Interactive / chemical synthesis route design method, system, medium and electronic device

Abstract

The present invention provides an interactive / chemical synthesis route design method, system, medium and electronic device, wherein the interactive chemical synthesis route design method comprises: obtaining a compound; performing a step-by-step reverse synthesis process on the compound according to a preset screening condition to obtain a chemical synthesis route of the compound; obtaining a preference screening condition; updating a certain step in the reverse synthesis process according to the preference screening condition, and updating the subsequent steps of a certain step in the reverse synthesis process based on the preference screening condition and the preset screening condition. The present invention can automatically complete the chemical synthesis route design of the compound, and at the same time allow chemical synthesis experts to interact with the system, so that the system can use the professional knowledge of the experts to continuously iterate in a short time to obtain an accurate and feasible chemical synthesis route.

Description

Technical Field

[0001] The present invention belongs to the technical field of computer chemistry and relates to a chemical synthesis route design method, in particular to an interactive / chemical synthesis route design method, system, medium and electronic equipment. Background Art

[0002] Chemical synthesis is the process of producing a desired compound through chemical reactions. Chemical synthesis begins with the selection of starting materials and reagents. A variety of chemical reactions can be used to obtain the desired product. For the conversion of compound A to product B, various pathways can be considered, referred to as "synthetic strategies." In multistage reactions, chemical transformations are performed sequentially on a single substrate, while in multicomponent reactions, a single product is obtained from multiple reactants.

[0003] There are two ways to express product quantity: yield expressed as mass, and yield expressed as a percentage of the theoretical amount obtained from the starting material. The process of preparing complex compounds from simple compounds requires multiple steps and a significant amount of time and effort until the desired product is synthesized. Therefore, how to quickly obtain an accurate and feasible compound synthesis route is an urgent problem to be solved in the present invention. Summary of the Invention

[0004] In view of the above-mentioned shortcomings of the prior art, the object of the present invention is to provide an interactive / chemical synthesis route design method, system, medium and electronic equipment for quickly obtaining accurate and feasible compound synthesis routes.

[0005] To achieve the above-mentioned objectives and other related objectives, the present invention provides an interactive chemical synthesis route design method, which includes: obtaining a compound; performing step-by-step reverse synthesis processing on the compound according to a preset screening condition to obtain a chemical synthesis route of the compound; obtaining a preference screening condition; updating a certain step in the reverse synthesis processing according to the preference screening condition, and updating the subsequent steps of a certain step in the reverse synthesis processing based on the preference screening condition and the preset screening condition.

[0006] In one embodiment of the present invention, an implementation process of performing step-by-step retrosynthesis on the compound according to a preset screening condition includes: searching for a retrosynthesis reaction template matching the compound in a reaction template library; obtaining a reaction precursor corresponding to the compound based on each of the retrosynthesis reaction templates; selecting a reaction precursor that meets the preset screening condition and a corresponding retrosynthesis reaction template as a step in designing a chemical synthesis route for the compound; if the reaction precursor is a compound that requires retrosynthesis, repeating the above process for the compound that requires retrosynthesis until the obtained reaction precursor is a compound that does not require retrosynthesis, thereby completing all steps in designing a chemical synthesis route for the compound.

[0007] In one embodiment of the present invention, a process for implementing the search for a retrosynthetic reaction template matching the compound in a reaction template library includes: the compound is quantized into a molecular graph G(V,E), where vertices V correspond to atoms and edges E correspond to bonds; each reaction template in the reaction template library is quantized into a connected graph G(V,E) after adding virtual chemical bonds to the corresponding atoms in its reactants and products; searching for the best retrosynthetic reaction template in the reaction template library based on a graph convolution method includes: calculating the embedding vector of the compound according to the molecular graph G(V,E) of the compound as:

[0008]

[0009]

[0010] Where K represents the total number of convolutional network layers, and the compound embedding vector C is aggregated by the highest layer, i.e., the Kth layer, of each atom in the compound; W k is the neural network parameter matrix of the kth layer, P i,j is the bond embedding vector of the bond between atoms i and j in the compound, f(·) is an arbitrary neural network activation function, Reduce(·) is an aggregation function that maps a set of vectors to a fixed-length vector; Concat{·} represents a vector concatenation operation; is the k-th layer embedding of the i-th atom in compound C;

[0011] The embedding vector of each retrosynthetic reaction template in the reaction template library is calculated based on the connectivity graph G(V, E) of the reaction template as T:

[0012]

[0013]

[0014] Where K represents the total number of convolutional network layers, and the embedding vector T of the reaction template is aggregated by the highest layer, i.e., the Kth layer, of each atom in the reaction template; W kis the neural network parameter matrix of the kth layer, P i,j is the bond embedding vector of the bond between atoms i and j in the reaction template, f(·) is an arbitrary neural network activation function, Reduce(·) is an aggregation function that maps a set of vectors to a fixed-length vector; Concat{·} represents a vector concatenation operation; is the kth layer embedding of the i-th atom in the reaction template;

[0015] According to the recommended score P of each retrosynthetic reaction template m =C·T m Output the recommended retrosynthesis reaction templates in descending order, T m is the embedding of the mth reaction template in the reaction template library; m is a positive integer greater than or equal to 1.

[0016] In one embodiment of the present invention, the preset screening condition includes synthesis cost; the synthesis cost of compound c is Cost(c): if compound c is available for purchase, the cost is Cost(c)=1; if the synthesis method of compound c has not been searched, its cost is estimated to be λ -SA(c) , where λ is a parameter that can be preset to 0.4; SA(c) is the estimated number of reaction steps required for the synthesis of compound c with an unknown route; if the synthesis method of compound c has been searched, and there are N known single-step synthesis methods, P i ={p ij} represents the set of all precursors of the i-th synthesis method, y i represents the estimated yield of the i-th synthesis method, then the synthesis cost of compound c is the lowest cost synthesis method among N synthesis methods, that is:

[0017]

[0018] In one embodiment of the present invention, an implementation process for estimating the number of reaction steps SA(c) required for the synthesis of compound c of an unknown route includes: pre-setting the number of synthesis steps for each element in compound c; the sum of the number of synthesis steps of each element is the number of reaction steps SA(c); wherein, the elements include the type and number of atoms, the type and number of functional groups, the length of the long carbon chain, and the number and type of rings; the type and number of atoms include the number of carbon atoms and the number of heteroatoms; the type and number of functional groups include carbonyl, carboxyl, hydroxyl, cyano, amide, or / and acyl chloride; the number and type of rings include aromatic rings, heterocyclic rings, ring size, and the number of fused rings.

[0019] In one embodiment of the present invention, an implementation process of updating a certain step in the retrosynthesis process according to the preference screening conditions includes: the preference screening conditions include a preference for a specific route; and according to the preference for the specific route, a certain step in the chemical synthesis route is designated to be completed using a specific retrosynthesis reaction template.

[0020] In one embodiment of the present invention, an implementation process of updating a certain step in the retrosynthesis process according to the preference screening condition includes: the preference screening condition includes disabling a specific route; the disabling of the specific route is to prohibit an inappropriate reaction from occurring in the chemical synthesis route; disabling an inappropriate reaction occurring in the chemical synthesis route according to the disabling of the specific route, and replacing the step in which the inappropriate reaction occurs with a retrosynthetic reaction according to the preset screening condition.

[0021] The present invention also provides a chemical synthesis route design method, comprising: obtaining a compound; performing a stepwise reverse synthesis process on the compound according to a preset screening condition to obtain a chemical synthesis route of the compound, comprising: searching a reaction template library for a reverse synthesis reaction template that matches the compound; obtaining a corresponding reaction precursor of the compound based on each of the reverse synthesis reaction templates; selecting a reaction precursor that meets the preset screening condition and a corresponding reverse synthesis reaction template as a step in designing a chemical synthesis route of the compound; if the reaction precursor is a compound that requires reverse synthesis, repeating the above process on the compound that requires reverse synthesis until the obtained reaction precursor is a compound that does not require reverse synthesis, thereby completing all steps in designing a chemical synthesis route of the compound.

[0022] The present invention also provides an interactive chemical synthesis route design system, which includes: an input module for obtaining a compound; a retrosynthesis processing module, which is communicatively connected to the input module and performs step-by-step retrosynthesis processing on the compound according to a preset screening condition to obtain a chemical synthesis route of the compound; an interaction module, which obtains a preference screening condition; and a retrosynthesis update module, which is communicatively connected to the interaction module and the retrosynthesis processing module respectively, and updates a certain step in the retrosynthesis processing according to the preference screening condition, and updates the subsequent steps of a certain step in the retrosynthesis processing based on the preference screening condition and the preset screening condition.

[0023] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the interactive chemical synthesis route design method or implements the steps of the chemical synthesis route design method.

[0024] The present invention also provides an electronic device, comprising: a memory storing a computer program; a processor communicatively connected to the memory, and configured to implement the steps of the interactive chemical synthesis route design method or the steps of the chemical synthesis route design method when calling the computer program.

[0025] As described above, the interactive / chemical synthesis route design method, system, medium, and electronic device of the present invention have the following beneficial effects:

[0026] The present invention can automatically complete the design of chemical synthesis routes for small molecules, while allowing chemical synthesis experts to interact with the system, so that the system can utilize the experts' professional knowledge and continuously iterate to obtain accurate and feasible chemical synthesis routes in a short period of time. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 Shown is a schematic diagram of an implementation flow of the interactive chemical synthesis route design method described in an embodiment of the present invention.

[0028] Figure 2 Shown is a schematic diagram of an implementation flow of step S102 of the interactive chemical synthesis route design method according to an embodiment of the present invention.

[0029] Figure 3A Shown is a schematic diagram of the exploration process for target compounds in a single iteration according to an embodiment of the present invention.

[0030] Figure 3B FIG. 1 is a schematic diagram of a complete path search process including five iterations according to an embodiment of the present invention.

[0031] Figure 4 Shown is an exemplary flow chart of the method for designing a chemical synthesis route according to an embodiment of the present invention.

[0032] Figure 5 Shown is a schematic diagram of an implementation process of the chemical synthesis route design method described in an embodiment of the present invention.

[0033] Figure 6 Shown is a schematic diagram of an implementation structure of the interactive chemical synthesis route design system according to an embodiment of the present invention.

[0034] Figure 7 Shown is a schematic diagram of an implementation structure of the chemical synthesis route design system described in an embodiment of the present invention.

[0035] Figure 8 Shown is a schematic diagram of an implementation structure of an electronic device according to an embodiment of the present invention.

[0036] Component number description

[0037] 600 Interactive Chemical Synthesis Route Design System

[0038] 610 Input Module

[0039] 620 Retrosynthesis Processing Module

[0040] 630 Interaction Module

[0041] 640 Retrosynthesis Update Module

[0042] 700 Chemical Synthesis Route Design System

[0043] 710 Input Module

[0044] 720 Retrosynthesis Processing Module

[0045] 800 Electronic Equipment

[0046] 810 Memory

[0047] 820 processor

[0048] Steps S101 to S104

[0049] Steps S201 to S204

[0050] Steps S401 to S410

[0051] Steps S510 to S524 DETAILED DESCRIPTION

[0052] The following describes the embodiments of the present invention through specific examples. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments. The details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the following embodiments and features in the embodiments can be combined with each other unless they conflict.

[0053] It should be noted that the illustrations provided in the following embodiments are merely schematic illustrations of the basic concept of the present invention. Therefore, the illustrations only show components related to the present invention and are not drawn according to the number, shape, and size of components in actual implementation. In actual implementation, the type, quantity, and proportion of each component may be changed arbitrarily, and the component layout may also be more complex.

[0054] See also Figure 1 The present invention provides an interactive chemical synthesis route design method, which comprises:

[0055] S101, obtaining a compound;

[0056] S102, performing a stepwise retrosynthesis process on the compound according to a preset screening condition to obtain a chemical synthesis route of the compound;

[0057] S103, obtaining a preference screening condition;

[0058] S104 , performing an update process on a certain step in the retrosynthesis process according to the preference screening condition, and performing an update process on a subsequent step of the certain step in the retrosynthesis process based on the preference screening condition and the preset screening condition.

[0059] This embodiment can automatically complete the design of chemical synthesis routes for small molecules, while allowing chemical synthesis experts to interact with the system, so that the system can use the expert's professional knowledge to continuously iterate and obtain accurate and feasible chemical synthesis routes in a short period of time. Among them, small molecules refer to chemically synthesized drugs, which correspond to the macromolecules in macromolecular drugs. Macromolecular drugs such as peptides, polysaccharides, nucleic acids, proteins, etc. are generally synthesized by biological means, while drugs synthesized by chemical synthesis in the pharmaceutical industry are small molecule drugs. Generally speaking, the molecular weight of small molecules is below 900 Daltons and the physical size is below 1 nm; small molecule drugs can bind to biological macromolecular targets to achieve regulatory effects.

[0060] See also Figure 2 As shown, in one embodiment of the present invention, the step S102 includes performing a stepwise retrosynthesis process on the compound according to a preset screening condition.

[0061] S201, searching a reaction template library for a retrosynthetic reaction template that matches the compound, wherein the retrosynthetic reaction template is referred to as a reaction template.

[0062] S202, obtaining a reaction precursor of the corresponding compound based on each of the retrosynthetic reaction templates;

[0063] S203, selecting a reaction precursor that meets the preset screening conditions and a corresponding retrosynthetic reaction template as a step in designing a chemical synthesis route for the compound;

[0064] S204, if the reaction precursor is a compound that requires retrosynthesis, repeat the above process on the compound that requires retrosynthesis until the obtained reaction precursor is a compound that does not require retrosynthesis, thus completing all steps of the chemical synthesis route design for the compound.

[0065] In one embodiment of the present invention, an implementation process of searching a reaction template library for a retrosynthetic reaction template that matches the compound in step S201 includes:

[0066] The compound is quantized into a molecular graph G(V,E), where vertices V correspond to atoms and edges E correspond to bonds; each reaction template in the reaction template library is quantized into a connected graph G(V,E) after adding virtual chemical bonds to the corresponding atoms in its reactants and products, where vertices V correspond to atoms and edges E correspond to bonds or / and virtual chemical bonds;

[0067] The method based on graph convolution searches for the best retrosynthetic reaction template in the reaction template library, including:

[0068] The embedding vector of the compound is calculated based on the molecular graph G(V,E) of the compound as follows:

[0069]

[0070]

[0071] Where K represents the total number of convolutional network layers, and the compound embedding vector C is aggregated by the highest layer, i.e., the Kth layer, of each atom in the compound; W k is the neural network parameter matrix of the kth layer, P i,j is the bond embedding vector of the bond between atoms i and j in the compound, f(·) is an arbitrary neural network activation function, Reduce(·) is an aggregation function that maps a set of vectors to a fixed-length vector; Concat{·} represents a vector concatenation operation; is the k-th level embedding of the i-th atom in the compound (i.e., the i-th vertex in the molecular graph G(V,E) of the compound);

[0072] The embedding vector of each retrosynthetic reaction template in the reaction template library is calculated based on the connectivity graph G(V, E) of the reaction template as T:

[0073]

[0074]

[0075] Where K represents the total number of convolutional network layers, and the embedding vector T of the reaction template is aggregated by the highest layer, i.e., the Kth layer, of each atom in the reaction template; W k is the neural network parameter matrix of the kth layer, P i,j is the bond embedding vector of the bond between atoms i and j in the reaction template, f(·) is an arbitrary neural network activation function, Reduce(·) is an aggregation function that maps a set of vectors to a fixed-length vector; Concat{·} represents a vector concatenation operation; is the k-th level embedding of the i-th atom in the reaction template (i.e., the i-th vertex in the connected graph G(V,E) of the reaction template);

[0076] According to the recommended score P of each retrosynthetic reaction template m =C·T m Output the recommended retrosynthesis reaction templates in descending order, T m is the embedding of the mth reaction template in the reaction template library, where m is a positive integer greater than or equal to 1.

[0077] In one embodiment of the present invention, the preset screening condition in step S102 includes synthesis cost;

[0078] The synthesis cost of compound c is Cost(c):

[0079] If compound c is available for purchase, the cost is Cost(c) = 1;

[0080] If the synthesis method of compound c has not been searched, its cost is estimated to be λ -SA(c) , where λ is a parameter that can be preset to 0.4; SA(c) is the estimated number of reaction steps required for the synthesis of compound c by an unknown route;

[0081] If the synthesis method of compound c has been searched, let there be N known single-step synthesis methods, P i ={p ij} represents the set of all precursors of the i-th synthesis method, y i represents the estimated yield of the i-th synthesis method, then the synthesis cost of compound c is the lowest cost synthesis method among N synthesis methods, that is:

[0082]

[0083] Furthermore, an implementation process of estimating the number of reaction steps SA(c) required for the synthesis of compound c of an unknown route includes: pre-setting the number of synthesis steps for each element in compound c; the sum of the number of synthesis steps of each element is the number of reaction steps SA(c); wherein, the elements include the type and number of atoms, the type and number of functional groups, the length of long carbon chains, and the number and type of rings; the type and number of atoms include the number of carbon atoms and the number of heteroatoms; the type and number of functional groups include carbonyl, carboxyl, hydroxyl, cyano, amide, or / and acyl chloride; the number and type of rings include aromatic rings, heterocyclic rings, ring size, and the number of fused rings.

[0084] Specifically, the reaction recommendation module is the basis of the retrosynthetic design method. For a given compound, the reaction recommendation module uses a graph convolution-based method to search for the best retrosynthetic reaction template (or simply reaction template). Only when the reaction site of the compound is fully compatible with the reaction template, the reaction template is applied to obtain the precursor of the reaction. This implementation strictly ensures that the reaction site is compatible with the reaction template and can use a reaction feasibility predictor through a discrete reaction template.

[0085] The reaction recommendation module is implemented based on a graph convolutional neural network. The input compound is first quantized into a molecular graph G(V,E), where vertices V correspond to atoms and edges E correspond to bonds. Each reaction template in the reaction template library is quantized into a connected graph G(V,E) after adding virtual chemical bonds to the corresponding atoms in its reactants and products. The initial atomic embedding is defined by the embedding layer of atomic elements. Each graph convolution layer consists of an embedding set of each atom in the compound and its bonding atoms. composition:

[0086]

[0087] Among them, W k is the neural network parameter matrix of the kth layer, P i，j is the bond embedding vector of the bond between atom i and atom j in the compound, f(·) is an arbitrary neural network activation function, Reduce(·) is an aggregation function that maps a set of vectors to a fixed-length vector, and Concat{·} represents a vector concatenation operation, which is used to concatenate multiple input vectors into a single vector for subsequent processing; is the k-th level embedding of the i-th atom in the compound (i.e., the i-th vertex in the molecular graph G(V,E) of the compound). Finally, the embedding vector of the compound is calculated based on the molecular graph G(V,E) of the compound as:

[0088]

[0089] Where K represents the total number of convolutional network layers, the embedding vector C of the compound is formed by the embedding aggregation of the highest layer (Kth layer, the size of K varies depending on the input compound) of each atom in the compound; V represents the atomic set of the compound.

[0090] Reaction templates in the reaction template library include the correspondence between reactants, products, and their atoms. For each known reaction template in the system, the corresponding atoms between the reactants and products are connected with virtual bonds to obtain a connected graph G(V,E) similar to the input compound. A graph convolutional neural network is then used to calculate the embedding vector T of the reaction template.

[0091] Similarly, the k-th layer embedding of the i-th atom in the reaction template (i.e., the i-th vertex in the connected graph G(V,E) of the reaction template) is

[0092]

[0093] Among them, W k is the neural network parameter matrix of the kth layer, P i,j is the bond embedding vector of the bond between atoms i and j in the reaction template, f(·) is an arbitrary neural network activation function, Reduce(·) is an aggregation function that maps a vector set to a fixed-length vector, and Concat{·} represents a vector concatenation operation, which is used to concatenate multiple input vectors into a single vector for subsequent processing. Finally, the embedding vector T of each retrosynthetic reaction template in the reaction template library is calculated based on the connectivity graph G(V,E) of the reaction template:

[0094]

[0095] Where K represents the total number of convolutional network layers, the embedding vector T of the reaction template is aggregated by embedding the highest layer (Kth layer, the size of K varies depending on the input compound) of each atom in the reaction template; V represents the atomic set of the reaction template.

[0096] Let the embedding vector of the input compound be C and the embedding vector of the reaction template numbered m be T m , the recommendation score of the reaction template is calculated as the dot product of these two vectors: P m =C·T m , T m is the embedding of the mth reaction template in the reaction template library, where m is a positive integer greater than or equal to 1.

[0097] The recommendation module enumerates all known reaction templates for the input compound and calculates the recommendation score. m After sorting, the recommended reaction templates are output in descending order.

[0098] The retrosynthesis planning process uses a directed acyclic graph (DAG) based search algorithm to efficiently search the synthesis space. Figure 3A As shown, in a single iteration, a target compound is explored by applying compatible reaction templates to the molecule to generate its possible precursors. The cost of each reaction is estimated using a compound complexity estimator and a reaction feasibility score. The step with the lowest cost is then selected, and the updated cost is fed back to the remaining compounds along the synthetic pathway in the graph.

[0099] The complexity estimator SA(c) estimates the number of reaction steps required to synthesize a molecule c with an unknown route. The estimation method is to calculate the following elements in the compound:

[0100] 1) Atom type and number, including the number of carbon atoms and heteroatoms;

[0101] 2) Type and number of functional groups, including carbonyl, carboxyl, hydroxyl, cyano, amide, acyl chloride, etc.;

[0102] 3) Long carbon chain length;

[0103] 4) Number and type of rings, including aromatic and heterocyclic rings, ring size and number of fused rings.

[0104] For each element, the estimated number of synthesis steps for the element is set in advance, and the final sum of all elements is the estimated synthesis cost of the compound.

[0105] For each compound c on a path, the synthesis cost Cost(c) is estimated as follows:

[0106] If compound c is available for purchase, the cost is 1;

[0107] If the synthesis method of compound c has not been searched, its cost is estimated to be λ -SA(c) , where λ is a parameter, which can be preset to 0.4;

[0108] If the synthesis method of compound c has been searched, let there be N known single-step synthesis methods, P i ={p ij} represents the set of all precursors of the i-th synthesis method, y i represents the estimated yield of the i-th synthesis method, and its synthesis cost is estimated to be the lowest cost synthesis method among all methods, that is:

[0109]

[0110] The ultimate goal of the path planning module is to find a synthetic route for the desired final product x that minimizes Cost(x). CostReduction(c) is defined as the reduction in Cost(x) when the synthesis cost of compound c is set to 1. During planning, the compound c with the highest CostReduction(c) is selected for expansion (i.e., for the next step of retrosynthesis).

[0111] Figure 3BA complete path search process consisting of five iterations is shown. Each iteration expands the compound in its corresponding box. During the search process, the updated cost may change the priority of all expansion directions after each iteration, as shown in Iteration 4 (Iter4), in this way, compounds that have not been explored in depth in other pathways can be selected. The cost estimator sometimes overestimates or underestimates the synthesis cost of a compound early in the search. By continuously iterating, incorrect estimates can be corrected midway through the planning. After Iteration 4, the cost of the new path is updated to be higher than the original path, so Iteration 5 (Iter5) is executed on the original path to complete the path by finding two purchasable precursor compounds. The complete calculation method can be seen in Figure 4 Shown, including:

[0112] S401, input target compound m;

[0113] S402, adding m to the compound collection S;

[0114] S403, find the current best synthesis route r in S cb ;

[0115] S404, looking for r cb The most complex molecule to be synthesized e ;

[0116] S405, for c e Use the reaction recommender to recommend multiple reaction templates and their corresponding precursor compounds R e ;

[0117] S406, will All precursor compounds of the chemical reactions in are added to the set S;

[0118] S407, determine whether the predetermined number of iterations has been reached? If not, return to step S403;

[0119] S408: If yes, then determine whether there is a complete route starting from the purchasable compound in S?

[0120] S409, if yes, output the synthesis route in S;

[0121] If not, output that no synthetic route is found in S410.

[0122] In one embodiment of the present invention, an implementation process of updating a certain step in the retrosynthesis process according to the preference screening condition described in step S104 includes: the preference screening condition includes a preference for a specific route; and according to the preference for the specific route, a certain step in the chemical synthesis route is designated to be completed using a specific retrosynthesis reaction template.

[0123] In one embodiment of the present invention, an implementation process of updating a certain step in the retrosynthesis process according to the preference screening condition described in step S104 includes: the preference screening condition includes disabling a specific route; the disabling of the specific route is to prohibit an inappropriate reaction from occurring in the chemical synthesis route; disabling an inappropriate reaction occurring in the chemical synthesis route according to the disabling of the specific route, and replacing the step in which the inappropriate reaction occurs with a retrosynthetic reaction according to the preset screening condition.

[0124] After browsing the synthesis routes provided by the system, users can enter their preference information for the route (i.e., preference filter conditions) into the system, and the system will use this information to iteratively optimize the synthesis route. For example, the preference information that can be entered includes:

[0125] 1) Preference for a specific route: Users can specify that a step in the path should be completed using a specific reaction. When re-iteration is performed, the system will force the use of this step and re-plan the subsequent retrosynthesis steps.

[0126] 2) Disabling Specific Routes: If a step in a route is deemed inappropriate, the user can specify that it be disabled. The system records all inappropriate steps specified by the user during the iteration process. During route planning, the system's reaction similarity assessment module compares the designed steps with all user-specified inappropriate steps. The reaction feasibility of similar steps is penalized, and the entire route planning process is then re-planned.

[0127] The scope of protection of the interactive chemical synthesis route design method described in the present invention is not limited to the order of execution of the steps listed in this embodiment. All solutions implemented by adding, subtracting, or replacing steps in the prior art based on the principles of the present invention are included in the scope of protection of the present invention.

[0128] See also Figure 5 As shown, an embodiment of the present invention further provides a method for designing a chemical synthesis route, comprising:

[0129] S510, obtaining a compound;

[0130] S520, performing a stepwise retrosynthesis process on the compound according to a preset screening condition to obtain a chemical synthesis route of the compound, including:

[0131] S521, searching a reaction template library for a retrosynthetic reaction template that matches the compound; wherein the retrosynthetic reaction template is referred to as a reaction template.

[0132] In one embodiment of the present invention, an implementation process of searching a reaction template library for a retrosynthetic reaction template that matches the compound in step S201 includes:

[0133] The compound is quantized into a molecular graph G(V,E), where vertices V correspond to atoms and edges E correspond to bonds; each reaction template in the reaction template library is quantized into a connected graph G(V,E) after adding virtual chemical bonds to the corresponding atoms in its reactants and products;

[0134] The method based on graph convolution searches for the best retrosynthetic reaction template in the reaction template library, including:

[0135] The embedding vector of the compound is calculated based on the molecular graph G(V,E) of the compound as follows:

[0136]

[0137]

[0138] Where K represents the total number of convolutional network layers, and the compound embedding vector C is aggregated by the highest layer, i.e., the Kth layer, of each atom in the compound; W k is the neural network parameter matrix of the kth layer, P i,j is the bond embedding vector of the bond between atoms i and j in the compound, f(·) is an arbitrary neural network activation function, Reduce(·) is an aggregation function that maps a set of vectors to a fixed-length vector; Concat{·} represents a vector concatenation operation; is the k-th level embedding of the i-th atom in the compound C (i.e., the i-th vertex in the molecular graph G(V,E) of the compound);

[0139] The embedding vector of each retrosynthetic reaction template in the reaction template library is calculated based on the connectivity graph G(V, E) of the reaction template as T:

[0140]

[0141]

[0142] Where K represents the total number of convolutional network layers, and the embedding vector T of the reaction template is aggregated by the highest layer, i.e., the Kth layer, of each atom in the reaction template; W k is the neural network parameter matrix of the kth layer, P i,j is the bond embedding vector of the bond between atoms i and j in the reaction template, f(·) is an arbitrary neural network activation function, Reduce(·) is an aggregation function that maps a set of vectors to a fixed-length vector; Concat{·} represents a vector concatenation operation; is the k-th level embedding of the i-th atom in the reaction template (i.e., the i-th vertex in the connected graph G(V,E) of the reaction template);

[0143] According to the recommended score P of each retrosynthetic reaction template m =C·T m Output the recommended retrosynthesis reaction templates in descending order, T m is the embedding of the mth reaction template in the reaction template library, where m is a positive integer greater than or equal to 1.

[0144] S522, obtaining a reaction precursor of the corresponding compound based on each of the retrosynthetic reaction templates;

[0145] S523, selecting a reaction precursor that meets the preset screening conditions and a corresponding retrosynthetic reaction template as a step in designing a chemical synthesis route for the compound.

[0146] In one embodiment of the present invention, the preset screening condition includes synthesis cost;

[0147] The synthesis cost of compound c is Cost(c):

[0148] If compound c is available for purchase, the cost is Cost(c) = 1;

[0149] If the synthesis method of compound c has not been searched, its cost is estimated to be λ -SA(c) , where λ is a parameter that can be preset to 0.4; SA(c) is the estimated number of reaction steps required for the synthesis of compound c by an unknown route;

[0150] If the synthesis method of compound c has been searched, let there be N known single-step synthesis methods, P i ={p ij} represents the set of all precursors of the i-th synthesis method, y i represents the estimated yield of the i-th synthesis method, then the synthesis cost of compound c is the lowest cost synthesis method among N synthesis methods, that is:

[0151]

[0152] Furthermore, an implementation process of estimating the number of reaction steps SA(c) required for the synthesis of compound c of an unknown route includes: pre-setting the number of synthesis steps for each element in compound c; the sum of the number of synthesis steps of each element is the number of reaction steps SA(c); wherein, the elements include the type and number of atoms, the type and number of functional groups, the length of long carbon chains, and the number and type of rings; the type and number of atoms include the number of carbon atoms and the number of heteroatoms; the type and number of functional groups include carbonyl, carboxyl, hydroxyl, cyano, amide, or / and acyl chloride; the number and type of rings include aromatic rings, heterocyclic rings, ring size, and the number of fused rings.

[0153] S524, if the reaction precursor is a compound that requires retrosynthesis, repeat the above process on the compound that requires retrosynthesis until the obtained reaction precursor is a compound that does not require retrosynthesis, thereby completing all steps of the chemical synthesis route design for the compound.

[0154] The scope of protection of the chemical synthesis route design method described in the present invention is not limited to the order of execution of the steps listed in this embodiment. All solutions implemented by adding, subtracting, or replacing steps in the prior art based on the principles of the present invention are included in the scope of protection of the present invention.

[0155] The present invention also provides an interactive chemical synthesis route design system, which can implement the interactive chemical synthesis route design method described in the present invention. However, the implementation device of the interactive chemical synthesis route design method described in the present invention includes but is not limited to the structure of the interactive chemical synthesis route design system listed in this embodiment. Any structural deformation and replacement of the existing technology made according to the principles of the present invention are included in the scope of protection of the present invention.

[0156] See also Figure 6 As shown, an embodiment of the present invention further provides an interactive chemical synthesis route design system, wherein the interactive chemical synthesis route design system 600 includes: an input module 610 , a retrosynthesis processing module 620 , an interaction module 630 , and a retrosynthesis update module 640 .

[0157] The input module 610 obtains a compound.

[0158] The retrosynthesis processing module 620 is in communication with the input module 610 and performs retrosynthesis processing on the compound step by step according to a preset screening condition to obtain a chemical synthesis route of the compound.

[0159] In one embodiment of the present invention, the retrosynthesis processing module 620 quantizes the compound into a molecular graph G(V, E), where vertices V correspond to atoms and edges E correspond to bonds; each reaction template in the reaction template library is quantized into a connected graph G(V, E) after adding virtual chemical bonds to the corresponding atoms in its reactants and products; the retrosynthesis processing module searches for the best retrosynthesis reaction template in the reaction template library based on a graph convolution method, including:

[0160] The retrosynthesis processing module calculates the embedding vector of the compound according to the molecular graph G(V, E) of the compound as:

[0161]

[0162]

[0163] Where K represents the total number of convolutional network layers, and the compound embedding vector C is aggregated by the highest layer, i.e., the Kth layer, of each atom in the compound; W k is the neural network parameter matrix of the kth layer, P i,j is the bond embedding vector of the bond between atoms i and j in the compound, f(·) is an arbitrary neural network activation function, Reduce(·) is an aggregation function that maps a set of vectors to a fixed-length vector; Concat{·} represents a vector concatenation operation; is the k-th level embedding of the i-th atom in the compound (i.e., the i-th vertex in the molecular graph G(V,E) of the compound);

[0164] The retrosynthesis processing module calculates the embedding vector T of each retrosynthesis reaction template in the reaction template library according to the connectivity graph G(V, E) of the reaction template:

[0165]

[0166]

[0167] Where K represents the total number of convolutional network layers, and the embedding vector T of the reaction template is aggregated by the highest layer, i.e., the Kth layer, of each atom in the reaction template; W k is the neural network parameter matrix of the kth layer, P i,j is the bond embedding vector of the bond between atoms i and j in the reaction template, f(·) is an arbitrary neural network activation function, Reduce(·) is an aggregation function that maps a set of vectors to a fixed-length vector; Concat{·} represents a vector concatenation operation; is the k-th level embedding of the i-th atom in the reaction template (i.e., the i-th vertex in the connected graph G(V,E) of the reaction template);

[0168] The retrosynthesis processing module is based on the recommendation score P of each retrosynthesis reaction template. m =C·T m Output the recommended retrosynthesis reaction templates in descending order, T m is the embedding of the mth reaction template in the reaction template library, where m is a positive integer greater than or equal to 1.

[0169] The interaction module 630 obtains a preference screening condition; the preference screening condition includes a preference for a specific route and / or a prohibition of a specific route; a step in the chemical synthesis route is designated to be completed using a specific retrosynthetic reaction template based on the preference for the specific route. The prohibition of the specific route is to prohibit an inappropriate reaction from occurring in the chemical synthesis route; the inappropriate reaction occurring in the chemical synthesis route is prohibited based on the prohibition of the specific route, and a retrosynthetic reaction is replaced in the step where the inappropriate reaction occurs based on the preset screening condition.

[0170] The retrosynthesis update module 640 is communicated with the interaction module 630 and the retrosynthesis processing module 620 respectively, and updates a certain step in the retrosynthesis processing according to the preference screening conditions, and updates the subsequent steps of a certain step in the retrosynthesis processing based on the preference screening conditions and the preset screening conditions. For example: the user can specify that a certain step in the path is completed using a specific reaction. At this time, when the re-iteration calculation is performed, the system will force the use of this step and re-complete the planning of the subsequent retrosynthesis steps. Or when performing path planning, the reaction similarity evaluation module in the system will compare the similarity of the designed steps with all the inappropriate steps specified by the user. The reaction feasibility scores of similar steps are penalized, and then the entire path planning is re-completed.

[0171] The present invention also provides a chemical synthesis route design system, which can implement the chemical synthesis route design method described in the present invention. However, the implementation device of the chemical synthesis route design method described in the present invention includes but is not limited to the structure of the chemical synthesis route design system listed in this embodiment. All structural modifications and replacements of the prior art made according to the principles of the present invention are included in the scope of protection of the present invention.

[0172] See also Figure 7 As shown, an embodiment of the present invention further provides a chemical synthesis route design system, and the chemical synthesis route design system 700 includes: an input module 710 and a retrosynthesis processing module 720.

[0173] The input module 710 obtains a compound.

[0174] The retrosynthesis processing module 720 is in communication with the input module 710 and performs retrosynthesis processing on the compound step by step according to a preset screening condition to obtain a chemical synthesis route of the compound.

[0175] In one embodiment of the present invention, the retrosynthesis processing module 720 quantizes the compound into a molecular graph G(V, E), where vertices V correspond to atoms and edges E correspond to bonds. Each reaction template in the reaction template library is quantized into a connected graph G(V, E) after adding virtual chemical bonds to the corresponding atoms in its reactants and products. The retrosynthesis processing module searches for the best retrosynthesis reaction template in the reaction template library based on a graph convolution method, including:

[0176] The retrosynthesis processing module calculates the embedding vector of the compound according to the molecular graph G(V, E) of the compound as:

[0177]

[0178]

[0179] Where K represents the total number of convolutional network layers, and the compound embedding vector C is aggregated by the highest layer, i.e., the Kth layer, of each atom in the compound; W k is the neural network parameter matrix of the kth layer, P i,j is the bond embedding vector of the bond between atoms i and j in the compound, f(·) is an arbitrary neural network activation function, Reduce(·) is an aggregation function that maps a set of vectors to a fixed-length vector; Concat{·} represents a vector concatenation operation; is the k-th level embedding of the i-th atom in the compound (i.e., the i-th vertex in the molecular graph G(V,E) of the compound);

[0180] The retrosynthesis processing module calculates the embedding vector T of each retrosynthesis reaction template in the reaction template library according to the connectivity graph G(V, E) of the reaction template:

[0181]

[0182]

[0183] Where K represents the total number of convolutional network layers, and the embedding vector T of the reaction template is aggregated by the highest layer, i.e., the Kth layer, of each atom in the reaction template; W k is the neural network parameter matrix of the kth layer, P i,jis the bond embedding vector of the bond between atoms i and j in the reaction template, f(·) is an arbitrary neural network activation function, Reduce(·) is an aggregation function that maps a set of vectors to a fixed-length vector; Concat{·} represents a vector concatenation operation; is the k-th level embedding of the i-th atom in the reaction template (i.e., the i-th vertex in the connected graph G(V,E) of the reaction template);

[0184] The retrosynthesis processing module is based on the recommendation score P of each retrosynthesis reaction template. m =C·T m Output the recommended retrosynthesis reaction templates in descending order, T m is the embedding of the mth reaction template in the reaction template library, where m is a positive integer greater than or equal to 1.

[0185] In one embodiment of the present invention, the preset screening condition includes synthesis cost;

[0186] The synthesis cost of compound c is Cost(c):

[0187] If compound c is available for purchase, the cost is Cost(c) = 1;

[0188] If the synthesis method of compound c has not been searched, its cost is estimated to be λ -SA(c) , where λ is a parameter that can be preset to 0.4; SA(c) is the estimated number of reaction steps required to synthesize compound c of unknown route;

[0189] If the synthesis method of compound c has been searched, let there be N known single-step synthesis methods, P i ={p ij} represents the set of all precursors of the i-th synthesis method, y i represents the estimated yield of the i-th synthesis method, then the synthesis cost of compound c is the lowest cost synthesis method among N synthesis methods, that is:

[0190]

[0191] Furthermore, an implementation process of estimating the number of reaction steps SA(c) required for the synthesis of compound c of an unknown route includes: pre-setting the number of synthesis steps for each element in compound c; the sum of the number of synthesis steps of each element is the number of reaction steps SA(c); wherein, the elements include the type and number of atoms, the type and number of functional groups, the length of long carbon chains, and the number and type of rings; the type and number of atoms include the number of carbon atoms and the number of heteroatoms; the type and number of functional groups include carbonyl, carboxyl, hydroxyl, cyano, amide, or / and acyl chloride; the number and type of rings include aromatic rings, heterocyclic rings, ring size, and the number of fused rings.

[0192] An embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the interactive chemical synthesis route design method described in the present invention or implements the steps of the chemical synthesis route design method described in the present invention.

[0193] See also Figure 8 As shown, an embodiment of the present invention further provides an electronic device, wherein the electronic device 800 includes a memory 810 and a processor 820. The memory 810 stores a computer program; the processor 820 is in communication with the memory, and when the computer program is invoked, the steps of the interactive chemical synthesis route design method described in the present invention or the steps of the chemical synthesis route design method described in the present invention are implemented.

[0194] In summary, the present invention effectively overcomes the various shortcomings of the prior art and has high industrial utilization value.

[0195] The above embodiments are merely illustrative of the principles and effects of the present invention and are not intended to limit the present invention. Anyone skilled in the art may modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by one of ordinary skill in the art without departing from the spirit and technical principles disclosed herein are intended to be covered by the claims of the present invention.

Claims

1. An interactive chemical synthesis route design method, characterized in that: The interactive chemical synthesis route design method comprises: obtaining a compound; Performing a step-by-step retrosynthesis process on the compound according to a preset screening condition, searching a retrosynthesis reaction template matching the compound in a reaction template library; wherein the cost of each reaction in the retrosynthesis process is obtained by a compound complexity estimator, and the complexity estimator estimates the number of reaction steps required for the synthesis of the compound of an unknown route; Obtaining a chemical synthesis route of the compound, wherein the compound is quantized into a molecular graph G(V,E), where the vertices V correspond to atoms and the edges E correspond to bonds; each reaction template in the reaction template library is quantized into a connected graph G(V,E) after adding virtual chemical bonds to the corresponding atoms in its reactants and products; Searching for the best retrosynthetic reaction template in the reaction template library based on a graph convolution method; Get a preferred filter condition; Performing an update process on a certain step in the retrosynthesis process according to the preference screening condition and performing an update process on a subsequent step of the certain step in the retrosynthesis process based on the preference screening condition and the preset screening condition; An implementation process of updating a certain step in the retrosynthesis process according to the preference screening condition includes: the preference screening condition includes a preference for a specific route; according to the preference for the specific route, a certain step in the chemical synthesis route is designated as being completed using a specific retrosynthesis reaction template; or The preferred screening conditions include disabling a specific route; the disabling of a specific route is to prohibit an inappropriate reaction from occurring in the chemical synthesis route; disabling an inappropriate reaction occurring in the chemical synthesis route according to the disabling of the specific route, and replacing the step where the inappropriate reaction occurs with a retrosynthetic reaction according to the preset screening conditions.

2. The interactive chemical synthesis route design method according to claim 1, characterized in that: A process for implementing the stepwise retrosynthesis of the compound according to a preset screening condition also includes: Obtaining a reaction precursor of the corresponding compound based on each of the retrosynthetic reaction templates; Selecting a reaction precursor that meets the preset screening conditions and a corresponding retrosynthetic reaction template as a step in the chemical synthesis route design of the compound; If the reaction precursor is a compound that needs to be retrosynthesized, the above treatment is repeated for the compound that needs to be retrosynthesized until the reaction precursor obtained is a compound that does not need to be retrosynthesized, thereby completing all steps of the chemical synthesis route design for the compound.

3. The interactive chemical synthesis route design method according to claim 1, characterized in that: The graph convolution-based method searches for the best retrosynthetic reaction template in the reaction template library, comprising: The embedding vector of the compound is calculated according to the molecular graph G(V,E) of the compound as follows: Where K represents the total number of convolutional network layers, and the compound embedding vector C is aggregated by the highest layer, i.e., the Kth layer, of each atom in the compound; W k is the neural network parameter matrix of the kth layer, P i，j is the bond embedding vector of the bond between atoms i and j in the compound, f(·) is an arbitrary neural network activation function, Reduce(·) is an aggregation function that maps a vector set to a fixed-length vector; Concat{·} represents a vector concatenation operation; is the k-th layer embedding of the ith atom in the compound; The embedding vector of each retrosynthetic reaction template in the reaction template library is calculated according to the connectivity graph G(V,E) of the reaction template as T: Where K represents the total number of convolutional network layers, and the embedding vector T of the reaction template is aggregated by the highest layer, i.e., the Kth layer, of each atom in the reaction template; W k is the neural network parameter matrix of the kth layer, P i,j is the bond embedding vector of the bond between atoms i and j in the reaction template, f(·) is an arbitrary neural network activation function, Reduce(·) is an aggregation function that maps a vector set to a fixed-length vector; Concat{·} represents a vector concatenation operation; is the k-th embedding of the ith atom in the reaction template; According to the recommended score P of each retrosynthetic reaction template m =C·T m Output the recommended retrosynthetic reaction templates in descending order, T m is the embedding of the mth reaction template in the reaction template library, where m is a positive integer greater than or equal to 1.

4. The interactive chemical synthesis route design method according to claim 1, characterized in that: The preset screening conditions include synthesis cost; The synthesis cost of compound c is Cost(c): If compound c is available for purchase, the cost is Cost(c) = 1; If the synthesis method of compound c has not been searched, its cost is estimated to be λ -SA(c) , where λ is a parameter, which can be preset to 0.4; SA(c) is the estimated number of reaction steps required for the synthesis of compound c of unknown route; If the synthesis method of compound c has been searched, and there are N known single-step synthesis methods, P i ={p ij } represents the set of all precursors of the i-th synthesis method, y i represents the estimated yield of the i-th synthesis method, then the synthesis cost of compound c is the synthesis method with the lowest cost among the N synthesis methods, that is:

5. The interactive chemical synthesis route design method according to claim 4, characterized in that: A process for implementing the estimation of the number of reaction steps SA(c) required for the synthesis of a compound c of an unknown route includes: The number of synthesis steps is pre-set for each element in compound c; the sum of the number of synthesis steps of each element is the number of reaction steps SA(c); Among them, the various elements include the type and number of atoms, the type and number of functional groups, the length of the long carbon chain and the number and type of rings; the type and number of atoms include the number of carbon atoms and the number of heteroatoms; the type and number of functional groups include carbonyl, carboxyl, hydroxyl, cyano, amide, or / and acyl chloride; the number and type of rings include aromatic rings, heterocyclic rings, ring size and the number of parallel rings.

6. A method for designing a chemical synthesis route, characterized in that: include: obtaining a compound; Performing a step-by-step retrosynthesis process on the compound according to a preset screening condition, searching a retrosynthesis reaction template matching the compound in a reaction template library; wherein the cost of each reaction in the retrosynthesis process is obtained by a compound complexity estimator, and the complexity estimator estimates the number of reaction steps required for the synthesis of the compound of an unknown route; Obtaining a chemical synthesis route of the compound, wherein the compound is quantized into a molecular graph G(V,E), where the vertices V correspond to atoms and the edges E correspond to bonds; each reaction template in the reaction template library is quantized into a connected graph G(V,E) after adding virtual chemical bonds to the corresponding atoms in its reactants and products; Searching for the best retrosynthetic reaction template in the reaction template library based on a graph convolution method; Obtaining a reaction precursor of the corresponding compound based on each of the retrosynthetic reaction templates; Selecting a reaction precursor that meets the preset screening conditions and a corresponding retrosynthetic reaction template as a step in the chemical synthesis route design of the compound; If the reaction precursor is a compound that needs to be retrosynthesized, the above treatment is repeated for the compound that needs to be retrosynthesized until the reaction precursor obtained is a compound that does not need to be retrosynthesized, thereby completing all steps of the chemical synthesis route design for the compound.

7. An interactive chemical synthesis route design system, characterized in that: The interactive chemical synthesis route design system comprises: Input module to obtain a compound; A retrosynthesis processing module is connected to the input module for communication, and performs retrosynthesis processing on the compound step by step according to a preset screening condition, and searches for a retrosynthesis reaction template matching the compound in a reaction template library; wherein the cost of each reaction in the retrosynthesis process is obtained by a compound complexity estimator, and the complexity estimator estimates the number of reaction steps required for the synthesis of the compound of an unknown route; obtains a chemical synthesis route of the compound, wherein the compound is quantified into a molecular graph G(V, E), where vertices V correspond to atoms and edges E correspond to bonds; each reaction template in the reaction template library is quantified into a connected graph G(V, E) after adding virtual chemical bonds to the corresponding atoms in its reactants and products; and searches for the best retrosynthesis reaction template in the reaction template library based on a graph convolution method; An interactive module obtains a preference screening condition; A retrosynthesis update module is respectively connected to the interaction module and the retrosynthesis processing module for communication, and updates a certain step in the retrosynthesis processing according to the preference screening condition and updates the subsequent steps of a certain step in the retrosynthesis processing based on the preference screening condition and the preset screening condition; wherein the preference screening condition includes a preference for a specific route; a certain step in the chemical synthesis route is designated to be completed using a specific retrosynthesis reaction template according to the preference specific route; or the preference screening condition includes a disabling specific route; the disabling specific route is to prohibit an inappropriate reaction from occurring in the chemical synthesis route; an inappropriate reaction occurring in the chemical synthesis route is disabled according to the disabling specific route, and a step in which the inappropriate reaction occurs is replaced by a retrosynthesis reaction according to the preset screening condition.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 5 or the steps of the method according to claim 6 are implemented.

9. An electronic device, characterized in that: The electronic device comprises: a memory storing a computer program; A processor is communicatively connected to the memory, and implements the steps of the method according to any one of claims 1 to 5 or the steps of the method according to claim 6 when calling the computer program.

Images