Method for predicting pharmacological action target of atractylin based on a graph neural network

Abstract

The application discloses a kind of based on graph neural network's pharmacological action target point prediction method of atractylin, comprising the following steps: S1. generating atractylin molecular structure standardization dataset;S2. the atractylin molecular structure standardization dataset is converted into molecular graph dataset;S3. constructing drug target point graph dataset;S4. based on graph neural network constructs atractylin pharmacological action target point prediction model;S5. using atractylin pharmacological action target point prediction model to the molecular graph dataset and interactive graph dataset are handled, output the interaction matching result between atractylin and each candidate target point, and determine potential pharmacological action target point according to matching result.The application can dynamically weight the characteristics of molecular atomic level and the characteristics of target point key amino acid residues, so that the prediction model can more accurately depict the local interaction characteristics of drug and target point.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artesunate, and particularly relates to a pharmacological action target point prediction method of artesunate based on a graph neural network. BACKGROUND

[0002] With the development of computer science and artificial intelligence technology, machine learning and deep learning technology are gradually introduced into the field of biological medicine to improve the efficiency and accuracy of new drug research and development. The prediction of small molecule drug targets is an important part of drug development, and accurately identifying the interaction between small molecules and specific targets is of great significance for drug screening, mechanism analysis and precision medicine. However, traditional drug target prediction methods have many limitations in practical application, and it is difficult to meet the needs of modern biological and medical research for efficient and accurate prediction.

[0003] At present, drug target prediction mainly relies on experimental verification, computer simulation and machine learning methods. Experimental verification includes in vitro binding experiments, cell experiments and animal experiments, which can provide relatively reliable results, but the cost is high, the cycle is long, and it is limited by experimental conditions, making it difficult to carry out large-scale screening. Computer simulation methods such as molecular docking and molecular dynamics simulation can assist drug screening, but the computational complexity is high, and there are certain assumptions in the modeling of target and small molecule interactions, making it difficult to fully reflect the dynamic changes in the real biological environment. In addition, traditional machine learning methods usually rely on artificially constructed molecular descriptors, and the feature extraction process is tedious and difficult to capture the complex nonlinear relationships between molecules.

[0004] In recent years, deep learning has made certain breakthroughs in drug research and development, especially methods based on graph neural networks, which can directly process molecular graph structure data and have shown good application prospects in drug screening and target prediction. However, existing graph neural network models still have the following technical bottlenecks in practical application: first, most models only use the topological structure information of molecules, ignoring the three-dimensional space information of molecules, which leads to limitations in modeling the molecule-target interaction; second, existing methods mainly rely on global feature mapping in the process of matching molecules and targets, making it difficult to accurately describe the interaction between local key sites; in addition, the feature aggregation method of some graph neural network models is too simple, and the interactive features between drug molecules and target proteins are not fully utilized, affecting the accuracy and generalization ability of the prediction.

[0005] In summary, the existing drug target prediction technology still has significant deficiencies in computational cost, prediction accuracy and biological interpretability, which is difficult to meet the demand of new drug research and development for efficient and accurate prediction methods. Therefore, a new method is needed to optimize the molecule-target interaction modeling by combining graph neural network technology and fully considering the topological structure and three-dimensional geometric characteristics of molecules to improve the accuracy and biological interpretability of drug target prediction. SUMMARY

[0006] One object of the present application is to provide a graph neural network-based pharmacological target prediction method for atractylin, which can dynamically weight the features of molecular atoms and the features of key amino acid residues of targets, so that the prediction model can more accurately depict the local interaction characteristics of drugs and targets.

[0007] According to the graph neural network-based pharmacological target prediction method for atractylin, the following steps are included:

[0008] S1. Collect atractylin molecular structure data and perform standardization processing on the atractylin molecular structure data to generate atractylin molecular structure standardized data set;

[0009] S2. Convert the atractylin molecular structure standardized data set into a molecular graph data set;

[0010] S3. Collect drug target data containing known drug targets and related biological information, and construct a drug target graph data set;

[0011] S4. Construct a graph neural network-based atractylin pharmacological target prediction model, integrate a graph matching attention mechanism in the atractylin pharmacological target prediction model, and use the graph matching attention mechanism to perform feature matching on the molecular graph data set and the drug target graph data set;

[0012] S5. Process the molecular graph data set and the interaction graph data set using the atractylin pharmacological target prediction model, output the interaction matching results between atractylin and each candidate target, and determine the potential pharmacological target according to the matching results.

[0013] Optionally, the S1 includes the following steps:

[0014] S11. Collect atractylin molecular structure data, which includes the atom type, chemical bond connection relationship, molecular configuration and molecular topological structure information of atractylin, and construct atractylin molecular data set D cangshu :

[0015]

[0016] M is atractylin, and D is the atractylin molecular data set.i represents the i-th atractylodin molecule, N is the total number of atractylodin molecule data, V i is the atomic set of atractylodin molecule M i i is the bond set in atractylodin molecule M i i is the topological structure matrix of atractylodin molecule M i i is the molecular configuration information of atractylodin molecule M i ;

[0017] S12. Normalizing the atractylodin molecule data set D cangshu to generate an atractylodin molecule structure standardized data set D

[0018] Optionally, the S2 comprises the following steps:

[0019] S21. Constructing an atractylodin molecule graph data set G according to the atractylodin molecule structure standardized data set D cangshu :

[0020]

[0021] wherein G i represents the molecular graph of atractylodin molecule M i , X i is the node feature matrix of atractylodin molecule M i , W i is the edge feature matrix of atractylodin molecule M i ;

[0022] S22. Defining the adjacency matrix A i of atractylodin molecule M i according to the topological structure information of each molecule in the atractylodin molecule structure standardized data set D i , the adjacency matrix A i records the topological connection relationship between atoms in atractylodin molecule M i ;

[0023] S23. Defining the three-dimensional coordinate matrix C i of atractylodin molecule M i,j according to the molecular configuration information of each molecule in the atractylodin molecule structure standardized data set D i , wherein c i in the three-dimensional coordinate matrix C i represents the three-dimensional space coordinates of the j-th atom in atractylodin molecule M i ; S26. For each atractylodin molecule M i ​, combined with its atomic feature matrix X i , edge feature matrix W i , adjacency matrix A i and three-dimensional coordinate matrix C i , the comprehensive representation of the nodes and edges of the molecule graph G i is constructed by function Φ:

[0025] Φ(X i ,W i ,A i ,C i )={v′ i,j ,e′ i,jk |j,k=1,2,…,n i};

[0026]

[0027] e′ i,jk =σ(γ·w i,jk +δ·(‖x i,j -x i,k ‖2+‖c i,j -c i, k‖2));

[0028] wherein Φ(·) represents the molecule graph construction function, v′ i,j represents the final feature of the jth atom in the atractylodin molecule M i , e′ i,jk represents the final edge feature of the edge connecting atoms v i,j and v i,k in the atractylodin molecule M i , j, k represent the atom index in the molecule, x i,j represents the original feature vector of the jth atom in the atractylodin molecule M i , w i,jk represents the chemical bond feature vector connecting atoms v i,j and v i,k , a i,jk is an element of the adjacency matrix A i , if there is a chemical bond between atoms v i,j and v i,k , then a i,jk =1, otherwise, ‖·‖2 represents the Euclidean norm, represents the element-wise product of vectors, used to fuse edge features and adjacent atom features, σ(·) is an activation function, α, β, γ, δ are learnable scaling parameters for balancing the contribution of each part of the feature, and λ is a parameter for controlling the influence degree of spatial distance;

[0029] Construct the atractylodin molecule graph dataset G cangshu :

[0030] where G i = Φ(X i ,W i ,A i ,C i ).

[0031] Optionally, the S3 comprises the following steps:

[0032] S31. Collecting drug target data containing known drug targets and related biological information, and constructing a drug target data set D target :

[0033]

[0034] where T i represents the i-th drug target, M is the total number of drug target data, R i represents the set of key amino acid residues in the drug target T i , U i represents the set of functional structure information of the drug target T i ;

[0035] S32. Based on the drug target data set D target , graph construction is performed for each drug target T i = (R i , U i ) to form a drug target graph data set G target :

[0036]

[0037] where, represents the graph structure representation of the drug target T i , represents the edge set of the internal interaction relationship of the drug target T i based on R i and U i and external biological information, or the interaction relationship between the drug target and other biological molecules, is the feature matrix of the node R i .

[0038] Optionally, the S4 comprises the following steps:

[0039] S41. Based on the atractylin molecule graph data set G cangshu and the drug target graph data set G target , a graph matching attention mechanism A GMSA that integrates topological information and spatial geometry is constructed, and an atractylin pharmacological action target prediction model F GNNThe integrated graph matching attention mechanism optimizes the feature interaction between the atractylodes molecule graph and the drug target graph.

[0040] S42. In the graph matching attention mechanism A GMSA In the middle, combined with the molecular diagram G of atractylodesin i Middle atomic node v i,j Drug target map key amino acid residues r k,m Based on topological similarity and spatial conformation constraints, calculate the dynamic matching weight α. i,jk,m :

[0041]

[0042] Where, α i,jk,m G represents atractylodes molecule i Middle atom v i,j With drug target key amino acid residues r k,m The dynamic matching weights between them, f(h) i,j ,g k,m The function is the basic feature similarity calculation function. Atractylodes molecule G i and drug targets Topological similarity measurement C is the three-dimensional coordinate matrix of the atractylodes molecule. i and the three-dimensional coordinate matrix of drug target Spatial geometric matching degree between them;

[0043] S43. Topological similarity measurement The molecular graph G of atractylodesin was calculated using a hierarchical subgraph matching method. i Drug target map Structural similarity at different scales:

[0044]

[0045] in, and These represent the G molecules of atractylodesin. i and drug targets In the local subgraph structure at layer l, Jaccard(·,·) is the Jaccard similarity coefficient, ω l The importance weights for matching at different levels, where L is the number of levels into which the subgraph is divided;

[0046] S44. Spatial Geometric Matching Metric The Laplacian regularized embedding method was used to calculate the positional similarity between atractylodes molecule and drug target in three-dimensional space:

[0047]

[0048] wherein, is a spatial adjacency matrix, representing the molecule G i of atom v i,j and the drug target of key amino acid residue r k,m is within a reasonable interaction range, if the spatial distance between them is less than a set threshold τ otherwise

[0049] S45. Combining the topological structure similarity and the spatial geometry matching measure adjust the feature aggregation way of the graph matching attention mechanism, to realize the optimal fusion of molecule-target interaction information:

[0050]

[0051] wherein, z ik is the matching feature vector between the molecule G i and the drug target , and represents the feature splicing operation.

[0052] S46. The calculated matching feature vector z ik is mapped through a fully connected layer, to output the interaction probability between the molecule G i and the drug target :

[0053] p ik = σ(W T z ik +b);

[0054] wherein, p ik represents the interaction probability between the molecule G i and the drug target , W is a trainable weight matrix, b is a bias term, and σ(·) is a sigmoid activation function.

[0055] S47. Obtain the pharmacological action target matching matrix P match of atractylodin:

[0056] P match = [p ik ] N×M ;

[0057] wherein, P match records the set of molecules G cangshu and the drug target graph dataset G targetThe matching relationship between the atractylodin molecules and the drug targets, N is the number of atractylodin molecules, and M is the number of drug targets;

[0058] S48. Utilizing the atractylodin pharmacological action target matching matrix P match As a supervision signal, the atractylodin pharmacological action target prediction model F GNN is trained, and a loss function L(F GNN ) is used for model optimization:

[0059]

[0060] Wherein, y ik is the real interaction label between the atractylodin molecule G i and the drug target , and p ik is the predicted interaction probability;

[0061] S49. Outputting the trained atractylodin pharmacological action target prediction model

[0062]

[0063] Optionally, the S includes the following steps:

[0064] S51. Utilizing the trained atractylodin pharmacological action target prediction model to process the inputted atractylodin molecule graph data set and the interaction graph data set, calculate the interaction matching feature vector z ik between the atractylodin molecule G i and the drug target , and obtain the interaction probability p through a full connection prediction layer mapping.

[0065] S52. The interaction probabilities between all atractylodin molecules and drug targets are composed into an atractylodin pharmacological action target matching matrix P

[0066] S53. The matching matrix P is subjected to threshold determination, a preset threshold τ is set, and a potential pharmacological action target set T pot is determined:

[0067]

[0068] Wherein, T k represents the kth drug target in the drug target graph data set G target , and τ is the set interaction probability threshold.

[0069] S54. Outputting the atractylodin pharmacological action target matching matrix P and a potential pharmacological action target set T pot as the final pharmacological action target prediction result of atractylodin.

[0070] The beneficial effects of the present application are:

[0071] (1) The present application integrates a graph matching attention mechanism in the graph neural network model, realizes accurate molecule-target feature mapping by calculating the dynamic matching weight between the atractylodin molecule graph and the drug target graph, and the graph matching attention mechanism can dynamically weight the features of the molecule atoms and the key amino acid residue features of the target, so that the prediction model can more accurately depict the local interaction characteristics of the drug and the target.

[0072] (2) The present application constructs a hierarchical topological structure similarity measure and a spatial geometric matching measure, fully considers the three-dimensional spatial coordinate information of the atractylodin molecule and the stereo conformation of the target protein structure, so that the model can more comprehensively depict the actual binding state between the molecule and the target.

[0073] (3) In the feature fusion process of molecule-target matching, the present application combines topological similarity weight, spatial geometric features and local interaction features, optimizes the interaction information aggregation through weighted feature splicing and full connection layer mapping, so that the molecule-target matching matrix is more biologically interpretable, and can more effectively distinguish high affinity targets from low affinity targets. BRIEF DESCRIPTION OF DRAWINGS

[0074] The accompanying drawings are included to provide a further understanding of the present application, and constitute a part of the specification, together with the embodiments of the present application, to explain the present application, and do not constitute a limitation on the present application. In the drawings:

[0075] Figure 1 A flowchart of a graph neural network-based atractylodin pharmacological action target prediction method according to the present application. DETAILED DESCRIPTION

[0076] The present application will now be described in further detail with reference to the accompanying drawings. These drawings are all simplified schematic diagrams, and only illustrate the basic structure of the present application in a schematic manner, and therefore only show the components related to the present application.

[0077] Reference Figure 1 A graph neural network-based atractylodin pharmacological action target prediction method, comprising the following steps:

[0078] S1. Collect atractylodin molecule structure data, and standardize the atractylodin molecule structure data to generate atractylodin molecule structure standardized data set;

[0079] S2. Convert the atractylodin molecule structure standardized data set into a molecule graph data set;

[0080] S3. Collecting drug target data containing known drug targets and related biological information, and constructing a drug target graph data set;

[0081] S4. Constructing a atractylin pharmacological target prediction model based on a graph neural network, integrating a graph matching attention mechanism in the atractylin pharmacological target prediction model, and using the graph matching attention mechanism to perform feature matching on the molecular graph data set and the drug target graph data set;

[0082] S5. Processing the molecular graph data set and the interaction graph data set using the atractylin pharmacological target prediction model, outputting the interaction matching results between atractylin and each candidate target, and determining the potential pharmacological target according to the matching results.

[0083] In this embodiment, S1 includes the following steps:

[0084] S11. Collecting atractylin molecular structure data, which includes the atomic type, chemical bond connection relationship, molecular configuration and molecular topology information of atractylin, and constructing an atractylin molecular data set D cangshu :

[0085]

[0086] wherein M i represents the i-th atractylin molecule, N is the total number of atractylin molecular data, V i is the atomic set of atractylin molecule M i , E i is the chemical bond set inside atractylin molecule M i , A i is the topology matrix of atractylin molecule M i , and T i is the molecular configuration information of atractylin molecule M i .

[0087] S12. Standardizing the atractylin molecular data set D cangshu to generate an atractylin molecular structure standardized data set

[0088] In this embodiment, S2 includes the following steps:

[0089] S21. Constructing an atractylin molecular graph data set G according to the atractylin molecular structure standardized data set cangshu :

[0090]

[0091] wherein G iM represents the atractylodes molecule i Molecular diagram, X i Atractylodes molecule M i The node feature matrix, W i Atractylodes molecule M i The edge feature matrix;

[0092] S22. Based on the standardized dataset of atractylodesin molecular structure The topological structure information of each molecule is used to define the atractylodes molecule M. i Adjacency matrix A i Adjacency matrix A i Recording the molecule M of atractylodes i The topological connections between atoms in the middle;

[0093] S23. Based on the standardized dataset of atractylodesin molecular structure Molecular configuration information of each molecule, defining the atractylodes molecule M. i The three-dimensional coordinate matrix, the three-dimensional coordinate matrix C i c inside i,j M represents the atractylodes molecule i The three-dimensional spatial coordinates of the j-th atom;

[0094] S26. For each atractylodesin molecule M i Combined with its atomic feature matrix X i Edge feature matrix W i Adjacency matrix A i and the three-dimensional coordinate matrix C i Molecular graph G is constructed using the function Φ. i A comprehensive representation of the nodes and edges:

[0095] Φ(X i W i A i C i )={v′ i,j ,e′ i,jk |j,k=1,2,…,n i};

[0096]

[0097] e′ i,jk =σ(γ·w i,jk +δ·(‖x i,j -x i,k ||2+||c i,j -c i,k ||2));

[0098] Where Φ(·) represents the molecular graph construction function, v′ i,j M represents the atractylodes moleculei the final feature of the jth atom in the molecule, e′ i,jk denotes the atractyloside molecule M i denotes the connecting atom v i,j and v i,k the final edge feature, j, k denote the atom indices in the molecule, x i,j denotes the atractyloside molecule M i denotes the original feature vector of the jth atom in the molecule, w i,jk denotes the connecting atom v i,j and v i,k the chemical bond feature vector of v i,jk is an element of the adjacency matrix A i , a i,j = 1 if there is a chemical bond between atom v i,k and v i,jk , otherwise, ‖·‖2 denotes the Euclidean norm, denotes the element-wise product of vectors, used to fuse edge features and neighboring atom features, σ(·) is an activation function, α, β, γ, δ are learnable scaling parameters to balance the contribution of each part of features, λ is a parameter to control the degree of influence of spatial distance;

[0099] constructing the atractyloside molecule graph dataset G cangshu :

[0100] where G i = Φ(X i , W i , A i , C i ).

[0101] In this embodiment, S3 includes the following steps:

[0102] S31. Collecting drug target data containing known drug targets and related biological information, and constructing a drug target data set D target :

[0103]

[0104] where T i denotes the ith drug target, M is the total number of drug target data, R i denotes the set of key amino acid residues in the drug target T i , U i denotes the set of functional structure information in the drug target T i ;

[0105] S32. Based on the drug target data set D target , for each drug target T i = (Ri U i ) Perform graph construction to form a drug target graph dataset G target :

[0106]

[0107] in, Indicates drug target T i The graph structure representation, Indicates based on R i with U i and drug target T identified by external biological information i A set of edges representing the interactions between an internal or drug target and other biomolecules. For node R i The characteristic matrix of .

[0108] In this embodiment, S4 includes the following steps:

[0109] S41. Based on the Atractylodesin Molecular Map Dataset G cangshu Drug target map dataset G target Construct a graph matching attention mechanism A that integrates topological information and spatial geometry. GMSA In the pharmacological target prediction model F of atractylodes lancea GNN The integrated graph matching attention mechanism optimizes the feature interaction between the atractylodes molecule graph and the drug target graph.

[0110] S42. In graph matching attention mechanism A GMSA In the middle, combined with the molecular diagram G of atractylodesin i Middle atomic node v i,j Drug target map key amino acid residues r k,m Based on topological similarity and spatial conformation constraints, calculate the dynamic matching weight α. i,jk,m :

[0111]

[0112] Where, α i,jk,m G represents atractylodes molecule i Middle atom v i,j With drug target key amino acid residues r k,m The dynamic matching weights between them, f(h) i,j ,g k,m The function is the basic feature similarity calculation function. Atractylodes molecule G i and drug targets Topological similarity measurement C is the three-dimensional coordinate matrix of the artactin molecule i and the three-dimensional coordinate matrix of the drug target The spatial geometric matching degree between them;

[0113] S43. Topological structure similarity measure Using the hierarchical subgraph matching method, the artactin molecule graph G i and the drug target graph Structural similarity at different scales:

[0114]

[0115] wherein, and respectively represent the artactin molecule G i and the drug target The local subgraph structure at the l-th layer, Jaccard(·,·) is the Jaccard similarity coefficient, ω l is the importance weight of matching at different levels, and L is the number of subgraph division layers;

[0116] S44. Spatial geometric matching measure Using Laplace regularization embedding method, the position similarity of artactin molecule and drug target in three-dimensional space is calculated:

[0117]

[0118] wherein, is the spatial adjacency matrix, which represents the artactin molecule G i The atom v i,j and the key amino acid residue r k,m in the drug target whether they are within a reasonable interaction range, if the spatial distance between them is less than the set threshold τ Otherwise

[0119] S45. Combined topological structure similarity and spatial geometric matching measure Adjust the feature aggregation method of the graph matching attention mechanism to realize the optimal fusion of molecule-target interaction information:

[0120]

[0121] wherein, z ik is the matching feature vector between the artactin molecule G i and the drug target , and represents the feature splicing operation;

[0122] S46. Obtain the calculated matching feature vector z ik After mapping through the fully connected layer, output the atractylin molecule G i and the drug target :

[0123] p ik = σ(W T z ik + b);

[0124] wherein p ik represents the interaction probability between the atractylin molecule G i and the drug target , W is a trainable weight matrix, b is a bias term, and σ(·) is a sigmoid activation function.

[0125] S47. Obtain the atractylin pharmacological action target matching matrix P match :

[0126] P match = [p ik ] N×M ;

[0127] wherein P match records the matching relationship between the atractylin molecule set G cangshu and the drug target graph data set G target , N is the number of atractylin molecules, and M is the number of drug targets.

[0128] S48. Train the atractylin pharmacological action target prediction model F match using the atractylin pharmacological action target matching matrix P GNN as a supervision signal, and perform model optimization using the loss function L(F GNN ):

[0129]

[0130] wherein y ik is the true interaction label between the atractylin molecule G i and the drug target , and p ik is the predicted interaction probability.

[0131] S49. Output the trained atractylin pharmacological action target prediction model

[0132]

[0133] In this embodiment, S includes the following steps:

[0134] S51. Utilize the trained pharmacological action target prediction model of atractylin The input atractylin molecular graph data set and the interaction graph data set are processed, and the atractylin molecular G i is matched with the drug target The interaction matching feature vector z ik is calculated, and the interaction probability is mapped through the full connection prediction layer.

[0135] S52. All interaction probabilities between atractylin molecules and drug targets are combined to form an atractylin pharmacological action target matching matrix.

[0136] S53. The matching matrix is thresholded, a preset threshold τ is set, and a set of potential pharmacological action targets T pot is determined.

[0137]

[0138] Wherein, T k represents the kth drug target in the drug target graph data set G target , and τ is the set interaction probability threshold.

[0139] S54. Output the atractylin pharmacological action target matching matrix and the set of potential pharmacological action targets T pot as the final atractylin pharmacological action target prediction result.

[0140] Example 1:

[0141] On April 15, 2024, researchers in a certain biological and pharmaceutical laboratory in City A were conducting a prediction study of the pharmacological action targets of atractylin, aiming to screen out biological targets of atractylin action to explore its potential pharmacological action. The laboratory is equipped with a high-performance computing server with an NVIDIA A100 GPU, and has deployed the atractylin target prediction system of the present application based on graph neural network and graph matching attention mechanism.

[0142] The research team downloaded the latest drug target data from the ChEMBL database, including 10234 known drug targets and their related biological characteristics, and at the same time, the team obtained the molecular structure data of atractylin from the PubChem database, including its chemical bond information, three-dimensional conformation, and electronic density key parameters. All data are standardized and converted into molecular graph data structure.

[0143] During the data cleaning process, researchers found that the topological structures of 3 atractylin variants were abnormal. The team used a molecular filling algorithm to correct them, ensuring data integrity. After processing, a standardized molecular graph dataset containing 50 atractylin molecular variants was formed.

[0144] Researchers used 80% of the data for training, 10% for validation, and 10% for testing. They trained the model using the Adam optimizer for 500 rounds. During training, the system dynamically adjusted the parameters of the graph matching attention mechanism to ensure the best balance between topological features and spatial geometric features.

[0145] At the 200th round of training, researchers observed a slowdown in the loss function's decline. After adjusting the learning rate, the model training stabilized and eventually converged to the best state, with an F1-score of 0.91.

[0146] On April 25, the research team used the trained model to predict targets for atractylin molecules. The model analyzed the interaction matching relationship between 50 atractylin molecular variants and 10234 drug targets, calculated the interaction probability between each pair of molecules-targets, and generated a target prediction report. The system ultimately selected 5 high-confidence targets, including:

[0147] PTGS2 (Cyclooxygenase-2), with a predicted interaction probability of 0.92;

[0148] PPARG (Peroxisome Proliferator-Activated Receptor Gamma), with a predicted interaction probability of 0.89;

[0149] EGFR (Epidermal Growth Factor Receptor), with a predicted interaction probability of 0.85;

[0150] MAPK1 (Mitogen-Activated Protein Kinase 1), with a predicted interaction probability of 0.81;

[0151] TNF (Tumor Necrosis Factor), with a predicted interaction probability of 0.79.

[0152] To verify the reliability of the prediction results, the research team compared the prediction results of traditional molecular docking methods and random forest machine learning methods and selected experimental data for verification. The results are as follows:

[0153]

[0154]

[0155] From the experimental results, we can see that:

[0156] The method of the present application reaches 89.7% in prediction accuracy, which is 15.5% higher than the traditional molecular docking method and 10.2% higher than the random forest method. In terms of calculation time, the method of the present application only takes 6 hours to complete the prediction, while the molecular docking method takes 48 hours, which is 8 times faster. In terms of misjudgment rate, the misjudgment rate of the present application is 2.3%, which is much lower than the 8.1% of the molecular docking method, indicating that the prediction is more reliable.

[0157] The research team contacted clinical pharmacology experts to experimentally verify the two predicted target points PTGS2 and PPARG. The laboratory used PTGS2 enzyme activity inhibition experiments to test the effect of atractylin, and the results showed that the inhibition rate of atractylin on PTGS2 reached 85.3% (concentration 50 μM), further proving the effectiveness of the prediction results of the present application.

[0158] In addition, the research team conducted PPARG action experiments on a mouse model and found that atractylin can significantly improve the expression level of PPARG (P<0.01), further supporting the feasibility of its use as a potential target point. This result shows that the method of the present application is not only accurate in calculation and prediction, but also supported by experimental verification.

[0159] This embodiment verifies the advantages of the method of the present application in drug target prediction by applying it in a real biological and medical research environment. Compared with traditional prediction methods, the method of the present application based on graph neural networks and graph matching attention mechanisms is faster in calculation, more accurate in prediction, and more reliable in experimental results, providing an efficient calculation tool for the modernization of traditional Chinese medicine ingredients.

[0160] The present application integrates a graph matching attention mechanism in the graph neural network model, which realizes accurate molecule-target feature mapping by calculating the dynamic matching weight between the atractylin molecule graph and the drug target graph. The graph matching attention mechanism can dynamically weight the features of the molecule atoms and the key amino acid residues of the target point, so that the prediction model can more accurately depict the local interaction characteristics of the drug and the target point.

[0161] The present application constructs a hierarchical topological structure similarity measure and a spatial geometric matching measure, fully considering the three-dimensional spatial coordinate information of atractylin molecules and the stereo conformation of target protein structures, so that the model can more comprehensively depict the actual binding state between molecules and targets.

[0162] In the feature fusion process of molecule-target matching, the present application combines topological similarity weight, spatial geometric features and local interaction features, optimizes the interaction information aggregation through weighted feature splicing and full connection layer mapping, so that the molecule-target matching matrix is more biologically interpretable and can more effectively distinguish between high affinity targets and low affinity targets.

[0163] The above merely provides the preferred embodiment of the present application, and the protection scope of the present application is not limited thereto, and any person skilled in the art, according to the technical scheme and the inventive concept of the present application, can make equivalent replacements or changes within the technical scope disclosed by the present application, which should be covered within the protection scope of the present application.

Claims

1. A method for predicting the pharmacological target of atractylodes based on graph neural networks, characterized in that, Includes the following steps: S1. Collect atractylodes molecular structure data and standardize the atractylodes molecular structure data to generate a standardized dataset of atractylodes molecular structure. S2. Convert the standardized dataset of atractylodes molecule structure into a molecular graph dataset; S3. Collect drug target data containing known drug targets and related biological information, and construct a drug target map dataset; S4. Construct a pharmacological target prediction model for atractylodes based on a graph neural network, integrate a graph matching attention mechanism into the pharmacological target prediction model for atractylodes, and use the graph matching attention mechanism to perform feature matching between the molecular graph dataset and the drug target graph dataset; S5. The molecular graph dataset and interaction graph dataset are processed using the atractylodes pharmacological target prediction model to output the interaction matching results between atractylodes and each candidate target, and potential pharmacological targets are determined based on the matching results. S4 includes the following steps: S41. Based on the Atractylodesin Molecular Map Dataset Drug target map dataset Construct a graph matching attention mechanism that integrates topological information and spatial geometry. In the pharmacological target prediction model of atractylodes lancea The integrated graph matching attention mechanism optimizes the feature interaction between the atractylodes molecule graph and the drug target graph. S42. In the graph matching attention mechanism In the middle, combined with the molecular map of atractylodesin Middle atomic nodes Drug target map key amino acid residues Based on topological similarity and spatial conformation constraints, calculate dynamic matching weights. : ； in, Indicates atractylodes molecule Middle Atom With drug target key amino acid residues Dynamic matching weights between them The basic feature similarity calculation function, Atractylodes molecule and drug targets Topological similarity measurement The three-dimensional coordinate matrix of atractylodes molecule and the three-dimensional coordinate matrix of drug target Spatial geometric matching degree between them; S43. Topological similarity measurement The molecular graph of atractylodesin was calculated using a hierarchical subgraph matching method. Drug target map Structural similarity at different scales: ； in, and They represent atractylodesin molecules respectively and drug targets In the Local subgraph structure of the layer The Jaccard similarity coefficient. Weights are assigned to the importance of matching at different levels. The number of layers to divide the subgraph into; S44. Spatial Geometric Matching Metric The Laplacian regularized embedding method was used to calculate the positional similarity between atractylodes molecule and drug target in three-dimensional space: ； in, Let be a spatial adjacency matrix, representing atractylodes molecules. Middle Atom With drug target key amino acid residues Whether they are within a reasonable range of interaction, if the spatial distance between them is less than a set threshold. but ,otherwise ; S45. Combining topological similarity Spatial geometric matching metric Adjusting the feature aggregation method of the graph matching attention mechanism to achieve optimal fusion of molecule-target interaction information: ； in, Atractylodes molecule With drug target Matching feature vectors between them Indicates feature concatenation operation; S46. Calculate the matching feature vector. After mapping through a fully connected layer, the atractylodes molecule is output. With drug target Interaction probability between them: ； in, Indicates atractylodes molecule With drug target The probability of interaction between them For trainable weight matrix, For bias terms, It is the sigmoid activation function; S47. Obtain the target matching matrix for the pharmacological action of atractylene. : ； in, Record the collection of atractylodes molecules Drug target map dataset The matching relationship between them The number of atractylene molecules. This represents the number of drug targets. S48. Utilizing the target matching matrix of atractylene pharmacological action As a supervisory signal, the pharmacological target prediction model of atractylodes lancea. Training is performed using a loss function. Perform model optimization: ； in, Atractylodes molecule With drug target Real interaction tags between them The predicted interaction probability; S49. Output the trained atractylodesin pharmacological target prediction model. : 。 2. The method for predicting the pharmacological target of atractylodes based on graph neural networks according to claim 1, characterized in that, S1 includes the following steps: S11. Collect atractylodes molecular structure data, including atractylodes atom types, chemical bond relationships, molecular configuration, and molecular topological structure information, and construct an atractylodes atom data set. : ； in, Indicates the first One atractylodes molecule, This represents the total number of atractylodes molecules. Atractylodes molecule A collection of atoms Atractylodes molecule The internal set of chemical bonds, Atractylodes molecule The topological structure matrix, Atractylodes molecule Molecular configuration information; S12. The atractylone molecular data set Standardization was performed to generate a standardized dataset of atractylodes lancea molecular structure. .

3. The method for predicting the pharmacological target of atractylodes based on graph neural networks according to claim 2, characterized in that, S2 includes the following steps: S21. Based on the standardized dataset of atractylodesin molecular structure Constructing atractylodesin molecular map dataset : ； in, Indicates atractylodes molecule Molecular diagram, Atractylodes molecule The node feature matrix, Atractylodes molecule The edge feature matrix; S22. Based on the standardized dataset of atractylodesin molecular structure The topological structure information of each molecule is used to define the atractylodes molecule. adjacency matrix Adjacency matrix Recording atractylodesin molecules The topological connections between atoms in the middle; S23. Based on the standardized dataset of atractylodesin molecular structure Molecular configuration information of each molecule, defining the atractylodes molecule. The three-dimensional coordinate matrix, the three-dimensional coordinate matrix within Indicates atractylodes molecule The Middle The three-dimensional spatial coordinates of an atom; S26. For each atractylone molecule Combined with its atomic feature matrix Edge feature matrix Adjacency matrix and three-dimensional coordinate matrix Through function Constructing molecular diagrams A comprehensive representation of the nodes and edges: ； ； ； in, This represents the molecular graph construction function. Indicates atractylodes molecule The Middle The final characteristics of each atom, Indicates atractylodes molecule Middle connecting atoms and The final edge features, Indicates the atomic index in the molecule. Indicates atractylodes molecule The Middle The original feature vector of each atom, Indicates connecting atoms and The chemical bond feature vector, Adjacency matrix Elements, if atoms and If chemical bonds exist, then ,otherwise, Denotes the Euclidean norm. This represents the element-wise product of vectors, used to fuse edge features and neighboring atom features. For activation function, These are learnable scaling parameters used to balance the contributions of different feature groups. Parameters used to control the degree of influence of spatial distance; Constructing atractylodesin molecular map dataset : 。 4. The method for predicting the pharmacological target of atractylodes based on graph neural networks according to claim 1, characterized in that, S3 includes the following steps: S31. Collect drug target data containing known drug targets and related biological information, and construct a drug target dataset. : ； in, Indicates the first One drug target, This represents the total number of drug target data. Indicates drug target A collection of key amino acid residues, Indicates drug target A collection of functional structure information; S32. Based on the drug target data set For each drug target Graph construction is performed to form a drug target graph dataset. : ； in, Indicates drug target The graph structure representation, Indicates based on and Drug targets identified by external biological information A set of edges representing the interactions between an internal or drug target and other biomolecules. For nodes The characteristic matrix of .

5. The method for predicting the pharmacological target of atractylodes based on graph neural networks according to claim 1, characterized in that, S5 includes the following steps: S51. Utilizing a trained pharmacological target prediction model for atractylodes lancea. The input atractylodes molecule graph dataset and interaction graph dataset are processed to calculate the atractylodes molecule. With drug target Interaction matching feature vectors The interaction probability is obtained by mapping through a fully connected prediction layer. ; S52. Interaction probabilities between all atractylene molecules and drug targets Composition of the pharmacological target matching matrix of atractylodes ; S53. For the matching matrix Perform threshold determination and set a preset threshold. To identify the set of potential pharmacological targets : ； in, Represents a drug target map dataset The first in One drug target, The set interaction probability threshold; S54. Output the target matching matrix of the pharmacological action of atractylodes. and a set of potential pharmacological targets This serves as the final prediction result for the pharmacological target of atractylodes.

Images