Processing method and device for generating drug molecules based on diffusion model

Abstract

The embodiment of the invention relates to a processing method and device for drug molecule generation based on a diffusion model. The method comprises the following steps: designing a model for predicting corresponding ligand mass center coordinates and ligand atom total number according to a protein pocket conformation input by the model as a molecule prediction model, and training the molecule prediction model; taking the well-trained DiffBP model, the well-trained TargetDiff model and the well-trained D3FG model as a corresponding ligand molecule generation model, an atomic-level optimization model and a component-level optimization model; receiving a drug molecule generation task data set input by a user as a corresponding first data set; and on the basis of the molecule prediction model, the ligand molecule generation model, the atomic-level optimization model and the component-level optimization model, performing ligand molecule generation task processing according to the first data set to obtain a corresponding first task report which is fed back to a user. According to the ligand molecule design mode based on the diffusion model, the design period can be shortened, and the design efficiency can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a processing method and device for generating drug molecules based on a diffusion model. Background Art

[0002] In the field of drug development, the structure-based drug design (SBDD) method is to design ligand molecules that can effectively bind to target proteins (also known as target proteins) by analyzing the three-dimensional conformation of the target protein molecule. The conventional analysis method is to process it through molecular motion simulation and quantum chemical calculation. However, through practical application, we found that the simulation / calculation cycle of this conventional analysis method is long and the analysis efficiency is not high. Summary of the invention

[0003] The purpose of the present invention is to provide a processing method, device, electronic device and computer-readable storage medium for drug molecule generation based on a diffusion model in view of the defects of the prior art. The present invention pre-designs a model for predicting the corresponding ligand mass center coordinates and the total number of ligand atoms according to the protein pocket conformation input by the model as a molecular prediction model and trains it, and uses three types of ligand molecule generation models (DiffBP model, TargetDiff model, D3FG model) that are implemented based on the diffusion model and have been trained to be mature as corresponding ligand molecule generation models, atomic-level optimization models and component-level optimization models; then the above four types of models (molecular prediction model, ligand molecule generation model, atomic-level optimization model and component-level optimization model) are used to automatically process the batch generation task of ligand molecules, and the five types of index scores (BA, QED, SA, SIM, LPSK) of each ligand molecule are calculated using the Computer-Aided Drug Design (CCAD) tool during the task processing, and the five types of index scores of the batch task are statistically evaluated before the task processing is completed. The ligand molecule design method based on the diffusion model of the present invention is used to replace the traditional ligand molecule design method based on molecular motion simulation and quantum chemical calculation, which can shorten the molecular design cycle and improve the design efficiency.

[0004] To achieve the above-mentioned purpose, a first aspect of an embodiment of the present invention provides a method for generating drug molecules based on a diffusion model, the method comprising: A model for predicting the corresponding ligand mass center coordinates and the total number of ligand atoms according to the protein pocket conformation input by the model is designed as a molecular prediction model; and the molecular prediction model is trained based on a preset first data set; and the maturely trained DiffBP model, TargetDiff model and D3FG model are used as corresponding ligand molecule generation models, atomic-level optimization models and component-level optimization models; the DiffBP model, the TargetDiff model and the D3FG model are each a type of ligand molecule generation model based on a diffusion model; Receive a drug molecule generation task data set input by a user as the corresponding first data set; the first data set includes a first number of generated molecules, a first protein pocket conformation, a first reference ligand molecular conformation and a first indicator set; the first number of generated molecules is a positive integer; the first protein pocket conformation is a three-dimensional molecular conformation corresponding to a protein pocket; the first reference ligand molecular conformation is the three-dimensional molecular conformation of a reference ligand molecule; the first indicator set at least includes a first BA indicator threshold, a first QED indicator threshold, a first SA indicator threshold, a first SIM indicator threshold and a first LPSK indicator threshold; the three-dimensional molecular conformation is composed of multiple atomic objects; the atomic object includes an atomic index, an atomic type and an atomic coordinate; Based on the molecular prediction model, the ligand molecule generation model, the atomic level optimization model and the component level optimization model, the ligand molecule generation task is processed according to the first data set to obtain a corresponding first task report which is fed back to the user.

[0005] Preferably, the model input of the molecular prediction model is a protein pocket conformation A0, and the model output is a parameter prediction vector P; the molecular prediction model is used to predict the center of mass coordinates and the total number of atoms of the ligand molecule corresponding to the protein pocket conformation A0 of the model input and output the corresponding parameter prediction vector P; the protein pocket conformation A0 is the three-dimensional molecular conformation corresponding to a protein pocket; the parameter prediction vector P includes the predicted center of mass coordinates and the predicted total number of atoms; The model input of the DiffBP model includes protein pocket conformation A1, ligand molecule input conformation B in,1 and the center of mass coordinates of the ligand molecule, the model output is a corresponding ligand molecule output conformation B out,1 ; The DiffBP model is used to input the ligand molecule input into the model conformation B in,1As a scrambled diffusion conformation, and according to the protein pocket conformation A1 and the center of mass coordinates of the ligand molecule input by the model, each atom type and coordinate in the current diffusion conformation is iteratively denoised to obtain the corresponding denoised ligand conformation, and the currently obtained denoised ligand conformation is used as the corresponding ligand molecule output conformation B out,1 Output; wherein the protein pocket conformation A1 is the three-dimensional molecular conformation corresponding to a protein pocket; the ligand molecule input conformation B in,1 and the ligand molecule outputs conformation B out,1 The three-dimensional molecular conformation of a ligand molecule bound to the current protein pocket before and after being processed by the DiffBP model; The model input of the TargetDiff model includes the protein pocket conformation A2 and the ligand molecule input conformation B in,2 , the model output is a corresponding ligand molecule output conformation B out,2 ; The TargetDiff model is used to input the ligand molecule input into the model conformation B in,2 As a scrambled diffusion conformation, and according to the protein pocket conformation A2 input by the model, each atom type and coordinate in the current diffusion conformation is iteratively denoised to obtain the corresponding denoised ligand conformation, and the currently obtained denoised ligand conformation is used as the corresponding ligand molecule output conformation B out,2 Output; wherein the protein pocket conformation A2 is the three-dimensional molecular conformation corresponding to a protein pocket; the ligand molecule input conformation B in,2 and the ligand molecule outputs conformation B out,2 The three-dimensional molecular conformation of a ligand molecule bound to the current protein pocket before and after being processed by the TargetDiff model; The model input of the D3FG model includes the protein pocket conformation A3 and the ligand molecule input conformation B in,3 , the model output is a corresponding ligand molecule output conformation B out,3 The D3FG model is used to identify the corresponding amino acid components of the protein pocket conformation A3 input by the model to obtain the corresponding amino acid component set, and to input the conformation B of the ligand molecule input by the model. in,2 The corresponding functional group and connector component are identified to obtain the corresponding functional group component set and connector component set, and the three-dimensional conformations of each functional group component and each connector component in the functional group component set and the connector component set are iteratively denoised according to the amino acid component set to obtain the corresponding denoised ligand conformation, and the currently obtained denoised ligand conformation is used as the corresponding ligand molecule output conformation B out,3Output; wherein the protein pocket conformation A3 is the three-dimensional molecular conformation corresponding to a protein pocket; the ligand molecule input conformation B in,3 and the ligand molecule outputs conformation B out,3 It is the three-dimensional molecular conformation of a ligand molecule bound to the current protein pocket before and after being processed by the D3FG model.

[0006] Preferably, the first BA index threshold is an affinity evaluation index threshold; The first QED indicator threshold is a quantitative evaluation indicator threshold for drug properties; The first SA indicator threshold is a synthesizability evaluation indicator threshold; The first SIM index threshold is a structural similarity evaluation index threshold; The first LPSK indicator threshold is an evaluation indicator threshold of the Lipinski five rules.

[0007] Preferably, the first data set includes a plurality of first data records; the first data record includes a first training conformation, a first pocket atom index set and a first ligand atom index set; the first training conformation is the three-dimensional molecular conformation corresponding to a complex molecule, and the complex molecule is composed of a protein pocket and a ligand molecule; the first pocket atom index set and the first ligand atom index set are each composed of a plurality of atom indices in the first training conformation; the first pocket atom index set is the atom index set of the protein pocket in the current complex molecule; the first ligand atom index set is the atom index set of the ligand molecule in the current complex molecule; The first task report includes a first ligand molecule list and a first indicator analysis list; The first ligand molecule list is composed of the first task quantity of first ligand molecule records; The first ligand molecule record includes a first ligand conformation field, a first BA indicator field, a first QED indicator field, a first SA indicator field, a first SIM indicator field and a first LPSK indicator field; The first indicator analysis list includes a plurality of first indicator analysis records; The first indicator analysis record includes a first indicator type field, a first indicator threshold field, a first indicator statistics field and a first indicator compliance status field; the first indicator type field includes BA type, QED type, SA type, SIM type and LPSK type; the first indicator threshold field is the first BA indicator threshold, the first QED indicator threshold, the first SA indicator threshold, the first SIM indicator threshold or the first LPSK indicator threshold corresponding to the first indicator type field; the first indicator statistics field is the indicator statistical parameter corresponding to the first indicator type field; the first indicator compliance status field includes met and not met.

[0008] Preferably, the training of the molecular prediction model based on a preset first data set specifically includes: Step 51, randomly splitting the first data set based on a preset first split ratio to obtain two corresponding sub-data sets recorded as a corresponding first training set and a first evaluation set; Wherein, both the first training set and the first evaluation set are composed of a plurality of the first data records; the ratio of the total number of records in the first training set to the total number of records in the first evaluation set satisfies the first segmentation ratio; Step 52, extracting the first first data record of the first training set as the corresponding current training record; Step 53, taking the first training conformation, the first pocket atom index set and the first ligand atom index set of the current training record as the corresponding current training conformation, the first atom index set and the second atom index set; and performing corresponding protein pocket and ligand molecule conformation extraction processing on the current training conformation based on the first and second atom index sets to obtain the corresponding first pocket conformation and first ligand conformation; and performing mean calculation on the atomic coordinates of all the atomic objects in the first ligand conformation and using the calculation result as the corresponding first label centroid coordinates; and performing statistics on the total number of atomic objects in the first ligand conformation and using the statistical result as the corresponding first label atom total number; and forming the corresponding first parameter label vector by the first label centroid coordinates and the first label atom total number; Step 54, using the molecular prediction model to predict the center of mass coordinates and the total number of atoms of the ligand molecule corresponding to the first pocket conformation to obtain a corresponding first parameter prediction vector; Step 55, bringing the first parameter label vector and the first parameter prediction vector into a preset first model loss function to calculate and obtain a corresponding first loss value; Wherein, the first model loss function includes at least an L1 loss function and an L2 loss function; Step 56, identifying whether the first loss value satisfies a preset first loss value range; if the first loss value satisfies the first loss value range, identifying whether the current training record is the last first data record of the first training set, if so, turning to step 57, if not, extracting the next first data record of the first training set as the new current training record and returning to step 53; if the first loss value does not satisfy the first loss value range, performing a round of modulation on the model parameters of the molecular prediction model in a direction to minimize the first model loss function based on a preset first model optimizer, and returning to step 54 at the end of this round of modulation; Wherein, the first model optimizer includes at least an SGD optimizer and an ADAM optimizer; Step 57, perform a round of traversal on all the first data records of the first evaluation set; and in this round of traversal, use the first data record currently traversed as the corresponding current evaluation record; and use the first training conformation, the first pocket atom index set and the first ligand atom index set of the current evaluation record as the corresponding current training conformation, the third atom index set and the fourth atom index set; and perform corresponding protein pocket and ligand molecule conformation extraction processing on the current training conformation based on the third and fourth atom index sets to obtain the corresponding second pocket conformation and the second ligand conformation; and perform mean calculation on the atomic coordinates of all the atomic objects in the second ligand conformation and use the calculation result as the The corresponding second label centroid coordinates; and the total number of atomic objects of the second ligand conformation is counted and the statistical result is used as the corresponding second label atom total number; and the second label centroid coordinates and the second label atom total number form a corresponding second parameter label vector; and the molecular prediction model predicts the centroid coordinates and the total number of atoms of the ligand molecule corresponding to the second pocket conformation to obtain the corresponding second parameter prediction vector; and the second parameter label vector obtained this time and the corresponding second parameter prediction vector form a corresponding first label-prediction pair; and at the end of this round of traversal, all the first prediction-label pairs obtained are brought into the preset first model evaluation function to calculate and obtain the corresponding first evaluation value; Wherein, the first model evaluation function at least includes an RMSE function; Step 58, identifying whether the first evaluation value satisfies a preset first evaluation value range; if the first evaluation value does not satisfy the first evaluation value range, returning to step 51 to continue training; if the first evaluation value satisfies the first evaluation value range, stopping training and confirming that the model training of the molecular prediction model is completed.

[0009] Preferably, the processing of the ligand molecule generation task based on the molecular prediction model, the ligand molecule generation model, the atomic-level optimization model and the component-level optimization model according to the first data set to obtain a corresponding first task report and to feed back to the user specifically includes: Step 6-1, extracting the corresponding first generated molecule quantity, the first protein pocket conformation, the first reference ligand molecule conformation and the first indicator set from the first data set; and initializing the corresponding first counter to 1; Step 6-2, predicting the center of mass coordinates and the total number of atoms of the ligand molecule corresponding to the first protein pocket conformation by the molecular prediction model to obtain a corresponding current parameter prediction vector; and using the predicted center of mass coordinates and the predicted total number of atoms in the parameter prediction vector as the corresponding first center of mass coordinates and the first total number of atoms; Step 6-3, initializing a ligand molecular conformation as a corresponding first ligand molecular conformation based on the first mass center coordinates and the first total number of atoms; wherein the total number of atomic objects in the first ligand molecular conformation matches the first total number of atoms; all of the atomic types in the first ligand molecular conformation are preset invalid atomic types; all of the atomic coordinates in the first ligand molecular conformation are random coordinates satisfying Gaussian distribution, and the coordinate mean of all of the atomic coordinates in the first ligand molecular conformation matches the first centroid coordinates; a line between any two of the atomic objects in the first ligand molecular conformation will not intersect a line between any two of the corresponding atomic objects in the first protein pocket conformation; Step 6-4, inputting the first protein pocket conformation, the first center of mass coordinates and the first ligand molecule conformation into the ligand molecule generation model; the ligand molecule generation model regards the first ligand molecule conformation as a scrambled post-diffusion conformation, and iteratively denoises each atom type and coordinate in the current post-diffusion conformation using the first protein pocket conformation and the first center of mass coordinates to obtain a corresponding denoised ligand conformation, and outputs the currently obtained denoised ligand conformation as the corresponding second ligand molecule conformation; Step 6-5, inputting the first protein pocket conformation and the second ligand molecular conformation into the atomic-level optimization model; the atomic-level optimization model regards the second ligand molecular conformation as a scrambled post-diffusion conformation, and performs iterative denoising on each atom type and coordinate in the current post-diffusion conformation according to the first protein pocket conformation to obtain a corresponding denoised ligand conformation, and outputs the currently obtained denoised ligand conformation as the corresponding third ligand molecular conformation; Step 6-6, inputting the first protein pocket conformation and the third ligand molecular conformation into the component-level optimization model; the component-level optimization model performs corresponding amino acid component identification on the first protein pocket conformation to obtain a corresponding first amino acid component set, and performs corresponding functional group and connector component identification on the third ligand molecular conformation to obtain a corresponding first functional group component set and a first connector component set, and performs iterative denoising on the three-dimensional conformations of each functional group component in the first functional group component set and the first connector component set and each connector component according to the first amino acid component set to obtain a corresponding denoised ligand conformation, and outputs the currently obtained denoised ligand conformation as the corresponding fourth ligand molecular conformation; Step 6-7, using a preset computer-aided drug design tool, the binding affinity score between the first protein pocket conformation and the fourth ligand molecular conformation is calculated to obtain a corresponding first index score; and the drug property quantitative evaluation score of the fourth ligand molecular conformation is calculated to obtain a corresponding second index score; and the synthesizability score of the fourth ligand molecular conformation is calculated to obtain a corresponding third index score; and the structural similarity score between the fourth ligand molecular conformation and the first reference ligand molecular conformation is calculated to obtain a corresponding fourth index score; and whether the fourth ligand molecular conformation satisfies the Lipinski five rules is identified, and if so, the corresponding fifth index score is set to 1, and if not, the corresponding fifth index score is set to 0; Wherein, the computer-aided drug design tools include at least DOCK, AutoDock, GOLD, FlexX, Glide, ICM, RDkits, and Discovery Studio; Step 6-8, using the fourth ligand molecule conformation, the first index score, the second index score, the third index score, the fourth index score and the fifth index score as the corresponding first ligand conformation field, the first BA score field, the first QED score field, the first SA score field, the first SIM score field and the first LPSK score field to form a corresponding first ligand molecule record; Step 6-9, adding 1 to the first counter; and identifying whether the first counter after adding 1 exceeds the corresponding first generated molecule quantity; if not, returning to step 6-3; if it has exceeded, forming the corresponding first ligand molecule list from the first ligand molecule records of the first generated molecule quantity, and going to step 6-10; Step 6-10, bring all the first BA score fields in the first ligand molecule list into a preset first evaluation function to calculate the corresponding first index evaluation value, bring all the first QED score fields into a preset second evaluation function to calculate the corresponding second index evaluation value, bring all the first SA score fields into a preset third evaluation function to calculate the corresponding third index evaluation value, and bring all the first SIM score fields into a preset fourth evaluation function to calculate the corresponding fourth index evaluation value; and count the total number of the first ligand molecule records whose first LPSK score field in the first ligand molecule list is 1 to obtain the corresponding first total number, and calculate the corresponding fifth index evaluation value based on the first total number and the first generated molecule number = (first total number / first generated molecule number) × 100%; Wherein, the first, second, third and fourth evaluation functions are all implemented based on the RMSE function; Step 6-11, taking the first, second, third, fourth and fifth indicator evaluation values ​​as the corresponding current indicator evaluation values ​​in sequence; and setting a corresponding current indicator type as the BA type, QED type, SA type, SIM type or LPSK type corresponding to the current indicator evaluation value; and taking the first BA indicator threshold, the first QED indicator threshold, the first SA indicator threshold, the first SIM indicator threshold or the first LPSK indicator threshold corresponding to the current indicator type in the first indicator set as the corresponding current indicator threshold; and identifying whether the current indicator evaluation value is lower than the corresponding current indicator threshold, if so, setting a corresponding current indicator compliance state as unsatisfactory, otherwise setting a corresponding current indicator compliance state as satisfactory; and taking the current indicator type, the current indicator threshold, the current indicator evaluation value and the current indicator compliance state as the corresponding first indicator type field, the first indicator threshold field, the first indicator statistics field and the first indicator compliance state field to form a corresponding first indicator analysis record; and forming a corresponding first indicator analysis list from all the obtained first indicator analysis records; Step 6-12: The first task report corresponding to the obtained first ligand molecule list and the first indicator analysis list is fed back to the user.

[0010] A second aspect of an embodiment of the present invention provides a device for implementing the processing method for drug molecule generation based on a diffusion model as described in the first aspect, the device comprising: a model preparation module, a task data receiving module and a task processing module; The model preparation module is used to design a model for predicting the corresponding ligand mass center coordinates and the total number of ligand atoms according to the protein pocket conformation input by the model as a molecular prediction model; and to train the molecular prediction model based on a preset first data set; and to use the maturely trained DiffBP model, TargetDiff model and D3FG model as the corresponding ligand molecule generation model, atomic-level optimization model and component-level optimization model; the DiffBP model, the TargetDiff model and the D3FG model are each a type of ligand molecule generation model based on a diffusion model; The task data receiving module is used to receive the drug molecule generation task data set input by the user as the corresponding first data set; the first data set includes the first generated molecule quantity, the first protein pocket conformation, the first reference ligand molecular conformation and the first indicator set; the first generated molecule quantity is a positive integer; the first protein pocket conformation is a three-dimensional molecular conformation corresponding to a protein pocket; the first reference ligand molecular conformation is the three-dimensional molecular conformation of a reference ligand molecule; the first indicator set at least includes a first BA indicator threshold, a first QED indicator threshold, a first SA indicator threshold, a first SIM indicator threshold and a first LPSK indicator threshold; the three-dimensional molecular conformation is composed of multiple atomic objects; the atomic object includes atomic index, atomic type and atomic coordinates; The task processing module is used to perform ligand molecule generation task processing according to the first data set based on the molecular prediction model, the ligand molecule generation model, the atomic level optimization model and the component level optimization model to obtain a corresponding first task report and provide feedback to the user.

[0011] A third aspect of an embodiment of the present invention provides an electronic device, including: a memory, a processor, and a transceiver; The processor is used to couple with the memory, read and execute instructions in the memory, so as to implement the method steps described in the first aspect above; The transceiver is coupled to the processor, and the processor controls the transceiver to send and receive messages.

[0012] A fourth aspect of an embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions. When the computer instructions are executed by a computer, the computer executes the instructions of the method described in the first aspect above.

[0013] The embodiment of the present invention provides a processing method, device, electronic device and computer-readable storage medium for drug molecule generation based on a diffusion model. As can be seen from the above content, the embodiment of the present invention pre-designs a model for predicting the corresponding ligand mass center coordinates and the total number of ligand atoms according to the protein pocket conformation input by the model as a molecular prediction model and trains it, and uses three types of ligand molecule generation models (DiffBP model, TargetDiff model, D3FG model) that are implemented based on the diffusion model and have been trained to be mature as corresponding ligand molecule generation models, atomic-level optimization models and component-level optimization models; then the above four types of models (molecular prediction model, ligand molecule generation model, atomic-level optimization model and component-level optimization model) are used to automatically process the batch generation task of ligand molecules, and the five types of index scores (BA, QED, SA, SIM, LPSK) of each ligand molecule are calculated during the task processing using a computer-aided drug design tool, and the five types of index scores of the batch task are statistically evaluated before the task processing is completed. The ligand molecule design method based on the diffusion model of the embodiment of the present invention is used to replace the traditional ligand molecule design method based on molecular motion simulation and quantum chemical calculation, which not only shortens the molecular design cycle but also improves the design efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 A schematic diagram of a processing method for generating drug molecules based on a diffusion model provided in Example 1 of the present invention; Figure 2 A module structure diagram of a processing device for generating drug molecules based on a diffusion model provided in the second embodiment of the present invention; Figure 3 A schematic diagram of the structure of an electronic device provided in Embodiment 3 of the present invention. DETAILED DESCRIPTION

[0015] In order to make the purpose, technical solution and advantages of the present invention clearer, the present invention will be further described in detail below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0016] Embodiment 1 of the present invention provides a method for generating drug molecules based on a diffusion model, such as Figure 1 A schematic diagram of a method for generating drug molecules based on a diffusion model provided in the first embodiment of the present invention is shown. The method mainly includes the following steps: Step 1, design a model for predicting the corresponding ligand mass center coordinates and the total number of ligand atoms according to the protein pocket conformation input by the model as a molecular prediction model; and train the molecular prediction model based on the preset first data set; and use the maturely trained DiffBP model, TargetDiff model and D3FG model as the corresponding ligand molecule generation model, atomic-level optimization model and component-level optimization model; Specifically, the method comprises: step 11, designing a model for predicting the corresponding ligand mass center coordinates and the total number of ligand atoms according to the protein pocket conformation input by the model as a molecular prediction model; Here, the model input of the molecular prediction model designed in the embodiment of the present invention is a protein pocket conformation A0, and the model output is a parameter prediction vector P; the molecular prediction model is used to predict the center of mass coordinates and the total number of atoms of the ligand molecule corresponding to the protein pocket conformation A0 input by the model and output the corresponding parameter prediction vector P; Wherein, the protein pocket conformation A0 is a three-dimensional molecular conformation corresponding to a protein pocket; the parameter prediction vector P includes the predicted center of mass coordinates and the predicted total number of atoms; it should be noted that the three-dimensional molecular conformation mentioned in the embodiment of the present invention is composed of multiple atomic objects, each of which includes an atomic index, an atomic type and an atomic coordinate; It should be noted that the functional module structure of the molecular prediction model of the embodiment of the present invention includes a data preprocessing module, a coding module and a prediction head module; wherein: 1) The input end of the data preprocessing module is connected to the input end of the model, and the output end is connected to the input end of the encoding module; the data preprocessing module is used to take each atomic object of the input protein pocket conformation A0 as the corresponding node element, and form a corresponding node set by all the obtained node elements; and use the preset quantum chemical calculation tools (such as Gaussian, GAMESS, CP2K, XTB, MOLPRO, Q-Chem, ADF, etc.) to simulate and calculate all chemical bonds of the protein pocket conformation A0 based on all atomic objects of the protein pocket conformation A0 to obtain a corresponding chemical bond set, which includes multiple chemical bond objects, each chemical bond object includes a chemical bond index, a chemical bond type and a chemical bond atom pair index group, and the chemical bond atom pair index group is composed of the atomic indexes of two atomic objects corresponding to the current chemical bond object; and each chemical bond object in the chemical bond set is taken as a corresponding edge element, and a corresponding edge set is formed by all the obtained edge elements; and a corresponding three-dimensional graph data is formed by the obtained node set and edge set, which is recorded as the corresponding protein pocket graph and sent to the encoding module; 2) The encoding module is composed of multiple EGNN models connected in sequence; the output end of the encoding module is connected to the input end of the prediction head module; the encoding module is used to encode the graph features of the received protein pocket graph to obtain the corresponding encoding tensor and send it to the prediction head module; 3) The prediction head module includes a first MLP model, a second MLP model and a prediction output unit; the input end of the first MLP model is connected to the output end of the encoding module, and the output end is connected to the first input end of the prediction output unit; the input end of the second MLP model is connected to the output end of the encoding module, and the output end is connected to the second input end of the prediction output unit; the output end of the prediction output unit is connected to the output end of the model; the first MLP model is used to predict the corresponding ligand molecule centroid coordinates according to the received encoding tensor, obtain the corresponding predicted centroid coordinates and send them to the prediction output unit; the first MLP model is used to predict the corresponding total number of ligand molecule atoms according to the received encoding tensor, obtain the corresponding predicted total number of atoms and send them to the prediction output unit; the prediction output unit forms a corresponding parameter prediction vector P composed of the received predicted centroid coordinates and the predicted total number of atoms and outputs it; Step 12, training the molecular prediction model based on a preset first data set; Wherein, the first data set includes multiple first data records; the first data record includes a first training conformation, a first pocket atom index set and a first ligand atom index set; the first training conformation is a three-dimensional molecular conformation corresponding to a complex molecule, and the complex molecule is composed of a protein pocket and a ligand molecule; the first pocket atom index set and the first ligand atom index set are each composed of multiple atomic indexes in the first training conformation; the first pocket atom index set is the atom index set of the protein pocket in the current complex molecule; the first ligand atom index set is the atom index set of the ligand molecule in the current complex molecule; Specifically, it includes: step 121, randomly splitting the first data set based on a preset first split ratio to obtain two corresponding sub-data sets recorded as a corresponding first training set and a first evaluation set; The first split ratio is a preset ratio value, such as 8:2; the first training set and the first evaluation set are both composed of a plurality of first data records; the ratio of the total number of records in the first training set to the total number of records in the first evaluation set satisfies the first split ratio; Step 122, extracting the first first data record of the first training set as the corresponding current training record; Step 123, taking the first training conformation, the first pocket atom index set and the first ligand atom index set of the current training record as the corresponding current training conformation, the first atom index set and the second atom index set; and performing corresponding protein pocket and ligand molecule conformation extraction processing on the current training conformation based on the first and second atom index sets to obtain the corresponding first pocket conformation and first ligand conformation; and performing mean calculation on the atomic coordinates of all atomic objects in the first ligand conformation and using the calculation result as the corresponding first label centroid coordinates; and performing statistics on the total number of atomic objects in the first ligand conformation and using the statistical result as the corresponding first label atom total number; and forming the corresponding first parameter label vector from the first label centroid coordinates and the first label atom total number; Step 124, using a molecular prediction model to predict the center of mass coordinates and the total number of atoms of the ligand molecule corresponding to the first pocket conformation to obtain a corresponding first parameter prediction vector; Step 125, bringing the first parameter label vector and the first parameter prediction vector into a preset first model loss function to calculate and obtain a corresponding first loss value; Wherein, the first model loss function includes at least an L1 loss function and an L2 loss function; Step 126, identifying whether the first loss value satisfies a preset first loss value range; if the first loss value satisfies the first loss value range, identifying whether the current training record is the last first data record of the first training set, if so, turning to step 127, if not, extracting the next first data record of the first training set as a new current training record and returning to step 123; if the first loss value does not satisfy the first loss value range, performing a round of modulation on the model parameters of the molecular prediction model in a direction to minimize the first model loss function based on a preset first model optimizer, and returning to step 124 at the end of this round of modulation; The first loss value range is a preset loss value range; the first model optimizer includes at least an SGD optimizer and an ADAM optimizer; Step 127, perform a round of traversal on all the first data records of the first evaluation set; and in this round of traversal, use the currently traversed first data record as the corresponding current evaluation record; and use the first training conformation, the first pocket atom index set and the first ligand atom index set of the current evaluation record as the corresponding current training conformation, the third atom index set and the fourth atom index set; and perform corresponding protein pocket and ligand molecule conformation extraction processing on the current training conformation based on the third and fourth atom index sets to obtain the corresponding second pocket conformation and the second ligand conformation; and perform mean calculation on the atomic coordinates of all atomic objects in the second ligand conformation and use the calculation result as the corresponding first The second label centroid coordinates; the total number of atomic objects of the second ligand conformation is counted and the statistical result is used as the corresponding second label atom total number; the second label centroid coordinates and the second label atom total number form a corresponding second parameter label vector; the molecular prediction model predicts the centroid coordinates and the total number of atoms of the ligand molecule corresponding to the second pocket conformation to obtain the corresponding second parameter prediction vector; the second parameter label vector obtained this time and the corresponding second parameter prediction vector form a corresponding first label-prediction pair; and at the end of this round of traversal, all the first prediction-label pairs obtained are brought into the preset first model evaluation function to calculate the corresponding first evaluation value; Wherein, the first model evaluation function at least includes the RMSE function; Step 128, identifying whether the first evaluation value satisfies a preset first evaluation value range; if the first evaluation value does not satisfy the first evaluation value range, returning to step 121 to continue training; if the first evaluation value satisfies the first evaluation value range, stopping training and confirming that the model training of the molecular prediction model is completed; Wherein, the first evaluation value range is a preset evaluation value range; Step 13, and use the maturely trained DiffBP model, TargetDiff model and D3FG model as the corresponding ligand molecule generation model, atomic level optimization model and component level optimization model; Here, the DiffBP model, TargetDiff model and D3FG model used in the embodiment of the present invention are each a type of ligand molecule generation model implemented based on a diffusion model; Among them, the forward scrambling / reverse denoising principle of the DiffBP model, the model training method, and the processing flow of generating corresponding ligand molecules for specified protein pockets based on the model have been disclosed in the technical document A "DiffBP: Generative Diffusion of 3D Molecules for Target Protein Binding". Specific model information can be obtained by reading the document, and no further details will be given here. In short, the model input of the DiffBP model used in the embodiment of the present invention includes protein pocket conformation A1, ligand molecule input conformation B in,1 and the center of mass coordinates of the ligand molecule, the model output is a corresponding ligand molecule output conformation B out,1 ; The DiffBP model is used to input the ligand molecule of the model into conformation B in,1 As a scrambled diffusion conformation, the atom types and coordinates in the current diffusion conformation are iteratively denoised according to the protein pocket conformation A1 and the center of mass coordinates of the ligand molecule input by the model to obtain the corresponding denoised ligand conformation, and the current denoised ligand conformation is used as the corresponding ligand molecule output conformation B out,1 Output; Among them, protein pocket conformation A1 is the three-dimensional molecular conformation corresponding to a protein pocket; ligand molecule input conformation B in,1 and the ligand molecule output conformation B out,1 It is the three-dimensional molecular conformation of a ligand molecule bound to the current protein pocket before and after DiffBP model processing; The forward scrambling / reverse denoising principle of the TargetDiff model, the model training method, and the processing flow of generating corresponding ligand molecules for specified protein pockets based on the model have been disclosed in the technical document B "3D EQUIVARIANTDIFFUSION FOR Target-AWARE MOLECULE GENERATION AND AFFINITY PREDICTION". The specific model information can be obtained by reading the document, and no further details will be given here. In short, the model input of the TargetDiff model used in the embodiment of the present invention includes the protein pocket conformation A2 and the ligand molecule input conformation B in,2 , the model output is a corresponding ligand molecule output conformation B out,2 ; The TargetDiff model is used to input the ligand molecule of the model into conformation B in,2 As a scrambled diffusion conformation, the atom types and coordinates in the current diffusion conformation are iteratively denoised according to the protein pocket conformation A2 input by the model to obtain the corresponding denoised ligand conformation, and the current denoised ligand conformation is used as the corresponding ligand molecule output conformation B out,2Output; Among them, protein pocket conformation A2 is the three-dimensional molecular conformation corresponding to a protein pocket; ligand molecule input conformation B in,2 and the ligand molecule output conformation B out,2 It is the three-dimensional molecular conformation of a ligand molecule bound to the current protein pocket before and after TargetDiff model processing; The component identification method, forward scrambling / reverse denoising principle, model training method and the processing flow of generating corresponding ligand molecules for specified protein pockets based on the model of the D3FG model have been disclosed in the technical document C "Functional-Group-Based Diffusion for Pocket-Specific Molecule Generation and Elaboration". The specific model information can be obtained by reading the document, and no further details will be given here. In short, the model input of the D3FG model used in the embodiment of the present invention includes the protein pocket conformation A3 and the ligand molecule input conformation B in,3 , the model output is a corresponding ligand molecule output conformation B out,3 The D3FG model is used to identify the corresponding amino acid components of the protein pocket conformation A3 input by the model to obtain the corresponding amino acid component set, and to input the conformation B of the ligand molecule input by the model. in,2 The corresponding functional group and connector components are identified to obtain the corresponding functional group component set and connector component set, and the three-dimensional conformations of each functional group component and each connector component in the functional group component set and the connector component set are iteratively denoised according to the amino acid component set to obtain the corresponding denoised ligand conformation, and the currently obtained denoised ligand conformation is used as the corresponding ligand molecule output conformation B out,3 Output; Among them, protein pocket conformation A3 is the three-dimensional molecular conformation corresponding to a protein pocket; ligand molecule input conformation B in,3 and the ligand molecule output conformation B out,3 It is the three-dimensional molecular conformation of a ligand molecule bound to the current protein pocket before and after D3FG model processing; It should also be noted that although these three types of models (DiffBP model, TargetDiff model and D3FG model) can handle the task of ligand molecule generation, they have different advantages; We call the ligand molecule generation task of generating a solid molecular conformation (i.e., a three-dimensional molecular conformation with a known total number of atomic objects but unknown types and coordinates of each atomic object) based on an empty molecular conformation (i.e., a three-dimensional molecular conformation with a known total number of atomic objects but unknown types and coordinates of each atomic object) a de novo molecular generation task, the ligand molecule generation task of further optimizing the types and coordinates of each atom in the solid molecular conformation of a ligand molecule a first-class lead molecule optimization task, and the ligand molecule generation task of further optimizing each atomic motif (amino acid, functional group, connector) in the solid molecular conformation of a ligand molecule a second-class lead molecule optimization task; Through public technical literature and practical verification, we found that: 1) The DiffBP model has the best performance in processing de novo molecule generation tasks, but its performance in processing Class I lead molecule generation tasks is average, and it cannot process Class II lead molecule generation tasks; 2) The TargetDiff model has the best performance in processing Class I lead molecule generation tasks, but its performance in processing de novo molecule generation tasks is average, and it cannot process Class II lead molecule generation tasks; 3) The D3FG model has the best performance in processing Class II lead molecule generation tasks, but its performance in processing de novo molecule generation tasks is average, and it cannot process Class I lead molecule generation tasks; Therefore, the embodiment of the present invention deliberately combines these three types of models together, first using the DiffBP model as the corresponding ligand molecule generation model to process the de novo molecule generation task, then using the TargetDiff model as the corresponding atomic-level optimization model to process the first type of lead molecule optimization task, and then using the D3FG model as the corresponding component-level optimization model to process the second type of lead molecule optimization task; the ligand molecule structure thus obtained is more reasonable, more functional, and more compatible with existing domain knowledge than the structure obtained using any of the above three types of models; Step 2: receiving a drug molecule generation task data set input by a user as a corresponding first data set.

[0017] Here, the first data set of the embodiment of the present invention includes a first generated molecule quantity, a first protein pocket conformation, a first reference ligand molecule conformation, and a first indicator set; Among them, the first number of generated molecules is a positive integer; the first protein pocket conformation is a three-dimensional molecular conformation corresponding to a protein pocket; the first reference ligand molecular conformation is a three-dimensional molecular conformation of a reference ligand molecule; the first indicator set at least includes a first BA indicator threshold, a first QED indicator threshold, a first SA indicator threshold, a first SIM indicator threshold and a first LPSK indicator threshold; The meaning of each indicator threshold in the first indicator set and the calculation method of the corresponding indicator evaluation value are as follows: 1) The first BA indicator threshold is the Binding Affinity (BA) evaluation indicator threshold; the BA indicator evaluation value corresponding to the evaluation indicator threshold is calculated by calculating the root mean square error (RMSE) of multiple (for example, number m) BA indicator scores, that is: , 1≤sample index i≤m,x max is the preset BA indicator maximum score, x i is the i-th BA index score; for each BA index score, if the affinity between the protein pocket and the corresponding ligand molecule is higher, the score is higher, and vice versa, if the affinity is lower, the score is lower; 2) The first QED indicator threshold is the quantitative estimate of drug-likeness (QED) indicator threshold; the QED indicator evaluation value corresponding to this evaluation indicator threshold is calculated by RMSE calculation of multiple QED indicator scores, which is similar to the calculation method of BA indicator evaluation value, except that x max It becomes another preset maximum score of the QED indicator; for each QED indicator score, if the drug property of the corresponding ligand molecule is better, the score is higher, and vice versa, if the drug property is worse, the score is lower; 3) The first SA indicator threshold is the synthetic accessibility (SA) evaluation indicator threshold; the calculation method of the SA indicator evaluation value corresponding to this evaluation indicator threshold is to calculate the RMSE of multiple SA indicator scores, which is similar to the calculation method of the BA indicator evaluation value, except that x max becomes another preset SA index maximum score; for each SA index score, if the synthesis difficulty of the corresponding ligand molecule is smaller, the score is higher, and vice versa, if the synthesis difficulty is greater, the score is lower; 4) The first SIM indicator threshold is the structural similarity (SIM) evaluation indicator threshold; the calculation method of the SIM indicator evaluation value corresponding to this evaluation indicator threshold is to calculate the RMSE of multiple SIM indicator scores, which is similar to the calculation method of the BA indicator evaluation value, except that x max becomes another preset SIM index maximum score; for each SIM index score, if the corresponding ligand molecule has a higher structural similarity with the reference ligand molecule, the score is higher, and conversely, if the structural similarity with the reference ligand molecule is lower, the score is lower; 5) The first LPSK indicator threshold is the evaluation indicator threshold of Lipinski's five rules; the calculation method of the LPSK indicator evaluation value corresponding to the evaluation indicator threshold is: whether the generated multiple (for example, number m) ligand molecules satisfy the Lipinski five rules is checked one by one, and the LPSK indicator score of the ligand molecules that satisfy the Lipinski five rules is set to 1, and the LPSK indicator score of the ligand molecules that do not satisfy the Lipinski five rules is set to 0, and the total number of LPSK indicator scores with a value of 1 is counted to obtain the statistical number n, and the LPSK indicator evaluation value is calculated based on n and m = (n / m) × 100%; for each LPSK indicator score, if the molecular structure quality of the corresponding ligand molecule is higher (that is, it satisfies Lipinski's five rules), the score is 1, otherwise, if the molecular structure quality is worse (that is, it does not satisfy Lipinski's five rules), the score is 0.

[0018] Step 3, based on the molecular prediction model, the ligand molecule generation model, the atomic level optimization model and the component level optimization model, the ligand molecule generation task is processed according to the first data set to obtain a corresponding first task report and provide feedback to the user; The first task report includes a first ligand molecule list and a first indicator analysis list; The first ligand molecule list is composed of the first task quantity of first ligand molecule records; the first ligand molecule record includes a first ligand conformation field, a first BA indicator field, a first QED indicator field, a first SA indicator field, a first SIM indicator field and a first LPSK indicator field; The first indicator analysis list includes multiple first indicator analysis records; the first indicator analysis record includes a first indicator type field, a first indicator threshold field, a first indicator statistics field and a first indicator compliance status field; the first indicator type field includes a BA type, a QED type, a SA type, a SIM type and an LPSK type; the first indicator threshold field is a first BA indicator threshold, a first QED indicator threshold, a first SA indicator threshold, a first SIM indicator threshold or a first LPSK indicator threshold corresponding to the first indicator type field; the first indicator statistics field is an indicator statistical parameter corresponding to the first indicator type field; the first indicator compliance status field includes compliance and non-compliance; The current step 3 specifically includes: Step 3-1, extracting the corresponding first generated molecule quantity, the first protein pocket conformation, the first reference ligand molecule conformation and the first indicator set from the first data set; and initializing the corresponding first counter to 1; Step 3-2, predicting the center of mass coordinates and the total number of atoms of the ligand molecule corresponding to the first protein pocket conformation by the molecular prediction model to obtain the corresponding current parameter prediction vector; and using the predicted center of mass coordinates and the predicted total number of atoms in the parameter prediction vector as the corresponding first center of mass coordinates and the first total number of atoms; Step 3-3, initializing a ligand molecular conformation as a corresponding first ligand molecular conformation based on the first mass center coordinates and the first total number of atoms; Among them, the total number of atomic objects in the first ligand molecular conformation matches the total number of first atoms; all atomic types in the first ligand molecular conformation are preset invalid atomic types; all atomic coordinates in the first ligand molecular conformation are random coordinates that satisfy Gaussian distribution, and the coordinate mean of all atomic coordinates in the first ligand molecular conformation matches the first centroid coordinates; the line between any two atomic objects in the first ligand molecular conformation will not intersect with the line between any two corresponding atomic objects in the first protein pocket conformation; Step 3-4, inputting the first protein pocket conformation, the first center of mass coordinates and the first ligand molecule conformation into the ligand molecule generation model; the ligand molecule generation model regards the first ligand molecule conformation as a scrambled post-diffusion conformation, and iteratively denoises each atom type and coordinate in the current post-diffusion conformation using the first protein pocket conformation and the first center of mass coordinates to obtain a corresponding denoised ligand conformation, and outputs the currently obtained denoised ligand conformation as the corresponding second ligand molecule conformation; Step 3-5, input the first protein pocket conformation and the second ligand molecular conformation into the atomic-level optimization model; the atomic-level optimization model regards the second ligand molecular conformation as a scrambled post-diffusion conformation, and performs iterative denoising on each atom type and coordinate in the current post-diffusion conformation according to the first protein pocket conformation to obtain a corresponding denoised ligand conformation, and outputs the currently obtained denoised ligand conformation as the corresponding third ligand molecular conformation; Step 3-6, inputting the first protein pocket conformation and the third ligand molecule conformation into the component-level optimization model; the component-level optimization model performs corresponding amino acid component identification on the first protein pocket conformation to obtain the corresponding first amino acid component set, and performs corresponding functional group and connector component identification on the third ligand molecule conformation to obtain the corresponding first functional group component set and first connector component set, and performs iterative denoising on the three-dimensional conformations of each functional group component in the first functional group component set and the first connector component set and each connector component according to the first amino acid component set to obtain the corresponding denoised ligand conformation, and outputs the currently obtained denoised ligand conformation as the corresponding fourth ligand molecule conformation; Step 3-7, using a preset computer-aided drug design tool, the binding affinity score between the first protein pocket conformation and the fourth ligand molecular conformation is calculated to obtain a corresponding first index score; and the drug property quantitative evaluation score of the fourth ligand molecular conformation is calculated to obtain a corresponding second index score; and the synthesizability score of the fourth ligand molecular conformation is calculated to obtain a corresponding third index score; and the structural similarity score between the fourth ligand molecular conformation and the first reference ligand molecular conformation is calculated to obtain a corresponding fourth index score; and whether the fourth ligand molecular conformation satisfies the Lipinski five rules is identified, and if so, the corresponding fifth index score is set to 1, and if not, the corresponding fifth index score is set to 0; Among them, computer-aided drug design tools include at least DOCK, AutoDock, GOLD, FlexX, Glide, ICM, RDkits, and Discovery Studio; Step 3-8, the fourth ligand molecule conformation, the first index score, the second index score, the third index score, the fourth index score and the fifth index score are used as the corresponding first ligand conformation field, the first BA score field, the first QED score field, the first SA score field, the first SIM score field and the first LPSK score field to form a corresponding first ligand molecule record; Step 3-9, adding 1 to the first counter; and identifying whether the first counter after adding 1 exceeds the corresponding first generated molecule quantity; if not, returning to step 3-3; if it has exceeded, forming a corresponding first ligand molecule list from the first ligand molecule records of the obtained first generated molecule quantity, and going to step 3-10; Step 3-10, bring all the first BA score fields in the first ligand molecule list into the preset first evaluation function to calculate the corresponding first index evaluation value, bring all the first QED score fields into the preset second evaluation function to calculate the corresponding second index evaluation value, bring all the first SA score fields into the preset third evaluation function to calculate the corresponding third index evaluation value, and bring all the first SIM score fields into the preset fourth evaluation function to calculate the corresponding fourth index evaluation value; and count the total number of first ligand molecule records whose first LPSK score field in the first ligand molecule list is 1 to obtain the corresponding first total number, and calculate the corresponding fifth index evaluation value based on the first total number and the first generated molecule number = (first total number / first generated molecule number) × 100%; Among them, the first, second, third and fourth evaluation functions are all implemented based on the RMSE function; Step 3-11, taking the first, second, third, fourth and fifth indicator evaluation values ​​as the corresponding current indicator evaluation values ​​in sequence; and setting a corresponding current indicator type as the BA type, QED type, SA type, SIM type or LPSK type corresponding to the current indicator evaluation value; and taking the first BA indicator threshold, the first QED indicator threshold, the first SA indicator threshold, the first SIM indicator threshold or the first LPSK indicator threshold corresponding to the current indicator type in the first indicator set as the corresponding current indicator threshold; and identifying whether the current indicator evaluation value is lower than the corresponding current indicator threshold, if so, setting a corresponding current indicator compliance status as unsatisfactory, otherwise setting a corresponding current indicator compliance status as satisfactory; and taking the current indicator type, the current indicator threshold, the current indicator evaluation value and the current indicator compliance status as the corresponding first indicator type field, the first indicator threshold field, the first indicator statistics field and the first indicator compliance status field to form a corresponding first indicator analysis record; and forming a corresponding first indicator analysis list from all the obtained first indicator analysis records; Step 3-12, the obtained first ligand molecule list and the first indicator analysis list form a corresponding first task report which is fed back to the user.

[0019] Figure 2 The module structure diagram of a processing device for drug molecule generation based on a diffusion model provided in the second embodiment of the present invention is a terminal device or server that implements the aforementioned method embodiment, and can also be a device that enables the aforementioned terminal device or server to implement the aforementioned method embodiment. For example, the device can be a device or chip system of the aforementioned terminal device or server. Figure 2 As shown, the device includes: a model preparation module 201, a task data receiving module 202 and a task processing module 203.

[0020] The model preparation module 201 is used to design a model for predicting the corresponding ligand center of mass coordinates and the total number of ligand atoms according to the protein pocket conformation input by the model as a molecular prediction model; and to train the molecular prediction model based on a preset first data set; and to use the maturely trained DiffBP model, TargetDiff model and D3FG model as the corresponding ligand molecule generation model, atomic-level optimization model and component-level optimization model; the DiffBP model, TargetDiff model and D3FG model are each a type of ligand molecule generation model based on a diffusion model.

[0021] The task data receiving module 202 is used to receive the drug molecule generation task data set input by the user as the corresponding first data set; the first data set includes the first number of generated molecules, the first protein pocket conformation, the first reference ligand molecular conformation and the first indicator set; the first number of generated molecules is a positive integer; the first protein pocket conformation is a three-dimensional molecular conformation corresponding to a protein pocket; the first reference ligand molecular conformation is a three-dimensional molecular conformation of a reference ligand molecule; the first indicator set at least includes a first BA indicator threshold, a first QED indicator threshold, a first SA indicator threshold, a first SIM indicator threshold and a first LPSK indicator threshold; the three-dimensional molecular conformation is composed of multiple atomic objects; the atomic object includes an atomic index, an atomic type and an atomic coordinate.

[0022] The task processing module 203 is used to perform ligand molecule generation task processing according to the first data set based on the molecular prediction model, the ligand molecule generation model, the atomic level optimization model and the component level optimization model to obtain the corresponding first task report and feedback it to the user.

[0023] An embodiment of the present invention provides a processing device for generating drug molecules based on a diffusion model, which can execute the method steps in the above method embodiment. Its implementation principle and technical effects are similar and will not be repeated here.

[0024] It should be noted that it should be understood that the division of the various modules of the above device is only a division of logical functions. In actual implementation, they can be fully or partially integrated into one physical entity, or they can be physically separated. And these modules can all be implemented in the form of software called by processing elements; they can also be all implemented in the form of hardware; some modules can also be implemented in the form of software called by processing elements, and some modules can be implemented in the form of hardware. For example, the model preparation module can be a separately established processing element, or it can be integrated in a chip of the above device. In addition, it can also be stored in the memory of the above device in the form of program code, and called and executed by a processing element of the above device. The function of the above-mentioned module is determined. The implementation of other modules is similar. In addition, these modules can be fully or partially integrated together, or they can be implemented independently. The processing element described here can be an integrated circuit with signal processing capabilities. In the implementation process, each step of the above method or each module above can be completed by an integrated logic circuit of hardware in the processor element or instructions in the form of software.

[0025] For example, the above modules may be one or more integrated circuits configured to implement the above methods, such as one or more application specific integrated circuits (ASIC), or one or more digital signal processors (DSP), or one or more field programmable gate arrays (FPGA). For another example, when a module above is implemented in the form of a processing element scheduling program code, the processing element may be a general-purpose processor, such as a central processing unit (CPU) or other processor that can call program code. For another example, these modules may be integrated together and implemented in the form of a system-on-a-chip (SOC).

[0026] In the above embodiments, all or part of the embodiments may be implemented by software, hardware, firmware or any combination thereof. When implemented by software, all or part of the embodiments may be implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the above method embodiments are generated. The above computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The above-mentioned computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the above-mentioned computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center through wired (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (such as infrared, wireless, Bluetooth, microwave, etc.) methods. The above-mentioned computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more available media integrated. The above-mentioned available medium can be a magnetic medium (such as a floppy disk, a hard disk, a magnetic tape), an optical medium (such as a DVD), or a semiconductor medium (such as a solid state disk (SSD)), etc.

[0027] Figure 3 This is a schematic diagram of the structure of an electronic device provided in Embodiment 3 of the present invention. The electronic device may be a terminal device or a server for implementing the method of the aforementioned embodiment, or may be a terminal device or a server for implementing the method of the aforementioned embodiment connected to the aforementioned terminal device or server. Figure 3As shown, the electronic device may include: a processor 301 (such as a CPU), a memory 302, and a transceiver 303; the transceiver 303 is coupled to the processor 301, and the processor 301 controls the transceiver 303. Various instructions may be stored in the memory 302 to complete various processing functions and implement the processing steps described in the aforementioned embodiment method. Preferably, the electronic device involved in the embodiment of the present invention also includes: a power supply 304, a system bus 305 and a communication port 306. The system bus 305 is used to realize the communication connection between components. The above-mentioned communication port 306 is used for connecting and communicating between the electronic device and other peripherals.

[0028] exist Figure 3 The system bus 305 mentioned in the figure can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The system bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus. The communication interface is used to realize the communication between the database access device and other devices (such as clients, read-write libraries, and read-only libraries). The memory may include random access memory (RAM) and may also include non-volatile memory (Non-Volatile Memory), such as at least one disk storage.

[0029] The above-mentioned processors can be general-purpose processors, including central processing units (CPU), network processors (NP), graphics processing units (GPU), etc.; they can also be digital signal processors (DSP), application-specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0030] It should be noted that an embodiment of the present invention further provides a computer-readable storage medium, in which instructions are stored. When the computer-readable storage medium is run on a computer, the computer executes the method and processing process provided in the above embodiments.

[0031] The embodiment of the present invention provides a processing method, device, electronic device and computer-readable storage medium for drug molecule generation based on a diffusion model. As can be seen from the above content, the embodiment of the present invention pre-designs a model for predicting the corresponding ligand mass center coordinates and the total number of ligand atoms according to the protein pocket conformation input by the model as a molecular prediction model and trains it, and uses three types of ligand molecule generation models (DiffBP model, TargetDiff model, D3FG model) that are implemented based on the diffusion model and have been trained to be mature as corresponding ligand molecule generation models, atomic-level optimization models and component-level optimization models; then the above four types of models (molecular prediction model, ligand molecule generation model, atomic-level optimization model and component-level optimization model) are used to automatically process the batch generation task of ligand molecules, and the five types of index scores (BA, QED, SA, SIM, LPSK) of each ligand molecule are calculated during the task processing using a computer-aided drug design tool, and the five types of index scores of the batch task are statistically evaluated before the task processing is completed. The ligand molecule design method based on the diffusion model of the embodiment of the present invention is used to replace the traditional ligand molecule design method based on molecular motion simulation and quantum chemical calculation, which not only shortens the molecular design cycle but also improves the design efficiency.

[0032] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0033] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for generating drug molecules based on a diffusion model, characterized in that: The method comprises: A model for predicting the corresponding ligand mass center coordinates and the total number of ligand atoms according to the protein pocket conformation input by the model is designed as a molecular prediction model; and the molecular prediction model is trained based on a preset first data set; and the maturely trained DiffBP model, TargetDiff model and D3FG model are used as corresponding ligand molecule generation models, atomic-level optimization models and component-level optimization models; the DiffBP model, the TargetDiff model and the D3FG model are each a type of ligand molecule generation model based on a diffusion model; Receive a drug molecule generation task data set input by a user as the corresponding first data set; the first data set includes a first number of generated molecules, a first protein pocket conformation, a first reference ligand molecular conformation and a first indicator set; the first number of generated molecules is a positive integer; the first protein pocket conformation is a three-dimensional molecular conformation corresponding to a protein pocket; the first reference ligand molecular conformation is the three-dimensional molecular conformation of a reference ligand molecule; the first indicator set at least includes a first BA indicator threshold, a first QED indicator threshold, a first SA indicator threshold, a first SIM indicator threshold and a first LPSK indicator threshold; the three-dimensional molecular conformation is composed of multiple atomic objects; the atomic object includes an atomic index, an atomic type and an atomic coordinate; Based on the molecular prediction model, the ligand molecule generation model, the atomic-level optimization model and the component-level optimization model, the ligand molecule generation task is processed according to the first data set to obtain a corresponding first task report to be fed back to the user.

2. The method for generating drug molecules based on a diffusion model according to claim 1, characterized in that: The model input of the molecular prediction model is a protein pocket conformation A0, and the model output is a parameter prediction vector P; the molecular prediction model is used to predict the centroid coordinates and the total number of atoms of the ligand molecule corresponding to the protein pocket conformation A0 input by the model and output the corresponding parameter prediction vector P; the protein pocket conformation A0 is the three-dimensional molecular conformation corresponding to a protein pocket; the parameter prediction vector P includes the predicted centroid coordinates and the predicted total number of atoms; The model input of the DiffBP model includes protein pocket conformation A1, ligand molecule input conformation B in,1 and the center of mass coordinates of the ligand molecule, the model output is a corresponding ligand molecule output conformation B out,1 ; The DiffBP model is used to input the ligand molecule input into the model conformation B in,1 As a scrambled diffusion conformation, and according to the protein pocket conformation A1 and the center of mass coordinates of the ligand molecule input by the model, each atom type and coordinate in the current diffusion conformation is iteratively denoised to obtain the corresponding denoised ligand conformation, and the currently obtained denoised ligand conformation is used as the corresponding ligand molecule output conformation B out,1 Output; wherein the protein pocket conformation A1 is the three-dimensional molecular conformation corresponding to a protein pocket; the ligand molecule input conformation B in,1 and the ligand molecule outputs conformation B out,1 The three-dimensional molecular conformation of a ligand molecule bound to the current protein pocket before and after being processed by the DiffBP model; The model input of the TargetDiff model includes the protein pocket conformation A2 and the ligand molecule input conformation B in,2 , the model output is a corresponding ligand molecule output conformation B out,2 ; The TargetDiff model is used to input the ligand molecule input into the model conformation B in,2 As a scrambled diffusion conformation, and according to the protein pocket conformation A2 input by the model, each atom type and coordinate in the current diffusion conformation is iteratively denoised to obtain the corresponding denoised ligand conformation, and the currently obtained denoised ligand conformation is used as the corresponding ligand molecule output conformation B out,2 Output; wherein the protein pocket conformation A2 is the three-dimensional molecular conformation corresponding to a protein pocket; the ligand molecule input conformation B in,2 and the ligand molecule outputs conformation B out,2 The three-dimensional molecular conformation of a ligand molecule bound to the current protein pocket before and after being processed by the TargetDiff model; The model input of the D3FG model includes the protein pocket conformation A3 and the ligand molecule input conformation B in,3 , the model output is a corresponding ligand molecule output conformation B out,3 The D3FG model is used to identify the corresponding amino acid components of the protein pocket conformation A3 input by the model to obtain the corresponding amino acid component set, and to input the conformation B of the ligand molecule input by the model. in,2 The corresponding functional group and connector component identification is performed to obtain the corresponding functional group component set and connector component set, and the three-dimensional conformations of each functional group component and each connector component in the functional group component set and the connector component set are iteratively denoised according to the amino acid component set to obtain the corresponding denoised ligand conformation, and the currently obtained denoised ligand conformation is used as the corresponding ligand molecule output conformation B out,3 Output; wherein the protein pocket conformation A3 is the three-dimensional molecular conformation corresponding to a protein pocket; the ligand molecule input conformation B in,3 and the ligand molecule outputs conformation B out,3 It is the three-dimensional molecular conformation of a ligand molecule bound to the current protein pocket before and after being processed by the D3FG model.

3. The method for generating drug molecules based on a diffusion model according to claim 1, characterized in that: The first BA index threshold is an affinity evaluation index threshold; The first QED indicator threshold is a threshold for quantitative evaluation of drug properties; The first SA indicator threshold is a synthesizability evaluation indicator threshold; The first SIM index threshold is a structural similarity evaluation index threshold; The first LPSK indicator threshold is an evaluation indicator threshold of the Lipinski five rules.

4. The method for generating drug molecules based on a diffusion model according to claim 2, characterized in that: The first data set includes multiple first data records; the first data record includes a first training conformation, a first pocket atom index set and a first ligand atom index set; the first training conformation is the three-dimensional molecular conformation corresponding to a complex molecule, and the complex molecule is composed of a protein pocket and a ligand molecule; the first pocket atom index set and the first ligand atom index set are each composed of multiple atom indices in the first training conformation; the first pocket atom index set is the atom index set of the protein pocket in the current complex molecule; the first ligand atom index set is the atom index set of the ligand molecule in the current complex molecule; The first task report includes a first ligand molecule list and a first indicator analysis list; The first ligand molecule list is composed of the first task quantity of first ligand molecule records; The first ligand molecule record includes a first ligand conformation field, a first BA indicator field, a first QED indicator field, a first SA indicator field, a first SIM indicator field and a first LPSK indicator field; The first indicator analysis list includes a plurality of first indicator analysis records; The first indicator analysis record includes a first indicator type field, a first indicator threshold field, a first indicator statistics field and a first indicator compliance status field; the first indicator type field includes BA type, QED type, SA type, SIM type and LPSK type; the first indicator threshold field is the first BA indicator threshold, the first QED indicator threshold, the first SA indicator threshold, the first SIM indicator threshold or the first LPSK indicator threshold corresponding to the first indicator type field; the first indicator statistics field is the indicator statistical parameter corresponding to the first indicator type field; the first indicator compliance status field includes met and not met.

5. The method for generating drug molecules based on a diffusion model according to claim 4, characterized in that: The training of the molecular prediction model based on the preset first data set specifically includes: Step 51, randomly splitting the first data set based on a preset first split ratio to obtain two corresponding sub-data sets recorded as a corresponding first training set and a first evaluation set; Wherein, both the first training set and the first evaluation set are composed of a plurality of the first data records; the ratio of the total number of records in the first training set to the total number of records in the first evaluation set satisfies the first segmentation ratio; Step 52, extracting the first first data record of the first training set as the corresponding current training record; Step 53, taking the first training conformation, the first pocket atom index set and the first ligand atom index set of the current training record as the corresponding current training conformation, the first atom index set and the second atom index set; and performing corresponding protein pocket and ligand molecule conformation extraction processing on the current training conformation based on the first and second atom index sets to obtain the corresponding first pocket conformation and first ligand conformation; and performing mean calculation on the atomic coordinates of all the atomic objects in the first ligand conformation and using the calculation result as the corresponding first label centroid coordinates; and performing statistics on the total number of atomic objects in the first ligand conformation and using the statistical result as the corresponding first label atom total number; and forming the corresponding first parameter label vector by the first label centroid coordinates and the first label atom total number; Step 54, using the molecular prediction model to predict the center of mass coordinates and the total number of atoms of the ligand molecule corresponding to the first pocket conformation to obtain a corresponding first parameter prediction vector; Step 55, bringing the first parameter label vector and the first parameter prediction vector into a preset first model loss function to calculate and obtain a corresponding first loss value; Wherein, the first model loss function includes at least an L1 loss function and an L2 loss function; Step 56, identifying whether the first loss value satisfies a preset first loss value range; if the first loss value satisfies the first loss value range, identifying whether the current training record is the last first data record of the first training set, if so, turning to step 57, if not, extracting the next first data record of the first training set as the new current training record and returning to step 53; if the first loss value does not satisfy the first loss value range, performing a round of modulation on the model parameters of the molecular prediction model in a direction to minimize the first model loss function based on a preset first model optimizer, and returning to step 54 at the end of this round of modulation; Wherein, the first model optimizer includes at least an SGD optimizer and an ADAM optimizer; Step 57, perform a round of traversal on all the first data records of the first evaluation set; and in this round of traversal, use the first data record currently traversed as the corresponding current evaluation record; and use the first training conformation, the first pocket atom index set and the first ligand atom index set of the current evaluation record as the corresponding current training conformation, the third atom index set and the fourth atom index set; and perform corresponding protein pocket and ligand molecule conformation extraction processing on the current training conformation based on the third and fourth atom index sets to obtain the corresponding second pocket conformation and the second ligand conformation; and perform mean calculation on the atomic coordinates of all the atomic objects in the second ligand conformation and use the calculation result as the The corresponding second label centroid coordinates; and the total number of atomic objects of the second ligand conformation is counted and the statistical result is used as the corresponding second label atom total number; and the second label centroid coordinates and the second label atom total number form a corresponding second parameter label vector; and the molecular prediction model predicts the centroid coordinates and the total number of atoms of the ligand molecule corresponding to the second pocket conformation to obtain the corresponding second parameter prediction vector; and the second parameter label vector obtained this time and the corresponding second parameter prediction vector form a corresponding first label-prediction pair; and at the end of this round of traversal, all the first prediction-label pairs obtained are brought into the preset first model evaluation function to calculate and obtain the corresponding first evaluation value; Wherein, the first model evaluation function at least includes an RMSE function; Step 58, identifying whether the first evaluation value satisfies a preset first evaluation value range; if the first evaluation value does not satisfy the first evaluation value range, returning to step 51 to continue training; if the first evaluation value satisfies the first evaluation value range, stopping training and confirming that the model training of the molecular prediction model is completed.

6. The method for generating drug molecules based on a diffusion model according to claim 2, characterized in that: The performing ligand molecule generation task processing based on the molecular prediction model, the ligand molecule generation model, the atomic level optimization model and the component level optimization model according to the first data set to obtain a corresponding first task report and feed it back to the user specifically includes: Step 6-1, extracting the corresponding first generated molecule quantity, the first protein pocket conformation, the first reference ligand molecule conformation and the first indicator set from the first data set; and initializing the corresponding first counter to 1; Step 6-2, predicting the center of mass coordinates and the total number of atoms of the ligand molecule corresponding to the first protein pocket conformation by the molecular prediction model to obtain a corresponding current parameter prediction vector; and using the predicted center of mass coordinates and the predicted total number of atoms in the parameter prediction vector as the corresponding first center of mass coordinates and the first total number of atoms; Step 6-3, initializing a ligand molecular conformation as a corresponding first ligand molecular conformation based on the first mass center coordinates and the first total number of atoms; wherein the total number of atomic objects in the first ligand molecular conformation matches the first total number of atoms; all of the atomic types in the first ligand molecular conformation are preset invalid atomic types; all of the atomic coordinates in the first ligand molecular conformation are random coordinates satisfying Gaussian distribution, and the coordinate mean of all of the atomic coordinates in the first ligand molecular conformation matches the first centroid coordinates; a line between any two of the atomic objects in the first ligand molecular conformation will not intersect a line between any two of the corresponding atomic objects in the first protein pocket conformation; Step 6-4, inputting the first protein pocket conformation, the first center of mass coordinates and the first ligand molecule conformation into the ligand molecule generation model; the ligand molecule generation model regards the first ligand molecule conformation as a scrambled post-diffusion conformation, and iteratively denoises each atom type and coordinate in the current post-diffusion conformation using the first protein pocket conformation and the first center of mass coordinates to obtain a corresponding denoised ligand conformation, and outputs the currently obtained denoised ligand conformation as the corresponding second ligand molecule conformation; Step 6-5, inputting the first protein pocket conformation and the second ligand molecular conformation into the atomic-level optimization model; the atomic-level optimization model regards the second ligand molecular conformation as a scrambled post-diffusion conformation, and performs iterative denoising on each atom type and coordinate in the current post-diffusion conformation according to the first protein pocket conformation to obtain a corresponding denoised ligand conformation, and outputs the currently obtained denoised ligand conformation as the corresponding third ligand molecular conformation; Step 6-6, inputting the first protein pocket conformation and the third ligand molecular conformation into the component-level optimization model; the component-level optimization model performs corresponding amino acid component identification on the first protein pocket conformation to obtain a corresponding first amino acid component set, and performs corresponding functional group and connector component identification on the third ligand molecular conformation to obtain a corresponding first functional group component set and a first connector component set, and performs iterative denoising on the three-dimensional conformations of each functional group component in the first functional group component set and the first connector component set and each connector component according to the first amino acid component set to obtain a corresponding denoised ligand conformation, and outputs the currently obtained denoised ligand conformation as the corresponding fourth ligand molecular conformation; Step 6-7, using a preset computer-aided drug design tool, the binding affinity score between the first protein pocket conformation and the fourth ligand molecular conformation is calculated to obtain a corresponding first index score; and the drug property quantitative evaluation score of the fourth ligand molecular conformation is calculated to obtain a corresponding second index score; and the synthesizability score of the fourth ligand molecular conformation is calculated to obtain a corresponding third index score; and the structural similarity score between the fourth ligand molecular conformation and the first reference ligand molecular conformation is calculated to obtain a corresponding fourth index score; and whether the fourth ligand molecular conformation satisfies the Lipinski five rules is identified, and if so, the corresponding fifth index score is set to 1, and if not, the corresponding fifth index score is set to 0; Wherein, the computer-aided drug design tools include at least DOCK, AutoDock, GOLD, FlexX, Glide, ICM, RDkits, and Discovery Studio; Step 6-8, using the fourth ligand molecule conformation, the first index score, the second index score, the third index score, the fourth index score and the fifth index score as the corresponding first ligand conformation field, the first BA score field, the first QED score field, the first SA score field, the first SIM score field and the first LPSK score field to form a corresponding first ligand molecule record; Step 6-9, adding 1 to the first counter; and identifying whether the first counter after adding 1 exceeds the corresponding first generated molecule quantity; if not, returning to step 6-3; if it has exceeded, forming the corresponding first ligand molecule list from the first ligand molecule records of the first generated molecule quantity, and going to step 6-10; Step 6-10, bring all the first BA score fields in the first ligand molecule list into a preset first evaluation function to calculate the corresponding first index evaluation value, bring all the first QED score fields into a preset second evaluation function to calculate the corresponding second index evaluation value, bring all the first SA score fields into a preset third evaluation function to calculate the corresponding third index evaluation value, and bring all the first SIM score fields into a preset fourth evaluation function to calculate the corresponding fourth index evaluation value; and count the total number of the first ligand molecule records whose first LPSK score field in the first ligand molecule list is 1 to obtain the corresponding first total number, and calculate the corresponding fifth index evaluation value based on the first total number and the first generated molecule number = (first total number / first generated molecule number) × 100%; Wherein, the first, second, third and fourth evaluation functions are all implemented based on the RMSE function; Step 6-11, taking the first, second, third, fourth and fifth indicator evaluation values ​​as the corresponding current indicator evaluation values ​​in sequence; and setting a corresponding current indicator type as the BA type, QED type, SA type, SIM type or LPSK type corresponding to the current indicator evaluation value; and taking the first BA indicator threshold, the first QED indicator threshold, the first SA indicator threshold, the first SIM indicator threshold or the first LPSK indicator threshold corresponding to the current indicator type in the first indicator set as the corresponding current indicator threshold; and identifying whether the current indicator evaluation value is lower than the corresponding current indicator threshold, if so, setting a corresponding current indicator compliance state as unsatisfactory, otherwise setting a corresponding current indicator compliance state as satisfactory; and taking the current indicator type, the current indicator threshold, the current indicator evaluation value and the current indicator compliance state as the corresponding first indicator type field, the first indicator threshold field, the first indicator statistics field and the first indicator compliance state field to form a corresponding first indicator analysis record; and forming a corresponding first indicator analysis list from all the obtained first indicator analysis records; Step 6-12: The first task report corresponding to the obtained first ligand molecule list and the first indicator analysis list is fed back to the user.

7. A device for executing the processing method for generating drug molecules based on a diffusion model according to any one of claims 1 to 6, characterized in that: The device comprises: a model preparation module, a task data receiving module and a task processing module; The model preparation module is used to design a model for predicting the corresponding ligand mass center coordinates and the total number of ligand atoms according to the protein pocket conformation input by the model as a molecular prediction model; and to train the molecular prediction model based on a preset first data set; and to use the maturely trained DiffBP model, TargetDiff model and D3FG model as the corresponding ligand molecule generation model, atomic-level optimization model and component-level optimization model; the DiffBP model, the TargetDiff model and the D3FG model are each a type of ligand molecule generation model based on a diffusion model; The task data receiving module is used to receive the drug molecule generation task data set input by the user as the corresponding first data set; the first data set includes the first generated molecule quantity, the first protein pocket conformation, the first reference ligand molecular conformation and the first indicator set; the first generated molecule quantity is a positive integer; the first protein pocket conformation is a three-dimensional molecular conformation corresponding to a protein pocket; the first reference ligand molecular conformation is the three-dimensional molecular conformation of a reference ligand molecule; the first indicator set at least includes a first BA indicator threshold, a first QED indicator threshold, a first SA indicator threshold, a first SIM indicator threshold and a first LPSK indicator threshold; the three-dimensional molecular conformation is composed of multiple atomic objects; the atomic object includes atomic index, atomic type and atomic coordinates; The task processing module is used to perform ligand molecule generation task processing according to the first data set based on the molecular prediction model, the ligand molecule generation model, the atomic level optimization model and the component level optimization model to obtain a corresponding first task report and provide feedback to the user.

8. An electronic device, characterized in that: include: memory, processors, and transceivers; The processor is used to couple with the memory, read and execute instructions in the memory, so as to implement the method according to any one of claims 1 to 6; The transceiver is coupled to the processor, and the processor controls the transceiver to send and receive messages.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and when the computer instructions are executed by a computer, the computer executes the method according to any one of claims 1 to 6.

Images