Microbial manufacturing yield improvement method based on gene dynamic down-regulation algorithm

Abstract

The application discloses a kind of microorganism manufacturing yield increasing methods based on gene dynamic down-regulation algorithm, it is related to the cross technical field of synthetic biology, metabolic engineering and computational biology, for rationally designing high-yield, robust microbial cell factory.The core of the method is to combine the static gene knockout target prediction, the data information collection of transcriptional regulatory element driven by large language model and the global metabolic correlation analysis based on graph theory, to realize the whole process rationalization design from "who to regulate" to "what to use", finally output the strain modification scheme that gene knockout (knockout) is converted into gene down-regulation (knockdown), to realize the efficient coupling of cell growth and product synthesis.

Description

Technical Field

[0001] This invention relates to the interdisciplinary fields of synthetic biology, metabolic engineering, and computational biology, and specifically to a method for increasing the yield of microbial production based on a dynamic gene downregulation algorithm. Background Technology

[0002] In the process of heterologous synthesis of high-value-added compounds by microorganisms, there is an imbalance in the allocation of metabolic resources between cell growth metabolism and product synthesis, which restricts the production efficiency. To address this technical bottleneck, various metabolic engineering strategies have been developed and utilized. For example: (i) Computer-aided metabolic engineering strategies: With the development of genome-scale metabolic network models and various metabolic flow simulation and optimization algorithms, it is possible to perform global and quantitative simulation and analysis of complex cellular metabolic networks, predict genetic modification targets that can enhance the synthesis of target products, and provide key guidance for the rational design of yield-enhancing strains. Represented by gene knockout algorithms (such as OptKnock), this algorithm couples growth and production through a two-layer optimization framework, systematically outputting gene knockout schemes to improve the allocation of metabolic flow between cell growth and production. However, this type of strategy is essentially a static and rigid "all or nothing" modification. The operation of completely eliminating the activity of the target gene irreversibly destroys the robustness of the metabolic network. The modified strains lack the dynamic response ability to self-regulate according to changes in the intracellular and extracellular environment, which easily leads to decreased cell viability and unstable yield during scale-up production. (ii) Gene downregulation strategies: To address the metabolic imbalances that may result from direct gene knockout, gene downregulation strategies regulate metabolic flux by partially reducing rather than completely eliminating gene expression levels. Common gene downregulation techniques include antisense RNA technology, RNA interference technology, CRISPRi technology, and transcription factor repression technology. These techniques can also be modularly integrated for more precise regulation. A common strategy is based on transcription factor responses to changes in the concentration of specific intracellular metabolites to regulate gene expression, exhibiting unique advantages due to its good biocompatibility and feedback regulation potential. Currently, there is considerable research on this type of regulatory system, but the implementation of these strategies mainly relies on researchers' experience and trial-and-error screening, lacking a systematic and standardized rational screening method.

[0003] While existing gene knockout algorithms for microbial yield enhancement (such as OptKnock) can improve the yield of target products, their strain modification strategies have limitations such as rigidity, irreversibility, lack of dynamic response, and susceptibility to metabolic network imbalance, which limit their potential for industrial scale-up. The design of regulatory systems relies on human experience and trial-and-error screening, lacking rational screening methods based on the global metabolic network, resulting in a lack of systematicness and predictability in the selection of regulatory elements and signaling molecules.

[0004] Therefore, there is currently a lack of fully automated algorithm tools that integrate "target prediction - signal screening - component matching", making it impossible to achieve rational design from computational prediction to gene regulatory circuit design. Summary of the Invention

[0005] In view of this, the present invention provides a method for improving microbial production based on a gene dynamic downregulation algorithm. This method is a systematic and automated transcription factor-mediated gene dynamic downregulation algorithm (TF-based OptDown) for improving microbial production. It overcomes the shortcomings of static knockout strategies and realizes dynamic control design of the entire process from "target prediction" to "signal screening" and then to "element matching". This improves the product synthesis capacity and robustness of microbial cell factories and ultimately constructs engineered strains that can adaptively regulate and efficiently produce target products.

[0006] To achieve the above objectives, this invention provides a method for increasing microbial production based on a dynamic gene downregulation algorithm, the technical solution of which includes the following steps: Step 1: Obtain a customized metabolic network model for the target strain and construct a dataset of transcriptional regulatory elements.

[0007] Step 2: Using a genome-scale customized metabolic network model as input, call the OptKnock algorithm to output gene knockout targets that theoretically improve the throughput of the target product. These are the candidate target gene schemes to be downregulated, and the global throughput distribution of the metabolic network of each scheme is recorded simultaneously.

[0008] Step 3: Convert the customized metabolic network model into a graph theory model. For each candidate gene scheme to be downregulated, use the global flux distribution as a constraint to generate a phenotype-dependent metabolic network topology. Finally, screen metabolite signaling molecules to select the optimal metabolite signaling molecules that are closely coupled to growth and have the least interference with product pathways.

[0009] Step 4: Based on the optimal metabolite signaling molecules obtained from the screening, match the corresponding transcription factor-promoter pairs in the constructed transcription regulatory element database, and design gene regulatory circuits.

[0010] Step 5: Output the target gene scheme to be downregulated, its global metabolic flux distribution, topology diagram, and the signaling molecule-transcription factor scheme for constructing the regulatory system.

[0011] Further, in step one, a customized metabolic network model for the target strain is obtained, specifically in the following manner: an initial metabolic network model is downloaded from a public database or obtained from other sources, and a customized metabolic network model for simulating the synthesis of the target compound is constructed by completing the boundary conditions; wherein the completed boundary conditions include the heterologous synthesis pathway of the target product, modification of substrate uptake rate, and modification of product secretion rate.

[0012] Furthermore, in step one, a transcriptional regulatory element dataset is constructed, specifically in the following manner: using large language modeling tools to extract the "compound-transcription factor-promoter" triplet information from literature and databases, and after sorting and summarizing, a transcriptional regulatory element dataset is obtained, which is used in subsequent transcription factor screening and regulatory strategy design.

[0013] Furthermore, in step two, a genome-scale customized metabolic network model is used as input, and the OptKnock algorithm is called to output gene knockout targets that theoretically increase the throughput of the target product: specifically, the inner layer optimization objective is set to maximize the biomass growth rate, and the outer layer optimization objective is set to maximize the target product synthesis rate.

[0014] Furthermore, in step three, the metabolic network model is converted into a graph theory model. Specifically, a directed graph model is constructed with metabolites as nodes and metabolic reactions as edges. During the model processing, circulating metabolites in the model are automatically removed.

[0015] Furthermore, in step three, metabolite signaling molecules are screened to identify the optimal metabolite signaling molecules that are closely coupled to growth and have minimal interference with product pathways. This is done using the following method: By calculating the topological distance X from each compound in the transcriptional regulatory element dataset to the biomass synthesis node Biomass and the topological distance Y from the target product node Product, the correlation ratio function F=X / Y is defined. Based on minimizing F, the optimal metabolite signaling molecules that are closely coupled to growth and have the least interference with the product pathway are screened out.

[0016] Furthermore, the following modules are set up to execute the microbial production yield enhancement method based on gene dynamic downregulation algorithm, including an input module, a core processing module, and an output module; the input module is used to execute step one; the core processing module is used to execute steps two, three, and four; and the output module is used to execute step five.

[0017] Beneficial effects: The present invention provides a method for increasing the yield of microbial production based on a dynamic gene downregulation algorithm, which has the following significant advantages: (i) Achieved a rational design closed loop for dynamic regulation: The algorithm overcomes the existing empirical mode of dynamic regulation design that relies on trial and error. By integrating computational biology and synthetic biology methods, it provides a rational design solution for the entire process from target gene prediction and signal molecule screening to regulatory element matching. (ii) Introducing a global metabolic correlation screening criterion: The topological distance calculation in graph theory is innovatively applied to metabolic network analysis, and the ratio function F=X / Y is defined. This provides an objective and quantitative global indicator for screening ideal signaling molecules that are highly coupled with cell growth and have minimal interference with product pathways, thus avoiding the limitations of local pathway experience. (iii) Improve strain performance: Through preliminary verification of the algorithm output scheme, the results show that the constructed engineered strain has achieved preliminary yield improvement effect, verifying the feasibility and optimization potential of the yield improvement method. Attached Figure Description

[0018] Figure 1 A flowchart of a microbial production enhancement method based on a dynamic gene downregulation algorithm provided by this invention. Figure 2 The basic flowchart of the core processing module of the microbial manufacturing yield enhancement method based on the dynamic gene downregulation algorithm provided by this invention is shown below. Figure 3 The graph shows the batch fermentation growth and production of gene knockout strains (EK1-EK6). Note: *Significant difference (p<0.05) is compared with the OA yield of strain OA07. Figure 4 A schematic diagram of the construction scheme for strain Y1; Figure 5 The graph shows the fermentation yield results of the strain. Detailed Implementation

[0019] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0020] This invention provides a transcription factor-mediated dynamic gene downregulation algorithm (TF-based OptDown) based on a genome-scale metabolic network model for the rational design of high-yield and robust microbial cell factories. The core of this method lies in combining static gene knockout target prediction, large language model-driven collection of transcriptional regulatory element data, and graph theory-based global metabolic correlation analysis. This enables a rational design process from "who to regulate" to "what to regulate," ultimately outputting a strain modification scheme that transforms gene knockout into gene downregulation, achieving efficient coupling of cell growth and product synthesis.

[0021] The method includes the following steps: Step 1: Obtain a customized metabolic network model for the target strain and construct a dataset of transcriptional regulatory elements; Step 2: Using a genome-scale customized metabolic network model as input, call the OptKnock algorithm to output gene knockout target points that theoretically improve the throughput of the target product, which are the candidate target gene schemes to be downregulated, and simultaneously record the global throughput distribution of the metabolic network of each scheme. Step 3: Convert the metabolic network model into a graph theory model. For the global flux distribution corresponding to each candidate target gene scheme to be downregulated, use it as a constraint to generate a phenotype-dependent metabolic network topology graph. Finally, screen metabolite signal molecules to select the optimal metabolite signal molecules that are closely coupled to growth and have the least interference with product pathways. Step 4: Based on the information of the signal molecules obtained from the screening, match the corresponding transcription factor-promoter pairs in the constructed transcription regulatory element database, and design gene regulatory circuits; Step 5: Output the target gene scheme to be downregulated, its global metabolic flux distribution, topology diagram, and the signaling molecule-transcription factor scheme for constructing the regulatory system.

[0022] This invention comprises the following modules for executing a microbial production yield enhancement method based on a dynamic gene downregulation algorithm: an input module, a core processing module, and an output module. The input module executes step one; the core processing module executes steps two, three, and four; and the output module executes step five. The overall development process of this method is as follows: Figure 1 As shown, it mainly includes the following parts: (a) TF-based OptDown algorithm input module: Obtain two types of core data required by the algorithm: (i) Customized target strain genome-scale metabolic network model: Initial metabolic network models can be downloaded from public databases or obtained from other research teams (where initial metabolic network models corresponding to different target strains are used). By completing the heterologous synthesis pathway of the target product, modifying the substrate uptake rate, modifying the product secretion rate, and other boundary conditions, a customized metabolic network model for simulating the synthesis of the target compound is constructed; (ii) Locally constructed transcriptional regulatory element dataset: Taking advantage of the advantages of large language model tools in text information extraction, the "compound-transcription factor-promoter" triple information is extracted from literature and databases. Each triple represents a transcriptional regulatory element. After sorting and summarizing, a transcriptional regulatory element dataset is obtained, which is used in subsequent transcription factor screening and regulatory strategy design.

[0023] (b) TF-based OptDown algorithm core processing module: The main task of the core processing module is to organically integrate the two parts of target gene identification and dynamic regulatory element screening, and output a strain modification plan that can be directly implemented. Its basic process is as follows: Figure 2 As shown, it is mainly divided into three parts: (i) Target gene identification based on OptKnock: Taking a customized genome-scale metabolic network model as input, the inner optimization objective is set to maximize the biomass growth rate, and the outer optimization objective is set to maximize the target product synthesis rate. The OptKnock algorithm is called to output gene knockout target points that can theoretically increase the target product throughput, which are the candidate target gene schemes to be downregulated. The global flux distribution of the metabolic network of each scheme is recorded simultaneously; (ii) Rational screening of signal molecules based on graph theory topological distance: First, the customized metabolic network is converted into a graph theory model. With metabolites as nodes and metabolic reactions as edges, a directed graph model is constructed. During the model processing, circulating metabolites in the customized metabolic network model are automatically removed. Circulating metabolites are general intermediates or energy carriers in metabolic reactions. Their presence will cause the calculated topological distance to be distorted; In addition, for the global flux distribution corresponding to each target scheme output by OptKnock, it is used as a constraint condition to generate a phenotype-dependent metabolic network topology. The graph is then used to screen metabolite signal molecules. By calculating the topological distance X to each biomass node and the topological distance Y to the target product node (the biomass node is the node in the graph theory model corresponding to the pseudo-metabolite biomass in the model, and the target product node is the node in the graph theory model corresponding to the target product in the model), the correlation ratio function F=X / Y is defined. Based on minimizing F, the optimal metabolite signal molecules that are closely coupled to growth and have the least interference with the product pathway are screened out. The biomass node (Biomass) is the node in the graph theory model corresponding to the pseudo-metabolite biomass in the model, and the target product node is the node in the graph theory model corresponding to the target product in the model. (iii) Based on the optimal metabolite signal molecules obtained by screening, the transcription factor-promoter pair scheme of the corresponding compound is matched in the constructed compound-transcription factor-promoter database.

[0024] (c) TF-based OptDown output module: The output module is used to present the final design results of the algorithm to guide wet experimental verification. The output includes the target gene scheme to be downregulated, its global flux distribution, the directed graph model, and the signal molecule-transcription factor scheme for constructing the regulatory system.

[0025] (d) Engineered strain modification based on TF-based OptDown algorithm: Based on the strategy output by TF-based OptDown and the given topological distance and growth-related metabolite signaling molecules, transcriptional regulatory elements (“compound-transcription factor-promoter” triplet) are introduced into engineered bacteria. For example, an activating / repressing promoter is introduced upstream of the target gene. The gene expression level is dynamically regulated according to the change of metabolite signaling molecule concentration, and the metabolic flow to cell growth and production is efficiently allocated to improve the production capacity of the strain to synthesize the target compound.

[0026] Specific application examples: Using the oleanolic acid-producing Saccharomyces cerevisiae strain OA07 as the research object, the TF-based OptDown algorithm was applied to output the strain's metabolic engineering scheme. The specific steps are as follows: (a) Input the genome-scale metabolic network model Yeast8-OA, which integrates the heterologous synthesis pathway of oleanolic acid in Saccharomyces cerevisiae. The corresponding metabolic reactions, metabolites, and gene information were searched from databases such as KEGG (https: / / www.kegg.jp / ) and NCBI (https: / / www.ncbi.nlm.nih.gov / ), and the boundary conditions such as substrate uptake rate and product secretion rate were modified according to the production situation.

[0027] (b) Using the Yeast8-OA model as input, the OptKnock function was called in the MATLAB environment to set the function parameters. Under the premise of allowing only single-reaction knockout, a series of target gene schemes that could theoretically increase oleanolic acid production were iteratively calculated and output, while simultaneously recording the global flux distribution corresponding to each scheme. For this application case, 12 theoretical single-reaction knockout schemes were finally obtained. These genes are distributed along the bypass pathway of the oleanolic acid synthesis pathway, involving the folic acid cycle (…). fol1 , fol2 , fol3 Aromatic amino acid synthesis ( abz1 , abz2 , pha2 ); Polyamine synthesis-related genes ( spe2 , spe4 ); Glycine metabolism-related genes ( agx1 ); Pantothenic acid synthesis-related genes ( ecm31 , pan6 ); Methionine recycling-related genes ( utr4 and adi1The knockout effect of these methods primarily affects the flux or enzyme activity at key nodes in the oleanolic acid synthesis pathway by reducing competition, relieving feedback inhibition, regulating cofactor balance (such as NADPH), or altering metabolite flow, ultimately potentially increasing oleanolic acid yield. Of the 12 output schemes mentioned above, six schemes (Δ...) fol1 Δ fol2 Δ fol3 Δ abz1 Δ abz2 Δ pha2 The research team has previously constructed direct gene knockout strains (EK1-EK6) in their existing research, and experimentally verified strain EK3 (Δ fol3 ), EK5 (Δ abz2 ), EK6 (Δ pha2 The yield-increasing performance is relatively significant. [1] ,like Figure 3 As shown, therefore, in the application examples of this invention, the preliminary application and verification of the TF-based OptDown algorithm mainly revolves around fol3 , abz2 , pha2 These three target genes were targeted.

[0028] (c) The Yeast8-OA was topologically structured and transformed into a graph theory model: (i) Removal of circulating metabolites: Some small currency metabolites were directly removed from the metabolic network. In addition, most circulating metabolites were paired into currency metabolite pairs (CMPs) according to their functional group transfer effects. Each type of CMP was prioritized according to its importance in the transfer functional group. When multiple CMPs appeared in the reaction at the same time, they were removed in order of priority to ensure that at least one real metabolite pair remained connected, preventing pathway breakage or accidental deletion. According to the literature review, some research groups have already used Python to form a script based on the above rules to realize the automatic removal of circulating metabolites in the model; (ii) Flux orientation constraint: Combined with the global metabolic flow under different schemes in the previous step, the reactions with a flux value of 0 in the metabolic network were removed. Then, a unique direction was assigned to the metabolic reaction according to the positive or negative flux to ensure that the direction of the metabolic reaction is consistent with the actual metabolic flow direction, and the phenotypic dependent metabolic network topology under multiple schemes was obtained.

[0029] (d) Topological distance calculation and signal molecule screening A constructed dataset of transcriptional regulatory elements is introduced. Each compound in the dataset is used as a starting node, and the target nodes are the biomass node (representing the product node of the pseudo-reaction of biomass synthesis) and the product node (the product node of the oleanolic acid secretion reaction). The topological distances X and Y, and the ratio F between the two distances (F=X / Y), are calculated. This invention uses Dijkstra's algorithm to calculate the topological distances from the compounds in the dataset to the two target nodes, specifically by calling the `shortest_path_length` function in the Python NetworkX library. After calculation, the compounds are sorted according to the F value, and usable compounds are selected as regulatory signaling molecules for strain modification. fol3 ↓、 abz2 ↓ and pha2 The calculation results for the three schemes are shown in Tables 1, 2, and 3. Based on the calculation results, pyruvate and its related transcription factors, as well as succinate and its related transcription factors, are considered to have certain application potential and can be further studied as candidate transcription factor regulatory elements. Through comprehensive analysis of the intracellular correlation between the two compounds and the target compound oleanolic acid, as well as their own metabolic characteristics, pyruvate was ultimately selected as a suitable regulatory signaling molecule.

[0030] Table 1 fol3 ↓ Calculation results of topological distance and ratio of the proposed schemes Table 1 Topological distance and ratio calculation results for the fol3 ↓ scheme

[0031] (Note: inf indicates that the path between nodes does not exist) Table 2 abz2 ↓ Calculation results of topological distance and ratio of the proposed schemes Table 2 Topological distance and ratio calculation results for the abz2 ↓ scheme

[0032] Table 2 (continued)

[0033] (Note: inf indicates that the path between nodes does not exist) Table 3 pha2 ↓ Calculation results of topological distance and ratio of the proposed schemes Table 3 Topological distance and ratio calculation results for thepha2↓ scheme

[0034] Table 3 (continued)

[0035] (Note: inf indicates that the path between nodes does not exist) (d) High-yield microbial engineered strains based on TF-based OptDown algorithm: Based on the calculation results, the algorithm outputs the downregulation of intracellular pyruvate as the signal molecule. fol3 , abz2 and pha2 The gene-based approach represents a subset of potential experimental approaches. For these three output approaches, wet experimental design and validation were conducted: the pyruvate-responsive transcription factor PdhR gene was integrated into the OA07 genome to construct a chassis strain OA07 (…). pdhR The inducible promoter pCYC1min_pdhOp, regulated by the PdhR protein, was inserted into the target gene in a reversed form (the reversed insertion promoter was named R_pCYC1min). fol3 , abz2 and pha2 Antisense RNA transcription units were constructed from the 3' untranslated region, resulting in strains Y1, Y2, and Y3. Preliminary fermentation validation showed that strain Y1 exhibited the best production performance among the three strains. Results indicated that strain Y1, after 120 h of fermentation, produced oleanolic acid at a yield of 124.73 ± 10.31 mg / L, significantly higher than that of the chassis strain OA07. pdhR The yield increased by 22.15%, and the unit cell yield increased by 37.65%, verifying the effectiveness of the algorithm's output scheme. The construction scheme diagram for strain Y1 and the fermentation results diagram are shown below. Figure 4 and Figure 5 .

[0036] In summary, the above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for increasing microbial production yield based on a dynamic gene downregulation algorithm, characterized in that, include: Step 1: Obtain a customized metabolic network model for the target strain and construct a dataset of transcriptional regulatory elements; Step 2: Using a genome-scale customized metabolic network model as input, the OptKnock algorithm is called to output gene knockout targets that theoretically improve the throughput of the target product. These are the candidate target gene schemes to be downregulated, and the global throughput distribution of the metabolic network of each scheme is recorded simultaneously. Step 3: Convert the customized metabolic network model into a graph theory model. For each candidate gene scheme to be downregulated, the global flux distribution is used as a constraint to generate a phenotype-dependent metabolic network topology. Finally, metabolite signal molecules are screened to select the optimal metabolite signal molecules that are closely coupled to growth and have the least interference with product pathways. Step 4: Based on the optimal metabolite signaling molecule information obtained from the screening, match the corresponding transcription factor-promoter pairs in the constructed transcription regulatory element database, and design gene regulatory circuits; Step 5: Output the target gene scheme to be downregulated, its global metabolic flux distribution, topology diagram, and the signaling molecule-transcription factor scheme for constructing the regulatory system.

2. The method for increasing microbial production yield based on a dynamic gene downregulation algorithm as described in claim 1, characterized in that, In step one, a customized metabolic network genome-scale metabolic model for the target strain is obtained, specifically in the following manner: an initial metabolic network model is downloaded from a public database or obtained from other sources, and a customized metabolic network model for simulating the synthesis of the target compound is constructed by completing the boundary conditions; wherein the completed boundary conditions include the heterologous synthesis pathway of the target product, modification of substrate uptake rate, and modification of product secretion rate.

3. The method for increasing microbial production yield based on a dynamic gene downregulation algorithm as described in claim 1, characterized in that, In step one, the transcriptional regulatory element dataset is constructed in the following manner: We used a large language model tool to extract "compound-transcription factor-promoter" triplet information from literature and databases, and after sorting and summarizing it, we obtained a dataset of transcriptional regulatory elements, which can be used for subsequent transcription factor screening and regulatory strategy design.

4. The method for increasing microbial production yield based on a dynamic gene downregulation algorithm as described in claim 1, characterized in that, In step two, a genome-scale customized metabolic network model is used as input, and the OptKnock algorithm is called to output gene knockout targets that theoretically increase the throughput of the target product: specifically, the inner optimization objective is set to maximize the biomass growth rate, and the outer optimization objective is set to maximize the target product synthesis rate.

5. The method for increasing microbial production yield based on a dynamic gene downregulation algorithm as described in claim 1, characterized in that, In step three, the metabolic network model is converted into a graph theory model. Specifically, a directed graph model is constructed with metabolites as nodes and metabolic reactions as edges. During the model processing, circulating metabolites in the model are automatically removed.

6. The method for increasing microbial production yield based on a dynamic gene downregulation algorithm as described in claim 5, characterized in that, In step three, metabolite signaling molecules are screened to identify the optimal metabolite signaling molecules that are closely coupled to growth and have minimal interference with product pathways. This is done using the following method: By calculating the topological distance X from each compound in the transcriptional regulatory element dataset to the biomass synthesis node Biomass and the topological distance Y from the target product node Product, the correlation ratio function F=X / Y is defined. Based on minimizing F, the optimal metabolite signaling molecules that are closely coupled to growth and have the least interference with the product pathway are screened out.

7. The method for increasing microbial production yield based on a dynamic gene downregulation algorithm as described in claim 1, characterized in that, The following modules are configured to execute the microbial manufacturing yield enhancement method based on the gene dynamic downregulation algorithm, including an input module, a core processing module, and an output module; The input module is used to perform step one; The core processing module is used to execute steps two, three, and four. The output module is used to perform step five.

Images