Method for designing product and synthesizing optimal strain through fusion of enzyme constraint model and evolutionary algorithm

Abstract

The invention discloses a method for designing a product to synthesize an optimal strain by fusing an enzyme constraint model with an evolutionary algorithm, which comprises the following steps of: S1, solving an enzyme constraint model by taking a maximized strain specific growth rate as a target function to obtain a simulated wild type strain metabolic flux; s2, performing dimension reduction analysis and labeling on genes directly related to enzyme synthesis in the model by solving a series of flux equilibrium analysis problems of a fixed biomass synthesis rate; s3, predicting the yield of single target point editing; s4, performing combined target spot search on the dimensionality-reduced target spots by adopting a genetic algorithm, and adjusting related parameters to obtain an optimal gene editing combination strategy; and S5, repeating the experiment to verify the stability of the algorithm. According to the method, a non-visual gene editing combination strategy is explored by utilizing a genetic algorithm and an enzyme constraint model, a new method is provided for constructing a microbial cell factory for efficiently synthesizing chemicals, and the potential of applying a heuristic algorithm to strain optimization design is shown.

Description

Technical Field

[0001] The present invention relates to the field of combining biology with computers, and in particular to a method for optimizing strains for product synthesis by integrating an enzyme constraint model with an evolutionary algorithm. Background Art

[0002] The rapid development of computer and biotechnology has greatly promoted the application of microorganisms in many fields, including the production of biofuels, antibiotics, food and its additives, biopesticides and various processing enzymes. Biological production pathways based on microbial fermentation technology are gradually replacing traditional chemical synthesis processes, which also marks the transformation of production methods towards green and sustainable development. However, the key to winning the fierce competition with traditional petroleum-based production methods is to develop high-performance and superior microbial cell factories. For industrial production, it is crucial to develop economical and efficient industrial strains. Under the evolutionary pressure of natural selection, the metabolic phenotype of microorganisms in natural habitats often manifests itself as maximizing cell growth rather than maximizing the synthesis of a certain product. This requires metabolic modification of the strain to obtain the desired cell phenotype. Therefore, metabolic engineering of industrial strains, that is, optimizing the yield and production efficiency of target products through gene editing, has become an important research direction in the field of bioengineering. The main challenges faced by this process include the complexity of the microbial metabolic network itself and the incomplete understanding of the regulatory mechanism of cellular metabolism.

[0003] At present, most successful cases of metabolic engineering rely on intuitive and qualitative design principles, lacking rational and systematic design and analysis tools. In addition, most strain design algorithms are used to predict single targets, lacking the prediction and analysis of multi-target combinations. The concept of constraint-based modeling has promoted the development of reliable quantitative metabolic models and provided a new perspective and analysis platform for metabolic engineering. Among them, the genome-scale metabolic network model (GSMM) as a comprehensive bioinformatics tool can describe all known metabolic pathways in cells in detail and predict how cells adjust their metabolic processes under different environments. However, it is difficult for GSMM to accurately simulate metabolic conversions that occur at high growth rates, such as overflow metabolism of bacteria, Crabtree effect of Saccharomyces cerevisiae, and Warburg effect of human cancer cells. The prediction ability of GSMM can be enhanced by integrating enzyme concentration and enzyme kinetic parameters into the stoichiometric model to construct an enzyme-constrained model. The proteomics data were integrated using the GECKO (GEM with Enzymatic Constraints using Kinetic and Omics data) method, which adds an upper bound to each metabolic flux by including the enzyme as part of the reaction - the enzyme abundance and enzyme turnover number (k cat ). While significantly reducing the flux variability of traditional GSMM, it also reveals the usage and distribution of enzymes in different reactions and different metabolic pathways.

[0004] The number of metabolites and reactions contained in GSMM and enzyme constraint models is huge, which leads to a large dimension of the solution space defined by them. A commonly used method to determine metabolic phenotypes is flux balance analysis (FBA). In addition, in order to simulate the metabolic flux distribution of mutant strains produced by genetic intervention, the Minimization of Metabolic Adjustment (MOMA) method has also been proposed. The former mainly relies on linear programming techniques for solution, and commonly used methods include the simplex method. FBA is based on two basic assumptions: the steady-state assumption and the mass conservation assumption. The steady-state assumption assumes that the growth of the strain has reached a stable state; the mass conservation assumption states that in a stable metabolic network, the sum of the production and consumption of each metabolite should be zero in terms of stoichiometry. The latter provides an approximate solution to the suboptimal state of mutant strains through mathematical methods: the distance minimization principle in the flux space is adopted to transform the problem into a quadratic programming problem. Flux variability analysis (FVA) is mainly used to determine the maximum and minimum flux values ​​of each metabolic reaction under specific growth conditions and environments. Used to understand the robustness of metabolic networks and their adaptability to environmental changes. The forced target flux scanning algorithm (FluxScanning based on Enforced Objective Flux, FSEOF) adds additional constraints to the path leading to the synthesis of the target product during the flux balance analysis, forcibly increases the flux of the path step by step, and then scans all reaction fluxes. ecFSEOF has been improved on this basis and adapted to the enzyme constraint model. This algorithm is implemented by solving a series of flux balance analysis problems with fixed biomass synthesis rates, scoring all reactions and giving suggestions for three gene editing directions: overexpression, knockdown, and knockout for related genes.

[0005] Given the limitations of biometric technology and the complexity of microbial metabolic regulation, it is currently impossible to obtain complete metabolic flux phenotype data of strains under specific growth conditions. However, the optimization design of strains requires a reference yield of the target product, and the MOMA method also requires a set of reference flux vectors to be input for solution. Therefore, it is necessary to use the model to predict the metabolic phenotype of wild-type strains with high accuracy.

[0006] At present, a variety of strain design algorithms and software toolboxes for integrated model building and optimization algorithms have been developed. There are also cases of integrating advanced algorithms such as intelligent algorithms, reinforcement learning, and machine learning into strain design. However, the development of algorithms for enzyme constraint models is still relatively small, and accurately identifying metabolic engineering targets under time and resource constraints is still a major challenge. Summary of the invention

[0007] The purpose of the present invention is to combine genetic algorithms and enzyme constraint models to obtain combined gene targets for transforming microorganisms and the editing direction of the targets. This is achieved by simulating the metabolic phenotype of wild-type strains, dimensionality reduction analysis and annotation of genes in the model, and searching with genetic algorithms, so as to solve the problems raised in the above-mentioned background technology.

[0008] To achieve the above-mentioned object of the invention, the present invention provides a method for designing an optimal strain for product synthesis by integrating an enzyme constraint model with an evolutionary algorithm, comprising the following steps:

[0009] Step S1, solving the enzyme constraint model with maximizing the strain specific growth rate as the objective function to obtain the simulated metabolic flux of the wild-type strain;

[0010] Step S2, by solving a series of flux balance analysis problems with fixed biomass synthesis rates, dimensionality reduction analysis and annotation of genes directly related to enzyme synthesis in the model;

[0011] Step S3, predicting the yield of single-target editing, modifying the model based on the results of dimensionality reduction and gene annotation, solving the strain phenotype of single-target mutation and calculating the yield of the target product;

[0012] Step S4, using a genetic algorithm to search for combined targets for the reduced-dimensional targets, designing and adjusting relevant parameters, and obtaining the optimal gene editing combination strategy;

[0013] Step S5, repeat the experiment to verify the stability of the algorithm, and perform statistical analysis and biological interpretation on the obtained combination strategy.

[0014] Furthermore, before the analysis of the enzyme constraint model and the combined optimization of the genetic algorithm, a series of necessary adjustments and verifications were made to the model to ensure that it could accurately simulate the metabolic state of the strain under different environmental conditions. This included releasing the absorption pathways of oxygen and the main carbon and nitrogen sources, and setting the upper bound of all reactions with an upper limit of infinity to b u (b u ∈[100,2000]) to support subsequent flux scanning, flux variability analysis and metabolic adjustment minimization calculation. Considering that the absolute flux value of protein synthesis during strain growth is very small, the accuracy of the model is set to Ω(Ω∈[10 -8 ,10 -10 ]), increasing insight into protein synthesis.

[0015] Furthermore, the flux balance analysis method solves the enzyme constraint model and simulates the metabolic flux of the wild-type strain, including the following three methods:

[0016] Method 1: After the minimum production limit of the target product is given, a single-step flux balance analysis is performed with maximizing the specific growth rate of the strain as the objective function;

[0017] Method 2: Based on method 1, the specific growth rate is limited to no less than the maximum growth rate λ

[0018] (λ∈[0.5,0.9]), FBA is solved with minimizing glucose consumption as the objective function, and the glucose consumption rate is limited to not exceed α(α∈[1.1,1.5]) times its minimum consumption, and FBA is solved with minimizing the total enzyme consumption as the objective function;

[0019] Method 3: Perform FBA three times according to Method 2 without limiting the yield of the target product, and then limit the total enzyme consumption to no more than β (β∈[1.1,1.5]) times its minimum value, and perform FBA with maximizing product yield as the objective function.

[0020] Furthermore, the flux balance analysis used is expressed by the following formula:

[0021] max c T ·v

[0022]

[0023] Where c is a constant vector of 0 and 1, and v is the flux vector representing all reactions. The transpose of c and the multiplication of v define the optimization direction of a model, that is, maximizing the flux of a reaction. S is the stoichiometric matrix of the model, and Represent the minimum and maximum constraint boundaries of the reaction, respectively.

[0024] Furthermore, step 2 requires screening out all genes related to enzyme synthesis. The enzyme constraint model under the GECKO framework adds new rows representing enzymes and new columns representing their usage on the basis of GSMM, and directly uses enzymes as reaction components to expand the model. Each reaction related to enzyme synthesis has a corresponding gene, and dimensionality reduction analysis and annotation are performed on these genes.

[0025] First, the biomass synthesis flux value is set to the theoretical maximum value. The algorithm will explore 16 different biomass synthesis rate conditions that gradually decrease from this maximum value (from 100% of its maximum predicted value to 25%) in order to explore the changes in metabolic reactions under different biomass synthesis rate conditions. Then, FBA is performed with the yield of the target product as the objective function to obtain the flux values ​​of all reactions. The changes in these reactions are quantified through a two-step scoring process, and reactions that have undergone significant changes are considered targets for engineering adjustments. The total flux score (V score) is the average value of the ratio of the flux of the reaction under all mandatory target flux setting conditions to the flux under the maximum biomass synthesis rate conditions. The specific formula is:

[0026]

[0027] in is the flux of reaction i under the conditions of maximum biomass synthesis rate, while the numerator in the formula is the sum of the fluxes of the reaction under all simulation conditions, and n is the number of simulations.

[0028] In addition, any V greater than Α(Α∈[1000,5000]) score Values ​​are truncated to 1000 to avoid infinite values. Undetermined values ​​(0 divided by 0) are assigned 1. Based on these flux scores, gene scores are then calculated. As the average reaction flux score for each gene:

[0029]

[0030] In the formula is the score of gene target i, the numerator is the sum of the reaction scores of all reactions catalyzed by the gene product, and m is the total number of reactions catalyzed by the gene product.

[0031] according to Genes with values ​​higher than 1 are suggested as overexpression targets. Between k a to k b The target is between k and k a The target is the knockout target. a ∈[0.02,0.08], k b ∈[0.2,0.8].

[0032] Furthermore, a representation method for implementing gene editing in the model is designed, where:

[0033] In the enzyme constraint model, the knockout operation of a gene can be intuitively implemented by setting the upper and lower bounds of the reaction flux of the corresponding enzyme synthesis path to 0, indicating that the synthesis path of the enzyme is closed. For the overexpression and knockdown operations of genes, the adjustment direction of the reaction flux and the degree of intervention need to be set manually. Based on the predicted wild-type flux, a reference dosage is provided for each enzyme. The overexpression operation is designed to increase the lower bound of the flux of the relevant enzyme synthesis reaction to N times the reference value, and the knockdown operation adjusts the upper bound of the flux to M times the reference value. If the reference flux value of the target to be overexpressed is 0, its lower bound is set to v low (v low ∈[10 -9 ,4×10 -9 ]).

[0034] The metabolic adjustment minimization method to solve the metabolic flux of the mutant strain defines the distance as:

[0035]

[0036] The vector w represents the metabolic flux of the wild-type strain as a reference vector, and the vector x represents the metabolic flux distribution of the mutant strain.

[0037] Furthermore, the yield prediction of single-target editing was performed one by one, and the metabolic adjustment minimization method was used to solve the edited enzyme constraint model to obtain the metabolic phenotype of the mutant strain and calculate the yield of the target product.

[0038] Furthermore, in step S4, a genetic algorithm is used to search for combined targets for the targets after dimensionality reduction. In the initialization stage, the following population is generated:

[0039] pop={X1,X2,...,X NP},

[0040] X i =[x i1 ,x i2 ,...,x iL ],

[0041] NP is the population size, L represents the dimension of the individual, that is, the number of genes contained in the individual, where i∈[1,NP].

[0042] There are two encoding forms:

[0043] In the binary code, the number on each dimension of the individual package can only be "0" or "1". "0" means that the gene is not selected by this set of gene editing strategies, and "1" means that it is selected, and then the editing method of the gene is obtained based on the annotation results. At this time, the individual initialization method is as follows:

[0044]

[0045] oi is a random number between 0 and 1. The O selected by the judgment condition can control the number of times 1 appears in the sequence of the individual. Taking a smaller value of O∈[0.05,0.2] can ensure that only a small number of targets need to be intervened in the gene editing combination strategy represented by the individual, ensuring the feasibility of the actual experiment.

[0046] The symbol-encoded genetic algorithm relies only on the results of gene dimensionality reduction, and finds the combined target and the editing direction of the target in a larger search space. In the case of symbol encoding, the individual coding sequence contains four situations: 0, 1, 2, and 3, which respectively represent that the gene is not selected, needs to be overexpressed, knocked down, and knocked out. The individual initialization method is as follows:

[0047]

[0048] Among them, 0<O1≤0.04<O2≤0.08<O3≤1.2. Each gene has the same probability of being initialized to three editing directions, and has a greater probability of not needing editing.

[0049] In the selection and population update, a roulette selection strategy is adopted. After each individual is evaluated, it is sorted according to the size of its fitness value, and the probability of each individual being selected is calculated according to the following formula.

[0050]

[0051] The individual evaluation function f is solved with the help of the model. This formula can be used to calculate the cumulative probability of all individuals being selected, and then generate a random number between 0 and 1, and select the corresponding individual according to the cumulative probability interval that the random number is in. The selected individuals will be regarded as candidates for the crossover operation.

[0052] After one iteration of the algorithm, the original population and the newly generated individuals from crossover mutation are mixed and sorted, and a portion of the top individuals are selected according to the set population size to enter the next generation search cycle.

[0053] The crossover operation of this algorithm adopts single-point crossover with a crossover probability V c (V c ∈[0.1,0.9]) to determine whether two parent individuals crossover. The two parent individuals selected for crossover exchange the coding sequence after a certain gene locus, where the gene locus is randomly selected.

[0054] In the case of binary coding, if an individual is about to mutate, a gene locus in the individual is randomly selected to mutate: from "0" to "1" or from "1" to "0".

[0055] For the case of symbol coding, first determine whether the individual needs to mutate, and then determine the direction of mutation bit by bit: a random number is generated for each position of the individual, and mutation is performed according to the interval where the random number is located. If its size exceeds the mutation probability V r (V r ∈[0.01,0.5]), then the point does not mutate.

[0056]

[0057] Among them, 0<M1≤0.02<M2≤0.04<M3≤0.06. The flux variability analysis of the enzyme constraint model was performed. The flux variation value of each reaction was calculated according to the formula:

[0058] flux variability i =max flux i -min flux i ,

[0059] Among them, max flux i and minimum flux i It is obtained by changing the solution target through FBA, and limiting the specific growth rate to a variation range of 1% of the maximum specific growth rate during the solution process.

[0060] Compared with the prior art, the present system and method have the following advantages:

[0061] (1) The metabolic phenotype of the wild-type strain with the lowest target product yield was predicted by the enzyme constraint model and used as a reference for subsequent design, thus solving the problem of lack of experimental data.

[0062] (2) We performed dimensionality reduction analysis and annotation on genes directly related to enzyme synthesis in the enzyme-constrained metabolic network model, effectively reducing the scale of the search space and ensuring that research resources and time are focused on the most promising candidate targets.

[0063] (3) The specific implementation method of gene editing in the model was designed, and a genetic algorithm was used to search for gene editing combination strategies, balancing the algorithm's global search capability and convergence speed.

[0064] (4) The target product yield predicted by the searched combined target editing model is significantly higher than the predicted yield after single target editing.

[0065] (5) The genetic algorithm in symbolic coding form can simultaneously search for combination targets and their editing directions, which has a higher improvement effect on the strain in the model calculation simulation results.

[0066] (6) The use of genetic algorithms to discover non-intuitive gene editing strategies provides powerful algorithmic support for solving complex metabolic engineering optimization problems and demonstrates the great potential of applying heuristic algorithms to strain optimization design. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 Schematic diagram of the method for designing optimal strains for product synthesis by fusing the enzyme constraint model with the evolutionary algorithm.

[0068] Figure 2 Flowchart of the method for designing optimal strains for product synthesis by integrating enzyme constraint model with evolutionary algorithm.

[0069] Figure 3 This is a variability analysis diagram of the Saccharomyces cerevisiae enzyme constraint model.

[0070] Figure 4 It is a scatter plot of the target product yield after single target editing in the present invention.

[0071] Figure 5 The search results of the genetic algorithm in binary and symbolic encoding form are shown in Figure 2.

[0072] Figure 6 This is a convergence curve diagram of ten repeated experiments in the present invention.

[0073] Figure 7 Stacked bar chart showing statistics of gene editing methods in replicate experiments. DETAILED DESCRIPTION

[0074] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0075] Genetic Algorithms (GA) can achieve rapid search in high-dimensional space by simulating the mechanism of natural selection and inheritance in the process of biological evolution, and is very effective in solving this type of combinatorial optimization problem. Genetic algorithms simulate biological genetic processes through crossover, mutation and selection operators, and perform iterative searches in populations to find the optimal solution. The goal of the present invention is to use the powerful search ability of genetic algorithms and the accurate simulation ability of enzyme constraint models on microorganisms to solve the combinatorial optimization search problem of large-scale gene editing strategies, so as to shorten the time of strain optimization design and improve the accuracy of model prediction targets, and reduce the number of targets involved in gene editing strategies to facilitate experimental verification, ensure the feasibility of experimental verification and the economy of microbial cell factories. By combining heuristic algorithms and advanced enzyme constraint models, the present invention breaks through the limitations of traditional methods and mechanism knowledge, so that large-scale gene editing strategy search problems can be solved within an acceptable time frame, opening up new ways to improve the production efficiency and stability of industrial microorganisms.

[0076] In the present embodiment, the latest Saccharomyces cerevisiae enzyme constraint metabolic network model ecYeast8.3.4 is used, and the influence of additional constraints on the model is applied to the enzyme constraint model by FVA analysis. This model contains a total of 1148 genes and 8144 reactions. According to the gene-protein-reaction correspondence of the model, there are 968 genes corresponding to the reactions directly related to enzyme synthesis. Taking 2-phenylethanol as the target product, the metabolic phenotype prediction schemes of three wild-type strains are compared. Gene dimensionality reduction is performed by using the ecFSEOF algorithm, and the score value of each reaction is calculated, so as to label the related genes. Before starting the search of the combined gene target editing strategy, MOMA is used to predict the editing results of a single target. Then the genetic algorithm test is performed according to different coding forms. Finally, the results and iterative curves are analyzed in detail to obtain valuable data on algorithm optimization and model adjustment and non-intuitive gene editing strategies. Figure 1 It is a schematic diagram of the present invention.

[0077] like Figure 2 Shown is a flow chart of the method of the present invention, and the specific steps are as follows:

[0078] Step S1, solving the enzyme constraint model with maximizing the strain specific growth rate as the objective function to obtain the simulated metabolic flux of the wild-type strain.

[0079] By comparing the three simulation methods introduced above, it can be seen from the results that the accuracy of the model solution has a more significant impact on the number of zero values ​​in the prediction results, while the results of single-step solution and multi-step single-objective solution are not much different. In addition, even repeated experiments of the same scheme are not exactly the same, which will affect the final analysis of the experimental results. Although all schemes can effectively predict the metabolic phenotype of the strain, the schemes show obvious differences in metabolic flux distribution, energy efficiency, and target product yield. These differences not only reflect the impact of model settings, but also reveal the potential impact of different growth conditions and biological constraints on the metabolic regulation of strains.

[0080] The results of flux variability analysis are shown in Figure 3 The red and blue curves representing additional constraints are both above and to the right of the green curve representing no additional constraints, indicating that by imposing appropriate biological constraints on the model, the variability of flux can be effectively reduced, the feasible solution space of the model can be narrowed, and the accuracy and reliability of the prediction can be improved.

[0081] Step S2, by solving a series of flux balance analysis problems with fixed biomass synthesis rates, dimensionality reduction analysis and annotation of genes directly related to enzyme synthesis in the model are performed.

[0082] The improved forced target flux scanning algorithm was used to effectively reduce the dimension and annotate the genes in the Saccharomyces cerevisiae enzyme constraint model. 968 genes were narrowed down to 52, and the recommended editing direction was annotated for each gene: 11 genes need to be overexpressed, 35 genes need to be knocked down, and 6 genes need to be knocked out.

[0083] Step S3, predicting the yield of single target editing.

[0084] The yield of a single target was predicted based on the ecFSEOF results combined with MOMA. The results are as follows: Figure 4 As shown. Here, the yield of the target product is defined as the output of 2-phenylethanol produced by the strain per unit time divided by the consumption of glucose. The red curve in the figure represents the 2-phenylethanol yield value of the wild-type strain. It can be found that the results of most single-target treatments are very similar to the reference values. After some targets are edited separately, the model cannot find the optimal solution. There is only one target that increases the yield of 2-phenylethanol after editing in the marked direction. This is because the synthesis of specific chemicals in organisms involves multiple reaction pathways, and changes in a single gene may not directly lead to an increase in product yield, indicating that it is necessary to search for combinatorial editing strategies.

[0085] Step S4: Genetic algorithm is used to search for combined targets after dimensionality reduction, and relevant parameters are adjusted to obtain the optimal gene editing combination strategy.

[0086] In the first experiment, the population size was set to 50, the maximum number of iterations was 500, the crossover probability was 0.8, the mutation probability was 0.05, and the other parameters remained unchanged. When the roulette wheel selection method was used to select the parent generation, the number of candidate individuals was set to 4. The binary encoding form is directly based on the gene dimension reduction and annotation results of the ecFSEOF algorithm, and the regulatory direction of the key genes is fixed as the basis of genetic encoding. The results are shown in Figure 5 As shown in the upper part of . From the results, it can be seen that the fitness value of the best individual in each generation generally shows an upward trend and eventually converges to a larger value. In the initialized population, the fitness value of the best individual in the first generation is 0.1374. After the algorithm iterates 500 times, the fitness value of the global best individual reaches 0.1412, which is 14.7% higher than the reference fitness value of 0.1231 of the wild-type strain. This result also significantly exceeds the prediction result of a single target. The blue curve shows the mean change of the fitness value of all individuals in each generation of the population. It generally shows an upward trend, but there are small fluctuations. This is because there are individuals in the population that make it impossible for the model to find the optimal solution (the fitness value is set to 0.01). As the genetic algorithm runs, these poor individuals are gradually eliminated, but crossover and mutation may produce new individuals that cannot be solved, lowering the mean fitness value.

[0087] The symbolic encoding form depends only on the gene dimensionality reduction results of ecFSEOF. While selecting the combination of genes to be edited, the editing methods of each gene are searched. Although this increases the computational complexity of the search, it may also reveal new regulatory relationships, especially the mutual influence relationship between genes. The population size is set to 100, the maximum number of iterations is 2000, and a convergence detection condition is introduced (the search is terminated early if the optimal fitness value does not change by more than 0.01 for 1000 consecutive generations). The results of the algorithm are shown in Figure 5 The results show that the fitness value of the best individual in the initial population is 0.1375, and the fitness value of the best individual finally obtained by the algorithm is 0.1630, which is 32.4% higher than the reference yield.

[0088] Comparing the results of the genetic algorithms under the two encoding forms, it can be concluded that both design ideas can search for gene editing combination strategies that make Saccharomyces cerevisiae have a higher 2-phenylethanol synthesis yield. The binary encoding genetic algorithm with the help of ecFSEOF's dimensionality reduction and gene annotation results can converge more quickly to obtain a set of optimal gene editing strategies, while the genetic algorithm based only on the symbolic encoding form of its dimensionality reduction results can break through the convergence limit of the former and obtain better gene editing strategies, but this requires a larger population size and more iterations, which consumes more time costs.

[0089] Step S5, repeat the experiment to verify the stability of the algorithm, and perform statistical analysis and biological interpretation on the obtained combination strategy.

[0090] In order to explore the impact of different parameters on the genetic algorithm, this experiment adjusted the number of parent individuals selected using the roulette wheel selection method. Increasing the number of candidate individuals to 20 increases the crossover probability of the algorithm. Selecting binary encoding and expanding the population size to 100, the maximum number of iterations allowed is 2000. The results are as follows: Figure 6 The upper part is shown. From the results, we can see that the convergence curve of the algorithm is steeper and the algorithm converges quickly. The fitness value of the optimal individual finally obtained is 0.1411, which is close to the previous experimental results. As the number of new individuals generated by crossover and mutation increases, the number of infeasible solutions generated in each iteration will also increase, which will lead to an increase in the volatility of the blue curve.

[0091] Genetic algorithm is a heuristic search algorithm, which has a certain degree of randomness. This is reflected in the initialization stage of the algorithm and the operation stage of genetic operators such as crossover and mutation. In order to study the stability of the algorithm, this study conducted ten repeated experiments for the same set of parameter settings. Figure 6The lower part shows the results of the ten algorithm runs, using the genetic algorithm in symbolic encoding. In most cases, the algorithm converges to 0.15 to 0.17. The red curve is the average of the fitness values ​​of the best individuals in each generation of the ten repeated experiments, and the average convergence is around 0.16.

[0092] In each experiment, the best individuals in the algorithm iteration process and the best fitness values ​​of each generation were retained. Based on the results of ten repeated experiments, the ten best gene editing strategies were statistically analyzed, and the number of times each gene target was coded as non-zero in the ten combined editing strategies was greater than or equal to 6. This means that the gene was selected as the object of gene intervention in multiple experiments. The results are as follows Figure 7 Most genes will be edited in three different ways in different combinatorial editing strategies. This is due to the diversity of biological systems. Different gene intervention methods may lead to the same optimization results.

[0093] In summary, the solution accuracy of the enzyme constraint model has a significant impact on the prediction results of the reference strain, and the imposition of additional constraints on the enzyme constraint model can effectively reduce the variability of the flux. The gene dimensionality reduction and annotation performed by the ecFSEOF algorithm not only streamlined the search space of the genetic algorithm, but also improved the search efficiency. The genetic algorithms under the two encoding forms can find gene editing strategies to improve the yield of 2-phenylethanol, and key genes in Saccharomyces cerevisiae such as Pdx1p are also effectively identified. The searched combined target results are significantly better than the results of single target treatment, which can provide more practical guidance for wet experiment design. These findings have important theoretical and practical significance for understanding and improving microbial metabolic engineering.

[0094] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for designing an optimal strain for product synthesis by integrating an enzyme constraint model with an evolutionary algorithm, characterized in that: The following steps are involved: Step S1, solving the enzyme constraint model with maximizing the strain specific growth rate as the objective function to obtain the simulated metabolic flux of the wild-type strain; Step S2, by solving a series of flux balance analysis problems with fixed biomass synthesis rates, dimensionality reduction analysis and annotation of genes directly related to enzyme synthesis in the model; Step S3, predicting the yield of single target editing; Step S4, using a genetic algorithm to search for combined targets for the reduced-dimensional targets, adjusting relevant parameters, and obtaining the optimal gene editing combination strategy; Step S5, repeat the experiment to verify the stability of the algorithm, and perform statistical analysis and biological interpretation on the obtained combination strategy.

2. The method for designing product synthesis optimized strains by using enzyme constraint model fusion evolution algorithm according to claim 1, characterized in that: The enzyme constraint model has made a series of necessary adjustments and verifications on the basis of GSMM to ensure that it accurately simulates the metabolic state of the strain under different environmental conditions, including releasing the absorption pathways of oxygen and major carbon and nitrogen sources, and setting all upper limits to b u , b u ∈[100,2000], set the accuracy of the model to Ω, Ω∈[10 -8 ,10 -10 ], increasing insight into protein synthesis.

3. The method for designing product synthesis optimized strains by using enzyme constraint model fusion evolution algorithm according to claim 1, characterized in that: The flux balance analysis method solves the enzyme constraint model and simulates the metabolic flux of the wild-type strain, including the following three methods: Method 1: After the minimum production limit of the target product is given, a single-step flux balance analysis is performed with maximizing the specific growth rate of the strain as the objective function; Method 2: Based on method 1, the specific growth rate is limited to not less than the maximum growth rate λ, λ∈[0.5,0.9], and the FBA solution is performed with the objective function of minimizing the consumption of glucose, and the glucose consumption rate is limited to not more than α times its minimum consumption, α∈[1.1,1.5], and the FBA solution is performed with the objective function of minimizing the total enzyme consumption; Method 3: Perform FBA three times according to Method 2 without limiting the yield of the target product, and then limit the total enzyme consumption to no more than β times its minimum value, β∈[1.1,1.5], and perform FBA with maximizing product yield as the objective function.

4. The method for designing optimal strains for product synthesis by using an enzyme constraint model fusion evolution algorithm according to claim 3, characterized in that: The flux balance analysis used is expressed by the following formula: max c T ·v Where c is a constant vector consisting of 0 and 1, v is the flux vector representing all reactions, the transpose of c multiplied by v defines the optimization direction of a model, that is, maximizing the flux of a reaction, S is the stoichiometric matrix of the model, and Represent the minimum and maximum constraint boundaries of the reaction, respectively.

5. The method for designing product synthesis optimized strains by fusion evolution algorithm using enzyme constraint model according to claim 1, characterized in that: The gene dimensionality reduction and annotation is to analyze the changes in metabolic reactions in the FBA solution results under 16 different conditions of gradually reducing the biomass synthesis rate for all genes related to enzyme synthesis, quantify the changes in reaction flux through two-step scoring, and annotate the corresponding genes according to different scoring thresholds, and the annotation methods include overexpression, knockdown and knockout; the dimensionality reduction of gene targets is to screen out the genes most relevant to the synthesis of target products, reduce background noise, and make the algorithm focus on the genes that have the greatest influence on improving the yield of target products; the annotation of gene targets is to additionally provide intervention methods for selected genes.

6. The method for designing optimal strains for product synthesis by using enzyme constraint model fusion evolution algorithm according to claim 1, characterized in that: Design a representation of gene editing in the model where: The knockout operation of a gene is to set the upper and lower limits of the reaction flux of the corresponding enzyme synthesis pathway to 0, indicating that the synthesis pathway of the enzyme is closed; The overexpression and knockdown operations of genes require manual adjustment. Taking the predicted wild-type flux as a reference, the overexpression operation is to increase the lower limit of the flux of the relevant enzyme synthesis reaction to N times the reference value, and the knockdown operation is to adjust the upper limit of the flux to M times the reference value. If the reference flux value of the target to be overexpressed is 0, its lower limit of flux is set to v low , v low ∈[10 -9 ,4×10 -9 ].

7. The method for designing product synthesis optimized strains by fusion evolution algorithm using enzyme constraint model according to claim 1, characterized in that: In step S3, based on the results of gene dimensionality reduction and annotation, the yield of single target editing is predicted one by one, and the edited enzyme constraint model is solved by the metabolic adjustment minimization method to obtain the metabolic phenotype of the mutant strain, and the yield of the target product is calculated.

8. The method for designing optimal strains for product synthesis by using enzyme constraint model fusion evolution algorithm according to claim 1, characterized in that: In step S4, a genetic algorithm is used to search for combined targets for the targets after dimensionality reduction, including two encoding forms: The binary coding form genetic algorithm relies on the results of gene dimensionality reduction and annotation to search for combinatorial targets to be edited; The symbolic coding form of the genetic algorithm only relies on the results of gene dimensionality reduction, and simultaneously finds the combination targets and the editing directions of the targets in a larger search space.

Images