A metabolic flux calculation tool, metabolic flux analysis method and application

Abstract

The application belongs to the technical field of biological information, and provides a metabolic flux calculation tool, a metabolic flux analysis method and application, and belongs to the technical field of biological information.The calculation steps of the metabolic flux calculation tool include: using the screened solver and the metabolic network model to fit the mass distribution vector of the labeled metabolite, taking the minimization of the residual sum of squares as the calculation endpoint, and realizing spatial metabolic flux calculation.The metabolic flux calculation tool and the analysis method realize the identification of labeled metabolites and the calculation of metabolic flux based on mass spectrometry imaging.By introducing a stable isotope tracer in vivo and in vitro, and simultaneously detecting the tissue slices by imaging technology, the metabolic flux calculation tool can quickly and accurately identify the labeled metabolites, realize the tracking of the metabolic pathway in vivo, and quantitatively describe the metabolic activity of the tissue, thereby providing a powerful analysis tool for studying the metabolic activity of tissue heterogeneity.

Description

Technical Field

[0001] This invention belongs to the field of bioinformatics technology, and in particular relates to a metabolic flux calculation tool, a metabolic flux analysis method and its application. Background Technology

[0002] Tissues and organs are composed of various cell types, and differentiated cells exhibit specific metabolic characteristics. Metabolic heterogeneity has become an important aspect of complex tissue physiology. Mass spectrometry imaging (MSI) can simultaneously detect changes in hundreds of in-situ metabolites at spatially resolved scales. Numerous reports have demonstrated the ability of MSI-based spatial metabolomics to interpret cellular metabolic states and their heterogeneity. However, metabolite levels reflect the final result of a series of metabolic reactions involved in their synthesis and consumption. Therefore, spatial metabolomics mainly focuses on the ion abundance and spatial distribution of metabolites, but lacks research on the activity and dynamic changes of metabolic pathways.

[0003] Accurate quantification of metabolic fluxes in biological systems is crucial for guiding metabolic needs, biotechnology, human health, and cell culture. To date, the analysis of metabolic fluxes in tissues has primarily relied on liquid chromatography-mass spectrometry (LC-MS), a technique that has yielded many groundbreaking discoveries. For example, Rabinowitz et al. discovered that the metabolic waste product lactate is a major substrate of the tricarboxylic acid (TCA) cycle in most tissues and tumors. Further research has shown that TCA cycle fluxes are suppressed in various solid tumors, with lower ATP synthesis rates than in normal tissues. Fu et al. used multi-organ basic metabolite units to simulate and quantify inter-tissue metabolic exchanges, finding that tissues maintain nucleotide homeostasis through blood exchange rather than synthesis. However, LC-MS-based metabolic flux analysis often loses information about tissue metabolic heterogeneity due to homogenization during sample pretreatment. Therefore, MSI combined with stable isotope tracing is increasingly used to study tissue metabolic activity at a spatially resolved level. For instance, Wang et al. revealed metabolic heterogeneity in rat kidney and brain tissue by combining matrix-assisted laser desorption / ionization (MALDI)-MSI with in vivo stable isotope tracing. Gary et al. used MALDI-MSI and isotope spectroscopy to assess the fractional fluxes of fatty acid synthesis and prolongation in the brains of mice with glioblastoma. While these studies targeted metabolites and metabolic pathways, they fall short in quantifying spatial metabolic fluxes in biological tissues. Therefore, there is an urgent need to develop a widely available tool for quantitative spatial metabolic flux analysis.

[0004] Compared to metabolomics, stable isotope tracing data additionally includes the dimension of isotopic isoforms, making data interpretation more complex. Furthermore, metabolic flux estimation requires three basic inputs: a metabolic network model, substrate uptake, and metabolite labeling, further complicating the calculation. To address this issue, several tools for metabolic flux calculation have been reported, including Metran, 13C-FLUX, 13C-FLUX2, INCA, FreeFlux, OpenMebius, OpenFLUX, and mfapy. However, these tools are primarily designed for LC-MS data and cannot be directly applied to MSI. Moreover, spatial quantitative metabolic flux analysis still faces challenges such as the identification of labeled metabolites and the construction of in vivo metabolic mathematical models. Summary of the Invention

[0005] Therefore, the purpose of this invention is to provide a metabolic flux calculation tool, a metabolic flux analysis method, and an application.

[0006] To achieve the above-mentioned objectives, the present invention provides the following technical solution:

[0007] This invention provides a metabolic flux calculation tool, the calculation steps of which are as follows:

[0008] (1) The m / z and intensity of the labeled metabolites to be tested by MSI are used as MSI data. A database containing labeled metabolites and their isotopic isotopes is used as a database. The labeled metabolites are automatically identified. AccuCor is used to correct the natural isotopes of the identified labeled metabolites and outputs the mass distribution vector of the labeled metabolites.

[0009] (2) Based on the identified labeled metabolites and combined with the chemical reactions in the KEGG database, a metabolic network model was established;

[0010] (3) Based on the metabolic network model and the mass distribution vector of the labeled metabolites, the solver is selected to obtain the selected solver;

[0011] (4) Using the screened solver and metabolic network model, fit the mass distribution vector of the labeled metabolites, and use the minimization of the sum of squared residuals as the calculation endpoint to realize the calculation of spatial metabolic flux.

[0012] Preferably, the method for automatically identifying labeled metabolites includes extracting and matching ion intensity information from MSI data and a database, screening the database multiple times to remove labeled metabolites with poor imaging effects, retaining labeled metabolites visible in MSI, and obtaining the identified labeled metabolites.

[0013] The method for selecting solvers includes preparing a metabolic network model and mass distribution vector data of labeled metabolites, calling the metabolic network model and mass distribution vector data of labeled metabolites, randomly generating initial values ​​of metabolic reaction rates of labeled metabolites, automatically calculating the mass distribution vector 1 of labeled metabolites simulated by the computer based on the initial values, and automatically calculating the RSS value a using the mass distribution vector described in step (1) and the mass distribution vector 1 of the computer simulation; based on the RSS value a, iterating through the solvers, repeating each solver 5 times, and obtaining a fitted metabolic reaction rate after each solver calculation, obtaining the mass distribution vector 2 of labeled metabolites simulated by the computer based on the metabolic reaction rate, automatically calculating and recording the RSS value b based on the mass distribution vector 2 of labeled metabolites simulated by the computer and the mass distribution vector described in step (1) each time, sorting and statistically analyzing the 5 RSS values ​​b, and selecting the median of the 5 RSS values ​​b, and selecting the solvers that are locally optimal and globally optimal based on the median of the RSS values;

[0014] The RSS value b is less than the RSS value a.

[0015] Preferably, the method for removing labeled metabolites with poor imaging performance includes the following steps: for a matched labeled metabolite, if the ion intensity of labeled metabolite M0 is less than a set intensity threshold in more than half of the samples, it is considered that the labeled metabolite cannot form an effective imaging signal in MSI, and therefore the labeled metabolite is removed; wherein labeled metabolite M0 is a labeled metabolite in which the number of C labeled is 0.

[0016] The method for retaining visible ions in MSI includes the following steps: calculating the ratio of the average ion intensity of the labeled metabolites M2 and M3 in the labeled group to the unlabeled group, respectively. If one of the ratios is greater than 10, it indicates that the labeled metabolite has formed a sufficient ion signal in the MSI data, and the labeled metabolite is retained; the labeled metabolite M2 is a labeled metabolite with 2 carbon atoms; the labeled metabolite M3 is a labeled metabolite with 3 carbon atoms.

[0017] Preferably, the solver is SLSQP, COBYLA, LN_COBYLA, LN_BOBYQA, LN_PRAXIS, LN_SBPLX, LN_NELDERMEAD, LN_NEWUOA, GN_CRS2_LM, GN_DIRECT_L, GN_AGS, GN_IRES, and GN_ESCH.

[0018] Preferably, the construction of the database containing labeled metabolites and their isotopic isoforms includes:

[0019] Based on the metabolites included in the metabolic network model, calculate the summation of all metabolites in the metabolic network model using U-... 13 The molecular weight of the isotope isoform after C6-glucose labeling was determined, and a database containing metabolite names, molecular formulas and isotope molecular weights was finally established, resulting in a database containing labeled metabolites and their isotope isoforms.

[0020] The chemical reactions referred to are all central carbon metabolic reactions associated with labeled metabolites in KEGG.

[0021] This invention provides an application of the above-mentioned metabolic flux calculation tool in at least one of the following a to e:

[0022] a. Identify labeled metabolites;

[0023] b. Calculation of metabolic flux;

[0024] c. Research on tissue metabolic regulation mechanisms;

[0025] d. Studies on metabolic activities in tissue heterogeneity;

[0026] e. Quantification of tissue metabolite conversion rate.

[0027] Preferably, the metabolic flux includes spatial metabolic flux or in situ metabolic flux.

[0028] This invention provides a method for analyzing metabolic flux, including the steps of stable isotope tracing, sample preparation, data acquisition, and calculating metabolic flux using the aforementioned metabolic flux calculation tool.

[0029] Preferably, the isotope tracing reagent includes U- 13 C6 glucose solution.

[0030] Preferably, the data acquisition is performed using the AFADESI-MSI system.

[0031] Compared with the prior art, the present invention has the following beneficial effects:

[0032] 1. The metabolic flux calculation tool and analysis method developed in this invention realizes the identification of labeled metabolites based on mass spectrometry imaging (MSI). By introducing stable isotope-labeled tracers in vitro and in vivo, and simultaneously using imaging technology to detect tissue sections, MSIFluxer can rapidly and accurately identify labeled metabolites, enabling the tracking of metabolic pathways in vivo and providing a powerful analytical tool for studying metabolic activities in tissue heterogeneity.

[0033] 2. The spatial metabolic flux calculation tool and analysis method developed in this invention realizes the quantitative calculation of in-situ metabolic flux. Based on the identification of labeled metabolite functions, a corresponding metabolic network model is established. Using the tool MSIFluxer of this invention, the mathematical solver is automatically selected and the metabolic flux is calculated, realizing the calculation of spatial metabolic flux based on MSI, and quantitatively describing the metabolic activity of tissues.

[0034] 3. The spatial metabolic flux calculation tool and analysis method developed in this invention enable the quantitative description of metabolic pathway activities in a spatially resolved manner. This method and tool can quantitatively characterize the metabolite conversion rate in tissue microregions and confirm the direction of reversible reactions, contributing to a comprehensive understanding of the metabolic state of tissue microregions and the study of metabolic regulatory mechanisms in these regions. Attached Figure Description

[0035] Figure 1 A schematic diagram of spatial metabolic flux calculation for MSIFluxer;

[0036] Figure 2 This is a schematic diagram of a spatial metabolic flux analysis method.

[0037] Figure 3 In the middle section, a represents the results of screening for the local optimal solver of LC-MS and AFADESI-MSI, b represents the results of screening for the global optimal solver of LC-MS and AFADESI-MSI, and c represents the comparison of metabolic fluxes of 12 reactions in the TCA cycle calculated by MSIFluxer based on LC-MS and AFADESI-MSI renal cortical data.

[0038] Figure 4 AFADESI-MSI images of isotopic isotopes of 10 labeled metabolites;

[0039] Figure 5 In the figure, a represents the metabolic flux of the renal cortex and renal medulla calculated using MSIFluxer; b represents the chemical reaction results in the metabolic network that show different reaction rates in the renal cortex and medulla; c represents the average enrichment image of glutamine; d represents the average enrichment image of glutamate; e represents the conversion rate between glutamate and glutamine; and f represents the conversion rate of the renal cortex and renal medulla in the reversible reaction of pyruvate and lactate.

[0040] Figure 6In the middle section, a represents the direction and related reaction rate of multiple reversible reactions in the TCA cycle in the cerebral cortex; b represents the reaction rate of ATP production in the cerebral cortex; c represents the reaction flux of glyceraldehyde-3-phosphate and glyceric acid-3-phosphate in the cerebral cortex and renal cortex; d represents the synthesis flux of AcAsp measured in the cerebral cortex and renal cortex; e represents the average enrichment image of AcAsp in the cerebral cortex; and F represents the average enrichment image of AcAsp in the renal cortex.

[0041] Figure 7 In Figure a, the tumor and normal regions in laryngeal cancer tissue were detected by hematoxylin-eosin staining. In Figure b, the AFADESI-MSI imaging of 5-phosphate ribose and related nucleotides M0 and M5 was shown. In Figure c, the metabolic pathway of 5-phosphate ribose incorporation into nucleotide synthesis was shown. In Figure d, the metabolic flux of purine nucleotide salvage synthesis was shown. In Figure e, the flux of NADPH production in normal tissue, tumor tissue, and cerebral cortex was shown. In Figure F, the flux of metabolic enzymes PFK, CS, G6PDH, and GAPDH in normal tissue, tumor tissue, and cerebral cortex was calculated. Detailed Implementation

[0042] This invention provides a metabolic flux calculation tool, the calculation steps of which are as follows:

[0043] (1) The m / z and intensity of the labeled metabolites to be tested by MSI are used as MSI data. A database containing labeled metabolites and their isotopic isotopes is used as a database. The labeled metabolites are automatically identified. AccuCor is used to correct the natural isotopes of the identified labeled metabolites and outputs the mass distribution vector of the labeled metabolites.

[0044] (2) Based on the identified labeled metabolites and combined with the chemical reactions in the KEGG database, a metabolic network model was established;

[0045] (3) Based on the metabolic network model and the mass distribution vector of the labeled metabolites, the solver is selected to obtain the selected solver;

[0046] (4) Using the screened solver and metabolic network model, fit the mass distribution vector of the labeled metabolites, and use the minimization of the sum of squared residuals as the calculation endpoint to realize the calculation of metabolic flux.

[0047] In this invention, the method for automatically identifying labeled metabolites includes extracting and matching ion intensity information from MSI data and a database, performing multiple screenings on the database to remove labeled metabolites with poor imaging effects, and retaining labeled metabolites visible in the MSI, thereby obtaining the identified labeled metabolites. The labeling preferably uses U-... 13C6-glucose labeling. The method for matching ion intensity includes the following steps: after importing the two-dimensional data matrix of m / z and ion intensity, iteratively calculate the error between the measured m / z value and the theoretical m / z value (the theoretical m / z value is the value in the database mentioned above). When the error is less than a set ppm threshold, the theoretical m / z is matched with the ion signal intensity corresponding to the measured m / z. The set ppm threshold can be selected as 5 ppm. The method for removing labeled metabolites with poor imaging performance includes the following steps: for the matched labeled metabolites, if the ion intensity of labeled metabolite M0 is less than the set intensity threshold in more than half of the samples, it is considered that the labeled metabolite cannot form an effective imaging signal in MSI, and therefore the labeled metabolite is removed; the labeled metabolite M0 is a labeled metabolite in which the number of C labeled is 0. The default value of the intensity threshold can be selected as 100. The method for retaining labeled metabolites visible in MSI includes the following steps: calculating the ratio of the average ionic intensity of the labeled metabolite isotopes M2 and M3 in the labeled group to the unlabeled group, respectively. If one of these ratios is greater than 10, it indicates that the labeled metabolite has formed a sufficient ionic signal in the MSI data, and the labeled metabolite is retained. The labeled metabolite isotope M2 is a labeled metabolite with 2 carbon atoms; the labeled metabolite isotope M3 is a labeled metabolite with 3 carbon atoms.

[0048] In this invention, the construction of the database containing labeled metabolites and their isotopic isoforms includes:

[0049] Based on the metabolites included in the metabolic network model, calculate the m / z of the summation of all metabolites in the metabolic network model and its U-value. 13 C6-glucose (abbreviation) 13 C) The molecular weight of the labeled isotopic isomers was determined, and a database containing the metabolite name, molecular formula, and isotopic isomer molecular weight was finally established, resulting in a database containing labeled metabolites and their isotopic isomers. The database containing labeled metabolites and their isotopic isomers is shown in Table 1.

[0050] In this invention, a metabolic network model is established based on the identified labeled metabolites and chemical reactions from the KEGG database. The chemical reactions refer to all central carbon metabolism reactions associated with the labeled metabolites in KEGG. The specific metabolic network model constructed is shown in Table 2.

[0051] In this invention, solvers are selected based on the metabolic network model and the mass distribution vector of the labeled metabolites to obtain the selected solvers. Initial values ​​of the metabolic reaction rates of the labeled metabolites are randomly generated. The mass distribution vector 1 of the labeled metabolites simulated by the computer is automatically calculated based on the initial values. The RSS value a is automatically calculated using the mass distribution vector described in step (1) and the mass distribution vector 1 of the computer simulation. Based on the RSS value a, the solvers are traversed and calculated. Each solver is repeated 5 times. After each calculation, a fitted metabolic reaction rate is obtained. The mass distribution vector 2 of the labeled metabolites simulated by the computer is obtained based on the metabolic reaction rate. The RSS value b is automatically calculated and recorded based on the mass distribution vector 2 of the labeled metabolites simulated by the computer and the mass distribution vector described in step (1) each time. The 5 RSS values ​​b are sorted and statistically analyzed, and the median of the 5 RSS values ​​b is selected. The solvers that are locally optimal and globally optimal (i.e., the selected solvers) are selected based on the median of the RSS values. The RSS value b is less than the RSS value a. During the screening process, if the median of the RSS value b is greater than the RSS value a, the solver is removed, and the remaining solvers are used as candidate solvers. The candidate solver with the smallest median of the RSS value b is selected as the solver for both local and global optima.

[0052] In this invention, the solvers are SLSQP, COBYLA, LN_COBYLA, LN_BOBYQA, LN_PRAXIS, LN_SBPLX, LN_NELDERMEAD, LN_NEWUOA, GN_CRS2_LM, GN_DIRECT_L, GN_AGS, GN_IRES, and GN_ESCH. The selected solvers are preferably divided into locally optimized solvers and globally optimized solvers. The locally optimized solver is LN_BOBYQA, and the globally optimized solver is GN_ESCH.

[0053] In this invention, a filtered solver and metabolic network model are used to fit the mass distribution vector of labeled metabolites, with minimizing the sum of squared residuals as the calculation endpoint to achieve metabolic flux calculation. The fitting method involves standardizing all metabolic reaction fluxes based on glucose uptake rates. Each calculation uses 2000 random seeds and 5 optimal seeds, with global and local algorithms performing 1 million iterations. Finally, the 5 flux solutions with the lowest RSS values ​​are reported.

[0054] This invention provides an application of the above-mentioned metabolic flux calculation tool in at least one of the following a to e:

[0055] a. Identify labeled metabolites;

[0056] b. Calculation of metabolic flux;

[0057] c. Research on tissue metabolic regulation mechanisms;

[0058] d. Studies on metabolic activities in tissue heterogeneity;

[0059] e. Quantification of tissue metabolite conversion rate.

[0060] In this invention, the metabolic flux preferably includes spatial metabolic flux or in situ metabolic flux.

[0061] To address the challenge of calculating spatial metabolic fluxes, this invention proposes an analytical method for such calculations and, by combining airflow-assisted desorption / electrospray ionization-MSI (AFADESI-MSI) with stable isotope tracing, develops an automated tool for calculating spatial metabolic fluxes, MSIFluxer. MSIFluxer can determine metabolite turnover rates, confirm the direction of reversible reactions, and quantitatively describe metabolic pathway activity in a spatially resolved manner. MSIFluxer is used to analyze U- 13 This study investigated the dynamic metabolic changes in tissues and in situ after C6-glucose tracing, quantitatively characterizing the metabolic flux in rat renal cortex, renal medulla, cerebral cortex, and clinical tumor samples. This provides key evidence and a new perspective for a deeper understanding of the mechanisms of metabolic heterogeneity within and between tissues.

[0062] This invention provides a method for analyzing metabolic flux, including the steps of stable isotope tracing, sample preparation and data acquisition, and calculating metabolic flux using the aforementioned spatial metabolic flux calculation tool.

[0063] In this invention, the isotope tracing reagent includes U- 13 C6 glucose solution, the U- 13 The concentration of the C6 glucose solution is 550–570 mM, more preferably 560 mM; the U- 13 The preferred treatment time for the C6 glucose solution is 2 to 3 hours, more preferably 2.5 hours.

[0064] In this invention, the samples include one or more of the following: renal cortex, renal medulla, cerebral cortex, and clinical tumor samples (such as laryngeal cancer). Data acquisition is performed using an AFADESI-MSI system. The parameters of the AFADESI-MSI system are as follows: spray voltage ±7kV, spray gas pressure 0.7MPa, pump flow rate 45L / min, spray solvent system acetonitrile-water solution (8:2, V:V), spray flow rate 6μL / min. The mass spectrometry parameters are as follows: positive and negative ion full scan detection mode, mass spectrometry scan range m / z 70–1000, ion source temperature 350℃, resolution 140000, maximum injection time 100ms, and AGCtarget parameter set to 3e6.

[0065] In this invention, unless otherwise specified, all raw material components are commercially available products well known to those skilled in the art.

[0066] The technical solutions provided by the present invention will be described in detail below with reference to the embodiments, but they should not be construed as limiting the scope of protection of the present invention.

[0067] In the following embodiments, AccuCor is an R package, the code of which can be found at: https: / / github.com / XiaoyangSu / AccuCor.git, and the detailed algorithm can be found in "Su, Xiaoyang, W.Lu, and JDRabinowitz."Metabolite Spectral Accuracy on Orbitraps."Analytical Chemistry(2017):5940-5948.".

[0068] The U- 13 C6 glucose was purchased from Cambridge Isotope Laboratories, catalog number CLM-1396-1.

[0069] The mouse drug administration / blood collection device was purchased from Sinocare Biotechnology (Beijing) Co., Ltd., model number: SK-PK2.

[0070] Example 1

[0071] MSIFluxer, a tool for calculating spatial metabolic flux

[0072] To quantitatively analyze the flux of metabolic pathways in biological tissues, a spatial metabolic flux calculation tool, MSIFluxer, was developed. MSIFluxer has the dual functions of identifying labeled metabolites and calculating metabolic flux.

[0073] The specific steps for calculating spatial metabolic flux using MSIFluxer are as follows:

[0074] (1) The m / z and intensity of the labeled metabolites to be detected by MSI are used as the basic input (also known as MSI data). A database containing labeled metabolites and their isotopic isotopes is used as the database. The first code is run to automatically identify the labeled metabolites. AccuCor is used to correct the natural isotopes of the identified labeled metabolites and output the mass distribution vector (MDV) of the labeled metabolites.

[0075] The first code is:

[0076] A1. Load the required R packages: readxl, accucor, openxlsx, and tidyverse.

[0077] A2. Clear all variables in the current environment.

[0078] A3. Define three functions:

[0079] Extractint(): Used to match mass-to-charge ratio and intensity. The ppm threshold can be adjusted, with a default value of 5 ppm. After importing the two-dimensional data matrix of m / z and ion intensity, it iterates and calculates the error between the measured m / z value and the theoretical m / z value (the theoretical m / z value is the value in the database mentioned above). When the error is less than the set ppm threshold, the theoretical m / z is matched with the ion signal intensity corresponding to the measured m / z.

[0080] Scanint(): Used to filter ions that do not produce good imaging results. The threshold can be adjusted, with a default value of 100. For matched ions, if the ionic intensity of metabolite M0 is less than 100 in more than half of the samples, it is considered that the ion cannot form an effective imaging signal in MSI, and therefore the ion and its corresponding isotope are removed (the threshold "100" can be changed by the user according to the instrument sensitivity).

[0081] Filterint(): Used to retain labeled ions visible to MSI. Calculates the ratio of the average ionic intensity of the isotopic isotopes M2 and M3 of the labeled metabolite to the ratio in the labeled group and the unlabeled group, respectively. If either ratio is greater than 10, it indicates that the metabolite... 13 The C-labeled form generated sufficient ion signal in the MSI data for the preservation of the isotopic isotope group. The threshold of 10 for ratio analysis can also be adjusted based on the applied... 13 The C-tracer type and isotope enrichment level can be customized.

[0082] A4. Import MSI data and Database

[0083] The Extractint() function was used to perform intensity extraction matching on the MSI data and database. Ions with poor imaging performance were filtered using the Scanint() and Filterint() functions. The matching and filtering results were saved as four worksheets, with default names "Extract_Int", "Scan_Int", "Filter_Int", and "Input_Int", and output to the current folder as a workbook named "Result_Int_example.xlsx". The R package "AccuCor" was used to perform 13C natural isotope abundance correction on "Input_Int", and the corrected isotope abundance, label fraction, and the name and molecular formula of the labeled compound were output as a workbook named "Result_Int_example_corrected.xlsx".

[0084] The construction of the database containing labeled metabolites and their isotopic isoforms includes:

[0085] Based on the metabolites included in the metabolic network model, calculate the common additive forms (hydrogenation, dehydrogenation, chlorination, etc.) of all metabolites in the metabolic network model. 13 C-labeled isotopic isomorphs (M1-M) n The molecular weight of the metabolite (where n is the number of labeled C atoms in the labeled metabolite) was used to establish a database containing the metabolite name, molecular formula, and isotope molecular weight (see Table 1).

[0086] Table 1 contains a database of labeled metabolites and their isotopic isoforms.

[0087]

[0088]

[0089]

[0090] (2) Based on the identified labeled metabolites, and in conjunction with the KEGG database, establish the corresponding metabolic network model;

[0091] The model is constructed using chemical reactions of labeled metabolites present in KEGG. Taking Glc as an example (the Chinese name of the labeled compound is glucose, i.e., glucose, according to the labeling compound and isotope database), the central carbon metabolism reactions of glucose in KEGG are searched. For example, the first reaction, kegg:R02848, is the production of G6P (glucose 6-phosphate) from glucose using ATP. The carbon atom transfer pathway in this reaction is listed (ABCDEF→ABCDEF), meaning that 6 carbon atoms of glucose are transferred to the compound glucose 6-phosphate. Similarly, listing all central carbon metabolism reactions related to the labeled metabolites forms a metabolic network model. This metabolic network model includes glycolysis, the TCA cycle, the pentose phosphate pathway, and purine synthesis, presented in tabular form for easy code reading.

[0092] The definition of the metabolic network model is shown in Table 2.

[0093] Table 2 Definition of Metabolic Network Model

[0094]

[0095]

[0096]

[0097]

[0098] Note: In Table 2, the underlined part in the ID-reaction name column is the abbreviation of the metabolic enzyme involved in the reaction; in the chemical reaction column, --> indicates the direction of the reaction; the carbon atom transfer matrix in the chemical reaction column only contains metabolites in which carbon atom conversion occurs, and nd indicates that no chemical change has occurred; in the specific carbon atom transfer pattern column, the letters A to J represent different carbon atoms, and the number of letters represents the number of carbon atoms, and nd indicates that no carbon atom is transferred.

[0099] (3) Based on the metabolic network model and the MDV of the labeled metabolites, the second code is used to screen the solver; and the screened solver and metabolic network model are automatically called to fit the MDV of the labeled metabolites, with the minimum residual sum of squares (RSS) as the calculation endpoint, to finally achieve accurate calculation of metabolic flux. The fitting method is to standardize all metabolic reaction fluxes according to the glucose uptake rate, set 2000 random seeds and 5 optimal optimization seeds for each calculation, and perform 1 million iterations of global and local algorithms, finally reporting the 5 flux solutions with the lowest RSS values.

[0100] The second code is:

[0101] B1. First, solver selection is performed, and then the selected solvers are used to perform flux calculation.

[0102] The results of the filter solver will be stored in five columns: "Method", "RSS1", "RSS2", "RSS3", "RSS4", and "RSS5".

[0103] A total of 100 random initial flux vectors were prepared, with 20 models and 5 repetitions.

[0104] Prepare a metabolic network model;

[0105] B2. Obtain relevant data for the metabolic network model in order to construct the metabolic network model object.

[0106] Information such as reaction, reversibility, metabolites, and target fragments is read from a text file, and then this information is used to construct an mfapy.metabolicmodel.MetabolicModel object, which represents a metabolic network model. The data file of the metabolic network model is stored in the "file_path" text file.

[0107] The main parameters and return values ​​of this metabolic network model are: reactions, reversible, metabolites, and target_fragments; where reactions: a list containing all reactions; reversible: a list of boolean values ​​indicating whether each reaction is reversible; metabolites: a list containing all metabolites; and target_fragments: a list containing all target fragments.

[0108] Use `set_configuration` to set the parameters in the metabolic network model described above;

[0109] `set_configuration` is also a function in the mfapy library. Since `model` is an `mfapy.metabolicmodel.MetabolicModel` object, consider the following code as `mfapy.metabolicmodel.MetabolicModel.set_configuration(...)`.

[0110] 1. Control the detail level of callback functions. During optimization, you can specify the detail level of callback functions to more comprehensively monitor and understand the optimization process;

[0111] 2. The maximum number of iterations is controlled, limiting the maximum number of iterations allowed during optimization to 10,000;

[0112] 3. Configure the model to use 4 local CPUs for parallel computing;

[0113] Load metabolite states from a CSV file to constrain flux vectors in the model;

[0114] Set the isotopic isotope distribution of the carbon source;

[0115] Using the previously obtained isotope ratios, i.e., different metabolites, the isotope distribution is set; this step loads the measured mass distribution vector (MDV) data from the file corresponding to the current file;

[0116] Set the name of the current experiment to 'ex1', set the observed MDV data to mdv_observed1, and set the carbon source to carbon_source1;

[0117] Print a message indicating that parallel processing will be used to generate the initial state.

[0118] Initialize two empty lists, one to store the generated initial flux and the other to store the corresponding residual sum of squares (RSS).

[0119] Start a loop and repeat it n times;

[0120] Generate 10 initial states using parallel processing, of which 10 states will be selected as initial states. Add the generated initial flux `flux_initial` to the `flux_initial_array` list.

[0121] Calculate the residual sum of squares (RSS) corresponding to the generated initial flux and add it to the rss_initial_array list;

[0122] Sort the RSS values ​​in rss_initial_array to obtain temparray. Print some statistical information about the RSS values, including the minimum value, 5th percentile, median, 95th percentile, and maximum value;

[0123] Configure the model by setting the callbacklevel parameter to 0;

[0124] Initialize an empty dictionary to store the RSS values ​​of different optimization methods;

[0125] A list is defined containing the names of various optimization algorithms: "SLSQP", "COBYLA", "LN_COBYLA", "LN_BOBYQA", "LN_PRAXIS", "LN_SBPLX", "LN_NELDERMEAD", "LN_NEWUOA", "GN_CRS2_LM", "GN_DIRECT_L", "GN_AGS", "GN_IRES", and "GN_ESCH". These algorithms will be used in the subsequent optimization process.

[0126] Start a loop that iterates through each optimization algorithm in the previously defined all_methods list;

[0127] Create an empty list in the RSS_dict dictionary for the current optimization algorithm to store the RSS values ​​of the algorithm;

[0128] Start an inner loop and repeat it n times;

[0129] Record the current time to limit the maximum running time of the optimization process;

[0130] Take the j-th initial flux from flux_initial_array as the starting point for the current optimization;

[0131] Start an inner loop that executes a maximum of 1000 times;

[0132] The initial flux flux_opt1 is optimized using the current optimization algorithm method to obtain the optimized state, the best-fit RSS value RSS_bestfit, and the optimized flux flux_opt1.

[0133] If the optimized flux list is empty, then exit the nested loop;

[0134] Calculate the p-value and RSS threshold of the current optimization result to determine whether the optimization has converged;

[0135] If the optimization process has been running for more than 60 seconds, exit the inner loop;

[0136] Add the RSS value obtained by the current optimization algorithm to the RSS_dict[method] list;

[0137] Sort the RSS values ​​in the RSS_dict[method] list;

[0138] Print out some RSS value statistics of the current optimization algorithm, including minimum, 5th percentile, median, 95th percentile and maximum;

[0139] This part of the code adds the RSS value statistics of the current optimization algorithm to a DataFrame, including the minimum value, 5th percentile, median, 95th percentile, and maximum value;

[0140] The pd.concat() function is used to add new rows of data to the results_df DataFrame;

[0141] ignore_index=True means that the original row index is not retained, and the default numeric index is used instead;

[0142] Save the results_df DataFrame to an Excel file named "method_comparison_results.xlsx";

[0143] Read data from the "method_comparison_results.xlsx" file and store it in a dfDataFrame;

[0144] Extract the data from the "Method" column of the df DataFrame and store it in the methods variable;

[0145] The data in the median_results variable is divided into two parts: median_results_top8 contains the first 8 elements; median_results_last5 contains the last 5 elements.

[0146] Print out the names of the local optimum and the global optimum optimization algorithms;

[0147] Write the names of the local and global optimal optimization algorithms to the file "method_comparison_results.txt". If an optimal algorithm does not exist, write the string "none".

[0148] Information such as reaction, reversibility, metabolites, and target fragments is read from a text file. The model represents a metabolic model. The following four configuration functions are used to set different parameters of the mfapy.metabolicmodel.MetabolicModel (i.e., the model above) object to affect subsequent metabolic flux analysis.

[0149] set_configuration is also a function in the mfapy library, because model is an mfapy.metabolicmodel.MetabolicModel object;

[0150] Consider the following code as mfapy.metabolicmodel.MetabolicModel.set_configuration(...);

[0151] Control the level of detail in callback functions. During optimization, you can specify the level of detail for callback functions to more comprehensively monitor and understand the optimization process.

[0152] Controls the maximum number of iterations. This parameter limits the maximum number of iterations allowed during optimization to 10,000;

[0153] The comment in the following line of code explains that when calling the self.fitting_flux method, its parameter method is set to 'deep';

[0154] In this case, the number of iterations is the value of number_of_repeat;

[0155] Specifically, the self.fitting_flux method is a function used for metabolic flux fitting;

[0156] Here, two different optimization methods are used: SLSQP and LN_PRAXIS.

[0157] SLSQP is a sequential least squares quadratic programming method, mainly used to solve constrained optimization problems;

[0158] LN_PRAXIS is a limited-memory Broyden-Fletcher-Goldfarb-Shanno (BFGS) algorithm that combines the Praxis algorithm.

[0159] Furthermore, these two methods are performed alternately multiple times. The total number of such alternating iterations is determined by the `number_of_repeat` parameter.

[0160] In short, this comment describes how the `self.fitting_flux` method is iterated multiple times when using the 'deep' method, with each iteration using two different optimization methods (SLSQP and LN_PRAXIS). The total number of iterations is equal to the value of `number_of_repeat`.

[0161] The comment "SLSQP => LN_PRAXIS" indicates that two different optimization methods are used in the metabolic flux analysis, and these two methods are performed alternately. The arrow "=>" indicates that the methods are executed sequentially.

[0162] Specifically, this means first using the SLSQP method for optimization, then switching to the LN_PRAXIS method, and then switching back to SLSQP, and so on.

[0163] The model has been configured to use four local CPUs for parallel computation.

[0164] Load metabolite states from a CSV file to constrain flux vectors in the model;

[0165] Iterate through the files in the "MDV_MFA / " directory;

[0166] Within the loop, perform metabolic flux analysis on each file, using global and local optimization methods, and then display the results;

[0167] Define the path where the results will be saved. `dest_dir1` is the target directory for storing the results;

[0168] Obtain the C12 and C13 isotope ratios corresponding to the current file from the carbon_ratio DataFrame;

[0169] It was obtained from the file GLC_Normal_KINDEY_Carbon_Source.csv mentioned above;

[0170] Use the generate_carbon_source_template method of the metabolic model to create a carbon source template;

[0171] Use the previously obtained isotope ratios, i.e., different metabolites, to set the isotope distribution; load the measured mass spectrometry distribution vector (MDV) data from the file corresponding to the current file;

[0172] Set up the experiments in the metabolic model by specifying the experiment name ('ex1'), MDV data, and carbon source template;

[0173] Define the global optimization method as "min_median_method2", which is the global optimization algorithm selected earlier;

[0174] The following are actually various optimization algorithms:

[0175] "SLSQP": Sequential Least Squares Quadratic Programming;

[0176] "COBYLA": Constrained optimization using linear approximation;

[0177] "LN_COBYLA": A limited-memory version of COBYLA;

[0178] "LN_BOBYQA": A limited-memory version of BOBYQA;

[0179] "LN_NEWUOA": A type of BFGS method for finite memory;

[0180] "LN_PRAXIS": A limited-memory version of PRAXIS;

[0181] "LN_SBPLX": A limited-memory version of Subplex;

[0182] "LN_NELDERMEAD": A finite-memory version of the Nelder-Mead method;

[0183] "GN_CRS2_LM": An optimization method for global optimization with constraints and sequential quadratic programming with Levenberg-Marquardt modifications;

[0184] The local optimization method is defined as "min_median_method1", which is the local optimization algorithm selected earlier.

[0185] The `generate_initial_states` method of the `model` object generates 2000 initial states, and selects 5 of them for subsequent flux fitting.

[0186] Create a list of results, initially containing a tuple ('template', flux_opt1[0]). This tuple represents the flux value used as the template in the initial state;

[0187] The flux is fitted using a global optimization method (method_global) to obtain the fitted state, the sum of squared residuals (RSS), and the optimal flux value. This is to more coarsely fit the model to find the global optimum.

[0188] Add the results of the global fit to the results list, and iterate 5 times.

[0189] The flux was fitted using a local optimization method;

[0190] The fitted state, the sum of squared residuals, and the optimal flux value are obtained. This is to fit the model more accurately in order to find local optima;

[0191] The results of the local fitting are evaluated, the p-value and the residual sum of squares threshold are calculated, and the results are added to the results list.

[0192] Since the absolute substrate uptake rate is difficult to determine, a glucose uptake rate (100 nmol / g / min) was manually set in the metabolic network model, and the fitted metabolic response rate was expressed as a value relative to the glucose uptake rate.

[0193] A schematic diagram of MSIFluxer's spatial metabolic flux calculation is shown below. Figure 1 .

[0194] Example 2

[0195] A method for spatial metabolic flux analysis

[0196] This method consists of three parts: stable isotope tracing, sample preparation and data acquisition, and metabolic flux calculation.

[0197] (1) In vivo and in vitro stable isotope tracing was performed by jugular vein infusion and in vitro culture.

[0198] The specific protocol for jugular vein infusion is as follows: First, administer 2604 mg U- 13 C6 glucose was dissolved in 25 mL of sterile physiological saline to prepare a 560 mM U- 13 C6 glucose solution (tracer solution). Next, connect the external connecting tube, rotating shaft, spring-loaded inner catheter, vest spring, and reducer connecting tube of the mouse drug delivery / blood collection device. Inject U- into the catheter. 13 C6 glucose injection solution was administered and air bubbles were removed. After weighing the animal (rats or mice), it was anesthetized in an isoflurane anesthesia machine and fitted with a vest. Blood was drawn through a pre-drilled jugular vein catheter using a syringe with a flat-tipped needle, ensuring the blood vessel was patent and not blocked. The animal's jugular vein catheter was then connected to an external large-diameter connecting tube. Next, the rotating shaft with the vest spring was fixed to a rotating support, and the syringe connected to the external connecting tube on the other side of the shaft was fixed to an infusion pump. U- was then continuously infused into the animal at a controlled rate. 13 C6 glucose was administered for 2.5 hours, with continuous monitoring during the infusion to ensure no leakage. After infusion, blood was drawn from the orbital cavity, allowed to stand at room temperature for at least 1 hour, and then centrifuged at 3000 rpm for 10 minutes at 4°C. The supernatant was then collected and stored at -80°C for LC-MS analysis. Immediately after blood collection, the rats were euthanized, rapidly dissected, and the required tissue samples (brain, kidney, liver, etc.) were collected and quenched with liquid nitrogen. The samples were then placed in 6-well plates and stored at -80°C for MSI analysis. In vitro stable isotope tracing was performed using U-... 13 C6 glucose-labeled DMEM medium enables continuous culture of isolated tissues, U- 13 U- in C6 glucose-labeled DMEM medium 13The concentration of C6 glucose was 4.5 mg / mL. After ex vivo, the tissue fractions used in experiments were immediately placed in labeled DMEM medium and cultured for 24 hours in an incubator containing 5% CO2 at 37°C. The medium was then removed, the tissue was washed three times with PBS, nitrogen-quenched, and stored at -80°C.

[0199] (2) Sample preparation and data acquisition

[0200] The serum sample pretreatment and LC-MS data acquisition conditions were as follows: 50 μL of serum was transferred to an EP tube, and 400 μL of pre-chilled extraction buffer (methanol:acetonitrile = 1:1 (v / v)) was added. The mixture was vortexed for 30 s; sonicated in an ice-water bath for 10 min; and allowed to stand at -20℃ for 1 h. The sample was centrifuged at 12000 rpm for 15 min at 4℃; the supernatant was collected and evaporated to dryness in a vacuum concentrator; the sample was reconstituted with 100 μL of a mixture of acetonitrile:water (1:1, v / v), centrifuged for the first time at 15000 rpm for 10 min, and the first supernatant was collected. The first supernatant was then centrifuged for the second time at 15000 rpm for 10 min, and the second supernatant was collected. The second supernatant was then aliquoted into vials for LC-MS analysis. Chromatographic analysis conditions: A Thermo UltiMate 3000 liquid chromatograph was used. The chromatographic column used was an ACQUITY UPLC BEH Amide Column (2.1 × 100 mm, 1.7 μm, Waters Corporation, USA). The column temperature was set to 30 °C. Mobile phase A consisted of water containing 25 mmol / L ammonium acetate and 25 mmol / L ammonia, and mobile phase B consisted of acetonitrile. The flow rate was set to 0.5 mL / min, and the injection volume was 5 μL. Pre-column equilibration was performed for 3.5 minutes. Linear gradient elution was used as follows: 0–0.5 min, 5% A; 0.5–7 min, 5–35% A; 7–8 min, 35–60% A; 8–9 min, 60% A; 9–9.1 min, 60–5% A. Mass spectrometry parameters were set as follows: A Q-Exactive quadrupole-orbit trap high-resolution tandem mass spectrometer with an ESI ion source was used in conjunction with the aforementioned UPLC / Q-Exactive liquid chromatograph for sample detection and analysis. Data acquisition was performed in full scan mode (Full MS). The scanning range is 70–1000 m / z, the resolution is 140,000, the maximum injection time is 100 ms, and the AGC target parameter is set to 3e6.

[0201] The tissue section preparation and AFADESI-MSI data acquisition process were as follows: All frozen tissue fractions were embedded in Surgipath CryoGel and then sectioned into 15 μm thick sections at -20°C using a Leica Cryostat CM1860 (Leica Microsystems, Wetzlar, Germany). Before MSI detection, the sections were dried in vacuum desiccators at -20°C and room temperature for 1 hour each. A Q-Exactive mass spectrometer from Thermo Fisher Scientific (USA) and a self-developed atmospheric pressure open-type AFADESI-MSI system were used. The imaging system parameters were as follows: spray voltage ±7 kV, spray gas pressure 0.7 MPa, and pump flow rate 45 L / min. The spray solvent system was acetonitrile-water solution (8:2, V:V), and the spray flow rate was 6 μL / min.

[0202] The sample's X-axis movement speed and Y-axis movement interval are set according to the sample size, using the following formulas: X-axis movement speed (mm / s) = scan time per line (min) × 60 ÷ sample length (mm); Y-axis movement interval = sample width (mm) ÷ number of lines; total scan time = scan time per line × number of lines. The mass spectrometry parameters are as follows: positive and negative ion full scan detection mode; mass spectrometry scan range m / z 70–1000; ion source temperature 350℃; resolution 140000; maximum injection time 100ms; AGC target parameter set to 3e6.

[0203] (3) Calculation of metabolic flux

[0204] First, the peak areas of glucose and its isotopic isoforms in serum samples detected by LC-MS were extracted. After natural isotope correction using AccuCor, the plasma enrichment of the tracer was obtained. Then, MassImager was used to reconstruct the MSI data, and the region of interest was selected without background subtraction to obtain a two-dimensional data matrix of the m / z and intensity of the traced metabolites. Using the plasma enrichment, the m / z and intensity of the traced metabolites as basic inputs, the MSIFluxer described in Example 1 was used to quantitatively calculate the spatially resolved metabolic flux of the corresponding tissue.

[0205] A schematic diagram of the spatial metabolic flux analysis method is shown below. Figure 2 .

[0206] Experimental results:

[0207] S1. Analyze the metabolic flux of the TCA cycle in the same renal cortex sample using both LC-MS and AFADESI-MSI methods.

[0208] 560 mM U- was continuously infused into SD rats via jugular vein at a rate of 0.1 μL / min / g. 13 C6 glucose solution was incubated for 2.5 h, and then serum and kidney tissue samples were collected. Each kidney tissue sample was longitudinally divided into two halves, and the renal cortex of one half was extracted and pretreated for LC-MS detection; the other half was sectioned and AFADESI-MSI data were collected. The m / z and peak area of ​​the labeled metabolites and their isotopic isotopes detected by LC-MS and AFADESI-MSI were extracted respectively, and the metabolic flux of the same renal cortex was calculated using MSIFluxer (see steps (1) to (3) above for specific steps).

[0209] First, the performance of 13 solvers was compared, and the solver with the smallest RSS value was selected as the best solver.

[0210] like Figure 3 a and Figure 3 As shown in b, based on the minimum residual sum of squares (RSS) value, LN_BOBYQA and GN_ESCH were selected as the best solvers for the local and global optimizations of LC-MS and AFADESI-MSI, respectively. Figure 3 As shown in c, the metabolic flux results of the 12 reactions in the TCA cycle calculated from LC-MS and AFADESI-MSI data are highly consistent, demonstrating the feasibility of spatial metabolic flux calculation based on MSIFluxer.

[0211] S2. Using MSIFluxer to identify labeled metabolites in brain tissue imaging.

[0212] 560 mM U- was continuously infused into SD rats via jugular vein at a rate of 0.1 μL / min / g. 13 C6 glucose solution for 2.5 h, then brain tissue samples were taken, and the brain tissue was sliced ​​and AFADESI-MSI data was acquired. After reconstruction of the MassImager image, the m / z and intensity of the brain tissue imaging were obtained. The labeled metabolites detected by the brain tissue imaging were identified using MSIFluxer (see steps (1) to (3) above for specific steps).

[0213] Ten labeled metabolites were identified from 21 metabolites in central carbon metabolism using MSIFluxer (see Table 3). AFADESI-MSI imaging of the isotopic isotopes of these 10 labeled metabolites is then presented (see Table 3). Figure 4By visualizing isotopic isotopes of labeled metabolites, the abundance of newly synthesized metabolites in different tissue microregions can be distinguished. AFADESI imaging results of these labeled metabolites show significant heterogeneous distribution in brain tissue. For example, M2 and M3 are highly labeled in glutamate M1-M5; succinate M1-M4 is more abundant in the cerebral cortex; and aspartate M1-M4 is more intense in the cerebral cortex than in the hippocampus and thalamus.

[0214] Table 3. 21 related metabolites of central carbon metabolism

[0215] ID m / z Molecular formula Ionic form Compound Name 1 215.0328 <![CDATA[C6H 12 O6]]> <![CDATA[[M+Cl] - ]]> glucose 2 259.0224 <![CDATA[C6H 13 O9P]]> <![CDATA[[M-H] - ]]> 6-phosphate glucose 3 168.9907 <![CDATA[C3H7O6P]]> <![CDATA[[M-H] - ]]> dihydroxyacetone phosphate 4 184.9857 <![CDATA[C3H7O7P]]> <![CDATA[[M-H] - ]]> Phosphoglyceric acid 5 166.9751 <![CDATA[C3H5O6P]]> <![CDATA[[M-H] - ]]> Phosphoenolpyruvate 6 87.0088 <![CDATA[C3H4O3]]> <![CDATA[[M-H] - ]]> Pyruvic acid 7 191.0197 <![CDATA[C6H8O7]]> <![CDATA[[M-H] - ]]> Citric acid 8 145.0142 <![CDATA[C5H6O5]]> <![CDATA[[M-H] - ]]> α-Ketoglutarate 9 117.0193 <![CDATA[C4H6O4]]> <![CDATA[[M-H] - ]]> Succinic acid 10 115.0037 <![CDATA[C4H4O4]]> <![CDATA[[M-H] - ]]> fumaric acid 11 133.0142 <![CDATA[C4H6O5]]> <![CDATA[[M-H] - ]]> malic acid 12 89.0244 <![CDATA[C3H6O3]]> <![CDATA[[M-H] - ]]> lactic acid 13 146.0459 <![CDATA[C5H9NO4]]> <![CDATA[[M-H] - ]]> glutamic acid 14 145.0619 <![CDATA[C5H 10 N2O3]]> <![CDATA[[M-H] - ]]> glutamine 15 132.0302 <![CDATA[C4H7NO4]]> <![CDATA[[M-H] - ]]> Aspartic acid 16 229.0119 <![CDATA[C5H 11 O8P]]> <![CDATA[[M-H] - ]]> 5-Phosphoribose 17 289.033 <![CDATA[C7H 15 O 10 P]]> <![CDATA[[M-H] - ]]> 7-Syringoheptulose phosphate 18 104.0352 <![CDATA[C3H7NO3]]> <![CDATA[[M-H] - ]]> Serine 20 259.0224 <![CDATA[C6H 13 O9P]]> <![CDATA[[M-H] - ]]> Fructose-6-phosphate 21 168.9907 <![CDATA[C3H7O6P]]> <![CDATA[[M-H] - ]]> Glyceraldehyde phosphate

[0216] S3. Calculate the metabolic flux of the renal cortex and renal medulla using MSIFluxer.

[0217] 560 mM U- was continuously infused into SD rats via jugular vein at a rate of 0.1 μL / min / g. 13 C6 glucose solution was used for 2.5 h, then serum and kidney tissue samples were collected. After the kidney tissue was sectioned, AFADESI-MSI data were acquired. After reconstruction of the images with MassImager, the renal cortex and renal medulla were selected to obtain the m / z and intensity of the corresponding microregions. The metabolic flux of the renal cortex and renal medulla was calculated using MSIFluxer. Figure 5 (a).

[0218] Figure 5 Figure 'a' shows the metabolic fluxes of substrate uptake and excretion, glycolysis, the TCA cycle, and the pentose phosphate pathway-related reactions in the renal cortex and medulla, quantified using MSIFluxer. Figure 5 As shown in b, 25 of the 86 reactions in the metabolic network exhibited different reaction rates in the renal cortex and medulla. The faster-reacting pathways in the renal cortex were mainly related to pyruvate metabolism, while the faster-reacting pathways in the renal medulla were mainly related to amino acid metabolism. The turnover of labeled metabolites was further determined using the MSIFluxer tool, such as... Figure 5 The average enrichment images of glutamate and glutamine shown in c-d indicate that the labeling levels in the medulla are higher than in the cortex. The conversion rate between glutamate and glutamine was further quantified using MSIFluxer. Figure 5 The results showed that the interconversion of glutamate and glutamine in the renal medulla was faster than in the renal cortex. Secondly, MSIFluxer was the first to determine the direction of reversible reactions in tissue microregions. For example... Figure 5 As shown in f, in the reversible reaction of pyruvate and lactate, the conversion rate of the renal cortex and renal medulla is very fast, but the net reaction rate is very slow and the reaction direction is always lactate to pyruvate.

[0219] S4. Calculate metabolic fluxes in the renal cortex and cerebral cortex using MSIFluxer.

[0220] 560 mM U- was continuously infused into SD rats via jugular vein at a rate of 0.1 μL / min / g. 13 C6 glucose solution was incubated for 2.5 hours. Then, serum, kidney and brain tissue samples were collected. After the kidney and brain tissue were sliced, AFADESI-MSI data were collected. Then, the metabolic flux of the renal cortex and cerebral cortex was calculated using MSIFluxer (see steps (1) to (3) above for details).

[0221] Figure 6 The 'a' in the figure shows the direction of multiple reversible reactions in the TCA cycle, and the relevant reaction rate of the TCA cycle in the cerebral cortex is significantly increased, reaching 2.8 to 6.6 times that in the renal cortex. Figure 6 Figure b shows that an active TCA cycle provides more NADH, and ATP production in the cerebral cortex is also significantly increased. The reversible reaction between glyceraldehyde-3-phosphate (GA3PA) and 3-phosphoglycerate (3PG) exhibits opposite directions in the renal cortex and cerebral cortex. Figure 6 As shown in c, the reaction in the renal cortex proceeds towards the formation of glyceraldehyde-3-phosphate, while in the cerebral cortex, the reaction proceeds towards the formation of 3-phosphoglycerate. Furthermore, N-acetylaspartic acid (AcAsp) is synthesized from acetyl-CoA and aspartic acid. The observation of various isotopic isotopes of AcAsp (M2 to M6) in the brain indicates that labeled aspartic acid and acetyl-CoA are incorporated into the synthesis of AcAsp. Figure 6 The data in the study showed that the synthesis flux of AcAsp in the cerebral cortex was 12 nmol / g / min, while the flux in the renal cortex was almost zero. Figure 6 The mean enrichment images of AcAsp in the cerebral cortex and renal cortex in e-f further confirmed the response flux results. These results indicate glucose production function in the renal cortex and active TCA circulation and specific AcAsp synthesis in the cerebral cortex.

[0222] S5. Metabolic flux analysis of laryngeal cancer samples

[0223] In vitro stable isotope tracing was performed on surgically resected laryngeal cancer tissue. 1% sodium pyruvate, 10% fetal bovine serum, 1% penicillin-antibody antibiotics, and 4.5 mg / mL U- were added to sugar-free DMEM medium. 13 C6 glucose was used to culture laryngeal cancer tissue samples in the above-mentioned culture medium for 24 hours, followed by washing and flash freezing in liquid nitrogen. AFADESI-MSI data were then acquired and metabolic flux was calculated after sectioning.

[0224] like Figure 7As shown in 'a', based on the hematoxylin-eosin staining results, T1, T2, and T3 are designated as tumor regions, and N1, N2, and N3 are the corresponding normal regions. Figure 7 As shown in b, the labeled isotopes of purine nucleotides such as inosine monophosphate (IMP), adenosine monophosphate (AMP), and guanosine monophosphate (GMP) appeared only in the M5 form. This indicates that laryngeal cancer tissue utilizes ribose-5-phosphate (R5P) produced by PPP to synthesize purine nucleotides in the tumor, reflecting the increased demand for nucleotides due to rapid tumor cell proliferation. The flux of 100 reactions in the metabolic network model was calculated, such as... Figure 7 Figures c-d show that AMP synthesis was faster than GMP synthesis in laryngeal cancer samples. R5P and reduced nicotinamide adenine dinucleotide phosphate (NADPH) are important products of PPP and play crucial roles in nucleotide and lipid synthesis. Figure 7 As shown in 'e', ​​PPP almost entirely provides NADPH in tumors. Phosphofructokinase (PFK), citrate synthase (CS), and glucose-6-phosphate dehydrogenase (G6PDH) are the rate-limiting enzymes of glycolysis, the TCA cycle, and the PPP pathway, respectively. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) is a key enzyme connecting glycolysis and the TCA cycle. Figure 7 The f-values ​​indicate that the PFK-involved response did not differ significantly between tumor and brain tissues, while CS, G6PDH, and GAPDH exhibited distinct patterns. In the brain, the CS-involved response was the fastest, while the G6PDH-involved response was the slowest. Conversely, in laryngeal cancer tissue, the G6PDH-involved response was the fastest, while the GAPDH-involved response was the slowest. Furthermore, the CS-involved response showed regional heterogeneity within the tumor.

[0225] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A metabolic flux calculation tool, characterized in that, The calculation steps of the metabolic flux calculation tool are as follows: (1) The m / z and intensity of the labeled metabolites to be tested by MSI are used as MSI data. A database containing labeled metabolites and their isotopic isotopes is used as a database. The labeled metabolites are automatically identified. After natural isotope correction of the labeled metabolites by AccuCor, the mass distribution vector of the labeled metabolites is output. (2) Based on the identified labeled metabolites and combined with the chemical reactions in the KEGG database, a metabolic network model was established; (3) Based on the metabolic network model and the mass distribution vector of the labeled metabolites, the solver is selected to obtain the selected solver; (4) Using the screened solver and metabolic network model, fit the mass distribution vector of the labeled metabolites, and use the minimization of the sum of squared residuals as the calculation endpoint to realize the calculation of metabolic flux.

2. The metabolic flux calculation tool according to claim 1, characterized in that, The method for automatically identifying labeled metabolites includes extracting and matching ion intensity information from MSI data and a database, filtering the database multiple times to remove labeled metabolites with poor imaging effects, retaining labeled metabolites visible in MSI, and obtaining the identified labeled metabolites. The method for selecting solvers includes preparing a metabolic network model and mass distribution vector data of labeled metabolites, calling the metabolic network model and mass distribution vector data of labeled metabolites, randomly generating initial values ​​of metabolic reaction rates of labeled metabolites, automatically calculating the mass distribution vector 1 of labeled metabolites simulated by the computer based on the initial values, and automatically calculating the RSS value a using the mass distribution vector in step (1) and the mass distribution vector 1 of the computer simulation; based on the RSS value a, iterating through the solvers, repeating each solver 5 times, and obtaining a fitted metabolic reaction rate after each solver calculation, obtaining the mass distribution vector 2 of labeled metabolites simulated by the computer based on the metabolic reaction rate, automatically calculating and recording the RSS value b based on the mass distribution vector 2 of labeled metabolites simulated by the computer and the mass distribution vector in step (1) returned each time, sorting and statistically analyzing the 5 RSS values ​​b, and selecting the median of the 5 RSS values ​​b, and selecting the solvers that are locally optimal and globally optimal based on the median of the RSS values; The RSS value b is less than the RSS value a.

3. The metabolic flux calculation tool according to claim 2, characterized in that, The method for removing labeled metabolites with poor imaging performance includes the following steps: For a matched labeled metabolite, if the ion intensity of labeled metabolite M0 is less than a set intensity threshold in more than half of the samples, it is considered that the labeled metabolite cannot form an effective imaging signal in MSI, and therefore the labeled metabolite is removed; the labeled metabolite M0 is a labeled metabolite in which the number of C labeled is 0. The method for retaining labeled metabolites visible in MSI includes the following steps: calculating the ratio of the average ionic intensity of the labeled metabolite isotopes M2 and M3 in the labeled group to the unlabeled group, respectively; if one of the ratios is greater than 10, it indicates that the labeled metabolite has formed a sufficient ionic signal in the MSI data, and the labeled metabolite is retained; wherein the labeled metabolite isotope M2 is a labeled metabolite with 2 C atoms; and the labeled metabolite isotope M3 is a labeled metabolite with 3 C atoms.

4. The metabolic flux calculation tool according to claim 2, characterized in that, The solvers mentioned are SLSQP, COBYLA, LN_COBYLA, LN_BOBYQA, LN_PRAXIS, LN_SBPLX, LN_NELDERMEAD, LN_NEWUOA, GN_CRS2_LM, GN_DIRECT_L, GN_AGS, GN_IRES, and GN_ESCH.

5. The metabolic flux calculation tool according to claim 1, characterized in that, The construction of the database containing labeled metabolites and their isotopic isoforms includes: Based on the metabolites included in the metabolic network model, calculate the summation of all metabolites in the metabolic network model using U-... 13 The molecular weight of the isotope isoform after C6-glucose labeling was determined, and a database containing metabolite names, molecular formulas and isotope molecular weights was finally established, resulting in a database containing labeled metabolites and their isotope isoforms. The chemical reactions described are all central carbon metabolic reactions associated with labeled metabolites in KEGG.

6. The application of the metabolic flux calculation tool according to any one of claims 1 to 5 in at least one of the following a to e: a. Identify labeled metabolites; b. Calculation of metabolic flux; c. Research on tissue metabolic regulation mechanisms; d. Studies on metabolic activities in tissue heterogeneity; e. Quantification of tissue metabolite conversion rate.

7. The application according to claim 6, characterized in that, The metabolic flux includes spatial metabolic flux or in situ metabolic flux.

8. A method for analyzing metabolic flux, characterized in that, The steps include stable isotope tracing, sample preparation, data acquisition, and calculating metabolic flux using the metabolic flux calculation tool described in any one of claims 1 to 5.

9. The method according to claim 8, characterized in that, The isotope tracing reagents include U- 13 C6 glucose solution.

10. The method according to claim 8, characterized in that, The data was acquired using the AFADESI-MSI system.

Images