Non-natural pathway generation method and system based on continuous biological template

Abstract

The invention belongs to the technical field of biology, and discloses a non-natural pathway generation method and system based on a continuous biological template, and the method comprises the following steps: obtaining an enzymatic reaction data set, and preprocessing the enzymatic reaction data set; constructing a directed reaction diagram on the basis of the preprocessed enzymatic reaction data set, performing similar compound search on a target compound by utilizing the directed reaction diagram, and obtaining a continuous group conversion template of the similar compound in the natural enzymatic pathway by taking the searched similar compound as a starting point and combining a depth-first search mode; the continuous group conversion template is applied to a target compound in a molecular linear input specification form, and an inverse synthesis process of the target compound is simulated to generate a non-natural synthesis pathway based on the continuous group conversion template. The method not only enhances the biological feasibility and construction success rate of the approach, but also has universality, expandability and engineering potential, and provides a generalizable normal form for the field of synthetic biology.

Description

Technical Field

[0001] This invention relates to the field of biological technology, and in particular to a method and system for generating non-natural pathways based on continuous biological templates. Background Technology

[0002] Currently, biotransformation is playing an increasingly crucial role in the green manufacturing of bulk and high-value-added chemicals due to its high selectivity, safety, environmental friendliness, and efficient ability to construct complex molecules. However, the evolution of natural metabolic pathways is driven by the physiological needs of organisms, rather than by optimal industrial production output. This often leads to challenges in industrial applications, such as limited pathway selection, low yields, numerous side reactions, and impaired kinetics. Therefore, designing and constructing non-natural metabolic pathways has become a key direction for achieving industrial-scale yields and efficiency.

[0003] In recent years, with the rapid development of interdisciplinary fields such as synthetic biology, systems biology and artificial intelligence, computational design of non-natural pathways has gradually become a research hotspot, especially showing great potential in synthetic systems that have never existed or are rare in nature, such as novel carbon fixation pathways and artificial photosynthesis. Traditional non-natural pathway development methods often suffer from problems such as low design efficiency, poor targeting and weak reproducibility.

[0004] Therefore, how to provide a method and system for generating non-natural pathways based on continuous biological templates is an urgent problem to be solved. Summary of the Invention

[0005] This invention provides a method and system for generating non-natural pathways based on continuous biological templates to solve the problems mentioned above in the prior art.

[0006] According to a first aspect of the present invention, a method for generating non-natural pathways based on continuous biological templates is provided.

[0007] In one embodiment, the method for generating non-natural pathways based on continuous biological templates includes: Obtain enzyme-catalyzed reaction datasets from a pre-defined enzyme-catalyzed reaction database and preprocess the enzyme-catalyzed reaction datasets; A directed reaction graph was constructed based on the preprocessed enzyme-catalyzed reaction dataset. The directed reaction graph was used to search for similar compounds of the target compound. Starting from the similar compounds found, a depth-first search method was used to obtain the continuous group transformation templates of similar compounds in the natural enzyme-catalyzed pathway. By applying continuous group transformation templates to target compounds in a linear molecular input canonical form and simulating the retrosynthetic process of the target compounds, a non-natural synthetic pathway based on continuous group transformation templates is generated.

[0008] In one embodiment, obtaining an enzyme-catalyzed reaction dataset from a preset enzyme-catalyzed reaction database and preprocessing the enzyme-catalyzed reaction dataset includes: The system retrieves raw enzyme-catalyzed reaction data from a pre-defined enzyme-catalyzed reaction database, updates the separators in the reaction formulas based on the reversibility status of the reactions in the raw enzyme-catalyzed reaction data, and filters the compound compartment information in the raw enzyme-catalyzed reaction data to obtain enzyme-catalyzed reaction data with reactivity significance. The reaction formulas are structured; at the same time, the linear input canonical structures of the metabolites are extracted based on the similarity of the compounds. The linear input canonical structures of the molecules are then subjected to structure cleaning and standardization processes to obtain reaction formulas with several linear input canonical structures of molecules. Reactions with completely identical molecular linear input canonical structures are merged to obtain several unique molecular linear input canonical reactions. Atom mapping is then performed on the unique molecular linear input canonical reactions to generate molecular linear input canonical reactions with mapping labels. The reaction template is extracted and corrected from the linear input canonical reaction with mapping labels, and a preprocessed enzymatic reaction dataset is generated.

[0009] In one embodiment, updating the separators in the reaction formula based on the reversibility state of the original enzymatic reaction data and filtering the compound compartment information in the original enzymatic reaction data to obtain enzymatic reaction data with reactivity significance includes: Obtain the reversibility state of the enzyme-catalyzed reaction from the raw enzyme-catalyzed reaction data, and based on the reversibility state, replace the separators in the reaction formula using a custom script; Analyze the compound lists on both sides of the reaction equation, iterate through each compound in the compound list to see if it appears in the preset auxiliary molecule list, and remove the compounds that appear in the auxiliary molecule list; Compounds lacking linear molecular input canonical representation in the raw enzymatic reaction data are filtered out, and compounds with identical exchange reactions on both sides are also filtered out to obtain enzymatic reaction data with structural reactivity significance.

[0010] In one embodiment, the reaction formula is structured; simultaneously, based on compound similarity, the molecular linear input canonical structures of metabolites are extracted, and the molecular linear input canonical structures are sequentially cleaned and standardized to obtain reaction formulas with several molecular linear input canonical structures, including: Based on the reaction formulas in the enzyme-catalyzed reaction data with reactivity significance, reaction formulas with unequal stoichiometry are split into equal forms according to the stoichiometric coefficients. At the same time, based on the number of compounds on the left and right sides of the reaction formula, compounds with many-to-many reactions are filtered out. The main substrate and product pairs of each reaction are determined by compound similarity, and the molecular linear input canonical structure of the metabolites is extracted based on the main substrate and product pairs to obtain several effective compound terms; The effective compound terms are sequentially subjected to structure cleaning and standardization to obtain several reaction formulas with linear input canonical structures of molecules.

[0011] In one embodiment, merging reactions with completely identical molecular linear input canonical structures to obtain several unique molecular linear input canonical reactions, and performing atom mapping on the unique molecular linear input canonical reactions to generate molecular linear input canonical reactions with mapping labels includes: Reactions with completely identical molecular linear input canonical structures are merged, and the annotation information of the molecular linear input canonical structures is merged into one information record, resulting in several unique molecular linear input canonical reactions. Construct a neural network model to format the linear input of a unique molecule into a standard input string using the neural network model; The system calls a preset function to perform atomic mapping on the standard input string, and outputs the probability distribution of the atomic mapping path through the encoding and decoding process of the neural network model; the atomic mapping path with the highest probability distribution is selected to obtain the molecular linear input canonical reaction with mapping label.

[0012] In one embodiment, extracting and modifying the reaction template from a molecular linear input canonical reaction with mapping labels includes: The linear input canonical reaction with mapping labels is used as input, and a predefined reaction class is called to parse the linear input canonical reaction with mapping labels. The template extraction interface is called to extract changes in reaction centers, bond breaking and formation, and stereo tags from the parsed reaction, and generate molecular linear input canonical reaction substructure template strings; The linear input canonical reactive substructure template string of molecules is filtered by neighborhood to obtain several reaction templates; the rationality of the reaction templates is evaluated, and unreasonable templates are screened out and corrected.

[0013] In one embodiment, evaluating the rationality of the reaction template and filtering out unreasonable templates from the reaction templates for correction includes: If the original substrate can be successfully reconstructed by applying the reaction template in reverse to the product molecules of the original reaction, the current reaction template is considered reasonable; otherwise, it is considered unreasonable, and the unreasonable reaction template is corrected.

[0014] In one embodiment, the construction of a directed reaction graph based on the preprocessed enzyme-catalyzed reaction dataset, the use of the directed reaction graph to search for similar compounds of the target compound, and the acquisition of continuous group transformation templates of similar compounds in natural enzymatic pathways using the searched similar compounds as starting points and combining a depth-first search method, includes: A directed reaction graph was constructed from the preprocessed enzyme-catalyzed reaction dataset. In the directed reaction graph, the linear input norm of molecules is used as nodes, the reaction template is used as directed edges, and the connectivity between nodes represents the actual enzyme-catalyzed reaction. Based on the directed reaction graph, the first round of similar compound search was performed on the target compound. Starting with the similar compounds obtained in the first round of search, and following the directed reaction graph, and combining it with a depth-first search method, we obtain the continuous group transformation templates of similar compounds in the natural enzymatic pathway.

[0015] In one embodiment, obtaining continuous group transformation templates of similar compounds in natural enzymatic pathways, starting from the similar compounds obtained in the first round of search, along the directed reaction graph, and combining a depth-first search approach, includes: Using the similar compounds obtained in the first round of search as the starting node, initialize the stack and mark the starting node as visited; Remove the current node from the stack and check all adjacent nodes connected to the current node through the reaction template. If the adjacent node has not been visited and is a compound in the natural enzymatic pathway, mark the adjacent node as visited and push it onto the stack. At the same time, record the reaction template connected to the adjacent node as the group transformation template. The search process is executed recursively until the stack is empty or the preset search depth is reached. All paths are explored through a backtracking mechanism, and the final collected continuous reaction template sequence is used as a continuous group transformation template.

[0016] According to a second aspect of the present invention, a non-natural pathway generation system based on a continuous biological template is provided.

[0017] In one embodiment, the non-natural pathway generation system based on continuous biological templates includes: The dataset processing module is used to obtain enzyme-catalyzed reaction datasets from a preset enzyme-catalyzed reaction database and to preprocess the enzyme-catalyzed reaction datasets. The group transformation template acquisition module is used to construct a directed reaction graph based on the preprocessed enzyme-catalyzed reaction dataset, and use the directed reaction graph to search for similar compounds of the target compound. Starting from the searched similar compounds, the module combines a depth-first search method to obtain continuous group transformation templates of similar compounds in the natural enzyme-catalyzed pathway. The pathway generation module applies the continuous group transformation template to the target compound in a linear molecular input specification and simulates the retrosynthetic process of the target compound to generate a non-natural synthetic pathway based on the continuous group transformation template.

[0018] According to a third aspect of the present invention, a computer device is provided.

[0019] In some embodiments, the computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of the method described above.

[0020] According to a fourth aspect of the present invention, a computer-readable storage medium is provided.

[0021] In one embodiment, a computer program is stored on the computer-readable storage medium, which, when executed by a processor, implements the steps of the above method.

[0022] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects: The non-natural pathway generation method provided by this invention discovers and optimizes novel metabolic pathways in a systematic and quantitative manner. This method not only significantly shortens the research and development cycle, reduces costs, and improves research efficiency, but also provides a new solution for the industrial biomanufacturing of complex functional molecules by improving product purity, yield, and reducing byproducts. It not only enhances the biological feasibility and construction success rate of the pathway, but also has universality, scalability, and engineering potential, providing a paradigm that can be promoted in the field of synthetic biology.

[0023] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit the invention. Attached Figure Description

[0024] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0025] Figure 1 This is a flowchart illustrating a method for generating non-natural pathways based on continuous biological templates according to an exemplary embodiment; Figure 2 This is a schematic diagram illustrating the principle of a non-natural pathway generation system based on a continuous biological template, according to an exemplary embodiment. Figure 3 This is a schematic diagram of the structure of a computer device according to an exemplary embodiment; Figure 4This is a schematic diagram illustrating a common combination of cofactors and circulating metabolites in a non-natural pathway generation method based on a continuous biological template, according to an exemplary embodiment. Figure 5 This is a flowchart illustrating the MetaCyc database reaction data processing and cleaning process in a non-natural pathway generation method based on a continuous biological template, according to an exemplary embodiment. Figure 6 This is a basic flowchart illustrating the generation of non-natural pathways in a method for generating non-natural pathways based on a continuous biological template, according to an exemplary embodiment. Figure 7 This is a schematic diagram of the homepage of NeoPathTP, an online website for non-natural pathway design, in a non-natural pathway generation method based on continuous biological templates, according to an exemplary embodiment. Figure 8 This is a schematic diagram of the Submission page of NeoPathTP, an online website for non-natural pathway design, in a non-natural pathway generation method based on continuous biological templates, according to an exemplary embodiment. Figure 9 This is a schematic diagram of the Result page of NeoPathTP, an online website for non-natural pathway design, in a non-natural pathway generation method based on continuous biological templates, according to an exemplary embodiment. Figure 10 This is a schematic diagram of a non-natural synthetic pathway of 2,4-DHB using L-malic acid as a synthetic precursor in a non-natural pathway generation method based on a continuous biological template, according to an exemplary embodiment. Figure 11 This is a schematic diagram of a non-natural synthetic pathway of pHS using L-tyrosine as a synthetic precursor in a non-natural pathway generation method based on a continuous biological template, according to an exemplary embodiment. Figure 12 This is a schematic diagram of a non-natural synthetic pathway of 1,3-PDO using malonyl-CoA as a synthetic precursor in a non-natural pathway generation method based on a continuous biological template, according to an exemplary embodiment. Detailed Implementation

[0026] The following description and accompanying drawings fully illustrate specific embodiments described herein to enable those skilled in the art to practice them. Some portions and features of certain embodiments may be included in or replace portions and features of other embodiments. The scope of the embodiments herein includes the entire scope of the claims and all available equivalents thereof. The various embodiments described herein are presented in a progressive manner, with each embodiment focusing on its differences from other embodiments; similar or identical parts between embodiments can be referred to interchangeably.

[0027] The modules in the apparatus or system of this application can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0028] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0029] Figure 1 An embodiment of a non-natural pathway generation method based on a continuous biological template according to the present invention is shown.

[0030] In this optional embodiment, the method for generating non-natural pathways based on continuous biological templates includes: Step S101: Obtain the enzyme-catalyzed reaction dataset from the preset enzyme-catalyzed reaction database and preprocess the enzyme-catalyzed reaction dataset; Step S102: Construct a directed reaction graph based on the preprocessed enzyme-catalyzed reaction dataset, and use the directed reaction graph to search for similar compounds of the target compound. Starting from the similar compounds found, combine the depth-first search method to obtain the continuous group transformation templates of similar compounds in the natural enzyme-catalyzed pathway. Step S103: Apply the continuous group transformation template to the target compound in the form of molecular linear input canonicalization, and simulate the retrosynthetic process of the target compound to generate a non-natural synthetic route based on the continuous group transformation template.

[0031] In this optional embodiment, obtaining the enzyme-catalyzed reaction dataset from a preset enzyme-catalyzed reaction database and preprocessing the enzyme-catalyzed reaction dataset includes: Raw enzyme-catalyzed reaction data is obtained from a pre-defined enzyme-catalyzed reaction database. The separators in the reaction formulas are updated based on the reversibility state of the reactions in the raw data. Compound compartmentalization information in the raw data is filtered to obtain enzyme-catalyzed reaction data with reactivity significance. The reaction formulas are then structured. Simultaneously, based on compound similarity, the linear input canonical structures of metabolites are extracted. These canonical structures are then subjected to structure cleaning and standardization to obtain several reaction formulas with identical linear input canonical structures. Reactions with completely identical linear input canonical structures are merged to obtain several unique linear input canonical reactions. Atom mapping is then performed on these unique linear input canonical reactions to generate linear input canonical reactions with mapping labels. Reaction templates are extracted and corrected from these labeled linear input canonical reactions, and a preprocessed enzyme-catalyzed reaction dataset is generated.

[0032] In this optional embodiment, updating the separators in the reaction formula based on the reversibility state of the original enzymatic reaction data and filtering the compound compartment information in the original enzymatic reaction data to obtain enzymatic reaction data with reactivity significance includes: The system obtains the reversibility state of the enzyme reaction in the raw enzyme reaction data and replaces the separators in the reaction formula based on the reversibility state using a custom script. It parses the compound lists on both sides of the reaction formula, iterates through each compound in the compound list to see if it appears in the preset auxiliary molecule list, and removes the compounds that appear in the auxiliary molecule list. It filters out compounds in the raw enzyme reaction data that lack linear molecular input canonical representation, and also filters out compounds with completely identical exchange reactions on both sides, to obtain enzyme reaction data with structural reactivity significance.

[0033] In this optional embodiment, the reaction formula is structured; simultaneously, based on compound similarity, the molecular linear input canonical structure of the metabolite is extracted, and the molecular linear input canonical structure is sequentially cleaned and standardized to obtain a reaction formula with several molecular linear input canonical structures, including: Based on the reaction formulas in the enzyme-catalyzed reaction data with reactivity significance, reaction formulas with unequal stoichiometry are split into equal-quantity forms according to the stoichiometric coefficients. At the same time, based on the number of compounds on both sides of the reaction formula, compounds with multiple-to-multiple reactions are filtered out. The main substrate and product pairs of each reaction are determined by the similarity of the compounds, and the linear input canonical structures of the metabolites are extracted based on the main substrate and product pairs to obtain several effective compound terms. The effective compound terms are then subjected to structure cleaning and standardization processes to obtain several reaction formulas with linear input canonical structures.

[0034] In this optional embodiment, the step of merging reactions with completely identical molecular linear input canonical structures to obtain several unique molecular linear input canonical reactions, and performing atom mapping on the unique molecular linear input canonical reactions to generate molecular linear input canonical reactions with mapping labels includes: Reactions with completely identical molecular linear input canonical structures are merged, and the annotation information of the molecular linear input canonical structures is merged into a single information record, resulting in several unique molecular linear input canonical reactions. A neural network model is constructed, and the unique molecular linear input canonical reactions are formatted into a standard input string through the neural network model. A preset function is called to perform atom mapping on the standard input string, and the probability distribution of the atom mapping path is output through the encoding and decoding process of the neural network model. The atom mapping path with the highest probability distribution is selected to obtain the molecular linear input canonical reactions with mapping labels.

[0035] In this optional embodiment, the extraction and modification of the reaction template from the molecular linear input canonical reaction with mapping labels includes: Taking linear input canonical reactions with mapping tags as input, the system calls a predefined reaction class to parse these reactions. It then uses a template extraction interface to extract changes in reaction centers, bond breaking and formation, and stereo tags from the parsed reactions, generating a substructure template string for the linear input canonical reaction. This substructure template string is then filtered by neighborhood to obtain several reaction templates. Finally, the rationality of the reaction templates is evaluated, and unreasonable templates are selected for correction.

[0036] In this optional embodiment, the step of evaluating the rationality of the reaction template and filtering out unreasonable templates from the reaction templates for correction includes: If the original substrate can be successfully reconstructed by applying the reaction template in reverse to the product molecules of the original reaction, the current reaction template is considered reasonable; otherwise, it is considered unreasonable, and the unreasonable reaction template is corrected.

[0037] In this optional embodiment, the construction of a directed reaction graph based on the preprocessed enzymatic reaction dataset, the use of the directed reaction graph to search for similar compounds of the target compound, and the acquisition of continuous group transformation templates of similar compounds in natural enzymatic pathways using the searched similar compounds as starting points and combining a depth-first search method, includes: A directed reaction graph is constructed from the preprocessed enzyme-catalyzed reaction dataset. In the directed reaction graph, the linear input norm of molecules is used as nodes, and the reaction template is used as directed edges. The connectivity between nodes represents the actual enzyme-catalyzed reaction. Based on the directed reaction graph, a first round of similar compound search is performed on the target compound. Starting from the similar compounds obtained in the first round of search, along the directed reaction graph and combined with a depth-first search method, continuous group transformation templates of similar compounds in the natural enzyme-catalyzed pathway are obtained.

[0038] In this optional embodiment, the step of using the similar compounds obtained in the first round of search as a starting point, following the directed reaction graph, and combining a depth-first search method to obtain continuous group transformation templates for similar compounds in natural enzymatic pathways includes: Using the similar compounds obtained in the first round of search as the starting node, initialize the stack and mark the starting node as visited; Remove the current node from the stack and check all adjacent nodes connected to the current node through the reaction template. If the adjacent node has not been visited and is a compound in the natural enzymatic pathway, mark the adjacent node as visited and push it onto the stack. At the same time, record the reaction template connected to the adjacent node as the group transformation template. The search process is executed recursively until the stack is empty or the preset search depth is reached. All paths are explored through a backtracking mechanism, and the final collected continuous reaction template sequence is used as a continuous group transformation template.

[0039] Figure 2 An embodiment of a non-natural pathway generation system based on a continuous biological template according to the present invention is shown.

[0040] In this optional embodiment, the non-natural pathway generation system based on continuous biological templates includes: The dataset processing module 201 is used to obtain enzyme-catalyzed reaction datasets from a preset enzyme-catalyzed reaction database and to preprocess the enzyme-catalyzed reaction datasets. The group transformation template acquisition module 202 is used to construct a directed reaction graph based on the preprocessed enzyme-catalyzed reaction dataset, and use the directed reaction graph to search for similar compounds of the target compound. Starting from the searched similar compounds, the module combines a depth-first search method to obtain continuous group transformation templates of similar compounds in the natural enzyme-catalyzed pathway. The pathway generation module 203 applies the continuous group transformation template to the target compound in a linear molecular input specification and simulates the retrosynthetic process of the target compound to generate a non-natural synthetic pathway based on the continuous group transformation template.

[0041] To facilitate understanding of the above technical solutions of the present invention, the following further describes the above technical solutions of the present invention from the perspectives of architecture and principle, as follows: Dataset Construction: This invention aims to design non-natural synthetic pathways for target products based on continuous group transformation templates from natural enzymatic pathways. Therefore, a comprehensive and rich database of known enzymatic reactions is required. This invention selects MetaCyc version 29.0 as the base database for the following reasons: (1) Complete molecular SMILES (Simplified molecular input line entry system, a standard that explicitly describes molecular structure using ASCII strings) representations and their correspondences: Most compounds and reactions in MetaCyc come with standardized SMILES representations, which can be used to directly construct directed networks of natural enzymatic reactions. In non-natural pathway analysis, both template-based and graph-based methods heavily rely on SMILES characterization. Using a database with complete and accurate SMILES data can effectively avoid the loss or mapping errors of SMILES due to inconsistencies in compound molecule IDs or names across different databases, thereby reducing the risk of missing metabolic network information.

[0042] (2) Abundant and comprehensive information on enzyme-catalyzed reactions: MetaCyc is a manually compiled database of metabolic pathways across multiple species, covering primary and secondary metabolic processes. All pathways included are derived from experimentally validated literature records, making it one of the world's largest databases of experimentally proven metabolic pathways. As of the latest version, the database contains thousands of pathways and tens of thousands of experimentally supported reactions and their corresponding enzyme information. Furthermore, MetaCyc has gradually integrated some non-natural reactions, further enhancing its practicality in discovering retrosynthetic pathways.

[0043] Therefore, the MetaCyc29.0 database not only provides comprehensive and standardized SMILES information that can be used to construct molecular reaction networks, but also covers a wide range of enzymatic reaction pathways, making it suitable as the data basis for generating non-natural synthetic pathways in this invention.

[0044] Dataset processing: Based on 39,696 enzymatic reaction data points from MetaCyc version 29.0, the separators in the reaction equations were first updated according to the information in the "reversible" column, and the compartmentalization information of the compounds was removed. Specifically, for updating the separators, the separators in the original reaction equations were all >>, representing the direction of the reaction from left to right; based on the value of the "reversible" column, if it was True (indicating that the reaction is reversible), the separators were replaced with <=> using a custom Python script; if it was False (indicating that the reaction is irreversible, i.e., from left to right), it was replaced with =>.

[0045] In non-natural pathways, compartmentalization information of compounds is typically ignored because this information is primarily used for intracellular localization and has no impact on the structural transformation nature of the reaction. Meanwhile, to more accurately characterize the main transformation processes of the reaction, this invention removes auxiliary components that do not directly participate in substrate transformation. This removal process is based on a manually collected and maintained list of auxiliary molecules, such as common cofactor pairs (e.g., NAD⁺ / NADH, ATP / ADP, etc.) and circulating metabolites. These substances are mostly involved in energy conversion or charge balance and do not constitute the core structural changes of the reaction; therefore, they are uniformly filtered during pathway planning, specifically as follows: Figure 4 As shown. A custom filtering algorithm is used: First, the lists of compounds on both sides of the reaction equation are parsed. Then, each compound in the list is iterated, and if it appears in the auxiliary molecule list, it is removed from the equation. Simultaneously, since some compounds lack available SMILES representations in MetaCyc, and these compounds are mostly complex macromolecules or polymers unsuitable for non-natural pathway design, they are also removed. Next, exchange reactions with completely identical compounds on both sides are identified, ultimately retaining 19264 reactions with structural reactivity significance.

[0046] In the reaction structuring process, reactions with unequal stoichiometry were split into equal-quantity forms based on stoichiometric coefficients (e.g., 2.0pyr→alac__S was processed into pyr+pyr→alac__S), and reversible reactions were split into two forward reactions for easier unified screening and template extraction. Subsequently, based on the number of compounds on both sides of the reaction, only one-to-one, one-to-many, and many-to-one reactions were retained, while many-to-many reactions were removed, as the latter are difficult to rationally decompose through reaction mechanisms and were temporarily disregarded. Next, the main substrate and product pairs of each reaction were determined by compound similarity, and the SMILES structures of metabolites were extracted from MetaCyc (excluding substances with R groups and unknown groups), resulting in 20,249 valid compound entries. After structural cleaning (normalizing charge, stereo information, removing explicit hydrogen ions, etc.), the compound IDs in the reaction were replaced with SMILES. If any compound lacked a SMILES, the reaction was discarded. Simultaneously, the SMILES structures of CoA and SAH were uniformly replaced with [La] and [Ce], respectively, ultimately yielding 19,463 reaction formulas represented by SMILES.

[0047] Next, reactions with completely identical SMILES structures were merged, and their annotation information (such as NAME, ID, GENE, DATALINK, etc.) was combined into a single information record, resulting in 15,346 unique SMILES reactions. Atom mapping was then performed using RXNmapper, a neural network model based on the Transformer architecture (such as an ALBERT variant). RXNmapper automatically infers and constructs coherent atomic correspondence rules from the original chemical reaction SMILES through unsupervised learning. Its core principle is to use an attention mechanism to capture the atomic-level structural similarity and change patterns between reactants and products, avoiding the limitations of traditional heuristic rules.

[0048] The specific implementation process of using RXNmapper for atom mapping is as follows: Install the rxnmapper library in the Python environment (via pip install rxnmapper), load the pre-trained model, then format the unique SMILES reaction as a standard input string ('reactant>>product', where reactants and products are represented by SMILES), and call the map_reaction function to perform atom mapping. This function outputs a probability distribution through the model's encoding-decoding process, selects the atom mapping path with the highest probability to generate SMILES output with mapping labels (such as corresponding atom numbers), thus providing accurate reaction center identification for subsequent template extraction; then use RDChiral to extract the reaction template. RDChiral is an open-source Python wrapper for RDKit, designed specifically for retrosynthetic template extraction and processing stereochemical information in applications. Its architecture extends RDKit's core molecular operation interfaces (such as Mol and chemical reaction classes) to achieve consistent stereo configuration tracking and template generalization. The principle is to identify reaction centers (bond changes) and their neighboring atoms (including chiral centers) from the atom-mapped reaction, generate abstract SMARTS patterns to capture general reaction rules, and retain stereo information to avoid template overgeneralization.

[0049] The specific implementation process for extracting reaction templates is as follows: install the rdchiral library (pip install rdchiral), load the RDKit environment, input the reaction SMILES mapped by RXNmapper, use the rdchiral reaction class to parse the molecules, call the extract_from_reaction function (or a similar template extraction interface) to analyze changes in reaction centers, bond breaking / formation, and stereo labels, automatically generate SMARTS template strings (e.g., [:1]-C:2[O:4]>>[:1]-C:2[:3], where: represents variable atoms), and ensure template robustness through neighborhood filtering. Due to limitations in character length and structural complexity of RDChiral, 12707 mappable reaction template rules were successfully extracted, of which 7741 were unique templates.

[0050] In the initial evaluation, the template was reversed and applied to the product molecule of the original reaction. If the original substrate could be successfully reconstructed, the template was considered valid. The initial evaluation showed that approximately 81% of the templates could be correctly verified. Further analysis of the templates that failed verification revealed that many erroneous templates originated from atomic mappings containing groups in the substrate and product whose structures did not change. In response, batch corrections of erroneous mappings were performed, significantly improving template accuracy. The verification pass rate of the corrected templates increased to approximately 93%, laying a more reliable foundation for reaction templates in subsequent non-natural pathway design and reaction prediction. The data processing flowchart is shown below. Figure 5 As shown.

[0051] exist Figure 5 In this context, MetaCyc refers to the MetaCyc database; Remove Compartments means removing compartments (in metabolic networks, this refers to removing cellular structural partitions such as cytoplasm or mitochondria); Reaction Without Compartments means reactions without compartments; Remove No Formula Metabolites means removing metabolites without formulas (meaning removing metabolites lacking molecular formula or structural definition); Reaction Without No Formula Metabolites means reactions without metabolites with defined molecular formulas (i.e., reactions that have been cleaned up to remove metabolites without formulas); Split Reaction and Represent Using SMILES means splitting reactions and representing them using SMILES (SMILES stands for Simplified Linear Input Specification, used to standardize molecular structure description); Single Reaction SMILES means single reaction SMILES; Atomic Mapping and Extract Template means mapping atoms and extracting templates (atomic mapping refers to tracking the correspondence between atoms in a reaction from reactants to products, and template extraction refers to abstracting a general reaction pattern); Reaction Template means reaction template; and Reaction Without Currency means reaction template. Metabolites indicates reactions without circulating metabolites (circulating metabolites refer to universal cofactors such as ATP or NADH), Remove Currency Metabolites indicates removing circulating metabolites, Save Template indicates saving the template, and Manual indicates manual (referring to manually handling incorrect reaction templates).

[0052] Non-natural pathway design process: First, the processed MetaCyc reaction dataset is constructed into a directed reaction graph, where molecular SMILES are used as nodes and reaction templates are used as directed edges. The connectivity between nodes represents the actual enzymatic reactions. Based on this, the first round of similar compound search is performed on the target compound: the molecules are encoded using the Morgan fingerprint (ECFP) widely used in the RDKit toolkit. The Tanimoto similarity between the target compound and known compounds in the database is calculated (the specific calculation formula is T=Nc / (Na+Nb-Nc), where Na and Nb represent the number of bits set to 1 in the fingerprints of the target compound and the known compounds in the database, respectively, and Nc represents the number of bits set to 1 in the same position in both fingerprints. The Tanimoto similarity value ranges from [0,1], and the closer the value is to 1, the more similar the structures of the two molecules are. The candidates that are most similar to the target molecule are selected (the 1st, 2nd and 3rd most similar molecules, or even all molecules with similarity in the range of 0.5 to 1.0, can be selected) to ensure that the selected templates are sufficiently representative.

[0053] Starting with these similar compounds, a depth-first search (DFS) approach is used to automatically obtain their continuous group transformation templates in natural enzymatic pathways along a constructed directed graph. The specific implementation steps include: using the screened similar compounds as starting nodes, initializing the access stack and marking the starting node as visited; removing the current node from the stack and checking all adjacent molecular nodes connected to it through the reaction template; if an adjacent node is not visited and is a compound in the natural enzymatic pathway, it is marked as visited and pushed onto the stack, while the reaction template connected to that node is recorded as the group transformation template; this process is recursively executed until the stack is empty or the preset search depth is reached. During this process, all possible paths are explored through a backtracking mechanism (i.e., restoring the current access state when returning to the previous node). The final collected continuous reaction template sequence is the required continuous group transformation template.

[0054] The obtained templates, expressed as SMILES / SMARTS, are applied to the target compound using RDKit to simulate its stepwise retrosynthetic process, thereby generating a rational synthetic pathway. The core of this process lies in the automatic extraction of reaction templates from known enzymatic reactions, which generate interpretable and traceable pattern rules based on atom mapping and contextual environment. Through this method, the framework can integrate natural reaction information from structurally similar leading molecules and a graph structural pathway retrieval mechanism to achieve retrosynthetic design of the target compound. The specific process is as follows: Figure 6 As shown, in Figure 6In this context, Target mol represents the target molecule; Similar mol represents similar molecules; Similarity represents the similarity level; and Multi-step template represents a multi-step template.

[0055] Building an online website through non-natural means: NeoPathTP encapsulates the generation of non-natural pathways into an online platform for non-natural pathway design. Based on continuous transformation templates of natural reactions, NeoPathTP constructs a system for analyzing and predicting non-natural biosynthetic pathways. It supports the automatic planning of multi-step non-natural retrosynthetic pathways starting from user-input target compound SMILES. Its core functions mainly include two aspects: firstly, a non-natural pathway prediction module, where the system automatically calculates and recommends the optimal non-natural synthesis scheme based on the user-input target molecular structure using internal algorithms; secondly, a non-natural enzyme recommendation service, where the platform matches highly efficient and usable non-natural catalytic enzymes (including relevant genes and EC numbers) for user reference based on the recommended chemical reaction templates. The entire platform is clearly structured and easy to operate in terms of its delivery interface and functional design, as shown on the website homepage. Figure 7 As shown, in Figure 7 NeoPathTP is a versatile, web-based tool designed for synthetic biologists and metabolic engineers. Utilizing natural, sequential reaction templates, NeoPathTP enables the rational design of non-natural biosynthetic pathways with high precision and efficiency. By simply providing the target compound or desired transformation, users can explore custom synthetic routes beyond natural metabolism. NeoPathTP is a vital resource for accelerating the discovery and engineering of innovative biosynthetic systems.

[0056] After entering the Submission page, users need to enter the SMILES representation of the target compound in the text box. The system also supports users selecting suitable precursor molecules from a pre-set library of commonly used Building Blocks. Furthermore, users can personalize the pathway planning process, such as setting the Similarity Range (the Tanimoto similarity range used to screen for similar known compounds) and Maximum Steps (the maximum number of reaction steps for retrosynthetic pathway searching), to control the search space and pathway complexity. Figure 8 As shown, Figure 8 This is the website task submission interface. Here, "Target Compound" represents the target molecule; "Similarity Range" represents the similarity range; "Maximum Steps" represents the maximum number of steps; and "Email" represents the email address. After the user confirms the parameters and clicks "Submit," the platform will automatically start the analysis process.

[0057] After generating the prediction results (averaging about 1 minute per molecule), NeoPathTP's Result page will graphically compare the user's reference template with the system's predicted non-natural pathways, allowing users to intuitively view the reaction information at each step. Furthermore, by right-clicking on a compound or reaction in the graph, detailed information for that node will pop up in real time, including the compound's NAME, ID, SMILES structure, and the NAME, GENE, EC Number associated with the reaction template, as well as the thermodynamic data GIBBS value, such as... Figure 9 As shown, the Reference Result represents the reference result, and the Predicted Result represents the predicted result. This interactive design not only makes the prediction results clear in hierarchy, but also provides detailed evidence for subsequent path verification and experimental design.

[0058] Performance of non-natural pathways on the test set: To verify the effectiveness of this embodiment in predicting real-world non-natural pathways, this invention manually collected 55 experimentally successfully constructed non-natural synthetic cases as a test set. Based on the natural pathways of all known compound molecules with a target compound similarity between 0.5 (inclusive) and 1.0 (exclusive), continuous group transformation templates were extracted for non-natural pathway generation. The prediction process terminated under two conditions: either the maximum number of steps set by the user (five steps in this case) was reached, or the generated intermediate product appeared in a preset substrate library. The substrate pool used in this experiment included 113 common molecules related to central metabolism and amino acid synthesis in *E. coli*, as starting materials.

[0059] Under the above settings, at least one complete synthetic pathway was successfully predicted for 23 out of 55 target compounds. However, 8 target compounds only generated one pathway or contained only a single-step reaction, indicating limited overall predictive ability. Further statistical analysis showed that the average number of solutions for the tested target compounds was 0.96, less than 1, indicating that nearly half of the molecules could not obtain an effective pathway. This suggests that the number and complexity of prediction results may be limited by the size of the substrate pool.

[0060] To verify this hypothesis, the substrate pool was expanded to include all known compound molecules (denoted as SMILES), and the pathway search was repeated while keeping other conditions unchanged. The results showed that the number of target compounds that could be successfully traced back to the initial Building Blocks significantly increased from 23 to 45, with a substantial increase in both the number and coverage of pathways. Among these, some molecules (such as 5-Aminovalerate, 6-Aminocaproic acid, 5-Hydroxytryptophan, 5-Hydroxyvalericacid, and 1,6-Hexanediamine) yielded more than 10 feasible pathways. The average number of solutions for all tested target molecules also increased to 3.3, indicating a significant enhancement in the quantity and diversity of pathway generation. This result further validates the crucial role of substrate pool setting in the prediction of non-natural synthetic pathways. The larger substrate coverage significantly improved the coverage and richness of the pathway search, consistent with the conclusion in previous studies that "multi-substrate strategies help improve prediction performance."

[0061] Analysis of some specific results: In the design of 2,4-dihydroxybutyric acid (2,4-DHB), based on 2,051 reference pathways, four efficient synthetic pathways tracing back to core building blocks of natural metabolism (such as aspartate family metabolic nodes) were successfully predicted within five steps. The first predicted pathway (L-malic acid → L-maleyl-4-phosphate → L-malic acid-4-semialdehyde → 2,4-dihydroxybutyric acid) is completely consistent with engineered synthetic pathways reported in existing literature. This pathway utilizes the synergistic action of transaminases and reductases from natural homoserine to homoserine metabolic branches to extend the carbon chain from the aspartic acid family, and then achieves the synthesis of the target product through a combination of phosphorylation and reduction. This is highly consistent with the natural-non-natural coupling pathways constructed in existing literature centered on homoserine transaminases and OHB reductases, such as... Figure 10 As shown ( Figure 10This diagram illustrates the non-natural pathway of 2,4-DHB starting from L-malic acid and the reference natural pathway, where L-aspartate, L-aspartyl-4-phosphate, L-aspartyl-4-semialdehyde, L-homoserine, L-malate, L-malyl-4-phosphate, L-malyl-4-semialdehyde, and L-2,4-DHB are all represented by L-2,4-dihydroxybutyric acid. Meanwhile, another pathway identified in this example (L-homoserine → 4-hydroxy-2-oxobutyric acid → 2,4-DHB) is consistent with pathways reported in existing literature, further validating the effectiveness of the "natural continuous template-guided non-natural pathway design" strategy and its feasibility in identifying the coupling between natural catalytic nodes and non-natural extended reactions.

[0062] Similarly, in the pathway design for p-hydroxystyrene (pHS), starting from the natural degradation pathway of phenylalanine / tyrosine, two core synthetic pathways were predicted. Pathway 1 (L-tyrosine → coumaric acid → pHS) involves the synergistic catalysis of tyrosine ammonia-lyase (TAL, EC4.3.1.23) and aromatic carboxylate decarboxylase (EC4.1.1.102) to achieve the formation and decarboxylation of p-hydroxycinnamic acid as the target product, such as... Figure 11 As shown ( Figure 11 A schematic diagram of the non-natural pathway for p-hydroxystyrene starting from tyrosine, where Glucose, Tyrosine, pHCA, pHS, Tyrosine ammonialyase, and pHCA decarboxylase are all involved. Pathway two (L-phenylalanine → trans-cinnamic acid → coumaric acid → pHS) relies on cytochrome P450 monooxygenase to hydroxylate the benzene ring, followed by decarboxylation to generate pHS. This pathway aligns with existing literature on the synthetic pathways of L-phenylalanine-derived aromatic compounds. Both pathways achieve highly selective biosynthesis of p-hydroxystyrene through the organic combination of natural aromatic modifying enzymes and non-natural decarboxylation steps.

[0063] Regarding 1,3-propanediol (1,3-PDO), the 3-hydroxypropionic acid (3-HP) pathway designed in this embodiment is highly consistent with the pathways reported in existing literature, such as... Figure 12 As shown ( Figure 12A schematic diagram of the non-natural pathway of 1,3-propanediol starting from malonyl-CoA, where Malonyl-CoA, 3-Hydroxypropionate, 3-hydroxypropionic acid, 3-Hydroxypropionyl-CoA, 3-Hydroxypropionaldehyde, and 1,3-Propanediol are all present. The predicted pathway in this embodiment (malonyl-CoA → 3-HP → 3-hydroxypropionyl-CoA → 3-hydroxypropionaldehyde → 1,3-propanediol) utilizes a cascade catalysis of reductase (EC1.1.1.298) and aldehyde reductase (EC1.1.1.59), starting from the natural intermediate malonyl-CoA, bypassing the ATP-dependent mechanism through atypical steps. Theoretically, this could improve ATP utilization efficiency and provide a new approach for optimizing synthetic pathways.

[0064] The prediction pathways for the three non-natural products described above are highly consistent with existing experimental verifications, fully demonstrating the reliability of the natural continuous template method in identifying structurally similar natural reaction modules. This design strategy is applicable not only to different types of carbon skeletons (such as aliphatic 2,4-DHB and 1,3-propanediol, and aromatic compounds like pHS), but also to precise functional group modifications (such as hydroxyl, carboxyl, and olefin groups). More importantly, by bypassing traditional ATP-dependent steps (such as the 1,3-PDO pathway), this embodiment demonstrates that this design strategy has significant potential for energy utilization optimization, indicating a direction for designing more efficient industrial production pathways.

[0065] In summary, this invention, based on the transformation of templates with continuous groups in natural enzymatic pathways, designs non-natural synthetic pathways, which has the following practical significance: 1. Biological feasibility: This strategy can significantly improve the biological feasibility of non-natural pathways because they are highly similar to natural pathways, effectively reducing the risk of failure in non-natural enzyme discovery and increasing the success rate of pathway construction.

[0066] 2. High efficiency and controllability: By precisely selecting enzymes and controlling the sequence of group transformations in the pathway, it helps to improve substrate selectivity and transformation efficiency, reduce by-product generation, and overcome kinetic and thermodynamic limitations, thereby enhancing the efficiency and controllability of the pathway.

[0067] 3. Industrial application potential: This templated design method has good versatility and scalability, and is applicable to a variety of hosts and target products. In particular, it shows significant industrial application potential in the biomanufacturing of high-value-added compounds without clear biosynthetic pathways.

[0068] 4. Promote engineering development: This strategy can serve as a standard paradigm for synthetic biology and metabolic engineering methodologies, helping to build a closed-loop system of "design-build-test-learn," improve engineering efficiency, and accelerate knowledge accumulation and technology iteration.

[0069] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 3 As shown, the computer device includes a processor, memory, and a network interface connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database stores static and dynamic information data. The network interface communicates with external terminals via a network connection. When the computer program is executed by the processor, it implements the steps in the above method embodiments.

[0070] Those skilled in the art will understand that Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present invention and does not constitute a limitation on the computer device to which the present invention is applied. A specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0071] In addition, the present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0072] In addition, 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 in the above method embodiments.

[0073] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the methods described above. Any references to memory, storage, databases, or other media used in the embodiments provided by this invention can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical storage, etc. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0074] This invention is not limited to the structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this invention is limited only by the appended claims.

Claims

1. A method for generating non-natural pathways based on continuous biological templates, characterized in that, include: Obtain enzyme-catalyzed reaction datasets from a pre-defined enzyme-catalyzed reaction database and preprocess the enzyme-catalyzed reaction datasets; A directed reaction graph was constructed based on the preprocessed enzyme-catalyzed reaction dataset. The directed reaction graph was used to search for similar compounds of the target compound. Starting from the similar compounds found, a depth-first search method was used to obtain continuous group transformation templates of similar compounds in the natural enzyme-catalyzed pathway. By applying continuous group transformation templates to target compounds in a linear molecular input canonical form and simulating the retrosynthetic process of the target compounds, a non-natural synthetic pathway based on continuous group transformation templates is generated.

2. The method for generating non-natural pathways based on continuous biological templates according to claim 1, characterized in that, The step of obtaining the enzyme-catalyzed reaction dataset from a preset enzyme-catalyzed reaction database and preprocessing the enzyme-catalyzed reaction dataset includes: The system retrieves raw enzyme-catalyzed reaction data from a pre-defined enzyme-catalyzed reaction database, updates the separators in the reaction formulas based on the reversibility status of the reactions in the raw enzyme-catalyzed reaction data, and filters the compound compartment information in the raw enzyme-catalyzed reaction data to obtain enzyme-catalyzed reaction data with reactivity significance. The reaction formulas are structured; at the same time, the linear input canonical structures of the metabolites are extracted based on the similarity of the compounds. The linear input canonical structures of the molecules are then subjected to structure cleaning and standardization processes to obtain reaction formulas with several linear input canonical structures of molecules. Reactions with completely identical molecular linear input canonical structures are merged to obtain several unique molecular linear input canonical reactions. Atom mapping is then performed on these unique molecular linear input canonical reactions to generate molecular linear input canonical reactions with mapping labels. The reaction template is extracted and corrected from the linear input canonical reaction with mapping labels, and a preprocessed enzymatic reaction dataset is generated.

3. The method for generating non-natural pathways based on continuous biological templates according to claim 2, characterized in that, The separator in the reaction formula is updated based on the reversibility status of the reaction in the original enzyme-catalyzed reaction data. Filtering the compound compartmentalization information from the raw enzymatic reaction data yields enzymatic reaction data with reactivity significance, including: Obtain the reversibility state of the enzyme-catalyzed reaction from the raw enzyme-catalyzed reaction data, and based on the reversibility state, replace the separators in the reaction formula using a custom script; Analyze the compound lists on both sides of the reaction equation, iterate through each compound in the compound list to see if it appears in the preset auxiliary molecule list, and remove the compounds that appear in the auxiliary molecule list; Compounds lacking linear molecular input canonical representation in the raw enzymatic reaction data are filtered out, and compounds with identical exchange reactions on both sides are also filtered out to obtain enzymatic reaction data with structural reactivity significance.

4. The method for generating non-natural pathways based on continuous biological templates according to claim 3, characterized in that, The reaction formulas are structured; simultaneously, based on compound similarity, the linear input canonical structures of metabolites are extracted, and the linear input canonical structures are sequentially cleaned and standardized to obtain several reaction formulas with linear input canonical structures, including: Based on the reaction formulas in the enzyme-catalyzed reaction data with reactivity significance, reaction formulas with unequal stoichiometry are split into equal-quantity forms according to the stoichiometric coefficients. At the same time, based on the number of compounds on the left and right sides of the reaction formula, compounds with multiple-to-multiple reactions are filtered out. The main substrate and product pairs of each reaction are determined by compound similarity, and the molecular linear input canonical structure of the metabolites is extracted based on the main substrate and product pairs to obtain several effective compound terms; The effective compound terms are sequentially subjected to structure cleaning and standardization to obtain several reaction formulas with linear input canonical structures of molecules.

5. The method for generating non-natural pathways based on continuous biological templates according to claim 4, characterized in that, The process of merging reactions with completely identical molecular linear input canonical structures to obtain several unique molecular linear input canonical reactions, and then performing atom mapping on these unique molecular linear input canonical reactions to generate molecular linear input canonical reactions with mapping labels, includes: Reactions with completely identical molecular linear input canonical structures are merged, and the annotation information of the molecular linear input canonical structures is merged into one information record, resulting in several unique molecular linear input canonical reactions. Construct a neural network model to format the linear input of a unique molecule into a standard input string using the neural network model; The system calls a preset function to perform atomic mapping on the standard input string, and outputs the probability distribution of the atomic mapping path through the encoding and decoding process of the neural network model; the atomic mapping path with the highest probability distribution is selected to obtain the molecular linear input canonical reaction with mapping label.

6. The method for generating non-natural pathways based on continuous biological templates according to claim 5, characterized in that, The extraction and modification of reaction templates from linear input canonical reactions with mapping labels includes: The linear input canonical reaction with mapping labels is used as input, and a predefined reaction class is called to parse the linear input canonical reaction with mapping labels. The template extraction interface is called to extract changes in reaction centers, bond breaking and formation, and stereo tags from the parsed reaction, and generate molecular linear input canonical reaction substructure template strings; The linear input canonical reactive substructure template string of molecules is filtered by neighborhood to obtain several reaction templates; the rationality of the reaction templates is evaluated, and unreasonable templates are screened out and corrected.

7. The method for generating non-natural pathways based on continuous biological templates according to claim 6, characterized in that, The process of evaluating the rationality of the reaction templates and filtering out unreasonable templates for correction includes: If the original substrate can be successfully reconstructed by applying the reaction template in reverse to the product molecules of the original reaction, the current reaction template is considered reasonable; otherwise, it is considered unreasonable, and the unreasonable reaction template is corrected.

8. The method for generating non-natural pathways based on continuous biological templates according to claim 1, characterized in that, The directed reaction graph is constructed based on the preprocessed enzyme-catalyzed reaction dataset. This graph is then used to search for similar compounds to the target compound. Starting with the found similar compounds, a depth-first search method is used to obtain continuous group transformation templates for similar compounds in natural enzymatic pathways, including: A directed reaction graph was constructed from the preprocessed enzyme-catalyzed reaction dataset. In the directed reaction graph, the linear input norm of molecules is used as nodes, the reaction template is used as directed edges, and the connectivity between nodes represents the actual enzyme-catalyzed reaction. Based on the directed reaction graph, the first round of similar compound search was performed on the target compound. Starting with the similar compounds obtained in the first round of search, and following the directed reaction graph, combined with a depth-first search method, we obtain continuous group transformation templates of similar compounds in natural enzymatic pathways.

9. The method for generating non-natural pathways based on continuous biological templates according to claim 8, characterized in that, The process of using the similar compounds obtained in the first round of search as a starting point, following the directed reaction graph, and combining a depth-first search approach to obtain continuous group transformation templates for similar compounds in natural enzymatic pathways includes: Using the similar compounds obtained in the first round of search as the starting node, initialize the stack and mark the starting node as visited; Remove the current node from the stack and check all adjacent nodes connected to the current node through the reaction template. If the adjacent node has not been visited and is a compound in the natural enzymatic pathway, mark the adjacent node as visited and push it onto the stack. At the same time, record the reaction template connected to the adjacent node as the group transformation template. The search process is executed recursively until the stack is empty or the preset search depth is reached. All paths are explored through a backtracking mechanism, and the final collected continuous reaction template sequence is used as a continuous group transformation template.

10. A non-natural pathway generation system based on a continuous biological template, characterized in that, The system includes: The dataset processing module is used to obtain enzyme-catalyzed reaction datasets from a preset enzyme-catalyzed reaction database and to preprocess the enzyme-catalyzed reaction datasets. The group transformation template acquisition module is used to construct a directed reaction graph based on the preprocessed enzyme-catalyzed reaction dataset, and use the directed reaction graph to search for similar compounds of the target compound. Starting from the searched similar compounds, the module combines a depth-first search method to obtain continuous group transformation templates of similar compounds in the natural enzyme-catalyzed pathway. The pathway generation module applies the continuous group transformation template to the target compound in a linear molecular input specification and simulates the retrosynthetic process of the target compound to generate a non-natural synthetic pathway based on the continuous group transformation template.

Images