Metabolic network processing system and metabolic network processing method

The metabolic network processing system addresses the challenge of predicting enzyme modification effects by generating sensitivity matrices and verifying results through biological experiments, ensuring accurate proposal of modified enzymes that enhance target substance yield.

WO2024236670A1PCT designated stage expired Publication Date: 2024-11-21HITACHI LTD
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
PCT/JP2023/018013
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Filing Date
2023-05-12
Publication Date
2024-11-21

AI Technical Summary

Technical Problem

Current metabolic network design technologies, such as flux balance analysis (FBA), struggle to propose enzyme modifications that improve target substance yield due to the inability to introduce reaction kinetics, making it difficult to predict and verify the effects of enzyme modifications on metabolic networks.

Method used

A metabolic network processing system that generates sensitivity matrices using a stoichiometric matrix, core basis, and probability distribution to extract and propose modified enzymes with high confidence, allowing for the prediction of yield increases or decreases, and verifies these predictions through biological experiments.

Benefits of technology

This approach consistently predicts the impact of enzyme modifications on target substance yield, improving proposal accuracy and enabling the selection of modified enzymes that achieve desired yields, while reducing dependence on initial probability distributions.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure JP2023018013_21112024_PF_FP_ABST
    Figure JP2023018013_21112024_PF_FP_ABST
Patent Text Reader

Abstract

The present disclosure proposes a metabolic network processing technology, one example of which being a metabolic network processing system that extracts and proposes a modified enzyme with which a target yield of a target substance in a metabolic network is achieved, in order to invariably allow prediction of an increase / decrease in yield of the target substance when proposing modification of an enzyme on the basis of structural sensitivity analysis. In the system, a control device executes (i) a process for generating N-number of sensitivity matrices using a stoichiometric matrix obtained by converting the metabolic network, a basis c of the kernel of the stoichiometric matrix, and a given probability distribution, (ii) a process for checking the sign of each matrix element of the N-number of sensitivity matrices, and calculating the proportion of signs in each of the matrix elements, thereby calculating the certainty factor of each of the matrix elements, (iii) a process for extracting, as modified enzymes, k-number of matrix elements having a high certainty factor from among the certainty factors of which the quantity is the same as the number of the matrix elements in the sensitivity matrix, and (iv) a process for outputting the extracted k-number of modified enzymes (see fig. 2).
Need to check novelty before this filing date? Find Prior Art

Description

Metabolic network processing system and metabolic network processing method

[0001] The present disclosure relates to a metabolic network processing system and a metabolic network processing method.

[0002] In biosynthesis, it is important to improve the yield of the target substance through metabolic network design. Here, biosynthesis is a method of producing substances through the biological activities of microorganisms. Biosynthesis has a lower environmental impact than chemical synthesis and can be carbon-neutral. Furthermore, there have been cases where biosynthesis is superior to chemical synthesis in terms of cost.

[0003] A metabolic network refers to the collection of enzymatic reactions that an organism possesses. Metabolic network design refers to the design of a metabolic network suitable for purposes such as substance production by deleting, inserting, or modifying those enzymes. Here, enzyme deletion, insertion, and modification refer to removing an enzymatic reaction from a metabolic network, adding an enzymatic reaction to a metabolic network, and changing the conditions for an enzymatic reaction, respectively, through genetic modification, genome editing, or the introduction of a plasmid. Therefore, metabolic network design can improve the efficiency of substance production. However, selecting enzymes to delete, insert, or modify is difficult. For this reason, technology that can suggest enzymes to delete, insert, or modify based on information such as metabolic networks is important. In particular, proposing modified enzymes is particularly important, as enzyme modification can be achieved using protein engineering techniques.

[0004] To overcome the difficulties of metabolic network design, for example, Patent Document 1 discloses a technique for proposing enzymes to be deleted or inserted using flux balance analysis (hereinafter referred to as FBA). Many other examples of deletions and insertions into metabolic networks have also been reported.

[0005] Japanese Patent Application Laid-Open No. 2021-16325

[0006] Fiedler, Bernold, and Atsushi Mochizuki. "Sensitivity of chemical reaction networks: a structural approach. 2. Regular monomolecular systems." Mathematical methods in the applied sciences 38.16 (2015): 3519-3537.

[0007] However, while FBA can suggest enzyme deletions and insertions, it cannot suggest enzyme modifications. This is because enzyme modification requires the introduction of reaction kinetics, but FBA simplifies the problem by analyzing without incorporating reaction kinetics. Furthermore, to improve the accuracy of the suggestions, it is desirable to verify the results of the suggestions through biological experiments and provide feedback.

[0008] However, because the deletion or insertion of enzymes changes the shape of the metabolic network, the predicted results cannot be compared with data obtained from biological experiments, making it difficult to improve the next proposal.

[0009] Meanwhile, as shown in Non-Patent Document 1, structural sensitivity analysis (SSA) has attracted attention as a technique for proposing enzyme modifications, which are important in metabolic network design. Structural sensitivity analysis makes it possible to predict the impact of enzyme modifications on substance production based solely on the structure of the metabolic network, without the need to determine parameters from data or other sources. Specifically, for all enzymes in a metabolic network, it is possible to predict whether the yield of a target substance will increase, decrease, or remain unchanged when the corresponding enzymatic reaction is promoted. Furthermore, because enzyme modifications do not change the shape of the metabolic network, the predicted results can be verified through biological experiments, allowing for feedback of the results.

[0010] On the other hand, SSA is inconsistent, meaning that there are cases where it is not possible to predict increases or decreases in yield. This is due to the strict analytical conditions of SSA. Therefore, it may not be possible to propose modified enzymes simply by using SSA with strict analytical conditions. In view of this situation, the present disclosure proposes a technology that always makes it possible to predict increases or decreases in yield of a target substance when proposing enzyme modifications based on structure sensitivity analysis.

[0011] In order to solve the above-mentioned problems, the present disclosure proposes, as an example, a metabolic network processing system that extracts and proposes modified enzymes that achieve a target yield of a target substance in a metabolic network, the metabolic network processing system comprising: a storage device that stores a metabolic network processing program that extracts and outputs modified enzymes; and a control device that processes the metabolic network in accordance with the metabolic network processing program read from the storage device and extracts modified enzymes. The control device performs the following processes: generating N sensitivity matrices using a stoichiometric matrix obtained by converting the metabolic network, a kernel basis c of the stoichiometric matrix, and a given probability distribution; calculating the certainty of each matrix element of the N sensitivity matrices by checking the sign of each matrix element and calculating the proportion of the sign for each matrix element; extracting k matrix elements with the highest certainty from the certainty levels corresponding to the number of matrix elements in the sensitivity matrix as modified enzymes; and outputting the extracted k modified enzymes.

[0012] Further features related to the present disclosure will become apparent from the description of this specification and the accompanying drawings. Also, aspects of the present disclosure are achieved and realized by the elements and combinations of various elements and the aspects of the following detailed description and the appended claims. The description of this specification is merely exemplary and does not limit the scope or application of the claims of the present disclosure in any way.

[0013] According to the technology of the present disclosure, when proposing enzyme modifications based on structure sensitivity analysis, it is possible to consistently predict increases or decreases in the yield of a target substance.

[0014] 1 is a diagram illustrating an outline of the hardware configuration of a metabolic network processing system 10 according to an embodiment of the present disclosure. It is a diagram illustrating an example sequence of processing executed in the metabolic network system according to an embodiment of the present disclosure. It is a diagram illustrating an example display of an information input GUI and an information output GUI in the processing sequence of FIG. 2. It is a flowchart illustrating an overview of metabolic network processing. It is a flowchart illustrating details of the structural sensitivity analysis processing (S401 in FIG. 4). It is a diagram illustrating an example of one sensitivity matrix generated from a metabolic network (using a simple one as an example) and a probability distribution p. It is a flowchart illustrating details of the extraction of a predetermined number (k) of modified enzymes based on confidence levels (S402). It is a flowchart illustrating details of the determination of the effectiveness of modified enzymes based on biological experiment results (S403). It is a flowchart illustrating details of the sequential Bayesian estimation processing (S406). It is a diagram illustrating an example of the probability distribution p updated by fitting a multivariate lognormal distribution to a sample distribution.

[0015] This embodiment proposes a metabolic network processing technology that introduces a new concept of certainty into structural sensitivity analysis, relaxes the conditions of structural sensitivity analysis, and extracts and proposes modified enzymes with high certainty, thereby enabling consistent prediction of increases or decreases in the yield of a target substance in a metabolic network. This embodiment will be described below with reference to the accompanying drawings. Note that in the accompanying drawings, functionally identical elements may be indicated by the same numbers. Note that the accompanying drawings show specific embodiments and implementation examples in accordance with the principles of the present disclosure, but these are intended to facilitate understanding of the present disclosure and should not be used to limit the interpretation of the present disclosure.

[0016] Furthermore, although the present embodiment has been described in sufficient detail to enable those skilled in the art to implement the present disclosure, it should be understood that other implementations and forms are possible, and that changes to the configuration and structure and substitutions of various elements are possible without departing from the scope and spirit of the technical ideas of the present disclosure. Therefore, the following description should not be interpreted as being limited thereto.

[0017] <Terminology> We will first provide a brief explanation of each term used in this specification. "Enzyme" refers to a molecule in a living organism that has the function of converting (catalyzing) one substance (compound) into another substance. "Metabolism" refers to the process of converting substances into energy through a series of enzymatic reactions. "Metabolic network" refers to a network representation of the transition in which one substance produced by another enzyme is converted into yet another substance by another enzyme. "Flux" refers to the flow, speed, and amount of a substance transition. Flux analysis refers to the analysis of the flow, speed, and amount of a substance transition. "Flux Balance Analysis (FBA)" refers to a technology that expresses the flux of a metabolic network as a matrix (stoichiometric matrix) and simulates the transition of substances that cannot be observed with the naked eye. "Structure Sensitivity Analysis (SSA)" is the name of a unique method that adds the word "structure" to the general term "sensitivity analysis," which describes a type of analysis. It is a technology that suggests enzymes that affect the amount of substances on a metabolic network based on the structure (= shape) of the metabolic network.

[0018] Furthermore, "enzyme modification" refers to changing the parameters that control the reaction rate of an enzyme. Here, parameters refer to enzyme concentration, turnover rate, substrate affinity, etc. Furthermore, "proposing a modified enzyme" refers to proposing an enzyme that should be modified to improve the efficiency of substance production.

[0019] By "biological experiments" herein is meant experiments that quantify the effect of enzyme modifications on the efficiency of biosynthesis, such as by comparing enzyme-modified organisms with those that are not.

[0020] "Probability distribution" refers to the ease of obtaining values ​​assumed to be obeyed by the random variables required for the algorithm proposed by the modified enzyme. The type is not particularly limited unless it is necessary to limit it by the algorithm. In this embodiment, for example, a log-normal distribution, a mixed log-normal distribution, a gamma distribution, etc. can be used as the probability distribution. "Confidence" refers to the value obtained by taking the expected value from the probability distribution for the algebraic expression of sensitivity in structural sensitivity analysis.

[0021] "Sequential Bayesian estimation" refers to updating a probability distribution to a posterior distribution using Bayesian estimation. The algorithm is not particularly limited. For example, the distribution may be fitted using sample statistics obtained from the posterior distribution using the Markov chain Monte Carlo method, or the sample itself may be used as a distribution to propose the next modified enzyme.

[0022] 1 is a diagram showing an outline of the hardware configuration of a metabolic network processing system 10 according to this embodiment. The metabolic network processing system 10 can be configured as a computer system, and includes, for example, a control device (processor) 11, a storage device 12, an input device 13, and an output device 14.

[0023] The control device 11 reads the metabolic network processing program from the storage device 12 and expands it into an internal memory (not shown), thereby realizing a modified enzyme proposal unit 100, an experimental data verification unit 200, and a sequential Bayesian estimation unit 300.

[0024] The storage device 12 stores a metabolic network processing program that identifies modified enzymes that increase the yield of a target substance in a specified metabolic network based on information about the metabolic network and its probability distribution, and proposes the modified enzymes. The metabolic network processing program corresponds to the flowchart in Figure 4 (including the flowcharts in Figures 5 and 7 to 9) described below.

[0025] The storage device 12 may also be configured to function as a metabolic information DB (database) that stores metabolic networks corresponding to metabolic network names and identification information. If the storage device 12 does not include a metabolic information DB, the control device 11 may access an external database (e.g., a public database such as KEGG or Metacyc) via a communication network (not shown) to obtain a metabolic network specified by the user.

[0026] The input device 13 is composed of, for example, a keyboard, a mouse, a touch panel, a switch, etc. For example, the input device 13 can be used when an operator (user) inputs (specifies) metabolic network information (name and identification information) to be processed, target substance information (information on a substance whose yield is to be improved), target yield information (a target value for the yield), information on a probability distribution to be used as an initial value (a probability distribution name and identification information), etc., to the metabolic network processing system 10 via, for example, a GUI.

[0027] The output device 14 is composed of a display device, a printer, etc., and displays, for example, the above-mentioned GUI, the results obtained by processing the metabolic network (e.g., suggestions for modified enzymes), and the experimental results described below (e.g., information on the yield obtained by the organism) on a display screen.

[0028] In Figure 1, the computer system is configured so that the control device 11, memory device 12, input device 13, and output device 14 are connected via data lines / control lines, but the metabolic network processing system 10 may also be configured so that each device is located remotely and connected via a communication network.

[0029] <Example of Processing Sequence> FIG. 2 is a diagram showing an example of a processing sequence executed in the metabolic network system according to this embodiment.

[0030] First, the user operates the input device 13 to input or specify (i) metabolic network information, (ii) target substance information, (iii) target yield information, and (iv) probability distribution information into the information input GUI (Graphical User Interface) (SQ1). The control device 11 then stores the input or specified information in the storage device 12 (SQ2). The control device 11 also reads the metabolic network and target substance input (specified) by the user from the storage device 12 (SQ3). As mentioned above, information may also be acquired from an external database. The control device 11 then performs a structural sensitivity analysis incorporating certainty factors (an analysis in which the application conditions in conventional SSA are less stringent) according to the metabolic network processing program, extracts a predetermined number (e.g., five) of modified enzymes (modified enzymes with high certainty) from the target metabolic network (as candidates), and displays (proposes) the modified enzymes on the information output GUI on the screen of the output device (display device) 14 (SQ4 and SQ5). Here, the modified enzyme means an enzyme that can potentially improve yields by modification.

[0031] The user performs a biological experiment (actually modifying the enzyme and confirming the change in yield) to verify the proposed modified enzyme (verifying whether the target yield can be achieved) (SQ6). The user also obtains the results of the enzyme modification determined by the biological experiment (SQ7) and operates the input device 13 to input the results of the enzyme modification into the information input GUI.

[0032] The control device 11 records the enzyme modification results (experimental results) entered by the user in the storage device 12 (SQ9). The control device 11 then reads from the storage device the metabolic network, target substance, and target yield information initially entered (specified) by the user, along with the enzyme modification results (experimental results) (SQ10). The control device 11 then compares the metabolic network, target substance, and enzyme modification results (experimental results) from a biological experiment with the metabolic network, target substance, and target yield specified by the user. If a predetermined number (e.g., 10) of modified enzymes achieve the target yield, the control device 11 terminates processing (SQ11). On the other hand, if the predetermined number of modified enzymes achieve the target yield, the control device 11 performs sequential Bayesian estimation to change (update) the probability distribution. Then, the control device 11 uses the modified probability distribution to again perform structural sensitivity analysis incorporating confidence levels to extract a predetermined number of modified enzymes (e.g., five), and displays (proposes) these as proposed modified enzymes on the output GUI of the output device 14 (SQ12 and SQ13).

[0033] The user then conducts another biological experiment to verify the newly proposed modified enzyme (SQ6). Thereafter, the processing sequence from SQ6 to SQ13 is repeated until a predetermined number (e.g., 10) or more modified enzymes that achieve the target yield are obtained.

[0034] <GUI Display Example in Processing Sequence> FIG. 3 is a diagram showing a display example of an information input GUI and an information output GUI in the processing sequence of FIG.

[0035] When a user operates the input device 13 to press a button to start the enzyme modification proposal process (e.g., a start button (not shown) provided on the information input GUI), the control device 11, in response to the button press (input of an instruction to start the enzyme modification proposal process), displays a metabolic network input screen 301 (corresponding to SQ1), a target substance input screen 302 (SQ1), a target yield input screen 303 (SQ1), and a probability distribution input screen (not shown: SQ1) at the timing of each information input. The order of display does not have to be the order shown in Figure 3. The GUI may also be configured so that the metabolic network, target substance, target yield, and probability distribution can be input on a single input screen.

[0036] For example, in the example of FIG. 3, a user refers to a file of metabolic network data on a metabolic network input screen 301, and inputs an increase in the yield of succinic acid (Succinate) and carbon dioxide (CO ) on a target substance input screen 302. 2 ) is selected. Also, the target yield input screen 303 specifies that the target yield is to increase succinate by 5.0 mol / mol (mol / mol: amount (amount of succinate) per mol of glucose).

[0037] After inputting the information required for the modified enzyme proposal process is completed, the control device 11 performs the structural sensitivity analysis incorporating the confidence factor as described above to extract a predetermined number of modified enzymes (e.g., five), and displays these as modified enzymes to be proposed on the modified enzyme proposal screen 304 (SQ4 and SQ12). 2 A list of proposed modified enzymes that reduce the amount of β-glucan is displayed. Although the modified enzyme proposal screen 304 displays a list, other display formats may be used.

[0038] The user conducts a biological experiment on the proposed modified enzyme or a modified enzyme that the user has separately considered, to verify the change (increase or decrease) in the yield of the target substance due to the enzyme, and then enters the experimental results, i.e., the numerical value of the increase or decrease in yield, into the biological experiment result input screen 305 (SQ8).

[0039] The display of modified enzymes on the modified enzyme proposal screen 304 and the input of experimental results on the biological experiment result input screen 305 are repeated until the desired number (e.g., 10) of modified enzymes that achieve the target yield are obtained (SSQ6 to SQ13).

[0040] <Outline of Metabolic Network Processing> Fig. 4 is a flowchart for explaining an outline of metabolic network processing. The operating subjects of each step in the flowcharts shown in Fig. 4 to Fig. 7 are the modified enzyme proposing unit 100, the experimental data verifying unit 200, or the sequential Bayesian inference unit 300, but since these are realized by the control device 11 executing the metabolic network processing program, the control device 11 may also be the operating subject.

[0041] (i) S401: The modified enzyme proposing unit 100 performs a structural sensitivity analysis process using the metabolic network, target substance, and probability distribution p specified by the user, and generates N (a predetermined number: changeable) sensitivity matrices. Details of the process of S401 will be described later with reference to FIGS. 5 and 6.

[0042] (ii) S402: The modified enzyme proposal unit 100 calculates a confidence level based on the sign (positive or negative) of each element in each of the N sensitivity matrices. Here, the confidence level refers to the expected value calculated by the probability distribution for the algebraic sensitivity expression in structural sensitivity analysis, as described above. Specifically, the confidence level here is the ratio of positive signs to negative signs for each element in the N sensitivity matrices. The positive ratio represents a positive confidence level, and the negative ratio represents a negative confidence level. The modified enzyme proposal unit 100 then selects k enzymes (e.g., k = 5) with high confidence levels and proposes them as modified enzyme candidates. For example, if the positive confidence level of a certain element (1, 1) in a sensitivity matrix exceeds 80% and is within the k most highly confident elements in the sensitivity matrix group, the enzyme corresponding to that element is proposed as a modified enzyme candidate. Details of the processing in S402 will be described later with reference to FIG. 7. Once the modified enzymes are proposed, the user actually modifies the target enzyme in the target metabolic network and conducts a biological experiment.

[0043] (iii) S403: The experimental data verification unit 200 acquires the results of an actual biological experiment input by the user and determines the effectiveness of the modified enzyme based on the experimental results. Details of the process of S403 will be described later with reference to FIG. 8.

[0044] (iv) S404 The modified enzyme suggestion unit 100 checks whether M or more (e.g., M = 10) effective modified enzymes determined in S403 have been obtained. If M or more effective modified enzymes have been obtained (Yes in S404), the process proceeds to S405. If M or more effective modified enzymes have not been obtained (No in S404), the process proceeds to S406.

[0045] The reason for proposing multiple (M) modified enzymes is to accommodate situations where the desired effect is not achieved even after modifying (increasing) a proposed enzyme and a different enzyme is desired to be tried, and also to accommodate requests to use another enzyme if the desired purpose can be achieved by modifying a specific enzyme when circumstances make it undesirable to modify that enzyme.

[0046] (v) S405 The modified enzyme proposing unit 100 creates a proposal list (modified enzyme proposal list) including the M or more effective modified enzymes obtained in S403, and outputs it (presents it to the user).

[0047] (vi) S406: The sequential Bayesian estimation unit 300 uses the results obtained from the biological experiment (verification experiment) to perform sequential Bayesian estimation of the probability distribution p, and generates a new probability distribution p. Details of the processing in S406 will be described later with reference to FIG. 9.

[0048] <Details of Structural Sensitivity Analysis Process (S401)> Fig. 5 is a flowchart for explaining the details of the structural sensitivity analysis process (S401 in Fig. 4). Fig. 6 is a diagram showing an example of a metabolic network (a simple one as an example) and one sensitivity matrix generated from a probability distribution p.

[0049] (i) S4011 The modified enzyme proposal unit 100 acquires the metabolic network name or metabolic network identification information, the target substance whose yield is to be changed (increased or decreased), the target yield of the target substance, and the initial probability distribution information, all of which are input by the user via the GUI (metabolic network input screen 301, target substance input screen 302, target yield input screen 303, and probability distribution input screen) by operating the input device 13. The modified enzyme proposal unit 100 also accesses a metabolic network DB (a DB installed inside or outside the system) and acquires the metabolic network (detailed information such as structure) based on the acquired metabolic network name, etc. Furthermore, the modified enzyme proposal unit 100 acquires data on the probability distribution p specified by the user from the storage device 12 or another DB.

[0050] (ii) In step S4012, the modified enzyme proposing unit 100 converts the metabolic network into a stoichiometric matrix representing the metabolic network. In the example of Figure 6, a simple metabolic network (1) is shown. The metabolic network is converted into a stoichiometric matrix (2).

[0051] (iii) S4013 The modified enzyme proposal unit 100 calculates the kernel (3) basis c from the (2) stoichiometric matrix. Subsequently, when the processing of S4013 is completed, the modified enzyme proposal unit 100 executes N (e.g., 10,000) sensitivity matrix generation loops (S4014 to S4018).

[0052] (iv) S4014: The modified enzyme proposing unit 100 performs random sampling on the probability distribution p to obtain multiple sample values. The number of samples corresponds to the number of elements showing negative values ​​in the (2) stoichiometric matrix. From the (4) probability distribution p in Figure 6, values ​​of 0.026, 0.75, and 0.87 are obtained by random sampling.

[0053] (v) S4015 The modified enzyme proposal unit 100 generates a matrix r by replacing matrix elements that take negative values ​​in the stoichiometric matrix with the sampled values ​​acquired in S4014. For example, the matrix r is a matrix generated by replacing the three "-1"s enclosed in a rectangle in the stoichiometric matrix (2) in Figure 6 with the sampled values ​​0.75, 0.87, and 0.026, respectively.

[0054] (vi) S4016 The modified enzyme proposing unit 100 replaces, in the matrix r, elements of the stoichiometric matrix that correspond to elements that take positive values ​​with “0.” For example, in FIG. 6, all positive values ​​“1” in the stoichiometric matrix are replaced with “0.”

[0055] (vii) S4017 The modified enzyme proposal unit 100 generates matrix A by arranging matrix r', which is obtained by replacing matrix elements corresponding to matrix elements that take positive values ​​in the stoichiometric matrix, with "0", and the base c. In Figure 6, (5) matrix A is a 4-row, 4-column matrix in which the base c is arranged in each element (4 × 3 matrix) of matrix r + the 4th column.

[0056] (viii) S4018 The modified enzyme proposing unit 100 takes the inverse matrix of the A matrix obtained in S4017 and then inverts the sign to generate a sensitivity matrix. Figure 6 shows how one (6) sensitivity matrix is ​​generated from the (5) A matrix.

[0057] As described above, one sensitivity matrix is ​​generated by the processes from S4014 to S4018. Then, the sampling process (S4014) for the probability distribution p is executed N times, and N sensitivity matrices are generated by using new sampled values. Once N sensitivity matrices have been generated, the process proceeds to S4019.

[0058] (ix) S4019 The modified enzyme proposing unit 100 acquires the sign (positive or negative) of each element of the generated N sensitivity matrices. For example, in the case of a 4x4 matrix such as the sensitivity matrix (6) in Figure 6, the N (e.g., 10,000) codes in the first row and first column to the N (10,000) codes in the fourth row and fourth column are acquired (collected). Note that the processing up to S4019 corresponds to the existing structural sensitivity analysis processing.

[0059] 7 is a flowchart for explaining the details of the process of extracting a predetermined number (k) of modified enzymes based on the confidence factor (S402). In this embodiment, unlike existing structural sensitivity analysis processes, the concept of confidence factor is introduced to relax the application conditions of the analysis.

[0060] (i) S4021 The modified enzyme proposal unit 100 calculates the proportion of positive signs and the proportion of negative signs for the N signs in each element based on the sign information of each element in the N sensitivity matrices acquired in S4019. The modified enzyme proposal unit 100 then defines the positive proportion value as a positive confidence level, and the negative proportion value as a negative confidence level. Therefore, the number of elements in the sensitivity matrix x 2 confidence levels are calculated.

[0061] Each element of the sensitivity matrix corresponds to an enzyme to be modified. Therefore, a high confidence level for a matrix element means that modification is likely to result in a change in the yield of the target substance. For example, if a matrix element has a high positive confidence level, increasing the enzyme corresponding to that element is likely to improve the yield of the target substance.

[0062] (ii) S4022: The modified enzyme proposing unit 100 selects k modified enzymes (k in descending order of confidence: for example, 5) that show high confidence values ​​and extracts them as modified enzymes (candidates). Then, the output device (display device) 14 outputs (displays) information on the modified enzymes (candidates) in response to an instruction from the modified enzyme proposing unit 100.

[0063] <Details of Determining Effectiveness of Modified Enzyme Based on Biological Experiment Results (S403)> FIG. 8 is a flowchart for explaining details of determining effectiveness of modified enzyme based on biological experiment results (S403).

[0064] (i) S4031 The experimental data verification unit 200 acquires information on yields (biological experimental results) when an experiment is conducted after a user modifies an enzyme in the metabolic network according to the modified enzyme proposed in S4022. The biological experimental results are acquired, for example, by the user inputting the experimental results via the biological experimental result input screen 305.

[0065] It is difficult to suggest how much enzyme to modify. Therefore, users must decide how much enzyme to increase or decrease as appropriate. However, it is necessary to use the same amount of modification (i.e., conduct biological experiments under the same conditions (same amount and same procedure)) for each proposed enzyme (and / or enzymes that the user decides to modify). If the amount of enzyme modification varies, it will be impossible to properly compare yields between modified enzymes, making the target yield meaningless.

[0066] (ii) S4032: The experimental data verification unit 200 assigns a "1" if the yield of the target substance due to the enzyme modification obtained in S4031 is increased compared to the yield of the wild type (comparing the yield before and after modification), and assigns a "0" if the yield is decreased, and records the information in the storage device 12 in association with the modified enzyme. Here, "wild type" refers to an organism having a metabolic network that operates with the amount of enzyme before modification through biological experiments. In the case of a wild type, the yield data (yield) of the target substance is known in advance.

[0067] <Details of Sequential Bayesian Inference Process (S406)> Figure 9 is a flowchart for explaining the details of the sequential Bayesian Inference process (S406). The confidence value is affected by the probability distribution p, which was used to calculate each sensitivity matrix. Therefore, the probability distribution p is thought to be the reason why a sufficient number of modified enzymes effective in achieving the target yield of the target substance could not be secured. Therefore, the sequential Bayesian Inference process is used to improve the probability distribution p based on the results of biological experiments (changing the probability distribution p so that modified enzymes effective in achieving the target yield can be extracted → preventing the confidence value from being affected by the initial probability distribution).

[0068] (i) S4061 The sequential Bayesian estimation unit 300 performs Monte Carlo sampling from the function of the product of the likelihood and the probability distribution p in order to improve the probability distribution p as described above. Specifically, the sequential Bayesian estimation unit 300 calculates a function (p(x 目的物質,酵素AMonte Carlo sampling is performed from p(θ) (p(θ) = 0|θ) × p(θ)). Here, Monte Carlo sampling is the process of repeatedly obtaining samples from an arbitrary probability distribution (or a function of a constant multiple thereof). Since this is a well-known calculation method, a detailed description will be omitted.

[0069] When focusing on one element of the sensitivity matrix, the prediction result changes depending on the sampled values ​​from the probability distribution p. Conceptually, the processing in this step can be said to be processing that changes the probability distribution p so as to weight the combination of sampled values ​​that reduces the yield in this way.

[0070] (ii) S4062 The sequential Bayesian estimation unit 300 performs fitting on the distribution of the samples obtained in S4061 to generate a new probability distribution p. As shown in Fig. 10 , the probability distribution p is updated by fitting a multivariate log-normal distribution to the sample distribution. Once the probability distribution p is updated, the processing following the N sensitivity matrix generation loops is executed again using the new probability distribution p.

[0071] (i) This embodiment proposes a metabolic network processing and method for extracting and proposing modified enzymes that achieve a target yield of a target substance in a metabolic network. Specifically, the control device 11 loads a metabolic network processing program (a program corresponding to a flowchart such as that shown in FIG. 4 ) from the storage device 12, and executes the following steps in accordance with the program: generating N sensitivity matrices using a stoichiometric matrix obtained by converting the metabolic network, a kernel basis c of the stoichiometric matrix, and a given probability distribution p (S401 in FIG. 4 ; S4011 to S4019 in FIG. 5 ); calculating the confidence factor of each matrix element of the N sensitivity matrices by checking the sign of each matrix element and calculating the proportion of the sign of each matrix element (S402 in FIG. 4 ; S4021 in FIG. 7 ); and extracting k (e.g., 5) matrix elements with the highest confidence factors from the confidence factors for the number of matrix elements in the sensitivity matrix as modified enzymes and outputting the k modified enzymes (S402 in FIG. 4 ; S4022 in FIG. 7 ). In this way, by using the confidence factor when proposing modified enzymes using SSA, it is possible to always propose modified enzymes. Note that, as the probability distribution p, for example, a log-normal distribution, a mixed log-normal distribution, or a gamma distribution can be applied.

[0072] When generating N (e.g., 10,000) sensitivity matrices, the control device 11 (i) generates matrix r by replacing negative matrix elements of the stoichiometric matrix with sampled values ​​of probability distribution p (S4014 and S4015), (ii) generates matrix r' by replacing matrix elements of matrix r that have positive values ​​in the stoichiometric matrix with zeros (S4016), (iii) generates matrix A by arranging matrix r' and base c (S4017), and (iv) generates a sensitivity matrix by taking the inverse matrix of matrix A and inverting the sign (S4018). The generation of the sensitivity matrix follows the structural sensitivity analysis method, thereby taking advantage of the benefits of structural sensitivity analysis. In other words, there is no need to determine parameters from data; the effects of enzyme modifications can be predicted from the structure of the metabolic network alone.

[0073] (ii) This embodiment utilizes biological experimental results to verify the yield of a target substance produced by a proposed modified enzyme (candidate). Specifically, the control device 11 executes a process (S403: S4031 and S4032) of comparing the yield of the target substance obtained as a result of a biological experiment using the modified enzyme with the yield of the target substance in the metabolic network (wild-type) before modification, and a process (S404 and S405) of outputting the modified enzyme with improved yield as a final proposal. In this way, by feeding back the results of biological experiments conducted to verify the proposal and taking the experimental results into consideration, the modified enzyme proposal can be made while mitigating its dependence on the initial probability distribution p due to the use of confidence factors. If the condition for the number of modified enzymes with improved yield is met (e.g., if there are 10 or more effective modified enzymes), the control device 11 outputs the modified enzyme as a final proposal.

[0074] On the other hand, if the condition for the number of modified enzymes that improve yield is not met (for example, if there are fewer than 10 effective modified enzymes or if the above modified enzymes (candidates) are not effective), the control device 11 performs a sequential Bayesian estimation process and a process of updating the probability distribution p (S406: S4061 and S4062).

[0075] (iii) When the probability distribution p is updated, the control device 11 uses the updated probability distribution p to again execute the following series of processes: sensitivity matrix generation → confidence calculation → proposed modified enzymes (k) → verification based on biological experiment results → output of a list of modified enzymes. That is, the control device 11 executes the following processes: generating new N sensitivity matrices using the stoichiometric matrix, the basis c, and the updated probability distribution p (S4014 to S4018); calculating the confidence of each matrix element in the newly generated N sensitivity matrices by checking the sign of each matrix element and calculating the proportion of signs in each matrix element (S4021); extracting k matrix elements with high confidence from the confidence levels of the matrix elements in the newly generated N sensitivity matrices as modified enzymes (S4022); and outputting the newly extracted k modified enzymes (S4022). In this way, repeated feedback gradually reduces the dependence on the initial probability distribution p, thereby improving the proposal accuracy.

[0076] (iv) The functions of the present embodiment can also be realized by software program code. In this case, a storage medium on which the program code is recorded is provided to a system or device, and the computer (or CPU or MPU) of the system or device reads the program code stored in the storage medium. In this case, the program code itself read from the storage medium realizes the functions of the above-mentioned embodiment, and the program code itself and the storage medium on which it is stored constitute the present disclosure. Examples of storage media for providing such program code include flexible disks, CD-ROMs, DVD-ROMs, hard disks, optical disks, magneto-optical disks, CD-Rs, magnetic tape, non-volatile memory cards, and ROMs.

[0077] Furthermore, an operating system (OS) running on a computer may perform some or all of the actual processing based on instructions in the program code, and the functions of the above-described embodiments may be realized by this processing.Furthermore, after the program code is read from a storage medium and written to a memory on the computer, a CPU of the computer may perform some or all of the actual processing based on instructions in the program code, and the functions of the above-described embodiments may be realized by this processing.

[0078] Furthermore, the program code of the software that realizes the functions of the embodiment may be distributed via a network and stored in a storage means such as a hard disk or memory of the system or device, or in a storage medium such as a CD-RW or CD-R, so that when used, the computer (or CPU or MPU) of the system or device reads and executes the program code stored in the storage means or storage medium.

[0079] The processes and techniques described herein are not inherently related to any specific device and can be implemented by a combination of components. Various types of general-purpose devices can also be added. A dedicated device may be constructed to perform the functions of the present embodiment. Various functions can also be formed by appropriately combining multiple components disclosed in the present embodiment. For example, some components may be omitted from all the components shown in the embodiment, or different components may be appropriately combined.

[0080] Although specific embodiments are described in this disclosure, they are in all respects for the purpose of explanation (understanding the technology of the present disclosure) and not for the purpose of limitation. Those skilled in the art will recognize that there are many combinations of hardware, software, and firmware suitable for implementing the technology of the present disclosure. For example, the software described can be implemented in a wide variety of programming or scripting languages, such as assembler, C / C++, Perl, Shell, PHP, Java (registered trademark), etc.

[0081] Furthermore, in the above-described embodiment, the control lines and information lines are those that are considered necessary for the explanation, and not all control lines and information lines in the product are necessarily shown. All components may be interconnected.

[0082] In addition, other implementations of the present disclosure will be apparent to those skilled in the art from consideration of the present embodiments and examples. The specification and examples are exemplary only, with the scope and spirit of the present disclosure being indicated by the following claims.

[0083] 10 Metabolic network processing system 11 Control device 12 Storage device 13 Input device 14 Output device 301 Metabolic network input screen 302 Target substance input screen 303 Target yield input screen 304 Modified enzyme proposal screen 305 Biological experiment result input screen

Claims

1. A metabolic network processing system that extracts and proposes modified enzymes that achieve a target yield of a target substance in a metabolic network, comprising: a storage device that holds a metabolic network processing program that extracts and outputs the modified enzymes; and a control device that processes the metabolic network in accordance with the metabolic network processing program read from the storage device and extracts the modified enzymes, wherein the control device executes the following processes: a process of generating N sensitivity matrices using a stoichiometric matrix obtained by converting the metabolic network, a basis c of the kernel of the stoichiometric matrix, and a given probability distribution; a process of calculating the confidence of each of the matrix elements of the N sensitivity matrices by checking the sign of each matrix element and calculating the ratio of the sign for each matrix element; a process of extracting k matrix elements with high confidence as modified enzymes from the confidence levels for the number of matrix elements of the sensitivity matrix; and a process of outputting the extracted k modified enzymes.

2. A metabolic network processing system as claimed in claim 1, wherein, when generating the N sensitivity matrices, the control device (i) generates a matrix r by replacing negative matrix elements of the stoichiometric matrix with sampling values ​​of the probability distribution, (ii) generates a matrix r' by replacing matrix elements of the matrix r that have positive values ​​in the stoichiometric matrix with zero values, (iii) generates an A matrix by arranging the matrix r' and the basis c, and (iv) generates the sensitivity matrix by taking the inverse matrix of the A matrix and inverting the sign.

3. A metabolic network processing system as claimed in claim 1, wherein the control device further executes the steps of: comparing the yield of the target substance obtained as a result of a biological experiment using the modified enzyme with the yield of the target substance in the metabolic network before modification; and outputting the modified enzyme with an improved yield as a final proposal as a result of the comparison process.

4. A metabolic network processing system according to claim 3, wherein the control device outputs the yield-improving modified enzyme number condition as a final proposal.

5. A metabolic network processing system according to claim 3, wherein the control device executes a sequential Bayesian estimation process and executes a process of updating the probability distribution when the condition for the number of modified enzymes for which the yield is improved is not satisfied.

6. A metabolic network processing system as claimed in claim 5, wherein the control device executes the following processes: generating N new sensitivity matrices using the stoichiometric matrix, the basis c, and the updated probability distribution; calculating a confidence level for each matrix element in the newly generated N sensitivity matrices by checking the sign of each matrix element of the newly generated N sensitivity matrices and calculating the proportion of the sign in each matrix element; extracting k matrix elements with high confidence levels as modified enzymes from the confidence levels corresponding to the number of matrix elements in the newly generated N sensitivity matrices; and outputting the newly extracted k modified enzymes.

7. The metabolic network processing system according to claim 1, wherein the given probability distribution is any one of a log-normal distribution, a mixed log-normal distribution, and a gamma distribution.

8. A metabolic network processing method for extracting and proposing modified enzymes that achieve a target yield of a target substance in a metabolic network, comprising: a control device that processes the metabolic network and extracts the modified enzymes in accordance with a metabolic network processing program read from a storage device, the control device generating N sensitivity matrices using a stoichiometric matrix obtained by converting the metabolic network, a basis c of the kernel of the stoichiometric matrix, and a given probability distribution; the control device calculating the confidence of each of the matrix elements of the N sensitivity matrices by checking the sign of each matrix element and calculating the proportion of the sign in each matrix element; the control device extracting k matrix elements with high confidence as modified enzymes from the confidence levels corresponding to the number of matrix elements of the sensitivity matrix; and the control device outputting the k extracted modified enzymes.

9. A metabolic network processing method as claimed in claim 8, wherein, when generating the N sensitivity matrices, the control device (i) generates a matrix r by replacing negative matrix elements of the stoichiometric matrix with sampling values ​​of the probability distribution, (ii) generates a matrix r' by replacing matrix elements of the matrix r that have positive values ​​in the stoichiometric matrix with zero values, (iii) generates an A matrix by arranging the matrix r' and the basis c, and (iv) generates the sensitivity matrix by taking the inverse matrix of the A matrix and inverting the sign.

10. The metabolic network processing method according to claim 8, further comprising: the control device comparing the yield of the target substance obtained as a result of a biological experiment using the modified enzyme with the yield of the target substance in the metabolic network before modification; and the control device outputting, as a final proposal, the modified enzyme that has improved the yield as a result of the comparison process.

11. The metabolic network processing method according to claim 10, further comprising the step of: when the yield-improving modified enzyme number condition is satisfied, the control device outputs the result as a final proposal.

12. The metabolic network processing method according to claim 10, further comprising the step of: when the condition for the number of modified enzymes for which the yield is improved is not satisfied, the control device executes a sequential Bayesian estimation process to update the probability distribution.

13. The metabolic network processing method according to claim 12, further comprising: the control device generating N new sensitivity matrices using the stoichiometric matrix, the basis c, and the updated probability distribution; the control device calculating a certainty of each matrix element in the newly generated N sensitivity matrices by checking the sign of each matrix element of the newly generated N sensitivity matrices and calculating the proportion of the sign in each matrix element; the control device newly extracting k matrix elements with high certainty as modified enzymes from the certainty levels corresponding to the number of matrix elements of the newly generated N sensitivity matrices; and the control device outputting the newly extracted k modified enzymes.

14. The metabolic network processing method according to claim 8, wherein the given probability distribution is any one of a log-normal distribution, a mixed log-normal distribution, and a gamma distribution.

Citation Information

Patent Citations

  • Information analysis system, method, and program

    JP2021016325A

  • Whole-cell manipulation using real-time metabolic flux analysis

    JP2005506840A