CN110767261A - A method to automate the construction of high-precision genome-scale metabolic network models - Google Patents

Abstract

The invention belongs to the field of systems biology, in particular to a method for automatically constructing a high-precision genome scale metabolic network model. Based on the genomic information of target microorganisms, this patent uses the API provided by KEGG to realize the automatic construction of the "gene-enzyme-reaction-compound" database, and automatically refine and calibrate the reaction matrix through the transcriptomic and proteomic data database, so as to quickly to construct an accurate genome-scale metabolic network model. The simulation of the experiment through the model provided by this patent can effectively reduce the workload and provide a reference for the wet experiment, which can be used for the simulation screening of key genes, the adaptation evaluation of exogenous genes and chassis organisms, and the mining of key metabolic pathways. , minimal genomic and the study of the stress mechanisms of microorganisms in response to external stimuli.

Description

A method to automate the construction of high-precision genome-scale metabolic network models

Technical field:

The invention relates to a method for constructing a metabolic network model, in particular to a method for automatically constructing a high-precision genome scale metabolic network model, which belongs to the field of systems biology.

Background technique:

Cellular metabolism is a complex network of thousands of metabolites, enzymes, and regulators. The complex composition and regulatory mechanisms of this network make it very difficult to engineer metabolic systems and rationally design microbial factories. How to accurately capture the actual metabolic behavior of cells from the countless possible metabolic patterns has become one of the core challenges in the field of microbiology.

One of the effective ways to solve the above problems is to construct a metabolic flux model. This method uses a stoichiometric matrix to represent the intracellular response under the quasi-steady state assumption. The specific method of the model is: assuming that the concentration of intracellular intermediate metabolites remains unchanged for a certain period of time, according to the measurement relationship of each reaction in the metabolic pathway and the substrate consumption rate or product generation rate measured in the experiment, based on mass balance and possible energy balance to determine unknown reaction rates and, in turn, flux allocations to metabolic networks. As one of the important parts of systems biology, genome-scale metabolic network models have been successfully applied in the fields of strain modification, microbial metabolic behavior prediction, and fermentation process monitoring.

The current metabolic flux modeling data mainly comes from three aspects: ① network model databases, such as KEGG, NCBI, Biomodel, etc.; ② existing network models in the literature; ③ based on existing metabolic models of other species Modification: In recent years, with the research of omics technology, the construction of metabolic network based on the actual genome sequence information of microorganisms has become a hot topic. For example, CN 103729576 B, CN 103276011 B, CN 102629304 B, CN 103279689 B and the like. However, the response of microbial cells to the environment is a very complex process, involving multi-scale biochemical reactions such as gene transcription, protein expression, and enzyme activity regulation; and it dynamically adjusts its metabolic behavior according to different environmental stimuli. Genomic information can only reflect the biochemical reactions of microorganisms, but cannot accurately reflect the reactions that microorganisms actually participate in under certain conditions. This modeling method, which ignores the diversity and dynamics of microorganisms, makes methods based solely on databases, literature data, or omics data unable to accurately reflect the specific processes that actually occur in cells.

In addition, the current genomic metabolic network uses manual methods to query and simplify data, and build a complete metabolic flux model, which often requires thousands of database comparisons, which consumes a lot of time and energy costs, which severely limits the Efficiency of model building. There is no effective automation solution yet. Therefore, developing new and automated methods for constructing high-quality models has become an urgent problem to be solved in related fields.

The API (Application Programming Interface) of the KEGG database provides technical support for the automated construction of genome-scale metabolic network models. KEGG (Kyoto Encyclopedia of Genes and Genomes) is currently the most authoritative database for genome deciphering. Its main purpose is to fully express and deduce cells and organisms on the computer, allowing computers to use genetic information to make computational inferences on higher-level and more complex cellular activities and organism behaviors. Given a complete set of genes in a chromosome, it can make predictions about the role of protein interaction (interaction) networks in various cellular activities. KEGG's PATHWAY database integrates current knowledge on molecular interaction networks (eg channels, consortia), KEGG's GENES/SSDB/KO database provides relevant knowledge about genes and proteins discovered in the Genome Project, KEGG's COMPOUND/GLYCAN/REACTION The database provides knowledge of biochemical complexes and reactions. The specific API call method KEGG official description: https://www.kegg.jp/kegg/rest/keggapi.html.

This patent will use the API provided by KEGG to realize the automatic construction of the "gene-enzyme-reaction-compound" database based on the genome information of the target microorganism, and automatically refine and calibrate the reaction matrix through the transcriptomic and proteomic data database, thereby Quickly build accurate genome-scale metabolic network models.

Invention content:

The purpose of the present invention is to provide a method for automatically constructing a high-precision genome-scale metabolic network model. Compared with the previous method, the method has the characteristics of higher efficiency, higher precision and better integrity. The specific steps are as follows:

(1) Genome sequencing: Sequence the genome of the target microorganism to obtain all the genetic information in the microorganism genome;

Described gene information includes its gene function annotation, and corresponding KEGG gene number;

(2) Establishment of the reaction network database: use the API interface provided by the KEGG database to obtain the genes, enzymes, compounds, reactions and their associated information of the "genus" of the target microorganism, and build a database;

Further, the construction of the above-mentioned database is to use the matlab language to write a program, and is automatically built according to the acquired associated information;

Then, according to the genome sequencing results of step (1), the above database is revised, and the genes, proteins, compounds and reaction information not included in the database are added, and redundant data information is deleted at the same time, and the genes of the metabolic network actually contained in the target microorganism are simplified- Enzyme-reaction-compound data information database (see Figure 2 for an example of an infographic);

Further, when a certain gene in the step (1) does not exist in the above-mentioned database, this gene sequence is compared in the KEGG database with the BLAST algorithm, and the highest similarity gene is selected from the obtained comparison result and its KO number is counted. , obtain the reaction number by comparing the database of KO and enzyme and enzyme and reaction in the KEGG database, and add it into the database;

Furthermore, using python language to write programs can automatically compare genes one by one with BLAST algorithm in KEGG database and count KO numbers; using VBA language to write programs can realize the mutual conversion of the above KO and enzyme and enzyme and reaction data. ;

Further, when a certain gene existing in the database does not exist in the detection result of step (1), the gene and the corresponding reaction are deleted from the database.

(3) Construction of the model: According to the data information obtained in step (2), the reaction is taken as the horizontal row, the compound is taken as the vertical row, and the coefficient of the compound in the reaction is taken as the numerical value (consumption is negative, synthesis is positive), Construct a metabolic flux matrix, that is, a metabolic model. This process can be automated with the help of languages such as Python, Matlab, Java, and C++ (see Figure 3 for an example of a flux matrix model);

(4) Calibration of the model: Measure the transcriptome and proteome data of the target microorganism at a specific time node, and compare the measured proteome and transcriptome data with the metabolic model of step (3) by designing a program to compare the transcriptome. Supplementing data that exists in the group and proteome but not in the model, and deleting data that is not available in the transcriptome and proteome but exists in the model, the resulting model can accurately describe the transient pathway behavior in cells.

Further, the proteomic and transcriptomic data are converted into the corresponding reaction equations in the KEGG information database and compared with the metabolic matrix, and the data that exists in the transcriptome and proteome but not in the model is supplemented. Data that are not present on the group and proteome but are present on the response matrix are deleted.

Beneficial effects:

At present, the genome metabolism network uses manual methods to query and simplify data. Generally, it takes about half a year to a year to construct a genome-wide scale network from scratch. The method provided by the present invention uses the API format provided by the KEGG database to obtain data. , the data capture can be completed within ten minutes, which greatly saves time and cost, and improves work efficiency and accuracy.

In this patent, after the model is constructed, the transcriptome and proteome data of a specific time node are determined by measurement, and the metabolic model is corrected, so as to accurately obtain the transient pathway behavior in the cell in turn.

By accurately grasping the transient pathway behavior in cells, this patent can be applied to the simulation screening of key genes, the evaluation of the fitness of exogenous genes and chassis organisms, the mining of key metabolic pathways, the minimal genome and the response of microorganisms to external stimuli. research on the mechanism of stimulation. By simulating the experiment through the model provided by this patent, the workload can be effectively reduced and a reference basis for the wet experiment can be provided.

Description of drawings:

Fig. 1 is the working flow chart of this patent to construct the metabolic network model;

Figure 2 is a schematic diagram of biochemical information obtained through API;

Figure 3 Schematic diagram of the flux matrix model.

Detailed ways:

In order to make the purpose, technical solutions and advantages of the present patent more clear, the present patent will be further described in detail below with reference to specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present patent, but not to limit the present invention.

One of the features of this patent is the establishment of a database integrating the biochemical reaction data of target microorganisms on KEGG through the central dogma. First, all the genes of the target strain are obtained from the KEGG database through the API interface, then the corresponding relationship between all the genes of the target microorganism and the enzymes is obtained according to the genes, and finally the specific reactions catalyzed by the enzymes and the compounds involved in the specific reactions are obtained according to the enzymes. A database of biochemical information of target strains from genes to enzymes to reactions and compounds. Since the KEGG database officially classifies homologous genes with similar functions into one category and assigns the corresponding KEGG Orthology number (KO number for short), this patent constructs a structure from the KO number to which the gene belongs to the corresponding enzyme to the corresponding exclusive reaction and compound. another route and added it to the database. So far, the biochemical information database is divided into two routes according to the data transformation or acquisition method, namely: 1) The biochemical reaction of the target microorganism is obtained by using the gene as the starting point through the corresponding relationship between the gene and the enzyme, the enzyme and the reaction, and the reaction and the compound. equation. 2) Using the KO number as the starting point, obtain a specific reaction equation through the correspondence between the KO number and the enzyme, the enzyme and the reaction, and the reaction and the compound.

The gene, protein, compound, reaction information involved in the process of constructing the metabolic network model of the present invention, as well as the transformation, conversion or correspondence between them all come from the following five database information in KEGG, and the specific data acquisition and The transformation can be realized automatically by programming the VBA language:

(1) The download interface of genetic data is: http://rest.kegg.jp/list/<org>;

(2) The data download interface of the enzyme-gene correspondence is: http://rest.kegg.jp/link/<org>/ec ;

(3) The data download interface of the enzyme-reaction correspondence is: http://rest.kegg.jp/link/ec/rn ;

(4) The data download interface of the reaction-compound correspondence is: http://rest.kegg.jp/link/rn/cpd ;

(5) The data download interface for the correspondence between KO numbers and enzymes is: http://rest.kegg.jp/link/ec/ko.

The method flow of the present invention for constructing a high-precision genome-scale metabolic network model is shown in FIG. 1 , and the present invention will be further explained below through specific embodiments.

Example 1 A method for automatically constructing a high-precision genome-scale metabolic network model of Clostridium acetobutylicum

Taking Clostridium acetobutylicum ATCC824 as the research object, a high-precision genome-scale metabolic network model was constructed. The specific construction process is as follows:

1) Sequence the gene of Clostridium acetobutylicum ATCC824 stored in the laboratory to obtain 2779 genes, record the gene function annotation, and the corresponding KEGG gene number; 2) Download the gene, enzyme- Gene correspondence, enzyme-reaction correspondence, and compound-reaction general information. The download interface of the gene data is: http://rest.kegg.jp/list/cac ; the download interface of the enzyme-reaction correspondence is: http://rest.kegg.jp/link/ec/rn ; the enzyme- The gene correspondence is: http://rest.kegg.jp/link/cac/ec ; The reaction-compound correspondence is: http://rest.kegg.jp/link/rn/cpd ; Using the above information, use A program written in matlab language automatically builds a database of all genes, proteins, compounds, and reaction information of Clostridium acetobutylicum. The database contains 1131 compounds, 2838 genes and 1114 reactions. 2) Add and delete the database information by comparing the genome sequencing results, for example, the bofA gene does not exist in the above database, but the whole genome sequencing results show that the bofA gene does exist in the target strain. Then use the python language to write a program to compare the bofA gene sequence in the KEGG database with the BLAST algorithm, then select the gene with the highest similarity from the comparison results and count its KO number K06317, and use the VBA language to write a program to compare the enzyme and KO. And the database of enzymes and reactions (the download interface for the correspondence between KO numbers and enzymes is: http://rest.kegg.jp/link/ec/ko ) to obtain the reaction rn: R10128 , and add it into the database. However, genes such as kdgK existing in the database have not been detected, and the gene and the corresponding reaction will be deleted from the database. The whole adjustment process deletes 69 genes that the target strain does not have in the database, adds 10 genes unique to the target strain, and updates the biochemical information data of the metabolic network actually contained in the target microorganism according to the gene annotation function. The model contains 1032 compounds, 2779 genes and 1065 responses. 3) According to the gene-enzyme-reaction-compound data information obtained by streamlining, and taking the reaction as the horizontal row, the compound as the vertical row, and the coefficient of the compound in the reaction as the value (consumption is negative, production is positive), construct a metabolic pathway. quantity matrix. 4) When the node (72h) of Clostridium acetobutylicum ATCC824 for high-speed production of butanol from acetone butanol was performed, the transcriptome and proteome were detected, and it was found that 680 enzymes and 2604 genes were involved in intracellular metabolism at this node. The proteomic and transcriptomic data are compared with the metabolic matrix after the biochemical information database is converted into the corresponding reaction equation, and the data that exists in the transcriptome and proteome but not in the model is supplemented. Delete the data that does not exist but exists in the reaction matrix, for example: count all the detected enzyme numbers according to the proteomics results, use the VBA language to write the program, get the reaction according to the database corresponding to the enzyme and the reaction, and add the occurrences in the protein group. But the reactions that do not appear in the reaction matrix, such as Rn00356 etc. Responses that only appeared in the response matrix but could not be obtained according to the transcriptome were deleted from the response matrix, such as Rn20226 and so on. A total of 69 responses were added to its response database and 209 responses were decreased. The dynamic metabolic network involved in the 72h high-yield solvent period was finally determined to contain 978 compounds, 2604 genes and 886 reactions. Metabolic flux calculations accurately describe the 72-h transient pathway behavior in cells.

Example 2 A method for automatically constructing a high-precision genome-scale metabolic network model of Escherichia coli

Taking Escherichia coli JM109 as the research object, a high-precision genome-scale metabolic network was constructed. The specific construction process is as follows:

1) Sequence the gene of Escherichia coli JM109 stored in the laboratory, and find that there are 2279 genes in JM109, record the gene function annotation, and the corresponding KEGG gene number; 2) Download the gene of the genus JM109 belongs to in the KEGG database through API , enzyme-gene correspondence, enzyme-reaction correspondence, and compound-reaction general information. The download interface of the gene data is: http://rest.kegg.jp/list/eco ; the download interface of the enzyme-reaction correspondence is: http://rest.kegg.jp/link/ec/rn ; the enzyme- The gene correspondence is: http://rest.kegg.jp/link/ eco/ec ; the reaction-compound correspondence is: http://rest.kegg.jp/link/rn/cpd ; Using the above information, use A program written in matlab language automatically builds a database of all genes, proteins, compounds and reaction information of the genus Escherichia coli JM109 belongs to. The database contains 915 enzymes, 1528 compounds, 3410 genes and 1656 reactions. The database was added and deleted by comparing the results of genome sequencing. For example, the gfcD gene did not exist in the database of reaction information, but the results of whole genome sequencing showed that the gfcD gene did exist in the target strain. A program written in python language was used to compare the gfcD gene sequence in the KEGG database with the BLAST algorithm, and then the gene with the highest similarity was selected from the comparison results and its KO number K02377 was counted. The program was written in the VBA language to compare the enzyme with KO and Reaction rn: R04128 was obtained from the database of enzymes and reactions (the download interface for the correspondence between KO-enzymes: http://rest.kegg.jp/link/ec/ko ), and added it into the database. However, the torR and other genes existing in the database were not detected, and the torR gene and the corresponding reaction were deleted from the database. Delete 1251 genes that the target strain does not have in the database, add 120 genes unique to the target strain, and update the biochemical information data of the metabolic network actually contained in the target microorganism according to the gene annotation function. The model contains 873 enzymes, 1325 compounds, 2779 genes and 1324 responses. 3) According to the gene-reaction-compound data information obtained by streamlining, and the reaction is a horizontal row, the compound is a vertical row, and the coefficient of the compound in the reaction is a numerical value (consumption is negative, generation is positive), and a metabolic flux matrix is constructed. 4) Transcriptome and proteome detection were performed at the node (8h) when E. coli fermentation entered the logarithmic phase, and the proteome and transcriptome data were converted into corresponding reaction equations in the biochemical information database and compared with the metabolic matrix. The data that exists in the transcriptome and proteome but not in the model is supplemented, and the data that is not in the transcriptome and proteome but exists in the reaction matrix is deleted, for example, all detected enzyme numbers are counted according to the proteomic results , using VBA language to write programs, according to the corresponding relationship between enzymes and reactions to get reactions in the database, add reactions that appear in the proteome but not in the reaction matrix, such as Rn02346 and so on. Responses that only appeared in the response matrix but were not available according to the transcriptome were removed from the response matrix, such as Rn22048. It was found that there are 543 enzymes and 2480 genes involved in intracellular metabolism at this node. An update to its response database added a total of 50 responses and decreased 377 responses. The dynamic metabolic network in log phase contained 1302 compounds, 2480 genes and 997 responses. Metabolic flux calculations accurately describe 8-h transient pathway behavior in cells.

Example 3 Method for automatically constructing a high-precision genome-scale metabolic network model of Saccharomyces cerevisiae

Using Saccharomyces Cerevisiae W3a as the research object, a high-precision genome-scale metabolic network was constructed. The specific construction process is as follows:

1) Sequence the gene of Saccharomyces Cerevisiae W3a stored in the laboratory to obtain 2989 genes, record the gene function annotation, and the corresponding KEGG gene number; 2) Download the genes and enzymes of the genus W3a belongs to from the KEGG database through API -Gene correspondence, enzyme-reaction correspondence, and compound-reaction general information. The download interface of the gene data is: http://rest.kegg.jp/list/sce ; the download interface of the enzyme-reaction correspondence is: http://rest.kegg.jp/link/ec/rn ; the enzyme- The gene correspondence is : http://rest.kegg.jp/link/sce/ec ; The reaction-compound correspondence is: http://rest.kegg.jp/link/rn/cpd ; Using the above information, use A program written in matlab language automatically builds a database of all genes, proteins, compounds and reaction information of the genus Saccharomyces cerevisiae W3a belongs to. The database contains 697 enzymes, 1780 compounds, 3221 genes and 1390 reactions. The database information was added and deleted by comparing the genome sequencing results. For example, the ychD gene did not exist in the database of the reaction information, but the whole genome sequencing results showed that the ychD gene did exist in the target strain. Then use the python language to write a program to compare the ychD gene sequence in the KEGG database with the BLAST algorithm, then select the gene with the highest similarity from the comparison results and count its KO number K02508, and use the VBA language to write a program to compare the enzyme and KO. And the database of enzymes and reactions (the KO-enzyme correspondence download interface is: http://rest.kegg.jp/link/ec/ko ) to obtain the reaction rn:R04406, and add it into the database. However, the tcrP and other genes present in the database were not detected, and they and their corresponding reactions were deleted from the database. Delete 250 genes that the target strain does not have in the database, add 18 genes unique to the target strain, and update the biochemical information data of the metabolic network actually contained in the target microorganism according to the gene annotation function to obtain the biochemical information of the metabolic network actually contained in the target microorganism Data, the model contains 578 enzymes, 1525 compounds, 2989 genes and 1224 reactions. 3) According to the gene-reaction-compound data information obtained by streamlining, and taking the reaction as the horizontal row, the compound as the vertical row, and the coefficient of the compound in the reaction as the value (consumption is negative, generation is positive), construct a metabolic flux matrix . 4) The transcriptome and proteome were detected at the high-speed butanol-producing node of alcohol (10h). It was found that 482 enzymes and 2694 genes were involved in intracellular metabolism at this node. The proteome and transcriptome data were included in the biochemical information. After the database is converted into the corresponding reaction equation, it is compared with the metabolic matrix, the data that exists in the transcriptome and proteome but not in the model is supplemented, and the data that is not in the transcriptome and proteome but exists in the model is deleted, For example: count all the detected enzymes according to the proteomics results, use the VBA language to write the program, get the reactions in the database according to the corresponding relationship between the enzymes and the reactions, and add the reactions that appear in the proteome but do not appear in the reaction matrix, such as Rn02886 Wait. Responses that only appeared in the response matrix but were not available according to the transcriptome were removed from the response matrix, such as Rn28672. The tuning results added a total of 82 reactions and decreased 219 reactions to its reaction database, and the dynamic metabolic network involved in the 10h ethanol production period contained 1236 compounds, 2694 genes and 1087 reactions. Metabolic flux calculations accurately describe the 10-h transient pathway behavior in cells.

The above-mentioned embodiments only represent several embodiments of the present invention, and the descriptions thereof are relatively specific and detailed, but should not be construed as a limitation on the scope of the patent. It should be noted that, for those skilled in the art, without departing from the concept of the present patent, the above-mentioned embodiments can also be modified, combined and improved, which all belong to the protection scope of the present patent. Therefore, the scope of protection of this patent should be subject to the claims.

Claims (7)

1. a method for automatically constructing a high-precision genome scale metabolic network model, is characterized in that, concrete steps are as follows: (1) Genome sequencing: Sequence the genome of the target microorganism to obtain all the genetic information in the microorganism genome; (2) Establishment of the reaction network database: use the API interface provided by the KEGG database to obtain the genes, enzymes, compounds, reactions and their associated information involved in the "genus" of the target microorganism, and build a database; According to the genome sequencing results of step (1), the above database is revised, the genes, proteins, compounds and reaction information not included in the database are added, and redundant data information is deleted to simplify the gene-enzyme- of the metabolic network actually contained in the target microorganism. Reaction-Compound Data Information Database; (3) Model construction: According to the data information obtained in step (2), and take the reaction as the horizontal row, the compound as the vertical row, and the coefficient of the compound in the reaction as the numerical value, construct the metabolic flux matrix, that is, the metabolic model; (4) Model calibration: measure the transcriptome and proteome data of the target microorganism at a specific time node, compare the measured proteome and transcriptomic data with the metabolic model of step (3), and compare the transcriptome and protein The data that exists on the group but not on the model is supplemented, and the data that does not exist on the transcriptome and proteome but exists on the model is deleted, and the obtained model can accurately describe the transient pathway behavior in the cell. 2 . The method for automatically constructing a high-precision genome-scale metabolic network model according to claim 1 , wherein the gene information in step (1) includes gene function annotation and the corresponding KEGG gene number. 3 . 3. a kind of method for automatically constructing a high-precision genome scale metabolic network model as claimed in claim 1, is characterized in that, the database in step (2) is written program by matlab language, and the acquired gene, enzyme, compound, The associated information of the reaction is automatically constructed. 4. The method for automatically constructing a high-precision genome scale metabolic network model according to claim 1, wherein in step (2), when a certain gene in step (1) does not exist in the database, The gene sequence was compared with the BLAST algorithm in the KEGG database, and the gene with the highest similarity was selected from the comparison results and its KO number was counted, and the reaction number was obtained by comparing the KO and enzyme and enzyme and reaction databases in the KEGG database, and add it to the database. 5. The method for automatically constructing a high-precision genome-scale metabolic network model according to claim 1, wherein in step (2), when a certain gene existing in the database does not exist in the same manner as in step (1) When the genome detection result is in, the gene and the corresponding reaction are deleted from the database. 6. The method for automatically constructing a high-precision genome scale metabolic network model as claimed in claim 1, wherein in step (3), the construction process of the metabolic flux matrix is by means of Python, Matlab, Java or C++ language Automation is done. 7. The method for automatically constructing a high-precision genome scale metabolic network model according to claim 1, wherein in step (4), the data of proteome and transcriptomics are transformed with the help of the information of the KEGG database In order to compare the corresponding reaction equation with the metabolic matrix, supplement the data that exists in the transcriptome and proteome but not in the model, and delete the data that is not in the transcriptome and proteome but exists in the response matrix.

Images