Processing method and device for optimizing drug molecules

Abstract

The embodiment of the invention relates to a processing method and device for optimizing drug molecules. The method comprises the following steps: step 1, constructing a chemical information base; 2, constructing a synthetic path regression prediction model; 3, constructing a model training data set training model; 4, receiving an old drug molecular map; 5, carrying out molecular structure optimization and synthetic route analysis on the old drug molecular map based on a synthetic route regression prediction model; 6, measuring and calculating the structural similarity and property difference of the new and old drug molecular maps, and evaluating the optimization effect according to a measurement and calculation report; step 7, if the structure evaluation result is that the structure has obvious difference or the property evaluation result is that the comprehensive molecule is smaller than the old molecule, returning to the step 5, otherwise, returning to the step 8; and step 8, feeding back the old drug molecular map, the newest new drug molecular map, the new molecule synthesis path and the new and old molecule measurement report to the user. According to the invention, the synthesizability of drug molecules 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 optimizing drug molecules. Background Art

[0002] The design process of traditional drug molecules is extremely complex and time-consuming. With the application and development of artificial intelligence technology in the field of drug molecule design, the application of various generative models has effectively improved the design efficiency of new drug molecules. However, we have also found in practical applications that due to the lack of full consideration of synthetic feasibility, some generative drug molecules output by generative models are difficult or even impossible to effectively synthesize in real scenarios. 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 optimizing drug molecules in view of the defects of the prior art. The present invention first creates a corresponding building block information library by collecting the real molecular fragments and / or unit structure information of various drug molecules, and creates a corresponding reaction rule information library by collecting the real synthesis reaction information occurring during the synthesis of various drug molecules, and the building block information library and the reaction rule information library form a corresponding chemical information library; then, a synthetic pathway regression prediction model is constructed in combination with the chemical information library, and a model training data set is constructed to train the synthetic pathway regression prediction model; then, after the model training is completed, the generative drug molecule graph input by the user each time is used as the corresponding old drug molecule graph, and the molecular structure optimization and synthetic pathway analysis of the old drug molecule graph are performed based on the synthetic pathway regression prediction model to obtain the corresponding new drug molecule graph and new molecule synthetic pathway, and The structural similarity and property difference of the new and old drug molecular maps are calculated to obtain the corresponding new and old molecule measurement reports, and the optimization effect is evaluated based on the new and old molecule measurement reports to obtain the corresponding comprehensive evaluation results, and when the structural evaluation result of the comprehensive evaluation result is that the structure has obvious differences or the property evaluation result is inferior to the old molecule in all aspects, the molecular structure optimization and synthesis path analysis are performed again based on the synthesis path regression prediction model until the latest structural evaluation result is that the structure is similar and the property evaluation result is not inferior to the old molecule in all aspects, and when the structural evaluation result is that the structure is similar and the property evaluation result is not inferior to the old molecule in all aspects, the old drug molecular map, the latest new drug molecular map, the new molecule synthesis path and the new and old molecule measurement reports are fed back to the current user. The synthesis pathway regression prediction model provided by the present invention can optimize the structure of drug molecules on the premise of meeting the synthesizability requirements, and can predict the synthesis reaction pathway of the optimized drug molecules; the present invention can not only improve the synthesizability of drug molecules, but also guide the synthesis process of optimized new drug molecules, and can also ensure the structural similarity of new and old drug molecules before and after optimization, and can also comprehensively evaluate the quality of the properties of new and old drug molecules before and after optimization.

[0004] To achieve the above object, a first aspect of an embodiment of the present invention provides a method for optimizing drug molecules, the method comprising:

[0005] Step 1, creating a corresponding building block information library by collecting real molecular fragments and / or unit structure information of various drug molecules; and creating a corresponding reaction rule information library by collecting real synthesis reaction information occurring during the synthesis of various drug molecules; and forming a corresponding chemical information library by the building block information library and the reaction rule information library;

[0006] Step 2, constructing a synthetic pathway regression prediction model in combination with the chemical information library;

[0007] Step 3, constructing a model training data set; and training the synthesis path regression prediction model based on the model training data set;

[0008] Step 4: After the model training is completed, the generated drug molecule graph input by the user is received as the corresponding old drug molecule graph;

[0009] Step 5, performing molecular structure optimization and synthesis path analysis on the old drug molecular graph based on the synthesis path regression prediction model to obtain a corresponding new drug molecular graph and a new molecular synthesis path;

[0010] Step 6, calculating the structural similarity and property difference of the new and old drug molecule graphs to obtain a corresponding new and old molecule calculation report; and evaluating the optimization effect according to the new and old molecule calculation report to obtain a corresponding comprehensive evaluation result;

[0011] The comprehensive evaluation results include structural evaluation results and property evaluation results; the structural evaluation results include similar structures and obvious differences in structures; the property evaluation results include overall superiority to the old molecule, partial superiority to the old molecule, and overall inferiority to the old molecule;

[0012] Step 7, identifying the structure evaluation result and the property evaluation result of the comprehensive evaluation result; if the structure evaluation result is that the structure is significantly different or the property evaluation result is inferior to the old molecule in all aspects, returning to step 5 for re-optimization; if the structure evaluation result is that the structure is similar and the property evaluation result is not inferior to the old molecule in all aspects, going to step 8;

[0013] Step 8, feeding back the old drug molecule map, the latest new drug molecule map, the new molecule synthesis pathway and the new and old molecule measurement report to the current user.

[0014] Preferably, the building block information library includes a plurality of building block information records; each of the building block information records corresponds to a real molecular fragment or a real molecular unit structure; the building block information record includes a building block identifier, a building block molecular graph and a building block molecular fingerprint; the building block identifier is a unique identifier of the current building block; the building block molecular graph is a molecular graph of the current building block; the building block molecular fingerprint is a molecular fingerprint of the current building block;

[0015] The reaction rule information library includes multiple reaction rule information records; each of the reaction rule information records corresponds to a type of chemical reaction that occurs during the synthesis of a type of drug molecule; the reaction rule information record includes a reaction rule type, an optional range of reactant A, and an optional range of reactant B; the optional ranges of reactant A / B each contain multiple reactant identifiers; each of the reactant identifiers in the reaction rule information library corresponds to one of the building block identifiers in the building block information library.

[0016] Preferably, the synthetic pathway regression prediction model is used to optimize the molecular structure of the original drug molecule graph input into the model, perform regression prediction processing on the synthetic pathway of the optimized structure, and output the corresponding optimized drug molecule graph and optimized molecular synthetic pathway;

[0017] The model input end of the synthetic pathway regression prediction model is used to receive the original drug molecule graph input by the model, and the model output end is used to output the corresponding optimized drug molecule graph and the optimized molecular synthetic pathway;

[0018] The synthetic pathway regression prediction model includes an encoder, a decoder, a feedback output module, a reactant processing module, a synthetic processing module, a chemical information library and a molecular graph stack;

[0019] The input end of the encoder is connected to the input end of the model, and the output end is connected to the first input end of the decoder; the second and third input ends of the decoder are respectively connected to the output end of the reactant processing module and the output end of the composite processing module, and the first and second output ends of the decoder are respectively connected to the first and second input ends of the feedback output module; the first, second and third output ends of the feedback output module are respectively connected to the input end of the reactant processing module, the input end of the composite processing module and the output end of the model; the reactant processing module and the composite processing module are also respectively connected to the chemical information library; the feedback output module, the reactant processing module and the composite processing module are also respectively connected to the molecular graph stack;

[0020] The encoder is implemented based on the Uni-Mol model; the encoder is used to perform atomic level embedding coding on the original drug molecule graph to obtain a corresponding coding tensor X and send it to the decoder;

[0021] The decoder is implemented based on the Decoder model of the Transformer model; the decoder is formed by sequentially connecting an embedded coding module, N decoding modules, a linear layer and a Softmax layer, and the preset total number of decoding layers N is a positive integer; the first and second input ends of the embedded coding module are respectively connected to the second and third input ends of the decoder; the first input end of the first decoding module is connected to the output end of the embedded coding module; except for the first decoding module, the first input end of each of the remaining decoding modules is connected to the output end of the previous decoding module; the second input end of each decoding module is connected to the first input end of the decoder; the output end of the last decoding module is respectively connected to the input end of the linear layer and the first output end of the decoder; the output end of the linear layer is connected to the input end of the Softmax layer; the output end of the Softmax layer is connected to the second output end of the decoder;

[0022] The embedded coding module is used to initialize the corresponding prediction marker s0 based on the preset start marker before the first step prediction; and to set the corresponding position encoding vector PE based on the preset position embedded coding rule during the i-th step prediction. i , the single-step index i is an integer greater than or equal to 1; and the predicted marker s corresponding to the i-1th step is obtained i-1 Identify; if the predicted marker s i-1 is the start marker, then the encoding vector e is embedded based on the preset start marker start Set the corresponding encoding vector e i ; If the predicted marker s i-1 is a reactant marker, a preset first MLP network is used to predict the marker s i-1 The corresponding reactant molecular fingerprint p i-1 Perform embedded coding to obtain the corresponding encoding vector e i ; If the predicted marker s i-1 is a reaction rule marker, then the prediction marker s is embedded in the coding rule based on the preset reaction rule. i-1 The corresponding reaction rule type r i-1 Perform embedded coding to obtain the corresponding encoding vector e i ; and the position encoding vector PE obtained this time i and the encoding vector e i The starting decoded tensor H that constitutes the i-th step i,0 Sending to the first decoding module;

[0023] H i,0 =PE i +e i ,

[0024]

[0025] Among them, f MLP1 () is the model inference function corresponding to the first MLP network, θ MLP1 is the model parameter of the first MLP network; Embed() is the encoding rule for embedding the reaction rule;

[0026] The first decoding module is used to decode the encoding tensor X and the starting decoding tensor H according to the encoding tensor X and the starting decoding tensor H i,0 Decode to get the corresponding decoding tensor H i,1 Send to the second decoding module; the second decoding module is used to generate a decoded tensor based on the encoding tensor X and the decoding tensor H i,1 Decode to get the corresponding decoding tensor H i,2 Send to the third decoding module; and so on, the Nth decoding module is used to decode the tensor H according to the encoding tensor X and the decoding tensor H sent by the N-1th decoding module i,N-1 Decode to get the corresponding decoding tensor H i,N Sending to the linear layer and the feedback output module;

[0027] The linear layer is used to decode the tensor H i,N Perform full connection calculation to obtain the corresponding full connection vector L i Send to the Softmax layer;

[0028] The Softmax layer is used to calculate the fully connected vector L i Predict the i-th step marker to get the corresponding predicted marker s i Send to the feedback output module; the predicted marker s i It includes reactant marker, reaction rule marker and end marker;

[0029] The feedback output module is used to predict the predicted marker s in step i. i When the predicted marker s is a reactant marker, i and the corresponding decoded tensor H i,N Sending to the reactant processing module;

[0030] The feedback output module is also used to predict the predicted marker s in step i. i is the reaction rule marker, and the prediction marker s i and the corresponding decoded tensor H i,N Sending to the composition processing module;

[0031] The feedback output module is also used to receive a predicted marker s i When , the current time is used as the corresponding timestamp t i , and the timestamp t obtained this time i and the corresponding predicted marker s i composing a corresponding first process record and caching it;

[0032] The feedback output module is also used to read out the molecular graph pushed into the stack as the corresponding reactant molecular graph g through a stack reading operation each time the molecular graph stack completes a stacking operation. i , and take the current time as the corresponding timestamp t i , and the timestamp t obtained this time i and the corresponding reactant molecule graph g i composing a corresponding record of the first process and caching it;

[0033] The feedback output module is also used to predict the predicted marker s in step i. i as the end marker, performing a pop operation on the molecular graph stack to obtain a popped molecular graph, and using the current popped molecular graph as the corresponding optimized drug molecular graph; and performing a pop operation on all cached first process records according to the timestamp t i Sorting the sequence from front to back to form a corresponding first process record sequence, and using the first process record sequence as the corresponding optimized molecular synthesis path; sending the optimized drug molecule graph and the optimized molecular synthesis path to the model output end; and deleting all the first process records in the cache after the sending is completed;

[0034] The reactant processing module is used to receive a set of prediction markers s i and the decoded tensor H i,N When a preset second MLP network is used according to the decoded tensor H i,N Perform molecular fingerprint prediction processing to obtain the corresponding predicted fingerprint p pre,i ; and compare the molecular fingerprints of each of the building blocks in the building block information library of the chemical information library with the predicted fingerprint p pre,i The similarity is calculated to obtain the corresponding fingerprint similarity; and the building block molecular graph and the building block molecular fingerprint of the building block information record corresponding to the largest fingerprint similarity are extracted as the corresponding current molecular graph and current molecular fingerprint; and the current molecular graph is pushed into the molecular graph stack by performing a push operation on the molecular graph stack; and the current molecular fingerprint is used as the corresponding reactant molecular fingerprint p i ; and the predicted marker si and the reactant molecular fingerprint p i Sending to the embedded coding module;

[0035] Among them, p pre,u =sigmoid(f MLP2 (H i,N θ MLP2 ));

[0036] The sigmoid() function is the activation function of the second MLP network, f MLP2 () is the model inference function corresponding to the second MLP network, θ MLP2 are the model parameters of the second MLP network;

[0037] The composition processing module is used for receiving a group of predicted markers s i and the decoded tensor H i,N When , a preset third MLP network is used according to the decoding tensor H i,N Perform reaction rule type prediction processing to obtain the corresponding prediction type r pre,i ; and perform two pop operations on the molecular graph stack to obtain two popped molecular graphs as corresponding A and B molecular graphs; and identify whether the A and B molecular graphs are both empty; if at least one of the A and B molecular graphs is empty, the prediction type r is determined based on the preset invalid reaction rule type. pre,i reset; if the A and B molecular graphs are not empty, then the reaction rule type in the reaction rule information library of the chemical information library and the prediction type r pre,i The matched reaction rule information record is recorded as the corresponding current record, and whether the current record is empty is identified; if the current record is empty, the prediction type r is changed based on the invalid reaction rule type. pre,i reset; if the current record is not empty, record the building block identifiers corresponding to the A and B molecular graphs as corresponding A and B identifiers, and identify the A and B identifiers and the reactant A optional range and the reactant B optional range of the current record; if the A identifier satisfies the reactant A optional range but the B identifier does not satisfy the reactant B optional range, or the A identifier satisfies the reactant B optional range but the B identifier does not satisfy the reactant A optional range, or the B identifier satisfies the reactant A optional range but the A identifier does not satisfy the reactant B optional range, or the B identifier satisfies the reactant B optional range but the A identifier does not satisfy the reactant A optional range, or the B identifier satisfies the reactant B optional range but the A identifier does not satisfy the reactant A optional range, or the A and B identifiers do not satisfy the reactant A / B optional range, then based on the invalid reaction rule type, the prediction type r pre,iReset; and in the prediction type r pre,i If it is not the invalid reaction rule type, the prediction type r is used based on the RDKit tool. pre,i The current reaction rule is used to perform a synthetic compound prediction calculation with the A and B molecular graphs as the current reactant combination, and the synthetic compound molecular graph obtained by this prediction calculation is used as the corresponding current molecular graph, and the current molecular graph is pushed into the molecular graph stack by performing a stacking operation on the molecular graph stack; and the prediction type r pre,i As the corresponding reaction rule type r i ; and the predicted marker s i and the reaction rule type r i Sending to the embedded coding module;

[0038] Among them, r pre,i =softmax(f MLP3 (H i,N θ MLP3 ));

[0039] The softmax() function is the activation function of the third MLP network, f MLP3 () is the model inference function corresponding to the third MLP network, θ MLP3 are model parameters of the third MLP network;

[0040] The molecular graph stack is used to store the molecular graph pushed into the stack and put it on the top of the stack when processing the push operation, and when the total number of other molecular graphs in the stack is not zero, the other molecular graphs are shifted one step toward the bottom of the stack in sequence;

[0041] The molecular graph stack is also used to output the molecular graph at the top of the stack as the corresponding popped molecular graph and remove the currently output molecular graph from the stack when processing a pop operation, and when the total number of remaining molecular graphs in the stack is not zero, shift the remaining molecular graphs one step toward the top of the stack in sequence;

[0042] The molecular graph stack is also used to output the molecular graph at the top of the stack as the corresponding read molecular graph when processing a stack read operation but not to remove or shift any molecular graph in the stack.

[0043] Preferably, the model training data set includes a plurality of first data records;

[0044] The first data record includes a first training molecular graph and a first label sequence;

[0045] The first tag sequence consists of a plurality of first tags;

[0046] The first tag includes a tag index j, a tag marker and label data The tag index j is a positive integer; the tag marker It includes a reactant marker, a reaction rule marker, and an end marker; the label marker When it is a reactant tag, the tag data is a corresponding reactant molecular fingerprint label p tag,j ; The label marker When it is a reaction rule tag, the tag data is a corresponding reaction rule type label r tag,j ; The label marker When it is the end tag, the tag data Is empty.

[0047] Preferably, the constructing of the model training data set specifically includes:

[0048] Step 51, obtaining a plurality of first collected data groups by collecting drug molecule graphs and corresponding synthesis path information from a public drug molecule synthesis database;

[0049] Wherein, the first collected data group includes a first drug molecule graph and a first synthesis path; the first synthesis path is formed by sequentially sorting a plurality of first path point information; the first path point information includes a first reactant molecule graph, a second reactant molecule graph, a first reactant molecule fingerprint, a second reactant molecule fingerprint, a synthetic product molecule graph, a synthetic product molecule fingerprint and a synthetic reaction rule type; the first reactant molecule fingerprint is a molecule fingerprint of the first reactant molecule graph; the second reactant molecule fingerprint is a molecule fingerprint of the second reactant molecule graph; the synthetic product molecule fingerprint is a molecule fingerprint of the synthetic product molecule graph; the synthetic reaction rule type satisfies the reaction rule type range of the chemical information library; one and only one of the first and second reactant molecule graphs in the latter first path point information is the synthetic product molecule graph in the previous first path point information; the synthetic product molecule graph in the last first path point information matches the first drug molecule graph;

[0050] Step 52, taking the first of the first collected data groups as the corresponding current data group;

[0051] Step 53, taking the first drug molecule graph and the first synthesis path of the current data group as the corresponding current drug molecule graph and current synthesis path; taking the first first path point information of the current synthesis path as the corresponding current path point information; initializing the first index to 1; and initializing an empty sequence as the corresponding first label sequence;

[0052] Step 54: Set the tag index j as the first index; and set a corresponding tag marker for the current tag index j. For the reactant marker, set a corresponding label data is the first reactant molecular fingerprint of the current path point information; and the label index j obtained this time and its corresponding label identifier and the label data Form a corresponding first tag and add it to the first tag sequence; and add 1 to the first index;

[0053] Step 55: Set the tag index j as the first index; and set a corresponding tag marker for the current tag index j. For the reactant marker, set a corresponding label data is the first reactant molecular fingerprint of the current path point information; and the label index j obtained this time and its corresponding label identifier and the label data Form a corresponding first tag and add it to the first tag sequence; and add 1 to the first index;

[0054] Step 56: Set the tag index j as the first index; and set a corresponding tag marker for the current tag index j. For the reaction rule marker, set a corresponding label data is the synthesis reaction rule type of the current path point information; and the label index j obtained this time and its corresponding label identifier and the label data Form a corresponding first tag and add it to the first tag sequence; and add 1 to the first index;

[0055] Step 57, identifying whether the current path point information is the last of the first path point information of the current synthetic path; if so, setting a label index j as the first index, and setting a corresponding label marker for the current label index j is the end marker, and sets a corresponding tag data for the current tag index j is empty, and the label index j obtained this time and its corresponding label marker and the label data Form a corresponding first tag and add it to the first tag sequence, and use the current drug molecule graph as a corresponding first training molecule graph, and use the first training molecule graph and the first tag sequence corresponding to the current synthesis path to form a corresponding first data record, and go to step 58; if not, extract the next first path point information of the current synthesis path as the new current path point information, and return to step 54;

[0056] Step 58, identifying whether the current data group is the last of the first collected data group; if not, taking the next first collected data group as the new current data group and returning to step 53; if so, forming the corresponding model training data set from all the first data records obtained.

[0057] Preferably, the training of the synthesis pathway regression prediction model based on the model training data set specifically includes:

[0058] Step 61, based on a preset first segmentation ratio, the model training data set is divided into two sub-data sets recorded as a corresponding first training set and a first evaluation set;

[0059] 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 of the first training set to the first evaluation set satisfies the first segmentation ratio;

[0060] Step 62, taking the first first data record of the first training set as the corresponding current training record;

[0061] Step 63, inputting the first training molecular graph of the current training record as the corresponding original drug molecular graph into the synthetic pathway regression prediction model to optimize the molecular structure and perform regression prediction processing on the synthetic pathway of the optimized structure; and in the process of model processing, the prediction marker s generated by each step of the model prediction i Data collection is performed, and for each of the predicted markers s specifically as a reactant marker i The corresponding predicted fingerprint p pre,i Data collection is performed, and the prediction marker s for each reaction rule marker is i The corresponding prediction type r pre,i Perform data collection; and perform prediction sequence assembly based on all collected data and the maximum label index Q of the first label sequence of the current training record to obtain a corresponding first prediction sequence;

[0062] Wherein, Q=max(j); the first prediction sequence is composed of a plurality of first prediction arrays; the first prediction array includes a prediction marker s j and predicted data d j ; The predicted marker s j It includes a reactant marker, a reaction rule marker, and an end marker; the prediction marker s j When it is a reactant marker, the predicted data d j is a corresponding predicted fingerprint p pre,i ; The predicted marker s j When it is a reaction rule marker, the prediction data d j is a corresponding prediction type r pre,i ; The predicted marker s j When it is the end marker, the predicted data d j Is empty; if the predicted marker s collected i If the total number of predicted tags is less than the maximum tag index Q, the first prediction sequence is supplemented with a prediction array based on a preset invalid prediction array until the first prediction sequence is consistent with the sequence length of the current first tag sequence; if the collected prediction tag s i If the total number of is greater than the maximum label index Q, the current first label sequence is label-filled based on the invalid prediction array until the current first label sequence is consistent with the first prediction sequence and the maximum label index Q is reset based on the sequence length of the first prediction sequence;

[0063] Step 64: Substitute the first prediction sequence and the current first label sequence into a preset first model loss function L ALL ; and based on the preset first model optimizer, move towards making the first model loss function L ALL Performing a round of modulation on the model parameters of the encoder, the decoder, the reactant processing module and the composite processing module of the synthesis path regression prediction model in the direction of reaching the minimum value;

[0064] Among them, the first model loss function L ALL By the type loss function L S , molecular fingerprint loss function L P and the reaction rule loss function L R composition;

[0065] L ALL =L S +L P +L R ,

[0066]

[0067] Indicator function

[0068] Indicator function

[0069] L CE () is the cross entropy loss function, L BCE () is the binary cross entropy loss function;

[0070] The first model optimizer includes at least an SGD optimizer and an Adam optimizer;

[0071] Step 65, identifying whether the current training record is the last first data record of the first training set; if so, going to step 66; if not, extracting the next first data record of the first training set as the new current training record, and returning to step 63;

[0072] Step 66, 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 input the first training molecular graph of the current evaluation record as the corresponding original drug molecular graph into the synthetic pathway regression prediction model to optimize the molecular structure and perform regression prediction processing on the synthetic pathway of the optimized structure; and in the process of model processing, the prediction marker s generated by each step of the model prediction i Data collection is performed, and for each of the predicted markers s specifically as a reactant marker i The corresponding predicted fingerprint p pre,i Data collection is performed, and the prediction marker s for each reaction rule marker is i The corresponding prediction type r pre,i Perform data collection; and assemble a prediction sequence based on all collected data and the maximum label index Q of the first label sequence of the current evaluation record to obtain a corresponding second prediction sequence; and form a corresponding prediction-label sequence pair by the second prediction sequence and the current first label sequence; and at the end of this round of traversal, estimate the accuracy, precision, recall and F1 score based on all the obtained prediction-label sequence pairs to obtain the corresponding first accuracy, first precision, first recall and first F1 score;

[0073] Step 67, identifying the first accuracy, the first precision, the first recall and the first F1 score; if the first accuracy does not meet the preset first accuracy range, or the first accuracy does not meet the preset first accuracy range, or the first recall does not meet the preset first recall range, or the first F1 score does not meet the preset first F1 score range, return to step 62 to continue training; if the first accuracy meets the first accuracy range, and the first accuracy meets the first accuracy range, and the first recall meets the first recall range, and the first F1 score meets the first F1 score range, stop training and confirm that the model training is completed.

[0074] Preferably, the molecular structure optimization and synthesis path analysis of the old drug molecule graph based on the synthesis path regression prediction model to obtain the corresponding new drug molecule graph and new molecule synthesis path specifically include:

[0075] The old drug molecule graph is used as the corresponding original drug molecule graph and input into the synthesis pathway regression prediction model to perform molecular structure optimization, and the synthesis pathway of the optimized structure is subjected to regression prediction processing to obtain the corresponding optimized drug molecule graph and the optimized molecular synthesis pathway; and the optimized drug molecule graph is used as the corresponding new drug molecule graph, and the optimized molecular synthesis pathway is used as the corresponding new molecular synthesis pathway.

[0076] Preferably, the structural similarity and property difference of the new and old drug molecule graphs are calculated to obtain the corresponding new and old molecule calculation report, specifically including:

[0077] Step 81, based on a preset chemical calculation tool, the molecular fingerprint similarity of the new and old drug molecular graphs is measured to obtain the corresponding structural similarity;

[0078] Wherein, the chemical calculation tool at least includes RDKit tool and CDK tool;

[0079] Step 82, estimating multiple types of properties of the new and old drug molecules corresponding to the new and old drug molecule graphs based on a preset property calculation tool to obtain a corresponding list of properties of new and old molecules;

[0080] Wherein, the property calculation tools at least include SwissADME tool, admetSAR tool, ADMETlab 2.0 tool, ProTox-II tool, BioTransformer tool, MetaPred tool, CyProduct tool, and RDKit tool;

[0081] The new and old molecular property list includes a plurality of new and old molecular property records; the new and old molecular property records include property type, new molecular estimated value and old molecular estimated value; the property type includes at least multiple types of physicochemical properties, biological activity and toxicity; the multiple types of physicochemical properties include at least molecular mass, melting point, boiling point, solubility, lipophilicity, permeability and hydrogen bonding;

[0082] Step 83, a new and old molecule measurement report corresponding to the structural similarity and the new and old molecule property list is formed.

[0083] Preferably, the evaluation of the optimization effect according to the new and old molecular measurement reports to obtain the corresponding comprehensive evaluation results specifically includes:

[0084] Step 91, extracting the corresponding structural similarity and the list of properties of the new and old molecules from the new and old molecule measurement report;

[0085] Step 92, identifying whether the structural similarity exceeds a preset similarity threshold; if so, setting the corresponding structural evaluation result as structural similarity; if not, setting the corresponding structural evaluation result as significant structural differences;

[0086] Step 93, counting the total number of records of the new and old molecular property records in the new and old molecular property lists to obtain the corresponding total number of properties K ALL ; and initialize the first counter to 0; and perform a round of traversal of all the new and old molecular property records in the new and old molecular property list; and in this round of traversal, use the new and old molecular property records currently traversed as the corresponding current records; and use the property type, the new molecule estimated value and the old molecule estimated value of the current record as the corresponding current property type, the current new molecule property value and the current old molecule property value; and select the property quality assessment rule corresponding to the current property type from the preset property quality assessment rule library as the corresponding current assessment rule; and use the current assessment rule as the assessment standard, and assess whether the new drug molecule is better than the old drug molecule in the current property type according to the current new molecule property value and the current old molecule property value to obtain the corresponding current property assessment result; and add 1 to the first counter when the current property assessment result is worse than the old molecule;

[0087] Step 94, after the current round of traversal of all the records of the properties of the new and old molecules in the property list of the new and old molecules is completed, the first counter is identified; if the first counter is 0, the corresponding property evaluation result is set to be better than the old molecule in all aspects; if the first counter is greater than 0 but less than the total number of properties K ALL, then set the corresponding property evaluation result as partially better than the old molecule; if the first counter is greater than 0 and equal to the total number of properties K ALL , then the corresponding property evaluation result is set to be comprehensively inferior to the old molecule;

[0088] Step 95, composing the corresponding comprehensive evaluation result by the obtained structural evaluation result and the obtained property evaluation result.

[0089] The second aspect of the embodiment of the present invention provides a device for implementing the processing method for optimizing drug molecules described in the first aspect, the device comprising: a chemical information library processing module, a model building module, a model training module, a data receiving module, a model prediction module, a prediction evaluation module, an evaluation judgment module and a prediction output module;

[0090] The chemical information library processing module is used to create a corresponding building block information library by collecting real molecular fragments and / or unit structure information of various drug molecules; and to create a corresponding reaction rule information library by collecting real synthesis reaction information occurring during the synthesis of various drug molecules; and the corresponding chemical information library is composed of the building block information library and the reaction rule information library;

[0091] The model building module is used to build a synthetic pathway regression prediction model in combination with the chemical information library;

[0092] The model training module is used to construct a model training data set; and to train the synthesis path regression prediction model based on the model training data set;

[0093] The data receiving module is used to receive the generated drug molecule graph input by the user as the corresponding old drug molecule graph after the model training is completed;

[0094] The model prediction module is used to perform molecular structure optimization and synthesis path analysis on the old drug molecule graph based on the synthesis path regression prediction model to obtain the corresponding new drug molecule graph and new molecule synthesis path;

[0095] The prediction and evaluation module is used to calculate the structural similarity and property difference of the new and old drug molecule graphs to obtain the corresponding new and old molecule calculation report; and evaluate the optimization effect according to the new and old molecule calculation report to obtain the corresponding comprehensive evaluation result; the comprehensive evaluation result includes the structural evaluation result and the property evaluation result; the structural evaluation result includes the structural similarity and the obvious difference in structure; the property evaluation result includes the overall superiority to the old molecule, the partial superiority to the old molecule and the overall inferiority to the old molecule;

[0096] The evaluation judgment module is used to identify the structural evaluation result and the property evaluation result of the comprehensive evaluation result; if the structural evaluation result is that the structure is significantly different or the property evaluation result is inferior to the old molecule in all aspects, then return to the model prediction module for re-optimization; if the structural evaluation result is that the structure is similar and the property evaluation result is not inferior to the old molecule in all aspects, then go to the prediction output module;

[0097] The prediction output module is used to feed back the old drug molecule map and the latest new drug molecule map, the new molecule synthesis path and the new and old molecule measurement report to the current user.

[0098] A third aspect of an embodiment of the present invention provides an electronic device, including: a memory, a processor, and a transceiver;

[0099] 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;

[0100] The transceiver is coupled to the processor, and the processor controls the transceiver to send and receive messages.

[0101] 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.

[0102] The embodiments of the present invention provide a processing method, device, electronic device and computer-readable storage medium for optimizing drug molecules. As can be seen from the above content, the embodiments of the present invention first create a corresponding building block information library by collecting the real molecular fragments and / or unit structure information of various drug molecules, and create a corresponding reaction rule information library by collecting the real synthesis reaction information occurring during the synthesis of various drug molecules, and the building block information library and the reaction rule information library form a corresponding chemical information library; then, a synthetic pathway regression prediction model is constructed in combination with the chemical information library, and a model training data set is constructed to train the synthetic pathway regression prediction model; then, after the model training is completed, the generated drug molecule graph input by the user each time is used as the corresponding old drug molecule graph, and the molecular structure optimization and synthetic pathway analysis of the old drug molecule graph are performed based on the synthetic pathway regression prediction model to obtain the corresponding new drug molecule graph and new molecule synthesis. The synthesis pathway is optimized, and the structural similarity and property difference between the new and old drug molecule graphs are calculated to obtain the corresponding new and old molecule calculation reports. The optimization effect is evaluated based on the new and old molecule calculation reports to obtain the corresponding comprehensive evaluation results. When the structural evaluation result of the comprehensive evaluation result is that the structure has obvious differences or the property evaluation result is inferior to the old molecule in all aspects, the molecular structure optimization and synthesis pathway analysis are performed again based on the synthesis pathway regression prediction model until the latest structural evaluation result is that the structure is similar and the property evaluation result is not inferior to the old molecule in all aspects. When the structural evaluation result is that the structure is similar and the property evaluation result is not inferior to the old molecule in all aspects, the old drug molecule graph, the latest new drug molecule graph, the new molecule synthesis pathway and the new and old molecule calculation reports are fed back to the current user. The synthesis pathway regression prediction model provided by the embodiment of the present invention can optimize the structure of drug molecules on the premise of meeting the synthesizability requirements, and can predict the synthesis reaction pathway of the optimized drug molecules; the embodiment of the present invention not only improves the synthesizability of drug molecules, but also provides guidance and suggestions for the synthesis process of optimized new drug molecules, improves the drug synthesis efficiency, effectively guarantees the structural similarity of new and old drug molecules before and after optimization, and provides an intuitive comprehensive evaluation of the quality of the properties of new and old drug molecules before and after optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0103] Figure 1 A schematic diagram of a processing method for optimizing drug molecules provided in Example 1 of the present invention;

[0104] Figure 2 A module structure diagram of the synthesis path regression prediction model provided in Example 1 of the present invention;

[0105] Figure 3 A module structure diagram of a processing device for optimizing drug molecules provided in the second embodiment of the present invention;

[0106] Figure 4 A schematic diagram of the structure of an electronic device provided in Embodiment 3 of the present invention. DETAILED DESCRIPTION

[0107] 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.

[0108] Embodiment 1 of the present invention provides a method for optimizing drug molecules, such as Figure 1 As shown in the schematic diagram of a processing method for optimizing drug molecules provided in Example 1 of the present invention, the method mainly includes the following steps:

[0109] Step 1, by collecting the real molecular fragments and / or unit structure information of various drug molecules, a corresponding building block information library is created; and by collecting the real synthesis reaction information occurring during the synthesis of various drug molecules, a corresponding reaction rule information library is created; and the building block information library and the reaction rule information library form a corresponding chemical information library.

[0110] Here, the building block information library of the embodiment of the present invention includes multiple building block information records; each building block information record corresponds to a real molecular fragment or a real molecular unit structure; the building block information record includes a building block identifier, a building block molecular graph and a building block molecular fingerprint; the building block identifier is a unique identifier of the current building block; the building block molecular graph is a molecular graph of the current building block; the building block molecular fingerprint is a molecular fingerprint of the current building block;

[0111] The reaction rule information library of the embodiment of the present invention includes multiple reaction rule information records; each reaction rule information record corresponds to a type of chemical reaction that occurs during the synthesis of a type of drug molecule; the reaction rule information record includes a reaction rule type, an optional range of reactant A and an optional range of reactant B; the optional ranges of reactant A / B each contain multiple reactant identifiers; each reactant identifier in the reaction rule information library corresponds to a building block identifier in the building block information library.

[0112] Step 2: Construct a synthetic pathway regression prediction model in combination with the chemical information library.

[0113] Here, the synthetic pathway regression prediction model of the embodiment of the present invention is used to optimize the molecular structure of the original drug molecule graph input by the model, perform regression prediction processing on the synthetic pathway of the optimized structure, and output the corresponding optimized drug molecule graph and optimized molecular synthetic pathway. The model input end of the synthetic pathway regression prediction model is used to receive the original drug molecule graph input by the model, and the model output end is used to output the corresponding optimized drug molecule graph and optimized molecular synthetic pathway, such as Figure 2 The module structure diagram of the synthesis path regression prediction model provided in Example 1 of the present invention is shown.

[0114] like Figure 2 As shown, the model components of the synthetic pathway regression prediction model include: encoder, decoder, feedback output module, reactant processing module, synthetic compound processing module, chemical information library and molecular graph stack.

[0115] The connection relationship of the model components is: the input end of the encoder is connected to the input end of the model, and the output end is connected to the first input end of the decoder; the second and third input ends of the decoder are respectively connected to the output end of the reactant processing module and the output end of the composite processing module, and the first and second output ends of the decoder are respectively connected to the first and second input ends of the feedback output module; the first, second and third output ends of the feedback output module are respectively connected to the input end of the reactant processing module, the input end of the composite processing module, and the output end of the model; the reactant processing module and the composite processing module are also respectively connected to the chemical information library; the feedback output module, the reactant processing module and the composite processing module are also respectively connected to the molecular graph stack.

[0116] The component functions of the model components are described below.

[0117] 1) Encoder:

[0118] The encoder of the embodiment of the present invention is implemented based on the Uni-Mol model. The encoder is used to perform atomic level embedding coding on the original drug molecule graph to obtain the corresponding coding tensor X and send it to the decoder.

[0119] Here, the Uni-Mol model is a three-dimensional molecular graph encoding model implemented by an Encoder model based on the Transformer model. The Uni-Mol model is an encoding model that can perform atomic-level embedding encoding on molecular graphs. For details, please refer to the technical document "Uni-Mol: A Universal 3D Molecular Representation Learning Framework". I will not elaborate on it again.

[0120] 2) Decoder:

[0121] The decoder of the embodiment of the present invention is implemented based on the Decoder model of the Transformer model. Figure 2 As shown, similar to the standard structure of the Decoder model based on the Transformer model, the decoder is composed of an embedded encoding module, N decoding modules, a linear layer and a Softmax layer connected in sequence; wherein the preset total number of decoding layers N is a positive integer.

[0122] The connection relationship of each component of the decoder is as follows: the first and second input ends of the embedded coding module are connected to the second and third input ends of the decoder respectively; the first input end of the first decoding module is connected to the output end of the embedded coding module; except for the first decoding module, the first input end of each decoding module is connected to the output end of the previous decoding module; the second input end of each decoding module is connected to the first input end of the decoder; the output end of the last decoding module is respectively connected to the input end of the linear layer and the first output end of the decoder; the output end of the linear layer is connected to the input end of the Softmax layer; the output end of the Softmax layer is connected to the second output end of the decoder.

[0123] The functional description of each decoder component is shown below.

[0124] a. Embedded coding module:

[0125] The embedded coding module is used to initialize the corresponding prediction marker s0 based on the preset start marker before the first step prediction; and to set the corresponding position encoding vector PE based on the preset position embedding coding rule during the i-th step prediction. i , the single-step index i is an integer greater than or equal to 1; and the predicted marker s corresponding to the i-1th step is obtained i-1 For identification; if the predicted marker s i-1 is the start marker, then the encoding vector e is embedded based on the preset start marker start Set the corresponding encoding vector e i ; If the predicted token s i-1 is the reactant marker, a preset first MLP network is used to predict the marker s i-1 The corresponding reactant molecular fingerprint p i-1 Perform embedded coding to obtain the corresponding encoding vector e i ; If the predicted token s i-1 is a reaction rule marker, then the prediction marker s is embedded in the encoding rule based on the preset reaction rule. i-1 The corresponding reaction rule type r i-1 Perform embedded coding to obtain the corresponding encoding vector e i ; And the position encoding vector PE obtained this time i and the encoding vector ei The starting decoded tensor H that constitutes the i-th step i,0 Send to the first decoding module.

[0126] Here, the starting decoded tensor H i,0 The encoding process is:

[0127] H i,0 =PE i +e i ,

[0128]

[0129] Among them, f MLP1 () is the model inference function corresponding to the first MLP network, θ MLP1 are the model parameters of the first MLP network; Embed() is the reaction rule embedding encoding rule.

[0130] b. N decoding modules:

[0131] The first decoding module is used to decode the tensor H based on the encoding tensor X and the starting tensor i,0 Decode to get the corresponding decoding tensor H i,1 Send to the second decoding module; the second decoding module is used to decode the tensor according to the encoding tensor X and the decoding tensor H i,1 Decode to get the corresponding decoding tensor H i,2 Send to the third decoding module; and so on, the Nth decoding module is used to decode the tensor H sent by the encoding tensor X and the N-1th decoding module i,N-1 Decode to get the corresponding decoding tensor H i,N Send to the linear layer and feedback output module.

[0132] c. Linear layer:

[0133] The linear layer is used to decode the tensor H i,N Perform full connection calculation to obtain the corresponding full connection vector L i Send to the Softmax layer.

[0134] d. Softmax layer:

[0135] The Softmax layer is used to calculate the fully connected vector L i Predict the i-th step marker to get the corresponding predicted marker s i Sent to the feedback output module. Among them, the predicted tag s i It includes reactant marker, reaction rule marker and end marker.

[0136] 3) Feedback output module:

[0137] The feedback output module of the embodiment of the present invention is used to predict the prediction marker s in the i-th step. i When it is a reactant marker, the marker s is predicted i and the corresponding decoded tensor H i,N Sent to the reactant processing module.

[0138] The feedback output module is also used to predict the predicted token s at step i. i The predicted token s is the response rule token. i and the corresponding decoded tensor H i,N Send to the composite processing module.

[0139] The feedback output module is also used to receive a predicted marker s i When , the current time is used as the corresponding timestamp t i , and the timestamp t obtained this time i and the corresponding predicted token s i A corresponding first process record is composed and cached.

[0140] The feedback output module is also used to read out the molecular graph pushed into the stack as the corresponding reactant molecular graph g through a stack read operation each time the molecular graph stack completes a stack push operation. i , and use the current time as the corresponding timestamp t i , and the timestamp t obtained this time i and the corresponding reactant molecule graph g i A corresponding first process record is composed and cached.

[0141] The feedback output module is also used to predict the predicted token s at step i. i As the end marker, a pop operation is performed on the molecular graph stack to obtain a popped molecular graph, and the current popped molecular graph is used as the corresponding optimized drug molecular graph; and all the first process records in the cache are sorted by timestamp t i The order from front to back forms a corresponding first process record sequence, and the first process record sequence is used as the corresponding optimized molecular synthesis path; the optimized drug molecule graph and the optimized molecular synthesis path are sent to the model output end; and after the sending is completed, all cached first process records are deleted.

[0142] 4) Reactant processing module:

[0143] The reactant processing module of the embodiment of the present invention is used for receiving a set of prediction markers s i and decoded tensor H i,N When , a preset second MLP network is used according to the decoded tensor H i,NPerform molecular fingerprint prediction processing to obtain the corresponding predicted fingerprint p pre,i ; and compare the molecular fingerprints of each building block in the chemical information library with the predicted fingerprint p pre,i The similarity is calculated to obtain the corresponding fingerprint similarity; the building block molecular graph and the building block molecular fingerprint of the building block information record corresponding to the largest fingerprint similarity are extracted as the corresponding current molecular graph and current molecular fingerprint; and the current molecular graph is pushed into the molecular graph stack by performing a push operation on the molecular graph stack; and the current molecular fingerprint is used as the corresponding reactant molecular fingerprint p i ; and predict the token s i and the reactant molecular fingerprint p i Send to the embedded coding module.

[0144] Here, the predicted fingerprint p pre,i The prediction process is:

[0145] p pre,i =sigmoid(f MLP2 (H i,N θ MLP2 ));

[0146] Among them, the sigmoid() function is the activation function of the second MLP network, f MLP2 () is the model inference function corresponding to the second MLP network, θ MLP2 are the model parameters of the second MLP network.

[0147] 5) Synthetic processing module:

[0148] The composition processing module of the embodiment of the present invention is used for receiving a set of prediction markers s i and decoded tensor H i,N When , a preset third MLP network is used according to the decoded tensor H i,N Perform reaction rule type prediction processing to obtain the corresponding prediction type r pre,i ; and perform two pop operations on the molecular graph stack to obtain two popped molecular graphs as the corresponding A and B molecular graphs; and identify whether both A and B molecular graphs are not empty; if at least one of the A and B molecular graphs is empty, the prediction type r is based on the preset invalid reaction rule type. pre,i Reset; if both A and B molecular graphs are not empty, the reaction rule type and prediction type r in the reaction rule information library of the chemical information library are reset. pre,i The matched reaction rule information record is recorded as the corresponding current record, and whether the current record is empty is identified; if the current record is empty, the prediction type r is based on the invalid reaction rule type. pre,ireset; if the current record is not empty, the building block identifiers corresponding to the A and B molecular graphs are recorded as the corresponding A and B identifiers, and the A and B identifiers and the optional range of reactant A and the optional range of reactant B in the current record are identified; if the A identifier satisfies the optional range of reactant A but the B identifier does not satisfy the optional range of reactant B, or the A identifier satisfies the optional range of reactant B but the B identifier does not satisfy the optional range of reactant A, or the B identifier satisfies the optional range of reactant A but the A identifier does not satisfy the optional range of reactant B, or the B identifier satisfies the optional range of reactant B but the A identifier does not satisfy the optional range of reactant A, or both the A and B identifiers do not satisfy the optional range of reactant A / B, then the prediction type r is based on the invalid reaction rule type. pre,i reset; and in the prediction type r pre,i If it is not an invalid reaction rule type, the RDKit tool is used to predict the type r pre,i For the current reaction rule, a synthetic compound prediction calculation is performed with A and B molecular graphs as the current reactant combination, and the synthetic molecular graph obtained by this prediction calculation is used as the corresponding current molecular graph, and the current molecular graph is pushed into the molecular graph stack by performing a stacking operation on the molecular graph stack; and the prediction type r pre,i As the corresponding reaction rule type r i ; and predict the token s i and reaction rule type r i Send to the embedded coding module.

[0149] Here, the prediction type r pre,i The prediction process is:

[0150] r pre,i =softmax(f MLP3 (H i,N θ MLP3 ));

[0151] Among them, the softmax() function is the activation function of the third MLP network, f MLP3 () is the model inference function corresponding to the third MLP network, θ MLP3 are the model parameters of the third MLP network.

[0152] 6) Molecular graph stack:

[0153] The molecular graph stack of the embodiment of the present invention is used to store the molecular graph pushed into the stack and place it on the top of the stack when processing the push operation, and to shift the other molecular graphs one step toward the bottom of the stack in sequence when the total number of other molecular graphs in the stack is not zero.

[0154] The molecular graph stack is also used to output the molecular graph at the top of the stack as the corresponding popped molecular graph when processing a pop operation and remove the currently output molecular graph from the stack at the same time, and when the total number of remaining molecular graphs in the stack is not zero, the remaining molecular graphs are shifted one step toward the top of the stack in sequence.

[0155] The molecular graph stack is also used to output the molecular graph at the top of the stack as the corresponding read molecular graph when processing a stack read operation but without removing or shifting any molecular graph in the stack.

[0156] Step 3, constructing a model training data set; and training a synthetic path regression prediction model based on the model training data set;

[0157] Specifically including: step 31, constructing a model training data set;

[0158] Here, the model training data set of the embodiment of the present invention includes a plurality of first data records; the first data record includes a first training molecular graph and a first label sequence; the first label sequence is composed of a plurality of first labels; the first label includes a label index j, a label identifier and label data The label index j is a positive integer; the label marker Including reactant marker, reaction rule marker, end marker; label marker When it is a reactant marker, the label data is a corresponding reactant molecular fingerprint label p tag,j ; Tag marker When the response rule marker is is a corresponding reaction rule type label r tag,j ; Tag marker When it is the end tag, the tag data is empty;

[0159] The current step 31 specifically includes:

[0160] Step 311, obtaining a plurality of first collected data groups by collecting drug molecule graphs and corresponding synthesis path information from a public drug molecule synthesis database;

[0161] Among them, the first collected data group includes the first drug molecular map and the first synthesis path; the first synthesis path is formed by sequentially sorting a plurality of first path point information; the first path point information includes the first reactant molecular map, the second reactant molecular map, the first reactant molecular fingerprint, the second reactant molecular fingerprint, the synthetic product molecular map, the synthetic product molecular fingerprint and the synthetic reaction rule type; the first reactant molecular fingerprint is the molecular fingerprint of the first reactant molecular map; the second reactant molecular fingerprint is the molecular fingerprint of the second reactant molecular map; the synthetic product molecular fingerprint is the molecular fingerprint of the synthetic product molecular map; the synthetic reaction rule type satisfies the reaction rule type range of the chemical information library; in the first and second reactant molecular maps in the latter first path point information, there is and only one reactant molecular map that is the synthetic product molecular map in the previous first path point information; the synthetic product molecular map in the last first path point information matches the first drug molecular map;

[0162] Step 312, taking the first first collected data group as the corresponding current data group;

[0163] Step 313, taking the first drug molecule graph and the first synthesis path of the current data group as the corresponding current drug molecule graph and the current synthesis path; taking the first first path point information of the current synthesis path as the corresponding current path point information; initializing the first index to 1; and initializing an empty sequence as the corresponding first label sequence;

[0164] Step 314: Set a tag index j as the first index; and set a corresponding tag marker for the current tag index j Set a corresponding label data for the reactant marker is the first reactant molecular fingerprint of the current path point information; and the label index j obtained this time and its corresponding label marker and label data Form a corresponding first tag and add it to the first tag sequence; and add 1 to the first index;

[0165] Step 315: Set a tag index j as the first index; and set a corresponding tag marker for the current tag index j Set a corresponding label data for the reactant marker is the first reactant molecular fingerprint of the current path point information; and the label index j obtained this time and its corresponding label marker and label data Form a corresponding first tag and add it to the first tag sequence; and add 1 to the first index;

[0166] Step 316: Set a tag index j as the first index; and set a corresponding tag marker for the current tag index j Set a corresponding label data for the reaction rule marker is the synthetic reaction rule type of the current path point information; and the label index j obtained this time and its corresponding label marker and label data Form a corresponding first tag and add it to the first tag sequence; and add 1 to the first index;

[0167] Step 317, identify whether the current path point information is the last first path point information of the current synthetic path; if so, set a label index j as the first index, and set a corresponding label marker for the current label index j End marker and set a corresponding tag data for the current tag index j is empty, and the label index j obtained this time and its corresponding label marker and label data A corresponding first tag is formed and added to the first tag sequence, and the current drug molecule graph is used as a corresponding first training molecule graph, and the first training molecule graph corresponding to the current synthesis path and the first tag sequence are formed into a corresponding first data record, and the process goes to step 318; if not, the next first path point information of the current synthesis path is extracted as the new current path point information, and the process goes back to step 314;

[0168] Step 318, identifying whether the current data group is the last first collected data group; if not, taking the next first collected data group as the new current data group and returning to step 313; if so, forming a corresponding model training data set from all the obtained first data records;

[0169] Step 32, training the synthesis path regression prediction model based on the model training data set;

[0170] Specifically, it includes: step 321, dividing the model training data set into two sub-data sets based on a preset first division ratio and recorded as a corresponding first training set and a first evaluation set;

[0171] The first segmentation ratio is a preset ratio parameter, such as 9:1; 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 of the first training set to the first evaluation set satisfies the first segmentation ratio;

[0172] Step 322, taking the first first data record of the first training set as the corresponding current training record;

[0173] Step 323, input the first training molecular graph of the current training record as the corresponding original drug molecular graph into the synthetic pathway regression prediction model to optimize the molecular structure and perform regression prediction processing on the synthetic pathway of the optimized structure; and in the process of model processing, the prediction marker s generated by each step of the model prediction is generated. i Perform data collection and predict markers for each specific reactant marker i The corresponding predicted fingerprint p pre,i Perform data collection and predict markers for each specific reaction rule marker i Corresponding prediction type r pre,i Perform data collection; and assemble the prediction sequence based on all collected data and the maximum label index Q of the first label sequence of the current training record to obtain the corresponding first prediction sequence;

[0174] Here, Q = max(j); the first prediction sequence is composed of a plurality of first prediction arrays; the first prediction array includes a prediction marker s j and predicted data d j ; Prediction tokens j Including reactant markers, reaction rule markers, end markers; prediction markers j When it is the reactant marker, the predicted data d j is a corresponding predicted fingerprint p pre,i ; Prediction tokens j When it is a reaction rule marker, the predicted data d j is a corresponding prediction type r pre,i ; Prediction tokens j When it is the end marker, the predicted data d j Empty; if the collected prediction marker s i If the total number of is less than the maximum label index Q, the first prediction sequence is predicted based on the preset invalid prediction array until the first prediction sequence is consistent with the sequence length of the current first label sequence; if the collected prediction marker s i If the total number of is greater than the maximum label index Q, the current first label sequence is padded based on the invalid prediction array until the current first label sequence is consistent with the first prediction sequence and the maximum label index Q is reset based on the sequence length of the first prediction sequence;

[0175] Step 324: Substitute the first prediction sequence and the current first label sequence into the preset first model loss function L ALL ; and based on the preset first model optimizer, move towards making the first model loss function L ALLA round of modulation is performed on the model parameters of the encoder, decoder, reactant processing module and compound processing module of the synthesis path regression prediction model in the direction of reaching the minimum value;

[0176] Here, the first model loss function L of the embodiment of the present invention is ALL By the type loss function L S , molecular fingerprint loss function L P and the reaction rule loss function L R The composition is specifically:

[0177] L ALL =L S +L P +L R ,

[0178]

[0179] Indicator function

[0180] Indicator function

[0181] L CE () is the cross entropy loss function, L BCE () is the binary cross entropy loss function;

[0182] The first model optimizer includes at least an SGD optimizer and an Adam optimizer;

[0183] Step 325, identifying whether the current training record is the last first data record of the first training set; if so, go to step 326; if not, extract the next first data record of the first training set as the new current training record, and return to step 323;

[0184] Step 326, 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 input the first training molecular graph of the current evaluation record as the corresponding original drug molecular graph into the synthetic pathway regression prediction model to optimize the molecular structure and perform regression prediction processing on the synthetic pathway of the optimized structure; and in the process of model processing, the prediction marker s generated by each step of the model prediction i Perform data collection and predict markers for each specific reactant marker i The corresponding predicted fingerprint p pre,i Perform data collection and predict markers for each specific reaction rule marker i Corresponding prediction type r pre,iPerform data collection; and assemble the prediction sequence based on all the collected data and the maximum label index Q of the first label sequence of the current evaluation record to obtain the corresponding second prediction sequence; and form a corresponding prediction-label sequence pair by the second prediction sequence and the current first label sequence; and at the end of this round of traversal, estimate the accuracy, precision, recall and F1 score based on all the obtained prediction-label sequence pairs to obtain the corresponding first accuracy, first precision, first recall and first F1 score;

[0185] Step 327, identify the first accuracy, the first precision, the first recall rate and the first F1 score; if the first accuracy does not meet the preset first accuracy range, or the first accuracy does not meet the preset first accuracy range, or the first recall rate does not meet the preset first recall rate range, or the first F1 score does not meet the preset first F1 score range, return to step 322 to continue training; if the first accuracy meets the first accuracy range, and the first accuracy meets the first accuracy range, and the first recall rate meets the first recall rate range, and the first F1 score meets the first F1 score range, stop training and confirm that the model training is completed.

[0186] Here, the first accuracy range, the first precision range, the first recall range, and the first F1 score range are four preset value ranges.

[0187] Step 4: After the model training is completed, the generated drug molecule graph input by the user is received as the corresponding old drug molecule graph.

[0188] Step 5, based on the synthetic pathway regression prediction model, the molecular structure optimization and synthetic pathway analysis of the old drug molecular graph are performed to obtain the corresponding new drug molecular graph and new molecular synthetic pathway;

[0189] Specifically, the method includes: inputting the old drug molecular graph as the corresponding original drug molecular graph into the synthesis pathway regression prediction model to optimize the molecular structure and performing regression prediction processing on the synthesis pathway of the optimized structure to obtain the corresponding optimized drug molecular graph and optimized molecular synthesis pathway; and using the optimized drug molecular graph as the corresponding new drug molecular graph and the optimized molecular synthesis pathway as the corresponding new molecular synthesis pathway.

[0190] Step 6, calculating the structural similarity and property difference of the new and old drug molecular graphs to obtain the corresponding new and old molecular calculation reports; and evaluating the optimization effect according to the new and old molecular calculation reports to obtain the corresponding comprehensive evaluation results;

[0191] Among them, the comprehensive evaluation results include structural evaluation results and property evaluation results; structural evaluation results include similar structures and obvious differences in structures; property evaluation results include overall superiority to old molecules, partial superiority to old molecules, and overall inferiority to old molecules;

[0192] Specifically, it includes: step 61, calculating the structural similarity and property difference between the new and old drug molecule graphs to obtain a corresponding new and old molecule calculation report;

[0193] Specifically, the method comprises: step 611, calculating the molecular fingerprint similarity of the new and old drug molecular graphs based on a preset chemical calculation tool to obtain the corresponding structural similarity;

[0194] Among them, chemical calculation tools include at least RDKit tools and CDK tools;

[0195] Step 612, estimating multiple types of properties of new and old drug molecules corresponding to the new and old drug molecule graphs based on a preset property calculation tool to obtain a corresponding list of properties of new and old molecules;

[0196] Among them, the property calculation tools include at least SwissADME tool, admetSAR tool, ADMETlab 2.0 tool, ProTox-II tool, BioTransformer tool, MetaPred tool, CyProduct tool, and RDKit tool;

[0197] The list of new and old molecular properties includes multiple records of new and old molecular properties; the records of new and old molecular properties include property type, new molecular estimated value and old molecular estimated value; property type at least includes multiple types of physicochemical properties, biological activity and toxicity; multiple types of physicochemical properties at least include molecular mass, melting point, boiling point, solubility, lipophilicity, permeability and hydrogen bonding;

[0198] Step 613, a new and old molecule calculation report corresponding to the structure similarity and the list of properties of the new and old molecules is formed;

[0199] Step 62, evaluating the optimization effect according to the new and old molecular measurement reports to obtain a corresponding comprehensive evaluation result;

[0200] Specifically, it includes: step 621, extracting the corresponding structural similarity and the list of properties of the new and old molecules from the new and old molecule measurement reports;

[0201] Step 622, identifying whether the structural similarity exceeds a preset similarity threshold; if so, setting the corresponding structural evaluation result as structural similarity; if not, setting the corresponding structural evaluation result as significant structural differences;

[0202] Here, the similarity threshold is a preset threshold parameter;

[0203] Step 623, the total number of records of the new and old molecular property records in the new and old molecular property lists is counted to obtain the corresponding total number of properties K ALL; and initialize the first counter to 0; and perform a round of traversal of all new and old molecular property records in the new and old molecular property list; and in this round of traversal, the new and old molecular property records currently traversed are used as the corresponding current records; and the property type, new molecule estimated value and old molecule estimated value of the current record are used as the corresponding current property type, current new molecule property value and current old molecule property value; and select the property quality evaluation rule corresponding to the current property type from the preset property quality evaluation rule library as the corresponding current evaluation rule; and use the current evaluation rule as the evaluation standard, and evaluate whether the new drug molecule is better than the old drug molecule in the current property type according to the current new molecule property value and the current old molecule property value to obtain the corresponding current property evaluation result; and when the current property evaluation result is inferior to the old molecule, add 1 to the first counter;

[0204] Here, the property quality evaluation rule base is a pre-set rule base, which is composed of multiple property quality evaluation rules. Each property quality evaluation rule corresponds to a type of physicochemical property index of the drug molecule (such as molecular weight, melting point, boiling point, solubility, lipophilicity, permeability, hydrogen bonding, etc.), a type of biological activity index or a type of toxicity index; each property quality evaluation rule will specify the corresponding relationship between the change trend of the measured value of the corresponding property (such as increasing, decreasing, etc.) and the change trend of the property quality level (such as better, worse, etc.); based on a certain property The merit evaluation rule is the evaluation standard. When evaluating whether the new drug molecule is better than the old drug molecule in the current property type according to the current new molecule property value and the current old molecule property value, it is only necessary to first calculate the difference between the current new / old molecule property values, and confirm the current new / old molecule's measured value change trend through the positivity / negativity of the difference, and then evaluate whether the current new molecule is better than the current old molecule based on the corresponding relationship between the measured value change trend and the merit level change trend given by the current property merit evaluation rule, so as to obtain the corresponding current property evaluation result;

[0205] Step 624, after the current round of traversal of all the new and old molecular property records in the new and old molecular property lists is completed, the first counter is identified; if the first counter is 0, the corresponding property evaluation result is set to be better than the old molecule in all aspects; if the first counter is greater than 0 but less than the total number of properties K ALL , then set the corresponding property evaluation result as partially better than the old molecule; if the first counter is greater than 0 and equal to the total number of properties K ALL , then the corresponding property evaluation result is set to be comprehensively inferior to the old molecule;

[0206] Step 625, composing a corresponding comprehensive evaluation result from the obtained structural evaluation result and property evaluation result.

[0207] Step 7, identify the structural evaluation results and property evaluation results of the comprehensive evaluation results; if the structural evaluation results show that the structure is significantly different or the property evaluation results are inferior to the old molecule in all aspects, return to step 5 for re-optimization; if the structural evaluation results show that the structure is similar and the property evaluation results are not inferior to the old molecule in all aspects, go to step 8.

[0208] Step 8, feeding back the old drug molecule map and the latest new drug molecule map, the new molecule synthesis path and the new and old molecule measurement report to the current user.

[0209] Figure 3 The module structure diagram of a processing device for optimizing drug molecules 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 3 As shown, the device includes: a chemical information library processing module 201, a model building module 202, a model training module 203, a data receiving module 204, a model prediction module 205, a prediction evaluation module 206, an evaluation judgment module 207 and a prediction output module 208.

[0210] The chemical information library processing module 201 is used to create a corresponding building block information library by collecting the real molecular fragments and / or unit structure information of various drug molecules; and to create a corresponding reaction rule information library by collecting the real synthesis reaction information occurring during the synthesis of various drug molecules; and the corresponding chemical information library is composed of the building block information library and the reaction rule information library.

[0211] The model building module 202 is used to build a synthetic pathway regression prediction model in combination with a chemical information library.

[0212] The model training module 203 is used to construct a model training data set; and to train the synthetic path regression prediction model based on the model training data set.

[0213] The data receiving module 204 is used to receive the generated drug molecule graph input by the user as the corresponding old drug molecule graph after the model training is completed.

[0214] The model prediction module 205 is used to perform molecular structure optimization and synthesis path analysis on the old drug molecule graph based on the synthesis path regression prediction model to obtain the corresponding new drug molecule graph and new molecule synthesis path.

[0215] The prediction and evaluation module 206 is used to calculate the structural similarity and property differences between the new and old drug molecule graphs to obtain corresponding new and old molecule measurement reports; and to evaluate the optimization effect based on the new and old molecule measurement reports to obtain corresponding comprehensive evaluation results; the comprehensive evaluation results include structural evaluation results and property evaluation results; the structural evaluation results include structural similarity and significant structural differences; the property evaluation results include overall superiority to the old molecule, partial superiority to the old molecule, and overall inferiority to the old molecule.

[0216] The evaluation judgment module 207 is used to identify the structural evaluation results and property evaluation results of the comprehensive evaluation results; if the structural evaluation result is that the structure has obvious differences or the property evaluation result is inferior to the old molecule in all aspects, it returns to the model prediction module 205 for re-optimization; if the structural evaluation result is that the structure is similar and the property evaluation result is not inferior to the old molecule in all aspects, it goes to the prediction output module 208.

[0217] The prediction output module 208 is used to feed back the old drug molecule map and the latest new drug molecule map, the new molecule synthesis path and the new and old molecule measurement report to the current user.

[0218] An embodiment of the present invention provides a processing device for optimizing drug molecules, 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.

[0219] 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 chemical information library processing 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 software form.

[0220] 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).

[0221] 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 tape), an optical medium (such as a DVD), or a semiconductor medium (such as a solid state disk (SSD)), etc.

[0222] Figure 4 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 4As 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.

[0223] exist Figure 4 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.

[0224] The above-mentioned processor can be a general-purpose processor, including a central processing unit CPU, a network processor (Network Processor, NP), a graphics processing unit (Graphics Processing Unit, GPU), etc.; it can also be a digital signal processor DSP, an application-specific integrated circuit ASIC, a field programmable gate array FPGA or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0225] 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.

[0226] The embodiments of the present invention provide a processing method, device, electronic device and computer-readable storage medium for optimizing drug molecules. As can be seen from the above content, the embodiments of the present invention first create a corresponding building block information library by collecting the real molecular fragments and / or unit structure information of various drug molecules, and create a corresponding reaction rule information library by collecting the real synthesis reaction information occurring during the synthesis of various drug molecules, and the building block information library and the reaction rule information library form a corresponding chemical information library; then, a synthetic pathway regression prediction model is constructed in combination with the chemical information library, and a model training data set is constructed to train the synthetic pathway regression prediction model; then, after the model training is completed, the generated drug molecule graph input by the user each time is used as the corresponding old drug molecule graph, and the molecular structure optimization and synthetic pathway analysis of the old drug molecule graph are performed based on the synthetic pathway regression prediction model to obtain the corresponding new drug molecule graph and new molecule synthesis. The synthesis pathway is optimized, and the structural similarity and property difference between the new and old drug molecule graphs are calculated to obtain the corresponding new and old molecule calculation reports. The optimization effect is evaluated based on the new and old molecule calculation reports to obtain the corresponding comprehensive evaluation results. When the structural evaluation result of the comprehensive evaluation result is that the structure has obvious differences or the property evaluation result is inferior to the old molecule in all aspects, the molecular structure optimization and synthesis pathway analysis are performed again based on the synthesis pathway regression prediction model until the latest structural evaluation result is that the structure is similar and the property evaluation result is not inferior to the old molecule in all aspects. When the structural evaluation result is that the structure is similar and the property evaluation result is not inferior to the old molecule in all aspects, the old drug molecule graph, the latest new drug molecule graph, the new molecule synthesis pathway and the new and old molecule calculation reports are fed back to the current user. The synthesis pathway regression prediction model provided by the embodiment of the present invention can optimize the structure of drug molecules on the premise of meeting the synthesizability requirements, and can predict the synthesis reaction pathway of the optimized drug molecules; the embodiment of the present invention not only improves the synthesizability of drug molecules, but also provides guidance and suggestions for the synthesis process of optimized new drug molecules, improves the drug synthesis efficiency, effectively guarantees the structural similarity of new and old drug molecules before and after optimization, and provides an intuitive comprehensive evaluation of the quality of the properties of new and old drug molecules before and after optimization.

[0227] 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.

[0228] 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 optimizing drug molecules, characterized in that: The method comprises: Step 1, creating a corresponding building block information library by collecting real molecular fragments and / or unit structure information of various drug molecules; and creating a corresponding reaction rule information library by collecting real synthesis reaction information occurring during the synthesis of various drug molecules; and forming a corresponding chemical information library by the building block information library and the reaction rule information library; Step 2, constructing a synthetic pathway regression prediction model in combination with the chemical information library; Step 3, constructing a model training data set; and training the synthesis path regression prediction model based on the model training data set; Step 4: After the model training is completed, the generated drug molecule graph input by the user is received as the corresponding old drug molecule graph; Step 5, performing molecular structure optimization and synthesis path analysis on the old drug molecular graph based on the synthesis path regression prediction model to obtain a corresponding new drug molecular graph and a new molecular synthesis path; Step 6, calculating the structural similarity and property difference of the new and old drug molecule graphs to obtain a corresponding new and old molecule calculation report; and evaluating the optimization effect according to the new and old molecule calculation report to obtain a corresponding comprehensive evaluation result; The comprehensive evaluation results include structural evaluation results and property evaluation results; the structural evaluation results include similar structures and obvious differences in structures; the property evaluation results include overall superiority to the old molecule, partial superiority to the old molecule, and overall inferiority to the old molecule; Step 7, identifying the structure evaluation result and the property evaluation result of the comprehensive evaluation result; if the structure evaluation result is that the structure is significantly different or the property evaluation result is inferior to the old molecule in all aspects, returning to step 5 for re-optimization; if the structure evaluation result is that the structure is similar and the property evaluation result is not inferior to the old molecule in all aspects, going to step 8; Step 8, feeding back the old drug molecule map, the latest new drug molecule map, the new molecule synthesis pathway and the new and old molecule measurement report to the current user.

2. The method for optimizing drug molecules according to claim 1, characterized in that: The building block information library includes a plurality of building block information records; each of the building block information records corresponds to a real molecular fragment or a real molecular unit structure; the building block information record includes a building block identifier, a building block molecular graph and a building block molecular fingerprint; the building block identifier is a unique identifier of the current building block; the building block molecular graph is a molecular graph of the current building block; the building block molecular fingerprint is a molecular fingerprint of the current building block; The reaction rule information library includes multiple reaction rule information records; each of the reaction rule information records corresponds to a type of chemical reaction that occurs during the synthesis of a type of drug molecule; the reaction rule information record includes a reaction rule type, an optional range of reactant A, and an optional range of reactant B; the optional ranges of reactant A / B each contain multiple reactant identifiers; each of the reactant identifiers in the reaction rule information library corresponds to one of the building block identifiers in the building block information library.

3. The method for optimizing drug molecules according to claim 2, characterized in that: The synthetic pathway regression prediction model is used to optimize the molecular structure of the original drug molecule graph input by the model, perform regression prediction processing on the synthetic pathway of the optimized structure, and output the corresponding optimized drug molecule graph and optimized molecular synthetic pathway; The model input end of the synthetic pathway regression prediction model is used to receive the original drug molecule graph input by the model, and the model output end is used to output the corresponding optimized drug molecule graph and the optimized molecular synthetic pathway; The synthetic pathway regression prediction model includes an encoder, a decoder, a feedback output module, a reactant processing module, a synthetic processing module, a chemical information library and a molecular graph stack; The input end of the encoder is connected to the input end of the model, and the output end is connected to the first input end of the decoder; the second and third input ends of the decoder are respectively connected to the output end of the reactant processing module and the output end of the composite processing module, and the first and second output ends of the decoder are respectively connected to the first and second input ends of the feedback output module; the first, second and third output ends of the feedback output module are respectively connected to the input end of the reactant processing module, the input end of the composite processing module and the output end of the model; the reactant processing module and the composite processing module are also respectively connected to the chemical information library; the feedback output module, the reactant processing module and the composite processing module are also respectively connected to the molecular graph stack; The encoder is implemented based on the Uni-Mol model; the encoder is used to perform atomic level embedding coding on the original drug molecule graph to obtain a corresponding coding tensor X and send it to the decoder; The decoder is implemented based on the Decoder model of the Transformer model; the decoder is formed by sequentially connecting an embedded coding module, N decoding modules, a linear layer and a Softmax layer, and the preset total number of decoding layers N is a positive integer; the first and second input ends of the embedded coding module are respectively connected to the second and third input ends of the decoder; the first input end of the first decoding module is connected to the output end of the embedded coding module; except for the first decoding module, the first input end of each of the remaining decoding modules is connected to the output end of the previous decoding module; the second input end of each decoding module is connected to the first input end of the decoder; the output end of the last decoding module is respectively connected to the input end of the linear layer and the first output end of the decoder; the output end of the linear layer is connected to the input end of the Softmax layer; the output end of the Softmax layer is connected to the second output end of the decoder; The embedded coding module is used to initialize the corresponding prediction marker s0 based on the preset start marker before the first step prediction; and to set the corresponding position encoding vector PE based on the preset position embedded coding rule during the i-th step prediction. i , the single-step index i is an integer greater than or equal to 1; and the predicted marker s corresponding to the i-1th step is obtained i-1 Identify; if the predicted marker s i-1 is the start marker, then the encoding vector e is embedded based on the preset start marker start Set the corresponding encoding vector e i ; If the predicted marker s i-1 is a reactant marker, a preset first MLP network is used to predict the marker s i-1 The corresponding reactant molecular fingerprint p i-1 Perform embedded coding to obtain the corresponding encoding vector e i ; If the predicted marker s i-1 is a reaction rule marker, then the prediction marker s is embedded in the coding rule based on the preset reaction rule. i-1 The corresponding reaction rule type r i-1 Perform embedded coding to obtain the corresponding encoding vector e i ; and the position encoding vector PE obtained this time i and the encoding vector e i The starting decoded tensor H that constitutes the i-th step i,0 Sending to the first decoding module; A i,0 =PE i +e i , Among them, f MLP1 () is the model inference function corresponding to the first MLP network, θ MLP1 is the model parameter of the first MLP network; Embed() is the encoding rule for embedding the reaction rule; The first decoding module is used to decode the encoding tensor X and the starting decoding tensor H according to the encoding tensor X and the starting decoding tensor H i,0 Decode to get the corresponding decoding tensor H i,1 Send to the second decoding module; the second decoding module is used to generate a decoded tensor based on the encoding tensor X and the decoding tensor H i,1 Decode to get the corresponding decoding tensor H i,2 Send to the third decoding module; and so on, the Nth decoding module is used to decode the tensor H according to the encoding tensor X and the decoding tensor H sent by the N-1th decoding module i,N-1 Decode to get the corresponding decoding tensor H i,N Sending to the linear layer and the feedback output module; The linear layer is used to decode the tensor H i,N Perform full connection calculation to obtain the corresponding full connection vector L i Send to the Softmax layer; The Softmax layer is used to calculate the fully connected vector L i Predict the i-th step marker to get the corresponding predicted marker s i Send to the feedback output module; the predicted marker s i It includes reactant marker, reaction rule marker and end marker; The feedback output module is used to predict the predicted marker s in step i. i When the predicted marker s is a reactant marker, i and the corresponding decoded tensor H i,N Sending to the reactant processing module; The feedback output module is also used to predict the predicted marker s in step i. i is the reaction rule marker, and the prediction marker s i and the corresponding decoded tensor H i,N Sending to the composition processing module; The feedback output module is also used to receive a predicted marker s i When , the current time is used as the corresponding timestamp t i , and the timestamp t obtained this time i and the corresponding predicted marker s i composing a corresponding first process record and caching it; The feedback output module is also used to read out the molecular graph pushed into the stack as the corresponding reactant molecular graph g through a stack reading operation each time the molecular graph stack completes a stacking operation. i , and take the current time as the corresponding timestamp t i , and the timestamp t obtained this time i and the corresponding reactant molecule graph g i composing a corresponding record of the first process and caching it; The feedback output module is also used to predict the predicted marker s in step i. i as the end marker, performing a pop operation on the molecular graph stack to obtain a popped molecular graph, and using the current popped molecular graph as the corresponding optimized drug molecular graph; and performing a pop operation on all cached first process records according to the timestamp t i Sorting the sequence from front to back to form a corresponding first process record sequence, and using the first process record sequence as the corresponding optimized molecular synthesis path; sending the optimized drug molecule graph and the optimized molecular synthesis path to the model output end; and deleting all the first process records in the cache after the sending is completed; The reactant processing module is used to receive a set of prediction markers s i and the decoded tensor H i,N When a preset second MLP network is used according to the decoded tensor H i,N Perform molecular fingerprint prediction processing to obtain the corresponding predicted fingerprint p pre,i ; and compare the molecular fingerprints of each of the building blocks in the building block information library of the chemical information library with the predicted fingerprint p pre,i The similarity is calculated to obtain the corresponding fingerprint similarity; and the building block molecular graph and the building block molecular fingerprint of the building block information record corresponding to the largest fingerprint similarity are extracted as the corresponding current molecular graph and current molecular fingerprint; and the current molecular graph is pushed into the molecular graph stack by performing a push operation on the molecular graph stack; and the current molecular fingerprint is used as the corresponding reactant molecular fingerprint p i ; and the predicted marker s i and the reactant molecular fingerprint p i Sending to the embedded coding module; Among them, p pre,i =sigmoid(f MLP2 (H i,B θ MLP2 )); The sigmoid() function is the activation function of the second MLP network, f MLP2 () is the model inference function corresponding to the second MLP network, θ MLP2 are the model parameters of the second MLP network; The composition processing module is used for receiving a group of predicted markers s i and the decoded tensor H i,N When , a preset third MLP network is used according to the decoding tensor H i,N Perform reaction rule type prediction processing to obtain the corresponding prediction type r pre,i ; and perform two pop operations on the molecular graph stack to obtain two popped molecular graphs as corresponding A and B molecular graphs; and identify whether the A and B molecular graphs are both empty; if at least one of the A and B molecular graphs is empty, the prediction type r is determined based on the preset invalid reaction rule type. pre,i reset; if the A and B molecular graphs are not empty, then the reaction rule type in the reaction rule information library of the chemical information library and the prediction type r pre,i The matched reaction rule information record is recorded as the corresponding current record, and whether the current record is empty is identified; if the current record is empty, the prediction type r is changed based on the invalid reaction rule type. pre,i reset; if the current record is not empty, record the building block identifiers corresponding to the A and B molecular graphs as corresponding A and B identifiers, and identify the A and B identifiers and the reactant A optional range and the reactant B optional range of the current record; if the A identifier satisfies the reactant A optional range but the B identifier does not satisfy the reactant B optional range, or the A identifier satisfies the reactant B optional range but the B identifier does not satisfy the reactant A optional range, or the B identifier satisfies the reactant A optional range but the A identifier does not satisfy the reactant B optional range, or the B identifier satisfies the reactant B optional range but the A identifier does not satisfy the reactant A optional range, or the B identifier satisfies the reactant B optional range but the A identifier does not satisfy the reactant A optional range, or the A and B identifiers do not satisfy the reactant A / B optional range, then based on the invalid reaction rule type, the prediction type r pre,i Reset; and in the prediction type r pre,i If it is not the invalid reaction rule type, the prediction type r is used based on the RDKit tool. pre,i The current reaction rule is used to perform a synthetic compound prediction calculation with the A and B molecular graphs as the current reactant combination, and the synthetic compound molecular graph obtained by this prediction calculation is used as the corresponding current molecular graph, and the current molecular graph is pushed into the molecular graph stack by performing a stacking operation on the molecular graph stack; and the prediction type r pre,i As the corresponding reaction rule type r i ; and the predicted marker s i and the reaction rule type r i Sending to the embedded coding module; Among them, r pre,i =softmax(f MLP3 (H i,N θ MLP3 )); The softmax() function is the activation function of the third MLP network, f MLP3 () is the model inference function corresponding to the third MLP network, θ MLP3 are model parameters of the third MLP network; The molecular graph stack is used to store the molecular graph pushed into the stack and put it on the top of the stack when processing the push operation, and when the total number of other molecular graphs in the stack is not zero, the other molecular graphs are shifted one step toward the bottom of the stack in sequence; The molecular graph stack is also used to output the molecular graph at the top of the stack as the corresponding popped molecular graph and remove the currently output molecular graph from the stack when processing a pop operation, and when the total number of remaining molecular graphs in the stack is not zero, shift the remaining molecular graphs one step toward the top of the stack in sequence; The molecular graph stack is also used to output the molecular graph at the top of the stack as the corresponding read molecular graph when processing a stack read operation but not to remove or shift any molecular graph in the stack.

4. The method for optimizing drug molecules according to claim 3, characterized in that: The model training data set includes a plurality of first data records; The first data record includes a first training molecular graph and a first label sequence; The first tag sequence consists of a plurality of first tags; The first tag includes a tag index j, a tag marker and label data The tag index j is a positive integer; the tag marker It includes a reactant marker, a reaction rule marker, and an end marker; the label marker When it is a reactant tag, the tag data is a corresponding reactant molecular fingerprint label p tag,j ; The label marker When it is a reaction rule tag, the tag data is a corresponding reaction rule type label r tag,j ; The label marker When it is the end tag, the tag data Is empty.

5. The method for optimizing drug molecules according to claim 4, characterized in that: The constructing of the model training data set specifically includes: Step 51, obtaining a plurality of first collected data groups by collecting drug molecule graphs and corresponding synthesis path information from a public drug molecule synthesis database; Wherein, the first collected data group includes a first drug molecule graph and a first synthesis path; the first synthesis path is formed by sequentially sorting a plurality of first path point information; the first path point information includes a first reactant molecule graph, a second reactant molecule graph, a first reactant molecule fingerprint, a second reactant molecule fingerprint, a synthetic product molecule graph, a synthetic product molecule fingerprint and a synthetic reaction rule type; the first reactant molecule fingerprint is a molecule fingerprint of the first reactant molecule graph; the second reactant molecule fingerprint is a molecule fingerprint of the second reactant molecule graph; the synthetic product molecule fingerprint is a molecule fingerprint of the synthetic product molecule graph; the synthetic reaction rule type satisfies the reaction rule type range of the chemical information library; one and only one of the first and second reactant molecule graphs in the latter first path point information is the synthetic product molecule graph in the previous first path point information; the synthetic product molecule graph in the last first path point information matches the first drug molecule graph; Step 52, taking the first of the first collected data groups as the corresponding current data group; Step 53, taking the first drug molecule graph and the first synthesis path of the current data group as the corresponding current drug molecule graph and current synthesis path; taking the first first path point information of the current synthesis path as the corresponding current path point information; initializing the first index to 1; and initializing an empty sequence as the corresponding first label sequence; Step 54: Set the tag index j as the first index; and set a corresponding tag marker for the current tag index j. For the reactant marker, set a corresponding label data is the first reactant molecular fingerprint of the current path point information; and the label index j obtained this time and its corresponding label identifier and the label data Form a corresponding first tag and add it to the first tag sequence; and add 1 to the first index; Step 55: Set the tag index j as the first index; and set a corresponding tag marker for the current tag index j. For the reactant marker, set a corresponding label data is the first reactant molecular fingerprint of the current path point information; and the label index j obtained this time and its corresponding label identifier and the label data Form a corresponding first tag and add it to the first tag sequence; and add 1 to the first index; Step 56: Set the tag index j as the first index; and set a corresponding tag marker for the current tag index j. For the reaction rule marker, set a corresponding label data is the synthesis reaction rule type of the current path point information; and the label index j obtained this time and its corresponding label identifier and the label data Form a corresponding first tag and add it to the first tag sequence; and add 1 to the first index; Step 57, identifying whether the current path point information is the last of the first path point information of the current synthetic path; if so, setting a label index j as the first index, and setting a corresponding label marker for the current label index j is the end marker, and sets a corresponding tag data for the current tag index j is empty, and the label index j obtained this time and its corresponding label marker and the label data Form a corresponding first tag and add it to the first tag sequence, and use the current drug molecule graph as a corresponding first training molecule graph, and use the first training molecule graph and the first tag sequence corresponding to the current synthesis path to form a corresponding first data record, and go to step 58; if not, extract the next first path point information of the current synthesis path as the new current path point information, and return to step 54; Step 58, identifying whether the current data group is the last of the first collected data group; if not, taking the next first collected data group as the new current data group and returning to step 53; if so, forming the corresponding model training data set from all the first data records obtained.

6. The method for optimizing drug molecules according to claim 4, characterized in that: The training of the synthesis path regression prediction model based on the model training data set specifically includes: Step 61, based on a preset first segmentation ratio, the model training data set is divided into two 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 of the first training set to the first evaluation set satisfies the first segmentation ratio; Step 62, taking the first first data record of the first training set as the corresponding current training record; Step 63, inputting the first training molecular graph of the current training record as the corresponding original drug molecular graph into the synthetic pathway regression prediction model to optimize the molecular structure and perform regression prediction processing on the synthetic pathway of the optimized structure; and in the process of model processing, the prediction marker s generated by each step of the model prediction i Data collection is performed, and for each of the predicted markers s specifically as a reactant marker i The corresponding predicted fingerprint p pre,i Data collection is performed, and the prediction marker s for each reaction rule marker is i The corresponding prediction type r pre,i Perform data collection; and perform prediction sequence assembly based on all collected data and the maximum label index Q of the first label sequence of the current training record to obtain a corresponding first prediction sequence; Wherein, Q=max(j); the first prediction sequence is composed of a plurality of first prediction arrays; the first prediction array includes a prediction marker s j and predicted data d j ; The predicted marker s j It includes reactant marker, reaction rule marker and end marker; the prediction marker s j When it is a reactant marker, the predicted data d j is a corresponding predicted fingerprint p pre,i ; The predicted marker s j When it is a reaction rule marker, the prediction data d j is a corresponding prediction type r pre,i ; The predicted marker s j When it is the end marker, the predicted data d j Is empty; if the predicted marker s collected i If the total number of predicted tags is less than the maximum tag index Q, the first prediction sequence is supplemented with a prediction array based on a preset invalid prediction array until the first prediction sequence is consistent with the sequence length of the current first tag sequence; if the collected prediction tag s i If the total number of is greater than the maximum label index Q, the current first label sequence is label-filled based on the invalid prediction array until the current first label sequence is consistent with the first prediction sequence and the maximum label index Q is reset based on the sequence length of the first prediction sequence; Step 64: Substitute the first prediction sequence and the current first label sequence into a preset first model loss function L ALL ; and based on the preset first model optimizer, move towards making the first model loss function L ALL Performing a round of modulation on the model parameters of the encoder, the decoder, the reactant processing module and the composite processing module of the synthesis path regression prediction model in the direction of reaching the minimum value; Among them, the first model loss function L ALL By the type loss function L S , molecular fingerprint loss function L P and the reaction rule loss function L R composition; L ALL =L S +L P +L R , L CE () is the cross entropy loss function, L BCE () is the binary cross entropy loss function; The first model optimizer includes at least an SGD optimizer and an Adam optimizer; Step 65, identifying whether the current training record is the last first data record of the first training set; if so, going to step 66; if not, extracting the next first data record of the first training set as the new current training record, and returning to step 63; Step 66, 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 input the first training molecular graph of the current evaluation record as the corresponding original drug molecular graph into the synthetic pathway regression prediction model to optimize the molecular structure and perform regression prediction processing on the synthetic pathway of the optimized structure; and in the process of model processing, the prediction marker s generated by each step of the model prediction i Data collection is performed, and for each of the predicted markers s specifically as a reactant marker i The corresponding predicted fingerprint p pre,i Data collection is performed, and the prediction marker s for each reaction rule marker is i The corresponding prediction type r pre,i Perform data collection; and assemble a prediction sequence based on all collected data and the maximum label index Q of the first label sequence of the current evaluation record to obtain a corresponding second prediction sequence; and form a corresponding prediction-label sequence pair by the second prediction sequence and the current first label sequence; and at the end of this round of traversal, estimate the accuracy, precision, recall and F1 score based on all the obtained prediction-label sequence pairs to obtain the corresponding first accuracy, first precision, first recall and first F1 score; Step 67, identifying the first accuracy, the first precision, the first recall and the first F1 score; if the first accuracy does not meet the preset first accuracy range, or the first accuracy does not meet the preset first accuracy range, or the first recall does not meet the preset first recall range, or the first F1 score does not meet the preset first F1 score range, return to step 62 to continue training; if the first accuracy meets the first accuracy range, and the first accuracy meets the first accuracy range, and the first recall meets the first recall range, and the first F1 score meets the first F1 score range, stop training and confirm that the model training is completed.

7. The method for optimizing drug molecules according to claim 3, characterized in that: The step of performing molecular structure optimization and synthesis path analysis on the old drug molecule graph based on the synthesis path regression prediction model to obtain the corresponding new drug molecule graph and new molecule synthesis path specifically includes: The old drug molecule graph is used as the corresponding original drug molecule graph and input into the synthesis pathway regression prediction model to perform molecular structure optimization, and the synthesis pathway of the optimized structure is subjected to regression prediction processing to obtain the corresponding optimized drug molecule graph and the optimized molecular synthesis pathway; and the optimized drug molecule graph is used as the corresponding new drug molecule graph, and the optimized molecular synthesis pathway is used as the corresponding new molecular synthesis pathway.

8. The method for optimizing drug molecules according to claim 1, characterized in that: The structural similarity and property difference of the new and old drug molecule graphs are calculated to obtain the corresponding new and old molecule calculation report, specifically including: Step 81, based on a preset chemical calculation tool, the molecular fingerprint similarity of the new and old drug molecular graphs is measured to obtain the corresponding structural similarity; Wherein, the chemical calculation tool at least includes RDKit tool and CDK tool; Step 82, estimating multiple types of properties of the new and old drug molecules corresponding to the new and old drug molecule graphs based on a preset property calculation tool to obtain a corresponding list of properties of new and old molecules; Wherein, the property calculation tools at least include SwissADME tool, admetSAR tool, ADMETlab 2.0 tool, ProTox-II tool, BioTransformer tool, MetaPred tool, CyProduct tool, and RDKit tool; The new and old molecular property list includes a plurality of new and old molecular property records; the new and old molecular property records include property type, new molecular estimated value and old molecular estimated value; the property type includes at least multiple types of physicochemical properties, biological activity and toxicity; the multiple types of physicochemical properties include at least molecular mass, melting point, boiling point, solubility, lipophilicity, permeability and hydrogen bonding; Step 83, a new and old molecule measurement report corresponding to the structural similarity and the new and old molecule property list is formed.

9. The method for optimizing drug molecules according to claim 8, characterized in that: The optimization effect is evaluated according to the new and old molecular measurement reports to obtain the corresponding comprehensive evaluation results, specifically including: Step 91, extracting the corresponding structural similarity and the list of properties of the new and old molecules from the new and old molecule measurement report; Step 92, identifying whether the structural similarity exceeds a preset similarity threshold; if so, setting the corresponding structural evaluation result as structural similarity; if not, setting the corresponding structural evaluation result as significant structural differences; Step 93, counting the total number of records of the new and old molecular property records in the new and old molecular property lists to obtain the corresponding total number of properties K ALL ; and initialize the first counter to 0; and perform a round of traversal of all the new and old molecular property records in the new and old molecular property list; and in this round of traversal, use the new and old molecular property records currently traversed as the corresponding current records; and use the property type, the new molecule estimated value and the old molecule estimated value of the current record as the corresponding current property type, the current new molecule property value and the current old molecule property value; and select the property quality assessment rule corresponding to the current property type from the preset property quality assessment rule library as the corresponding current assessment rule; and use the current assessment rule as the assessment standard, and assess whether the new drug molecule is better than the old drug molecule in the current property type according to the current new molecule property value and the current old molecule property value to obtain the corresponding current property assessment result; and add 1 to the first counter when the current property assessment result is worse than the old molecule; Step 94, after the current round of traversal of all the records of the properties of the new and old molecules in the property list of the new and old molecules is completed, the first counter is identified; if the first counter is 0, the corresponding property evaluation result is set to be better than the old molecule in all aspects; if the first counter is greater than 0 but less than the total number of properties K ALL , then the corresponding property evaluation result is set to be partially better than the old molecule; if the first counter is greater than 0 and equal to the total number of properties K ALL , then the corresponding property evaluation result is set to be comprehensively inferior to the old molecule; Step 95, composing the corresponding comprehensive evaluation result by the obtained structural evaluation result and the obtained property evaluation result.

10. A device for executing the method for optimizing drug molecules according to any one of claims 1 to 9, characterized in that: The device comprises: a chemical information library processing module, a model building module, a model training module, a data receiving module, a model prediction module, a prediction evaluation module, an evaluation judgment module and a prediction output module; The chemical information library processing module is used to create a corresponding building block information library by collecting real molecular fragments and / or unit structure information of various drug molecules; and to create a corresponding reaction rule information library by collecting real synthesis reaction information occurring during the synthesis of various drug molecules; and the corresponding chemical information library is composed of the building block information library and the reaction rule information library; The model building module is used to build a synthetic pathway regression prediction model in combination with the chemical information library; The model training module is used to construct a model training data set; and to train the synthesis path regression prediction model based on the model training data set; The data receiving module is used to receive the generated drug molecule graph input by the user as the corresponding old drug molecule graph after the model training is completed; The model prediction module is used to perform molecular structure optimization and synthesis path analysis on the old drug molecule graph based on the synthesis path regression prediction model to obtain the corresponding new drug molecule graph and new molecule synthesis path; The prediction and evaluation module is used to calculate the structural similarity and property difference of the new and old drug molecule graphs to obtain the corresponding new and old molecule calculation report; and evaluate the optimization effect according to the new and old molecule calculation report to obtain the corresponding comprehensive evaluation result; the comprehensive evaluation result includes the structural evaluation result and the property evaluation result; the structural evaluation result includes the structural similarity and the obvious difference in structure; the property evaluation result includes the overall superiority to the old molecule, the partial superiority to the old molecule and the overall inferiority to the old molecule; The evaluation judgment module is used to identify the structural evaluation result and the property evaluation result of the comprehensive evaluation result; if the structural evaluation result is that the structure is significantly different or the property evaluation result is inferior to the old molecule in all aspects, then return to the model prediction module for re-optimization; if the structural evaluation result is that the structure is similar and the property evaluation result is not inferior to the old molecule in all aspects, then go to the prediction output module; The prediction output module is used to feed back the old drug molecule map and the latest new drug molecule map, the new molecule synthesis path and the new and old molecule measurement report to the current user.

11. 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 9; The transceiver is coupled to the processor, and the processor controls the transceiver to send and receive messages.

12. 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 is enabled to execute the method according to any one of claims 1 to 9.

Images