Method for generating chemical reaction rule base and method for generating chemical reaction network
By generating a chemical reaction rule base and a topological matrix to represent the reaction process, a reaction network is automatically generated, which solves the limitations and inefficiencies of generators in existing technologies, enables the prediction of new paths and new reactions, and improves the efficiency and reliability of the generated reaction network.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- DALIAN INSTITUTE OF CHEMICAL PHYSICS CHINESE ACADEMY OF SCIENCES
- Filing Date
- 2021-12-14
- Publication Date
- 2026-06-09
AI Technical Summary
Existing reaction network generators have limitations in complex systems. They cannot automatically generate complete and reliable reaction networks, rely on experience, are inefficient, and cannot extract patterns from reaction mechanisms to predict new pathways and reactions.
By generating a chemical reaction rule base, using topological matrices to represent reaction processes, establishing reaction rule models, automatically generating reaction networks, extracting reaction patterns and storing them in the reaction rule base, and continuously strengthening them through training, the prediction of new paths and new reactions can be achieved.
It improves the efficiency of generating reaction networks, can automatically extract and predict new paths and reactions, reduces reliance on experience, and ensures the integrity and reliability of reaction networks.
Smart Images
Figure CN116264106B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of chemical reaction network generation algorithms, and in particular to a method for generating a chemical reaction rule base and a method for generating a chemical reaction network. Background Technology
[0002] For complex reaction systems, manually drawing reaction networks is an extremely arduous task, almost impossible. However, reaction networks for complex systems can be automatically generated by computers. The specific steps include: vectorizing species and reactions; establishing vector transformation algorithms to describe chemical reactions based on reaction categories; and automatically generating the reaction network through program iteration based on specific environmental and constraint conditions.
[0003] The method of automatically generating reaction networks by computer has been continuously developing in recent years and has been applied to various reaction systems such as catalytic cracking, isomerization, combustion, and biological processes. Reaction network generators for different systems are also under development, enabling us to detect not only known reaction pathways but also new ones without any human intervention. However, existing reaction network generators have certain limitations in achieving network generation. On the one hand, specialized reaction generators are only applicable to their respective systems, and these generation methods are usually based on experience, with a one-to-one correspondence between the established chemical reaction rules and reaction transformation algorithms. On the other hand, researchers need to establish reaction pathways based on experience, such as characterization results from tests and simulation calculations, making mechanistic studies not only time-consuming and inefficient but also unable to guarantee the integrity and reliability of the reaction network.
[0004] Currently, there is a lack of a method to automatically generate chemical reaction networks that can extract patterns from past reaction mechanism knowledge, form new reaction rules, and predict new pathways and reactions. Summary of the Invention
[0005] This invention proposes a novel method for the computer-automated generation of reaction networks driven by reaction knowledge. First, reaction equations are extracted from mechanisms in the literature, determining the reaction process and transition state changes, and represented using a topological matrix. Then, a topological matrix representation of the reaction process and a reaction process feature recognition algorithm are established; finally, these are transformed into reaction rules. In application, the reaction network is automatically generated based on input species information and corresponding constraints. This method can extract the reaction patterns from the initial reaction network and store them in a reaction rule model. The reaction rule model becomes the carrier of reaction patterns and can be trained by using available mechanistic reactions as input, continuously training them into reaction patterns stored in a reaction rule base. Researchers can automatically generate limited reaction network pathways based on newly discovered species or intermediate processes according to the reaction rule base, rather than relying on past experience, which improves their research efficiency and reduces the difficulty of finding reaction paths.
[0006] A method for generating a chemical reaction rule base includes the following steps:
[0007] 1) Based on the atoms and chemical bonds contained in the species, convert the reactant species and product species in the known chemical reaction equations into a topological matrix expression;
[0008] 2) Combine the topological matrices of the reactant species to obtain the topological matrix TS-f of the reaction state; combine the topological matrices of the product species to obtain the topological matrix TS-p of the product state; transform TS-f into TS-p according to the simplest transformation of the topological matrix, and record the transformation process as a reaction matrix;
[0009] The reaction center atom and substructure in the reaction species are obtained based on the bond breaking changes in the reaction matrix. The substructure includes the reaction center atom of the reaction species, other atoms within a set topological distance, and the connection mode between the atoms.
[0010] The substructure and reaction matrix are stored in the reaction rule submodule;
[0011] 3) The reaction rule sub-modules generated from each chemical reaction equation are combined to form a reaction rule library.
[0012] Optionally, the transformation process from TS-f to TS-p in step 2) is obtained using the following method:
[0013] Based on the existing TS-p matrix and the corresponding element sequence of TS-f, and following the principle of element conservation (i.e., corresponding elements of the same kind), TS-f is transformed into TS-p through congruent transformation; the simplest transformation is defined as the sum of the absolute values of each row and column of the reaction matrix that minimizes the sum of the absolute values.
[0014] Optionally, the reaction center atom in step 2) is determined using the following method:
[0015] In the obtained reaction matrix, 1 represents a bonding position, -1 represents a bond breaking position, and 0 represents a non-reacting position; the row or column number of the non-zero position in the reaction matrix is the central atom of the reaction.
[0016] Optionally, the method of converting the reaction species and product species in the known chemical reaction equation into a topological matrix expression based on the atoms and chemical bonds contained in the species in step 1) is as follows: number each atom of the species molecule, establish an adjacency matrix based on the number, and the numbers in the adjacency matrix represent chemical bond information;
[0017] Optionally, in the chemical bond information of the adjacency matrix: the value corresponding to a single bond connection is 1, the value corresponding to a double bond connection is 2, the value corresponding to a triple bond connection is 3, and the value corresponding to no connection is 0.
[0018] Optionally, the species topology matrix expression includes one of the expressions of gaseous molecules, molecular sieve active centers, and active intermediates.
[0019] Optionally, during the process of converting molecular sieves into a topological matrix representation, the atomic structure contained in the entire active site is numbered as a whole.
[0020] Optionally, the topological distance of the substructure is set to 1.
[0021] Optionally, a chemical reaction equation can be represented by six topological matrices, where two topological matrices represent the reactant species, two topological matrices represent the product species, the topological matrices of the two reactant species are combined to obtain the reaction state topological matrix TS-f, and the topological matrices of the two product species are combined to obtain the product state topological matrix TS-p.
[0022] Optionally, in step 3), two substructures corresponding to two reactant species are obtained respectively; if the reaction equation is a unimolecular reaction, then one substructure is set as an empty set.
[0023] Optionally, in step 4), reaction rule sub-modules with the same reaction center atom are merged into one type of reaction rule; then, the various types of reaction rules are combined to obtain a reaction rule library.
[0024] This application also proposes a method for analyzing reaction systems that require the generation of chemical reaction networks using the chemical reaction rule library obtained by the above generation method.
[0025] Optionally, the method includes the following steps:
[0026] a) Transform the reaction species, intermediate species, and product species that need to generate chemical reaction networks into topological matrix representations to form a species library;
[0027] b) Call the chemical reaction rule library and compare the topology matrix of each species in the species library; if the topology matrix in the species library contains all the substructures in a reaction rule submodule, then obtain the product topology matrix TS-p of the transition state based on the topology matrix and the reaction matrix in the reaction rule submodule, and obtain the product based on the product topology matrix TS-p.
[0028] c) After comparing all species in the species library using the chemical reaction rule base, list all the chemical reactions that occurred in step b) to obtain the chemical reaction network.
[0029] Optionally, if the product obtained in step b) is not a species already existing in the species library, the product is added to the species library, and the chemical reaction rule library is called again to compare the topological matrix of each species in the species library; until all the generated products are species already existing in the species library.
[0030] Optionally, step a) further includes adding the predicted possible chemical equations as a reaction rule submodule to the reaction rule library.
[0031] The beneficial effects that this application can produce include:
[0032] 1) By automatically extracting transformation patterns from existing chemical reaction equations, reaction rules are obtained, including both existing and unknown rules. This reaction rule model can also be trained and continuously strengthened to cover a wider range of more precise reaction rules.
[0033] 2) Based on the input reactants and the extracted reaction patterns, reaction networks can be generated to predict new pathways and reactions. Attached Figure Description
[0034] Figure 1 This is a schematic diagram illustrating the basic process of training and prediction of the reaction network in this invention.
[0035] Figure 2 This is a schematic diagram illustrating the topological matrix representation principle of chemical species in this invention.
[0036] Figure 3 This is a schematic diagram illustrating the topological matrix representation of chemical reactions in this invention.
[0037] Figure 4 This is a schematic diagram illustrating the storage and operation principles of the chemical reaction rules in this invention.
[0038] Figure 5This is a schematic diagram of the algorithm for automatically generating chemical reaction networks in this invention.
[0039] Figure 6 This is a schematic diagram of the automatic generation of the chemical reaction network in Example 2 of the present invention.
[0040] Figure 7 This is a schematic diagram of the automatic generation of the chemical reaction network in Example 3 of the present invention. Detailed Implementation
[0041] The present invention is described in detail below through examples, but the present invention is not limited to these examples. The following examples are implemented in MATLAB, but can also be implemented in other languages.
[0042] Example 1
[0043] like Figure 1 As shown, this embodiment proposes a method for generating a chemical reaction rule base and using this chemical reaction rule base to generate a chemical reaction network.
[0044] First, generate a chemical reaction rule base:
[0045] (I) Establish input files, including training files and execution files. Training files typically include, but are not limited to, known reaction equations. Execution files are the files that need to be used as input after the reaction rule base is established, and typically include species, intermediates, and products.
[0046] (II) Extract chemical reaction equations from the training files and convert the reactant and product species into topological matrix representations. Each species corresponds one-to-one with a topological matrix; standardization is required when constructing the species' topological matrix. See [link to documentation]. Figure 2 and Figure 3 Species and reactions are represented using a topological matrix. The topological matrix consists of the element sequence of the species' topological structure and the corresponding adjacency matrix of the species' structure diagram. The representation method involves numbering each atom in the species molecule, with the element sequence corresponding to an atom in the species molecule's topological structure, and the adjacency matrix representing its connectivity information: a single bond corresponds to a value of 1, a double bond to 2, a triple bond to 3, and no connection to 0. The topological matrix clearly represents the relationships between species atoms and facilitates the search for substructures. It is important to note that the entire active site of the molecular sieve is represented by the symbol "Z," which can be understood as one character representing the active site of an acidic site, such as... Figure 2 , which is the topological matrix representation of the species “Z-CH3”.
[0047] (III) Considering unimolecular and bimolecular reactions, a chemical reaction process is represented by six topological matrices. The topological matrix of the transition state is obtained by searching using the simplest topological matrix transformation method. Two topological matrices represent reactants, two represent products, and two represent the reaction state and product state of the transition state. For example... Figure 3 TS-f and TS-p represent the reaction state and product state of the transition state, respectively. Their transformation process can describe the change of the transition state and detail the bond breaking and recombination process. The topological matrix TS-f is obtained by combining reactant 1 and reactant 2, and the topological matrix TS-p is obtained by combining product 1 and product 2. According to element conservation, the topological matrices TS-f and TS-p have the same dimension. The transformation from TS-f to TS-p follows the principle of the simplest transformation, that is, the minimum number of bonding and bond breaking methods. This satisfies both the simplest transformation and the requirement that after a finite number of congruent transformations, TS-p can yield product 1 and product 2 through a separation algorithm. The reaction matrix R-mat = TS-f – TS-p is recorded.
[0048] The specific calculation method for the simplified transformation matrix is as follows: Based on the existing TS-p matrix and the element sequence corresponding to TS-f, and adhering to element conservation (i.e., corresponding elements of the same type), the TS-p matrix indices are arranged and combined to obtain multiple sets of indices. Each set of indices corresponds to a rearranged TS-p matrix (obtained from the original matrix through congruent transformation; the rearranged matrix and the original matrix express the same structure). Each set of TS-p matrices corresponds to a reaction matrix. The transformation corresponding to the reaction matrix is minimized (i.e., the sum of the absolute values of each row and column of the matrix is minimized), which is defined as the simplified transformation.
[0049] (IV) Based on the bond-breaking transformation of the transition state in the chemical reaction topology matrix, the reaction center atom is obtained, and the reactant substructures within a certain topological distance are obtained. The substructures and the reaction matrix are stored in the submodule of the reaction rule. The method for determining the reaction center atom is as follows: the reaction matrix is obtained by congruent transformation and contains only -1, 0, and 1, where -1 is the bond-breaking position, 0 is the non-reacting position, and 1 is the bond-forming position; therefore, the row or column number of the non-zero position in the reaction matrix is the reaction center atom.
[0050] See Figure 4Based on the established reaction topology matrix expression, after obtaining the transition state information containing the reaction, the characteristics of the reaction process are extracted and analyzed. These characteristics include the types of bonds broken and formed atoms, bond types and their indices, and the substructure information of the reactants. The atom forming the bond is defined as the reaction center, and the topological distance of the atom from the reaction center is defined as the reaction distance. The reaction distance characterizes the actual size and range of the substructure. The region included by reactants at a specific reaction center and a specific reaction distance is defined as the substructure of the reaction topology matrix (the reaction distance can be set according to needs; the larger the distance, the more precise the limitation of the chemical reaction). The substructure describes a part of a species. The topology matrix expression of a reaction rule consists of three parts: substructure 1, substructure 2, and the reaction matrix. If it is a unimolecular reaction, substructure 2 is an empty set. A reaction rule can have multiple topology matrix expressions. When the extraction process encounters the same reaction center, as long as the substructures are different, different matrix expressions will be corresponding to it, but they all belong to the same type of reaction.
[0051] (V) Combine submodules with the same reaction center to obtain reaction rules; then combine multiple reaction rules to obtain a reaction rule library. See also Figure 4 When running the model, the first step is to determine whether the topology matrix corresponding to the reactant species contains the topology corresponding to substructure 1 and substructure 2 in the corresponding reaction rule. If it does, the product topology matrix (TS-p) of the transition state is obtained based on the correspondence of the topology structures and combined with the topology reaction matrix in the model. Then, the product topology matrix is obtained through congruence transformation and matrix separation algorithm. If there is no corresponding structure, the reactant species cannot carry out the corresponding reaction.
[0052] Then, a chemical reaction network is generated using a chemical reaction rule base.
[0053] (VI) Extract the reaction species, intermediates and products, and reaction equations from the run file, and convert all species into a topological matrix representation.
[0054] (VII) Using the species from the previous step as input, and based on the constructed reaction rule base, generate a reaction network using an algorithm that automatically generates reaction networks. See also Figure 5 Reaction networks are established based on reaction rules. The automatic generation of reaction networks is divided into two methods, depending on whether new species generated by the reaction are used as reactants. Figure 5As shown on the left, newly formed species can be used as reactants again to continue the reaction, which is called a cyclic reaction network. The specific process is as follows: First, species are selected from the species library as reactants as input. Based on the reaction rules stored in the reaction rule model, an automatic matrix transformation is performed to generate the corresponding topological matrix representation of the product, while simultaneously recording the reaction process. Then, the product needs to be standardized. If it is a new species compared to those in the original species library, it is added to the species library. When the entire species library is traversed, the program ends, and the reaction network for the entire process is obtained.
[0055] The diagram on the right is called a primary reaction network, characterized by newly formed species not being used as new reactants, but being directly added to the final species and reaction library.
[0056] This invention can automatically extract the transformation patterns of existing reaction formulas to obtain reaction rules. This reaction rule model can be trained and continuously strengthened to cover a wider range of more accurate reaction rules. Based on the input reaction species and the extracted reaction rules, a reaction network can be generated to predict new paths and reactions. The following section lists and identifies both known and unknown reaction rules for prediction.
[0057] Example 2
[0058] This example demonstrates the extraction and verification of known reaction rules. For example... Figure 6 The responses selected from the prediction file include:
[0059] (1) Deprotonation reaction: [R1]ZCH(CH3)CH3==ZH+CH2=CHCH3;
[0060] (2) Protonation reaction: [R2]ZH+CH2=CHCH3==ZCH(CH3)CH3;
[0061] (3) Methylation reaction: [R3]CH2=CH2+ZCH3==ZCH2CH2CH3;
[0062] (4) Alkylation reaction: [R4]CH2=CH2+ZCH2CH3==ZCH2CH2CH2CH3;
[0063] (5) β-cleavage reaction: [R5]ZCH2CH2CH2CH3==CH2=CH2+ZCH2CH3.
[0064] To verify the prediction results, the species selected in the running file were all the reaction species in the prediction file, including: ZH, CH2=CH2, CH2=CHCH3, ZCH3, ZCH2CH3, ZCH(CH3)CH3, and ZCH2CH2CH2CH3. The topological distance of the substructure was set to 1.
[0065] According to the method for automatically generating chemical reaction networks by computer, the reaction equations in the prediction file are first extracted, and their reaction rules are used to establish a reaction rule library, consisting of four types of reaction rules. Then, based on the species in the input execution file and the trained reaction rule library, all possible reactions are obtained through a one-time reaction network generation method. The final reaction network includes 10 species, 4 types of reactions, and a total of 7 chemical reaction equations. The 5 reactions in the training file were all reproduced, and methylation and alkylation were unified into the same reaction type because they share the same reaction center.
[0066] Example 3
[0067] This example demonstrates the extraction and verification of unknown reaction rules. For example... Figure 7 The responses selected from the prediction file include:
[0068] [R1]ZH+CH3OH==ZO(H)(H)CH3、
[0069] [R2]ZH+CH3OCH3==ZO(CH3)(H)CH3、
[0070] [R3]ZO(CH3)(H)CH3+CH3OH==ZO(CH3)(CH3)CH3+H2O,
[0071] [R4]ZO(CH3)(CH3)CH3==ZH+CH3OCH2CH3、
[0072] [R5]ZO(H)(H)CH3+CH3OCH3=ZO(CH3)(CH3)CH3+H2O,
[0073] [R6]CH3OCH2CH3==CH2=CH2+CH3OH、
[0074] [R7]ZO(CH3)(CH3)CH3+ZCH3==ZO(CH3)(CH3)CH2CH3+ZH.
[0075] Select the species for the running file, including: CH3OH, ZH, H2O, ZO(H)(CH3)CH3, ZCH3, CH3OCH3. The topological distance of the substructure is set to 1.
[0076] According to the method for automatically generating chemical reaction networks by computer, the reaction equations in the prediction file are first extracted, and their reaction rules are obtained to establish a reaction rule library, consisting of 6 types of reaction rules. Then, based on the species in the input execution file and the trained reaction rule library, all possible reactions are obtained through a one-time generation of the reaction network. The final reaction network includes 8 species, 3 types of reactions, and a total of 5 chemical reaction equations.
[0077] The above description is merely a few embodiments of this application and is not intended to limit this application in any way. Although this application discloses preferred embodiments as described above, it is not intended to limit this application. Any changes or modifications made by those skilled in the art without departing from the scope of the technical solution of this application using the disclosed technical content are equivalent to equivalent implementation cases and fall within the scope of the technical solution.
Claims
1. A method for generating a chemical reaction rule base, characterized in that, Includes the following steps: 1) Based on the atoms and chemical bonds contained in the species, convert the reactant species and product species in the known chemical reaction equations into a topological matrix expression; 2) Combine the topological matrices of the reactant species to obtain the topological matrix TS-f of the reaction state; combine the topological matrices of the product species to obtain the topological matrix TS-p of the product state; transform TS-f into TS-p according to the simplest transformation of the topological matrix, and record the transformation process as a reaction matrix; The reaction center atom and substructure in the reaction species are obtained based on the bond breaking changes in the reaction matrix. The substructure includes the reaction center atom of the reaction species, other atoms within a set topological distance, and the connection mode between the atoms. The substructure and reaction matrix are stored in the reaction rule submodule; 3) The reaction rule sub-modules generated from each chemical reaction equation are combined to form a reaction rule library.
2. The method for generating a chemical reaction rule base according to claim 1, characterized in that, The transformation process from TS-f to TS-p in step 2) is obtained using the following method: Based on the existing TS-p matrix and the corresponding element sequence of TS-f, and following the principle of element conservation (i.e., corresponding elements of the same kind), TS-f is transformed into TS-p through congruent transformation. The simplest transformation is defined as the sum of the absolute values of each row and column of the reaction matrix that minimizes the sum of the absolute values.
3. The method for generating a chemical reaction rule base according to claim 1, characterized in that, The reaction center atom in step 2) is determined using the following method: In the obtained reaction matrix, 1 represents a bonding position, -1 represents a bond breaking position, and 0 represents a non-reacting position; the row or column number of the non-zero position in the reaction matrix is the central atom of the reaction.
4. The method for generating a chemical reaction rule base according to claim 1, characterized in that, In step 1), the method of converting the reaction species and product species in the known chemical reaction equation into a topological matrix expression based on the atoms and chemical bonds contained in the species is as follows: number each atom of the species molecule, establish an adjacency matrix based on the number, and the numbers in the adjacency matrix represent chemical bond information.
5. The method for generating a chemical reaction rule base according to claim 4, characterized in that, In the adjacency matrix, the chemical bond information is as follows: a single bond connection corresponds to a value of 1, a double bond connection corresponds to a value of 2, a triple bond connection corresponds to a value of 3, and no connection corresponds to a value of 0.
6. The method for generating a chemical reaction rule base according to claim 1, characterized in that, The species topology matrix expression includes one of the following: expression of gas phase molecules, molecular sieve active centers, and active intermediates.
7. The method for generating a chemical reaction rule base according to claim 1, characterized in that, In the process of converting molecular sieves into a topological matrix representation, the atomic structure contained in the entire active site is numbered as a whole.
8. The method for generating a chemical reaction rule base according to claim 1, characterized in that, A chemical reaction equation is represented by six topological matrices, where two topological matrices represent the reactant species, two topological matrices represent the product species, the topological matrices of the two reactant species are combined to obtain the reaction state topological matrix TS-f, and the topological matrices of the two product species are combined to obtain the product state topological matrix TS-p.
9. The method for generating a chemical reaction rule base according to claim 8, characterized in that, In step 2), two substructures corresponding to two reactant species are obtained respectively; if the reaction equation is a unimolecular reaction, then one substructure is set as an empty set.
10. The method for generating a chemical reaction rule base according to claim 1, characterized in that, In step 3), reaction rule sub-modules with the same reaction center atom are merged into one type of reaction rule; then, the various types of reaction rules are combined to obtain a reaction rule library.
11. A method for generating a chemical reaction network, characterized in that, The chemical reaction rule base obtained by any one of the generation methods in claims 1-10 is used to analyze the reaction system for which a chemical reaction network needs to be generated, thereby generating the chemical reaction network.
12. The method for generating a chemical reaction network according to claim 11, characterized in that, The method includes the following steps: a) Transform the reaction species, intermediate species, and product species that need to generate chemical reaction networks into topological matrix representations to form a species library; b) Call the chemical reaction rule base to compare the topological matrices of each species in the species library; If the topology matrix in the species library contains all the substructures in a reaction rule submodule, then the product topology matrix TS-p of the transition state is obtained based on the topology matrix and the reaction matrix in the reaction rule submodule, and the product is obtained based on the product topology matrix TS-p. c) After comparing all species in the species library using the chemical reaction rule base, list all the chemical reactions that occurred in step b) to obtain the chemical reaction network.
13. The method for generating a chemical reaction network according to claim 12, characterized in that, If the product obtained in step b) is not a species already in the species library, the product is added to the species library, and the chemical reaction rule library is called again to compare the topological matrix of each species in the species library; until all the generated products are species already in the species library.
14. The method for generating a chemical reaction network according to claim 12, characterized in that, Step a) further includes adding the predicted possible chemical equations as a reaction rule submodule to the reaction rule library.