is related to the preset constant and temperature, and can be specifically expressed as: ΔG=RTlnK d , where R is a preset constant, usually the ideal gas constant, which can be 8.314 J / (mol·K), T is the temperature, which can be 298 K, ΔG is the binding free energy of the protein, K d is the dissociation constant, which is used to characterize the affinity of the protein. The smaller the value, the stronger the affinity. In addition, the affinity of the protein can also be expressed by the inhibition constant K i Indicates that Ki It is a thermodynamic parameter that directly reflects the affinity between the inhibitor and the target protein. The smaller its value, the stronger the affinity. For the binding of the inhibitor to the free protein, K i Can be regarded as a special K d The affinity of a protein can also be expressed as IC50, which is an indicator of the functional activity of an inhibitor. However, its value may be indirectly affected by the affinity between the protein and the ligand (such as an inhibitor). IC50 and K i The relationship between IC50 and K is: i ×(1+[S] / K m ), where K m is the Michaelis constant, and [S] is the substrate concentration in the experiment.

[0122] In order to more clearly reflect the effect of the method provided in this embodiment, this embodiment uses MAE (Mean Absolute Error) and R (Pearson Correlation Coefficient) as evaluation indicators.

[0123] 1. Mean Absolute Error (MAE): It is the average of the absolute values ​​of the differences between the predicted values ​​and the true values.

[0124] 2. Pearson correlation (R): The correlation between the predicted value and the actual value.

[0125] This example was tested on two benchmark datasets. Dataset 1 is from a subset of the PDB. This database contains structural information for protein-protein complexes, which may be composed of two or more protein chains, and has a total of 79 data points. Dataset 2 consists of only two protein chains and includes a homodimer complex. This test set has a total of 90 data points. Finally, a fusion function was set up to merge these two datasets and perform the test.

[0126] Table 1 compares the results of various methods.

[0127]

[0128] As can be seen from the results in the table, the ProAffinity-GNN method has an MAE of 1.65 and an R of 0.72 on Dataset 1, which is an excellent performance, indicating that it is very reliable in terms of prediction accuracy and relevance. However, the method of this embodiment further reduced the MAE to 1.42 on Dataset 1, while maintaining the R value at 0.76, outperforming ProAffinity-GNN and surpassing PRODIGY, showing the best prediction accuracy. On Dataset 2, the performance of ProAffinity-GNN is still strong, and the method of this embodiment is close to the optimal performance, which shows that the model of the method of this embodiment is still effective when processing complexes composed of two chains. At the same time, the method of this embodiment further improved the performance on the data of the merged set of Datasets 1 and 2, with an MAE of 1.54 and an R of 0.66, showing that the method of this embodiment has better prediction ability on the merged set of Datasets 1 and 2.

[0129] This example also compares it with other models, please refer to Figure 4 , Figure 4 This is an example diagram comparing the training and testing results of different models provided in this embodiment. Figure 4 Among them, ProAffinity-GNN is currently the best model; GAT_GNN is a method that superimposes graph attention network (GAT) and graph neural network (GNN). The graph attention mechanism learns the importance of nodes and inputs it into the graph neural network for learning; HeteroGraphConv is a network model of heterogeneous graphs. The original author added covalent bonds between nodes, atomic interaction networks and K-nearest neighbor networks between atoms to represent the attributes of edges in the graph; RCNN is a sequence-based model that uses a bidirectional gated recurrent unit (GRU) to learn sequence relationships.

[0130] Figure 4 The training and testing performance of different models in the PDBbind database are shown respectively. Figure 4 The MSE comparison chart in the figure compares the MSE in the training and validation stages. The yellow line (corresponding to the Validation MSE in the figure) represents the MSE in the application stage, and the blue line (corresponding to the Train MSE in the figure) represents the MSE in the training stage. The smaller the MSE value, the higher the prediction accuracy. It can be seen that the method provided in this embodiment shows the best performance. Figure 4The R comparison chart in the figure compares the Pearson correlation coefficient R in the training and validation stages. The pink line (corresponding to the Validation R in the figure) represents the R in the application stage, and the green line (corresponding to the R in the figure) represents the R in the training stage. The higher the R value, the more relevant the prediction results. It can be seen that the method provided in this example is similar to the ProAffinity-GNN method and also has good performance.

[0131] In order to execute the corresponding steps in the above embodiments and various possible implementations, a method for implementing the protein affinity prediction device 100 is given below. Figure 5 , Figure 5 This is a block diagram of the protein affinity prediction device provided in this embodiment. It should be noted that the basic principles and technical effects of the protein affinity prediction device 100 provided by the present invention are the same as those of the corresponding above-mentioned embodiments. For the sake of brief description, they are not mentioned in this embodiment.

[0132] The protein affinity prediction device 100 includes an acquisition module 110 , a determination module 120 , and a prediction module 130 .

[0133] An acquisition module 110 is used to acquire a three-dimensional structure of a protein, where the three-dimensional structure includes multiple protein chains, and each protein chain includes multiple amino acids;

[0134] a determination module 120 for determining a feature vector for each amino acid based on a connection relationship between amino acids, wherein the connection relationship includes intra-chain connections between amino acids belonging to the same protein chain and inter-chain connections between amino acids belonging to different protein chains;

[0135] The prediction module 130 is used to input the feature vectors of all amino acids into a pre-trained prediction model to obtain the binding free energy of the protein, so as to calculate the affinity of the protein based on the binding free energy.

[0136] In an optional implementation manner, the determination module 120 is specifically configured to:

[0137] Taking amino acids as nodes, the edges and edge feature vectors between corresponding nodes are determined according to the connection relationship between amino acids to generate a graph structure;

[0138] Update the feature vector of each amino acid according to the graph structure.

[0139] In an optional embodiment, when the determination module 120 is used to determine the edges and edge feature vectors between the corresponding nodes based on the connection relationship between the amino acids to generate the graph structure, it is specifically used to:

[0140] For the connection relationship between any two amino acids in the three-dimensional structure, obtaining the spatial distance between the two amino acids in the three-dimensional structure;

[0141] Determine a weight vector of the connection relationship between the two amino acids based on the spatial distance, the preset interval segmentation, and the atomic types of the two amino acids;

[0142] A graph structure is generated with amino acids as nodes, the connection relationships between amino acids as edges, and the weight vectors of the connection relationships as edge feature vectors.

[0143] In an optional implementation manner, the determining module 120 is further configured to:

[0144] For any two amino acids in the three-dimensional structure, calculating the spatial distance between the two amino acids in the three-dimensional structure;

[0145] If the two amino acids belong to the same protein chain, determining the intrachain connection of the two amino acids based on the spatial distance and the first preset value;

[0146] If the two amino acids belong to different protein chains, the interchain connection of the two amino acids is determined based on the spatial distance and a second preset value, where the first preset value is less than the second preset value.

[0147] In an optional embodiment, when determining the intra-chain connection of two amino acids based on the distance between them and the first preset value if the two amino acids belong to the same protein chain, the determination module 120 is specifically configured to:

[0148] If the spatial distance is less than or equal to the first preset value, it is determined that there is an intrachain connection between the two amino acids; otherwise, it is determined that there is no intrachain connection between the two amino acids;

[0149] When the two amino acids belong to different protein chains, the determination module 120 is used to determine the interchain connection between the two amino acids based on the spatial distance and the second preset value:

[0150] If the spatial distance is less than or equal to the second preset value, it is determined that there is an interchain connection between the two amino acids; otherwise, it is determined that there is no interchain connection between the two amino acids.

[0151] In an optional embodiment, each amino acid corresponding to a node in the graph structure has an initial vector, and the determination module 120 is specifically configured to update the feature vector of each amino acid according to the graph structure:

[0152] For each node in the graph structure, the initial vector of the amino acid corresponding to each node is updated according to the nodes associated with each node in the graph structure to obtain the feature vector of the amino acid corresponding to each node.

[0153] In an optional embodiment, the determination module 120 is configured to update, for each node in the graph structure, the initial vector of the amino acid corresponding to each node according to the nodes associated with each node in the graph structure to obtain the feature vector of the amino acid corresponding to each node, specifically for:

[0154] For any target node, the target node is taken as the current node, and the initial vector of amino acids corresponding to the target node is taken as the current feature vector;

[0155] The nodes directly connected to the current node in the graph structure are regarded as neighbor nodes;

[0156] Select a preset number of nodes from neighboring nodes as selected nodes;

[0157] According to the initial vector of each selected node and the connection relationship between the current node and each selected node, the current feature vector is updated to obtain an updated current feature vector;

[0158] The updated feature vector is used as the current feature vector again, all selected nodes are used as the current node again, and the process of using the nodes directly connected to the current node in the graph structure as neighbor nodes is returned until the preset conditions are met;

[0159] The current feature vector obtained last time is used as the feature vector of the amino acid corresponding to the target node.

[0160] In an optional embodiment, the prediction module 130 is further configured to:

[0161] Calculate the predicted affinity of a protein based on binding free energy, preset constants, and temperature.

[0162] The embodiment of the present invention also provides a block diagram of an electronic device 10, which implements the protein affinity prediction method of the above embodiment. Figure 6 , Figure 6 This is a block diagram of an electronic device 10 provided in this embodiment. The electronic device 10 includes a processor 11 , a memory 12 , and a bus 13 . The processor 11 and the memory 12 are connected via the bus 13 .

[0163] The processor 11 can be an integrated circuit chip with signal processing capabilities. During implementation, each step of the protein affinity prediction method of the above embodiment can be completed by a hardware integrated logic circuit in the processor 11 or by software instructions. The above-mentioned processor 11 can be a general-purpose processor, including a CPU (Central Processing Unit), an NP (Network Processor), etc.; it can also be a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Logic Gate Array), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0164] The memory 12 is used to store a program for implementing the protein affinity prediction method. The program can be a software function module stored in the memory 12 in the form of software or firmware or fixed in the OS (Operating System) of the electronic device 10.

[0165] After receiving the execution instruction, the processor 11 executes the program to implement the protein affinity prediction method of this embodiment.

[0166] This embodiment provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the protein affinity prediction method described in the aforementioned embodiment is implemented.

[0167] In summary, embodiments of the present invention provide a method, apparatus, electronic device, and computer-readable storage medium for predicting protein affinity. The method comprises: obtaining a three-dimensional structure of a protein, where the three-dimensional structure comprises multiple protein chains, and each protein chain comprises multiple amino acids; determining a characteristic vector for each amino acid based on the connection relationship between the amino acids, where the connection relationship comprises intra-chain connections between amino acids belonging to the same protein chain and inter-chain connections between amino acids belonging to different protein chains; and inputting the characteristic vectors of all amino acids into a pre-trained prediction model to obtain the binding free energy of the protein, so as to calculate the affinity of the protein based on the binding free energy. Compared with the prior art, this embodiment has at least the following advantages: (1) It determines the characteristics of amino acids more comprehensively through intra-chain and inter-chain connections between amino acids, and then uses the more comprehensive characteristics of amino acids to achieve accurate prediction of protein binding free energy, reduce the false alarm rate, and improve the reliability and economy of the prediction; (2) By using the close correlation between protein affinity and protein binding free energy, the prediction of protein affinity is converted into the prediction of protein binding free energy, and then the protein affinity is obtained according to the protein binding free energy, thereby achieving indirect prediction of protein affinity and accurate prediction of affinity even in scenarios with scarce data and large data noise; (3) By progressively selecting selected nodes from neighboring nodes, the computational cost is controlled and the information of multi-level neighboring nodes can be captured, achieving a balance between computational efficiency and comprehensiveness of feature information; (4) By predicting the dissociation constant K d Strongly correlated binding free energy accelerates the research progress of protein-protein interactions (PPIs), can support the rapid screening and verification of potential drug targets in the drug discovery process, and thus accelerate the development of new drugs.

[0168] The above descriptions are merely various embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications or substitutions that can be readily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A protein affinity prediction method, characterized in that: The method comprises: Obtaining a three-dimensional structure of a protein, wherein the three-dimensional structure includes a plurality of protein chains, each of the protein chains including a plurality of amino acids; Determining a feature vector for each amino acid based on a connection relationship between the amino acids, wherein the connection relationship includes intra-chain connections between amino acids belonging to the same protein chain and inter-chain connections between amino acids belonging to different protein chains; The feature vectors of all the amino acids are input into a pre-trained prediction model to obtain the binding free energy of the protein, so as to calculate the affinity of the protein according to the binding free energy.

2. The method according to claim 1, characterized in that The step of determining the characteristic vector of each amino acid according to the connection relationship between the amino acids comprises: Taking the amino acids as nodes and determining edges and edge feature vectors between corresponding nodes according to the connection relationship between the amino acids, thereby generating a graph structure; The feature vector of each amino acid is updated according to the graph structure.

3. The method according to claim 2, characterized in that The step of generating a graph structure by taking the amino acids as nodes and determining the edges and edge feature vectors between the corresponding nodes according to the connection relationship between the amino acids comprises: For the connection relationship between any two amino acids in the three-dimensional structure, obtaining the spatial distance between the two amino acids in the three-dimensional structure; Determining a weight vector of the connection relationship between the two amino acids based on the spatial distance, the preset interval segmentation, and the atomic types of the two amino acids; The graph structure is generated by taking the amino acids as nodes, the connection relationships between the amino acids as edges, and the weight vectors of the connection relationships as edge feature vectors.

4. The method according to claim 2, characterized in that Before the step of generating a graph structure by taking the amino acids as nodes and determining the edges and edge feature vectors between the corresponding nodes according to the connection relationship between the amino acids, the method includes: For any two amino acids in the three-dimensional structure, calculating the spatial distance between the two amino acids in the three-dimensional structure; If the two amino acids belong to the same protein chain, determining the intra-chain connection of the two amino acids based on the spatial distance and a first preset value; If the two amino acids belong to different protein chains, the interchain connection of the two amino acids is determined based on the spatial distance and a second preset value, where the first preset value is less than the second preset value.

5. The method according to claim 4, characterized in that If the two amino acids belong to the same protein chain, the step of determining the intra-chain connection of the two amino acids according to the spatial distance and the first preset value includes: If the spatial distance is less than or equal to the first preset value, it is determined that the two amino acids have an intra-chain connection; otherwise, it is determined that the two amino acids do not have an intra-chain connection; If the two amino acids belong to different protein chains, the step of determining the interchain connection of the two amino acids according to the spatial distance and the second preset value includes: If the spatial distance is less than or equal to the second preset value, it is determined that the interchain connection exists between the two amino acids; otherwise, it is determined that the interchain connection does not exist between the two amino acids.

6. The method according to claim 2, characterized in that Each amino acid corresponding to each node in the graph structure has an initial vector, and the step of updating the feature vector of each amino acid according to the graph structure includes: For each node in the graph structure, the initial vector of the amino acid corresponding to each node is updated according to the nodes associated with each node in the graph structure to obtain the feature vector of the amino acid corresponding to each node.

7. The method according to claim 6, characterized in that The step of updating, for each node in the graph structure, the initial vector of the amino acid corresponding to each node according to the node associated with each node in the graph structure to obtain the feature vector of the amino acid corresponding to each node includes: For any target node, the target node is used as the current node, and the initial vector of amino acids corresponding to the target node is used as the current feature vector; Taking nodes in the graph structure that are directly connected to the current node as neighbor nodes; Selecting a preset number of nodes from the neighboring nodes as selected nodes; updating the current feature vector according to the initial vector of each selected node and the edge feature vector between the current node and each selected node to obtain an updated current feature vector; Re-using the updated feature vector as the current feature vector, re-using all the selected nodes as the current node, and returning to the step of using nodes directly connected to the current node in the graph structure as neighbor nodes until a preset condition is met; The current feature vector obtained last time is used as the feature vector of the amino acid corresponding to the target node.

8. The method according to claim 1, characterized in that After the step of inputting the feature vectors of all the amino acids into a pre-trained prediction model to obtain the binding free energy of the protein, the method further includes: The predicted value of the affinity of the protein is calculated based on the binding free energy, a preset constant and temperature.

9. A protein affinity prediction device, characterized in that: The device comprises: an acquisition module, configured to acquire a three-dimensional structure of a protein, wherein the three-dimensional structure includes a plurality of protein chains, and each of the protein chains includes a plurality of amino acids; a determination module, configured to determine a feature vector for each of the amino acids based on a connection relationship between the amino acids, wherein the connection relationship includes intra-chain connections between amino acids belonging to the same protein chain and inter-chain connections between amino acids belonging to different protein chains; The prediction module is used to input the feature vectors of all the amino acids into a pre-trained prediction model to obtain the binding free energy of the protein, so as to calculate the affinity of the protein according to the binding free energy.

10. An electronic device, characterized in that: The method comprises a processor and a memory, wherein the memory is used to store a program, and the processor is used to implement the protein affinity prediction method according to any one of claims 1 to 8 when executing the program.

11. A computer-readable storage medium, characterized in that A computer program is stored thereon, and when the computer program is executed by a processor, the protein affinity prediction method according to any one of claims 1 to 8 is implemented.

Images